General news from the MIT - Massachusetts Institute of Technology University

MIT News
MIT News is dedicated to communicating to the media and the public the news and achievements of the students, faculty, staff and the greater MIT community.
Novel method detects microbial contamination in cell cultures

Ultraviolet light “fingerprints” on cell cultures and machine learning can provide a definitive yes/no contamination assessment within 30 minutes.



The chemistry of creativity

Senior Madison Wang blends science, history, and art to probe how the world works and the tools we use to explore and understand it.

Senior Madison Wang, a double major in creative writing and chemistry, developed her passion for writing in middle school. Her interest in chemistry fit nicely alongside her commitment to producing engaging narratives. 

Wang believes that world-building in stories supported by science and research can make for a more immersive reader experience.

“In science and in writing, you have to tell an effective story,” she says. “People respond well to stories.”  

A native of Buffalo, New York, Wang applied early action for admission to MIT and learned quickly that the Institute was where she wanted to be. “It was a really good fit,” she says. “There was positive energy and vibes, and I had a great feeling overall.”

The power of science and good storytelling

“Chemistry is practical, complex, and interesting,” says Wang. “It’s about quantifying natural laws and understanding how reality works.”

Chemistry and writing both help us “see the world’s irregularity,” she continues. Together, they can erase the artificial and arbitrary line separating one from the other and work in concert to tell a more complete story about the world, the ways in which we participate in building it, and how people and objects exist in and move through it. 

“Understanding magnetism, material properties, and believing in the power of magic in a good story … these are why we’re drawn to explore,” she says. “Chemistry describes why things are the way they are, and I use it for world-building in my creative writing.”

Wang lauds MIT’s creative writing program and cites a course she took with Comparative Media Studies/Writing Professor and Pulitzer Prize winner Junot Díaz as an affirmation of her choice. Seeing and understanding the world through the eyes of a scientist — its building blocks, the ways the pieces fit and function together — help explain her passion for chemistry, especially inorganic and physical chemistry.

Wang cites the work of authors like Sam Kean and Knight Science Journalism Program Director Deborah Blum as part of her inspiration to study science. The books “The Disappearing Spoon” by Kean and “The Poisoner’s Handbook” by Blum “both present historical perspectives, opting for a story style to discuss the events and people involved,” she says. “They each put a lot of work into bridging the gap between what can sometimes be sterile science and an effective narrative that gets people to care about why the science matters.”

Genres like fantasy and science fiction are complementary, according to Wang. “Constructing an effective world means ensuring readers understand characters’ motivations — the ‘why’ — and ensuring it makes sense,” she says. “It’s also important to show how actions and their consequences influence and motivate characters.” 

As she explores the world’s building blocks inside and outside the classroom, Wang works to navigate multiple genres in her writing, as with her studies in chemistry. “I like romance and horror, too,” she says. “I have gripes with committing to a single genre, so I just take whatever I like from each and put them in my stories.”

In chemistry, Wang favors an environment in which scientists can regularly test their ideas. “It’s important to ground chemistry in the real world to create connections for students,” she argues. Advancements in the field have occurred, she notes, because scientists could exit the realm of theory and apply ideas practically.

“Fritz Haber’s work on ammonia synthesis revolutionized approaches to food supply chains,” she says, referring to the German chemist and Nobel laureate. “Converting nitrogen and hydrogen gas to ammonia for fertilizer marked a dramatic shift in how farming could work.” This kind of work could only result from the consistent, controlled, practical application of the theories scientists consider in laboratory environments.

A future built on collaboration and cooperation

Watching the world change dramatically and seeing humanity struggle to grapple with the implications of phenomena like climate change, political unrest, and shifting alliances, Wang emphasizes the importance of deconstructing silos in academia and the workplace. Technology can be a tool for harm, she notes, so inviting more people inside previously segregated spaces helps everyone.

Criticism in both chemistry and writing, Wang believes, are valuable tools for continuous improvement. Effective communication, explaining complex concepts, and partnering to develop long-term solutions are invaluable when working at the intersection of history, art, and science. In writing, Wang says, criticism can help define areas to improve writers’ stories and shape interesting ideas.

“We’ve seen the positive results that can occur with effective science writing, which requires rigor and fact-checking,” she says. “MIT’s cross-disciplinary approach to our studies, alongside feedback from teachers and peers, is a great set of tools to carry with us regardless of where we are.”

Wang explores connections between science and stories in her leisure time, too. “I’m a member of MIT’s Anime Club and I enjoy participating in MIT’s Sport Taekwondo Club,” she says. The competitive aspect in tae kwon do allows for her to feed her competitive drive and gets her out of her head. Her participation in DAAMIT (Digital Art and Animation at MIT) creates connections with different groups of people and gives her ideas she can use to tell better stories. “It’s fascinating exploring others’ minds,” she says.

Wang argues that there’s a false divide between science and the humanities and wants the work she does after graduation to bridge that divide. “Writing and learning about science can help,” she asserts. “Fields like conservation and history allow for continued exploration of that intersection.”

Ultimately, Wang believes it’s important to examine narratives carefully and to question notions of science’s inherent superiority over humanities fields. “The humanities and science have equal value,” she says.

Artificial intelligence enhances air mobility planning

Lincoln Laboratory is transitioning tools to the 618th Air Operations Center to streamline global transport logistics.

Every day, hundreds of chat messages flow between pilots, crew, and controllers of the Air Mobility Command's 618th Air Operations Center (AOC). These controllers direct a thousand-wide fleet of aircraft, juggling variables to determine which routes to fly, how much time fueling or loading supplies will take, or who can fly those missions. Their mission planning allows the U.S. Air Force to quickly respond to national security needs around the globe.

"It takes a lot of work to get a missile defense system across the world, for example, and this coordination used to be done through phone and email. Now, we are using chat, which creates opportunities for artificial intelligence to enhance our workflows," says Colonel Joseph Monaco, the director of strategy at the 618th AOC, which is the Department of Defense's largest air operations center.

The 618th AOC is sponsoring Lincoln Laboratory to develop these artificial intelligence tools, through a project called Conversational AI Technology for Transition (CAITT).

During a visit to Lincoln Laboratory from the 618th AOC's headquarters at Scott Air Force Base in Illinois, Colonel Monaco, Lieutenant Colonel Tim Heaton, and Captain Laura Quitiquit met with laboratory researchers to discuss CAITT. CAITT is a part of a broader effort to transition AI technology into a major Air Force modernization initiative, called the Next Generation Information Technology for Mobility Readiness Enhancement (NITMRE).

The type of AI being used in this project is natural language processing (NLP), which allows models to read and process human language. "We are utilizing NLP to map major trends in chat conversations, retrieve and cite specific information, and identify and contextualize critical decision points," says Courtland VanDam, a researcher in Lincoln Laboratory's AI Technology and Systems Group, which is leading the project. CAITT encompasses a suite of tools leveraging NLP.

One of the most mature tools, topic summarization, extracts trending topics from chat messages and formats those topics in a user-friendly display highlighting critical conversations and emerging issues. For example, a trending topic might read, "Crew members missing Congo visas, potential for delay." The entry shows the number of chats related to the topic and summarizes in bullet points the main points of conversations, linking back to specific chat exchanges.

"Our missions are very time-dependent, so we have to synthesize a lot of information quickly. This feature can really cue us as to where our efforts should be focused," says Monaco.

Another tool in production is semantic search. This tool improves upon the chat service's search engine, which currently returns empty results if chat messages do not contain every word in the query. Using the new tool, users can ask questions in a natural language format, such as why a specific aircraft is delayed, and receive intelligent results. "It incorporates a search model based on neural networks that can understand the user intent of the query and go beyond term matching," says VanDam.

Other tools under development aim to automatically add users to chat conversations deemed relevant to their expertise, predict the amount of ground time needed to unload specific types of cargo from aircraft, and summarize key processes from regulatory documents as a guide to operators as they develop mission plans.

The CAITT project grew out of the DAF–MIT AI Accelerator, a three-pronged effort between MIT, Lincoln Laboratory, and the Department of the Air Force (DAF) to develop and transition AI algorithms and systems to advance both the DAF and society. "Through our involvement in the AI Accelerator via the NITMRE project, we realized we could do something innovative with all of the unstructured chat information in the 618th AOC," says Heaton.

As laboratory researchers advance their prototypes of CAITT tools, they have begun to transition them to the 402nd Software Engineering Group, a software provider for the Department of Defense. That group will implement the tools into the operational software environment in use by the 618th AOC. 

Designing a new way to optimize complex coordinated systems

Using diagrams to represent interactions in multipart systems can provide a faster way to design software improvements.

Coordinating complicated interactive systems, whether it’s the different modes of transportation in a city or the various components that must work together to make an effective and efficient robot, is an increasingly important subject for software designers to tackle. Now, researchers at MIT have developed an entirely new way of approaching these complex problems, using simple diagrams as a tool to reveal better approaches to software optimization in deep-learning models.

They say the new method makes addressing these complex tasks so simple that it can be reduced to a drawing that would fit on the back of a napkin.

The new approach is described in the journal Transactions of Machine Learning Research, in a paper by incoming doctoral student Vincent Abbott and Professor Gioele Zardini of MIT’s Laboratory for Information and Decision Systems (LIDS).

“We designed a new language to talk about these new systems,” Zardini says. This new diagram-based “language” is heavily based on something called category theory, he explains.

It all has to do with designing the underlying architecture of computer algorithms — the programs that will actually end up sensing and controlling the various different parts of the system that’s being optimized. “The components are different pieces of an algorithm, and they have to talk to each other, exchange information, but also account for energy usage, memory consumption, and so on.” Such optimizations are notoriously difficult because each change in one part of the system can in turn cause changes in other parts, which can further affect other parts, and so on.

The researchers decided to focus on the particular class of deep-learning algorithms, which are currently a hot topic of research. Deep learning is the basis of the large artificial intelligence models, including large language models such as ChatGPT and image-generation models such as Midjourney. These models manipulate data by a “deep” series of matrix multiplications interspersed with other operations. The numbers within matrices are parameters, and are updated during long training runs, allowing for complex patterns to be found. Models consist of billions of parameters, making computation expensive, and hence improved resource usage and optimization invaluable.

Diagrams can represent details of the parallelized operations that deep-learning models consist of, revealing the relationships between algorithms and the parallelized graphics processing unit (GPU) hardware they run on, supplied by companies such as NVIDIA. “I’m very excited about this,” says Zardini, because “we seem to have found a language that very nicely describes deep learning algorithms, explicitly representing all the important things, which is the operators you use,” for example the energy consumption, the memory allocation, and any other parameter that you’re trying to optimize for.

Much of the progress within deep learning has stemmed from resource efficiency optimizations. The latest DeepSeek model showed that a small team can compete with top models from OpenAI and other major labs by focusing on resource efficiency and the relationship between software and hardware. Typically, in deriving these optimizations, he says, “people need a lot of trial and error to discover new architectures.” For example, a widely used optimization program called FlashAttention took more than four years to develop, he says. But with the new framework they developed, “we can really approach this problem in a more formal way.” And all of this is represented visually in a precisely defined graphical language.

But the methods that have been used to find these improvements “are very limited,” he says. “I think this shows that there’s a major gap, in that we don’t have a formal systematic method of relating an algorithm to either its optimal execution, or even really understanding how many resources it will take to run.” But now, with the new diagram-based method they devised, such a system exists.

Category theory, which underlies this approach, is a way of mathematically describing the different components of a system and how they interact in a generalized, abstract manner. Different perspectives can be related. For example, mathematical formulas can be related to algorithms that implement them and use resources, or descriptions of systems can be related to robust “monoidal string diagrams.” These visualizations allow you to directly play around and experiment with how the different parts connect and interact. What they developed, he says, amounts to “string diagrams on steroids,” which incorporates many more graphical conventions and many more properties.

“Category theory can be thought of as the mathematics of abstraction and composition,” Abbott says. “Any compositional system can be described using category theory, and the relationship between compositional systems can then also be studied.” Algebraic rules that are typically associated with functions can also be represented as diagrams, he says. “Then, a lot of the visual tricks we can do with diagrams, we can relate to algebraic tricks and functions. So, it creates this correspondence between these different systems.”

As a result, he says, “this solves a very important problem, which is that we have these deep-learning algorithms, but they’re not clearly understood as mathematical models.” But by representing them as diagrams, it becomes possible to approach them formally and systematically, he says.

One thing this enables is a clear visual understanding of the way parallel real-world processes can be represented by parallel processing in multicore computer GPUs. “In this way,” Abbott says, “diagrams can both represent a function, and then reveal how to optimally execute it on a GPU.”

The “attention” algorithm is used by deep-learning algorithms that require general, contextual information, and is a key phase of the serialized blocks that constitute large language models such as ChatGPT. FlashAttention is an optimization that took years to develop, but resulted in a sixfold improvement in the speed of attention algorithms.

Applying their method to the well-established FlashAttention algorithm, Zardini says that “here we are able to derive it, literally, on a napkin.” He then adds, “OK, maybe it’s a large napkin.” But to drive home the point about how much their new approach can simplify dealing with these complex algorithms, they titled their formal research paper on the work “FlashAttention on a Napkin.”

This method, Abbott says, “allows for optimization to be really quickly derived, in contrast to prevailing methods.” While they initially applied this approach to the already existing FlashAttention algorithm, thus verifying its effectiveness, “we hope to now use this language to automate the detection of improvements,” says Zardini, who in addition to being a principal investigator in LIDS, is the Rudge and Nancy Allen Assistant Professor of Civil and Environmental Engineering, and an affiliate faculty with the Institute for Data, Systems, and Society.

The plan is that ultimately, he says, they will develop the software to the point that “the researcher uploads their code, and with the new algorithm you automatically detect what can be improved, what can be optimized, and you return an optimized version of the algorithm to the user.”

In addition to automating algorithm optimization, Zardini notes that a robust analysis of how deep-learning algorithms relate to hardware resource usage allows for systematic co-design of hardware and software. This line of work integrates with Zardini’s focus on categorical co-design, which uses the tools of category theory to simultaneously optimize various components of engineered systems.

Abbott says that “this whole field of optimized deep learning models, I believe, is quite critically unaddressed, and that’s why these diagrams are so exciting. They open the doors to a systematic approach to this problem.”

“I’m very impressed by the quality of this research. ... The new approach to diagramming deep-learning algorithms used by this paper could be a very significant step,” says Jeremy Howard, founder and CEO of Answers.ai, who was not associated with this work. “This paper is the first time I’ve seen such a notation used to deeply analyze the performance of a deep-learning algorithm on real-world hardware. ... The next step will be to see whether real-world performance gains can be achieved.”

“This is a beautifully executed piece of theoretical research, which also aims for high accessibility to uninitiated readers — a trait rarely seen in papers of this kind,” says Petar Velickovic, a senior research scientist at Google DeepMind and a lecturer at Cambridge University, who was not associated with this work. These researchers, he says, “are clearly excellent communicators, and I cannot wait to see what they come up with next!”

The new diagram-based language, having been posted online, has already attracted great attention and interest from software developers. A reviewer from Abbott’s prior paper introducing the diagrams noted that “The proposed neural circuit diagrams look great from an artistic standpoint (as far as I am able to judge this).” “It’s technical research, but it’s also flashy!” Zardini says.

Martina Solano Soto wants to solve the mysteries of the universe, and MIT Open Learning is part of her plan

The 17-year-old student from Spain uses MIT resources to deepen her understanding of math and physics.



Luna: A moon on Earth

MIT students and faculty designed and fabricated a control room for the first lunar landing mission since the Apollo era — an achievement in design and engineering.

On March 6, MIT launched its first lunar landing mission since the Apollo era, sending three payloads — the AstroAnt, the RESOURCE 3D camera, and the HUMANS nanowafer — to the moon’s south polar region. The mission was based out of Luna, a mission control space designed by MIT Department of Architecture students and faculty in collaboration with the MIT Space Exploration Initiative, Inploration, and Simpson Gumpertz and Heger. It is installed in the MIT Media Lab ground-floor gallery and is open to the public as part of Artfinity, MIT’s Festival for the Arts. The installation allows visitors to observe payload operators at work and interact with the software used for the mission, thanks to virtual reality.

A central hub for mission operations, the control room is a structural and conceptual achievement, balancing technical challenges with a vision for an immersive experience, and the result of a multidisciplinary approach. “This will be our moon on Earth,” says Mateo Fernandez, a third-year MArch student and 2024 MAD Design Fellow, who designed and fabricated Luna in collaboration with Nebyu Haile, a PhD student in the Building Technology program in the Department of Architecture, and Simon Lesina Debiasi, a research assistant in the SMArchS Computation program and part of the Self-Assembly Lab. “The design was meant for people — for the researchers to be able to see what’s happening at all times, and for the spectators to have a 360 panoramic view of everything that’s going on,” explains Fernandez. “A key vision of the team was to create a control room that broke away from the traditional, closed-off model — one that instead invited the public to observe, ask questions, and engage with the mission,” adds Haile.

For this project, students were advised by Skylar Tibbits, founder and co-director of the Self-Assembly Lab, associate professor of design research, and the Morningside Academy for Design (MAD)’s assistant director for education; J. Roc Jih, associate professor of the practice in architectural design; John Ochsendorf, MIT Class of 1942 Professor with appointments in the departments of Architecture and Civil and Environmental Engineering, and founding director of MAD; and Brandon Clifford, associate professor of architecture. The team worked closely with Cody Paige, director of the Space Exploration Initiative at the Media Lab, and her collaborators, emphasizing that they “tried to keep things very minimal, very simple, because at the end of the day,” explains Fernandez, “we wanted to create a design that allows the researchers to shine and the mission to shine.”

“This project grew out of the Space Architecture class we co-taught with Cody Paige and astronaut and MIT AeroAstro [Department of Aeronautics and Astronautics] faculty member Jeff Hoffman” in the fall semester, explains Tibbits. “Mateo was part of that studio, and from there, Cody invited us to design the mission control project. We then brought Mateo onboard, Simon, Nebyu, and the rest of the project team.” According to Tibbits, “this project represents MIT’s mind-and-hand ethos. We had designers, architects, artists, computational experts, and engineers working together, reflecting the polymath vision — left brain, right brain, the creative and the technical coming together to make this possible.”

Luna was funded and informed by Tibbits and Jih’s Professor Amar G. Bose Research Grant Program. “J. Jih and I had been doing research for the Bose grant around basalt and mono-material construction,” says Tibbits, adding that they “had explored foamed glass materials similar to pumice or foamed basalt, which are also similar to lunar regolith.” “FOAMGLAS is typically used for insulation, but it has diverse applications, including direct ground contact and exterior walls, with strong acoustic and thermal properties,” says Jih. “We helped Mateo understand how the material is used in architecture today, and how it could be applied in this project, aligning with our work on new material palettes and mono-material construction techniques.”

Additional funding came from Inploration, a project run by creative director, author, and curator Lawrence Azerrad, as well as expeditionary artist, curator, and analog astronaut artist Richelle Ellis, and Comcast, a Media Lab member company. It was also supported by the MIT Morningside Academy for Design through Fernandez’s Design Fellowship. Additional support came from industry members such as Owens Corning (construction materials), Bose (communications), as well as MIT Media Lab member companies Dell Technologies (operations hardware) and Steelcase (operations seating). 

A moon on Earth

While the lunar mission ended prematurely, the team says it achieved success in the design and construction of a control room embodying MIT’s design approach and capacity to explore new technologies while maintaining simplicity. Luna looks like variations of the moon, offering different perspectives of the moon’s round or crescent shape, depending on the viewer’s position.

“What’s remarkable is how close the final output is to Mateo’s original sketches and renderings,” Tibbits notes. “That often doesn’t happen — where the final built project aligns so precisely with the initial design intent.”

Luna’s entire structure is built from FOAMGLAS, a durable material composed of glass cells usually used for insulation. “FOAMGLAS is an interesting material,” says Lesina Debiasi, who supported fabrication efforts, ensuring a fast and safe process. “It’s relatively durable and light, but can easily be crumbled with a sharp edge or blade, requiring every step of the fabrication process — cutting, texturing, sealing — to be carefully controlled.”

Fernandez, whose design experience was influenced by the idea that “simple moves” are most powerful, explains: “We’re giving a second life to materials that are not thought of for building construction … and I think that’s an effective idea. Here, you don’t need wood, concrete, rebar — you can build with one material only.” While the interior of the dome-shaped construction is smooth, the exterior was hand textured to evoke the basalt-like surface of the moon.

The lightweight cellular glass produced by Owens Corning, which sponsored part of the material, comes as an unexpected choice for a compression structure — a type of architectural design where stability is achieved through the natural force of compression, usually implying heavy materials. The control room doesn’t use connections or additional supports, and depends upon the precise placement, size, and weight of individual blocks to create a stable form from a succession of arches.

“Traditional compression structures rely on their own weight for stability, but using a material that is more than 10 times lighter than masonry meant we had to rethink everything. It was about finding the perfect balance between design vision and structural integrity,” reflects Haile, who was responsible for the structural calculations for the dome and its support.

Compression relies on gravity, and wouldn’t be a viable construction method on the moon itself. “We’re building using physics, loads, structures, and equilibrium to create this thing that looks like the moon, but depends on Earth’s forces to be built. I think people don’t see that at first, but there’s something cheeky and ironic about it,” confides Fernandez, acknowledging that the project merges historical building methods with contemporary design.

The location and purpose of Luna — both a work space and an installation engaging the public — implied balancing privacy and transparency to achieve functionality. “One of the most important design elements that reflected this vision was the openness of the dome,” says Haile. “We worked closely from the start to find the right balance — adjusting the angle and size of the opening to make the space feel welcoming, while still offering some privacy to those working inside.”

The power of collaboration

With the FOAMGLAS material, the team had to invent a fabrication process that would achieve the initial vision while maintaining structural integrity. Sourcing a material with radically different properties compared to conventional construction implied collaborating closely on the engineering front, the lightweight nature of the cellular glass requiring creative problem-solving: “What appears perfect in digital models doesn’t always translate seamlessly into the real world,” says Haile. “The slope, curves, and overall geometry directly determine whether the dome will stand, requiring Mateo and me to work in sync from the very beginning through the end of construction.” While the engineering was primarily led by Haile and Ochsendorf, the structural design was officially reviewed and approved by Paul Kassabian at Simpson Gumpertz and Heger (SGH), ensuring compliance with engineering standards and building codes.

“None of us had worked with FOAMGLAS before, and we needed to figure out how best to cut, texture, and seal it,” says Lesina Debiasi. “Since each row consists of a distinct block shape and specific angles, ensuring accuracy and repeatability across all the blocks became a major challenge. Since we had to cut each individual block four times before we were able to groove and texture the surface, creating a safe production process and mitigating the distribution of dust was critical,” he explains. “Working inside a tent, wearing personal protective equipment like masks, visors, suits, and gloves made it possible to work for an extended period with this material.”

In addition, manufacturing introduced small margins of error threatening the structural integrity of the dome, prompting hands-on experimentation. “The control room is built from 12 arches,” explains Fernandez. “When one of the arches closes, it becomes stable, and you can move on to the next one … Going from side to side, you meet at the middle and close the arch using a special block — a keystone, which was cut to measure,” he says. “In conversations with our advisors, we decided to account for irregularities in the final keystone of each row. Once this custom keystone sat in place, the forces would stabilize the arch and make it secure,” adds Lesina Debiasi.

“This project exemplified the best practices of engineers and architects working closely together from design inception to completion — something that was historically common but is less typical today,” says Haile. “This collaboration was not just necessary — it ultimately improved the final result.”

Fernandez, who is supported this year by the MAD Design Fellowship, expressed how “the fellowship gave [him] the freedom to explore [his] passions and also keep [his] agency.”

“In a way, this project embodies what design education at MIT should be,” Tibbits reflects. “We’re building at full scale, with real-world constraints, experimenting at the limits of what we know — design, computation, engineering, and science. It’s hands-on, highly experimental, and deeply collaborative, which is exactly what we dream of for MAD, and MIT’s design education more broadly.”

“Luna, our physical lunar mission control, highlights the incredible collaboration across the Media Lab, Architecture, and the School of Engineering to bring our lunar mission to the world. We are democratizing access to space for all,” says Dava Newman, Media Lab director and Apollo Professor of Astronautics.

A full list of contributors and supporters can be found at the Morningside Academy for Design's website.

Six from MIT elected to American Academy of Arts and Sciences for 2025

The prestigious honor society announces nearly 250 new members.

Six MIT faculty members are among the nearly 250 leaders from academia, the arts, industry, public policy, and research elected to the American Academy of Arts and Sciences, the academy announced April 23.

One of the nation’s most prestigious honorary societies, the academy is also a leading center for independent policy research. Members contribute to academy publications, as well as studies of science and technology policy, energy and global security, social policy and American institutions, the humanities and culture, and education.

Those elected from MIT in 2025 are:

• Lotte Bailyn, T. Wilson Professor of Management Emerita;
• Gareth McKinley, School of Engineering Professor of Teaching Innovation;
• Nasser Rabbat, Aga Khan Professor;
• Susan Silbey, Leon and Anne Goldberg Professor of Humanities and professor of sociology and anthropology;
• Anne Whiston Spirn, Cecil and Ida Green Distinguished Professor of Landscape Architecture and Planning; and
• Catherine Wolfram, William Barton Rogers Professor in Energy and professor of applied economics.

“These new members’ accomplishments speak volumes about the human capacity for discovery, creativity, leadership, and persistence. They are a stellar testament to the power of knowledge to broaden our horizons and deepen our understanding,” says Academy President Laurie L. Patton. “We invite every new member to celebrate their achievement and join the Academy in our work to promote the common good.”

Since its founding in 1780, the academy has elected leading thinkers from each generation, including George Washington and Benjamin Franklin in the 18th century, Maria Mitchell and Daniel Webster in the 19th century, and Toni Morrison and Albert Einstein in the 20th century. The current membership includes more than 250 Nobel and Pulitzer Prize winners.

Robotic system zeroes in on objects most relevant for helping humans

A new approach could enable intuitive robotic helpers for household, workplace, and warehouse settings.

For a robot, the real world is a lot to take in. Making sense of every data point in a scene can take a huge amount of computational effort and time. Using that information to then decide how to best help a human is an even thornier exercise.

Now, MIT roboticists have a way to cut through the data noise, to help robots focus on the features in a scene that are most relevant for assisting humans.

Their approach, which they aptly dub “Relevance,” enables a robot to use cues in a scene, such as audio and visual information, to determine a human’s objective and then quickly identify the objects that are most likely to be relevant in fulfilling that objective. The robot then carries out a set of maneuvers to safely offer the relevant objects or actions to the human.

The researchers demonstrated the approach with an experiment that simulated a conference breakfast buffet. They set up a table with various fruits, drinks, snacks, and tableware, along with a robotic arm outfitted with a microphone and camera. Applying the new Relevance approach, they showed that the robot was able to correctly identify a human’s objective and appropriately assist them in different scenarios.

In one case, the robot took in visual cues of a human reaching for a can of prepared coffee, and quickly handed the person milk and a stir stick. In another scenario, the robot picked up on a conversation between two people talking about coffee, and offered them a can of coffee and creamer.

Overall, the robot was able to predict a human’s objective with 90 percent accuracy and to identify relevant objects with 96 percent accuracy. The method also improved a robot’s safety, reducing the number of collisions by more than 60 percent, compared to carrying out the same tasks without applying the new method.

“This approach of enabling relevance could make it much easier for a robot to interact with humans,” says Kamal Youcef-Toumi, professor of mechanical engineering at MIT. “A robot wouldn’t have to ask a human so many questions about what they need. It would just actively take information from the scene to figure out how to help.”

Youcef-Toumi’s group is exploring how robots programmed with Relevance can help in smart manufacturing and warehouse settings, where they envision robots working alongside and intuitively assisting humans.

Youcef-Toumi, along with graduate students Xiaotong Zhang and Dingcheng Huang, will present their new method at the IEEE International Conference on Robotics and Automation (ICRA) in May. The work builds on another paper presented at ICRA the previous year.

Finding focus

The team’s approach is inspired by our own ability to gauge what’s relevant in daily life. Humans can filter out distractions and focus on what’s important, thanks to a region of the brain known as the Reticular Activating System (RAS). The RAS is a bundle of neurons in the brainstem that acts subconsciously to prune away unnecessary stimuli, so that a person can consciously perceive the relevant stimuli. The RAS helps to prevent sensory overload, keeping us, for example, from fixating on every single item on a kitchen counter, and instead helping us to focus on pouring a cup of coffee.

“The amazing thing is, these groups of neurons filter everything that is not important, and then it has the brain focus on what is relevant at the time,” Youcef-Toumi explains. “That’s basically what our proposition is.”

He and his team developed a robotic system that broadly mimics the RAS’s ability to selectively process and filter information. The approach consists of four main phases. The first is a watch-and-learn “perception” stage, during which a robot takes in audio and visual cues, for instance from a microphone and camera, that are continuously fed into an AI “toolkit.” This toolkit can include a large language model (LLM) that processes audio conversations to identify keywords and phrases, and various algorithms that detect and classify objects, humans, physical actions, and task objectives. The AI toolkit is designed to run continuously in the background, similarly to the subconscious filtering that the brain’s RAS performs.

The second stage is a “trigger check” phase, which is a periodic check that the system performs to assess if anything important is happening, such as whether a human is present or not. If a human has stepped into the environment, the system’s third phase will kick in. This phase is the heart of the team’s system, which acts to determine the features in the environment that are most likely relevant to assist the human.

To establish relevance, the researchers developed an algorithm that takes in real-time predictions made by the AI toolkit. For instance, the toolkit’s LLM may pick up the keyword “coffee,” and an action-classifying algorithm may label a person reaching for a cup as having the objective of “making coffee.” The team’s Relevance method would factor in this information to first determine the “class” of objects that have the highest probability of being relevant to the objective of “making coffee.” This might automatically filter out classes such as “fruits” and “snacks,” in favor of “cups” and “creamers.” The algorithm would then further filter within the relevant classes to determine the most relevant “elements.” For instance, based on visual cues of the environment, the system may label a cup closest to a person as more relevant — and helpful — than a cup that is farther away.

In the fourth and final phase, the robot would then take the identified relevant objects and plan a path to physically access and offer the objects to the human.

Helper mode

The researchers tested the new system in experiments that simulate a conference breakfast buffet. They chose this scenario based on the publicly available Breakfast Actions Dataset, which comprises videos and images of typical activities that people perform during breakfast time, such as preparing coffee, cooking pancakes, making cereal, and frying eggs. Actions in each video and image are labeled, along with the overall objective (frying eggs, versus making coffee).

