MIT Latest News

Subscribe to MIT Latest News feed
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.
Updated: 11 hours 16 min ago

MIT engineers develop a magnetic transistor for more energy-efficient electronics

Wed, 09/23/3035 - 10:32am

Transistors, the building blocks of modern electronics, are typically made of silicon. Because it’s a semiconductor, this material can control the flow of electricity in a circuit. But silicon has fundamental physical limits that restrict how compact and energy-efficient a transistor can be.

MIT researchers have now replaced silicon with a magnetic semiconductor, creating a magnetic transistor that could enable smaller, faster, and more energy-efficient circuits. The material’s magnetism strongly influences its electronic behavior, leading to more efficient control of the flow of electricity. 

The team used a novel magnetic material and an optimization process that reduces the material’s defects, which boosts the transistor’s performance.

The material’s unique magnetic properties also allow for transistors with built-in memory, which would simplify circuit design and unlock new applications for high-performance electronics.

“People have known about magnets for thousands of years, but there are very limited ways to incorporate magnetism into electronics. We have shown a new way to efficiently utilize magnetism that opens up a lot of possibilities for future applications and research,” says Chung-Tao Chou, an MIT graduate student in the departments of Electrical Engineering and Computer Science (EECS) and Physics, and co-lead author of a paper on this advance.

Chou is joined on the paper by co-lead author Eugene Park, a graduate student in the Department of Materials Science and Engineering (DMSE); Julian Klein, a DMSE research scientist; Josep Ingla-Aynes, a postdoc in the MIT Plasma Science and Fusion Center; Jagadeesh S. Moodera, a senior research scientist in the Department of Physics; and senior authors Frances Ross, TDK Professor in DMSE; and Luqiao Liu, an associate professor in EECS, and a member of the Research Laboratory of Electronics; as well as others at the University of Chemistry and Technology in Prague. The paper appears today in Physical Review Letters.

Overcoming the limits

In an electronic device, silicon semiconductor transistors act like tiny light switches that turn a circuit on and off, or amplify weak signals in a communication system. They do this using a small input voltage.

But a fundamental physical limit of silicon semiconductors prevents a transistor from operating below a certain voltage, which hinders its energy efficiency.

To make more efficient electronics, researchers have spent decades working toward magnetic transistors that utilize electron spin to control the flow of electricity. Electron spin is a fundamental property that enables electrons to behave like tiny magnets.

So far, scientists have mostly been limited to using certain magnetic materials. These lack the favorable electronic properties of semiconductors, constraining device performance.

“In this work, we combine magnetism and semiconductor physics to realize useful spintronic devices,” Liu says.

The researchers replace the silicon in the surface layer of a transistor with chromium sulfur bromide, a two-dimensional material that acts as a magnetic semiconductor.

Due to the material’s structure, researchers can switch between two magnetic states very cleanly. This makes it ideal for use in a transistor that smoothly switches between “on” and “off.”

“One of the biggest challenges we faced was finding the right material. We tried many other materials that didn’t work,” Chou says.

They discovered that changing these magnetic states modifies the material’s electronic properties, enabling low-energy operation. And unlike many other 2D materials, chromium sulfur bromide remains stable in air.

To make a transistor, the researchers pattern electrodes onto a silicon substrate, then carefully align and transfer the 2D material on top. They use tape to pick up a tiny piece of material, only a few tens of nanometers thick, and place it onto the substrate.

“A lot of researchers will use solvents or glue to do the transfer, but transistors require a very clean surface. We eliminate all those risks by simplifying this step,” Chou says.

Leveraging magnetism

This lack of contamination enables their device to outperform existing magnetic transistors. Most others can only create a weak magnetic effect, changing the flow of current by a few percent or less. Their new transistor can switch or amplify the electric current by a factor of 10.

They use an external magnetic field to change the magnetic state of the material, switching the transistor using significantly less energy than would usually be required.

The material also allows them to control the magnetic states with electric current. This is important because engineers cannot apply magnetic fields to individual transistors in an electronic device. They need to control each one electrically.

The material’s magnetic properties could also enable transistors with built-in memory, simplifying the design of logic or memory circuits.

A typical memory device has a magnetic cell to store information and a transistor to read it out. Their method can combine both into one magnetic transistor.

“Now, not only are transistors turning on and off, they are also remembering information. And because we can switch the transistor with greater magnitude, the signal is much stronger so we can read out the information faster, and in a much more reliable way,” Liu says.

Building on this demonstration, the researchers plan to further study the use of electrical current to control the device. They are also working to make their method scalable so they can fabricate arrays of transistors.

This research was supported, in part, by the Semiconductor Research Corporation, the U.S. Defense Advanced Research Projects Agency (DARPA), the U.S. National Science Foundation (NSF), the U.S. Department of Energy, the U.S. Army Research Office, and the Czech Ministry of Education, Youth, and Sports. The work was partially carried out at the MIT.nano facilities.

This featherweight, flying robot carries its own sensors and computers

Wed, 11/18/2026 - 2:00pm

Over the past decade, MIT researchers have developed aerial microrobots with bumblebee-like speed, agility, and durability, for tasks like precision agriculture or search-and-rescue. But these featherweight microrobots couldn’t carry the electronics needed for navigation, forcing them to remain connected to a motion-capture system in the lab.

Now, the researchers have designed a flapping-wing microrobot with onboard sensors and computation, which can follow a complex trajectory and land precisely on a sunflower outside the lab’s motion capture environment.

While the robot is still connected to an offboard power source, this demonstration highlights the potential of aerial microrobots to be used for precision agriculture in the future. 

For instance, flying microrobots could weave through a field of almond trees or blueberry bushes to carefully pollinate delicate blossoms in regions where natural pollinator populations have been decimated by habitat loss. They could also hover near plants without causing damage, monitoring crop health and spotting pests before they become widespread.

To enable onboard sensing and computation, the researchers developed a method to expand the artificial muscles that power the robot’s wings, increasing the net payload so the tiny robot could carry the necessary electronics.

They also designed a sensor suite that can rapidly estimate the robot’s position and orientation without relying on an offboard motion capture system, and a miniature electronics package to securely hold the tiny computational components on the nimble vehicle.

“This is a really important milestone for the research community. This is the first demonstration of an aerial microrobot that is sensory and compute autonomous. There are many situations, like pollination or industrial surveillance, where you can’t use a larger robot. Small-scale robots are uniquely positioned for these types of tasks, and if this thread of technology becomes successful in the future, it could be very promising,” says Kevin Chen, an associate professor in the Department of Electrical Engineering and Computer Science (EECS), head of the Soft and Micro Robotics Laboratory within the Research Laboratory of Electronics (RLE), and co-corresponding author of a paper on the robot.

He is joined on the paper by co-lead authors and MIT graduate students Yi-Hsuan (Nemo) Hsiao, Quang Phuc Kieu, and Zhongtao Guan; co-corresponding authors Jonathan P. How, a Ford Professor of Aeronautics and Astronautics and a principal investigator in the Laboratory for Information and Decision Systems (LIDS) and the Aerospace Controls Laboratory (ACL) at MIT; Elizabeth Farrell Helbling, an assistant professor at Cornell University; and others at MIT. The research appears today in Science Robotics.

Increasing the payload

The MIT researchers previously developed a swift, agile aerial microrobot with four wings, each powered by a set of soft actuators, or artificial muscles, that flap the wings rapidly so the bot can perform detailed maneuvers. 

The robot, roughly the same size and shape as a microcassette, weighs less than a paperclip.

To bring this tiny robot closer to real-world applications, the researchers need it to carry the additional weight from electronics for onboard sensing and computation, while still being fast, maneuverable, and robust to collisions.

In this research, they developed a method to scale up the artificial muscles without hampering the robot’s performance. They carefully tuned the radius and length of the soft actuators, increasing their size by 50 percent while boosting the robot’s net lift forces by 46 percent. 

“It is very difficult to scale up the actuators without causing the output power of the system to go down, which reduces the lift forces. We tuned the geometry of the robot very precisely so lift proportionally increases with the additional actuator weight,” Chen explains. 

They were able to achieve a lift-to-weight ratio of 1.9, giving the robot more than enough lift force to carry the onboard electronics package they developed, which weighs only 244 milligrams. 

The tiny electronics package contains the sensors, and microcontroller needed for onboard sensing and computation. 

The researchers incorporated an inertial measurement unit to estimate attitude, an optical flow sensor to estimate lateral flight speed and position, and a time-of-flight sensor to measure height.

But without the precision of the attached motion capture system, the robot does not know its absolute position, causing it to drift off course.

Speedy sensing

To overcome this challenge, the team designed a sensing algorithm to calculate the robot orientation, angular velocity, and position extremely quickly, simplifying the calculations to reduce the computational cost.

They found that estimating position continuously at a faster rate yielded higher performance than a more accurate, slower estimation.

“We spent a lot of time optimizing the selection of the sensors and position estimators so the robot would have very low drift. Our robot can now fly with positional accuracy that is comparable to flying with offboard cameras,” says Hsiao. 

They integrated the sensors and computing components into a tiny electronics package, which they attached to a fiberglass substrate so it wouldn’t be thrown off by vibrations of the robot.

“When you are talking about a computer which is that small, everything must be thought through carefully. There are so many factors to consider; it must do both perception and control, while fitting onto the robot and running fast enough that the vehicle will remain stable during flight,” How says.

They tested the offboard computation capabilities of the aerial microrobot through a series of flight demonstrations, including hovering at different heights and following point-to-point trajectories. The robot maneuvered quickly and precisely, with positional errors of only a few centimeters.

In addition, they demonstrated the robot’s potential for artificial pollination by using a joystick to send high-level position setpoints, guiding the aerial vehicle around several plants before carefully landing on a sunflower. 

“This is quite difficult to do, even if you do have a motion-capture system,” Hsiao says.

The researchers analyzed the vehicle’s power expenditures and found the electronics only consume about 6 percent of the total power, with the rest fueling the robot’s artificial muscles and wings. 

They identified ways to further reduce power consumption from the electronics. This would allow for the addition of more components, like a camera that would enable the robot to recognize objects.

The researchers are now working toward two goals: putting a camera onboard for autonomous object recognition and obstacle avoidance; and integrating power electronics and batteries onto the tiny robot, which would enable untethered flight in an outdoor environment. 

This research was funded by the U.S. National Science Foundation (NSF), U.S. Office of Naval Research, U.S. Air Force Office of Scientific Research, MathWorks, and a Zakhartchenko Fellowship.

Joe Higgins to step down as VP for campus services and stewardship

Wed, 11/18/2026 - 11:20am

Joe Higgins will step down as MIT vice president for campus services and stewardship (VPCSS) at the end of January, Executive Vice President and Treasurer Glen Shor announced today in an email to MIT faculty and staff. Higgins is relocating to the West Coast to be closer to family. 

“Joe is an engineer by trade, and he is adept at building extraordinary things,” Shor says. “He has made an enormous mark on our campus and its future, and we are profoundly grateful for his vision, efforts, and highly engaging and collaborative style.”

Higgins joined MIT in 2016 as director of infrastructure operations in the Department of Facilities. His leadership portfolio quickly expanded to include overseeing campus maintenance, utilities, and facilities engineering and implementing the adoption of technology platforms to better support stewardship decisions, improve operations, and more effectively manage risks and costs.

Higgins assumed the newly created role of VPCSS in 2019. The VPCSS portfolio includes the Atlas Service Center; Department of Facilities; Environment, Health and Safety Office; MIT Welcome Center and Open Space Programming; Office of Campus Planning; and the Office of Sustainability.

Over the course of his seven-year tenure, Higgins has adeptly guided MIT’s physical campus through a transformative period, bringing MIT’s 2030 Capital Plan projects from the drawing board to reality; systematically improving the conditions of critical facilities; and making significant strides to advance campus decarbonization efforts.

“The biggest gift has been getting to work with an amazing team of people”

As Higgins prepares to relocate to the West Coast to be closer to his family, he has been reflecting on his MIT “family.”

“The biggest gift has been getting to work with an amazing team of people who are so dedicated to MIT,” says Higgins. “CSS senior leaders are exceptional, and they’ve built exceptional teams that really are the backbone of our campus.”

Case in point for Higgins is how CSS teams responded to the Covid-19 pandemic. They were instrumental to ensuring essential research and operations continued safely at the height of the pandemic, and they were equally imperative to reimagining new ways of working and operating the campus in a post-Covid world.

“No matter how hard the challenge or big the roadblock, the teams at CSS are here to support the Institute 24/7, 365 days a year,” says Higgins. “I’m going to really miss that shared sense of purpose and community that my colleagues display every day.”

As part of Higgins’ family-driven relocation, he has accepted the role of vice president of facilities management at Chapman University in Orange, California.

Stewarding MIT’s physical campus

Higgins has skillfully advanced MIT’s 2030 Capital Plan, helping to elevate the capabilities of Institute buildings to support cutting-edge academic and research endeavors, improve the residential experience for students, and establish a new level of discipline in keeping MIT facilities in suitable condition.

Signature projects that came to life during his tenure include the construction of the MIT Stephen A. Schwarzman College of Computing, the Graduate Junction residence hall, and the Edward and Joyce Linde Music Building; the recently completed renovation of the Metropolitan Warehouse; and a substantial upgrade to the Central Utilities Plant. And there’s more transformation on the horizon: Higgins and his team have been at the forefront of turning Building 39 into a headquarters and hub for the new MIT Quantum Initiative (QMIT) and a home for a new Quantum Systems Lab (QSL). 

“QMIT and QSL represent the next great chapter in MIT’s historic leadership in quantum science and engineering, and Joe and his team made it possible for us to take this leap,” says Ian A. Waitz, vice president for research. “Throughout every phase of the project, they have been incredible partners, creative problem-solvers, and deeply invested in creating an unprecedented facility to enable MIT to pursue transformative applications of quantum science as well as to catalyze a growing regional ecosystem.”

The capital plan is also an important vehicle for addressing deferred maintenance, an area where the Institute has made considerable progress on Higgins’ watch. Facilities condition is tracked through a measure called the “Facilities Condition Index” — the ratio of deferred maintenance to the total cost of replacing building systems. MIT 2030 Capital Plan investments in renewal have reduced MIT’s overall campus Facilities Condition Index by about 40 percent, making the Institute a peer leader on this front in higher education.  

Making progress on campus climate goals

Higgins has also brought passion, vision, and a collaborative spirit to MIT’s climate goals, helping to put the Institute on a path to dramatically reduce its carbon footprint. Over the past decade, the Institute has decreased energy use per square foot by more than 10 percent, even as the campus has grown and research activity has intensified. Rooftop solar power generation has increased by more than five times in the same period, and 33 MIT building projects have earned Leadership in Energy and Environmental Design (LEED) certification. 

“We completed more than 300 energy-efficiency projects across campus, focusing on our most energy-intensive research buildings, and ultimately touching nearly every corner of MIT,” notes Higgins.

Higgins forged extensive partnerships with leading organizations across Greater Boston and with the City of Cambridge to enable over 450 megawatts of large-scale wind and solar energy projects in regions of the United States heavily reliant on fossil fuels. Together, these steps have put MIT on the verge of achieving net-zero carbon emissions.

“Joe has represented MIT interests and objectives through extremely complex Cambridge processes such as the development of the Building Energy Use Disclosure Ordinance (BEUDO),” says Sarah Gallop, director of MIT’s Office of Government and Community Relations. “His signature collaborative style, combined with his high level of expertise, have helped MIT and the city to achieve greater mutual understanding and to pursue joint initiatives. These include the creation of a virtual power purchase agreement (VPPA) and the implementation of the Greater Cambridge Energy Program (GCEP). Joe is a creative and constructive thinker and really makes things happen.”

Higgins has also laid important groundwork for MIT’s long-term campus decarbonization strategy. This has included his work earlier this year to spearhead MIT becoming an anchor institution in the BosTEN Project, a yearlong study to explore the feasibility of creating what could become the first city-scale thermal network in the United States. The network would help decrease the carbon footprints of major buildings across Boston and Cambridge by harnessing heat from the soil and rock under the Charles River and Boston Harbor, as well as waste heat from buildings and industrial facilities. It would also provide a renewable source of energy that can stabilize and even reduce the costs to heat and cool buildings.

“Joe’s vision and leadership extend beyond MIT’s campus,” says Lindsey Butler, executive director of the Boston Green Ribbon Commission. “He’s been pushing farther so that we can ignite broader strategies and technologies for other institutions and organizations to use. He understands the scale of change we need, and he’s worked overtime to help us go bigger.”

Tapping a seasoned interim leader, launching a national search 

Shor has selected Jon Alvarez, director of campus planning, to oversee the CSS areas while the search for Higgins’ permanent successor is ongoing.

“I am grateful to Jon for accepting the interim role of VPCSS while we conduct this search,” Shor says. “Jon currently leads a team of planners, designers, archivists, and technical professionals. He is a respected, long-serving member of Joe’s leadership team, deeply connected to a range of campus partners, and engaged in several strategic initiatives on which we will continue to make progress during this transition period.”

National executive search firm Isaacson, Miller will lead the search for the Institute’s next VPCSS. Shor is aiming to make an appointment during the upcoming spring semester. 

“I am confident all the CSS areas will continue to serve MIT with dedication and skill while we search for Joe’s successor,” says Shor.

Designing the next chapter for professional athletes

Wed, 11/18/2026 - 3:30am

The 2026 National Football League (NFL) season is well underway, but you won’t find Collin Johnson on the gridiron. After six seasons as a professional football wide receiver, he elected to crisscross the MIT campus as an MBA student at the MIT Sloan School of Management instead.

While in the NFL, Johnson served on the NFL’s Players Association, the union serving players and their families. There, he found his next passion: supporting players’ transition from their playing days to their post-playing lives. Planning his own shift for when he would retire from football, Johnson founded Beyond-Sports, an education platform that develops programs for current and former professional athletes. 

“I realized a lot of players struggle for different reasons after they stop playing football,” says Johnson. “It may be financial reasons, or finding community, or finding a new occupation.”

After surveying NFL players about possible career paths, Johnson was surprised to find that “entrepreneurship” was at the top of the list. Intrigued, he decided that his fledgling company needed an entrepreneurship program. His research led him to MIT DesignX, the accelerator program based in the MIT Morningside Academy for Design, and Svafa Grönfeldt and Norbert Chang, the faculty director and lecturer, respectively. 

At 10 years old, DesignX has grown from its original program on campus to collaborating with universities and incubation centers around the world, including programs in Italy, France, Dubai, Mexico, Hong Kong, Sweden, and Switzerland. DesignX was recently named by Time magazine as one of the best incubators and accelerators in the United States, ranking highest among design-based programs.

“The field of entrepreneurship is not for the faint of heart, and with the rate of change and the ability to compete, the competition is fiercer than ever,” says Grönfeldt. “At DesignX, we look at how to de-risk the innovation and entrepreneurship process. We take a deep dive into understanding the people we aim to serve with what we create. What are their pain points, what are their needs, and what does any solution have to actually fulfill those needs or address those pain points? That is where the tools of design come in.”

Chang, who took the lead on developing the program with Johnson, says he found Johnson to be mission-driven, in large part because Johnson had just left pro football and was experiencing the transition himself.

“He knew firsthand what it feels like to step away from the game and ask yourself, ‘What’s next?’” says Chang. “He also knew how few resources exist to help athletes answer that question in a serious way. That lived experience shaped everything about the program.”

The result is Founders Academy, a venture-building executive education program designed to help current and former NFL players bring their ideas to life. The cornerstone of the program — a four-day workshop at MIT — was held earlier this summer.

Undaunted by hard work

“Building this program together meant taking the curriculum DesignX normally runs over a semester and condensing it into four days, while keeping its rigor,” says Chang. “That was highly ambitious, and Collin still pushed for more. He was the athletes’ biggest believer.”

Johnson says he recognized the similar mindsets between athletes and entrepreneurs. 

“There is a constant feedback loop that players are exposed to nearly every day,” he says. “We practice and then watch film to see how and where we can improve. We get obsessed with our craft and we iterate and do it again the next day and try to improve. We have agency over our craft — much like entrepreneurs.” 

If the metaphor that the rigors of an MIT education is akin to drinking from a fire hose causes some to hesitate, it did not faze the six current and former professional football players who attended the inaugural program. Grönfeldt says she was delighted with the passion and rigor they brought to the program, and she learned a lot about the life of a football player.

“We only see the glamour side of their life, but to hear them talk about the mental work, determination, and the stamina required is impressive,” she says. “DesignX goes through rapid prototyping, and the process goes very fast. They’re used to speed and not being in their comfort zone. They took to the material, sat down at eight o’clock in the morning, and they were still going at it at 10 o’clock at night. They had an impressive work ethic.”

Been there, done that

Because the DesignX curriculum was condensed and originally not designed for solo entrepreneurial efforts, Chang added a mentorship component to the program. Each student was paired with a mentor who had been through the experience and had tangible experience in their industry. 

“That meant someone who could guide them, pressure-test their thinking, and share hard-won lessons from building companies themselves,” says Chang. “Mentors were also able to open doors and make connections to help accelerate their ventures.”

Johnson did not want to develop a feel-good program for the athletes. He wanted them to see what entrepreneurship actually involves, from defining the problem to building the product to learning the economics of fundraising — including the hard, unglamorous parts. Applicants were required to share a business interest and proposed solution or an existing business they planned to develop. Applications were reviewed by Johnson and DesignX faculty to determine readiness and to match mentors — sourced heavily through Grönfeldt’s and Chang’s networks — with each venture. 

“I found the athletes’ business ideas to be directly related to the curriculum that was taught,” says Johnson. “That was the game changer for them, because they could talk directly with Norbert and Svafa about their ideas. The curriculum really unlocked everything for them.”

To launch the academy shortly before beginning his own journey at MIT Sloan was a huge thrill for Johnson. Making it more rewarding were the rave reviews from the participants. That feedback has given Johnson a path to refining next year’s program with Grönfeldt and Chang. He has already procured Founders Academy as an approved program by the NFL Players Association, which allows future applicants to attend the program with little to no out-of-pocket cost. Participants earn a DesignX certificate when they complete the program.

“Ultimately, I’d like to integrate the mentorship model into MIT’s ecosystem,” says Johnson. “I can see a time when MIT students are serving as mentors to the athletes. It could be a mutually beneficial collaboration. When you think about an MIT doctoral student and an NFL player, at face value, you don’t think they have anything in common. But when they sit down and bond over a problem — we found they have a lot in common.”

Helping AI text detectors keep up with changing conditions

Wed, 11/18/2026 - 12:00am

A teacher who believes a student’s term paper may have been written by a large language model (LLM) can use a tool for detecting artificial intelligence-generated text to help confirm their suspicions. But the teacher may want to think twice before putting all their trust in the detector. 

Researchers from MIT and Cornell University showed that such detection tools, which are AI models themselves, can be fooled by AI-generated text disguised to appear as if it were written by a human. Detectors can also fail to identify AI text due to natural shifts in writing habits over time or the emergence of new LLMs.

In a new paper, the researchers demonstrated how a special training technique makes an AI detector more accurate and resilient to deception, even as human and AI writing constantly changes.

They employed a method called test-time adaptation, which allows a deployed detection model to learn from new, unlabeled data by adjusting its inner workings on the fly. In this way, the model adapts to unseen inputs without being retrained from scratch.

In experiments, a commercial detector could only identify about 27 percent of AI-generated text that had been disguised, while the new approach achieved 93 percent accuracy.

This method could help educators, editors, publishers, and hiring managers flag LLM-generated writing that people try to pass off as human-written. It could also be used by cybersecurity experts to prevent email phishing scams or identify search engine optimization fraud.

“There are many high-stakes contexts where people really care about the identity of the author, from scientific publishing to short story competitions. We’ve shown that test-time adaptation allows your AI detector to be much more robust, even with continual adversarial behavior by AI users,” says study co-author Manish Raghavan, the Drew Houston (2005) Career Development Professor at the MIT Sloan School of Management, who also holds an MIT Schwarzman College of Computing shared position in the Department of Electrical Engineering and Computer Science (EECS), and is a principal investigator in the Laboratory for Information and Decision Systems (LIDS).

He is joined on the paper by lead author Kevin Ren, graduate student at Cornell Tech, and senior author Nikhil Garg, an assistant professor at Cornell Tech. The research will be presented at the Conference on Neural Information Processing Systems.

Duping the detectors

The proliferation of LLMs that generate human-like text has led to a plethora of commercial AI-detection tools to help people identify whether a piece of writing was authored by a human.

But these detection tools can fail when the text they evaluate significantly differs from the text they were taught to identify. This is known as a distribution shift.

Detector developers typically train their models by gathering text from the internet that they know to be either AI- or human-generated. They train the model to tell the difference. 

However, once the detector is released, people might change the way they write or the way they use AI to generate text. 

For instance, a human in 2030 may write about different topics using new analogies that weren’t common for someone writing in 2010. A detector trained with 2010 data might be unfamiliar with this new writing style and incorrectly mark human-written text as AI-generated.

The release of a detection model also creates a cat-and-mouse game between its developers and users who want to bypass it. These individuals can run AI-generated text through an automated “humanizer” software tool that modifies it to appear human-written to the detector.

“After the company releases a model, a bunch of ‘humanizers’ pop up to try to beat it, so the company tries to incorporate this humanized text into their training so the detector can identify it, and so on. It’s like rinse-and-repeat. The cycle can go on forever,” says Ren.

The researchers explored whether test-time adaptation (TTA) can make an AI detector resilient to these distribution shifts.

Adaptation in action

In TTA, instead of freezing the model so it can’t change during deployment, researchers allow that model to adapt itself to incoming data after it has been released. For an AI text detector, these data are new pieces of text uploaded by users. 

These data are unlabeled, so the detector developers don’t know which data were written by humans versus AI models. But these new data represent contemporary human and AI writing — exactly what the developers want the model to be able to predict. 

With enough examples, the detector learns to identify patterns to determine which data are AI-generated and which are human-generated.

The researchers studied TTA by creating datasets to contain a certain amount of AI-generated and human-generated text, but they don’t provide those labels to the model. They also used LLMs to generate some examples that resemble text written by AI to specifically evade a detector. 

Then they performed the TTA training process and tested the model on ground-truth data.

In experiments designed to mimic the natural shift in human language over time, a detector that used TTA was 85 percent accurate, while traditional methods only achieved 73 percent accuracy.

And when it came to writing designed to thwart detection tools, existing methods could only identify about 27 percent, while TTA enabled the model to catch more than 93 percent of disguised text.

“TTA has the potential to really help deployed AI-detection models in terms of being able to proactively address this iterative ‘gaming’ behavior by adversarial AI users,” Ren says.

Adapting to shifts in human writing behavior is a more difficult TTA problem, since the shifts are slower and occur in more subtle ways, he adds.

The researchers caution that TTA may not always be an effective solution as future LLMs produce more natural-sounding language and it becomes harder to find verifiable human text to use for training.