Using this dataset, the team tested various algorithms in their AI toolkit, such that, when receiving actions of a person in a new scene, the algorithms could accurately label and classify the human tasks and objectives, and the associated relevant objects.

In their experiments, they set up a robotic arm and gripper and instructed the system to assist humans as they approached a table filled with various drinks, snacks, and tableware. They found that when no humans were present, the robot’s AI toolkit operated continuously in the background, labeling and classifying objects on the table.

When, during a trigger check, the robot detected a human, it snapped to attention, turning on its Relevance phase and quickly identifying objects in the scene that were most likely to be relevant, based on the human’s objective, which was determined by the AI toolkit.

“Relevance can guide the robot to generate seamless, intelligent, safe, and efficient assistance in a highly dynamic environment,” says co-author Zhang.

Going forward, the team hopes to apply the system to scenarios that resemble workplace and warehouse environments, as well as to other tasks and objectives typically performed in household settings.

“I would want to test this system in my home to see, for instance, if I’m reading the paper, maybe it can bring me coffee. If I’m doing laundry, it can bring me a laundry pod. If I’m doing repair, it can bring me a screwdriver,” Zhang says. “Our vision is to enable human-robot interactions that can be much more natural and fluent.”

This research was made possible by the support and partnership of King Abdulaziz City for Science and Technology (KACST) through the Center for Complex Engineering Systems at MIT and KACST.

Wearable device tracks individual cells in the bloodstream in real time

The technology, which achieves single-cell resolution, could help in continuous, noninvasive patient assessment to guide medical treatments.

Researchers at MIT have developed a noninvasive medical monitoring device powerful enough to detect single cells within blood vessels, yet small enough to wear like a wristwatch. One important aspect of this wearable device is that it can enable continuous monitoring of circulating cells in the human body.

The technology was presented online on March 3 by the journal npj Biosensing and is forthcoming in the journal’s print version.

The device — named CircTrek — was developed by researchers in the Nano-Cybernetic Biotrek research group, led by Deblina Sarkar, assistant professor at MIT and AT&T Career Development Chair at the MIT Media Lab. This technology could greatly facilitate early diagnosis of disease, detection of disease relapse, assessment of infection risk, and determination of whether a disease treatment is working, among other medical processes.

Whereas traditional blood tests are like a snapshot of a patient’s condition, CircTrek was designed to present real-time assessment, referred to in the npj Biosensing paper as having been “an unmet goal to date.” A different technology that offers monitoring of cells in the bloodstream with some continuity, in vivo flow cytometry, “requires a room-sized microscope, and patients need to be there for a long time,” says Kyuho Jang, a PhD student in Sarkar’s lab.

CircTrek, on the other hand, which is equipped with an onboard Wi-Fi module, could even monitor a patient’s circulating cells at home and send that information to the patient’s doctor or care team.

“CircTrek offers a path to harnessing previously inaccessible information, enabling timely treatments, and supporting accurate clinical decisions with real-time data,” says Sarkar. “Existing technologies provide monitoring that is not continuous, which can lead to missing critical treatment windows. We overcome this challenge with CircTrek.”

The device works by directing a focused laser beam to stimulate cells beneath the skin that have been fluorescently labeled. Such labeling can be accomplished with a number of methods, including applying antibody-based fluorescent dyes to the cells of interest or genetically modifying such cells so that they express fluorescent proteins.

For example, a patient receiving CAR T cell therapy, in which immune cells are collected and modified in a lab to fight cancer (or, experimentally, to combat HIV or Covid-19), could have those cells labeled at the same time with fluorescent dyes or genetic modification so the cells express fluorescent proteins. Importantly, cells of interest can also be labeled with in vivo labeling methods approved in humans. Once the cells are labeled and circulating in the bloodstream, CircTrek is designed to apply laser pulses to enhance and detect the cells’ fluorescent signal while an arrangement of filters minimizes low-frequency noise such as heartbeats.

“We optimized the optomechanical parts to reduce noise significantly and only capture the signal from the fluorescent cells,” says Jang.

Detecting the labeled CAR T cells, CircTrek could assess whether the cell therapy treatment is working. As an example, persistence of the CAR T cells in the blood after treatment is associated with better outcomes in patients with B-cell lymphoma.

To keep CircTrek small and wearable, the researchers were able to miniaturize the components of the device, such as the circuit that drives the high-intensity laser source and keeps the power level of the laser stable to avoid false readings.

The sensor that detects the fluorescent signals of the labeled cells is also minute, and yet it is capable of detecting a quantity of light equivalent to a single photon, Jang says.

The device’s subcircuits, including the laser driver and the noise filters, were custom-designed to fit on a circuit board measuring just 42 mm by 35 mm, allowing CircTrek to be approximately the same size as a smartwatch.

CircTrek was tested on an in vitro configuration that simulated blood flow beneath human skin, and its single-cell detection capabilities were verified through manual counting with a high-resolution confocal microscope. For the in vitro testing, a fluorescent dye called Cyanine5.5 was employed. That particular dye was selected because it reaches peak activation at wavelengths within skin tissue’s optical window, or the range of wavelengths that can penetrate the skin with minimal scattering.

The safety of the device, particularly the temperature increase on experimental skin tissue caused by the laser, was also investigated. An increase of 1.51 degrees Celsius at the skin surface was determined to be well below heating that would damage tissue, with enough of a margin that even increasing the device’s area of detection, and its power, in order to ensure the observation of at least one blood vessel could be safely permitted.

While clinical translation of CircTrek will require further steps, Jang says its parameters can be modified to broaden its potential, so that doctors could be provided with critical information on nearly any patient.

New electronic “skin” could enable lightweight night-vision glasses

MIT engineers developed ultrathin electronic films that sense heat and other signals, and could reduce the bulk of conventional goggles and scopes.

MIT engineers have developed a technique to grow and peel ultrathin “skins” of electronic material. The method could pave the way for new classes of electronic devices, such as ultrathin wearable sensors, flexible transistors and computing elements, and highly sensitive and compact imaging devices. 

As a demonstration, the team fabricated a thin membrane of pyroelectric material — a class of heat-sensing material that produces an electric current in response to changes in temperature. The thinner the pyroelectric material, the better it is at sensing subtle thermal variations.

With their new method, the team fabricated the thinnest pyroelectric membrane yet, measuring 10 nanometers thick, and demonstrated that the film is highly sensitive to heat and radiation across the far-infrared spectrum.

The newly developed film could enable lighter, more portable, and highly accurate far-infrared (IR) sensing devices, with potential applications for night-vision eyewear and autonomous driving in foggy conditions. Current state-of-the-art far-IR sensors require bulky cooling elements. In contrast, the new pyroelectric thin film requires no cooling and is sensitive to much smaller changes in temperature. The researchers are exploring ways to incorporate the film into lighter, higher-precision night-vision glasses.

“This film considerably reduces weight and cost, making it lightweight, portable, and easier to integrate,” Xinyuan Zhang, a graduate student in MIT’s Department of Materials Science and Engineering (DMSE). “For example, it could be directly worn on glasses.”

The heat-sensing film could also have applications in environmental and biological sensing, as well as imaging of astrophysical phenomena that emit far-infrared radiation.

What’s more, the new lift-off technique is generalizable beyond pyroelectric materials. The researchers plan to apply the method to make other ultrathin, high-performance semiconducting films.

Their results are reported today in a paper appearing in the journal Nature. The study’s MIT co-authors are first author Xinyuan Zhang, Sangho Lee, Min-Kyu Song, Haihui Lan, Jun Min Suh, Jung-El Ryu, Yanjie Shao, Xudong Zheng, Ne Myo Han, and Jeehwan Kim, associate professor of mechanical engineering and of materials science and engineering, along with researchers at the University Wisconsin at Madison led by Professor Chang-Beom Eom and authors from multiple other institutions.

Chemical peel

Kim’s group at MIT is finding new ways to make smaller, thinner, and more flexible electronics. They envision that such ultrathin computing “skins” can be incorporated into everything from smart contact lenses and wearable sensing fabrics to stretchy solar cells and bendable displays. To realize such devices, Kim and his colleagues have been experimenting with methods to grow, peel, and stack semiconducting elements, to fabricate ultrathin, multifunctional electronic thin-film membranes.

One method that Kim has pioneered is “remote epitaxy” — a technique where semiconducting materials are grown on a single-crystalline substrate, with an ultrathin layer of graphene in between. The substrate’s crystal structure serves as a scaffold along which the new material can grow. The graphene acts as a nonstick layer, similar to Teflon, making it easy for researchers to peel off the new film and transfer it onto flexible and stacked electronic devices. After peeling off the new film, the underlying substrate can be reused to make additional thin films.

Kim has applied remote epitaxy to fabricate thin films with various characteristics. In trying different combinations of semiconducting elements, the researchers happened to notice that a certain pyroelectric material, called PMN-PT, did not require an intermediate layer assist in order to separate from its substrate. Just by growing PMN-PT directly on a single-crystalline substrate, the researchers could then remove the grown film, with no rips or tears to its delicate lattice.

“It worked surprisingly well,” Zhang says. “We found the peeled film is atomically smooth.”

Lattice lift-off

In their new study, the MIT and UW Madison researchers took a closer look at the process and discovered that the key to the material’s easy-peel property was lead. As part of its chemical structure, the team, along with colleagues at the Rensselaer Polytechnic Institute, discovered that the pyroelectric film contains an orderly arrangement of lead atoms that have a large “electron affinity,” meaning that lead attracts electrons and prevents the charge carriers from traveling and connecting to another materials such as an underlying substrate. The lead acts as tiny nonstick units, allowing the material as a whole to peel away, perfectly intact.

The team ran with the realization and fabricated multiple ultrathin films of PMN-PT, each about 10 nanometers thin. They peeled off pyroelectric films and transfered them onto a small chip to form an array of 100 ultrathin heat-sensing pixels, each about 60 square microns (about .006 square centimeters). They exposed the films to ever-slighter changes in temperature and found the pixels were highly sensitive to small changes across the far-infrared spectrum.

The sensitivity of the pyroelectric array is comparable to that of state-of-the-art night-vision devices. These devices are currently based on photodetector materials, in which a change in temperature induces the material’s electrons to jump in energy and briefly cross an energy “band gap,” before settling back into their ground state. This electron jump serves as an electrical signal of the temperature change. However, this signal can be affected by noise in the environment, and to prevent such effects, photodetectors have to also include cooling devices that bring the instruments down to liquid nitrogen temperatures.

Current night-vision goggles and scopes are heavy and bulky. With the group’s new pyroelectric-based approach, NVDs could have the same sensitivity without the cooling weight.

The researchers also found that the films were sensitive beyond the range of current night-vision devices and could respond to wavelengths across the entire infrared spectrum. This suggests that the films could be incorporated into small, lightweight, and portable devices for various applications that require different infrared regions. For instance, when integrated into autonomous vehicle platforms, the films could enable cars to “see” pedestrians and vehicles in complete darkness or in foggy and rainy conditions. 

The film could also be used in gas sensors for real-time and on-site environmental monitoring, helping detect pollutants. In electronics, they could monitor heat changes in semiconductor chips to catch early signs of malfunctioning elements.

The team says the new lift-off method can be generalized to materials that may not themselves contain lead. In those cases, the researchers suspect that they can infuse Teflon-like lead atoms into the underlying substrate to induce a similar peel-off effect. For now, the team is actively working toward incorporating the pyroelectric films into a functional night-vision system.

“We envision that our ultrathin films could be made into high-performance night-vision goggles, considering its broad-spectrum infrared sensitivity at room-temperature, which allows for a lightweight design without a cooling system,” Zhang says. “To turn this into a night-vision system, a functional device array should be integrated with readout circuitry. Furthermore, testing in varied environmental conditions is essential for practical applications.”

This work was supported by the U.S. Air Force Office of Scientific Research.

New model predicts a chemical reaction’s point of no return

Chemists could use this quick computational method to design more efficient reactions that yield useful compounds, from fuels to pharmaceuticals.

When chemists design new chemical reactions, one useful piece of information involves the reaction’s transition state — the point of no return from which a reaction must proceed.

This information allows chemists to try to produce the right conditions that will allow the desired reaction to occur. However, current methods for predicting the transition state and the path that a chemical reaction will take are complicated and require a huge amount of computational power.

MIT researchers have now developed a machine-learning model that can make these predictions in less than a second, with high accuracy. Their model could make it easier for chemists to design chemical reactions that could generate a variety of useful compounds, such as pharmaceuticals or fuels.

“We’d like to be able to ultimately design processes to take abundant natural resources and turn them into molecules that we need, such as materials and therapeutic drugs. Computational chemistry is really important for figuring out how to design more sustainable processes to get us from reactants to products,” says Heather Kulik, the Lammot du Pont Professor of Chemical Engineering, a professor of chemistry, and the senior author of the new study.

Former MIT graduate student Chenru Duan PhD ’22, who is now at Deep Principle; former Georgia Tech graduate student Guan-Horng Liu, who is now at Meta; and Cornell University graduate student Yuanqi Du are the lead authors of the paper, which appears today in Nature Machine Intelligence.

Better estimates

For any given chemical reaction to occur, it must go through a transition state, which takes place when it reaches the energy threshold needed for the reaction to proceed. These transition states are so fleeting that they’re nearly impossible to observe experimentally.

As an alternative, researchers can calculate the structures of transition states using techniques based on quantum chemistry. However, that process requires a great deal of computing power and can take hours or days to calculate a single transition state.

“Ideally, we’d like to be able to use computational chemistry to design more sustainable processes, but this computation in itself is a huge use of energy and resources in finding these transition states,” Kulik says.

In 2023, Kulik, Duan, and others reported on a machine-learning strategy that they developed to predict the transition states of reactions. This strategy is faster than using quantum chemistry techniques, but still slower than what would be ideal because it requires the model to generate about 40 structures, then run those predictions through a “confidence model” to predict which states were most likely to occur.

One reason why that model needs to be run so many times is that it uses randomly generated guesses for the starting point of the transition state structure, then performs dozens of calculations until it reaches its final, best guess. These randomly generated starting points may be very far from the actual transition state, which is why so many steps are needed.

The researchers’ new model, React-OT, described in the Nature Machine Intelligence paper, uses a different strategy. In this work, the researchers trained their model to begin from an estimate of the transition state generated by linear interpolation — a technique that estimates each atom’s position by moving it halfway between its position in the reactants and in the products, in three-dimensional space.

“A linear guess is a good starting point for approximating where that transition state will end up,” Kulik says. “What the model’s doing is starting from a much better initial guess than just a completely random guess, as in the prior work.”

Because of this, it takes the model fewer steps and less time to generate a prediction. In the new study, the researchers showed that their model could make predictions with only about five steps, taking about 0.4 seconds. These predictions don’t need to be fed through a confidence model, and they are about 25 percent more accurate than the predictions generated by the previous model.

“That really makes React-OT a practical model that we can directly integrate to the existing computational workflow in high-throughput screening to generate optimal transition state structures,” Duan says.

“A wide array of chemistry”

To create React-OT, the researchers trained it on the same dataset that they used to train their older model. These data contain structures of reactants, products, and transition states, calculated using quantum chemistry methods, for 9,000 different chemical reactions, mostly involving small organic or inorganic molecules.

Once trained, the model performed well on other reactions from this set, which had been held out of the training data. It also performed well on other types of reactions that it hadn’t been trained on, and could make accurate predictions involving reactions with larger reactants, which often have side chains that aren’t directly involved in the reaction.

“This is important because there are a lot of polymerization reactions where you have a big macromolecule, but the reaction is occurring in just one part. Having a model that generalizes across different system sizes means that it can tackle a wide array of chemistry,” Kulik says.

The researchers are now working on training the model so that it can predict transition states for reactions between molecules that include additional elements, including sulfur, phosphorus, chlorine, silicon, and lithium.

“To quickly predict transition state structures is key to all chemical understanding,” says Markus Reiher, a professor of theoretical chemistry at ETH Zurich, who was not involved in the study. “The new approach presented in the paper could very much accelerate our search and optimization processes, bringing us faster to our final result. As a consequence, also less energy will be consumed in these high-performance computing campaigns. Any progress that accelerates this optimization benefits all sorts of computational chemical research.”

The MIT team hopes that other scientists will make use of their approach in designing their own reactions, and have created an app for that purpose.

“Whenever you have a reactant and product, you can put them into the model and it will generate the transition state, from which you can estimate the energy barrier of your intended reaction, and see how likely it is to occur,” Duan says.

The research was funded by the U.S. Army Research Office, the U.S. Department of Defense Basic Research Office, the U.S. Air Force Office of Scientific Research, the National Science Foundation, and the U.S. Office of Naval Research.

MIT engineers print synthetic “metamaterials” that are both strong and stretchy

A new method could enable stretchable ceramics, glass, and metals, for tear-proof textiles or stretchy semiconductors.

In metamaterials design, the name of the game has long been “stronger is better.”

Metamaterials are synthetic materials with microscopic structures that give the overall material exceptional properties. A huge focus has been in designing metamaterials that are stronger and stiffer than their conventional counterparts. But there’s a trade-off: The stiffer a material, the less flexible it is.

MIT engineers have now found a way to fabricate a metamaterial that is both strong and stretchy. The base material is typically highly rigid and brittle, but it is printed in precise, intricate patterns that form a structure that is both strong and flexible.

The key to the new material’s dual properties is a combination of stiff microscopic struts and a softer woven architecture. This microscopic “double network,” which is printed using a plexiglass-like polymer, produced a material that could stretch over four times its size without fully breaking. In comparison, the polymer in other forms has little to no stretch and shatters easily once cracked.

The researchers say the new double-network design can be applied to other materials, for instance to fabricate stretchy ceramics, glass, and metals. Such tough yet bendy materials could be made into tear-resistant textiles, flexible semiconductors, electronic chip packaging, and durable yet compliant scaffolds on which to grow cells for tissue repair.

“We are opening up this new territory for metamaterials,” says Carlos Portela, the Robert N. Noyce Career Development Associate Professor at MIT. “You could print a double-network metal or ceramic, and you could get a lot of these benefits, in that it would take more energy to break them, and they would be significantly more stretchable.”

Portela and his colleagues report their findings today in the journal Nature Materials. His MIT co-authors include first author James Utama Surjadi as well as Bastien Aymon and Molly Carton.

Inspired gel

Along with other research groups, Portela and his colleagues have typically designed metamaterials by printing or nanofabricating microscopic lattices using conventional polymers similar to plexiglass and ceramic. The specific pattern, or architecture, that they print can impart exceptional strength and impact resistance to the resulting metamaterial.

Several years ago, Portela was curious whether a metamaterial could be made from an inherently stiff material, but be patterned in a way that would turn it into a much softer, stretchier version.

“We realized that the field of metamaterials has not really tried to make an impact in the soft matter realm,” he says. “So far, we’ve all been looking for the stiffest and strongest materials possible.”

Instead, he looked for a way to synthesize softer, stretchier metamaterials. Rather than printing microscopic struts and trusses, similar to those of conventional lattice-based metamaterials, he and his team made an architecture of interwoven springs, or coils. They found that, while the material they used was itself stiff like plexiglass, the resulting woven metamaterial was soft and springy, like rubber.

“They were stretchy, but too soft and compliant,” Portela recalls.

In looking for ways to bulk up their softer metamaterial, the team found inspiration in an entirely different material: hydrogel. Hydrogels are soft, stretchy, Jell-O-like materials that are composed of mostly water and a bit of polymer structure. Researchers including groups at MIT have devised ways to make hydrogels that are both soft and stretchy, and also tough. They do so by combining polymer networks with very different properties, such as a network of molecules that is naturally stiff,  which gets chemically cross-linked with another molecular network that is inherently soft. Portela and his colleagues wondered whether such a double-network design could be adapted to metamaterials.

“That was our ‘aha’ moment,” Portela says. “We thought: Can we get inspiration from these hydrogels to create a metamaterial with similar stiff and stretchy properties?”

Strut and weave

For their new study, the team fabricated a metamaterial by combining two microscopic architectures. The first is a rigid, grid-like scaffold of struts and trusses. The second is a pattern of coils that weave around each strut and truss. Both networks are made from the same acrylic plastic and are printed in one go, using a high-precision, laser-based printing technique called two-photon lithography.

The researchers printed samples of the new double-network-inspired metamaterial, each measuring in size from several square microns to several square millimeters. They put the material through a series of stress tests, in which they attached either end of the sample to a specialized nanomechanical press and measured the force it took to pull the material apart. They also recorded high-resolution videos to observe the locations and ways in which the material stretched and tore as it was pulled apart.

They found their new double-network design was able stretch three times its own length, which also happened to be 10 times farther compared to a conventional lattice-patterned metamaterial printed with the same acrylic plastic. Portela says the new material’s stretchy resistance comes from the interactions between the material’s rigid struts and the messier, coiled weave as the material is stressed and pulled.

“Think of this woven network as a mess of spaghetti tangled around a lattice. As we break the monolithic lattice network, those broken parts come along for the ride, and now all this spaghetti gets entangled with the lattice pieces,” Portela explains. “That promotes more entanglement between woven fibers, which means you have more friction and more energy dissipation.”

In other words, the softer structure wound throughout the material’s rigid lattice takes on more stress thanks to multiple knots or entanglements promoted by the cracked struts. As this stress spreads unevenly through the material, an initial crack is unlikely to go straight through and quickly tear the material. What’s more, the team found that if they introduced strategic holes, or “defects,” in the metamaterial, they could further dissipate any stress that the material undergoes, making it even stretchier and more resistant to tearing apart.

“You might think this makes the material worse,” says study co-author Surjadi. “But we saw once we started adding defects, we doubled the amount of stretch we were able to do, and tripled the amount of energy that we dissipated. That gives us a material that’s both stiff and tough, which is usually a contradiction.”

The team has developed a computational framework that can help engineers estimate how a metamaterial will perform given the pattern of its stiff and stretchy networks. They envision such a blueprint will be useful in designing tear-proof textiles and fabrics.

“We also want to try this approach on more brittle materials, to give them multifunctionality,” Portela says. “So far we’ve talked of mechanical properties, but what if we could also make them conductive, or responsive to temperature? For that, the two networks could be made from different polymers, that respond to temperature in different ways, so that a fabric can open its pores or become more compliant when it’s warm and can be more rigid when it’s cold. That’s something we can explore now.”

This research was supported, in part, by the U.S. National Science Foundation, and the MIT MechE MathWorks Seed Fund. This work was performed, in part, through the use of MIT.nano’s facilities.

MIT D-Lab spinout provides emergency transportation during childbirth

Moving Health has developed an emergency transportation network using motorized ambulances in rural regions of Ghana.

Amama has lived in a rural region of northern Ghana all her life. In 2022, she went into labor with her first child. Women in the region traditionally give birth at home with the help of a local birthing attendant, but Amama experienced last-minute complications, and the decision was made to go to a hospital. Unfortunately, there were no ambulances in the community and the nearest hospital was 30 minutes away, so Amama was forced to take a motorcycle taxi, leaving her husband and caregiver behind.

Amama spent the next 30 minutes traveling over bumpy dirt roads to get to the hospital. She was in pain and afraid. When she arrived, she learned her child had not survived.

Unfortunately, Amama’s story is not unique. Around the world, more than 700 women die every day due to preventable pregnancy and childbirth complications. A lack of transportation to hospitals contributes to those deaths.

Moving Health was founded by MIT students to give people like Amama a safer way to get to the hospital. The company, which was started as part of a class at MIT D-Lab, works with local communities in rural Ghana to offer a network of motorized tricycle ambulances to communities that lack emergency transportation options.

The locally made ambulances are designed for the challenging terrain of rural Ghana, equipped with medical supplies, and have space for caregivers and family members.

“We’re providing the first rural-focused emergency transportation network,” says Moving Health CEO and co-founder Emily Young ’18. “We’re trying to provide emergency transportation coverage for less cost and with a vehicle tailored to local needs. When we first started, a report estimated there were 55 ambulances in the country of over 30 million people. Now, there is more coverage, but still the last mile areas of the country do not have access to reliable emergency transportation.”

Today, Moving Health’s ambulances and emergency transportation network cover more than 100,000 people in northern Ghana who previously lacked reliable medical transportation.

One of those people is Amama. During her most recent pregnancy, she was able to take a Moving Health ambulance to the hospital. This time, she traveled in a sanitary environment equipped with medical supplies and surrounded by loved ones. When she arrived, she gave birth to healthy twins.

From class project to company

Young and Sade Nabahe ’17, SM ’21 met while taking Course 2.722J (D-Lab: Design), which challenges students to think like engineering consultants on international projects. Their group worked on ways to transport pregnant women in remote areas of Tanzania to hospitals more safely and quickly. Young credits D-Lab instructor Matt McCambridge with helping students explore the project outside of class. Fellow Moving Health co-founder Eva Boal ’18 joined the effort the following year.

The early idea was to build a trailer that could attach to any motorcycle and be used to transport women. Following the early class projects, the students received funding from MIT’s PKG Center and the MIT Undergraduate Giving Campaign, which they used to travel to Tanzania in the following year’s Independent Activities Period (IAP). That’s when they built their first prototype in the field.

The founders realized they needed to better understand the problem from the perspective of locals and interviewed over 250 pregnant women, clinicians, motorcycle drivers, and birth attendants.

“We wanted to make sure the community was leading the charge to design what this solution should be. We had to learn more from the community about why emergency transportation doesn’t work in these areas,” Young says. “We ended up redesigning our vehicle completely.”

Following their graduation from MIT in 2018, the founders bought one-way tickets to Tanzania and deployed a new prototype. A big part of their plans was creating a product that could be manufactured by the community to support the local economy.

Nabahe and Boal left the company in 2020, but word spread of Moving Health’s mission, and Young received messages from organizations in about 15 different countries interested in expanding the company’s trials.

Young found the most alignment in Ghana, where she met two local engineers, Ambra Jiberu and Sufiyanu Imoro, who were building cars from scratch and inventing innovative agricultural technologies. With these two engineers joining the team, she was confident they had the team to build a solution in Ghana.

Taking what they’d learned in Tanzania, the new team set up hundreds of interviews and focus groups to understand the Ghanaian health system. The team redesigned their product to be a fully motorized tricycle based on the most common mode of transportation in northern Ghana. Today Moving Health focuses solely on Ghana, with local manufacturing and day-to-day operations led by Country Director and CTO Isaac Quansah.

Moving Health is focused on building a holistic emergency transportation network. To do this, Moving Health’s team sets up community-run dispatch systems, which involves organizing emergency phone numbers, training community health workers, dispatchers, and drivers, and integrating all of that within the existing health care system. The company also conducts educational campaigns in the communities it serves.

Moving Health officially launched its ambulances in 2023. The ambulance has an enclosed space for patients, family members, and medical providers and includes a removable stretcher along with supplies like first aid equipment, oxygen, IVs, and more. It costs about one-tenth the price of a traditional ambulance.

“We’ve built a really cool, small-volume manufacturing facility, led by our local engineering team, that has incredible quality,” Young says. “We also have an apprenticeship program that our two lead engineers run that allows young people to learn more hard skills. We want to make sure we’re providing economic opportunities in these communities. It’s very much a Ghanaian-made solution.”

Unlike the national ambulances, Moving Health’s ambulances are stationed in rural communities, at community health centers, to enable faster response times.

“When the ambulances are stationed in these people’s communities, at their local health centers, it makes all the difference,” Young says. “We’re trying to create an emergency transportation solution that is not only geared toward rural areas, but also focused on pregnancy and prioritizing women’s voices about what actually works in these areas.”

A lifeline for mothers

When Young first got to Ghana, she met Sahada, a local woman who shared the story of her first birth at the age of 18. Sahada had intended to give birth in her community with the help of a local birthing attendant, but she began experiencing so much pain during labor the attendant advised her to go to the nearest hospital. With no ambulances or vehicles in town, Sahada’s husband called a motorcycle driver, who took her alone on the three-hour drive to the nearest hospital.

“It was rainy, extremely muddy, and she was in a lot of pain,” Young recounts. “She was already really worried for her baby, and then the bike slips and they crash. They get back on, covered in mud, she has no idea if the baby survived, and finally gets to the maternity ward.”

Sahada was able to give birth to a healthy baby boy, but her story stuck with Young.

“The experience was extremely traumatic, and what’s really crazy is that counts as a successful birth statistic,” Young says. “We hear that kind of story a lot.”

This year, Moving Health plans to expand into a new region of northern Ghana. The team is also exploring other ways their network can provide health care to rural regions. But no matter how the company evolves, the team remain grateful to have seen their D-Lab project turn into such an impactful solution.

“Our long-term vision is to prove that this can work on a national level and supplement the existing health system,” Young says. “Then we’re excited to explore mobile health care outreach and other transportation solutions. We’ve always been focused on maternal health, but we’re staying cognizant of other community ideas that might be able to help improve health care more broadly.”

“Periodic table of machine learning” could fuel AI discovery

Researchers have created a unifying framework that can help scientists combine existing ideas to improve AI models or create new ones.

MIT researchers have created a periodic table that shows how more than 20 classical machine-learning algorithms are connected. The new framework sheds light on how scientists could fuse strategies from different methods to improve existing AI models or come up with new ones.

For instance, the researchers used their framework to combine elements of two different algorithms to create a new image-classification algorithm that performed 8 percent better than current state-of-the-art approaches.

The periodic table stems from one key idea: All these algorithms learn a specific kind of relationship between data points. While each algorithm may accomplish that in a slightly different way, the core mathematics behind each approach is the same.

Building on these insights, the researchers identified a unifying equation that underlies many classical AI algorithms. They used that equation to reframe popular methods and arrange them into a table, categorizing each based on the approximate relationships it learns.

Just like the periodic table of chemical elements, which initially contained blank squares that were later filled in by scientists, the periodic table of machine learning also has empty spaces. These spaces predict where algorithms should exist, but which haven’t been discovered yet.

The table gives researchers a toolkit to design new algorithms without the need to rediscover ideas from prior approaches, says Shaden Alshammari, an MIT graduate student and lead author of a paper on this new framework.

“It’s not just a metaphor,” adds Alshammari. “We’re starting to see machine learning as a system with structure that is a space we can explore rather than just guess our way through.”

She is joined on the paper by John Hershey, a researcher at Google AI Perception; Axel Feldmann, an MIT graduate student; William Freeman, the Thomas and Gerd Perkins Professor of Electrical Engineering and Computer Science and a member of the Computer Science and Artificial Intelligence Laboratory (CSAIL); and senior author Mark Hamilton, an MIT graduate student and senior engineering manager at Microsoft. The research will be presented at the International Conference on Learning Representations.

An accidental equation

The researchers didn’t set out to create a periodic table of machine learning.

After joining the Freeman Lab, Alshammari began studying clustering, a machine-learning technique that classifies images by learning to organize similar images into nearby clusters.