“It isn’t entirely clear if this will be a solvable problem,” Raghavan says. “There are some big, open questions about how you should think about science, writing, and art in a world where it is infeasible to tell the difference between human-generated content and AI content.”

In the future, the researchers want to explore similar strategic domains where TTA could be useful and build better benchmarking tools to help engineers improve these AI detection models in a dynamic fashion.

This project used API credits provided by Pangram to use their AI text detectors. 

Building blocks for a greener skyline

Wed, 11/18/2026 - 12:00am

The construction industry is often cited as contributing over one-third of global greenhouse gas emissions, in large measure because buildings use more material than they need to. Miana Smith ’21, SM ’23, PhD ’26, a recent graduate and visiting researcher who is part of the Center for Bits and Atoms research team in MIT's Program in Media Arts and Sciences, has spent years trying to fix that by reimagining how buildings get put together.

Engineers already know how to design more structurally efficient buildings. For example, in buildings framed with structural steel, it's estimated that 30-40 percent of the steel could be removed without weakening it, simply by customizing the cross sections of the beams. 

Although this solution seems straightforward, implementing it is not. “The process of producing a custom part requires not just one-off design and engineering work, but also often impractically high manufacturing costs,” Smith notes. Once produced, all the custom parts require highly skilled laborers to install them on site. As a result, builders default to using more material than they need, because oversized, uniform parts are simply cheaper to produce and assemble at scale.

Another option is digital fabrication: automated methods like 3D printing and robotic assembly. While this offers adaptability — a way to decouple design complexity from the overall cost of production — the trade-off is precision. Extruding concrete or placing framing members across a 10-meter span for a building often demands a machine accurate to a few millimeters, and machines that precise are both difficult to manufacture and expensive.

Lego logic

Smith's research inverts that logic. Instead of a precise machine and a simple material, she builds a complex material, a small lattice block Smith’s lab calls a voxel, that even a basic, imprecise robot can assemble correctly. 

She compares the concept to the design principle behind Lego bricks. “You can take a very imprecise assembler — in this case, a child — and have them build something that's very precise,” Smith says, “not because a child is somehow uniquely able to build out of Lego blocks and nothing else, but because the blocks themselves have a clever mechanical design that prevents error accumulation as the structure grows.”

Smith’s voxels might look vastly different, but like Legos, they effortlessly snap into place when placed on top of one another. Her robots' margin of error is about 10 millimeters, far greater than the roughly 100-micron tolerances of most industrial robots — a number that would look like a failure to most engineers. 

It isn't one, Smith says, because her blocks are built to absorb it. And because the robots don't need industrial-grade precision, they're also comparatively cheaper — a few hundred dollars apiece instead of tens of thousands — which means many of them can work in parallel instead of one doing everything slowly.

Motorcycles and microbiology

Smith's interest in building things goes back to high school, when she rebuilt motorcycle engines through her school's STEM program and tinkered with electrical engineering work at a plasma physics lab at the University of California at Los Angeles. 

As a mechanical engineering undergraduate at MIT, a research position with professor of architecture Larry Sass introduced her to digital fabrication for buildings. Her perspective broadened further in the class 4.140 (How to Make [almost] Anything), taught by Neil Gershenfeld, professor of media arts and sciences, director of the Center for Bits and Atoms, and Smith's PhD advisor. She was captivated by the idea that an entire world, buildings included, could be built out of an engineering equivalent of Lego bricks. 

Before turning to robotics, Smith spent two years in a different Media Lab group working with living systems, engineering bacteria into woven textile fibers, and sending honeybees into microgravity to see whether they'd survive (they did). 

“The thing with living systems is, essentially, they’re also distributed assembly systems,” she says — the same idea, in her telling, that eventually inspired her to switch from using small bugs to using robots. 

The vision behind the voxel

The underlying lattice geometry that the voxel is founded on is not new. Adapting it to build structures using robots, however, took time. There did not yet exist a way to automate the voxel’s production and assembly so it could perform at scale. Developing this automation has been Smith’s primary research focus.

For a long stretch, her team built with small, individual units, each requiring as many as 24 connectors. Each of those joints provided a fresh opportunity for structural weakness and assembly failure, preventing the team from being able to robotically build larger structures. The fix Smith contributed, once she found it, felt to her almost embarrassingly obvious: Make the blocks bigger, and pre-attach as many of the internal joints as possible before a robot ever touches them. “It's introducing an error-correction feature that is structurally active” she says. “That's the main thing that enables the robots to build something big that's still load-bearing.” 

The lattice design itself mattered, too. Her team ultimately moved to a more heavily internally connected structure, because “in terms of a weight-to-stiffness or weight-to-strength ratio, a stretch-dominated lattice is going to outperform a bending-dominated one,” she says. 

By Smith's estimate, buildings made this way could cut embodied carbon emissions by roughly 80 percent compared to standard concrete construction. And because the system isn't tied to one material, it doesn't have to compromise on architectural ambition. Her team has built voxels out of plywood, bio-based plastics, aluminum, and steel, all using the same underlying joint design. 

The next era of builders 

Smith’s work is now being tested in the field. Her team is collaborating with Druk Holding and Investments, a major Bhutan company focused on creating a sustainable economy and future, on their next housing development project: Gelephu Mindfulness City. Bhutan officials wanted construction methods built from local materials that could be reconfigured over time — a natural fit, in Smith's view, for the voxel work already underway at MIT. Using plywood blocks, her team has built room-scale, load-bearing demonstrations, including a small house, with plans to test the system outdoors at a similar scale.

Her next target is concrete itself, using the same reconfigurable block logic as formwork, to shape concrete more efficiently. Making complex molds for concrete is very expensive, but with the advents of their flexible, reconfigurable building approach, she expects to be able to do reduce that cost significantly. “I want to feel like what I'm working on has an impact,” she says. “I think everyone in our generation, at least, feels that the climate crisis is the impending doom.”

To that end, in addition to her ongoing work at MIT, Smith is pursuing a startup focused on more sustainable concrete construction through an Activate Fellowship, a prestigious science entrepreneurship fellowship that helps recipients commercialize their research. “I did not originally think of her as an entrepreneur, remarks Gershenfeld, “but she is the best kind, driven by a vision of change she wants to create, rather than by a business plan.”

Back in the lab, the research will continue, even after Smith moves on. A new cohort of students, she says, is already looking toward a broader range of robotic behaviors; various robot shapes, working together, building bigger. It's a fitting place to leave the project: a set of greener building blocks, designed to be picked up and reconfigured by whoever comes next — much like the Lego bricks that started it all.

Putting a spotlight on AI, society, and humanity

Tue, 11/17/2026 - 3:40pm

MIT students are going to office hours less often. They’re spending less time working through problem sets, and with one another. Those were among the findings of MIT’s recent report on artificial intelligence and education, and they are one reason a growing number of MIT faculty are concerned about how generative AI is both changing how people work and affecting how they think, learn, and relate to one another.

For the MIT Human Insight Collaborative (MITHIC), those questions have become a central focus, and MITHIC is working with collaborators across the Institute, including the Social and Ethical Responsibilities of Computing (SERC), to help MIT address them.

Launched in 2024 as a strategic initiative, MITHIC brings together scholars, artists, educators, and researchers to explore the human dimensions of the defining challenges of our time. Since its launch, MITHIC has supported 67 projects across 31 MIT units, offering a total of $5.5 million in funding.

Roughly a third of those projects are focused on AI, society, and humanity — a topic that MITHIC faculty co-lead Heather Paxson says has become an increasingly important part of the initiative.

“When we refer to AI and society, we’re not only speaking about individual ethical decision-making or how to retrain workers. We’re looking at how AI is reshaping sociality, how we relate to one another,” says Paxson, a professor of anthropology and the associate dean of the School of Humanities, Arts, and Social Sciences (SHASS). 

Eroding “creative friction” 

Building upon this focus, an AI and Society Forum took place on campus earlier this year. Co-organized by MITHIC, SHASS, SERC, and the MIT Generative AI Impact Consortium (MGAIC), the event featured presentations and panels concentrating on jobs, democracy, and AI. 

On campus, the report on AI and education warned that AI is eroding the “creative friction” that has long defined an MIT education: students debating ideas and challenging one another, their professors, and their teaching assistants, in class and out of it.

MITHIC faculty co-lead Seth Mnookin, a science writing professor and head of Comparative Media Studies Program/Writing (CMS/W), said, “The report wasn’t saying that AI is here and we all need to adapt. It was saying that we need to be asking more fundamental questions.”  

“When the United States entered World War II in 1941, the Institute’s newly minted scientists and engineers helped shape that era through science and technology,” Mnookin said. “We’re at a similarly consequential moment with AI. The difference is that this time, the questions that matter most can’t be answered by technologists alone. That’s where the humanities, arts, and social sciences come in.”

Crossing disciplines in the pursuit of answers

MITHIC is supporting a range of interdisciplinary projects through its SHASS+ Connectivity Fund, in which a SHASS scholar partners with a researcher in another part of MIT. 

In the context of the 2026 U.S. midterm elections, political science professors Adam Berinsky and Charles Stewart III and MIT Sloan School of Management Assistant Professor Chara Podimata are studying how large language models (LLMs) have become intermediaries of democratic knowledge. 

The ongoing project, which has demonstrated how voters in Alaska’s midterms who conducted their pre-election research using Anthropic’s AI model, Claude, may have received incomplete information about candidates and their views, was recently featured in The New York Times. 

“AI is becoming part of the way people encounter and make sense of political information, and yet we know relatively little about what that information environment actually looks like, how it differs for different user demographics, political identities, and geographies,” Podimata told the Times.

In another project, CMS/W Professor Eric Klopfer and brain and cognitive sciences Professor Pawan Sinha are co-leading research that’s measuring AI’s near- and long-term effects on cognition.

Meeting at the intersection of AI, education, and neuroscience, the professors are examining whether generative AI is increasing the human brain’s consumption of information but offloading cognitive function, or supporting an increase in cognitive function when used as part of the creative process. They recently gathered experts on the topic for an event on the MIT campus, “AI and the Brain.”

In the field of arts and technology, Anna Huang, a professor of music technology holding an MIT Schwarzman College of Computing shared position with the Department of Electrical Engineering and Computer Science, Joe Paradiso, a professor at the MIT Media Lab, are examining how incorporating AI impacts the creative process and live performance experience.

Their project, “Live Musician-AI Improvisation, Tuning the Social Aesthetics in Human-AI Partnerships” explores the question: what happens when human composers and contemporary machine learning systems improvise to make music together? 

As part of their project, a collective of MIT musicians, ethnomusicologists, and engineers meets for regular sessions to experiment with improvisation frameworks like stopping to discuss how technology impacts authorship and musical direction, ensemble dynamics, and other cornerstones of improvisation. Faculty will ultimately incorporate their takeaways into a future course or practicum on AI-improvisation.

Teaching and scholarship in an AI era

SHASS faculty and academic staff are also leading new MITHIC initiatives to help broaden instruction and community dialogue about the intersection of AI and society. 

Graham Jones, professor of anthropology, and Arvind Satyanarayan, associate professor of computer science, used MITHIC’s SHASS+ Connectivity funding to help launch a course, 6.S061/21A.S02 (Humane User Experience Design), in which students examine how AI systems are designed and grapple with the consequences of AI agents entering into socio-cultural relationships with people through their ability to mimic human thought and speech.

“The crux of our collaboration is the term 'humane.' It’s a bit of a double-entendre, meaning both human-serving or benevolent, and humanlike,” Jones said at the inaugural MITHIC Annual Event. “What’s interesting is that our students are discovering that there’s a fundamental tension between building a system that’s good for people and one that seems human, because real people can develop dangerous dependencies.”

Another project, “Rebuilding Computing and Society at MIT,” is supported by MITHIC’s Education Impact Fund (EIF). Led by Dwai Banerjee, associate professor of science, technology, and society, the initiative revitalized the Computing and Society concentration launched in 2019. The HASS concentration had gone dormant, listing classes no longer being offered and with minimal student interest. The concentration now features current classes across 11 SHASS units in which students can explore computing through different humanistic and social science angles.

Banerjee, who serves as the concentration’s advisor, is also working with lecturer and research affiliate Mariel Garcia-Montes PhD ’26 to find ways to connect the concentration instructors and campus initiatives on computing and society. MIT senior Daniel Gonzalez is also part of the project, developing a research and outreach program that centers undergraduate voices in computing and society efforts across the Institute.

Another MITHIC project funded by the EIF, “Review of Generative AI Assistive and Resistant Communication Assignments and Supporting Workshops,” is led by Michael Trice, a lecturer in MIT’s Writing, Rhetoric, and Professional Communication (WRAP) program.

The project is responding to a consistent SHASS faculty request for additional training and skills needed “to fully address generative AI concerns in their classes.” 

Trice is creating a bank of faculty resources and facilitating workshops and visiting lectures on generative AI and writing. These include a talk given earlier this fall by John Gallagher from the University of Illinois, who presented findings from a large-scale survey on AI-usage and writing conducted with undergraduates at his institution in 2026.

Meeting the moment

MIT leaders agree that MITHIC, now in its third year, is playing a central role in the Institute meeting the moment, as AI presents both extraordinary potential and disturbing risks. 

“There has never been a more important time for constructive collaboration between science and technology and the human-centered disciplines,” MIT President Sally Kornbluth recently said. “Through MITHIC, MIT is giving faculty the support to reach out across disciplines and pursue their dreams.”

Agustín Rayo, dean of SHASS and co-chair of MITHIC, agrees. “MIT is uniquely placed to help us understand this extraordinary moment and to generate the values, tools, and policies that we will need to shape AI so that it becomes a force for good,” he said. “It is essential that we deploy AI technology in a way that elevates human existence, rather than diminishing it.”

At the recent SHASS 75th Anniversary Symposium, Provost Anantha Chandrakasan said given the rapid development of AI, an understanding of the ethical implications and societal impact of implementing new technologies is especially urgent.

“SHASS provides an essential perspective on questions concerning how technology can be developed thoughtfully and deployed responsibly, with the aim of improving human well-being,” said Chandrakasan, who also serves as MITHIC co-chair.

Looking ahead, MITHIC and SERC are developing a series of events that pair humanists or social scientists with colleagues in a single field, starting with mathematics, to examine what AI is changing about how that field produces knowledge and trains its next generation.

“The question isn’t whether AI will change these fields,” Mnookin said. “It’s whether we understand what’s at stake before it does.”

Q&A: Why building math muscles remains important in the AI era

Tue, 11/17/2026 - 12:00am

John Urschel PhD ’21, the Class of 1956 Career Development Assistant Professor in the MIT Department of Mathematics, is no stranger to building muscle: Before earning his PhD at MIT, Urschel strengthened his offensive muscles as a lineman for the Baltimore Ravens. He remains committed to helping his students exercise, but now his focus is on critical thinking skills. In the classroom, Urschel shares with his students his lifelong love for solving puzzles, while honing their abilities to be strong qualitive thinkers — skills he argues will be essential in an artificial intelligence-enabled world. 

Urschel recently published two preprints on arXiv, one on the growth factor of random matrices, the other on worst-case behavior in Gaussian elimination, in which he used a combination of human insight and AI tools to solve problems that had stumped mathematicians for decades. That experience provided Urschel with new insights into the future of mathematics and higher education, and the enduring importance of the in-person, residential education offered on college campuses, where students can experience the importance of struggle, problem solving, and human connection. Urschel recently chatted with MIT News about education, AI, and mathematics; excerpts from the conversation follow. 

Q: What is it about studying math that you hope to share with your students?

A: When I was very little, I loved solving puzzles, whether that’s Sudoku puzzles or Martin Gardner math puzzles or crossword puzzles. I think there’s something really enjoyable about being faced with some problem that you don’t quite understand, trying to get your bearings, trying to make your way through it, and finally gleaning some sort of understanding from it. That’s a process that I’ve enjoyed for as long as I can remember, and that’s the thing that I’ve really enjoyed most about math.

When it comes to teaching, first, I want to help our students train themselves and hone their ability to solve problems and their ability to be strong quantitative thinkers.

The second part, which is more specific to whichever course I’m teaching, is to help them assimilate the big ideas or tools or techniques or themes that help them become better problem solvers in that specific discipline. If they’re taking a course on a specific subject, the idea is for me to try to convey important fundamentals so that they can use that learned expertise to become better problem solvers out in the world.

Q: You recently used AI to assist in solving two proofs. Can you talk a bit about the experience?

A: These are the first two papers where I’ve ever touched the machine, and it does feel different. It’s not unlike having a collaborator, but it’s a collaborator who has access to unlimited resources, seemingly, and does certain things much, much faster than I ever could. 

In the same way that computers can do calculations much faster than any of us can, I think at this point it’s very clear that artificial intelligence systems can do certain types of mathematics much, much more efficiently than any of us can. 

Q: What was it like using AI to solve math problems? Was it just as rewarding?

A: For me, personally, I would say that the feeling that I talked about of problem solving, of thinking your way through a problem, I find that that’s most joyful for me when I do that by myself, without AI. 

One thing that I think is really important is to have an independent sense of problem-solving ability, a sense of curiosity — this idea of building those mathematics muscles, which I think are important, even in the era of AI.

Especially when you are talking with a machine, asking the machine for things, I think it’s very important to elevate the idea of understanding. If you’re talking to a machine, let’s say, and it gives you a proof of some mathematical fact, I think the true value in that proof comes from a human understanding and assimilating the way in which the machine combines certain mathematical ideas, and then internalizing that and communicating that knowledge to others.

Understanding is the big thing that I stress to my students who use AI systems. And then for the students who do not use AI systems, I focus on building their problem-solving abilities, which is what we’ve done for centuries in mathematics. 

I think in the age of modern AI machines, as long as there is a role for humans to think and try to solve problems, which I believe there is, then building these skills remains important.

Q: The recent report from MIT’s Ad Hoc Committee on AI Use in Teaching, Learning and Research Training stressed the importance of experiencing the necessary and productive struggle of learning, along with the collaborative aspect of learning. I’m curious what you think about the role of AI in math education going forward. 

A: On one hand, the development of AI technologies should be viewed as a giant win for education, a giant win for leveling playing fields when you have technology that’s easily accessible, that gives you access to vast amounts of knowledge. 

Neither of my parents are professors; I remember when I was a kid, I didn’t know what a mathematician was until I got to college. And the way I learned about advanced mathematical concepts in high school, but also when I got to college, was with a laptop, and with Google. I was able to Google some topic I want to learn about, and it was amazing that I could find papers written by some of the world’s leading experts, just freely available online at my fingertips to learn from. That was really important for me and so helpful for my education.

But I think we’re at a point where modern technology is a bit of a double-edged sword. Because on one hand, with AI systems you can learn something faster than ever, and the resources I had when I was a kid would seem like small potatoes compared to the sorts of things you can do to help yourself learn nowadays. 

But the issue is that this also can take away your ability to think and learn for yourself. I think it’s hard if you’re a high school student or you’re a college student. You have a lot going on, you’re tired, and thanks to AI, you can just finish your homework with the push of a button. You can write your 10-page essay with the push of a button. Also, you can get the false impression that you more or less understand what’s going on when you really don’t. I think that’s one of the big dangers when you don’t try to learn things for yourself. 

People don’t like the word “struggle.” But when you don’t struggle with a certain material on your own, I think you can get the impression of learning without actually learning. Connecting with the material and really pushing yourself, that’s quite an important skill. 

Certainly, there’s work to do from the faculty and instructor side: to make sure that our recitations are maximally useful; to make sure that our office hours are serving students in the best way possible; to encourage the students and try to remind them that they have some responsibility for their education, and that the best way is to come and show up and interact with the material. But we should also recognize our responsibility to the students to deliver them the best learning experience possible and to give them an educational system that challenges them but also is a quality product they want to come and take part in.

Q: Why would you still encourage students to pursue studies in mathematics given the advances AI technologies have made? Do you still see the future of mathematics as exciting?

A: Yes, for a multitude of reasons. The act of learning how to think is going to be more important than ever. This idea of being an expert, of knowing all these things, which are easily available and at your fingertips, is perhaps a little less important when you have a machine that has all these resources. I think something that may become even more important is the ability to think for yourself. Because with the existence of these machines, I think a natural fear is that that skill set is going to lessen in the general population as generations move on and on, and so I think that ability will be at a premium. I think training in mathematics and quantitative sciences more generally is quite important for that. 

The second thing is community. I think learning, not by yourself, but with a group of peers, greatly improves the overall experience. It’s an integral part of the experience. I would say those two things are the reason why I think it’s still important to study math, to study quantitative sciences, and also why it’s still important to come and go to class. 

The world is changing fairly rapidly. I think we as an institution have a responsibility to prepare students for that. We’re having lots of conversations about this right now because the ability to think, reason, and solve problems are core skills that our students, I think, will need now more than ever. I also think we need to make sure that we’re preparing our students for the world that they’re about to enter into. The exact way that we teach our students, or prepare our students for the world, absolutely cannot look like what it was 10 years ago. Because this world simply looks quite different. 

Learning the language of machine learning

Tue, 11/17/2026 - 12:00am

When deploying deep learning models in high-stakes scenarios, researchers would like to understand how they work. Today, this requires unraveling an enormously complex network of internal processing nodes to understand what the model knows and how it will likely behave in practice.

Jacob Andreas would rather have the models explain themselves. 

The newly tenured associate professor develops techniques that enable artificial intelligence models to communicate accurately about what they learned from their training data, in language humans can understand.

“Rather than needing to use all these external tools to try and understand what is going on under the hood, maybe we can train models to talk about themselves in a way that is accurate,” says Andreas, an associate professor in the Department of Electrical Engineering and Computer Science (EECS) and a member of the Computer Science and Artificial Intelligence Laboratory (CSAIL).

By studying how machines come to understand language, Andreas and his group work toward building intelligent systems that can learn about the world from human guidance and communicate effectively with human collaborators. They develop methods to make the training process more natural and efficient, so AI models learn in a way that is more like human learning.

Ultimately, he and his group want to help people design more accurate and robust deep-learning models that are better suited for their intended application. 

“We used to debug systems just by looking at their parameters and figuring out whether they make sense or not. Now, we need ‘mind-reading tools’ that can take this really complicated neural computation and translate it for us because it is just not in a language that we, as humans, know how to speak,” he says.

A love of language

Andreas’ fascination with language blossomed at a young age. As a child growing up in the San Francisco Bay Area, he was a voracious reader of any book he could get his hands on. 

One such book, passed along to him by a family friend, contained a pivotal scene where the main character hacked into a computer system to save the day.

That piqued his interest in computers. A few years later, his father brought home a simple BASIC Engine that Andreas could run on their home computer, and he started immersing himself in the world of computer programming.

Andreas chose to major in computer science as an undergraduate at Columbia University but wasn’t sure if he wanted to pursue a career as a researcher, as a software engineer, or at a startup.

His friends signed up for research projects in faculty labs, so Andreas figured he would try that, too.

“I just blast emailed everyone in the computer science department who was working on anything vaguely AI-related, and the natural language processing faculty got back to me first. That is how I wound up in the field where I am now,” he says.

Working in the lab of Michael Collins, the Vikram S. Pandit Professor of Computer Science at Columbia University, Andreas completed an undergraduate thesis project on machine translation. His project involved adapting existing statistical techniques for representing the meaning of language into more powerful AI systems that could translate passages of text.

“I always felt like these deeper questions about how thought works or how to model the world were closer to the surface in language processing problems than in the other tasks that AI people were working on at the time,” he recalls.

A seismic shift

He wanted to keep exploring deep questions, so Andreas enrolled in graduate school at the University of California at Berkeley to pursue a career as an academic researcher. 

Before starting his PhD studies, he spent a year earning a master’s degree at Cambridge University in England. In that historic environment, he was inspired to contemplate the nature of language learning as he worked on computational translation problems.

Those studies fueled his research when he joined the lab of Berkeley professor of computer science Dan Klein. 

Andreas began his studies at the time when neural networks were starting to take off. The field of natural language processing was making a seismic shift from statistical methods to this new world of deep learning. 

Statistical methods utilize transparent, mathematical formulas to analyze text, while deep learning techniques use opaque algorithms to automatically learn complex meanings from massive amounts of text data.

“It was fun to be there as we made this transition to the modern toolkit, and learning and inventing all of that stuff when it was still new,” Andreas says.

He began collaborating with scientists in computer vision labs to tackle increasingly complex questions about language learning. Together, they designed AI systems that excelled at generating natural language by teaching those systems to “see” and understand the world around them. 

“This was all super-exciting because there really hadn’t been a common set of tools for working with both language and vision data. Then, all of a sudden, neural networks gave us this common language,” he recalls.

He completed a thesis project on the use of neural module networks, which are small fragments of neural networks that perform specific tasks, like localizing objects or logical reasoning, in a larger AI system. But it was the more speculative projects he tackled alongside his thesis, like interpretability and creating natural language descriptions of an AI model’s internal states, that would drive the focus of his group at MIT.

Faculty focus

After spending a year at the startup Semantic Machines (which spun out of his lab at Berkeley), Andreas joined the MIT faculty in 2019.

“It felt like this was a fantastic opportunity to be in the homeland of AI research and collaborate with all these amazing, established people, but also do something new and figure out what the modern era in machine learning was going to look like,” he says.

His group focuses on two main areas. First, they are studying the science of large language models to understand the algorithms these models deploy as they learn and process information. 

The tools they develop help scientists and researchers better predict how a trained model will behave in the real world. For instance, one family of techniques they developed trains large language models to explain what is going on inside other neural networks using language a human could understand. 

These techniques can assign a natural language description to every information-processing node in a massive neural network. For example, one such description might represent an individual node that identifies “patches of grass near animals” in input images. 

“That work serves as a jumping off point for a bunch of higher-level questions we ask about how these models represent meaning,” he says.

On the other side, his team designs new algorithms for training machine-learning models, and for predicting how specific datasets or examples will affect different aspects of model behavior. 

“We don’t have a good scientific theory of what is going to happen when you train a model on a particular dataset. This is a big challenge for the field right now,” he says.

A changing environment

The speed of AI advancement presents one of the biggest challenges his lab faces. The field seems to completely change every three to five years, Andreas says, forcing him to stay nimble and ready to learn something new.

That same dilemma carries over into the courses 6.461 and 6.861 (undergraduate and graduate Natural Language Processing), which Andreas co-teaches. He and his fellow instructors must redesign the course each year to keep up with the newest developments.

It’s also challenging to teach the technical pieces of machine learning in an era where every assignment could be completed by a large language model rather than a human pupil.

“I try to impress upon the students that having a really deep foundation in how all this technology works under the hood is the only way they are going to understand it, and will differentiate them in a world where everyone has access to a coding agent,” he says.

With so much change, it is difficult to predict where the field is headed next, Andreas says. He foresees that his group’s work on interpretability will become even more important for making advanced systems like AI agents safe enough to deploy in high-stakes situations.

This will be especially true as researchers design deep-learning models that can aid in scientific discovery. These models will be trained on troves of raw data from research labs that may not be well-understood by scientists yet. 