She realized the clustering algorithm she was studying was similar to another classical machine-learning algorithm, called contrastive learning, and began digging deeper into the mathematics. Alshammari found that these two disparate algorithms could be reframed using the same underlying equation.

“We almost got to this unifying equation by accident. Once Shaden discovered that it connects two methods, we just started dreaming up new methods to bring into this framework. Almost every single one we tried could be added in,” Hamilton says.

The framework they created, information contrastive learning (I-Con), shows how a variety of algorithms can be viewed through the lens of this unifying equation. It includes everything from classification algorithms that can detect spam to the deep learning algorithms that power LLMs.

The equation describes how such algorithms find connections between real data points and then approximate those connections internally.

Each algorithm aims to minimize the amount of deviation between the connections it learns to approximate and the real connections in its training data.

They decided to organize I-Con into a periodic table to categorize algorithms based on how points are connected in real datasets and the primary ways algorithms can approximate those connections.

“The work went gradually, but once we had identified the general structure of this equation, it was easier to add more methods to our framework,” Alshammari says.

A tool for discovery

As they arranged the table, the researchers began to see gaps where algorithms could exist, but which hadn’t been invented yet.

The researchers filled in one gap by borrowing ideas from a machine-learning technique called contrastive learning and applying them to image clustering. This resulted in a new algorithm that could classify unlabeled images 8 percent better than another state-of-the-art approach.

They also used I-Con to show how a data debiasing technique developed for contrastive learning could be used to boost the accuracy of clustering algorithms.

In addition, the flexible periodic table allows researchers to add new rows and columns to represent additional types of datapoint connections.

Ultimately, having I-Con as a guide could help machine learning scientists think outside the box, encouraging them to combine ideas in ways they wouldn’t necessarily have thought of otherwise, says Hamilton.

“We’ve shown that just one very elegant equation, rooted in the science of information, gives you rich algorithms spanning 100 years of research in machine learning. This opens up many new avenues for discovery,” he adds.

“Perhaps the most challenging aspect of being a machine-learning researcher these days is the seemingly unlimited number of papers that appear each year. In this context, papers that unify and connect existing algorithms are of great importance, yet they are extremely rare. I-Con provides an excellent example of such a unifying approach and will hopefully inspire others to apply a similar approach to other domains of machine learning,” says Yair Weiss, a professor in the School of Computer Science and Engineering at the Hebrew University of Jerusalem, who was not involved in this research.

This research was funded, in part, by the Air Force Artificial Intelligence Accelerator, the National Science Foundation AI Institute for Artificial Intelligence and Fundamental Interactions, and Quanta Computer.

Kripa Varanasi named faculty director of the Deshpande Center for Technological Innovation

The interfacial engineering expert and prolific entrepreneur will help faculty and students take breakthroughs from lab to market.

Kripa Varanasi, professor of mechanical engineering, was named faculty director of the MIT Deshpande Center for Technological Innovation, effective March 1.

“Kripa is widely recognized for his significant contributions in the field of interfacial science, thermal fluids, electrochemical systems, and advanced materials. It’s remarkable to see the tangible impact Kripa’s ventures have made across such a wide range of fields,” says Anantha P. Chandrakasan, dean of the School of Engineering, chief innovation and strategy officer, and Vannevar Bush Professor of Electrical Engineering and Computer Science. “From energy and water conservation to consumer products and agriculture, his solutions are making a real difference. The Deshpande Center will benefit greatly from both his entrepreneurial expertise and deep technical insight.”

The MIT Deshpande Center for Technological Innovation is an interdepartmental center that empowers MIT students and faculty to make a difference in the world by helping them bring their innovative technologies from the lab to the marketplace in the form of breakthrough products and new companies. The center was established through a gift from philanthropist Guruaj “Desh” Deshpande and his wife, Jaishree.

“Kripa brings an entrepreneurial spirit, innovative thinking, and commitment to mentorship that has always been central to the Deshpande Center’s mission,” says Deshpande. “He is exceptionally well-positioned to help the next generation of MIT innovators turn bold ideas into real-world solutions that make a difference.”

Varanasi has seen the Deshpande Center’s influence on the MIT community since its founding in 2002, when he was a graduate student.

“The Deshpande Center was founded when I was a graduate student, and it truly inspired many of us to think about entrepreneurship and commercialization — with Desh himself being an incredible role model,” says Varanasi. “Over the years, the center has built a storied legacy as a one-of-a-kind institution for propelling university-invented technologies to commercialization. Many amazing companies have come out of this program, shaping industries and making a real impact.”

A member of the MIT faculty since 2009, Varanasi leads the interdisciplinary Varanasi Research Group, which focuses on understanding physico-chemical and biological phenomena at the interfaces of matter. His group develops novel surfaces, materials, and technologies that improve efficiency and performance across industries, including energy, decarbonization, life sciences, water, agriculture, transportation, and consumer products.

In addition to his academic work, Varanasi is a prolific entrepreneur who has co-founded six companies, including AgZen, Alsym Energy, CoFlo Medical, Dropwise, Infinite Cooling, and LiquiGlide, which was a Deshpande Center grantee in 2009. These ventures aim to translate research breakthroughs into products with global reach.

His companies have been widely recognized for driving innovation across a range of industries. LiquiGlide, which produces frictionless liquid coatings, was named one of Time and Forbes’ “Best Inventions of the Year” in 2012. Infinite Cooling, which offers a technology to capture and recycle power plant water vapor, has won the U.S. Department of Energy’s National Cleantech University Prize and top prizes at MassChallenge and the MIT $100K competition. It is also a participating company at this year’s IdeaStream: Next Gen event, hosted by the Deshpande Center.

Another company that Varanasi co-founded, AgZen, is pioneering feedback optimization for agrochemical application that allows farmers to use 30-90 percent less pesticides and fertilizers while achieving 1-10 percent more yield. Meanwhile, Alsym Energy is advancing nonflammable, high-performance batteries for energy storage solutions that are lithium-free and capable of a wide range of storage durations. 

Throughout his career, Varanasi has been recognized for both research excellence and mentorship. His honors include the National Science Foundation CAREER Award, DARPA Young Faculty Award, SME Outstanding Young Manufacturing Engineer Award, ASME’s Bergles-Rohsenow Heat Transfer Award and Gustus L. Larson Memorial Award, Boston Business Journal’s 40 Under 40, and MIT’s Frank E. Perkins Award for Excellence in Graduate Advising​.

Varanasi earned his undergraduate degree in mechanical engineering from the Indian Institute of Technology Madras, and his master’s degree and PhD from MIT. Prior to joining the Institute’s faculty, he served as lead researcher and project leader at the GE Global Research Center, where he received multiple internal awards for innovation and technical excellence​.

"It’s an honor to lead the Deshpande Center, and in collaboration with the MIT community, I look forward to building on its incredible foundation — fostering bold ideas, driving real-world impact from cutting-edge innovations, and making it a powerhouse for commercialization,” adds Varanasi.

As faculty director, Varanasi will work closely with Deshpande Center executive director Rana Gupta to guide the center’s support of MIT faculty and students developing technology-based ventures.

“With Kripa’s depth and background, we will capitalize on the initiatives started with Angela Koehler. Kripa shares our vision to grow and expand the center’s capabilities to serve more of MIT,” adds Gupta.

Varanasi succeeds Angela Koehler, associate professor of biological engineering, who served as faculty director from July 2023 through March 2025.

“Angela brought fresh vision and energy to the center,” he says. “She expanded its reach, introduced new funding priorities in climate and life sciences, and re-imagined the annual IdeaStream event as a more robust launchpad for innovation. We’re deeply grateful for her leadership.”

Koehler, who was recently appointed faculty lead of the MIT Health and Life Sciences Collaborative, will continue to play a key role in the Institute’s innovation and entrepreneurship ecosystem​.

3D modeling you can feel

TactStyle, a system developed by CSAIL researchers, uses image prompts to replicate both the visual appearance and tactile properties of 3D models.

Essential for many industries ranging from Hollywood computer-generated imagery to product design, 3D modeling tools often use text or image prompts to dictate different aspects of visual appearance, like color and form. As much as this makes sense as a first point of contact, these systems are still limited in their realism due to their neglect of something central to the human experience: touch.

Fundamental to the uniqueness of physical objects are their tactile properties, such as roughness, bumpiness, or the feel of materials like wood or stone. Existing modeling methods often require advanced computer-aided design expertise and rarely support tactile feedback that can be crucial for how we perceive and interact with the physical world.

With that in mind, researchers at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) have created a new system for stylizing 3D models using image prompts, effectively replicating both visual appearance and tactile properties.

The CSAIL team’s “TactStyle” tool allows creators to stylize 3D models based on images while also incorporating the expected tactile properties of the textures. TactStyle separates visual and geometric stylization, enabling the replication of both visual and tactile properties from a single image input.

PhD student Faraz Faruqi, lead author of a new paper on the project, says that TactStyle could have far-reaching applications, extending from home decor and personal accessories to tactile learning tools. TactStyle enables users to download a base design — such as a headphone stand from Thingiverse — and customize it with the styles and textures they desire. In education, learners can explore diverse textures from around the world without leaving the classroom, while in product design, rapid prototyping becomes easier as designers quickly print multiple iterations to refine tactile qualities.

“You could imagine using this sort of system for common objects, such as phone stands and earbud cases, to enable more complex textures and enhance tactile feedback in a variety of ways,” says Faruqi, who co-wrote the paper alongside MIT Associate Professor Stefanie Mueller, leader of the Human-Computer Interaction (HCI) Engineering Group at CSAIL. “You can create tactile educational tools to demonstrate a range of different concepts in fields such as biology, geometry, and topography.”

Traditional methods for replicating textures involve using specialized tactile sensors — such as GelSight, developed at MIT — that physically touch an object to capture its surface microgeometry as a “heightfield.” But this requires having a physical object or its recorded surface for replication. TactStyle allows users to replicate the surface microgeometry by leveraging generative AI to generate a heightfield directly from an image of the texture.

On top of that, for platforms like the 3D printing repository Thingiverse, it’s difficult to take individual designs and customize them. Indeed, if a user lacks sufficient technical background, changing a design manually runs the risk of actually “breaking” it so that it can’t be printed anymore. All of these factors spurred Faruqi to wonder about building a tool that enables customization of downloadable models on a high level, but that also preserves functionality.

In experiments, TactStyle showed significant improvements over traditional stylization methods by generating accurate correlations between a texture’s visual image and its heightfield. This enables the replication of tactile properties directly from an image. One psychophysical experiment showed that users perceive TactStyle’s generated textures as similar to both the expected tactile properties from visual input and the tactile features of the original texture, leading to a unified tactile and visual experience.

TactStyle leverages a preexisting method, called “Style2Fab,” to modify the model’s color channels to match the input image’s visual style. Users first provide an image of the desired texture, and then a fine-tuned variational autoencoder is used to translate the input image into a corresponding heightfield. This heightfield is then applied to modify the model’s geometry to create the tactile properties.

The color and geometry stylization modules work in tandem, stylizing both the visual and tactile properties of the 3D model from a single image input. Faruqi says that the core innovation lies in the geometry stylization module, which uses a fine-tuned diffusion model to generate heightfields from texture images — something previous stylization frameworks do not accurately replicate.

Looking ahead, Faruqi says the team aims to extend TactStyle to generate novel 3D models using generative AI with embedded textures. This requires exploring exactly the sort of pipeline needed to replicate both the form and function of the 3D models being fabricated. They also plan to investigate “visuo-haptic mismatches” to create novel experiences with materials that defy conventional expectations, like something that appears to be made of marble but feels like it’s made of wood.

Faruqi and Mueller co-authored the new paper alongside PhD students Maxine Perroni-Scharf and Yunyi Zhu, visiting undergraduate student Jaskaran Singh Walia, visiting masters student Shuyue Feng, and assistant professor Donald Degraen of the Human Interface Technology (HIT) Lab NZ in New Zealand.

Norma Kamali is transforming the future of fashion with AI

The renowned designer embraces generative AI to preserve and propel her legacy.

What happens when a fashion legend taps into the transformative power of artificial intelligence? For more than five decades, fashion designer and entrepreneur Norma Kamali has pioneered bold industry shifts, creating iconic silhouettes worn by celebrities including Whitney Houston and Jessica Biel. Now, she is embracing a new frontier — one that merges creativity with algorithms and AI to redefine the future of her industry.

Through MIT Professional Education’s online “Applied Generative AI for Digital Transformation” course, which she completed in 2023, Kamali explored AI’s potential to serve as creative partner and ensure the longevity and evolution of her brand.

Kamali’s introduction to AI began with a meeting in Abu Dhabi, where industry experts, inspired by her Walmart collection, suggested developing an AI-driven fashion platform. Intrigued by the idea, but wary of the concept of “downloading her brain,” Kamali instead envisioned a system that could expand upon her 57-year archive — a closed-loop AI tool trained solely on her work. “I thought, AI could be my Karl Lagerfeld,” she says, referencing the designer’s reverence for archival inspiration.

To bring this vision to life, Kamali sought a deeper understanding of generative AI — so she headed to MIT Professional Education, an arm of MIT that has taught and inspired global professionals for more than 75 years. “I wasn’t sure how much I could actually do,” she recalls. “I had all these preconceived notions, but the more I learned, the more ideas I had.” Initially intimidated by the technical aspects of AI, she persevered, diving into prompts and training data, and exploring its creative potential. “I was determined,” she says. “And then suddenly, I was playing.”

Experimenting with her proprietary AI model, created by Maison Meta, Kamali used AI to reinterpret one of her signature styles — black garments adorned with silver studs. By prompting AI with iterations of her existing silhouettes, she witnessed unexpected and thrilling results. “It was magic,” she says. “Art, technology, and fashion colliding in ways I never imagined.” Even AI’s so-called “hallucinations” — distortions often seen as errors — became a source of inspiration. “Some of the best editorial fashion is absurd,” she notes. “AI-generated anomalies created entirely new forms of art.”

Kamali’s approach to AI reflects a broader shift across industries, where technology is not just a tool but a catalyst for reinvention. Bhaskar Pant, executive director of MIT Professional Education, underscores this transformation. “While everyone is speculating about the impact of AI, we are committed to advancing AI’s role in helping industries and leaders achieve breakthroughs, higher levels of productivity, and, as in this case, unleash creativity. Professionals must be empowered to harness AI’s potential in ways that not only enhance their work, but redefine what’s possible. Norma’s journey is a testament to the power of lifelong learning — demonstrating that innovation is ageless, fueled by curiosity and ambition.”

The experience also deepened Kamali’s perspective on AI’s role in the creative process. “AI doesn’t have a heartbeat,” she asserts. “It can’t replace human passion. But it can enhance creativity in ways we’re only beginning to understand.” Kamali also addressed industry fears about job displacement, arguing that the technology is already reshaping fashion’s labor landscape. “Sewing talent is harder to find. Designers need new tools to adapt.”

Beyond its creative applications, Kamali sees AI as a vehicle for sustainability. A longtime advocate for reducing dry cleaning — a practice linked to chemical exposure — she envisions AI streamlining fabric selection, minimizing waste, and enabling on-demand production. “Imagine a system where you design your wedding dress online, and a robot constructs it, one garment at a time,” she says. “The possibilities are endless.”

Abel Sanchez, MIT research scientist and lead instructor for MIT Professional Education’s Applied Generative AI for Digital Transformation course, emphasizes the transformative potential of AI across industries. “AI is a force reshaping the foundations of every sector, including fashion. Generative AI is unlocking unprecedented digital transformation opportunities, enabling organizations to rethink processes, design, and customer engagement. Norma is at the forefront of this shift, exploring how AI can propel the fashion industry forward, spark new creative frontiers, and redefine how designers interact with technology.”

Kamali’s experience in the course sparked an ongoing exchange of ideas with Sanchez, further fueling her curiosity. “AI is evolving so fast, I know I’ll need to go back,” she says. “MIT gave me the foundation, but this is just the beginning.” For those hesitant to embrace AI, she offers a striking analogy: “Imagine landing in a small town, in a foreign country, where you don’t speak the language, don’t recognize the food, and feel completely lost. That’s what it will be like if you don’t learn AI. The train has left the station — it’s time to get on board.”

With her AI-generated designs now featured on her website alongside her traditional collections, Kamali is proving that technology and creativity aren’t at odds — they’re collaborators. And as she continues to push the boundaries of both, she remains steadfast in her belief: “Learning is the adventure of life. Why stop now?”

Julie Lucas to step down as MIT’s vice president for resource development

Lucas has led MIT’s fundraising since 2014, including the record-setting MIT Campaign for a Better World.

Julie A. Lucas has decided to step down as MIT’s vice president for resource development, President Sally Kornbluth announced today. Lucas has set her last day as June 30, which coincides with the close of the Institute’s fiscal year, to ensure a smooth transition for staff and donors. 

Lucas has led fundraising at the Institute since 2014. During that time, MIT’s average annual fundraising has increased 96 percent to $611 million, up from $313 million in the decade before her arrival. MIT’s annual fundraising totals have exceeded the Institute’s annual $500 million fundraising target for nine straight fiscal years, including a few banner fiscal years with results upward of $700 to $900 million.

“Before I arrived at MIT, Julie built a fundraising operation worthy of the Institute’s world-class stature,” Kornbluth says. “I have seen firsthand how Julie’s expertise, collegial spirit, and commitment to our mission resonates with alumni and friends, motivating them to support the Institute.”

Lucas spearheaded the MIT Campaign for a Better World, which concluded in 2021 and raised $6.2 billion, setting a record as the Institute’s largest fundraising initiative. Emphasizing the Institute’s hands-on approach to solving the world’s toughest challenges — and centered on its strengths in education, research, and innovation — the campaign attracted participation from more than 112,000 alumni and friends around the globe, including nearly 56,000 new donors.  

“From the moment I met Julie Lucas, I knew she was the right person to serve as MIT’s chief philanthropic leader of our capital campaign,” says MIT President Emeritus L. Rafael Reif. “Julie is both a ‘maker’ and a ‘doer,’ well attuned to our ‘mens et manus’ motto. The Institute has benefited immensely from her impressive set of skills and ability to convey a coherent message that has inspired and motivated alumni and friends, foundations and corporations, to support MIT.” 

Under Lucas, MIT’s Office of Resource Development (RD) created new fundraising programs and processes, and introduced expanded ways of giving. For example, RD established the Institute’s planned giving program, which supports donors who want to make a lasting impact at MIT through philanthropic vehicles such as bequests, retirement plan distributions, life-income gifts, and gifts of complex assets. She also played a lead role in creating a donor-advised fund at MIT that, since its inception in 2017, has seen almost $120 million in contributions.  

“Julie is a remarkable fundraiser and leader — and when it comes to Julie’s leadership of Resource Development, the results speak for themselves,” says Mark Gorenberg ’76, chair of the MIT Corporation, who has participated in multiple MIT committees and campaigns over the last two decades. “These tangible fundraising outcomes have helped to facilitate innovations and discoveries, expand educational programs and facilities, support faculty and researchers, and ensure that an MIT education is affordable and accessible to the brightest minds from around the world.”

Prior to joining MIT, Lucas served in senior fundraising roles at the University of Southern California and Fordham Law School, as well as New York University and its business and law schools. 

While Lucas readies herself for the next phase in her career, she remains grateful for her time at the Institute. 

“Philanthropy is a powerful fuel for good in our world,” Lucas says. “My decision to step down was difficult. I feel honored and thankful that my work — and the work of the team of professionals I lead in Resource Development — has helped continue the amazing trajectory of MIT research and innovation that benefits all of us by solving humanity’s greatest challenges, both now and in the future.”

Lucas currently serves on the steering committee and is the immediate past chair of CASE 50, the Council for Advancement and Support of Education group that includes the top 50 fundraising institutions in the world. In addition, she is chair of the 2025 CASE Summit for Leaders in Advancement and a founding member of Aspen Leadership Group’s Chief Development Officer Network.

Astronomers discover a planet that’s rapidly disintegrating, producing a comet-like tail

The small and rocky lava world sheds an amount of material equivalent to the mass of Mount Everest every 30.5 hours.

MIT astronomers have discovered a planet some 140 light-years from Earth that is rapidly crumbling to pieces.

The disintegrating world is about the mass of Mercury, although it circles about 20 times closer to its star than Mercury does to the sun, completing an orbit every 30.5 hours. At such close proximity to its star, the planet is likely covered in magma that is boiling off into space. As the roasting planet whizzes around its star, it is shedding an enormous amount of surface minerals and effectively evaporating away.

The astronomers spotted the planet using NASA’s Transiting Exoplanet Survey Satellite (TESS), an MIT-led mission that monitors the nearest stars for transits, or periodic dips in starlight that could be signs of orbiting exoplanets. The signal that tipped the astronomers off was a peculiar transit, with a dip that fluctuated in depth every orbit.

The scientists confirmed that the signal is of a tightly orbiting rocky planet that is trailing a long, comet-like tail of debris.

“The extent of the tail is gargantuan, stretching up to 9 million kilometers long, or roughly half of the planet’s entire orbit,” says Marc Hon, a postdoc in MIT’s Kavli Institute for Astrophysics and Space Research.

It appears that the planet is disintegrating at a dramatic rate, shedding an amount of material equivalent to one Mount Everest each time it orbits its star. At this pace, given its small mass, the researchers predict that the planet may completely disintegrate in about 1 million to 2 million years.

“We got lucky with catching it exactly when it’s really going away,” says Avi Shporer, a collaborator on the discovery who is also at the TESS Science Office. “It’s like on its last breath.”

Hon and Shporer, along with their colleagues, have published their results today in the Astrophysical Journal Letters. Their MIT co-authors include Saul Rappaport, Andrew Vanderburg, Jeroen Audenaert, William Fong, Jack Haviland, Katharine Hesse, Daniel Muthukrishna, Glen Petitpas, Ellie Schmelzer, Sara Seager, and George Ricker, along with collaborators from multiple other institutions.

Roasting away

The new planet, which scientists have tagged as BD+05 4868 Ab, was detected almost by happenstance.

“We weren’t looking for this kind of planet,” Hon says. “We were doing the typical planet vetting, and I happened to spot this signal that appeared very unusual.”

The typical signal of an orbiting exoplanet looks like a brief dip in a light curve, which repeats regularly, indicating that a compact body such as a planet is briefly passing in front of, and temporarily blocking, the light from its host star.

This typical pattern was unlike what Hon and his colleagues detected from the host star BD+05 4868 A, located in the constellation of Pegasus. Though a transit appeared every 30.5 hours, the brightness took much longer to return to normal, suggesting a long trailing structure still blocking starlight. Even more intriguing, the depth of the dip changed with each orbit, suggesting that whatever was passing in front of the star wasn’t always the same shape or blocking the same amount of light.

“The shape of the transit is typical of a comet with a long tail,” Hon explains. “Except that it’s unlikely that this tail contains volatile gases and ice as expected from a real comet — these would not survive long at such close proximity to the host star. Mineral grains evaporated from the planetary surface, however, can linger long enough to present such a distinctive tail.”

Given its proximity to its star, the team estimates that the planet is roasting at around 1,600 degrees Celsius, or close to 3,000 degrees Fahrenheit. As the star roasts the planet, any minerals on its surface are likely boiling away and escaping into space, where they cool into a long and dusty tail.

The dramatic demise of this planet is a consequence of its low mass, which is between that of Mercury and the moon. More massive terrestrial planets like the Earth have a stronger gravitational pull and therefore can hold onto their atmospheres. For BD+05 4868 Ab, the researchers suspect there is very little gravity to hold the planet together.

“This is a very tiny object, with very weak gravity, so it easily loses a lot of mass, which then further weakens its gravity, so it loses even more mass,” Shporer explains. “It’s a runaway process, and it’s only getting worse and worse for the planet.”

Mineral trail

Of the nearly 6,000 planets that astronomers have discovered to date, scientists know of only three other disintegrating planets beyond our solar system. Each of these crumbling worlds were spotted over 10 years ago using data from NASA’s Kepler Space Telescope. All three planets were spotted with similar comet-like tails. BD+05 4868 Ab has the longest tail and the deepest transits out of the four known disintegrating planets to date.

“That implies that its evaporation is the most catastrophic, and it will disappear much faster than the other planets,” Hon explains.

The planet’s host star is relatively close, and thus brighter than the stars hosting the other three disintegrating planets, making this system ideal for further observations using NASA’s James Webb Space Telescope (JWST), which can help determine the mineral makeup of the dust tail by identifying which colors of infrared light it absorbs.

This summer, Hon and graduate student Nicholas Tusay from Penn State University will lead observations of BD+05 4868 Ab using JWST. “This will be a unique opportunity to directly measure the interior composition of a rocky planet, which may tell us a lot about the diversity and potential habitability of terrestrial planets outside our solar system,” Hon says.

The researchers also will look through TESS data for signs of other disintegrating worlds.

“Sometimes with the food comes the appetite, and we are now trying to initiate the search for exactly these kinds of objects,” Shporer says. “These are weird objects, and the shape of the signal changes over time, which is something that’s difficult for us to find. But it’s something we’re actively working on.”

This work was supported, in part, by NASA.

MIT’s McGovern Institute is shaping brain science and improving human lives on a global scale

A quarter century after its founding, the McGovern Institute reflects on its discoveries in the areas of neuroscience, neurotechnology, artificial intelligence, brain-body connections, and therapeutics.

In 2000, Patrick J. McGovern ’59 and Lore Harp McGovern made an extraordinary gift to establish the McGovern Institute for Brain Research at MIT, driven by their deep curiosity about the human mind and their belief in the power of science to change lives. Their $350 million pledge began with a simple yet audacious vision: to understand the human brain in all its complexity, and to leverage that understanding for the betterment of humanity.
 
Twenty-five years later, the McGovern Institute stands as a testament to the power of interdisciplinary collaboration, continuing to shape our understanding of the brain and improve the quality of life for people worldwide.

In the beginning

“This is, by any measure, a truly historic moment for MIT,” said MIT’s 15th president, Charles M. Vest, during his opening remarks at an event in 2000 to celebrate the McGovern gift agreement. “The creation of the McGovern Institute will launch one of the most profound and important scientific ventures of this century in what surely will be a cornerstone of MIT scientific contributions from the decades ahead.”
 
Vest tapped Phillip A. Sharp, MIT Institute professor emeritus of biology and Nobel laureate, to lead the institute, and appointed six MIT professors — Emilio Bizzi, Martha Constantine-Paton, Ann Graybiel PhD ’71, H. Robert Horvitz ’68, Nancy Kanwisher ’80, PhD ’86, and Tomaso Poggio — to represent its founding faculty.  Construction began in 2003 on Building 46, a 376,000 square foot research complex at the northeastern edge of campus. MIT’s new “gateway from the north” would eventually house the McGovern Institute, the Picower Institute for Learning and Memory, and MIT’s Department of Brain and Cognitive Sciences.

Robert Desimone, the Doris and Don Berkey Professor of Neuroscience at MIT, succeeded Sharp as director of the McGovern Institute in 2005, and assembled a distinguished roster of 22 faculty members, including a Nobel laureate, a Breakthrough Prize winner, two National Medal of Science/Technology awardees, and 15 members of the American Academy of Arts and Sciences.
 
A quarter century of innovation

On April 11, 2025, the McGovern Institute celebrated its 25th anniversary with a half-day symposium featuring presentations by MIT Institute Professor Robert Langer, alumni speakers from various McGovern labs, and Desimone, who is in his 20th year as director of the institute.

Desimone highlighted the institute’s recent discoveries, including the development of the CRISPR genome-editing system, which has culminated in the world’s first CRISPR gene therapy approved for humans — a remarkable achievement that is ushering in a new era of transformative medicine. In other milestones, McGovern researchers developed the first prosthetic limb fully controlled by the body’s nervous system; a flexible probe that taps into gut-brain communication; an expansion microscopy technique that paves the way for biology labs around the world to perform nanoscale imaging; and advanced computational models that demonstrate how we see, hear, use language, and even think about what others are thinking. Equally transformative has been the McGovern Institute’s work in neuroimaging, uncovering the architecture of human thought and establishing markers that signal the early emergence of mental illness, before symptoms even appear.

Synergy and open science
 
“I am often asked what makes us different from other neuroscience institutes and programs around the world,” says Desimone. “My answer is simple. At the McGovern Institute, the whole is greater than the sum of its parts.”
 
Many discoveries at the McGovern Institute have depended on collaborations across multiple labs, ranging from biological engineering to human brain imaging and artificial intelligence. In modern brain research, significant advances often require the joint expertise of people working in neurophysiology, behavior, computational analysis, neuroanatomy, and molecular biology. More than a dozen different MIT departments are represented by McGovern faculty and graduate students, and this synergy has led to insights and innovations that are far greater than what any single discipline could achieve alone.
 
Also baked into the McGovern ethos is a spirit of open science, where newly developed technologies are shared with colleagues around the world. Through hospital partnerships for example, McGovern researchers are testing their tools and therapeutic interventions in clinical settings, accelerating their discoveries into real-world solutions.

The McGovern legacy  

Hundreds of scientific papers have emerged from McGovern labs over the past 25 years, but most faculty would argue that it’s the people — the young researchers — that truly define the McGovern Institute. Award-winning faculty often attract the brightest young minds, but many McGovern faculty also serve as mentors, creating a diverse and vibrant scientific community that is setting the global standard for brain research and its applications. Kanwisher, for example, has guided more than 70 doctoral students and postdocs who have gone on to become leading scientists around the world. Three of her former students, Evelina Fedorenko PhD ’07, Josh McDermott PhD ’06, and Rebecca Saxe PhD ’03, the John W. Jarve (1978) Professor of Brain and Cognitive Sciences, are now her colleagues at the McGovern Institute. Other McGovern alumni shared stories of mentorship, science, and real-world impact at the 25th anniversary symposium.

Looking to the future, the McGovern community is more committed than ever to unraveling the mysteries of the brain and making a meaningful difference in lives of individuals at a global scale.
 
“By promoting team science, open communication, and cross-discipline partnerships,” says institute co-founder Lore Harp McGovern, “our culture demonstrates how individual expertise can be amplified through collective effort. I am honored to be the co-founder of this incredible institution — onward to the next 25 years!”

Equipping living cells with logic gates to fight cancer

Founded by MIT researchers, Senti Bio is giving immune cells the ability to distinguish between healthy and cancerous cells.

One of the most exciting developments in cancer treatment is a wave of new cell therapies that train a patient’s immune system to attack cancer cells. Such therapies have saved the lives of patients with certain aggressive cancers and few other options. Most of these therapies work by teaching immune cells to recognize and attack specific proteins on the surface of cancer cells.