“Figuring out how to enable these models to verbalize all of this hidden knowledge they have about things people have never written about explicitly will be a huge challenge and opportunity for people who work in machine learning,” he says.

In addition to tackling challenges in the lab, Andreas enjoys pushing himself as a hiker and backpacker. He has loved spending time in the outdoors since his childhood in California and goes on at least one or two backpacking trips each summer.

He and a colleague in the Department of Mathematics are working their way across all the 4,000-foot peaks in the White Mountains of New Hampshire, making slow but steady progress.

And now that he has earned tenure, Andreas is looking forward to getting back into an old hobby — singing in a chamber choir. He’s always enjoyed music and became enamored with chapel choirs and bell ringing during his time at Cambridge University.

MIT School of Engineering faculty and staff receive awards in summer 2026

Mon, 11/16/2026 - 3:45pm

Each year, faculty and researchers across the MIT School of Engineering are recognized with prestigious awards for their contributions to research, technology, society, and education. To celebrate these achievements, the school periodically highlights select honors received by members of its departments, institutes, labs, and centers. The following individuals were recognized in summer 2026:

Faez Ahmed, the Esther and Harold E. Edgerton Associate Professor in the Department of Mechanical Engineering, received a 2025 Air Force Office of Scientific Research Young Investigator Program Award. The award provides early-career U.S. scientists and engineers up to $450,000 over three years to support innovative research.

Srini Devadas, the Edwin Sibley Webster Professor in the Department of Electrical Engineering and Computer Science, received the 2026 ACM-IEEE CS Eckert-Mauchly Award. The award recognizes pioneering contributions to computer and digital systems architecture and honors individuals whose work has had broad industrial and academic impact.

Elfatih Eltahir, the H. M. King Bhumibol Professor in the Department of Civil and Environmental Engineering, received the 2026 Arab Water Council Prize. Awarded under the theme “Leading Resilient and Climate-Adaptive Water Management Systems,” the prize recognizes significant contributions to understanding climate variability, water systems, and the Nile River basin.

William Green, the Hoyt C. Hottel Professor in the Department of Chemical Engineering and the director of the MIT Energy Initiative, was awarded the prestigious Jürgen Warnatz Gold Medal. The Warnatz Medal is one of the highest awards of the Combustion Institute and is presented biennially to one scientist in recognition of exceptional scientific contributions having a major impact on combustion applications or research.

Neville Hogan, the Sun Jae Professor in Mechanical Engineering, Emeritus, received the IEEE International Conference on Cyborg and Bionic Systems Best Paper Award for the paper, “Force Control Requires Stiffness Modulation During Unstable Contact Tasks.”

Dina Katabi, the Thuan (1990) and Nicole Pham Professor in the Department of Electrical Engineering and Computer Science, was named to Forbes and Know Your Value’s 2026 “50 Over 50” list. The recognition honors innovators whose work is shaping the future of science, technology, business, and society.

Robert Langer, the David H. Koch (1962) Institute Professor in the departments of Biological Engineering and Chemical Engineering, received the 2026 Global Distinguished Engineer Award. Presented by the European Academy of Engineering, the award recognizes exceptional engineering leadership, original innovation, and sustained impact on technology, industry, health, sustainability, and society. Langer was also selected to receive the 2026 AIChE Shaping the Future Award, which recognizes a career that has transformed the impact of chemical engineering on human health and society.

Jelena Notaros, an associate professor in the Department of Electrical Engineering and Computer Science, received an Air Force Office of Scientific Research Young Investigator Research Program Award for her project, “Quantum Computing: Integrated-Photonics-Based Systems for Selective Addressing of Trapped Ions in Multi-Ion Chains.” Notaros also received an SRC Young Faculty Award, which recognizes outstanding early-career faculty members whose research demonstrates novelty, technical excellence, and significant contributions to the semiconductor research community. Additionally, she was named to MIT Technology Review’s “35 Innovators Under 35” list for 2026 — a recognition that honors young innovators whose work is changing the future of science and technology.

Kristala L. Jones Prather, the Arthur Dehon Little Professor, MacVicar Faculty Fellow, and head of the Department of Chemical Engineering, received the 2026 AIChE Food, Pharmaceutical and Bioengineering Division Award in Chemical Engineering. The award recognizes outstanding chemical engineering contributions in the food, pharmaceutical, and bioengineering fields.

Lindsey Raymond, an assistant professor in the Department of Electrical Engineering and Computer Science, was named a 2025 Early Career Fellow by Schmidt Sciences’ AI2050 program. The fellowship supports ambitious, socially beneficial research addressing important challenges in artificial intelligence and its impacts on society.

Daniela Rus, the Panasonic Professor in the Department of Electrical Engineering and Computer Science and the director of the MIT Computer Science and Artificial Intelligence Laboratory, was named to TIME’s 2026 list of the 100 Most Influential People in AI in the “Thinkers” category. The annual list recognizes individuals shaping the future of artificial intelligence. Rus also received the 2026 High-Tech Prize of the Bavarian Minister-President. Jointly awarded by the Bavarian State Government and the Bavarian Academy of Sciences and Humanities, it is Germany’s most highly endowed award for technology and engineering.

César Terrer, the TianFu Group Career Development Professor and associate professor in the Department of Civil and Environmental Engineering, received both the AGU James B. Macelwane Medal and the Global Environmental Change Early Career Award. Together, the honors recognize exceptional early-career contributions to earth and environmental sciences, including research on large-scale environmental change.

Federico Tessari, a research scientist and incoming tenure-track assistant professor in the Department of Mechanical Engineering, received the IEEE International Conference on Cyborg and Bionic Systems Best Paper Award for the paper, “Force Control Requires Stiffness Modulation During Unstable Contact Tasks.”

Q&A: How ocean carbon sequestration changes under different glacial conditions

Mon, 11/16/2026 - 2:00pm

Over the past 800,000 years, Earth’s history has been marked by dramatic shifts in climate, swinging from warm periods to cold, when large ice sheets covered much of the Northern Hemisphere and sea ice greatly expanded. The transition between the two climates is paced by the Earth’s orbit with respect to the sun, but changes in the amount of carbon dioxide (CO2), a greenhouse gas, also affects global temperatures. By understanding how CO2 is exchanged between the atmosphere and the ocean, scientists hope to apply that knowledge to future climate scenarios.

In a new Science Advances paper, researchers at MIT and the Université du Québec à Rimouski have taken an approach to understanding how processes that occur in the Southern Ocean around Antarctica during glacial periods, such as ocean cooling, sea-ice expansion, and changes in ocean circulation, contribute to global ocean carbon storage and atmospheric CO2 drawdown. By looking at how these climate processes work together, rather than how they act independently, they’re piecing together the complex, interconnected systems that feed into it. 

In this Q&A, one author on the paper, Jonathan Lauderdale, a research scientist in MIT’s Department of Earth, Atmospheric and Planetary Sciences (EAPS) and MIT’s Center for Sustainability Science and Strategy (CS3), explains how and where the ocean stores carbon, and why coupling glaciation processes changes the outcome.

Q: How does the ocean contribute to the global carbon cycle?

A: It’s a key player. Trying to figure out how CO2 gets into the ocean and stays there for thousands of years is a question that scientists have been working on for over 40 years. We generally think about three important ocean carbon stores. One is solubility. If you have a warm soda versus a cold soda, the fizzy one has more CO2 dissolved in it because it's cooler. We can calculate how much more CO2 cold seawater can hold and, because cold waters are dense and sink, the carbon is stored in the deep ocean.

Second is biology. Phytoplankton take up CO2 near the surface when they photosynthesize, and then they die or get eaten by zooplankton and packaged into fecal pellets. These organic particles sink into the deep ocean and are converted back into dissolved CO2 by bacteria. 

Third is disequilibrium. Solubility gives us an estimate of how much CO2 could be in the water if it is in long-term balance with the air, but there’s lots of other processes that affect CO2 levels, like biology, ocean circulation, and slow gas exchange with the atmosphere. Disequilibrium is the difference between how much carbon the water mass could hold and how much carbon is actually there.

Q: Are those processes different than the currents and overturning that happen on global scales?

A: These processes are tightly linked to ocean currents. The Southern Ocean Overturning Circulation is a major highway that brings carbon-rich waters back to the surface, where they interact with the atmosphere and phytoplankton. How long that water remains at the surface determines how much CO2 it releases and how much gets returned into the ocean interior as disequilibrium.

Q: When scientists have explored different mechanisms like seawater cooling, sea ice expansion, and ocean circulation change during ice ages, they looked at them separately from each other. Why is that?

A: Studying one process at a time is useful because it helps isolate cause and effect. Early papers were based on very simple, idealized box models, which couldn't really capture the tight linkage between physical and biogeochemical processes that we can with our more complicated models today. The reverse is also true: We can run experiments using complex climate models, but then it's very difficult to say what process is responsible for a result unless you only change one thing at a time.

Q: When you look at your model’s outcomes separately, what do you see?

A: When we change Antarctic sea ice on its own, we see a slight increase in atmospheric CO2, and the same thing happened when we just run with a glacial circulation. This might imply that sea ice and circulation are unimportant for ice age ocean carbon storage. When we run a “full” experiment with all three mechanisms — extensive sea ice, glacial circulation, and cooler temperatures — we see atmospheric CO2 decline, i.e., ocean carbon increases. This is consistent with what other scientists have been seeing in different models. 

Yet, how the carbon was stored didn’t quite make sense to us, because we saw only a small net increase in carbon due to solubility, which we expected to dominate, while most of the added storage appeared in the disequilibrium carbon pool.

Q: With coupling, how does this change?

A: When we coupled sea ice and circulation we saw a significant CO2 drawdown and the disequilibrium carbon fingerprint of our “full” glacial ocean. Cooling still matters, but it is not the complete picture; it's like we now know how to get the right answer for the right reasons.

We can also explain why the disequilibrium change was so important: it depends on how long the water mass spends in contact with the atmosphere in the Southern Ocean. More sea ice, or the glacial circulation on their own, decrease exposure. Together, atmospheric exposure is significantly reduced, so there’s a small window for CO2 exchange. That's why much more of the upwelled deep-ocean carbon is sent back into the ocean interior. 

Of course, there are caveats, but our findings point to interactions among physical processes, like sea ice and circulation, as an important driver of ice age ocean carbon storage.

Q: There's a lot of talk recently about the possibility of the Atlantic Meridional Overturning Circulation (AMOC) collapsing. If that happened, looking at these results, how might that alter current carbon sequestration?

A: We were concentrating on the Southern Ocean, but in our scenarios the AMOC did slow down. The ocean is like a layer cake, with North Atlantic waters above denser Southern Ocean waters. Our AMOC reductions caused the boundary between Southern Ocean and North Atlantic waters to get shallower. This is important because rapid mixing near the seafloor is another, albeit slower, way for carbon to escape from isolated Southern Ocean waters into the better-ventilated North Atlantic waters, and back to the atmosphere. Reducing mixing across the water-mass boundary by raising it slows that escape. That's why sea ice alone didn't lead to carbon drawdown — because mixing across the deep boundary kept the two waters connected. So even though you've got this lid of sea ice in the Southern Ocean that's locally trapping carbon, it’s still got a route back to the atmosphere. 

It’s difficult because climate is changing in other ways at the same time. On one hand, a collapse of the AMOC could shoal that water mass boundary and isolate the Southern Ocean waters more. On the other hand, warming and a decrease in Antarctic sea ice cover might increase exposure time, leading to ocean carbon loss instead.

Lauderdale coauthored the paper with EAPS Orofessor Raffaele Ferrari and Sandy Grégorio and Louis-Philippe Nadeau at the Université du Québec à Rimouski in Canada. This research was supported, in part, by the Natural Sciences and Engineering Research Council of Canada; the Canada First Research Excellence Fund; Schmidt Sciences LLC and the MIT Climate Grand Challenge project Bringing Computation to the Climate Challenge (BC3); the Simons Collaboration on Computational BIOgeochemical Modeling of marine EcosystemS (CBIOMES); and the U.S. National Science Foundation.

Pipilotti Rist receives the 2027 Eugene McDermott Award in the Arts at MIT

Mon, 11/16/2026 - 1:15pm

Artist Pipilotti Rist is the recipient of the 2027 Eugene McDermott Award in the Arts at MIT, which will be awarded at her MIT public presentation on April 8, 2027, and includes an artist residency at MIT in spring 2027. 

Rist is a pioneer of spatial video art whose work explores the deep connections between the human body, nature, technology, and sensory perception using immersive video and light installations.

Rist said upon receiving the award, “I am kissing your Institute and your feet. I am immensely touched and grateful to have this trust and investment in ideas I’ve yet to conceive. I’m honored to receive an award that has been granted to so many of my heroes, and look forward to collaborating closely with the brilliant minds at MIT.”

The Eugene McDermott Award in the Arts at MIT recognizes innovative artists working in any field or cross-disciplinary activity. The prize represents an investment in the recipient’s future creative work, rather than a particular project or lifetime of achievement. 

The official announcement was made at the Council for the Arts at MIT’s 52nd annual meeting at MIT in October. Past recipients include Santiago Calatrava, Es Devlin, Gustavo Dudamel, Olafur Eliasson, Robert Lepage, Audra McDonald, Suzan-Lori Parks, Bill Viola, and Pamela Z, among others.

MIT Vice Provost for the Arts Keeril Makan says, “Pipilotti Rist has spent four decades using digital tools to dissolve the line between screen, body, building, and landscape. Her immersive multimedia installations slow us down, return us to our bodies and the spaces we occupy, and gather us into shared experience, rather than fragmenting our attention. We look forward to presenting Pipilotti Rist with MIT’s highest award in the arts, and it comes at an exciting moment as we gather our students and faculty to activate our new building for the School of Architecture and Planning, the MET Warehouse.”

A distinctive feature of the award is a short residency at MIT, which includes a public presentation of the artist’s work, as well as substantial interaction with students and faculty. The goal of the residency is to provide the recipient with unparalleled access to the creative energy and cutting-edge research at the Institute and to develop mutually enlightening relationships in the MIT community. 

The Eugene McDermott Award in the Arts at MIT was established in 1974 by Margaret McDermott (1912–2018) in honor of her husband, Eugene McDermott (1899–1973), a co-founder of Texas Instruments and long-time friend and benefactor of MIT. The award is presented by the Council for the Arts at MIT, a volunteer group of alumni and friends, fulfilling their mission “to foster the arts at MIT and to act as a catalyst for the development of a broadly based, highly participatory program in the arts.”

The award is bestowed upon individuals whose artistic trajectory and body of work have achieved the highest distinction in their field and indicate they will remain leaders for years to come. The award recognizes innovative talents in any arts discipline, and, thanks to the McDermott family gift, offers its recipient a $100,000 cash prize and a campus residency. The McDermott Award reflects MIT’s commitment to risk-taking, problem-solving, and connecting creative minds across disciplines. Since it was established in 1974, the award has been bestowed upon 40 individuals who work in performing, visual, literary, and media arts.

Pipilotti Rist, a pioneer of spatial video art, was born 1962 in Grabs in the Swiss Rhine Valley on the Austrian border and has been a central figure within the international art scene since the mid-1980s. Astounding the art world with the energetic exorcistic statement of her now-famous single-channel videos, such as “I’m Not The Girl Who Misses Much” (1986) and “Pickelporno” (1992), her artistic work has co-developed with technical advancements and in playful exploration of its new possibilities to propose footage resembling a collective brain. Through large video projections and digital manipulation, she has developed immersive installations that draw life from slow caressing showers of vivid color tones, like her works “Sip My Ocean” (1996) or “Worry Will Vanish” (2014).

For Rist, showing vulnerability is a sign of strength on which she draws for inspiration. With her curious and lavish recordings of nature (to which humans belong as an animal), and her investigative editing, Rist seeks to justify the privileged position we are born with, simply by being human. Her installations and exhibition concepts are expansive, finding within the mind, senses, and body the possibility for endless discovery and poetical invention. “Pixel Forest” (2016), made from 3,000 LEDs hung on strings, resembles a movie screen that has exploded into the room, allowing viewers an immersive walk through three-dimensional video. As she herself puts it, “beside the energy-intensive exploration of the geographical world, pictures, films and sounds have been and are the spaces into which we can escape ... The projector is the flamethrower, the space is the vortex, and you are the pearl within.”

Since 1984, Rist has had countless solo and group exhibitions, and video screenings worldwide. Her recent solo exhibitions are “Your Palm is My Universe” at UCCA Beijing (2025), “Electric Idyll” at the Fire Station Doha (2024), “Prickling Goosebumps & A Humming Horizon” at Hauser & Wirth New York and Luhring Augustine Chelsea (2023-24), “Behind Your Eyelid” at Tai Kwun Hong Kong (2022), “Big Heartedness, Be My Neighbor” at The Geffen Contemporary, MOCA, Museum of Contemporary Art Los Angeles (2021–22), and “Your Eye Is My Island” at MoMAK, The National Museum of Modern Art Kyoto and ART TOWER MITO (2021). “Åbn min Lysning. Open my Glade” at Louisiana Museum of Modern Art Humlebæk Denmark (2019), “Sip My Ocean” at the Museum of Contemporary Art Sydney (2017–18), “Pixel Forest” at New Museum New York (2016–17), and “Your Saliva is My Diving Suit of the Ocean of Pain” at Kunsthaus Zürich (2016) all resulted in record-breaking attendance for each institution. 

Arianna Doss named a 2027 Rhodes Scholar

Sat, 11/14/2026 - 9:00pm

MIT senior Arianna Doss has been selected as a 2027 Rhodes Scholar and will begin fully funded postgraduate studies at Oxford University in the U.K. next fall. 

Doss, who is from Houston, is double-majoring in biological engineering and political science. At Oxford, she plans to pursue a master’s degree in comparative social policy, followed by a master’s in global health science and epidemiology.

She aspires to become a physician and health policy leader who advances health equity, particularly to improve maternal health and chronic disease outcomes among marginalized groups.

Doss was supported by Associate Dean Kim Benard and the Distinguished Fellowships team in Career Advising and Professional Development. She received additional mentorship and guidance from the Presidential Committee on Distinguished Fellowships.

“No two applicants have the same story, and MIT students never fail to surprise us with their ingenuity,” says Professor Taylor Perron, who co-chairs the committee along with Professor Nancy Kanwisher. “Seeing Arianna earn this recognition and opportunity is exhilarating. We don’t know what lies ahead, but I’m confident the world will be better as the result of her work.”

At MIT, Doss worked in the Bryson Lab for more than two years, studying protein degradation within the phagosomes of mammalian macrophages. At Rice University’s Segatori Lab, she investigated gene circuits in retinal pigment epithelial cells to reduce chemotherapy toxicity. She has also researched rural-urban disparities in maternal mortality and analyzed the effects of abortion clinic closures on maternal health.

Doss’s policy and clinical experiences include serving as a maternal health intern for Representative Lauren Underwood, supporting the Black Maternal Health Caucus, and interning at Northwestern Medicine’s Bronzeville Outpatient Center, where she shadowed in family medicine and studied medication hesitancy among hypertensive patients. As a patient navigator with Link Health, she connects patients in East Boston with public benefits. 

Doss is the founding president of the MIT Alliance for the Advancement of Black Athletes, and previously served as attorney general of MIT’s Black Students’ Union. She is also a three-year starter for MIT’s nationally ranked women’s soccer team. Her honors include NCAA Scholar All-American, First Team All-Region, the Princeton Prize in Race Relations Certificate, and Scholastic writing awards.

Waves that clear waste from the brain during sleep are impaired in older adults

Fri, 11/13/2026 - 5:00am

As people age, they may sleep for shorter periods and wake up more frequently during the night. In addition to those disruptions, MIT researchers have now discovered that a process responsible for flushing waste from the brain during sleep is commonly impaired in older people.

The researchers found that waves of cerebrospinal fluid (CSF), which help remove debris from the brain, are significantly reduced in older adults. This impairment is linked to a decrease in the slow brain waves that normally occur during deep sleep, the researchers found, and is not just a result of getting less sleep.

The researchers now hope to explore whether the disruption of CSF flow increases the risk of neurodegenerative disorders, and if any of the sleep patterns they observed might offer a way to predict an individual’s risk.

“We know that people age really differently. Some people age very successfully with great cognitive outcomes, and other people develop cognitive decline. If we can figure out who has great trajectories of aging and what are the factors that are associated with that, we might be able to use that to predict future outcomes,” says Laura Lewis, the Athinoula A. Martinos Associate Professor of Electrical Engineering and Computer Science, a member of MIT’s Institute for Medical Engineering and Science and the Research Laboratory of Electronics, and an associate member of the Picower Institute for Learning and Memory.

Her lab also plans to investigate whether there might be any interventions that could improve CSF flow in older adults.

Lewis is the senior author of the study, which appears today in Nature Aging. Sydney Bailes, a postdoc in the MIT Laboratory for Information and Decision Systems, and Stephanie Williams, a former visiting student at MIT who earned her PhD from Boston University in 2026, are the lead authors of the paper.

Smaller waves

During the deepest stages of sleep, also called non-REM sleep, high-amplitude, low-frequency waves contribute to physical recovery, memory consolidation, and many other functions. In a 2019 study, Lewis found that these slow waves are also critical for generating waves of CSF that help flush waste from the brain.

That finding led her to wonder whether reductions in slow waves in older people might also reduce CSF flow. Previous studies of aging have yielded conflicting evidence; some found that CSF flow goes down in older adults, while others found it goes up, but these studies did not monitor flow during sleep.

In the latest study, the researchers recruited volunteers in two age groups: 18 to 39 and 60 to 84. The tests lasted for about two hours and began late at night, so that participants would be sleepy. 

The researchers used traditional fMRI to measure participants’ blood flow in the brain, along with a modified MRI technique, developed by Lewis, that measures CSF flow. While the subjects were in the scanner, the researchers also measured their brain waves using EEG. 

Many of the participants fell asleep in the scanner, allowing the researchers to compare their brain activity while awake and while sleeping. 

As they had seen in earlier work, the researchers found that when the younger adults fell into non-REM sleep, their EEG slow waves became stronger, and their CSF waves became larger. However, in the older group, the average CSF waves were significantly smaller than those seen in the younger group. 

Impaired flow

In additional experiments done while subjects were fully awake, the researchers tried to figure out what was causing the age-related declines in CSF flow.

First, they explored the possibility that changes in a phenomenon known as neurovascular coupling might play a role. When localized brain activity increases the demand for oxygen, nearby blood vessels dilate and deliver more oxygen and nutrients. That increase in blood flow can be measured with fMRI.

The researchers tested each subject’s neurovascular coupling by measuring blood flow while the subjects looked at a flickering checkerboard image — a task that typically drives an increase in neural activity, along with an increase in blood volume.

In many of the older subjects, while their neural activity did go up, the blood flow didn’t increase much, suggesting that the blood vessels’ ability to respond to increased oxygen demand has diminished.

During another task, the researchers tested whether the brain’s vascular system might be impaired in the older subjects. They asked the subjects to hold their breath, then measured what happened inside the brain. Normally, this leads to an increase in carbon dioxide levels in the blood, which triggers an increase in cerebral blood flow. However, in the older participants, the average blood flow increase was smaller than the increase seen in the younger subjects.

These two factors — reduced neurovascular coupling and diminished blood flow — accounted for much of the decline in CSF flow but could not explain all of it, according to a statistical model developed by the researchers. This suggests that there may be other age-related factors that contribute to the reduction in CSF flow, they say.

“Aging is so complex, and so many different things change, that it’s really hard to disentangle the individual factors,” Bailes says. “We try to account for some of the mechanisms, but there definitely are a lot of other factors related to vascular health and lifestyle. If you were able to measure all of them, you might be able to break it apart into different explanations for why an individual’s CSF flow might be more or less impaired.” 

Enhancing CSF flow

Previous studies have found that sleep disturbances are linked to a greater risk of neurodegenerative disorders such as Alzheimer’s and other forms of dementia.

None of the subjects in this study had been diagnosed with a neurodegenerative disorder, but the researchers found that some of the older adults showed signs of gray matter atrophy, which was correlated with a greater decrease in CSF flow. They hope to continue studying some of the participants in hopes of discovering whether changes to CSF flow are associated with a higher risk of developing a neurodegenerative disorder.

They also hope to test possible interventions that might help increase CSF flow. In a recent study, Lewis’ lab found that short bursts of “pink noise” during sleep could increase CSF flow in younger adults, but they have not yet tested that intervention in older adults. 

In other work, they plan to study why some older adults don’t experience a dropoff in CSF flow. In this study, the researchers found that while as a group, older adults showed weaker CSF waves during sleep, some individuals in the group maintained normal CSF waves.

“It’s really intriguing that for some of these older adults, their CSF fell right in the range of the healthy young adults,” Lewis says. “We are really interested in what’s different in those people and how that might be related to future health outcomes.”

The research was funded by the National Institutes of Health, the Simons Collaboration on Plasticity in the Aging Brain, a McKnight Scholar Award, a Pew Biomedical Scholar Award, an MIT EECS Transformative Research Award, the Picower Institute Neurological Disorder Discovery Research Fund, a Sloan Fellowship, and the Department of Defense National Defense Science and Engineering Graduate Fellowship Program.

New AI “copilot” helps engineers design better structures

Fri, 11/13/2026 - 12:00am

The use of algorithms to design structures has led to new parts optimized for ultralight aircraft, high-bandwidth antennas, and F1 racing cars.

But the best algorithmic design techniques still require lots of computation and time: Users only see the final design after it’s complete, after which they often must start the design process all over again to make changes. That process also makes it harder to trust that the final designs are structurally sound.

MIT researchers are making algorithmic designs more collaborative with a system that acts like an AI copilot. As the algorithm is generating the structure, it presents regions to the human designer for feedback.

In an open-access Communications Engineering paper describing the system, the researchers showed it led to more easily manufacturable structures with higher buckling loads while adding only 15 seconds to the total design time compared to all-algorithmic approaches.

The researchers believe the approach, which uses a design technique known as topology optimization, could make algorithmic design more accessible and manufacturable for a broader class of design engineers.

“Right now, the human doesn’t have any agency or control over the design process as it’s playing out,” explains lead author Dat Ha PhD ’26, a recent MIT civil and environmental engineering doctoral student who is now a structural designer at LeMessurier. “For these really complex designs, we wanted to also tap into our brain, which is the most complex machine there is, to solve these design problems in a way that’s more practical to industry.”

Joining Ha on the paper are Ferdous Alam, a former MIT postdoc who is now an assistant professor at Georgia Tech; Markus Buehler, the MIT Jerry McAfee (1940) Professor of Engineering; Faez Ahmed, MIT associate professor of mechanical engineering; and senior author Josephine Carstensen, MIT associate professor of civil and environmental engineering.