Unfortunately, most proteins found on cancer cells aren’t unique to tumors. They’re also often present on healthy cells, making it difficult to target cancer aggressively without triggering dangerous attacks on other tissue. The problem has limited the application of cell therapies to a small subset of cancers.

Now Senti Bio is working to create smarter cell therapies using synthetic biology. The company, which was founded by former MIT faculty member and current MIT Research Associate Tim Lu ’03, MEng ’03, PhD ’08 and Professor James Collins, is equipping cells with gene circuits that allow the cells to sense and respond to their environments.

Lu, who studied computer science as an undergraduate at MIT, describes Senti’s approach as programming living cells to behave more like computers — responding to specific biological cues with “if/then” logic, just like computer code.

“We have innovated a cell therapy that says, ‘Kill anything displaying the cancer target, but spare anything that has this healthy target,’” Lu explains. “Despite the promise of certain cancer targets, problems can arise when they are expressed on healthy cells that we want to protect. Our logic gating technology was designed to recognize and avoid killing those healthy cells, which introduces a whole spectrum of additional cancers that don’t have a single clean target that we can now potentially address. That’s the power of embedding these cells with logic.”

The company’s lead drug candidate aims to help patients with acute myeloid leukemia (AML) who have experienced a relapse or are unresponsive to other therapies. The prognosis for such patients is poor, but early data from the company’s first clinical trial showed that two of the first three patients Senti treated experienced complete remission, where subsequent bone marrow tests couldn’t detect a single cancer cell.

“It’s essentially one of the best responses you can get in this disease, so we were really excited to see that,” says Lu, who served on MIT’s faculty until leaving to lead Senti in 2022.

Senti is expecting to release more patient data at the upcoming American Association for Cancer Research (AACR) meeting at the end of April.

“Our groundbreaking work at Senti is showing that one can harness synthetic biology technologies to create programmable, smart medicines for treating patients with cancer,” says Collins, who is currently MIT’s Termeer Professor of Medical Engineering and Science. “This is tremendously exciting and demonstrates how one can utilize synthetic biological circuits, in this case logic gates, to design highly effective, next-generation living therapeutics.”

From computer science to cancer care

Lu was inspired as an undergraduate studying electrical engineering and computer science by the Human Genome Project, an international race to sequence the human genome. Later, he entered the Harvard-MIT Health Sciences and Technology (HST) program, through which he earned a PhD from MIT in electrical and biomedical imaging and an MD from Harvard. During that time, he worked in the lab of his eventual Senti co-founder James Collins, a synthetic biology pioneer.

In 2010, Lu joined MIT as an assistant professor with a joint appointment in the departments of Biological Engineering and of Electrical Engineering and Computer Science. Over the course of the next 14 years, Lu led the Synthetic Biology Group at MIT and started several biotech companies, including Engine Biosciences and Tango Therapeutics, which are also developing precision cancer treatments.

In 2015, a group of researchers including Lu and MIT Institute Professor Phillip Sharp published research showing they could use gene circuits to get immune cells to selectively respond to tumor cells in their environment.

“One of the first things we published focused on the idea of logic gates in living cells,” Lu says. “A computer has ‘and’ gates, ‘or’ gates, and ‘not’ gates that allow it to perform computations, and we started publishing gene circuits that implement logic into living cells. These allow cells to detect signals and then make logical decisions like, ‘Should we switch on or off?’”

Around that time, the first cell therapies and cancer immunotherapies began to be approved by the Food and Drug Administration, and the founders saw their technology as a way to take those approaches to the next level. They officially founded Senti Bio in 2016, with Lu taking a sabbatical from MIT to serve as CEO.

The company licensed technology from MIT and subsequently advanced the cellular logic gates so they could work with multiple types of engineered immune cells, including T cells and “natural killer” cells. Senti’s cells can respond to specific proteins that exist on the surface of both cancer and healthy cells to increase selectivity.

“We can now create a cell therapy where the cell makes a decision as to whether to kill a cancer cell or spare a healthy cell even when those cells are right next to each other,” Lu says. “If you can’t distinguish between cancerous and healthy cells, you get unwanted side effects, or you may not be able to hit the cancer as hard as you’d like. But once you can do that, there’s a lot of ways to maximize your firepower against the cancer cells.”

Hope for patients

Senti’s lead clinical trial is focusing on patients with relapsed or refractory blood cancers, including AML.

“Obviously the most important thing is getting a good response for patients,” Lu says. “But we’re also doing additional scientific work to confirm that the logic gates are working the way we expect them to in humans. Based on that information, we can then deploy logic gates into additional therapeutic indications such as solid tumors, where you have a lot of the same problems with finding a target.”

Another company that has partnered with Senti to use some of Senti’s technology also has an early clinical trial underway in liver cancer. Senti is also partnering with other companies to apply its gene circuit technology in areas like regenerative medicine and neuroscience.

“I think this is broader than just cell therapies,” Lu says. “We believe if we can prove this out in AML, it will lead to a fundamentally new way of diagnosing and treating cancer, where we’re able to definitively identify and target cancer cells and spare healthy cells. We hope it will become a whole new class of medicines moving forward.”

Making AI-generated code more accurate in any language

A new technique automatically guides an LLM toward outputs that adhere to the rules of whatever programming language or other format is being used.

Programmers can now use large language models (LLMs) to generate computer code more quickly. However, this only makes programmers’ lives easier if that code follows the rules of the programming language and doesn’t cause a computer to crash.

Some methods exist for ensuring LLMs conform to the rules of whatever language they are generating text in, but many of these methods either distort the model’s intended meaning or are too time-consuming to be feasible for complex tasks.

A new approach developed by researchers at MIT and elsewhere automatically guides an LLM to generate text that adheres to the rules of the relevant language, such as a particular programming language, and is also error-free. Their method allows an LLM to allocate efforts toward outputs that are most likely to be valid and accurate, while discarding unpromising outputs early in the process. This probabilistic approach boosts computational efficiency.

Due to these efficiency gains, the researchers’ architecture enabled small LLMs to outperform much larger models in generating accurate, properly structured outputs for several real-world use cases, including molecular biology and robotics.

In the long run, this new architecture could help nonexperts control AI-generated content. For instance, it could allow businesspeople to write complex queries in SQL, a language for database manipulation, using only natural language prompts.

“This work has implications beyond research. It could improve programming assistants, AI-powered data analysis, and scientific discovery tools by ensuring that AI-generated outputs remain both useful and correct,” says João Loula, an MIT graduate student and co-lead author of a paper on this framework.

Loula is joined on the paper by co-lead authors Benjamin LeBrun, a research assistant at the Mila-Quebec Artificial Intelligence Institute, and Li Du, a graduate student at John Hopkins University; co-senior authors Vikash Mansinghka ’05, MEng ’09, PhD ’09, a principal research scientist and leader of the Probabilistic Computing Project in the MIT Department of Brain and Cognitive Sciences; Alexander K. Lew SM ’20, an assistant professor at Yale University; Tim Vieira, a postdoc at ETH Zurich; and Timothy J. O’Donnell, an associate professor at McGill University and a Canada CIFAR AI Chair at Mila, who led the international team; as well as several others. The research will be presented at the International Conference on Learning Representations.

Enforcing structure and meaning

One common approach for controlling the structured text generated by LLMs involves checking an entire output, like a block of computer code, to make sure it is valid and will run error-free. If not, the user must start again, racking up computational resources.

On the other hand, a programmer could stop to check the output along the way. While this can ensure the code adheres to the programming language and is structurally valid, incrementally correcting the code may cause it to drift from the meaning the user intended, hurting its accuracy in the long run.

“It is much easier to enforce structure than meaning. We can quickly check whether something is in the right programming language, but to check its meaning you have to execute the code. Our work is also about dealing with these different types of information,” Loula says.

The researchers’ approach involves engineering knowledge into the LLM to steer it toward the most promising outputs. These outputs are more likely to follow the structural constraints defined by a user, and to have the meaning the user intends.

“We are not trying to train an LLM to do this. Instead, we are engineering some knowledge that an expert would have and combining it with the LLM’s knowledge, which offers a very different approach to scaling than you see in deep learning,” Mansinghka adds.

They accomplish this using a technique called sequential Monte Carlo, which enables parallel generation from an LLM to compete with each other. The model dynamically allocates resources to different threads of parallel computation based on how promising their output appears.

Each output is given a weight that represents how likely it is to be structurally valid and semantically accurate. At each step in the computation, the model focuses on those with higher weights and throws out the rest.

In a sense, it is like the LLM has an expert looking over its shoulder to ensure it makes the right choices at each step, while keeping it focused on the overall goal. The user specifies their desired structure and meaning, as well as how to check the output, then the researchers’ architecture guides the LLM to do the rest.

“We’ve worked out the hard math so that, for any kinds of constraints you’d like to incorporate, you are going to get the proper weights. In the end, you get the right answer,” Loula says.

Boosting small models

To test their approach, they applied the framework to LLMs tasked with generating four types of outputs: Python code, SQL database queries, molecular structures, and plans for a robot to follow.

When compared to existing approaches, the researchers’ method performed more accurately while requiring less computation.

In Python code generation, for instance, the researchers’ architecture enabled a small, open-source model to outperform a specialized, commercial closed-source model that is more than double its size.

“We are very excited that we can allow these small models to punch way above their weight,” Loula says.

Moving forward, the researchers want to use their technique to control larger chunks of generated text, rather than working one small piece at a time. They also want to combine their method with learning, so that as they control the outputs a model generates, it learns to be more accurate.

In the long run, this project could have broader applications for non-technical users. For instance, it could be combined with systems for automated data modeling, and querying generative models of databases.

The approach could also enable machine-assisted data analysis systems, where the user can converse with software that accurately models the meaning of the data and the questions asked by the user, adds Mansinghka.

“One of the fundamental questions of linguistics is how the meaning of words, phrases, and sentences can be grounded in models of the world, accounting for uncertainty and vagueness in meaning and reference. LLMs, predicting likely token sequences, don’t address this problem. Our paper shows that, in narrow symbolic domains, it is technically possible to map from words to distributions on grounded meanings. It’s a small step towards deeper questions in cognitive science, linguistics, and artificial intelligence needed to understand how machines can communicate about the world like we do,” says O’Donnell.

This research is funded and supported, in part, by the Canada CIFAR AI Chairs Program, the MIT Quest for Intelligence, and Convergent Research. 

Student spotlight: YongYan (Crystal) Liang

The senior, majoring in electrical engineering and computer science, has participated in SuperUROP, NEET, MISTI GTL, and multiple labs focusing on biological EECS.

The following is part of a series of short interviews from the Department of Electrical Engineering and Computer Science (EECS). Each spotlight features a student answering questions about themselves and life at MIT. Today’s interviewee, YongYan (Crystal) Liang, is a senior majoring in EECS with a particular interest in bioengineering and medical devices — which led her to join the Living Machines track as part of New Engineering Education Transformation (NEET) at MIT. An Advanced Undergraduate Research Opportunities Program (SuperUROP) scholar, Liang was supported by the Nadar Foundation Undergraduate Research and Innovation Scholar award for her project, which focused on steering systems for intravascular drug delivery devices. A world traveler, Liang has also taught robotics to students in MISTI Global Teaching Labs (GTL) programs in Korea and Germany — and is involved with the Terrascope and MedLinks communities. 

Q: Do you have a bucket list? If so, share one or two of the items on it.

A: I’d like to be proficient in at least five languages in a conversational sense (though probably not at a working proficiency level). Currently, I’m fluent in English, and can speak Cantonese and Mandarin. I also have a 1,600-plus day Duolingo streak where I’m trying to learn the foundations of a few languages, including German, Korean, Japanese, and Russian. 

Another bucket list item I have is to try every martial art/combat sport there is, even if it’s just an introduction class. So far, I’ve practiced taekwondo for a few years, taken a few lessons in boxing/kickboxing, and dabbled in beginners’ classes for karate, Krav Maga, and Brazilian jiujitsu. I’ll probably try to take judo, aikido, and other classes this upcoming year! It would also be pretty epic to be a fourth dan black belt one day, though that may take a decade or two.

Q: If you had to teach a really in-depth class about one niche topic, what would you pick?

A: Personally, I think artificial organs are pretty awesome! I would probably talk about the fusion of engineering with our bodies, and organ enhancement. This might include adding functionalities and possible organ regeneration, so that those waiting for organ donations can be helped without being morally conflicted by waiting for another person’s downfall. I’ve previously done research in several BioEECS-related labs that I’d love to talk about as well. This includes the Traverso Lab at Pappalardo, briefly in the Edelman Lab at the [Institute for Medical Engineering and Science], the Langer Lab at the Koch Institute of Integrative Cancer Research, as well as in the MIT Media Lab with the Conformable Decoders and BioMechatronics group. I also contributed to a recently published paper related to gastrointestinal devices: OSIRIS.  

Q: If you suddenly won the lottery, what would you spend some of the money on? 

A: I would make sure my mom got most of the money. The first thing we’d do is probably go house shopping around the world and buy properties in great travel destinations — then go around and live in said properties. We would do this on rotation with our friends until we ran out of money, then put the properties up for rent and use the money to open a restaurant with my mom’s recipes as the menu. Then I’d get to eat her food forever.

Q: What do you believe is an underrated invention or technology?

A: I feel like many people wear glasses or put on contacts nowadays and don’t really think twice about it, glossing over how cool it is that we can fix bad sight and how critical sight is for our survival. If a zombie apocalypse happened and my glasses broke, it would be over for me. And don’t get me started about the invention of the indoor toilet and plumbing systems!

Q: Are you a re-reader or a re-watcher? If so, what are your comfort books, shows, or movies? 

A: I’m both a re-reader and a re-watcher! I have a lot of fun binging webtoons and dramas. I’m also a huge Marvel fan, although recently, it’s been a hit or miss. Action and romcoms are my kinda vibes, and occasionally I do watch some anime. If I’m bored I usually re-watch some [Marvel Cinematic Universe] movies, or Fairy Tail, or read some Isekai genre stories. 

Q: It’s time to get on the shuttle to the first Mars colony, and you can only bring one personal item. What are you going to bring along with you?

A: My first thought was my phone, but I feel like that may be too standard of an answer. If we were talking about the fantasy realm, I might ask Stephen Strange to borrow his sling ring to open more portals to link the Earth and Mars. As to why he wouldn’t have just come with us in the first place, I don’t know; maybe he’s too busy fighting aliens, or something?

Q: What are you looking forward to about life after graduation? What do you think you’ll miss about MIT? 

A: I won’t be missing dining hall food very much, that’s for sure — except for the amazing oatmeal from one of the Maseeh dining hall chefs, Sum! I am, however, excited to live the nine-to-five life for a few years and have my weekends back. I’ll miss my friends dearly, since everyone will be so spread out across the States and abroad. I’ll miss the nights we spent watching movies, playing games, cooking, eating, and yapping away. I’m excited to see everyone grow and take another step closer to their dreams. It will be fun visiting them and being able to explore the world at the same time! For more immediate plans, I’ll be going back to Apple this summer to intern again, and will finish my MEng with the 6A program at Cadence. Afterwards, I shall see where life takes me!

Adam Berinsky awarded Carnegie fellowship

MIT political science professor among cohort of fellows who will focus on building a body of research on political polarization.

MIT political scientist Adam Berinsky has been named to the 2025 class of Andrew Carnegie Fellows, a high-profile honor for scholars pursuing research in the social sciences and humanities.

The fellowship is provided by The Carnegie Corp. of New York. Berinsky, the Mitsui Professor of Political Science, and 25 other fellows were selected from more than 300 applicants. They will each receive stipends of $200,000 for research that seeks to understand how and why our society has become so polarized, and how we can strengthen the forces of cohesion to fortify our democracy.

“Through these fellowships Carnegie is harnessing the unrivaled brainpower of our universities to help us to understand how our society has become so polarized,” says Carnegie President Louise Richardson. “Our future grant-making will be informed by what we learn from these scholars as we seek to mitigate the pernicious effects of political polarization.”

Berinsky said he is “incredibly honored to be named an Andrew Carnegie Fellow for the coming year. This fellowship will allow me to work on critical issues in the current political moment.”

During his year as a Carnegie Fellow, Berinsky will be working on a project, “Fostering an Accurate Information Ecosystem to Mitigate Polarization in the United States.”

“For a functioning democracy, it is essential that citizens share a baseline of common facts,” says Berinsky. “However, in today’s politically polarized climate, ‘alternative facts,’ and other forms of misinformation — from political rumors to conspiracy theories — distort how people see reality, and damage our social fabric.”

“I’ve spent the last 15 years investigating why individuals accept misinformation and how to counter misperceptions. But there is still a lot of work to be done. My project aims to tackle the serious problem of misinformation in the United States by bringing together existing approaches in new, more powerful combinations. I’m hoping that the whole can be more than the sum of its parts.”

Berinsky has been a member of the MIT faculty since 2003. He is the author of “Political Rumors: Why We Accept Misinformation and How to Fight It” (Princeton University Press, 2023).

Other MIT faculty who have received the Carnegie Fellowship in recent years include economists David Autor and Daron Acemoglu and political scientists Fotini Christia, Taylor Fravel, Richard Nielsen, and Charles Stewart.

New study reveals how cleft lip and cleft palate can arise

MIT biologists have found that defects in some transfer RNA molecules can lead to the formation of these common conditions.

Cleft lip and cleft palate are among the most common birth defects, occurring in about one in 1,050 births in the United States. These defects, which appear when the tissues that form the lip or the roof of the mouth do not join completely, are believed to be caused by a mix of genetic and environmental factors.

In a new study, MIT biologists have discovered how a genetic variant often found in people with these facial malformations leads to the development of cleft lip and cleft palate.

Their findings suggest that the variant diminishes cells’ supply of transfer RNA, a molecule that is critical for assembling proteins. When this happens, embryonic face cells are unable to fuse to form the lip and roof of the mouth.

“Until now, no one had made the connection that we made. This particular gene was known to be part of the complex involved in the splicing of transfer RNA, but it wasn’t clear that it played such a crucial role for this process and for facial development. Without the gene, known as DDX1, certain transfer RNA can no longer bring amino acids to the ribosome to make new proteins. If the cells can’t process these tRNAs properly, then the ribosomes can’t make protein anymore,” says Michaela Bartusel, an MIT research scientist and the lead author of the study.

Eliezer Calo, an associate professor of biology at MIT, is the senior author of the paper, which appears today in the American Journal of Human Genetics.

Genetic variants

Cleft lip and cleft palate, also known as orofacial clefts, can be caused by genetic mutations, but in many cases, there is no known genetic cause.

“The mechanism for the development of these orofacial clefts is unclear, mostly because they are known to be impacted by both genetic and environmental factors,” Calo says. “Trying to pinpoint what might be affected has been very challenging in this context.”

To discover genetic factors that influence a particular disease, scientists often perform genome-wide association studies (GWAS), which can reveal variants that are found more often in people who have a particular disease than in people who don’t.

For orofacial clefts, some of the genetic variants that have regularly turned up in GWAS appeared to be in a region of DNA that doesn’t code for proteins. In this study, the MIT team set out to figure out how variants in this region might influence the development of facial malformations.

Their studies revealed that these variants are located in an enhancer region called e2p24.2. Enhancers are segments of DNA that interact with protein-coding genes, helping to activate them by binding to transcription factors that turn on gene expression.

The researchers found that this region is in close proximity to three genes, suggesting that it may control the expression of those genes. One of those genes had already been ruled out as contributing to facial malformations, and another had already been shown to have a connection. In this study, the researchers focused on the third gene, which is known as DDX1.

DDX1, it turned out, is necessary for splicing transfer RNA (tRNA) molecules, which play a critical role in protein synthesis. Each transfer RNA molecule transports a specific amino acid to the ribosome — a cell structure that strings amino acids together to form proteins, based on the instructions carried by messenger RNA.

While there are about 400 different tRNAs found in the human genome, only a fraction of those tRNAs require splicing, and those are the tRNAs most affected by the loss of DDX1. These tRNAs transport four different amino acids, and the researchers hypothesize that these four amino acids may be particularly abundant in proteins that embryonic cells that form the face need to develop properly.

When the ribosomes need one of those four amino acids, but none of them are available, the ribosome can stall, and the protein doesn’t get made.

The researchers are now exploring which proteins might be most affected by the loss of those amino acids. They also plan to investigate what happens inside cells when the ribosomes stall, in hopes of identifying a stress signal that could potentially be blocked and help cells survive.

Malfunctioning tRNA

While this is the first study to link tRNA to craniofacial malformations, previous studies have shown that mutations that impair ribosome formation can also lead to similar defects. Studies have also shown that disruptions of tRNA synthesis — caused by mutations in the enzymes that attach amino acids to tRNA, or in proteins involved in an earlier step in tRNA splicing — can lead to neurodevelopmental disorders.

“Defects in other components of the tRNA pathway have been shown to be associated with neurodevelopmental disease,” Calo says. “One interesting parallel between these two is that the cells that form the face are coming from the same place as the cells that form the neurons, so it seems that these particular cells are very susceptible to tRNA defects.”

The researchers now hope to explore whether environmental factors linked to orofacial birth defects also influence tRNA function. Some of their preliminary work has found that oxidative stress — a buildup of harmful free radicals — can lead to fragmentation of tRNA molecules. Oxidative stress can occur in embryonic cells upon exposure to ethanol, as in fetal alcohol syndrome, or if the mother develops gestational diabetes.

“I think it is worth looking for mutations that might be causing this on the genetic side of things, but then also in the future, we would expand this into which environmental factors have the same effects on tRNA function, and then see which precautions might be able to prevent any effects on tRNAs,” Bartusel says.

The research was funded by the National Science Foundation Graduate Research Program, the National Cancer Institute, the National Institute of General Medical Sciences, and the Pew Charitable Trusts.

How should we prioritize patients waiting for kidney transplants?

A comprehensive study of the U.S. system could help policymakers analyze methods of matching donated kidneys and their recipients.

At any given time, about 100,000 people in the U.S. are waiting to become kidney transplant recipients. Roughly one-fifth of those get a new kidney each year, but others die while waiting. In short, the demand for kidneys makes it important to think about how we use the limited supply.

A study co-authored by an MIT economist brings new data to this issue, providing nuanced estimates of the lifespan-lengthening effect of kidney transplants. That can be hard to measure well, but the study is the first to account for some of the complexities involved, including the decisions patients make when accepting kidney transplants, and some of their pre-existing health factors.

The research concludes the system in use produces an additional 9.29 life-years from transplantation (LYFT) for kidney recipients. (LYFT is the difference in median survival for those with and without transplants.) If the organs were assigned randomly to patients, the study finds, that LYFT average would only be 7.54 overall. From that perspective, the current transplant system is a net positive for patients. However, the study also finds that the LYFT figure could potentially be raised as high as 14.08, depending on how the matching system is structured.

In any case, more precise estimates about the benefits of kidney transplants can help inform policymakers about the dynamics of the matching system in use.

“There’s always this question about how to take the scarce number of organs being donated and place them efficiently, and place them well,” says MIT economist Nikhil Agarwal, co-author of a newly published paper detailing the study’s results. As he emphasizes, the point of the paper is to inform the ongoing refinement of the matching system, rather than advocate one viewpoint or another.

The paper, “Choices and Outcomes in Assignment Mechanisms: The Allocation of Deceased Donor Kidneys,” is published in the latest issue of Econometrica. The authors are Agarwal, who is a professor in MIT’s Department of Economics; Charles Hodgson, an assistant professor of economics at Yale University; and Paulo Somaini, an associate professor of economics in Stanford University’s Graduate School of Business.

After people die, there is a period lasting up to 48 hours when they could be viable organ donors. Potential kidney recipients are prioritized by time spent on wait-lists as well as tissue-type similarity, and can accept or reject any given transplant offer.

Over the last decade-plus, Agarwal has conducted significant empirical research on matching systems for organ donations, especially kidney transplants. To conduct this study, the researchers used comprehensive data about patients on the kidney wait-list from 2000-2010, made available by the Organ Procurement and Transplantation Network, the national U.S. registry. This allowed the scholars to analyze both the matching system and the health effects of transplants; they track patient survival until February 2020.

The work is the first quasiexperimental study of kidney transplants; by carefully examining the decision-making tendencies of kidney recipients, along with many other health factors, the scholars are able to evaluate the effects of a transplant, other things being equal. Recipients are more likely to select kidney offers from donors who are younger, lacked hypertension, died of head trauma (suggesting their internal organs were healthy), and with whom they have perfect tissue-type matches.

“The [previous] methodology of estimating what are the life-years benefits was not incorporating this selection issue,” Agarwal says.

Additionally, overall, a key empirical feature of kidney transplants is that recipients who are healthier overall tend to have the largest realized life-years benefits from a transplant, meaning that the greatest increase in LYFT is not found in the set of patients with the worst health.

“You might think people who are the sickest and who are most likely to die without an organ are going to benefit the most from it [in added life-years],” Agarwal says. “But there might be some other comorbidity or factor that made them sick, and their body’s going to take a toll on the new organ, so the benefits might not be as large.”

With this in mind, the maximal LYFT number of 14.08 in the study comes from, broadly, a hypothetical scenario in which an increased number of otherwise healthy people receive transplants. Again, the current system tends to prioritize time spent on a wait-list. And some observers might advocate for a system that prioritizes those who are sickest. With all that in mind, the policymaking process for kidney transplants may still involve recognition that the biggest gains in patient life-years are not necessarily aligned with other prioritization factors.

“Our results indicate … a dilemma rooted in the tension between these two goals,” the authors write in the paper.

To be clear, Agarwal is not advocating for any one system over another, but conducting data-driven research so that policy officials can make more fully informed decisions in the ongoing, long-term process of trying to refine valuable transplant networks.

“I don’t necessarily think it’s my comparative advantage to make the ethical decisions, but we can at least think about and quantify what some of the tradeoffs are,” Agarwal adds.

Support for the research was provided in part by the National Science Foundation and by the Alfred P. Sloan Foundation. 

A chemist who tinkers with molecules’ structures

By changing how atoms in a molecule are arranged relative to each other, Associate Professor Alison Wendlandt aims to create compounds with new chemical properties.

Many biological molecules exist as “diastereomers” — molecules that have the same chemical structure but different spatial arrangements of their atoms. In some cases, these slight structural differences can lead to significant changes in the molecules’ functions or chemical properties.

As one example, the cancer drug doxorubicin can have heart-damaging side effects in a small percentage of patients. However, a diastereomer of the drug, known as epirubicin, which has a single alcohol group that points in a different direction, is much less toxic to heart cells.

“There are a lot of examples like that in medicinal chemistry where something that seems small, such as the position of a single atom in space, may actually be really profound,” says Alison Wendlandt, an associate professor of chemistry at MIT.

Wendlandt’s lab is focused on designing new tools that can convert these molecules into different forms.  Her group is also working on similar tools that can change a molecule into a different constitutional isomer — a molecule that has an atom or chemical group located in a different spot, even though it has the same chemical formula as the original.

“If you have a target molecule and you needed to make it without such a tool, you would have to go back to the beginning and make the whole molecule again to get to the final structure that you wanted,” Wendlandt says.

These tools can also lend themselves to creating entirely new molecules that might be difficult or even impossible to build using traditional chemical synthesis techniques.

“We’re focused on a broad suite of selective transformations, the goal being to make the biggest impact on how you might envision making a molecule,” she says. “If you are able to open up access to the interconversion of molecular structures, you can then think completely differently about how you would make a molecule.”

From math to chemistry

As the daughter of two geologists, Wendlandt found herself immersed in science from a young age. Both of her parents worked at the Colorado School of Mines, and family vacations often involved trips to interesting geological formations.

In high school, she found math more appealing than chemistry, and she headed to the University of Chicago with plans to major in mathematics. However, she soon had second thoughts, after encountering abstract math.

“I was good at calculus and the kind of math you need for engineering, but when I got to college and I encountered topology and N-dimensional geometry, I realized I don’t actually have the skills for abstract math. At that point I became a little bit more open-minded about what I wanted to study,” she says.

Though she didn’t think she liked chemistry, an organic chemistry course in her sophomore year changed her mind.

“I loved the problem-solving aspect of it. I have a very, very bad memory, and I couldn’t memorize my way through the class, so I had to just learn it, and that was just so fun,” she says.

As a chemistry major, she began working in a lab focused on “total synthesis,” a research area that involves developing strategies to synthesize a complex molecule, often a natural compound, from scratch.

Although she loved organic chemistry, a lab accident — an explosion that injured a student in her lab and led to temporary hearing loss for Wendlandt — made her hesitant to pursue it further. When she applied to graduate schools, she decided to go into a different branch of chemistry — chemical biology. She studied at Yale University for a couple of years, but she realized that she didn’t enjoy that type of chemistry and left after receiving a master’s degree.

She worked in a lab at the University of Kentucky for a few years, then applied to graduate school again, this time at the University of Wisconsin. There, she worked in an organic chemistry lab, studying oxidation reactions that could be used to generate pharmaceuticals or other useful compounds from petrochemicals.

After finishing her PhD in 2015, Wendlandt went to Harvard University for a postdoc, working with chemistry professor Eric Jacobsen. There, she became interested in selective chemical reactions that generate a particular isomer, and began studying catalysts that could perform glycosylation — the addition of sugar molecules to other molecules — at specific sites.

Editing molecules

Since joining the MIT faculty in 2018, Wendlandt has worked on developing catalysts that can convert a molecule into its mirror image or an isomer of the original.

In 2022, she and her students developed a tool called a stereo-editor, which can alter the arrangement of chemical groups around a central atom known as a stereocenter. This editor consists of two catalysts that work together to first add enough energy to remove an atom from a stereocenter, then replace it with an atom that has the opposite orientation. That energy input comes from a photocatalyst, which converts captured light into energy.

“If you have a molecule with an existing stereocenter, and you need the other enantiomer, typically you would have to start over and make the other enantiomer. But this new method tries to interconvert them directly, so it gives you a way of thinking about molecules as dynamic,” Wendlandt says. “You could generate any sort of three-dimensional structure of that molecule, and then in an independent step later, you could completely reorganize the 3D structure.”

She has also developed tools that can convert common sugars such as glucose into other isomers, including allose and other sugars that are difficult to isolate from natural sources, and tools that can create new isomers of steroids and alcohols. She is now working on ways to convert six-membered carbon rings to seven or eight-membered rings, and to add, subtract, or replace some of the chemical groups attached to the rings.

“I’m interested in creating general tools that will allow us to interconvert static structures. So, that may be taking a certain functional group and moving it to another part of the molecule entirely, or taking large rings and making them small rings,” she says. “Instead of thinking of molecules that we assemble as static, we’re thinking about them now as potentially dynamic structures, which could change how we think about making organic molecules.”

This approach also opens up the possibility of creating brand new molecules that haven’t been seen before, Wendlandt says. This could be useful, for example, to create drug molecules that interact with a target enzyme in just the right way.

“There’s a huge amount of chemical space that’s still unknown, bizarre chemical space that just has not been made. That’s in part because maybe no one has been interested in it, or because it’s just too hard to make that specific thing,” she says. “These kinds of tools give you access to isomers that are maybe not easily made.”

Anders Sejr Hansen named Edgerton Award winner

MIT associate professor recognized for exceptional distinction in teaching, research, and service at MIT.

Anders Sejr Hansen, Class of 1943 Career Development Professor in the Department of Biological Engineering, has been named as the recipient of the 2024-25 Harold E. Edgerton Faculty Achievement Award.

The annual award was established in fall 1982 as a permanent tribute to Institute Professor Emeritus Harold E. Edgerton for his great and enduring support for younger faculty members over the years. The purpose of the award is to recognize exceptional distinction in teaching, in research, and in service.

Hansen is the principal investigator of the Hansen Lab, which develops new methods to resolve 3D genome structure at high spatiotemporal resolution to understand how DNA looping and 3D folding regulates gene expression in health and disease. His areas of research include cancer biology, computational systems biology, instrumentation and measurement, and synthetic biology.

“My research focuses on how the expression of our genes is regulated,” says Hansen. “All the cells in our body have the same DNA and the same genes. Thus, the software or applications to each cell are the same. What’s different between a neuron and a blood cell is what genes they choose to express. My research focuses on understanding how this regulation takes place.”

Those who nominated Anders for the award emphasized his remarkable productivity, mentioning his two “highly cited, paradigm-shifting research articles in Science and Nature Genetics,” and his research presentations at 50 invited talks, including two keynotes, at universities and conferences worldwide. They also highlighted his passion for mentorship and career development for the 20 current members of his laboratory.

“Anders is an outstanding role model and ambassador of biological engineering, combining a powerful research program, run as a caring mentor, and innovative undergraduate education,” says Christopher Voigt, the Daniel I.C. Wang Professor in Biological Engineering and head of the Department of Biological Engineering.

Adds Laurie Boyer, a professor of biology and biological engineering, “His work reveals new insights into how we think about the dynamics of gene regulation that would not otherwise be possible. The Hansen Lab’s work provides a unified framework rapidly adopted by the field to learn how conserved regulators provide exquisite spatial and temporal control of gene expression in the context of 3D genome architecture.”

During the nomination process, students praised Hansen’s passion for his work, along with his ability to prepare them to apply their education outside the classroom.

“He always strives to guide each lab member towards both short-term scientific success and long-term career planning through regular one-on-one meetings, facilitating collaborations and access to scientific resources, and sharing his own experiences,” says Jin Yang, a graduate student in biological engineering and member of the Hansen Lab.

“Dr. Hansen's infectious excitement for the course material made it very enjoyable to come to class and envision potential applications of the fundamental topics he taught,” adds another one of his students. “Excellent lecturer!”

Hansen obtained his undergraduate and master’s degree in chemistry at Oxford University. He received his PhD in chemistry and chemical biology from Harvard University, where he applied systems biology approaches to understand how cells can encode and transmit information in the dynamics of transcription factor activation. For his postdoc at the University of California at Berkeley, Hansen developed new imaging approaches for dissecting the dynamics of architectural proteins with single-molecule resolution in living cells. Hansen joined MIT as an assistant professor of biological engineering in early 2020.

His recognitions include an NIH K99 Pathway to Independence Award (2019), NIH Director’s New Innovator Award (2020), a Pew-Stewart Scholar for Cancer Research Award (2021), an NSF CAREER Award (2024), and an NIH Director’s Transformative Research Award (2024).

Hansen has served on several committees at MIT, including the MIT Biological Engineering Graduate Program Admissions Committee, the MIT Computational and Systems Biology Graduate Admissions Committee, and the MIT Biological Engineering Graduate Recruiting Committee, of which he has been chair since 2023.

“I have known about the Edgerton Award since I started at MIT, and I think the broad focus on both research, teaching, and service really captures what makes MIT such a unique and wonderful place,” says Hansen. “I was therefore absolutely thrilled to receive the news that I would receive the Edgerton Award this year, and I am very grateful to all the wonderful colleagues here at MIT who have supported me over the years, and all the exceptional people in my lab whose work is being recognized.”

The Edward and Joyce Linde Music Building opens with Sonic Jubilance

An exuberant performance included five premieres by MIT composers, a fitting tribute to open the new home of MIT Music and launch the MIT arts festival Artfinity.

Johann Wolfgang von Goethe (1749-1832), the German polymath whose life and work embodied the connections between the arts and sciences, is said to have described architecture as “frozen music.” 

When the new Edward and Joyce Linde Music Building at MIT had its public opening earlier this year, the temperature outside may have been below freezing but the performances inside were a warm-up for the inaugural concert that took place in the evening. During the afternoon, visitors were invited to workshops in Balinese gamelan and Senegalese drumming, alongside performances by the MIT Chamber Music Society, MIT Festival Jazz Ensemble, and the MIT Laptop Ensemble (FaMLE), demonstrating the synergy between global music traditions and contemporary innovation in music technology. The building was filled with visitors from the MIT community and the Boston area, keen to be among the first to enter the new building and discover what MIT Music had planned for the opening occasion.

The evening’s landmark concert, Sonic Jubilance, celebrated the building’s completion and the pivotal role of MIT Music and Theater Arts (MTA) at the center of life on campus. The program was distinguished by five world premieres by MIT composers: “Summit and Mates,” by assistant professor in jazz Miguel Zenón; “Grace,” by senior lecturer in music Charles Shadle; “Two Noble Kinsmen,” by professor emeritus in music John Harbison; and “Madrigal,” by Keeril Makan, the Michael (1949) and Sonja Koerner Music Composition Professor. 

The premieres were interwoven through the program with performances by MIT ensembles demonstrating the breadth and depth of the conservatory-level music program — from the European classical tradition to Brazilian beats to Boston jazz (the full list of participating ensembles can be found below). 

Each performance demonstrated the different ways the space could be used to create new relationships between musicians and audiences. Designed in the round by the architecture firm SANAA, the Thomas Tull Concert Hall allows sound to resonate from the circular stage or from the aisles above the tiered seating; performers might be positioned below, above, or even in the midst of the audience.

“Music has been a part of MIT's curriculum and culture from the beginning,” said Chancellor Melissa Nobles in her opening address. “Arriving at this magnificent space has taken the collective efforts of past presidents, provosts, deans, faculty, alumni, and students, all working to get us here this evening.” 

Jay Scheib, the Class of 1949 Professor and MIT MTA section head, emphasized the vital role of Music at MIT as a source of cohesion and creativity for students, faculty, and the wider MIT community. 

“The new building is an extraordinary home for us. As a destination to convene communities around world musics and cultures, to engage in emerging music technologies, and to experience concerts and premieres featuring our extraordinary students and our internationally renowned faculty — the Edward and Joyce Linde Music Building is truly a transformational thing." 

The concert was also the launch event of Artfinity, MIT’s largest public festival of the arts since 2011, featuring more than 80 free performing and visual arts events. The concert hall will host performances throughout the spring, ranging from classical to jazz to rap, and more.

Institute Professor Marcus Thompson — the faculty co-lead for Artfinity alongside Azra Akšamija, director and associate professor of the Art, Culture, and Technology Program (ACT) at MIT — shared thoughts on the Edward and Joyce Linde Music Building as a point of orientation for the festival. 

“Our building offers the opportunity to point to the presence and importance of other art forms, media, practices, and experiences that can bring us together as practitioners and audiences, lifting our spirits and our sights,” Thompson reflected. “An ensemble of any kind is a community as well as a metaphor for what connects us, applying different talents to create more than we can do alone.”

The new compositions by the four faculty members were a case in point. The program opened with “Summit,” a brass fanfare projected from the top of the hall with ceremonial zeal. “The piece was specifically written as an opener for the concert,” Zenón explained. “My aim was to compose something that would make a statement straight away, while also using the idea of the ‘groove’ as a driving force. The title has two meanings. The first is a mountaintop, or the top of a structure — which is where the ensemble will be placed for the performance. The second is a gathering of great minds and great leaders, which is what MIT feels like for me.” Later in the program, Zenón premiered a jazz contrafact, “Mates,” playing on Benny Golson’s Stablemates, a tribute to Herb Pomeroy, founder of MIT’s jazz program. “The idea here is to use something connected to the jazz tradition — and to Boston’s history — and approach it from a more personal perspective,” said Zenón.

“Two Noble Kinsmen,” by Harbison, was composed as a benediction for the new home of MIT Music. “In choosing to set Shakespeare’s final words in this new piece for choir and strings, I wanted to convey the sense of an invocation, an introduction, an address to unseen forces,” said Harbison. “In this case, I wanted to leave the musical structure as plain as possible so that we understand why these words are chosen. I hoped to capture the stoic balance of these lines — they are in themselves a kind of verbal music.”

In setting the words of the poem “Grace,” by the Chickasaw poet Linda Hogan, Shadle — a composer of Choctaw heritage — envisioned a “sonic extension” of the MIT Land Acknowledgement. “‘Grace’ intended to speak to the Indigenous presence at the Institute and to open the new building with a reminder of the balm music that can bring to a troubled world,” said Shadle. “I hope that I have composed music that links Indigenous and Western traditions in ways that are compelling and thoughtful and that, while recognizing the ‘pieces of hurt,’ still makes a place for grace.”

Before the concert’s euphoric finale — a performance by Rambax Senegalese Drum Ensemble directed by Lamine Touré — “Madrigal” (the evening’s fourth world premiere) served to demonstrate the spatial dimensions of sound made possible by the design of the concert hall. 

Makan’s composition was performed by four student violinists positioned at the top of each aisle and a fifth, Professor Natalie Lin Douglas, at the center of the stage, simultaneously showcasing the geometry of the hall and referencing the ever-shifting perspectives of the sculpture that stands at the north entrance of the building — “Madrigal (2024),” by Sanford Biggers.

“My piece aims to capture the multifaceted quality of Sanford Biggers’ sculpture. From whichever vantage point we might look at it, we see the same patterns in new relationships with one another. In other words, there is no one point of view that is privileged over another.”

As faculty lead for the building project, Makan developed a friendship with Joyce Linde, who provided the principal gift that led to the building. “Joyce and I were on the selection committee to choose an artist to create a site-specific sculpture outside the building. She was very excited about the process, and very engaged with Sanford,” said Makan. “Joyce passed away before she was able to see the building’s completion, and I wanted to honor her legacy by writing an original piece of music in her memory.”

That sense of relationship, pattern-making, and new beginnings was articulated by Frederick Harris, director and senior lecturer in music and the co-producer of the concert, alongside Andy Wilds, program manager in music. “The hall is an instrument; we’re communing with this incredible space and getting to know it,” said Harris. “It’s a relationship. The circular form of the hall is very welcoming, not only to immersive experiences but also to shared experiences.”

The role of music in cultivating community will ensure that the building will become an integral part of MIT life. The work taking place in rehearsal rooms matches the innovation of the Institute’s labs — proving that the arts are a necessary counterpart to science and technology, continuous with the human instinct to express and invent. Sonic Jubilance sets the tone of what’s to come. 

MIT Music ensembles (in order of concert appearance):

• MIT Concert Choir

• MIT Chamber Chorus

• MIT Chamber Music Society

• MIT Vocal Jazz Ensemble

• MIT Jazz Advanced Music Performance Ensemble

• MIT Axiom Ensemble

• MIT Wind Ensemble

• MIT Gamelan Galak Tika

• Rambax MIT
   

Restoring healthy gene expression with programmable therapeutics

CAMP4 Therapeutics is targeting regulatory RNA, whose role in gene expression was first described by co-founder and MIT Professor Richard Young.

Many diseases are caused by dysfunctional gene expression that leads to too much or too little of a given protein. Efforts to cure those diseases include everything from editing genes to inserting new genetic snippets into cells to injecting the missing proteins directly into patients.

CAMP4 is taking a different approach. The company is targeting a lesser-known player in the regulation of gene expression known as regulatory RNA. CAMP4 co-founder and MIT Professor Richard Young has shown that by interacting with molecules called transcription factors, regulatory RNA plays an important role in controlling how genes are expressed. CAMP4’s therapeutics target regulatory RNA to increase the production of proteins and put patients’ levels back into healthy ranges.

The company’s approach holds promise for treating diseases caused by defects in gene expression, such as metabolic diseases, heart conditions, and neurological disorders. Targeting regulatory RNAs as opposed to genes could also offer more precise treatments than existing approaches.

“If I just want to fix a single gene’s defective protein output, I don’t want to introduce something that makes that protein at high, uncontrolled amounts,” says Young, who is also a core member of the Whitehead Institute. “That’s a huge advantage of our approach: It’s more like a correction than sledgehammer.”

CAMP4’s lead drug candidate targets urea cycle disorders (UCDs), a class of chronic conditions caused by a genetic defect that limits the body’s ability to metabolize and excrete ammonia. A phase 1 clinical trial has shown CAMP4’s treatment is safe and tolerable for humans, and in preclinical studies the company has shown its approach can be used to target specific regulatory RNA in the cells of humans with UCDs to restore gene expression to healthy levels.

“This has the potential to treat very severe symptoms associated with UCDs,” says Young, who co-founded CAMP4 with cancer genetics expert Leonard Zon, a professor at Harvard Medical School. “These diseases can be very damaging to tissues and causes a lot of pain and distress. Even a small effect in gene expression could have a huge benefit to patients, who are generally young.”

Mapping out new therapeutics

Young, who has been a professor at MIT since 1984, has spent decades studying how genes are regulated. It’s long been known that molecules called transcription factors, which orchestrate gene expression, bind to DNA and proteins. Research published in Young’s lab uncovered a previously unknown way in which transcription factors can also bind to RNA. The finding indicated RNA plays an underappreciated role in controlling gene expression.

CAMP4 was founded in 2016 with the initial idea of mapping out the signaling pathways that govern the expression of genes linked to various diseases. But as Young’s lab discovered and then began to characterize the role of regulatory RNA in gene expression around 2020, the company pivoted to focus on targeting regulatory RNA using therapeutic molecules known as antisense oligonucleotides (ASOs), which have been used for years to target specific messenger RNA sequences.

CAMP4 began mapping the active regulatory RNAs associated with the expression of every protein-coding gene and built a database, which it calls its RAP Platform, that helps it quickly identify regulatory RNAs to target  specific diseases and select ASOs that will most effectively bind to those RNAs.

Today, CAMP4 is using its platform to develop therapeutic candidates it believes can restore healthy protein levels to patients.

“The company has always been focused on modulating gene expression,” says CAMP4 Chief Financial Officer Kelly Gold MBA ’09. “At the simplest level, the foundation of many diseases is too much or too little of something being produced by the body. That is what our approach aims to correct.”

Accelerating impact

CAMP4 is starting by going after diseases of the liver and the central nervous system, where the safety and efficacy of ASOs has already been proven. Young believes correcting genetic expression without modulating the genes themselves will be a powerful approach to treating a range of complex diseases.

“Genetics is a powerful indicator of where a deficiency lies and how you might reverse that problem,” Young says. “There are many syndromes where we don’t have a complete understanding of the underlying mechanism of disease. But when a mutation clearly affects the output of a gene, you can now make a drug that can treat the disease without that complete understanding.”

As the company continues mapping the regulatory RNAs associated with every gene, Gold hopes CAMP4 can eventually minimize its reliance on wet-lab work and lean more heavily on machine learning to leverage its growing database and quickly identify regRNA targets for every disease it wants to treat.

In addition to its trials in urea cycle disorders, the company plans to launch key preclinical safety studies for a candidate targeting seizure disorders with a genetic basis, this year. And as the company continues exploring drug development efforts around the thousands of genetic diseases where increasing protein levels are can have a meaningful impact, it’s also considering collaborating with others to accelerate its impact.

“I can conceive of companies using a platform like this to go after many targets, where partners fund the clinical trials and use CAMP4 as an engine to target any disease where there’s a suspicion that gene upregulation or downregulation is the way to go,” Young says.

Beneath the biotech boom

MIT historian Robin Scheffler’s research shows how local regulations helped create certainty and safety principles that enabled an industry’s massive growth.

It’s considered a scientific landmark: A 1975 meeting at the Asilomar Conference Center in Pacific Grove, California, shaped a new safety regime for recombinant DNA, ensuring that researchers would apply caution to gene splicing. Those ideas have been so useful that in the decades since, when new topics in scientific safety arise, there are still calls for Asilomar-type conferences to craft good ground rules.

There’s something missing from this narrative, though: It took more than the Asilomar conference to set today’s standards. The Asilomar concepts were created with academic research in mind — but the biotechnology industry also makes products, and standards for that were formulated after Asilomar.

“The Asilomar meeting and Asilomar principles did not settle the question of the safety of genetic engineering,” says MIT scholar Robin Scheffler, author of a newly published research paper on the subject.

Instead, as Scheffler documents in the paper, Asilomar helped generate further debate, but those industry principles were set down later in the 1970s — first in Cambridge, Massachusetts, where politicians and concerned citizens wanted local biotech firms to be good neighbors. In response, the city passed safety laws for the emerging industry. And rather than heading off to places with zero regulations, local firms — including a fledgling Biogen — stayed put. Over the decades, the Boston area became the world leader in biotech.

Why stay? In essence, regulations gave biotech firms the certainty they needed to grow — and build. Lenders and real-estate developers needed signals that long-term investment in labs and facilities made sense. Generally, as Scheffler notes, even though “the idea that regulations can be anchoring for business does not have a lot of pull” in economic theory, in this case, regulations did matter.

“The trajectory of the industry in Cambridge, including biotechnology companies deciding to accommodate regulation, is remarkable,” says Scheffler. “It’s hard to imagine the American biotechnology industry without this dense cluster in Boston and Cambridge. These things that happened on a very local scale had huge echoes.”

Scheffler’s article, “Asilomar Goes Underground: The Long Legacy of Recombinant DNA Hazard Debates for the Greater Boston Area Biotechnology Industry,” appears in the latest issue of the Journal of the History of Biology. Scheffler is an associate professor in MIT’s Program in Science, Technology, and Society.

Business: Banking on certainty

To be clear, the Asilomar conference of 1975 did produce real results. Asilomar led to a system that helped evaluate projects’ potential risk and determine appropriate safety measures. The U.S. federal government subsequently adopted Asilomar-like principles for research it funded.

But in 1976, debate over the subject arose again in Cambridge, especially following a cover story in a local newspaper, the Boston Phoenix. Residents became concerned that recombinant DNA projects would lead to, hypothetically, new microorganisms that could damage public health.

“Scientists had not considered urban public health,” Scheffler says. “The Cambridge recombinant DNA debate in the 1970s made it a matter of what your neighbors think.”

After several months of hearings, research, and public debate (sometimes involving MIT faculty) stretching into early 1977, Cambridge adopted a somewhat stricter framework than the federal government had proposed for the handling of materials used in recombinant DNA work.

“Asilomar took on a new life in local regulations,” says Scheffler, whose research included government archives, news accounts, industry records, and more.

But a funny thing happened after Cambridge passed its recombinant DNA rules: The nascent biotech industry took root, and other area towns passed their own versions of the Cambridge rules.

“Not only did cities create more safety regulations,” Scheffler observes, “but the people asking for them switched from being left-wing activists or populist mayors to the Massachusetts Biotechnology Council and real estate development concerns.”

Indeed, he adds, “What’s interesting is how quickly safety concerns about recombinant DNA evaporated. Many people against recombinant DNA came to change their thinking.” And while some local residents continued to express concerns about the environmental impact of labs, “those are questions people ask when they no longer worry about the safety of the core work itself.”

Unlike federal regulations, these local laws applied to not only lab research but also products, and as such they let firms know they could work in a stable business environment with regulatory certainty. That mattered financially, and in a specific way: It helped companies build the buildings they needed to produce the products they had invented.

“The venture capital cycle for biotechnology companies was very focused on the research and exciting intellectual ideas, but then you have the bricks and mortar,” Scheffler says, referring to biotech production facilities. “The bricks and mortar is actually the harder problem for a lot of startup biotechnology companies.”

After all, he notes, “Venture capital will throw money after big discoveries, but a banker issuing a construction loan has very different priorities and is much more sensitive to things like factory permits and access to sewers 10 years from now. That’s why all these towns around Massachusetts passed regulations, as a way of assuring that.”

To grow globally, act locally

Of course, one additional reason biotech firms decided to land in the Boston area was the intellectual capital: With so many local universities, there was a lot of industry talent in the region. Local faculty co-founded some of the high-flying firms.

“The defining trait of the Cambridge-Boston biotechnology cluster is its density, right around the universities,” Scheffler says. “That’s a unique feature local regulations encouraged.”

It’s also the case, Scheffler notes, that some biotech firms did engage in venue-shopping to avoid regulations at first, although that was more the case in California, another state where the industry emerged. Still, the Boston-area regulations seemed to assuage both industry and public worries about the subject.

The foundations of biotechnology regulation in Massachusetts contain some additional historical quirks, including the time in the late 1970s when the city of Cambridge mistakenly omitted the recombinant DNA safety rules from its annually published bylaws, meaning the regulations were inactive. Officials at Biogen sent them a reminder to restore the laws to the books.

Half a century on from Asilomar, its broad downstream effects are not just a set of research principles — but also, refracted through the Cambridge episode, key ideas about public discussion and input; reducing uncertainty for business; the particular financing needs of industries; the impact of local and regional regulation; and the openness of startups to recognizing what might help them thrive.

“It’s a different way to think about the legacy of Asilomar,” Scheffler says. “And it’s a real contrast with what some people might expect from following scientists alone.” 

A faster way to solve complex planning problems

By eliminating redundant computations, a new data-driven method can streamline processes like scheduling trains, routing delivery drivers, or assigning airline crews.

When some commuter trains arrive at the end of the line, they must travel to a switching platform to be turned around so they can depart the station later, often from a different platform than the one at which they arrived.

Engineers use software programs called algorithmic solvers to plan these movements, but at a station with thousands of weekly arrivals and departures, the problem becomes too complex for a traditional solver to unravel all at once.

Using machine learning, MIT researchers have developed an improved planning system that reduces the solve time by up to 50 percent and produces a solution that better meets a user’s objective, such as on-time train departures. The new method could also be used for efficiently solving other complex logistical problems, such as scheduling hospital staff, assigning airline crews, or allotting tasks to factory machines.

Engineers often break these kinds of problems down into a sequence of overlapping subproblems that can each be solved in a feasible amount of time. But the overlaps cause many decisions to be needlessly recomputed, so it takes the solver much longer to reach an optimal solution.

The new, artificial intelligence-enhanced approach learns which parts of each subproblem should remain unchanged, freezing those variables to avoid redundant computations. Then a traditional algorithmic solver tackles the remaining variables.

“Often, a dedicated team could spend months or even years designing an algorithm to solve just one of these combinatorial problems. Modern deep learning gives us an opportunity to use new advances to help streamline the design of these algorithms. We can take what we know works well, and use AI to accelerate it,” says Cathy Wu, the Thomas D. and Virginia W. Cabot Career Development Associate Professor in Civil and Environmental Engineering (CEE) and the Institute for Data, Systems, and Society (IDSS) at MIT, and a member of the Laboratory for Information and Decision Systems (LIDS).

She is joined on the paper by lead author Sirui Li, an IDSS graduate student; Wenbin Ouyang, a CEE graduate student; and Yining Ma, a LIDS postdoc. The research will be presented at the International Conference on Learning Representations.

Eliminating redundance

One motivation for this research is a practical problem identified by a master’s student Devin Camille Wilkins in Wu’s entry-level transportation course. The student wanted to apply reinforcement learning to a real train-dispatch problem at Boston’s North Station. The transit organization needs to assign many trains to a limited number of platforms where they can be turned around well in advance of their arrival at the station.

This turns out to be a very complex combinatorial scheduling problem — the exact type of problem Wu’s lab has spent the past few years working on.

When faced with a long-term problem that involves assigning a limited set of resources, like factory tasks, to a group of machines, planners often frame the problem as Flexible Job Shop Scheduling.

In Flexible Job Shop Scheduling, each task needs a different amount of time to complete, but tasks can be assigned to any machine. At the same time, each task is composed of operations that must be performed in the correct order.

Such problems quickly become too large and unwieldy for traditional solvers, so users can employ rolling horizon optimization (RHO) to break the problem into manageable chunks that can be solved faster.

With RHO, a user assigns an initial few tasks to machines in a fixed planning horizon, perhaps a four-hour time window. Then, they execute the first task in that sequence and shift the four-hour planning horizon forward to add the next task, repeating the process until the entire problem is solved and the final schedule of task-machine assignments is created.

A planning horizon should be longer than any one task’s duration, since the solution will be better if the algorithm also considers tasks that will be coming up.

But when the planning horizon advances, this creates some overlap with operations in the previous planning horizon. The algorithm already came up with preliminary solutions to these overlapping operations.

“Maybe these preliminary solutions are good and don’t need to be computed again, but maybe they aren’t good. This is where machine learning comes in,” Wu explains.

For their technique, which they call learning-guided rolling horizon optimization (L-RHO), the researchers teach a machine-learning model to predict which operations, or variables, should be recomputed when the planning horizon rolls forward.

L-RHO requires data to train the model, so the researchers solve a set of subproblems using a classical algorithmic solver. They took the best solutions — the ones with the most operations that don’t need to be recomputed — and used these as training data.

Once trained, the machine-learning model receives a new subproblem it hasn’t seen before and predicts which operations should not be recomputed. The remaining operations are fed back into the algorithmic solver, which executes the task, recomputes these operations, and moves the planning horizon forward. Then the loop starts all over again.

“If, in hindsight, we didn’t need to reoptimize them, then we can remove those variables from the problem. Because these problems grow exponentially in size, it can be quite advantageous if we can drop some of those variables,” she adds.

An adaptable, scalable approach

To test their approach, the researchers compared L-RHO to several base algorithmic solvers, specialized solvers, and approaches that only use machine learning. It outperformed them all, reducing solve time by 54 percent and improving solution quality by up to 21 percent.

In addition, their method continued to outperform all baselines when they tested it on more complex variants of the problem, such as when factory machines break down or when there is extra train congestion. It even outperformed additional baselines the researchers created to challenge their solver.

“Our approach can be applied without modification to all these different variants, which is really what we set out to do with this line of research,” she says.

L-RHO can also adapt if the objectives change, automatically generating a new algorithm to solve the problem — all it needs is a new training dataset.

In the future, the researchers want to better understand the logic behind their model’s decision to freeze some variables, but not others. They also want to integrate their approach into other types of complex optimization problems like inventory management or vehicle routing.

This work was supported, in part, by the National Science Foundation, MIT’s Research Support Committee, an Amazon Robotics PhD Fellowship, and MathWorks.

MIT Lincoln Laboratory is a workhorse for national security

The US Air Force and MIT renew contract for operating the federally funded R&D center, a long-standing asset for defense innovation and prototyping.

In 1949, the U.S. Air Force called upon MIT with an urgent need. Soviet aircraft carrying atomic bombs were capable of reaching the U.S. homeland, and the nation was defenseless. A dedicated center — MIT Lincoln Laboratory — was established. The brightest minds from MIT came together in service to the nation, making scientific and engineering leaps to prototype the first real-time air defense system. The commercial sector and the U.S. Department of Defense (DoD) then produced and deployed the system, called SAGE, continent-wide.

The SAGE story still describes MIT Lincoln Laboratory’s approach to national security innovation today. The laboratory works with DoD agencies to identify challenging national security gaps, determines if technology can contribute to a solution, and then executes an R&D program to advance critical technologies. The principal products of these programs are advanced technology prototypes, which are often rapidly fabricated and demonstrated through test and evaluation.

Throughout this process, the laboratory closely coordinates with the DoD and other federal agency sponsors, and then transfers the technology in many forms to industry for manufacturing at scale to meet national needs. For nearly 75 years, these technologies have saved lives, responded to emergencies, fueled the nation’s economy, and impacted the daily life of Americans and our allies. 

"Lincoln Laboratory accelerates the pace of national security technology development, in partnership with the government, private industry, and the broader national security ecosystem," says Melissa Choi, director of MIT Lincoln Laboratory. "We integrate high-performance teams with advanced facilities and the best technology available to bring novel prototypes to life, providing lasting benefits to the United States."

The Air Force and MIT recently renewed their contract for the continued operation of Lincoln Laboratory. The contract was awarded by the Air Force Lifecycle Management Center Strategic Services Division on Hanscom Air Force Base for a term of five years, with an option for an additional five years. Since Lincoln Laboratory’s founding, MIT has operated the laboratory in the national interest for no fee and strictly on a cost-reimbursement basis. The contract award is indicative of the DoD’s continuing recognition of the long-term value of, and necessity for, cutting-edge R&D in service of national security.

Critical contributions to national security

MIT Lincoln Laboratory is the DoD’s largest federally funded research and development center R&D laboratory. Sponsored by the under secretary of defense for research and engineering, it contributes to a broad range of national security missions and domains.

Among the most critical domains are air and missile defense. Laboratory researchers pioneer advanced radar systems and algorithms crucial for detecting, tracking, and targeting ballistic missiles and aircraft, and serve as scientific advisors to the Reagan Test Site. They also conduct comprehensive studies on missile defense needs, such as the recent National Defense Authorization Act–directed study on the defense of Guam, and provide actionable insights to Congress.  

MIT Lincoln Laboratory is also at the forefront of space systems and technologies, enabling the military to monitor space activities and communicate at very high bandwidths. Laboratory engineers developed the innovatively curved detector within the Space Surveillance Telescope that allows the U.S. Space Force to track tiny space objects. It also operates the world's highest-resolution long-range radar for imaging satellites. Recently, the laboratory worked closely with NASA to demonstrate laser communications systems in space, setting a record for the fastest satellite downlink and farthest lasercom link ever achieved. These breakthroughs are heralding a new era in satellite communications for defense and civil missions.

Perhaps most importantly, MIT Lincoln Laboratory is asked to rapidly prototype solutions to urgent and emerging threats. These solutions are both transferred to industry for production and fielded directly to war-fighters, saving lives. To combat improvised explosive devices in Iraq and Afghanistan, the laboratory quickly and iteratively developed several novel systems to detect and defeat explosive devices and insurgent networks. When insurgents were attacking forward-operating bases at night, the laboratory developed an advanced infrared camera system to prevent the attacks. Like other multi-use technologies developed at the laboratory, that system led to a successful commercial startup, which was recently acquired by Anduril.