Bringing a human in the loop

Topology optimization is one of the most advanced and promising algorithmic design techniques. It uses formal optimization techniques to iteratively fill in the space given for a design in a way that maximally achieves the stated goals. For most applications, the process requires a lot of computation and time. When it’s finally done, the designs can be impractical to build or have other issues the engineer didn’t consider. For instance, some designs might have eight or more beam connections at a single point, which would be difficult to make in the real world.

“Engineers are always very skeptical of computer programs, especially if they give you very weird-looking or unexpected designs,” Ha says. “On paper, topology optimization is really powerful and can achieve great, high-performing designs, but its black box nature has made it less widely adopted in industry.”

Topology optimization has gotten traction in the automotive and aerospace industries, but it’s been used less to design buildings or bridges, partly because design engineers need to know they can trust the output.

“A big limiting factor has been that the design engineer has very little influence over how the design evolves and therefore can’t control undesirable features,” Carstensen explains. “The lack of control and lack of ownership also inhibits its use.”

To address those problems, the researchers made topology optimization more collaborative in an approach that Carstensen compares to having a design copilot.

Prior attempts at adding human judgment slowed things down dramatically. The researchers’ new system uses machine learning to quickly identify the most likely regions of an algorithmic design that a human would want to change. To train the machine-learning model, the researchers created a synthetic dataset of probable human feedback to thousands of designs that critiqued things like aesthetics, performance criteria, and manufacturability.

“We thought if we could create a record of where humans have made changes to these programs, then perhaps there’s a pattern to those choices, which is not very different from other types of machine-learning approaches,” Ha says.

Carstensen sees the system as bringing together the best of algorithmic and human design approaches.

“The idea is to tap into the brains of the designers and get them to infuse their domain knowledge as the design evolves,” Carstensen says.

In demonstrations, the researchers showed their human-in-the-loop approach improved manufacturability and linear buckling load by 39 percent while only modestly increasing the total design time.

Right now, the approach only allows users to make structures within identified regions thicker or thinner, but they hope to add other options for modification in future versions.

“The hope is that by addressing these hurdles, we open up topology optimization to a much wider audience, not only in industry but also to researchers and students in academia,” Carstensen says.

From designs to structures

The researchers see their approach complementing AI only-approaches to design.

“We are arguing there’s room for both,” Carstensen says. “If you are in a discipline that has a lot of computational resources, you can use the other approach. But if you really need the human input, you can use this collaborative approach.”

Adding human input also serves to increase trust.

“In some settings, the humans need to make the final decisions before advancing the final design,” Carstensen says. “By involving humans, we can get higher-performing designs to the real world to create higher-performing structures. Because it’s not just about designing good artifacts; it’s about creating good artifacts that humans will actually accept and make.”

The work was supported, in part, by the MathWorks Engineering Fellowship Fund, TATA Steel, and the MIT Generative AI Impact Consortium.

Helping robots fit into the human workplace

Fri, 11/13/2026 - 12:00am

Despite fears about robots replacing humans at work, there are still plenty of things humans can do far better than machines. That’s fine with Robust AI, a startup building robots designed to complement what humans do best.

The company’s mobile robot, named Carter, resembles a utility cart on wheels and is designed to fit seamlessly alongside humans in fast-changing warehouse and factory environments. Carter features a force-sensitive handlebar for easy guidance, an AI-powered vision system, and holonomic wheels for instant movement in any direction.

As a result, humans can cut Carter off, push and pull it in any direction with one hand, and direct it with gestures or the onboard touch screen.

The company’s robots are currently deployed in logistics warehouses and on factory floors, where they help humans while picking or stocking shelves and moving items across facilities autonomously.

“In these environments, what really determines success is how the overall system performs,” explains Anthony Jules ’92, SM ’92, who co-founded the company with MIT Professor Emeritus Rodney Brooks. “The robot does some of the work, the people do some of the work, and the backend systems do some of the work. It’s that combination that determines how well your system performs. We believe it’s really important that people have an appropriate level of agency and input so that they’re always moving the system forward.”

Collaborative robots

Brooks was a faculty member at MIT from 1984 until 2010 and served as the director of the Computer Science and Artificial Intelligence Laboratory (CSAIL) from 1997 until 2007. While at MIT, he co-founded iRobot, which invented the popular Roomba vacuum and went public in 2005. Later, he co-founded the company Rethink Robotics, which also designed collaborative robots.

Jules met Brooks as a graduate student at CSAIL in the 1990s. The two never worked together, but their labs were on the same floor. The pair kept in touch after Jules graduated and co-founded the company Sapient, which went public before being acquired by Publicis in 2015.

In 2019, Brooks bumped into Jules at a robotics conference. He told Jules he was thinking of starting a robotics software company, which he compared to a video game engine for training robots. Robust AI was founded later that year.

The Covid-19 pandemic caused the founders to shift to creating robots that could autonomously disinfect surfaces, but when it was discovered that Covid-19 was not transmitted via surfaces, they had to pivot again.

“That forced us to decide how to use this technology we had been building, which was really about helping robots understand the world more fully,” Jules recalls. “Long story short, I said to Rodney, ‘I have an idea. You’re going to love it and you’re going to hate it. You’re going to love it because it’s solving problems for customers that we’ve been hearing about, and you’re going to hate it because we have to build hardware.”

The founders had been interviewing potential customers for years through Robust AI’s early pivots. They finally decided on logistics and manufacturing as the industries to target early on.

“When people think about automation in factories, they think about pictures of the 5 percent that are fully automated with $50 to $200 million dollar systems,” says Jules, who now serves as Robust AI’s chief innovation officer. “They don’t think about the vast majority of factories that aren’t automated at all. These factories don’t even have a conveyor belt. They are not going to tear everything out and replace it with a $200 million system.”

Robust AI’s robots can be phased into operations as needed and offer several workflows for customers. In one workflow, the robot can guide human workers to specific spaces to retrieve items. In another it can follow humans to important spots for loading. The primary workflows involve Carter robots moving things from one place to another, often in large warehouses or factories, where they say human workers can walk on average 30,000 steps a day.

“These are huge spaces, so a person might load the robot with things and then need to bring them to another station, but that could be 400 feet away,” says, Brooks, Robust AI’s chief technology officer. “Now the person can just say ‘done’ and the robot brings it instead. How the robots are integrated into the workspace is entirely up to the customer.”

Getting work done

Leila Takayama, Robust AI’s vice president of human robot interaction, recently discovered a peer-reviewed study of robots in hospitals, which were used to do things like gather dishes, move sheets, and transport other equipment. The researchers discovered that, in many hospitals, a large number of robots were turned off and against the walls. The nurses said they broke.

“What really happened was the nurses were rushing with a patient on a gurney, and the robot sees them in the hallway and stops dead, and they couldn’t get the gurney around them,” Brooks says. “The robots were getting in the way of nurses’ primary tasks.”

Robust AI’s approach to technology development is to understand workers’ primary tasks deeply and build features that accelerate those tasks, whether a human or a robot is doing them.

“We let humans do what humans are good at in the physical world, which we’ve had millions of years of evolution to get good at, and have the digital system respond to that and provide us with all the things that it is good at,” Brooks explains.

Addressing nuclear challenges, from radiation detection to power in space

Thu, 11/12/2026 - 5:10pm

As humans expand their ambitions in deep space, energy-intensive missions can’t afford to rely solely on combustion or solar power. They will need nuclear energy for propulsion and surface habitation.

Jake Hecla ’17, who holds a dual appointment as assistant professor in the Department of Nuclear Science and Engineering (NSE) and Charles Stark Draper Career Development Professor of Aeronautics and Astronautics (AeroAstro), is researching ways to address one of the principal heat transfer challenges in such deployments of nuclear power in space: a lack of atmosphere. He is also working on technologies to enable the deployment of aerospace reactors, observe their operational signatures from long range, and create the technical basis for future arms-control agreements limiting the weaponization of these systems.

Thermal management in space

The next generation of space nuclear reactors demand considerably higher power than past missions, requiring the conversion of heat to power, typically using a Brayton or Rankine cycle. Such conversions deliver efficiencies of around 30-40 percent in terrestrial applications — and a lot of heat that needs to be removed. Because there is no atmosphere to rely on for convection, engineers use radiators instead. But traditional solid metal or composite radiator panels are mass-intensive, a significant impediment for space applications.

One of Hecla’s research projects at MIT addresses the limitations of an alternate solution: the liquid droplet radiator (LDR). First proposed by researchers at the University of Washington in the late 1970s, the LDR could be just as effective as conventional radiators, especially because their considerably lower weight is an advantage, helping radiators constitute just a fraction of the spacecraft’s mass.

In place of solid panels, LDRs use sheets of liquid droplets to radiate heat away. The LDR works by spraying coolant from an emitter to a collector as sheets of tiny droplets. As these droplets travel through space, they radiate and cool. They then impact a collector, which then pumps the coolant back to the heat source in a cyclical operation.

The principle sounds straightforward, but droplet retrieval is often complicated by drifts introduced by charging, splashing at the collector, and emitter misalignment. Engineers can’t afford to lose too many of them — acceptable loss rates could be one droplet in every million for a 10-year mission — because that will increase the coolant load that will need to be transported to space. (The rate might be small, but the radiator cycles the coolant every few seconds, meaning even a minuscule loss rate adds up over time.) Hecla is investigating a variety of collection techniques including electromagnetic collectors that can improve retrieval efficiency, and even allow maneuverability while the radiator is in operation.

If successful, this could reduce the mass of the cooling system for a future nuclear electric power space reactor by an order of magnitude. The work will vastly improve the potential of nuclear power in space and potentially expand its utility to include low-Earth orbit projects as well.

An early interest in radiation detection — at MIT

With his appointment as an assistant professor, Hecla’s career has come full circle since his undergraduate days as a student at NSE.

Early on, he was especially interested in radiation detection techniques because of their use in many practical applications.

That specific interest was seeded during a formative childhood trip to New Mexico, which included a visit to a shop selling rocks and minerals. Hecla remembers being fascinated by the goods, which included uranium ore. He was hooked. Hecla researched the field and bought and built a do-it-yourself Geiger counter. “Throughout high school I was buying and repairing radiation equipment in my home. I collected high-voltage equipment, built a homemade X-ray system, and it was all downhill from there,” Hecla says.

That passion for measuring radiation spilled over to MIT. One of Hecla’s undergraduate research projects involved developing a prototype of a photocathode-driven X-ray unit, in collaboration with Massachusetts General Hospital. The research team he worked with formed a startup called Advanced Tomography Systems. As part of another undergraduate research opportunities program (UROP), Hecla worked with Areg Danagoulian, associate professor at NSE, on a novel approach for treaty verification.

A visit to Chernobyl

While at MIT, another formative event would cement Hecla’s deep commitment to measuring radiation: a friend, Carl Willis, now a professor at the University of New Mexico, invited Hecla to travel to Chernobyl. “It was to better understand the realities of living and working in a contaminated environment,” Hecla says.

The experience shook him — and he visited several times more during his undergraduate years. Since 2017, Hecla has served as scientific adviser to Clean Futures Fund, a U.S.-based nonprofit that ensures activities in the exclusion zone and at the power plant are as well supplied as they can be, providing basic essentials, personal protective equipment, and detection equipment.

“Chernobyl has a lot to teach us, both about the necessity of a dedication to safety in nuclear power and about managing the consequences of radiation release,” Hecla says. “It’s an enormously challenging environment where a small amount of well-informed work can have an outsized impact.”

Chernobyl also provided the context for Hecla’s graduate studies at the University of California at Berkeley, where he studied radiation imaging techniques. For his master’s, the goal was to use a free-moving radiation detector to reconstruct a three-dimensional image that would inform workers about radiation intensity so they could be judicious about exposure.

Another area of radiation detection, to detect antineutrinos as an indicator of nuclear fission, anchored his doctoral research — Hecla’s PhD dissertation was titled “Optical Diagnostics for Antineutrino Detector Media” — advised by Kai Vetter. The goal was to use the particles as an essentially unshieldable signature of reactor operation.

Guarding against the misuse of nuclear energy

As Hecla’s graduate research showed, using radiation signatures is an effective way to track when nuclear energy is being misused.

Hecla’s work on aerospace applications of nuclear power extends to both improving the viability of space reactors and improving our ability to monitor orbital and sub-orbital applications of nuclear energy.

His current research focuses on understanding and tracking air-breathing nuclear propulsion, which is a technique in which heat from a reactor is used for propulsion. Hecla became especially interested in the project when, in 2018, the Russian government announced that they were developing a nuclear-powered, nuclear-armed missile, the 9M730 Burevestnik, that would pose a direct security threat to Europe.

The missile was said to be powered by an air-breathing nuclear engine, enabling extreme duration flight. In 2019, a mishap with the system killed at least five scientists, and generated a plume of radionuclides that drifted across the White Sea, indirectly confirming that the Russians were serious about developing the missile.

Between 2023-25, as a Stanton Postdoctoral Fellow at MIT’s Laboratory for Nuclear Security and Policy, working with R. Scott Kemp, Hecla developed techniques through which he could understand the performance envelope and flight signatures of the nuclear propulsion system in the missile. That information could be used to limit or ban use of such military systems, or contribute to new countermeasures.

Along with colleagues at MIT, Hecla is devising techniques to detect the presence of nuclear weapons by studying the gamma-induced neutral field from satellites, which can share information about the presence of fissile material. Such a system is especially relevant given the Kosmos 2553, a satellite launched by Russia in 2022 that is believed by American intelligence to contain components of a nuclear weapon. 

The threat of nuclear weapons in space is immense. “Someone may find it advantageous to destroy most assets in low and medium Earth orbit if it gives them an edge during combat,” Hecla says. As part of this work, researchers are looking to develop a highly sensitive compact neutron telescope to detect and locate the source of neutrons as a way to detect nuclear activity.