Responding to domestic crises is also a key part of the laboratory’s mission. After the attacks of 9/11/2001, the laboratory quickly integrated a system to defend the airspace around critical locations in the capital region. More recently, the laboratory’s application of AI to video forensics and physical screening has resulted in commercialized systems deployed in airports and mass transit settings. Over the last decade, the laboratory has adapted its technology for many other homeland security needs, including responses to natural disasters. As one example, researchers repurposed a world-class lidar system first used by the military for terrain mapping to quickly quantify damage after hurricanes.

For all of these efforts, the laboratory exercises responsible stewardship of taxpayer funds, identifying multiple uses for the technologies it develops and introducing disruptive approaches to reduce costs for the government. Sometimes, the system architecture or design results in cost savings, as is the case with the U.S. Air Force's SensorSat; the laboratory’s unique sensor design enabled a satellite 10 times smaller and cheaper than those typically used for space surveillance. Another approach is by creating novel systems from low-cost components. For instance, laboratory researchers discovered a way to make phased-array radars using cell phone electronics instead of traditional expensive components, greatly reducing the cost of deploying the radars for weather and aircraft surveillance.

The laboratory also pursues emerging technology to bring about transformative solutions. In the 1960s, such vision brought semiconductor lasers into the world, and in the 1990s shrunk transistors more than industry imagined possible. Today, laboratory staff are pursuing other new realms: making imagers reconfigurable at the pixel level, designing quantum sensors to transform navigation technology, and developing superconducting electronics to improve computing efficiency.

A long, beneficial relationship between MIT and the DoD

"Lincoln Laboratory has created a deep understanding and knowledge base in core national security missions and associated technologies. We look forward to continuing to work closely with government sponsors, industry, and academia through our trusted, collaborative relationships to address current and future national security challenges and ensure technological superiority," says Scott Anderson, assistant director for operations at MIT Lincoln Laboratory.

"MIT has always been proud to support the nation through its operation of Lincoln Laboratory. The long-standing relationship between MIT and the Department of Defense through this storied laboratory has been a difference-maker for the safety, economy, and industrial power of the United States, and we look forward to seeing the innovations ahead of us," notes Ian Waitz, MIT vice president for research.

Under the terms of the renewed contract, MIT will ensure that Lincoln Laboratory remains ready to meet R&D challenges that are critical to national security.

A visual pathway in the brain may do more than recognize objects

New research using computational vision models suggests the brain’s “ventral stream” might be more versatile than previously thought.

When visual information enters the brain, it travels through two pathways that process different aspects of the input. For decades, scientists have hypothesized that one of these pathways, the ventral visual stream, is responsible for recognizing objects, and that it might have been optimized by evolution to do just that.

Consistent with this, in the past decade, MIT scientists have found that when computational models of the anatomy of the ventral stream are optimized to solve the task of object recognition, they are remarkably good predictors of the neural activities in the ventral stream.

However, in a new study, MIT researchers have shown that when they train these types of models on spatial tasks instead, the resulting models are also quite good predictors of the ventral stream’s neural activities. This suggests that the ventral stream may not be exclusively optimized for object recognition.

“This leaves wide open the question about what the ventral stream is being optimized for. I think the dominant perspective a lot of people in our field believe is that the ventral stream is optimized for object recognition, but this study provides a new perspective that the ventral stream could be optimized for spatial tasks as well,” says MIT graduate student Yudi Xie.

Xie is the lead author of the study, which will be presented at the International Conference on Learning Representations. Other authors of the paper include Weichen Huang, a visiting student through MIT’s Research Science Institute program; Esther Alter, a software engineer at the MIT Quest for Intelligence; Jeremy Schwartz, a sponsored research technical staff member; Joshua Tenenbaum, a professor of brain and cognitive sciences; and James DiCarlo, the Peter de Florez Professor of Brain and Cognitive Sciences, director of the Quest for Intelligence, and a member of the McGovern Institute for Brain Research at MIT.

Beyond object recognition

When we look at an object, our visual system can not only identify the object, but also determine other features such as its location, its distance from us, and its orientation in space. Since the early 1980s, neuroscientists have hypothesized that the primate visual system is divided into two pathways: the ventral stream, which performs object-recognition tasks, and the dorsal stream, which processes features related to spatial location.

Over the past decade, researchers have worked to model the ventral stream using a type of deep-learning model known as a convolutional neural network (CNN). Researchers can train these models to perform object-recognition tasks by feeding them datasets containing thousands of images along with category labels describing the images.

The state-of-the-art versions of these CNNs have high success rates at categorizing images. Additionally, researchers have found that the internal activations of the models are very similar to the activities of neurons that process visual information in the ventral stream. Furthermore, the more similar these models are to the ventral stream, the better they perform at object-recognition tasks. This has led many researchers to hypothesize that the dominant function of the ventral stream is recognizing objects.

However, experimental studies, especially a study from the DiCarlo lab in 2016, have found that the ventral stream appears to encode spatial features as well. These features include the object’s size, its orientation (how much it is rotated), and its location within the field of view. Based on these studies, the MIT team aimed to investigate whether the ventral stream might serve additional functions beyond object recognition.

“Our central question in this project was, is it possible that we can think about the ventral stream as being optimized for doing these spatial tasks instead of just categorization tasks?” Xie says.

To test this hypothesis, the researchers set out to train a CNN to identify one or more spatial features of an object, including rotation, location, and distance. To train the models, they created a new dataset of synthetic images. These images show objects such as tea kettles or calculators superimposed on different backgrounds, in locations and orientations that are labeled to help the model learn them.

The researchers found that CNNs that were trained on just one of these spatial tasks showed a high level of “neuro-alignment” with the ventral stream — very similar to the levels seen in CNN models trained on object recognition.

The researchers measure neuro-alignment using a technique that DiCarlo’s lab has developed, which involves asking the models, once trained, to predict the neural activity that a particular image would generate in the brain. The researchers found that the better the models performed on the spatial task they had been trained on, the more neuro-alignment they showed.

“I think we cannot assume that the ventral stream is just doing object categorization, because many of these other functions, such as spatial tasks, also can lead to this strong correlation between models’ neuro-alignment and their performance,” Xie says. “Our conclusion is that you can optimize either through categorization or doing these spatial tasks, and they both give you a ventral-stream-like model, based on our current metrics to evaluate neuro-alignment.”

Comparing models

The researchers then investigated why these two approaches — training for object recognition and training for spatial features — led to similar degrees of neuro-alignment. To do that, they performed an analysis known as centered kernel alignment (CKA), which allows them to measure the degree of similarity between representations in different CNNs. This analysis showed that in the early to middle layers of the models, the representations that the models learn are nearly indistinguishable.

“In these early layers, essentially you cannot tell these models apart by just looking at their representations,” Xie says. “It seems like they learn some very similar or unified representation in the early to middle layers, and in the later stages they diverge to support different tasks.”

The researchers hypothesize that even when models are trained to analyze just one feature, they also take into account “non-target” features — those that they are not trained on. When objects have greater variability in non-target features, the models tend to learn representations more similar to those learned by models trained on other tasks. This suggests that the models are using all of the information available to them, which may result in different models coming up with similar representations, the researchers say.

“More non-target variability actually helps the model learn a better representation, instead of learning a representation that’s ignorant of them,” Xie says. “It’s possible that the models, although they’re trained on one target, are simultaneously learning other things due to the variability of these non-target features.”

In future work, the researchers hope to develop new ways to compare different models, in hopes of learning more about how each one develops internal representations of objects based on differences in training tasks and training data.

“There could be still slight differences between these models, even though our current way of measuring how similar these models are to the brain tells us they’re on a very similar level. That suggests maybe there’s still some work to be done to improve upon how we can compare the model to the brain, so that we can better understand what exactly the ventral stream is optimized for,” Xie says.

The research was funded by the Semiconductor Research Corporation and the U.S. Defense Advanced Research Projects Agency.

Bringing manufacturing back to America, one fab lab at a time

A collaborative network of makerspaces has spread from MIT across the country, helping communities make their own products.

Reindustrializing America will require action from not only businesses but also a new wave of people that have the skills, experience, and drive to make things. While many efforts in this area have focused on top-down education and manufacturing initiatives, an organic, grassroots movement has been inspiring a new generation of makers across America for the last 20 years.

The first fab lab was started in 2002 by MIT’s Center for Bits and Atoms (CBA). To teach students to use the digital fabrication research facility, CBA’s leaders began teaching a rapid-prototyping class called MAS.863 (How To Make (almost) Anything). In response to overwhelming demand, CBA collaborated with civil rights activist and MIT adjunct professor Mel King to create a community-scale version of the lab, integrating tools for 3D printing and scanning, laser cutting, precision and large-format machining, molding and casting, and surface-mount electronics, as well as design software.

That was supposed to be the end of the story; they didn’t expect a maker movement. Then another community reached out to get help building their own fab lab. Then another. Today there are hundreds of U.S. fab labs, in nearly every state, in locations ranging from community college campuses to Main Street. The fab labs offer open access to tools and software, as well as education, training, and community to people from all backgrounds.

“In the fab labs you can make almost anything,” says Professor and CBA Director Neil Gershenfeld. “That doesn’t mean everybody will make everything, but they can make things for themselves and their communities. The success of the fab labs suggests the real way to bring manufacturing back to America is not as it was. This is a different notion of agile, just-in-time manufacturing that’s personalized, distributed, and doesn’t have a sharp boundary between producer and consumer.”

Communities of makers

A fab lab opened at Florida A&M University about a year ago, but it didn’t take long for faculty and staff to notice its impact on their students. Denaria Pringley, an elementary education teacher with no experience in STEM, first came to the lab as part of a class requirement. That’s when she realized she could build her own guitar. In a pattern that has repeated itself across the country, Pringley began coming to the lab on nights and weekends, 3D-printing the body of the guitar, drilling together the neck, sanding and polishing the finish, laser engraving pick guards, and stringing everything together. Today, she works in the fab lab and knows how to run every machine in the space.

“Her entire disposition transformed through the fab lab,” says FAMU Dean of Education Sarah Price. “Every day, students make something new. There’s so much creativity going on in the lab it astounds me.”

Gershenfeld says describing how the fab labs work is a bit like describing how the internet works. At a high level, fab labs are spaces to play, create, learn, mentor, and invent. As they started replicating, Gershenfeld and his colleague Sherry Lassiter started the Fab Foundation, a nonprofit that provides operational, technical, and logistical assistance to labs. Last year, The Boston Globe called the global network of thousands of fab labs one of MIT’s most influential contributions of the last 25 years.

Some fab labs are housed in colleges. Others are funded by local governments, businesses, or through donations. Even fab labs operated in part by colleges can be open to anyone, and many of those fab labs partner with surrounding K-12 schools and continuing education programs.

Increasingly, corporate social responsibility programs are investing in fab labs, giving their communities spaces for STEM education, workforce development, and economic development. For instance, Chevron supported the startup of the fab lab at FAMU. Lassiter, the president of the Fab Foundation, notes, “Fab labs have evolved to become community anchor organizations, building strong social connections and resilience in addition to developing technical skills and providing public access to manufacturing capabilities.”

“We’re a community resource,” says Eric Saliim, who serves as a program manager at the fab lab housed in North Carolina Central University. “We have no restrictions for how you can use our fab lab. People make everything from art to car parts, products for their home, fashion accessories, you name it.”

Many fab lab instructors say the labs are a powerful way to make abstract concepts real and spark student interest in STEM subjects.

“More schools should be using fab labs to get kids interested in computer science and coding,” says Scott Simenson, former director of the fab lab at Century College in Minnesota. “This world is going to get a lot more digitally sophisticated, and we need a workforce that’s not only highly trained but also educated around subjects like computer science and artificial intelligence.”

Century College opened its fab lab in 2004 amid years of declining enrollment in its engineering and design programs.

“It’s a great bridge between the theoretical and the applied,” Simenson explains. “Frankly, it helped a lot of engineering students who were disgruntled because they felt like they didn’t get to make enough things with their hands.”

The fab lab has since helped support the creation of Century College programs in digital and additive manufacturing, welding, and bioprinting.

"Working in fab labs establishes a growth mindset for our community as well as our students,” says Kelly Zelesnik, the dean of Lorain County Community College in Ohio. “Students are so under-the-gun to get it right and the grade that they lose sight of the learning. But when they’re in the fab lab, they’re iterating, because nothing ever works the first time."

In addition to offering access to equipment and education, fab labs foster education, mentorship, and innovation. Businesses often use local fab labs to make prototypes or test new products. Students have started businesses around their art and fashion creations.

Rick Pollack was a software entrepreneur and frequent visitor to the fab lab at Lorain County Community College. Pollack became fascinated with 3D printers and eventually started the additive manufacturing company MakerGear after months of tinkering with the machines in the lab in 2009. MakerGear quickly became one of the most popular producers of 3D printers in the country.

“Everyone wants to talk about innovation with STEM education and business incubation,” Gershenfeld says. “This is delivering on that by filling in the missing scaffolding: the means of production.”

Manufacturing reimagined

Many fab labs begin with tiny spaces in forgotten corners of buildings and campuses. Over time, they attract a motley crew of people that have often struggled in structured, hierarchical classroom settings. Eventually, they become hubs for people of all backgrounds driven by making.

“Fab labs provide access to tools, but what’s really driving their success is the culture of peer-to-peer, project-based learning and production,” Gershenfeld says. “Fab labs don’t separate basic and applied work, short- and long-term goals, play and problem solving. The labs are a very bottom-up distribution of the culture at MIT.”

While the local maker movement won’t replace mass manufacturing, Gershenfeld says that mass manufacturing produces goods for consumers who all want the same thing, while local production can make more interesting things that differ for individuals.

Moreover, Gershenfeld doesn’t believe you can measure the impact of fab labs by looking only at the things produced.

“A significant part of the benefit of these labs is the act of making itself,” he says. “For instance, a fab lab in Detroit led by Blair Evans worked with at-risk youth, delivering better life outcomes than conventional social services. These labs attract interest and then build skills and communities, and so along with the things that get made, the community-building, the knowledge, the connecting, is all as important as the immediate economic impact.”

Hundred-year storm tides will occur every few decades in Bangladesh, scientists report

With projected global warming, the frequency of extreme storms will ramp up by the end of the century, according to a new study.

Tropical cyclones are hurricanes that brew over the tropical ocean and can travel over land, inundating coastal regions. The most extreme cyclones can generate devastating storm tides — seawater that is heightened by the tides and swells onto land, causing catastrophic flood events in coastal regions. A new study by MIT scientists finds that, as the planet warms, the recurrence of destructive storm tides will increase tenfold for one of the hardest-hit regions of the world.

In a study appearing today in One Earth, the scientists report that, for the highly populated coastal country of Bangladesh, what was once a 100-year event could now strike every 10 years — or more often — by the end of the century. 

In a future where fossil fuels continue to burn as they do today, what was once considered a catastrophic, once-in-a-century storm tide will hit Bangladesh, on average, once per decade. And the kind of storm tides that have occurred every decade or so will likely batter the country’s coast more frequently, every few years.

Bangladesh is one of the most densely populated countries in the world, with more than 171 million people living in a region roughly the size of New York state. The country has been historically vulnerable to tropical cyclones, as it is a low-lying delta that is easily flooded by storms and experiences a seasonal monsoon. Some of the most destructive floods in the world have occurred in Bangladesh, where it’s been increasingly difficult for agricultural economies to recover.

The study also finds that Bangladesh will likely experience tropical cyclones that overlap with the months-long monsoon season. Until now, cyclones and the monsoon have occurred at separate times during the year. But as the planet warms, the scientists’ modeling shows that cyclones will push into the monsoon season, causing back-to-back flooding events across the country.

“Bangladesh is very active in preparing for climate hazards and risks, but the problem is, everything they’re doing is more or less based on what they’re seeing in the present climate,” says study co-author Sai Ravela, principal research scientist in MIT’s Department of Earth, Atmospheric and Planetary Sciences (EAPS). “We are now seeing an almost tenfold rise in the recurrence of destructive storm tides almost anywhere you look in Bangladesh. This cannot be ignored. So, we think this is timely, to say they have to pause and revisit how they protect against these storms.”

Ravela’s co-authors are Jiangchao Qiu, a postdoc in EAPS, and Kerry Emanuel, professor emeritus of atmospheric science at MIT.

Height of tides

In recent years, Bangladesh has invested significantly in storm preparedness, for instance in improving its early-warning system, fortifying village embankments, and increasing access to community shelters. But such preparations have generally been based on the current frequency of storms.

In this new study, the MIT team aimed to provide detailed projections of extreme storm tide hazards, which are flooding events where tidal effects amplify cyclone-induced storm surge, in Bangladesh under various climate-warming scenarios and sea-level rise projections.

“A lot of these events happen at night, so tides play a really strong role in how much additional water you might get, depending on what the tide is,” Ravela explains.

To evaluate the risk of storm tide, the team first applied a method of physics-based downscaling, which Emanuel’s group first developed over 20 years ago and has been using since to study hurricane activity in different parts of the world. The technique involves a low-resolution model of the global ocean and atmosphere that is embedded with a finer-resolution model that simulates weather patterns as detailed as a single hurricane. The researchers then scatter hurricane “seeds” in a region of interest and run the model forward to observe which seeds grow and make landfall over time.

To the downscaled model, the researchers incorporated a hydrodynamical model, which simulates the height of a storm surge, given the pattern and strength of winds at the time of a given storm. For any given simulated storm, the team also tracked the tides, as well as effects of sea level rise, and incorporated this information into a numerical model that calculated the storm tide, or the height of the water, with tidal effects as a storm makes landfall.

Extreme overlap

With this framework, the scientists simulated tens of thousands of potential tropical cyclones near Bangladesh, under several future climate scenarios, ranging from one that resembles the current day to one in which the world experiences further warming as a result of continued fossil fuel burning. For each simulation, they recorded the maximum storm tides along the coast of Bangladesh and noted the frequency of storm tides of various heights in a given climate scenario.

“We can look at the entire bucket of simulations and see, for this storm tide of say, 3 meters, we saw this many storms, and from that you can figure out the relative frequency of that kind of storm,” Qiu says. “You can then invert that number to a return period.”

A return period is the time it takes for a storm of a particular type to make landfall again. A storm that is considered a “100-year event” is typically more powerful and destructive, and in this case, creates more extreme storm tides, and therefore more catastrophic flooding, compared to a 10-year event.

From their modeling, Ravela and his colleagues found that under a scenario of increased global warming, the storms that previously were considered 100-year events, producing the highest storm tide values, can recur every decade or less by late-century. They also observed that, toward the end of this century, tropical cyclones in Bangladesh will occur across a broader seasonal window, potentially overlapping in certain years with the seasonal monsoon season.

“If the monsoon rain has come in and saturated the soil, a cyclone then comes in and it makes the problem much worse,” Ravela says. “People won’t have any reprieve between the extreme storm and the monsoon. There are so many compound and cascading effects between the two. And this only emerges because warming happens.”

Ravela and his colleagues are using their modeling to help experts in Bangladesh better evaluate and prepare for a future of increasing storm risk. And he says that the climate future for Bangladesh is in some ways not unique to this part of the world.

“This climate change story that is playing out in Bangladesh in a certain way will be playing out in a different way elsewhere,” Ravela notes. “Maybe where you are, the story is about heat stress, or amplifying droughts, or wildfires. The peril is different. But the underlying catastrophe story is not that different.”

This research is supported in part by the MIT Climate Resilience Early Warning Systems Climate Grand Challenges project, the Jameel Observatory JO-CREWSNet project; MIT Weather and Climate Extremes Climate Grand Challenges project; and Schmidt Sciences, LLC. 

Engineered bacteria emit signals that can be spotted from a distance

These bacteria, which could be designed to detect pollution or nutrients, could act as sensors to help farmers monitor their crops.

Bacteria can be engineered to sense a variety of molecules, such as pollutants or soil nutrients. In most cases, however, these signals can only be detected by looking at the cells under a microscope, making them impractical for large-scale use.

Using a new method that triggers cells to produce molecules that generate unique combinations of color, MIT engineers have shown that they can read out these bacterial signals from as far as 90 meters away. Their work could lead to the development of bacterial sensors for agricultural and other applications, which could be monitored by drones or satellites.

“It’s a new way of getting information out of the cell. If you’re standing next to it, you can’t see anything by eye, but from hundreds of meters away, using specific cameras, you can get the information when it turns on,” says Christopher Voigt, head of MIT’s Department of Biological Engineering and the senior author of the new study.

In a paper appearing today in Nature Biotechnology, the researchers showed that they could engineer two different types of bacteria to produce molecules that give off distinctive wavelengths of light across the visible and infrared spectra of light, which can be imaged with hyperspectral cameras. These reporting molecules were linked to genetic circuits that detect nearby bacteria, but this approach could also be combined with any existing sensor, such as those for arsenic or other contaminants, the researchers say.

“The nice thing about this technology is that you can plug and play whichever sensor you want,” says Yonatan Chemla, an MIT postdoc who is one of the lead authors of the paper. “There is no reason that any sensor would not be compatible with this technology.”

Itai Levin PhD ’24 is also a lead author of the paper. Other authors include former undergraduate students Yueyang Fan ’23 and Anna Johnson ’22, and Connor Coley, an associate professor of chemical engineering at MIT.

Hyperspectral imaging

There are many ways to engineer bacterial cells so that they can sense a particular chemical. Most of these work by connecting detection of a molecule to an output such as green fluorescent protein (GFP). These work well for lab studies, but such sensors can’t be measured from long distances.

For long-distance sensing, the MIT team came up with the idea to engineer cells to produce hyperspectral reporter molecules, which can be detected using hyperspectral cameras. These cameras, which were first invented in the 1970s, can determine how much of each color wavelength is present in any given pixel. Instead of showing up as simply red or green, each pixel contains information on hundreds different wavelengths of light.

Currently, hyperspectral cameras are used for applications such as detecting the presence of radiation. In the areas around Chernobyl, these cameras have been used to measure slight color changes that radioactive metals produce in the chlorophyll of plant cells. Hyperspectral cameras are also used to look for signs of malnutrition or pathogen invasion in plants.

That work inspired the MIT team to explore whether they could engineer bacterial cells to produce hyperspectral reporters when they detect a target molecule.

For a hyperspectral reporter to be most useful, it should have a spectral signature with peaks in multiple wavelengths of light, making it easier to detect. The researchers performed quantum calculations to predict the hyperspectral signatures of about 20,000 naturally occurring cell molecules, allowing them to identify those with the most unique patterns of light emission. Another key feature is the number of enzymes that would need to be engineered into a cell to get it to produce the reporter — a trait that will vary for different types of cells.

“The ideal molecule is one that’s really different from everything else, making it detectable, and requires the fewest number of enzymes to produce it in the cell,” Voigt says.

In this study, the researchers identified two different molecules that were best suited for two types of bacteria. For a soil bacterium called Pseudomonas putida, they used a reporter called biliverdin — a pigment that results from the breakdown of heme. For an aquatic bacterium called Rubrivivax gelatinosus, they used a type of bacteriochlorophyll. For each bacterium, the researchers engineered the enzymes necessary to produce the reporter into the host cell, then linked them to genetically engineered sensor circuits.

“You could add one of these reporters to a bacterium or any cell that has a genetically encoded sensor in its genome. So, it might respond to metals or radiation or toxins in the soil, or nutrients in the soil, or whatever it is you want it to respond to. Then the output of that would be the production of this molecule that can then be sensed from far away,” Voigt says.

Long-distance sensing

In this study, the researchers linked the hyperspectral reporters to circuits designed for quorum sensing, which allow cells to detect other nearby bacteria. They have also shown, in work done after this paper, that these reporting molecules can be linked to sensors for chemicals including arsenic.

When testing their sensors, the researchers deployed them in boxes so they would remain contained. The boxes were placed in fields, deserts, or on the roofs of buildings, and the cells produced signals that could be detected using hyperspectral cameras mounted on drones. The cameras take about 20 to 30 seconds to scan the field of view, and computer algorithms then analyze the signals to reveal whether the hyperspectral reporters are present.

In this paper, the researchers reported imaging from a maximum distance of 90 meters, but they are now working on extending those distances.

They envision that these sensors could be deployed for agricultural purposes such as sensing nitrogen or nutrient levels in soil. For those applications, the sensors could also be designed to work in plant cells. Detecting landmines is another potential application for this type of sensing.

Before being deployed, the sensors would need to undergo regulatory approval by the U.S. Environmental Protection Agency, as well as the U.S. Department of Agriculture if used for agriculture. Voigt and Chemla have been working with both agencies, the scientific community, and other stakeholders to determine what kinds of questions need to be answered before these technologies could be approved.

“We’ve been very busy in the past three years working to understand what are the regulatory landscapes and what are the safety concerns, what are the risks, what are the benefits of this kind of technology?” Chemla says.

The research was funded by the U.S. Department of Defense; the Army Research Office, a directorate of the U.S. Army Combat Capabilities Development Command Army Research Laboratory (the funding supported engineering of environmental strains and optimization of genetically-encoded sensors and hyperspectral reporter biosynthetic pathways); and the Ministry of Defense of Israel.

New method efficiently safeguards sensitive AI training data

The approach maintains an AI model’s accuracy while ensuring attackers can’t extract secret information.

Data privacy comes with a cost. There are security techniques that protect sensitive user data, like customer addresses, from attackers who may attempt to extract them from AI models — but they often make those models less accurate.

MIT researchers recently developed a framework, based on a new privacy metric called PAC Privacy, that could maintain the performance of an AI model while ensuring sensitive data, such as medical images or financial records, remain safe from attackers. Now, they’ve taken this work a step further by making their technique more computationally efficient, improving the tradeoff between accuracy and privacy, and creating a formal template that can be used to privatize virtually any algorithm without needing access to that algorithm’s inner workings.

The team utilized their new version of PAC Privacy to privatize several classic algorithms for data analysis and machine-learning tasks.

They also demonstrated that more “stable” algorithms are easier to privatize with their method. A stable algorithm’s predictions remain consistent even when its training data are slightly modified. Greater stability helps an algorithm make more accurate predictions on previously unseen data.

The researchers say the increased efficiency of the new PAC Privacy framework, and the four-step template one can follow to implement it, would make the technique easier to deploy in real-world situations.

“We tend to consider robustness and privacy as unrelated to, or perhaps even in conflict with, constructing a high-performance algorithm. First, we make a working algorithm, then we make it robust, and then private. We’ve shown that is not always the right framing. If you make your algorithm perform better in a variety of settings, you can essentially get privacy for free,” says Mayuri Sridhar, an MIT graduate student and lead author of a paper on this privacy framework.

She is joined in the paper by Hanshen Xiao PhD ’24, who will start as an assistant professor at Purdue University in the fall; and senior author Srini Devadas, the Edwin Sibley Webster Professor of Electrical Engineering at MIT. The research will be presented at the IEEE Symposium on Security and Privacy.

Estimating noise

To protect sensitive data that were used to train an AI model, engineers often add noise, or generic randomness, to the model so it becomes harder for an adversary to guess the original training data. This noise reduces a model’s accuracy, so the less noise one can add, the better.

PAC Privacy automatically estimates the smallest amount of noise one needs to add to an algorithm to achieve a desired level of privacy.

The original PAC Privacy algorithm runs a user’s AI model many times on different samples of a dataset. It measures the variance as well as correlations among these many outputs and uses this information to estimate how much noise needs to be added to protect the data.

This new variant of PAC Privacy works the same way but does not need to represent the entire matrix of data correlations across the outputs; it just needs the output variances.

“Because the thing you are estimating is much, much smaller than the entire covariance matrix, you can do it much, much faster,” Sridhar explains. This means that one can scale up to much larger datasets.

Adding noise can hurt the utility of the results, and it is important to minimize utility loss. Due to computational cost, the original PAC Privacy algorithm was limited to adding isotropic noise, which is added uniformly in all directions. Because the new variant estimates anisotropic noise, which is tailored to specific characteristics of the training data, a user could add less overall noise to achieve the same level of privacy, boosting the accuracy of the privatized algorithm.

Privacy and stability

As she studied PAC Privacy, Sridhar hypothesized that more stable algorithms would be easier to privatize with this technique. She used the more efficient variant of PAC Privacy to test this theory on several classical algorithms.

Algorithms that are more stable have less variance in their outputs when their training data change slightly. PAC Privacy breaks a dataset into chunks, runs the algorithm on each chunk of data, and measures the variance among outputs. The greater the variance, the more noise must be added to privatize the algorithm.

Employing stability techniques to decrease the variance in an algorithm’s outputs would also reduce the amount of noise that needs to be added to privatize it, she explains.

“In the best cases, we can get these win-win scenarios,” she says.

The team showed that these privacy guarantees remained strong despite the algorithm they tested, and that the new variant of PAC Privacy required an order of magnitude fewer trials to estimate the noise. They also tested the method in attack simulations, demonstrating that its privacy guarantees could withstand state-of-the-art attacks.

“We want to explore how algorithms could be co-designed with PAC Privacy, so the algorithm is more stable, secure, and robust from the beginning,” Devadas says. The researchers also want to test their method with more complex algorithms and further explore the privacy-utility tradeoff.

“The question now is: When do these win-win situations happen, and how can we make them happen more often?” Sridhar says.

“I think the key advantage PAC Privacy has in this setting over other privacy definitions is that it is a black box — you don’t need to manually analyze each individual query to privatize the results. It can be done completely automatically. We are actively building a PAC-enabled database by extending existing SQL engines to support practical, automated, and efficient private data analytics,” says Xiangyao Yu, an assistant professor in the computer sciences department at the University of Wisconsin at Madison, who was not involved with this study.

This research is supported, in part, by Cisco Systems, Capital One, the U.S. Department of Defense, and a MathWorks Fellowship.

Hopping gives this tiny robot a leg up

MIT engineers developed an insect-sized jumping robot that can traverse challenging terrains and carry heavy payloads.

Insect-scale robots can squeeze into places their larger counterparts can’t, like deep into a collapsed building to search for survivors after an earthquake.

However, as they move through the rubble, tiny crawling robots might encounter tall obstacles they can’t climb over or slanted surfaces they will slide down. While aerial robots could avoid these hazards, the amount of energy required for flight would severely limit how far the robot can travel into the wreckage before it needs to return to base and recharge.

To get the best of both locomotion methods, MIT researchers developed a hopping robot that can leap over tall obstacles and jump across slanted or uneven surfaces, while using far less energy than an aerial robot.

The hopping robot, which is smaller than a human thumb and weighs less than a paperclip, has a springy leg that propels it off the ground, and four flapping-wing modules that give it lift and control its orientation.

The robot can jump about 20 centimeters into the air, or four times its height, at a lateral speed of about 30 centimeters per second, and has no trouble hopping across ice, wet surfaces, and uneven soil, or even onto a hovering drone. All the while, the hopping robot consumes about 60 percent less energy than its flying cousin.

Due to its light weight and durability, and the energy efficiency of the hopping process, the robot could carry about 10 times more payload than a similar-sized aerial robot, opening the door to many new applications.

“Being able to put batteries, circuits, and sensors on board has become much more feasible with a hopping robot than a flying one. Our hope is that one day this robot could go out of the lab and be useful in real-world scenarios,” says Yi-Hsuan (Nemo) Hsiao, an MIT graduate student and co-lead author of a paper on the hopping robot.

Hsiao is joined on the paper by co-lead authors Songnan Bai, a research assistant professor at The University of Hong Kong; and Zhongtao Guan, an incoming MIT graduate student who completed this work as a visiting undergraduate; as well as Suhan Kim and Zhijian Ren of MIT; and senior authors Pakpong Chirarattananon, an associate professor of the City University of Hong Kong; and Kevin Chen, an associate professor in the MIT Department of Electrical Engineering and Computer Science and head of the Soft and Micro Robotics Laboratory within the Research Laboratory of Electronics. The research appears today in Science Advances.

Maximizing efficiency

Jumping is common among insects, from fleas that leap onto new hosts to grasshoppers that bound around a meadow. While jumping is less common among insect-scale robots, which usually fly or crawl, hopping affords many advantages for energy efficiency.

When a robot hops, it transforms potential energy, which comes from its height off the ground, into kinetic energy as it falls. This kinetic energy transforms back to potential energy when it hits the ground, then back to kinetic as it rises, and so on.

To maximize efficiency of this process, the MIT robot is fitted with an elastic leg made from a compression spring, which is akin to the spring on a click-top pen. This spring converts the robot’s downward velocity to upward velocity when it strikes the ground.

“If you have an ideal spring, your robot can just hop along without losing any energy. But since our spring is not quite ideal, we use the flapping modules to compensate for the small amount of energy it loses when it makes contact with the ground,” Hsiao explains.

As the robot bounces back up into the air, the flapping wings provide lift, while ensuring the robot remains upright and has the correct orientation for its next jump. Its four flapping-wing mechanisms are powered by soft actuators, or artificial muscles, that are durable enough to endure repeated impacts with the ground without being damaged.

“We have been using the same robot for this entire series of experiments, and we never needed to stop and fix it,” Hsiao adds.

Key to the robot’s performance is a fast control mechanism that determines how the robot should be oriented for its next jump. Sensing is performed using an external motion-tracking system, and an observer algorithm computes the necessary control information using sensor measurements.

As the robot hops, it follows a ballistic trajectory, arcing through the air. At the peak of that trajectory, it estimates its landing position. Then, based on its target landing point, the controller calculates the desired takeoff velocity for the next jump. While airborne, the robot flaps its wings to adjust its orientation so it strikes the ground with the correct angle and axis to move in the proper direction and at the right speed.

Durability and flexibility

The researchers put the hopping robot, and its control mechanism, to the test on a variety of surfaces, including grass, ice, wet glass, and uneven soil — it successfully traversed all surfaces. The robot could even hop on a surface that was dynamically tilting.

“The robot doesn’t really care about the angle of the surface it is landing on. As long as it doesn’t slip when it strikes the ground, it will be fine,” Hsiao says.

Since the controller can handle multiple terrains, the robot can easily transition from one surface to another without missing a beat.

For instance, hopping across grass requires more thrust than hopping across glass, since blades of grass cause a damping effect that reduces its jump height. The controller can pump more energy to the robot’s wings during its aerial phase to compensate.

Due to its small size and light weight, the robot has an even smaller moment of inertia, which makes it more agile than a larger robot and better able to withstand collisions.

The researchers showcased its agility by demonstrating acrobatic flips. The featherweight robot could also hop onto an airborne drone without damaging either device, which could be useful in collaborative tasks.

In addition, while the team demonstrated a hopping robot that carried twice its weight, the maximum payload may be much higher. Adding more weight doesn’t hurt the robot’s efficiency. Rather, the efficiency of the spring is the most significant factor that limits how much the robot can carry.

Moving forward, the researchers plan to leverage its ability to carry heavy loads by installing batteries, sensors, and other circuits onto the robot, in the hopes of enabling it to hop autonomously outside the lab.

“Multimodal robots (those combining multiple movement strategies) are generally challenging and particularly impressive at such a tiny scale. The versatility of this tiny multimodal robot — flipping, jumping on rough or moving terrain, and even another robot — makes it even more impressive,” says Justin Yim, assistant professor at the University of Illinois at Urbana-Champagne, who was not involved with this work. “Continuous hopping shown in this research enables agile and efficient locomotion in environments with many large obstacles.”

This research is funded, in part, by the U.S. National Science Foundation and the MIT MISTI program. Chirarattananon was supported by the Research Grants Council of the Hong Kong Special Administrative Region of China. Hsiao is supported by a MathWorks Fellowship, and Kim is supported by a Zakhartchenko Fellowship.

Could LLMs help design our next medicines and materials?

A new method lets users ask, in plain language, for a new molecule with certain properties, and receive a detailed description of how to synthesize it.

The process of discovering molecules that have the properties needed to create new medicines and materials is cumbersome and expensive, consuming vast computational resources and months of human labor to narrow down the enormous space of potential candidates.

Large language models (LLMs) like ChatGPT could streamline this process, but enabling an LLM to understand and reason about the atoms and bonds that form a molecule, the same way it does with words that form sentences, has presented a scientific stumbling block.

Researchers from MIT and the MIT-IBM Watson AI Lab created a promising approach that augments an LLM with other machine-learning models known as graph-based models, which are specifically designed for generating and predicting molecular structures.

Their method employs a base LLM to interpret natural language queries specifying desired molecular properties. It automatically switches between the base LLM and graph-based AI modules to design the molecule, explain the rationale, and generate a step-by-step plan to synthesize it. It interleaves text, graph, and synthesis step generation, combining words, graphs, and reactions into a common vocabulary for the LLM to consume.

When compared to existing LLM-based approaches, this multimodal technique generated molecules that better matched user specifications and were more likely to have a valid synthesis plan, improving the success ratio from 5 percent to 35 percent.

It also outperformed LLMs that are more than 10 times its size and that design molecules and synthesis routes only with text-based representations, suggesting multimodality is key to the new system’s success.

“This could hopefully be an end-to-end solution where, from start to finish, we would automate the entire process of designing and making a molecule. If an LLM could just give you the answer in a few seconds, it would be a huge time-saver for pharmaceutical companies,” says Michael Sun, an MIT graduate student and co-author of a paper on this technique.

Sun’s co-authors include lead author Gang Liu, a graduate student at the University of Notre Dame; Wojciech Matusik, a professor of electrical engineering and computer science at MIT who leads the Computational Design and Fabrication Group within the Computer Science and Artificial Intelligence Laboratory (CSAIL); Meng Jiang, associate professor at the University of Notre Dame; and senior author Jie Chen, a senior research scientist and manager in the MIT-IBM Watson AI Lab. The research will be presented at the International Conference on Learning Representations.

Best of both worlds

Large language models aren’t built to understand the nuances of chemistry, which is one reason they struggle with inverse molecular design, a process of identifying molecular structures that have certain functions or properties.

LLMs convert text into representations called tokens, which they use to sequentially predict the next word in a sentence. But molecules are “graph structures,” composed of atoms and bonds with no particular ordering, making them difficult to encode as sequential text.

On the other hand, powerful graph-based AI models represent atoms and molecular bonds as interconnected nodes and edges in a graph. While these models are popular for inverse molecular design, they require complex inputs, can’t understand natural language, and yield results that can be difficult to interpret.

The MIT researchers combined an LLM with graph-based AI models into a unified framework that gets the best of both worlds.

Llamole, which stands for large language model for molecular discovery, uses a base LLM as a gatekeeper to understand a user’s query — a plain-language request for a molecule with certain properties.

For instance, perhaps a user seeks a molecule that can penetrate the blood-brain barrier and inhibit HIV, given that it has a molecular weight of 209 and certain bond characteristics.

As the LLM predicts text in response to the query, it switches between graph modules.

One module uses a graph diffusion model to generate the molecular structure conditioned on input requirements. A second module uses a graph neural network to encode the generated molecular structure back into tokens for the LLMs to consume. The final graph module is a graph reaction predictor which takes as input an intermediate molecular structure and predicts a reaction step, searching for the exact set of steps to make the molecule from basic building blocks.

The researchers created a new type of trigger token that tells the LLM when to activate each module. When the LLM predicts a “design” trigger token, it switches to the module that sketches a molecular structure, and when it predicts a “retro” trigger token, it switches to the retrosynthetic planning module that predicts the next reaction step.

“The beauty of this is that everything the LLM generates before activating a particular module gets fed into that module itself. The module is learning to operate in a way that is consistent with what came before,” Sun says.

In the same manner, the output of each module is encoded and fed back into the generation process of the LLM, so it understands what each module did and will continue predicting tokens based on those data.

Better, simpler molecular structures

In the end, Llamole outputs an image of the molecular structure, a textual description of the molecule, and a step-by-step synthesis plan that provides the details of how to make it, down to individual chemical reactions.

In experiments involving designing molecules that matched user specifications, Llamole outperformed 10 standard LLMs, four fine-tuned LLMs, and a state-of-the-art domain-specific method. At the same time, it boosted the retrosynthetic planning success rate from 5 percent to 35 percent by generating molecules that are higher-quality, which means they had simpler structures and lower-cost building blocks.

“On their own, LLMs struggle to figure out how to synthesize molecules because it requires a lot of multistep planning. Our method can generate better molecular structures that are also easier to synthesize,” Liu says.

To train and evaluate Llamole, the researchers built two datasets from scratch since existing datasets of molecular structures didn’t contain enough details. They augmented hundreds of thousands of patented molecules with AI-generated natural language descriptions and customized description templates.

The dataset they built to fine-tune the LLM includes templates related to 10 molecular properties, so one limitation of Llamole is that it is trained to design molecules considering only those 10 numerical properties.

In future work, the researchers want to generalize Llamole so it can incorporate any molecular property. In addition, they plan to improve the graph modules to boost Llamole’s retrosynthesis success rate.

And in the long run, they hope to use this approach to go beyond molecules, creating multimodal LLMs that can handle other types of graph-based data, such as interconnected sensors in a power grid or transactions in a financial market.

“Llamole demonstrates the feasibility of using large language models as an interface to complex data beyond textual description, and we anticipate them to be a foundation that interacts with other AI algorithms to solve any graph problems,” says Chen.

This research is funded, in part, by the MIT-IBM Watson AI Lab, the National Science Foundation, and the Office of Naval Research.

Exploring the impacts of technology on everyday citizens

Associate Professor Dwai Banerjee examines topics ranging from cancer care to the history of computing.

Give Dwai Banerjee credit: He doesn’t pick easy topics to study.

Banerjee is an MIT scholar who in a short time has produced a wide-ranging body of work about the impact of technology on society — and who, as a trained anthropologist, has a keen eye for people’s lived experience.

In one book, “Enduring Cancer,” from 2020, Banerjee studies the lives of mostly poor cancer patients in Delhi, digging into their psychological horizons and interactions with the world of medical care. Another book, “Hematologies,” also from 2020, co-authored with anthropologist Jacob Copeman, examines common ideas about blood in Indian society.

And in still another book, forthcoming later this year, Banerjee explores the history of computing in India — including the attempt by some to generate growth through domestic advances, even as global computer firms were putting the industry on rather different footing.

“I enjoy having the freedom to explore new topics,” says Banerjee, an associate professor in MIT’s Program in Science, Technology, and Society (STS). “For some people, building on their previous work is best, but I need new ideas to keep me going. For me, that feels more natural. You get invested in a subject for a time and try to get everything out of it.”

What largely links these disparate topics together is that Banerjee, in his work, is a people person: He aims to illuminate the lives and thoughts of everyday citizens as they interact with the technologies and systems of contemporary society.

After all, a cancer diagnosis can be life-changing not just in physical terms, but psychologically. For some, having cancer creates “a sense of being unmoored from prior certainties about oneself and one’s place in the world,” as Banerjee writes in “Enduring Cancer.”

The technology that enables diagnoses does not meet all our human needs, so the book traces the complicated inner lives of patients, and a medical system shifting to meet psychological and palliative-care challenges. Technology and society interact beyond blockbuster products, as the book deftly implies.

For his research and teaching, Banerjee was awarded tenure at MIT last year.

Falling for the humanities

Banerjee grew up in Delhi, and as a university student he expected to work in computing, before changing course.

“I was going to go to graduate school for computer engineering,” Banerjee says. “Then I just fell in love with the humanities, and studied the humanities and social sciences.” He received an MPhil and an MA in sociology from the Delhi School of Economics, then enrolled as a PhD student at New York University.

At NYU, Banerjee undertook doctoral studies in cultural anthropology, while performing some of the fieldwork that formed the basis of “Enduring Cancer.” At the same time, he found the people he was studying were surrounded by history — shaping the technologies and policies they encountered, and shaping their own thought. Ultimately even Banerjee’s anthropological work has a strong historical dimension.

After earning his PhD, Banerjee became a Mellon Fellow in the Humanities at Dartmouth College, then joined the MIT faculty in STS. It is a logical home for someone who thinks broadly and uses multiple research methods, from the field to the archives.

“I sometimes wonder if I am an anthropologist or if I am an historian,” Banerjee allows. “But it is an interdisciplinary program, so I try to make the most of that.”

Indeed, the STS program draws on many fields and methods, with its scholars and students linked by a desire to rigorously examine the factors shaping the development and application of technology — and, if necessary, to initiate difficult discussions about technology’s effects.

“That’s the history of the field and the department at MIT, that it’s a kind of moral backbone,” Banerjee says.

Finding inspiration

As for where Banerjee’s book ideas come from, he is not simply looking for large issues to write about, but things that spark his intellectual and moral sensibilities — like disadvantaged cancer patients in Delhi.

“‘Enduring Cancer,’ in my mind, is a sort of a traditional medical anthropology text, which came out of finding inspiration from these people, and running with it as far as I could,” Banerjee says.

Alternately, “‘Hematologies’ came out of a collaboration, a conversation with Jacob Copeman, with us talking about things and getting excited about it,” Banerjee adds. “The intellectual friendship became a driving force.” Copeman is now an anthropologist on the faculty at the University of Santiago de Compostela, in Spain.

As for Banerjee’s forthcoming book about computing in India, the spark was partly his own remembered enjoyment of seeing the internet reach the country, facilitated though it was by spotty dial-up modems and other now-quaint-seeming tools.

“It’s coming from an old obsession,” Banerjee says. “When the internet had just arrived, at that time when something was just blowing up, it was exciting. This project is [partly about] recovering my early enjoyment of what was then a really exciting time.”

The subject of the book itself, however, predates the commercial internet. Rather, Banerjee chronicles the history of computing during India’s first few decades after achieving independence from Britain, in 1947. Even into the 1970s, India’s government was interested in creating a strong national IT sector, designing and manufacturing its own machines. Eventually those efforts faded, and the multinational computing giants took ahold of India’s markets.

The book details how and why this happened, in the process recasting what we think we know about India and technology. Today, Banerjee notes, India is an exporter of skilled technology talent and an importer of tech tools, but that wasn’t predestined. It is more that the idea of an autonomous tech sector in the country ran into the prevailing forces of globalization.

“The book traces this moment of this high confidence in the country’s ability to do these things, producing manufacturing and jobs and economic growth, and then it traces the decline of that vision,” Banerjee says.

“One of the aims is for it to be a book anyone can read,” Banerjee adds. In that sense, the principle guiding his interests is now guiding his scholarly output: People first.

Study: Burning heavy fuel oil with scrubbers is the best available option for bulk maritime shipping

Researchers analyzed the full lifecycle of several fuel options and found this approach has a comparable environmental impact, overall, to burning low-sulfur fuels.

When the International Maritime Organization enacted a mandatory cap on the sulfur content of marine fuels in 2020, with an eye toward reducing harmful environmental and health impacts, it left shipping companies with three main options.

They could burn low-sulfur fossil fuels, like marine gas oil, or install cleaning systems to remove sulfur from the exhaust gas produced by burning heavy fuel oil. Biofuels with lower sulfur content offer another alternative, though their limited availability makes them a less feasible option.

While installing exhaust gas cleaning systems, known as scrubbers, is the most feasible and cost-effective option, there has been a great deal of uncertainty among firms, policymakers, and scientists as to how “green” these scrubbers are.

Through a novel lifecycle assessment, researchers from MIT, Georgia Tech, and elsewhere have now found that burning heavy fuel oil with scrubbers in the open ocean can match or surpass using low-sulfur fuels, when a wide variety of environmental factors is considered.

The scientists combined data on the production and operation of scrubbers and fuels with emissions measurements taken onboard an oceangoing cargo ship.

They found that, when the entire supply chain is considered, burning heavy fuel oil with scrubbers was the least harmful option in terms of nearly all 10 environmental impact factors they studied, such as greenhouse gas emissions, terrestrial acidification, and ozone formation.

“In our collaboration with Oldendorff Carriers to broadly explore reducing the environmental impact of shipping, this study of scrubbers turned out to be an unexpectedly deep and important transitional issue,” says Neil Gershenfeld, an MIT professor, director of the Center for Bits and Atoms (CBA), and senior author of the study.

“Claims about environmental hazards and policies to mitigate them should be backed by science. You need to see the data, be objective, and design studies that take into account the full picture to be able to compare different options from an apples-to-apples perspective,” adds lead author Patricia Stathatou, an assistant professor at Georgia Tech, who began this study as a postdoc in the CBA.

Stathatou is joined on the paper by Michael Triantafyllou, the Henry L. and Grace Doherty Professor in Ocean Science and Engineering in the Department of Mechanical Engineering and others at the National Technical University of Athens in Greece, Naias Laboratories, and the maritime shipping firm Oldendorff Carriers. The research appears today in Environmental Science and Technology.

Slashing sulfur emissions

Heavy fuel oil, traditionally burned by bulk carriers that make up about 30 percent of the global maritime fleet, usually has a sulfur content around 2 to 3 percent. This is far higher than the International Maritime Organization’s 2020 cap of 0.5 percent in most areas of the ocean and 0.1 percent in areas near population centers or environmentally sensitive regions.

Sulfur oxide emissions contribute to air pollution and acid rain, and can damage the human respiratory system.

In 2018, fewer than 1,000 vessels employed scrubbers. After the cap went into place, higher prices of low-sulfur fossil fuels and limited availability of alternative fuels led many firms to install scrubbers so they could keep burning heavy fuel oil.

Today, more than 5,800 vessels utilize scrubbers, the majority of which are wet, open-loop scrubbers.

“Scrubbers are a very mature technology. They have traditionally been used for decades in land-based applications like power plants to remove pollutants,” Stathatou says.

A wet, open-loop marine scrubber is a huge, metal, vertical tank installed in a ship’s exhaust stack, above the engines. Inside, seawater drawn from the ocean is sprayed through a series of nozzles downward to wash the hot exhaust gases as they exit the engines.

The seawater interacts with sulfur dioxide in the exhaust, converting it to sulfates — water-soluble, environmentally benign compounds that naturally occur in seawater. The washwater is released back into the ocean, while the cleaned exhaust escapes to the atmosphere with little to no sulfur dioxide emissions.

But the acidic washwater can contain other combustion byproducts like heavy metals, so scientists wondered if scrubbers were comparable, from a holistic environmental point of view, to burning low-sulfur fuels.

Several studies explored toxicity of washwater and fuel system pollution, but none painted a full picture.

The researchers set out to fill that scientific gap.

A “well-to-wake” analysis

The team conducted a lifecycle assessment using a global environmental database on production and transport of fossil fuels, such as heavy fuel oil, marine gas oil, and very-low sulfur fuel oil. Considering the entire lifecycle of each fuel is key, since producing low-sulfur fuel requires extra processing steps in the refinery, causing additional emissions of greenhouse gases and particulate matter.

“If we just look at everything that happens before the fuel is bunkered onboard the vessel, heavy fuel oil is significantly more low-impact, environmentally, than low-sulfur fuels,” she says.

The researchers also collaborated with a scrubber manufacturer to obtain detailed information on all materials, production processes, and transportation steps involved in marine scrubber fabrication and installation.

“If you consider that the scrubber has a lifetime of about 20 years, the environmental impacts of producing the scrubber over its lifetime are negligible compared to producing heavy fuel oil,” she adds.

For the final piece, Stathatou spent a week onboard a bulk carrier vessel in China to measure emissions and gather seawater and washwater samples. The ship burned heavy fuel oil with a scrubber and low-sulfur fuels under similar ocean conditions and engine settings.

Collecting these onboard data was the most challenging part of the study.

“All the safety gear, combined with the heat and the noise from the engines on a moving ship, was very overwhelming,” she says.

Their results showed that scrubbers reduce sulfur dioxide emissions by 97 percent, putting heavy fuel oil on par with low-sulfur fuels according to that measure. The researchers saw similar trends for emissions of other pollutants like carbon monoxide and nitrous oxide.

In addition, they tested washwater samples for more than 60 chemical parameters, including nitrogen, phosphorus, polycyclic aromatic hydrocarbons, and 23 metals.

The concentrations of chemicals regulated by the IMO were far below the organization’s requirements. For unregulated chemicals, the researchers compared the concentrations to the strictest limits for industrial effluents from the U.S. Environmental Protection Agency and European Union.

Most chemical concentrations were at least an order of magnitude below these requirements.

In addition, since washwater is diluted thousands of times as it is dispersed by a moving vessel, the concentrations of such chemicals would be even lower in the open ocean.

These findings suggest that the use of scrubbers with heavy fuel oil can be considered as equal to or more environmentally friendly than low-sulfur fuels across many of the impact categories the researchers studied.

“This study demonstrates the scientific complexity of the waste stream of scrubbers. Having finally conducted a multiyear, comprehensive, and peer-reviewed study, commonly held fears and assumptions are now put to rest,” says Scott Bergeron, managing director at Oldendorff Carriers and co-author of the study.

“This first-of-its-kind study on a well-to-wake basis provides very valuable input to ongoing discussion at the IMO,” adds Thomas Klenum, executive vice president of innovation and regulatory affairs at the Liberian Registry, emphasizing the need “for regulatory decisions to be made based on scientific studies providing factual data and conclusions.”

Ultimately, this study shows the importance of incorporating lifecycle assessments into future environmental impact reduction policies, Stathatou says.

“There is all this discussion about switching to alternative fuels in the future, but how green are these fuels? We must do our due diligence to compare them equally with existing solutions to see the costs and benefits,” she adds.

This study was supported, in part, by Oldendorff Carriers.

MIT graduate engineering and business programs ranked highly by U.S. News for 2025-26

Graduate engineering program is No. 1 in the nation; MIT Sloan is No. 5.

U.S. News and World Report has again placed MIT’s graduate program in engineering at the top of its annual rankings, released today. The Institute has held the No. 1 spot since 1990, when the magazine first ranked such programs.

The MIT Sloan School of Management also placed highly, in rankings announced April 8. It occupies the No. 5 spot for the best graduate business programs.

Among individual engineering disciplines, MIT placed first in six areas: aerospace/aeronautical/astronautical engineering, chemical engineering, computer engineering (tied with the University of California at Berkeley), electrical/electronic/communications engineering (tied with Stanford University and Berkeley), materials engineering, and mechanical engineering. It placed second in nuclear engineering and third in biomedical engineering/bioengineering.

In the rankings of individual MBA specialties, MIT placed first in four areas: information systems, production/operations, project management, and supply chain/logistics. It placed second in business analytics and third in entrepreneurship.

U.S. News bases its rankings of graduate schools of engineering and business on two types of data: reputational surveys of deans and other academic officials, and statistical indicators that measure the quality of a school’s faculty, research, and students. The magazine’s less-frequent rankings of graduate programs in the sciences, social sciences, and humanities are based solely on reputational surveys. Among the peer-review disciplines ranked this year, MIT placed first in computer science, and its doctoral program in economics also placed first (tied with Harvard University, Stanford, Berkeley, and the University of Chicago).

Supersize me

Political scientist Kathleen Thelen’s new book explains how America’s large retailers got very, very large.

Well into the late 19th century, the U.S. retail sector was overwhelmingly local, consisting of small, independent merchants throughout the country. That started changing after Sears and Roebuck’s famous catalog became popular, allowing the firm to grow, while a rival, Montgomery Ward, also expanded. By the 1930s, the U.S. had 130,000 chain stores, topped by Atlantic and Pacific supermarkets (the A&P), with over 15,000 stores.

A century onward, the U.S. retail landscape is dominated by retail giants. Today, 90 percent of Americans live within 10 miles of a Walmart, while five of the country’s 10 biggest employers — Walmart, Amazon, Home Depot, Kroger, and Target— are retailers. Two others in the top 10, UPS and FedEx, are a major part of the retail economy.

The ubiquity of these big retailers, and the sheer extent of the U.S. shopping economy as a whole, is unusual compared to the country’s European counterparts. Domestic consumption plays an outsized role in driving growth in the United States, and credit plays a much larger role in supporting that consumption than in Europe. The U.S. has five times as much retail space per capita as Japan and the U.K., and 10 times as much as Germany. Unlike in Europe, shopping hours are largely unregulated.

How did this happen? To be sure, Walmart, Amazon, Target, and other massive chains have plenty of business acumen. But the full story involves a century or more of political tectonics and legal debates, which helped shape the size of U.S. retailing and the prominence of its large discount chains. 

“The markets that we take as given, that we think of as the natural outcome of supply and demand, are heavily shaped by policy and by politics,” says MIT political scientist Kathleen Thelen.

Thelen examines the subject in a new book, “Attention, Shoppers! American Retail Capitalism and the Origins of the Amazon Economy,” published today by Princeton University Press. In it, she examines the growth of the particular model of supersized, low-cost, low-wage retailing that now features so prominently in the U.S. economy.

Prioritizing prices

While a great deal has been written about specific American companies, Thelen’s book has some distinctive features. One is a comparison to the economies of Europe, where she has focused much of her scholarship. Another is her historical lens, extending back to the start of chain retailing.

“It seems like every time I set out to explain something in the present, I’m thrown back to the 19th century,” Thelen says.

For instance, as both Sears and Montgomery Ward grew, producers and consumers were still experimenting with alternate commercial arrangements, like cooperatives, which pooled suppliers together, but they ultimately ran into economic and legal headwinds. Especially, at the time, legal headwinds.

“Antitrust laws in the United States were very forbearing toward big multidivisional corporations and very punitive toward alternative types of arrangements like cooperatives, so big retailers got a real boost in that period,” Thelen says. Separately, the U.S. Postal Service was also crucial, since big mail order houses like Sears relied on not just on its delivery services but also its money order system, to sell goods to the company’s many customers who lacked bank accounts.

Smaller retailers fought large chains during the Depression, especially in the South and the West, which forms another phase of the story. But low-cost discounters worked around some laws through regulatory arbitrage, finding friendlier regulations in some states — and sometimes though outright rule-breaking. Ultimately, larger retailers have thrived again in the last half century, especially as antitrust law increasingly prioritized consumer prices as its leading measuring stick.

Most antitrust theorizing since the 1960s “valorizes consumer welfare, which is basically defined as price, so anything that delivers the lowest price to consumers is A-OK,” Thelen says. “We’re in this world where the large, low-cost retailers are delivering consumer welfare in the way the courts are defining it.”

That emphasis on prices, she notes, then spills over into other areas of the economy, especially wages and labor relations.

“If you prioritize prices, one of the main ways to reduce prices is to reduce labor costs,” Thelen says. “It’s no coincidence that low-cost discounters are often low-wage employers. Indeed, they often squeeze their vendors to deliver goods at ever-lower prices, and by extension they’re pressing down on wages in their supplier networks as well.”

As Thelen’s book explains, legal views supporting large chains were also common during the first U.S. wave of chain-retail growth. She writes, “large, low-cost retailers have almost always enjoyed a privileged position in the American antitrust regime.”

In the “deep equilibrium”

“Attention, Shoppers!” makes clear that this tendency toward lower prices, lower employee pay, and high consumer convenience is particularly pronounced in the U.S., where 22.6 percent of employees count as low-wage workers (making two-thirds or less of the country’s median wage). In the other countries that belong to the Organization for Economic Cooperation and Development, 13.9 percent of workers fit that description. About three-quarters of U.S. retail workers are in the low-wage category.

In other OECD countries, on aggregate, manufacturers and producers make up bigger chunks of the economy and, correspondingly, often have legal frameworks more friendly to manufacturers and to labor. But in the U.S., large retailers have gained more leverage, if anything, in the last half-century, Thelen notes.

“You might think mass retailers and manufacturers would have a symbiotic relationship, but historically there has been great tension between them, especially on price,” Thelen says. “In the postwar period, the balance of power became tilted toward retailers, and away from manufacturers and labor. Retailers also had consumers on their side, and had more power over data to dictate the terms on which their vendors would supply goods to them.”

Currently, as Thelen writes in the book, the U.S. is in a “deep equilibrium” on this front, in that many low-wage workers now rely on these low-cost retailers to make ends meet — and because Americans as a whole now find it normal to have their purchases delivered at lightning speed. Things might be different, Thelen suggests, if there are changes to U.S. antitrust enforcement, or, especially, major reforms to labor law, such as allowing workers to organize for higher wages across companies, not just at individual stores. Short of that, the equilibrium is likely to hold.

“Attention, Shoppers!” has received praise from other scholars. Louis Hyman, a historian at Johns Hopkins University, has called it a “pathbreaking study that provides insight into not only the past but also the future of online retail.”

For her part, Thelen hopes readers will learn more about an economic landscape we might take for granted, even while we shop at big chains, around us and online.

“The triumph of these types of retailers was not inevitable,” Thelen says. “It was a function of politics and political choice.”

3Q: MIT’s Lonnie Petersen on the first medical X-ray taken in space

Performed in microgravity, 200 miles above the Earth’s surface, the imaging procedure could help keep astronauts safe and healthy on long-term missions.

Many of us have gotten an X-ray at one time or another, either at the dentist’s or the doctor’s office. Now, astronauts orbiting Earth have shown it’s possible to take an X-ray in space. The effort will help future space explorers diagnose and monitor medical conditions, from fractures and sprains to signs of bone decalcification, while in orbit.

Last week, crew members aboard the Fram2 mission posted to social media and shared the first-ever medical X-ray image taken in space. The image is a black-and-white scan of a hand with a ring, echoing the very first X-ray image ever taken, 130 years ago, by the physicist Wilhelm Roentgen, of his wife’s hand. The new X-ray image was taken in microgravity, inside a four-person space capsule flying at orbital speeds of 17,500 miles per hour, about 200 miles above the Earth’s surface.

The in-flight body scan was part of the SpaceXray project, one of 22 science experiments that astronauts conducted during the Fram2 mission. Operated by SpaceX, Fram2 was the first human spaceflight mission to travel in a polar orbit, looping around the planet from pole to pole. Fram2 gets its mission name from the Norwegian ship “Fram,” which was the first to carry explorers to the Arctic and Antarctic regions in the late 19th century.

The body scans are a first demonstration that medical X-ray imaging can be done within the confines and conditions in space. Lonnie Petersen, a co-investigator on the SpaceXray project, is an associate professor in MIT’s Department of Aeronautics and Astronautics who studies space physiology and the effects of spaceflight on the human body. Petersen helped to define and design the protocol around the SpaceXray project, in collaboration with institutional partners such as Stanford University and the Mayo Clinic, and X-ray hardware companies KA and MinXray. Petersen talked with MIT News about how these first in-orbit X-ray images can help enable safe and healthy longer-term missions in space.

Q: What are the challenges in taking an X-ray in space, versus here on Earth?

A: There are several challenges regarding the hardware, methods, and subjects being X-rayed.

To get hardware certified for spaceflight, it should be miniaturized and as lightweight as possible. There are also increased safety requirements because all devices work in a confined space. The increased requirements drive technology development. I always say space is our best technology accelerator — this is also true for medical technology.

For this project we used a portable, specialized X-ray generator and detector developed by MinXray and KA Imaging for the battlefield and made it applicable for spaceflight.

In terms of methods, one of my concerns was that the increased background radiation might reduce the quality of the image so that it would fall below clinical standards. From the first images we have received from space, it seems that the quality is great. I am very excited to further analyze the full set of images.

We want the X-rays to travel straight through the body part of interest. This requires alignment of equipment and patient.  As you can imagine, a floating subject will be harder to position. We will be quantifying any potential impact of this and using it for future updates.

We also do not have radiologists or technicians in space. The methods need to be simple and robust enough for a layperson to operate them.

And, finally, regarding subjects: Entry into space has huge impact on the human body. Blood and fluid are no longer pulled down toward the feet by gravity; they are evenly distributed, and thus there are regional pressures and perfusion changes. The cardiovascular system and the brain are impacted by this over time. Mechanical unloading of the body leads to muscle atrophy and bone decalcification as well reduction in exercise capacity. This mission was only 3.5 days, so the crew will likely not have experienced many negative effects, but with an X-ray, we can now monitor bone health in space. We have never been able to do that before. We can monitor potential fluid buildup in the lungs or check for diseases in the abdomen.

I’ll also take off my physician hat and put on my engineering hat: X-rays are a useful tool in nondestructive hardware tests in aviation (and other areas). This project increases our diagnostic capabilities in space, not just for patients, but also for hardware.

Q: How did the Fram2 crew do it?

A: The crew learned how to take X-rays in one afternoon. It was done as a train-the-trainer model. The protocol was created in advance and the crew took images of each other, checked the quality, and stored the images. We have only seen one image so far, but from that, we are very impressed with the quality, the skills, and the dedication to advancing science by the crew.

Q: What will you learn from these first images?

A: First and foremost, this was a technology demonstration: Can we even do this in space? We are looking forward to analyzing all the images, but from preliminary data it looks like we absolutely can. Now comes a detailed analysis to tease out all the lessons we possibly can learn from this both with regard to current capabilities but also the next steps. The team is, of course, very excited to carry this research forward and break even more ground.

Molecules that fight infection also act on the brain, inducing anxiety or sociability

New research on a cytokine called IL-17 adds to growing evidence that immune molecules can influence behavior during illness.

Immune molecules called cytokines play important roles in the body’s defense against infection, helping to control inflammation and coordinating the responses of other immune cells. A growing body of evidence suggests that some of these molecules also influence the brain, leading to behavioral changes during illness.

Two new studies from MIT and Harvard Medical School, focused on a cytokine called IL-17, now add to that evidence. The researchers found that IL-17 acts on two distinct brain regions — the amygdala and the somatosensory cortex — to exert two divergent effects. In the amygdala, IL-17 can elicit feelings of anxiety, while in the cortex it promotes sociable behavior.

These findings suggest that the immune and nervous systems are tightly interconnected, says Gloria Choi, an associate professor of brain and cognitive sciences, a member of MIT’s Picower Institute for Learning and Memory, and one of the senior authors of the studies.

“If you’re sick, there’s so many more things that are happening to your internal states, your mood, and your behavioral states, and that’s not simply you being fatigued physically. It has something to do with the brain,” she says.

Jun Huh, an associate professor of immunology at Harvard Medical School, is also a senior author of both studies, which appear today in Cell. One of the papers was led by Picower Institute Research Scientist Byeongjun Lee and former Picower Institute research scientist Jeong-Tae Kwon, and the other was led by Harvard Medical School postdoc Yunjin Lee and Picower Institute postdoc Tomoe Ishikawa.

Behavioral effects

Choi and Huh became interested in IL-17 several years ago, when they found it was involved in a phenomenon known as the fever effect. Large-scale studies of autistic children have found that for many of them, their behavioral symptoms temporarily diminish when they have a fever.

In a 2019 study in mice, Choi and Huh showed that in some cases of infection, IL-17 is released and suppresses a small region of the brain’s cortex known as S1DZ. Overactivation of neurons in this region can lead to autism-like behavioral symptoms in mice, including repetitive behaviors and reduced sociability.

“This molecule became a link that connects immune system activation, manifested as a fever, to changes in brain function and changes in the animals’ behavior,” Choi says.

IL-17 comes in six different forms, and there are five different receptors that can bind to it. In their two new papers, the researchers set out to map which of these receptors are expressed in different parts of the brain. This mapping revealed that a pair of receptors known as IL-17RA and IL-17RB is found in the cortex, including in the S1DZ region that the researchers had previously identified. The receptors are located in a population of neurons that receive proprioceptive input and are involved in controlling behavior.

When a type of IL-17 known as IL-17E binds to these receptors, the neurons become less excitable, which leads to the behavioral effects seen in the 2019 study.

“IL-17E, which we’ve shown to be necessary for behavioral mitigation, actually does act almost exactly like a neuromodulator in that it will immediately reduce these neurons’ excitability,” Choi says. “So, there is an immune molecule that’s acting as a neuromodulator in the brain, and its main function is to regulate excitability of neurons.”

Choi hypothesizes that IL-17 may have originally evolved as a neuromodulator, and later on was appropriated by the immune system to play a role in promoting inflammation. That idea is consistent with previous work showing that in the worm C. elegans, IL-17 has no role in the immune system but instead acts on neurons. Among its effects in worms, IL-17 promotes aggregation, a form of social behavior. Additionally, in mammals, IL-17E is actually made by neurons in the cortex, including S1DZ.

“There’s a possibility that a couple of forms of IL-17 perhaps evolved first and foremost to act as a neuromodulator in the brain, and maybe later were hijacked by the immune system also to act as immune modulators,” Choi says.

Provoking anxiety

In the other Cell paper, the researchers explored another brain location where they found IL-17 receptors — the amygdala. This almond-shaped structure plays an important role in processing emotions, including fear and anxiety.

That study revealed that in a region known as the basolateral amygdala (BLA), the IL-17RA and IL-17RE receptors, which work as a pair, are expressed in a discrete population of neurons. When these receptors bind to IL-17A and IL-17C, the neurons become more excitable, leading to an increase in anxiety.

The researchers also found that, counterintuitively, if animals are treated with antibodies that block IL-17 receptors, it actually increases the amount of IL-17C circulating in the body. This finding may help to explain unexpected outcomes observed in a clinical trial of a drug targeting the IL-17-RA receptor for psoriasis treatment, particularly regarding its potential adverse effects on mental health.

“We hypothesize that there’s a possibility that the IL-17 ligand that is upregulated in this patient cohort might act on the brain to induce suicide ideation, while in animals there is an anxiogenic phenotype,” Choi says.

During infections, this anxiety may be a beneficial response, keeping the sick individual away from others to whom the infection could spread, Choi hypothesizes.

“Other than its main function of fighting pathogens, one of the ways that the immune system works is to control the host behavior, to protect the host itself and also protect the community the host belongs to,” she says. “One of the ways the immune system is doing that is to use cytokines, secreted factors, to go to the brain as communication tools.”

The researchers found that the same BLA neurons that have receptors for IL-17 also have receptors for IL-10, a cytokine that suppresses inflammation. This molecule counteracts the excitability generated by IL-17, giving the body a way to shut off anxiety once it’s no longer useful.

Distinctive behaviors

Together, the two studies suggest that the immune system, and even a single family of cytokines, can exert a variety of effects in the brain.

“We have now different combinations of IL-17 receptors being expressed in different populations of neurons, in two different brain regions, that regulate very distinct behaviors. One is actually somewhat positive and enhances social behaviors, and another is somewhat negative and induces anxiogenic phenotypes,” Choi says.

Her lab is now working on additional mapping of IL-17 receptor locations, as well as the IL-17 molecules that bind to them, focusing on the S1DZ region. Eventually, a better understanding of these neuro-immune interactions may help researchers develop new treatments for neurological conditions such as autism or depression.

“The fact that these molecules are made by the immune system gives us a novel approach to influence brain function as means of therapeutics,” Choi says. “Instead of thinking about directly going for the brain, can we think about doing something to the immune system?”

The research was funded, in part, by Jeongho Kim and the Brain Impact Foundation Neuro-Immune Fund, the Simons Foundation Autism Research Initiative, the Simons Center for the Social Brain, the Marcus Foundation, the N of One: Autism Research Foundation, the Burroughs Wellcome Fund, the Picower Institute Innovation Fund, the MIT John W. Jarve Seed Fund for Science Innovation, Young Soo Perry and Karen Ha, and the National Institutes of Health.

New method assesses and improves the reliability of radiologists’ diagnostic reports

The framework helps clinicians choose phrases that more accurately reflect the likelihood that certain conditions are present in X-rays.

Due to the inherent ambiguity in medical images like X-rays, radiologists often use words like “may” or “likely” when describing the presence of a certain pathology, such as pneumonia.

But do the words radiologists use to express their confidence level accurately reflect how often a particular pathology occurs in patients? A new study shows that when radiologists express confidence about a certain pathology using a phrase like “very likely,” they tend to be overconfident, and vice-versa when they express less confidence using a word like “possibly.”

Using clinical data, a multidisciplinary team of MIT researchers in collaboration with researchers and clinicians at hospitals affiliated with Harvard Medical School created a framework to quantify how reliable radiologists are when they express certainty using natural language terms.

They used this approach to provide clear suggestions that help radiologists choose certainty phrases that would improve the reliability of their clinical reporting. They also showed that the same technique can effectively measure and improve the calibration of large language models by better aligning the words models use to express confidence with the accuracy of their predictions.

By helping radiologists more accurately describe the likelihood of certain pathologies in medical images, this new framework could improve the reliability of critical clinical information.

“The words radiologists use are important. They affect how doctors intervene, in terms of their decision making for the patient. If these practitioners can be more reliable in their reporting, patients will be the ultimate beneficiaries,” says Peiqi Wang, an MIT graduate student and lead author of a paper on this research.

He is joined on the paper by senior author Polina Golland, a Sunlin and Priscilla Chou Professor of Electrical Engineering and Computer Science (EECS), a principal investigator in the MIT Computer Science and Artificial Intelligence Laboratory (CSAIL), and the leader of the Medical Vision Group; as well as Barbara D. Lam, a clinical fellow at the Beth Israel Deaconess Medical Center; Yingcheng Liu, at MIT graduate student; Ameneh Asgari-Targhi, a research fellow at Massachusetts General Brigham (MGB); Rameswar Panda, a research staff member at the MIT-IBM Watson AI Lab; William M. Wells, a professor of radiology at MGB and a research scientist in CSAIL; and Tina Kapur, an assistant professor of radiology at MGB. The research will be presented at the International Conference on Learning Representations.

Decoding uncertainty in words

A radiologist writing a report about a chest X-ray might say the image shows a “possible” pneumonia, which is an infection that inflames the air sacs in the lungs. In that case, a doctor could order a follow-up CT scan to confirm the diagnosis.

However, if the radiologist writes that the X-ray shows a “likely” pneumonia, the doctor might begin treatment immediately, such as by prescribing antibiotics, while still ordering additional tests to assess severity.

Trying to measure the calibration, or reliability, of ambiguous natural language terms like “possibly” and “likely” presents many challenges, Wang says.

Existing calibration methods typically rely on the confidence score provided by an AI model, which represents the model’s estimated likelihood that its prediction is correct.

For instance, a weather app might predict an 83 percent chance of rain tomorrow. That model is well-calibrated if, across all instances where it predicts an 83 percent chance of rain, it rains approximately 83 percent of the time.

“But humans use natural language, and if we map these phrases to a single number, it is not an accurate description of the real world. If a person says an event is ‘likely,’ they aren’t necessarily thinking of the exact probability, such as 75 percent,” Wang says.

Rather than trying to map certainty phrases to a single percentage, the researchers’ approach treats them as probability distributions. A distribution describes the range of possible values and their likelihoods — think of the classic bell curve in statistics.

“This captures more nuances of what each word means,” Wang adds.

Assessing and improving calibration

The researchers leveraged prior work that surveyed radiologists to obtain probability distributions that correspond to each diagnostic certainty phrase, ranging from “very likely” to “consistent with.”

For instance, since more radiologists believe the phrase “consistent with” means a pathology is present in a medical image, its probability distribution climbs sharply to a high peak, with most values clustered around the 90 to 100 percent range.

In contrast the phrase “may represent” conveys greater uncertainty, leading to a broader, bell-shaped distribution centered around 50 percent.

Typical methods evaluate calibration by comparing how well a model’s predicted probability scores align with the actual number of positive results.

The researchers’ approach follows the same general framework but extends it to account for the fact that certainty phrases represent probability distributions rather than probabilities.

To improve calibration, the researchers formulated and solved an optimization problem that adjusts how often certain phrases are used, to better align confidence with reality.

They derived a calibration map that suggests certainty terms a radiologist should use to make the reports more accurate for a specific pathology.

“Perhaps, for this dataset, if every time the radiologist said pneumonia was ‘present,’ they changed the phrase to ‘likely present’ instead, then they would become better calibrated,” Wang explains.

When the researchers used their framework to evaluate clinical reports, they found that radiologists were generally underconfident when diagnosing common conditions like atelectasis, but overconfident with more ambiguous conditions like infection.

In addition, the researchers evaluated the reliability of language models using their method, providing a more nuanced representation of confidence than classical methods that rely on confidence scores. 

“A lot of times, these models use phrases like ‘certainly.’ But because they are so confident in their answers, it does not encourage people to verify the correctness of the statements themselves,” Wang adds.

In the future, the researchers plan to continue collaborating with clinicians in the hopes of improving diagnoses and treatment. They are working to expand their study to include data from abdominal CT scans.

In addition, they are interested in studying how receptive radiologists are to calibration-improving suggestions and whether they can mentally adjust their use of certainty phrases effectively.

“Expression of diagnostic certainty is a crucial aspect of the radiology report, as it influences significant management decisions. This study takes a novel approach to analyzing and calibrating how radiologists express diagnostic certainty in chest X-ray reports, offering feedback on term usage and associated outcomes,” says Atul B. Shinagare, associate professor of radiology at Harvard Medical School, who was not involved with this work. “This approach has the potential to improve radiologists’ accuracy and communication, which will help improve patient care.”

The work was funded, in part, by a Takeda Fellowship, the MIT-IBM Watson AI Lab, the MIT CSAIL Wistrom Program, and the MIT Jameel Clinic.

Surprise discovery could lead to improved catalysts for industrial reactions

Upending a long-held supposition, MIT researchers find a common catalyst works by cycling between two different forms.

The process of catalysis — in which a material speeds up a chemical reaction — is crucial to the production of many of the chemicals used in our everyday lives. But even though these catalytic processes are widespread, researchers often lack a clear understanding of exactly how they work.

A new analysis by researchers at MIT has shown that an important industrial synthesis process, the production of vinyl acetate, requires a catalyst to take two different forms, which cycle back and forth from one to the other as the chemical process unfolds.

Previously, it had been thought that only one of the two forms was needed. The new findings are published today in the journal Science, in a paper by MIT graduate students Deiaa Harraz and Kunal Lodaya, Bryan Tang PhD ’23, and MIT professor of chemistry and chemical engineering Yogesh Surendranath.

There are two broad classes of catalysts: homogeneous catalysts, which consist of dissolved molecules, and heterogeneous catalysts, which are solid materials whose surface provides the site for the chemical reaction. “For the longest time,” Surendranath says, “there’s been a general view that you either have catalysis happening on these surfaces, or you have them happening on these soluble molecules.” But the new research shows that in the case of vinyl acetate — an important material that goes into many polymer products such as the rubber in the soles of your shoes — there is an interplay between both classes of catalysis.

“What we discovered,” Surendranath explains, “is that you actually have these solid metal materials converting into molecules, and then converting back into materials, in a cyclic dance.”

He adds: “This work calls into question this paradigm where there’s either one flavor of catalysis or another. Really, there could be an interplay between both of them in certain cases, and that could be really advantageous for having a process that’s selective and efficient.”

The synthesis of vinyl acetate has been a large-scale industrial reaction since the 1960s, and it has been well-researched and refined over the years to improve efficiency. This has happened largely through a trial-and-error approach, without a precise understanding of the underlying mechanisms, the researchers say.

While chemists are often more familiar with homogeneous catalysis mechanisms, and chemical engineers are often more familiar with surface catalysis mechanisms, fewer researchers study both. This is perhaps part of the reason that the full complexity of this reaction was not previously captured. But Harraz says he and his colleagues are working at the interface between disciplines. “We’ve been able to appreciate both sides of this reaction and find that both types of catalysis are critical,” he says.

The reaction that produces vinyl acetate requires something to activate the oxygen molecules that are one of the constituents of the reaction, and something else to activate the other ingredients, acetic acid and ethylene. The researchers found that the form of the catalyst that worked best for one part of the process was not the best for the other. It turns out that the molecular form of the catalyst does the key chemistry with the ethylene and the acetic acid, while it’s the surface that ends up doing the activation of the oxygen.

They found that the underlying process involved in interconverting the two forms of the catalyst is actually corrosion, similar to the process of rusting. “It turns out that in rusting, you actually go through a soluble molecular species somewhere in the sequence,” Surendranath says.

The team borrowed techniques traditionally used in corrosion research to study the process. They used electrochemical tools to study the reaction, even though the overall reaction does not require a supply of electricity. By making potential measurements, the researchers determined that the corrosion of the palladium catalyst material to soluble palladium ions is driven by an electrochemical reaction with the oxygen, converting it to water. Corrosion is “one of the oldest topics in electrochemistry,” says Lodaya, “but applying the science of corrosion to understand catalysis is much newer, and was essential to our findings.”

By correlating measurements of catalyst corrosion with other measurements of the chemical reaction taking place, the researchers proposed that it was the corrosion rate that was limiting the overall reaction. “That’s the choke point that’s controlling the rate of the overall process,” Surendranath says.

The interplay between the two types of catalysis works efficiently and selectively “because it actually uses the synergy of a material surface doing what it’s good at and a molecule doing what it’s good at,” Surendranath says. The finding suggests that, when designing new catalysts, rather than focusing on either solid materials or soluble molecules alone, researchers should think about how the interplay of both may open up new approaches.

“Now, with an improved understanding of what makes this catalyst so effective, you can try to design specific materials or specific interfaces that promote the desired chemistry,” Harraz says. Since this process has been worked on for so long, these findings may not necessarily lead to improvements in this specific process of making vinyl acetate, but it does provide a better understanding of why the materials work as they do, and could lead to improvements in other catalytic processes.

Understanding that “catalysts can transit between molecule and material and back, and the role that electrochemistry plays in those transformations, is a concept that we are really excited to expand on,” Lodaya says.

Harraz adds: “With this new understanding that both types of catalysis could play a role, what other catalytic processes are out there that actually involve both? Maybe those have a lot of room for improvement that could benefit from this understanding.”

This work is “illuminating, something that will be worth teaching at the undergraduate level," says Christophe Coperet, a professor of inorganic chemistry at ETH Zurich, who was not associated with the research. “The work highlights new ways of thinking. ... [It] is notable in the sense that it not only reconciles homogeneous and heterogeneous catalysis, but it describes these complex processes as half reactions, where electron transfers can cycle between distinct entities.”

The research was supported, in part, by the National Science Foundation as a Phase I Center for Chemical Innovation; the Center for Interfacial Ionics; and the Gordon and Betty Moore Foundation.

Engineers develop a way to mass manufacture nanoparticles that deliver cancer drugs directly to tumors

Scaling up nanoparticle production could help scientists test new cancer treatments.

Polymer-coated nanoparticles loaded with therapeutic drugs show significant promise for cancer treatment, including ovarian cancer. These particles can be targeted directly to tumors, where they release their payload while avoiding many of the side effects of traditional chemotherapy.

Over the past decade, MIT Institute Professor Paula Hammond and her students have created a variety of these particles using a technique known as layer-by-layer assembly. They’ve shown that the particles can effectively combat cancer in mouse studies.

To help move these nanoparticles closer to human use, the researchers have now come up with a manufacturing technique that allows them to generate larger quantities of the particles, in a fraction of the time.

“There’s a lot of promise with the nanoparticle systems we’ve been developing, and we’ve been really excited more recently with the successes that we’ve been seeing in animal models for our treatments for ovarian cancer in particular,” says Hammond, who is also MIT’s vice provost for faculty and a member of the Koch Institute for Integrative Cancer Research. “Ultimately, we need to be able to bring this to a scale where a company is able to manufacture these on a large level.”

Hammond and Darrell Irvine, a professor of immunology and microbiology at the Scripps Research Institute, are the senior authors of the new study, which appears today in Advanced Functional Materials. Ivan Pires PhD ’24, now a postdoc at Brigham and Women’s Hospital and a visiting scientist at the Koch Institute, and Ezra Gordon ’24 are the lead authors of paper. Heikyung Suh, an MIT research technician, is also an author.

A streamlined process

More than a decade ago, Hammond’s lab developed a novel technique for building nanoparticles with highly controlled architectures. This approach allows layers with different properties to be laid down on the surface of a nanoparticle by alternately exposing the surface to positively and negatively charged polymers.

Each layer can be embedded with drug molecules or other therapeutics. The layers can also carry targeting molecules that help the particles find and enter cancer cells.

Using the strategy that Hammond’s lab originally developed, one layer is applied at a time, and after each application, the particles go through a centrifugation step to remove any excess polymer. This is time-intensive and would be difficult to scale up to large-scale production, the researchers say.

More recently, a graduate student in Hammond’s lab developed an alternative approach to purifying the particles, known as tangential flow filtration. However, while this streamlined the process, it still was limited by its manufacturing complexity and maximum scale of production.

“Although the use of tangential flow filtration is helpful, it’s still a very small-batch process, and a clinical investigation requires that we would have many doses available for a significant number of patients,” Hammond says.

To create a larger-scale manufacturing method, the researchers used a microfluidic mixing device that allows them to sequentially add new polymer layers as the particles flow through a microchannel within the device. For each layer, the researchers can calculate exactly how much polymer is needed, which eliminates the need to purify the particles after each addition.

“That is really important because separations are the most costly and time-consuming steps in these kinds of systems,” Hammond says.

This strategy eliminates the need for manual polymer mixing, streamlines production, and integrates good manufacturing practice (GMP)-compliant processes. The FDA’s GMP requirements ensure that products meet safety standards and can be manufactured in a consistent fashion, which would be highly challenging and costly using the previous step-wise batch process. The microfluidic device that the researchers used in this study is already used for GMP manufacturing of other types of nanoparticles, including mRNA vaccines.

“With the new approach, there’s much less chance of any sort of operator mistake or mishaps,” Pires says. “This is a process that can be readily implemented in GMP, and that’s really the key step here. We can create an innovation within the layer-by-layer nanoparticles and quickly produce it in a manner that we could go into clinical trials with.”

Scaled-up production

Using this approach, the researchers can generate 15 milligrams of nanoparticles (enough for about 50 doses) in just a few minutes, while the original technique would take close to an hour to create the same amount. This could enable the production of more than enough particles for clinical trials and patient use, the researchers say.

“To scale up with this system, you just keep running the chip, and it is much easier to produce more of your material,” Pires says.

To demonstrate their new production technique, the researchers created nanoparticles coated with a cytokine called interleukin-12 (IL-12). Hammond’s lab has previously shown that IL-12 delivered by layer-by-layer nanoparticles can activate key immune cells and slow ovarian tumor growth in mice.

In this study, the researchers found that IL-12-loaded particles manufactured using the new technique showed similar performance as the original layer-by-layer nanoparticles. And, not only do these nanoparticles bind to cancer tissue, but they show a unique ability to not enter the cancer cells. This allows the nanoparticles to serve as markers on the cancer cells that activate the immune system locally in the tumor. In mouse models of ovarian cancer, this treatment can lead to both tumor growth delay and even cures.

The researchers have filed for a patent on the technology and are now working with MIT’s Deshpande Center for Technological Innovation in hopes of potentially forming a company to commercialize the technology. While they are initially focusing on cancers of the abdominal cavity, such as ovarian cancer, the work could also be applied to other types of cancer, including glioblastoma, the researchers say.

The research was funded by the U.S. National Institutes of Health, the Marble Center for Nanomedicine, the Deshpande Center for Technological Innovation, and the Koch Institute Support (core) Grant from the National Cancer Institute.

Vana is letting users own a piece of the AI models trained on their data

More than 1 million people are contributing their data to Vana’s decentralized network, which started as an MIT class project.

In February 2024, Reddit struck a $60 million deal with Google to let the search giant use data on the platform to train its artificial intelligence models. Notably absent from the discussions were Reddit users, whose data were being sold.

The deal reflected the reality of the modern internet: Big tech companies own virtually all our online data and get to decide what to do with that data. Unsurprisingly, many platforms monetize their data, and the fastest-growing way to accomplish that today is to sell it to AI companies, who are themselves massive tech companies using the data to train ever more powerful models.

The decentralized platform Vana, which started as a class project at MIT, is on a mission to give power back to the users. The company has created a fully user-owned network that allows individuals to upload their data and govern how they are used. AI developers can pitch users on ideas for new models, and if the users agree to contribute their data for training, they get proportional ownership in the models.

The idea is to give everyone a stake in the AI systems that will increasingly shape our society while also unlocking new pools of data to advance the technology.

“This data is needed to create better AI systems,” says Vana co-founder Anna Kazlauskas ’19. “We’ve created a decentralized system to get better data — which sits inside big tech companies today — while still letting users retain ultimate ownership.”

From economics to the blockchain

A lot of high school students have pictures of pop stars or athletes on their bedroom walls. Kazlauskas had a picture of former U.S. Treasury Secretary Janet Yellen.

Kazlauskas came to MIT sure she’d become an economist, but she ended up being one of five students to join the MIT Bitcoin club in 2015, and that experience led her into the world of blockchains and cryptocurrency.

From her dorm room in MacGregor House, she began mining the cryptocurrency Ethereum. She even occasionally scoured campus dumpsters in search of discarded computer chips.

“It got me interested in everything around computer science and networking,” Kazlauskas says. “That involved, from a blockchain perspective, distributed systems and how they can shift economic power to individuals, as well as artificial intelligence and econometrics.”

Kazlauskas met Art Abal, who was then attending Harvard University, in the former Media Lab class Emergent Ventures, and the pair decided to work on new ways to obtain data to train AI systems.

“Our question was: How could you have a large number of people contributing to these AI systems using more of a distributed network?” Kazlauskas recalls.

Kazlauskas and Abal were trying to address the status quo, where most models are trained by scraping public data on the internet. Big tech companies often also buy large datasets from other companies.

The founders’ approach evolved over the years and was informed by Kazlauskas’ experience working at the financial blockchain company Celo after graduation. But Kazlauskas credits her time at MIT with helping her think about these problems, and the instructor for Emergent Ventures, Ramesh Raskar, still helps Vana think about AI research questions today.

“It was great to have an open-ended opportunity to just build, hack, and explore,” Kazlauskas says. “I think that ethos at MIT is really important. It’s just about building things, seeing what works, and continuing to iterate.”

Today Vana takes advantage of a little-known law that allows users of most big tech platforms to export their data directly. Users can upload that information into encrypted digital wallets in Vana and disburse it to train models as they see fit.

AI engineers can suggest ideas for new open-source models, and people can pool their data to help train the model. In the blockchain world, the data pools are called data DAOs, which stands for decentralized autonomous organization. Data can also be used to create personalized AI models and agents.

In Vana, data are used in a way that preserves user privacy because the system doesn’t expose identifiable information. Once the model is created, users maintain ownership so that every time it’s used, they’re rewarded proportionally based on how much their data helped trained it.

“From a developer’s perspective, now you can build these hyper-personalized health applications that take into account exactly what you ate, how you slept, how you exercise,” Kazlauskas says. “Those applications aren’t possible today because of those walled gardens of the big tech companies.”

Crowdsourced, user-owned AI

Last year, a machine-learning engineer proposed using Vana user data to train an AI model that could generate Reddit posts. More than 140,000 Vana users contributed their Reddit data, which contained posts, comments, messages, and more. Users decided on the terms in which the model could be used, and they maintained ownership of the model after it was created.

Vana has enabled similar initiatives with user-contributed data from the social media platform X; sleep data from sources like Oura rings; and more. There are also collaborations that combine data pools to create broader AI applications.

“Let’s say users have Spotify data, Reddit data, and fashion data,” Kazlauskas explains. “Usually, Spotify isn’t going to collaborate with those types of companies, and there’s actually regulation against that. But users can do it if they grant access, so these cross-platform datasets can be used to create really powerful models.”

Vana has over 1 million users and over 20 live data DAOs. More than 300 additional data pools have been proposed by users on Vana’s system, and Kazlauskas says many will go into production this year.

“I think there’s a lot of promise in generalized AI models, personalized medicine, and new consumer applications, because it’s tough to combine all that data or get access to it in the first place,” Kazlauskas says.

The data pools are allowing groups of users to accomplish something even the most powerful tech companies struggle with today.

“Today, big tech companies have built these data moats, so the best datasets aren’t available to anyone,” Kazlauskas says. “It’s a collective action problem, where my data on its own isn’t that valuable, but a data pool with tens of thousands or millions of people is really valuable. Vana allows those pools to be built. It’s a win-win: Users get to benefit from the rise of AI because they own the models. Then you don’t end up in scenario where you don’t have a single company controlling an all-powerful AI model. You get better technology, but everyone benefits.”