Most of Hecla’s work, including the space thermal management projects, sits at the intersection of nuclear and the aerospace or aeronautical fields.

Hecla is delighted to be back at MIT where it all started. “We have the right combination of talent here that is strong and growing,” he says. “And between world-class facilities like the MITR-II (MIT Nuclear Research Reactor), the PSFC (Plasma Science Fusion Center), the people, this is absolutely the best place for research,” Hecla says.

Q&A: From seed funding to scientific breakthroughs

Thu, 11/12/2026 - 4:45pm

The MIT Global Seed Funds (GSF) are designed to do more than launch research projects — they build international collaborations that make ambitious science possible. By supporting early-stage collaborations with researchers around the world, the program helps faculty test new ideas, exchange expertise, and lay the groundwork for discoveries that often grow far beyond the initial award. 

Matt Shoulders, professor of chemistry and department head, has experienced that impact firsthand. Since his first GSF grant in 2013, the collaborations these grants enabled have led to major research programs in protein engineering, photosynthesis, collagen biology, and gene-editing technologies, as well as subsequent funding. 

The GSF help Institute faculty and principle investigators create exciting new connections with colleagues abroad. GSF supports early-stage, open and publishable research collaborations with researchers at academic universities and public research institutions around the world. Administered by the Center for International Studies, GSF has awarded more than $30 million to over 1,300 faculty research projects since its inception in 2008. The current call for proposals from MIT faculty and researchers with principal investigator status is open until Dec. 3. 

Q: What makes the Global Seed Funds unique?

A: This has to be one of the standout ways to develop international collaborations that any school has anywhere. It's an extraordinary way to build serious relationships with people who have science that you can work on together, get to synergistic and exciting outcomes, and start projects that you could never get going otherwise.

I've found it to be an enormously valuable bridge-building mechanism for creating opportunities to work with other groups.

I started as an assistant professor in 2012 and received my first Global Seed Funds grant in 2013, probably within eight months of starting. That opportunity let me get to know an expert whose field of research was exactly what we needed. We published only one paper together, but that collaboration led directly to my first major NIH [National Institutes of Health] funding — a Director's New Innovator Award that brought nearly more than $2 million to the lab. It became the basis of what's still one of our lab's major research projects.

We never would have built that relationship otherwise. We visited their lab, they visited ours, our students exchanged visits, and together we built a whole platform for protein engineering that we still use all the time. Fourteen years later, that first GSF grant is still shaping the work we do.

That story has repeated itself over and over again. The collaborations have led us in lots of directions and created opportunities for different science along the way. Looking back, I think the GSF funding we've received can be directly connected to roughly $18 to $20 million in outside research funding. Of course, it wasn't only the GSF grants, but without them much of that later funding probably wouldn't have happened.

Q: Can you share a few examples of research that grew out of those collaborations?

A: One project is about engineering photosynthesis. The world needs more food, the climate is warming, fertilizer is becoming harder to obtain, water is increasingly scarce, and yet we need to grow more crops than ever. We need step-function improvements in agriculture, not just gradual ones.

We've focused on an enzyme called Rubisco, which takes carbon dioxide from the air and turns it into the sugars that build plants — and ultimately, all of us. Improving Rubisco has been a dream in agriculture for decades, but it's also been viewed as almost impossible.

One of the first pieces of funding we received for this project was a Global Seed Funds grant that allowed us to build a collaboration with a lab in Australia. They had technical assay expertise that existed in only a handful of laboratories worldwide. They came here, we went there, and we brought those capabilities to MIT. Now MIT is one of the few places that can perform those assays, and they've become part of the engine behind our entire Rubisco project.

That collaboration has since led to major foundation funding, publications, patents, and a great deal of exciting science. Whether we ultimately succeed in making Rubisco in a plant significantly better or not, we've already learned an enormous amount.

Another example involves collagen, the most abundant protein in the human body. There are dozens of diseases caused by collagen dysfunction, and we still can't treat them.

I met a researcher from Australia at a conference, and we realized our expertise complemented each other. The Global Seed Funds gave us the opportunity to build that collaboration. Since then, we received funding from the Mathers Foundation and multiple NIH grants. Today we've developed cartilage organoids that model these diseases, allowing us to study them and screen potential drug therapies. That entire research program exists because we were able to build that partnership.

The third example goes back to my very first Global Seed Funds grant. We wanted to develop a new way of doing directed evolution inside human cells, rather than bacteria. We had the idea, but we didn't have the adenovirus expertise to make it work.

That GSF-enabled collaboration gave us exactly the knowledge we needed. We recently published the first paper demonstrating CRISPR-Cas9 evolution inside human cells using the resulting platform. That work traces directly back to the GSF grant we received in 2013.

Q: How have MIT students benefited from these collaborations?

A: Most of the projects have involved graduate students, and they've really been the biggest beneficiaries. The collaborations often become the basis for their papers and dissertations.

When we've had the right undergraduate with the right interests, it's also been tremendously valuable, because they can spend an entire summer abroad supported by MISTI [MIT International Science and Technology Initiatives]. Michelle Huang is probably the flagship example. She spent several months in Australia, learning techniques that she brought back to MIT. She's a superstar who has now earned her PhD from Stanford, supported by an NSF [National Science Foundation] Graduate Fellowship awarded during her senior year at MIT.

Another great example is Julie McDonald, a superb MIT biology graduate student who worked on our Rubisco collaboration in Australia. She helped get that project going and has been tremendously successful. She's now a postdoctoral researcher at Rockefeller.

For our graduate students, these collaborations have been scientifically invaluable because they directly led to publications and launched major parts of their research careers.

Q: Have there been benefits beyond the research itself?

A: There are definitely personal benefits. Both the students and I have had opportunities to spend time in countries we otherwise wouldn't have visited. We've been exposed to different cultures and different ways of approaching science. India was a particularly memorable experience for me, in that respect.

I also think building partnerships with researchers in other countries has value beyond any individual project. On the scale of the Global Seed Funds program as a whole, strengthening those international relationships really matters.

Q: What would you tell other faculty members who are considering applying for Global Seed Funds?

A: If you've really got a plan and you're committed to building the collaboration, you should apply. The program can truly make the difference between an idea and a lasting research partnership. The resulting support can turn those partnerships — and the ideas behind them — into reality.

Energy, prosperity, emissions, and innovation: A decarbonization progress report

Thu, 11/12/2026 - 4:30pm

While rising global prosperity increases energy demand and resulting greenhouse gas emissions, promising low-carbon technologies offer reason for optimism. 

Eric Toone, Breakthrough Energy’s chief technical officer, gave this assessment in his keynote address at the recent Energizing@MIT, the annual research conference of the MIT Energy Initiative (MITEI).

“Energy is how we make things … build and heat our homes … grow and prepare our food. So, energy is prosperity,” said Toone.

Toone oversees the technological roadmap and evaluation at Breakthrough, which is a network of organizations founded by Microsoft’s Bill Gates to accelerate clean energy innovation. Toone also serves as a managing partner of Breakthrough Energy Ventures, the venture capital arm of Breakthrough Energy.

“This correlation between energy use and prosperity persists across a huge group of countries, and importantly, there are no high-wealth/low energy-use countries,” said Toone. Worldwide income parallels electricity consumption as they both climb, according to data shared by Toone. And in the past five decades, the middle class has been expanding worldwide, so prosperity has been increasing. 

Given that 80 percent of global energy use is energy from fossil fuels and 35 percent of electricity comes from coal, that prosperity results in emissions, declared Toone. Moreover, a rise in prosperity induces a set of behaviors that contribute to the production of emissions. We build more, drive more, eat more, and so on. Still, we need to decarbonize, and we know how to do it, Toone said.

“Of course we can decarbonize. It’s not hard to do. The question is: How much will it cost?” One approach is to think about how much it would cost to replace each of our business-as-usual products with equivalent zero-carbon products. Considering all products worldwide, and how much extra we’d pay per ton of carbon eliminated, we can calculate the total cost to the global economy to decarbonize.

But, he said, what it would cost is only half the question. The other half is, what are people willing to pay? While wealthy people in the developed world will say some percent, observations of their behavior — “when nobody’s looking” — show that they won’t pay anything at all. Toone added, “And that’s in the OECD world, where we can actually afford to do it!” OECD is the Organization for Economic Cooperation and Development, a world trade organization.

Breakthrough Energy and its collaborators enlisted artificial intelligence to analyze what it would cost to decarbonize the global economy, and AI estimated it would cost about $7.5 trillion. While that’s a big number, global GDP is about $130 trillion, so it’s about 5 percent of global GDP. But Toone also noted that people — even wealthy people — aren’t willing to pay. So the challenge is figuring out how to do it with zero “green premium,” that is, with no extra charge to use a zero-carbon product.

Breakthrough Energy was created in 2015 to do exactly that. Breakthrough started with a $1 billion venture fund, then raised about $5 billion more. Breakthrough has deployed about $3 billion of that into about 140 companies. Toone described the technologies of some of the companies Breakthrough Energy Ventures has invested in to demonstrate the difference they are making.

  • Luxwall has developed and is now manufacturing novel windows that are as insulating as walls. The windows are sized to fit into conventional frames. One commercial building recorded a 30 percent energy savings as a result of installing the new windows. Toone predicted that this product would be the “new LED bulb. We all swapped out incandescent for LED bulbs because the energy saving was so big and so obvious. This is what's next.”
  • Brimstone is aiming to reduce the 8 percent of emissions that come from making cement. And as the emerging economies come along, they do a lot of building. So, finding a way to decarbonize cement is critical. The company is mining a different source of critical raw materials that contains a lower concentration of the conventional cement feedstock used today, but also yields valuable metals, including aluminum.
  • Ecocem is investigating a recent discovery as it also seeks to reduce emissions related to making cement. Controlling the particle size of the cement enables much more effective packing, thus reducing the total amount of cement needed.
  • CarbonCure Technologies is adding carbon dioxide to cement as it’s being poured to yield cement that’s harder, resulting in a 10 percent reduction in the total amount of cement that’s needed.

Another challenge in the energy transition is how we make and distribute electricity, and how much that costs. Solar and wind are some of the cheapest sources today, but both are intermittent. Breakthrough companies are addressing two ways to solve that problem: First, by moving the electricity from where it’s abundant to places where it's needed; and second, by storing it when it’s abundant so that it's available when it’s needed. Here are some examples:

  • TS Conductor is building transmission wires with a solid core that keeps them from sagging, touching vegetation, and potentially starting fires. “That means I can put up to three times as much energy down an existing transmission corridor,” said Toone.
  • VEIR, a company founded by a MITEI Energy Scholar, is using superconducting cables that can transmit 10 times as much power down an existing corridor.
  • Form Energy, a spinout from MIT, is developing iron-air batteries that can store power for as long as a month or even a couple of months.

The other thing we could do would be to put our solar collectors where there isn’t an intermittency — for example, space. “If I go to space, the sun always shines,” said Toone. The idea of space-based solar has been around for a long time. The challenge is the cost, particularly the cost of “getting stuff into space,” he said. A Breakthrough company called Stoke Space Technologies is building reusable rockets to solve that problem.

Geothermal energy — that is, the energy that's contained as heat in the Earth — is another option for generating electricity. Fervo Energy, another Breakthrough portfolio company, is drawing on advances in drilling technology that came about for fracking. The company is now building the world's first large-scale geothermal power plant in southwest Utah, the Cape Station power plant. The plant is already generating 500 megawatts — continuously, regardless of the weather at the Earth’s surface.

Another source of always-available power is fusion energy, but the main challenge with fusion is containing materials at temperatures equivalent to temperatures in the core of the sun. A physical vessel is not an option. Containing them in a magnetic field is the approach that’s the furthest along. Commonwealth Fusion Systems, an MIT startup and Breakthrough-supported company, is a leader in this approach. There, researchers have created high-temperature superconducting tapes that generate the enormous magnetic fields needed to contain those high-temperature materials.

The potential of all of these technologies, if successful, could reduce the cost to decarbonize the world economy to $3.5 trillion — half as much as the initial estimate, Toone said.

The $5 billion that Breakthrough is investing is only a fraction of the total global investment in clean energy, Toone said. About $220 billion has been invested in clean technology during that same period of time, and venture capitalists have invested about $2.5 trillion, he said. 

The biggest challenge that Toone sees is that the world has an enormous existing infrastructure that has to be turned over. So, it's critical that those new technologies be deployed in the developing and emerging world so there’s not another enormous infrastructure that has to be replaced.

“We can do this! We can absolutely develop new technologies that provide the exact same service, but at zero carbon,” Toone said. “We can absolutely do this.”

Three Lincoln Laboratory technologies receive 2026 R&D 100 Awards

Thu, 11/12/2026 - 4:15pm

Three technologies developed by MIT Lincoln Laboratory are among recipients of 2026 R&D 100 Awards. The awards recognize the 100 most significant innovations that have transitioned to use or been made available for license in the past year.

R&D World, an online publication, manages the awards program, which has run annually since 1963. The worldwide competition is judged by an independent panel of science and technology experts and industry professionals.

The laboratory's winners overcome obstacles to improving the security and capability of systems. One technology is an operating system designed to protect embedded computers from cyberattacks. Another technology is a platform for integrating superconductor and semiconductor chips in a wafer-scale system for quantum or high-performance computing. The final winner is an algorithm that removes interference 5G base stations incur on co-channel radio frequency (RF) devices. 

"These awarded technologies address critical problems in cybersecurity, quantum computing, and communications. We're proud of our dedicated staff who developed and transitioned these technologies for the nation's advantage," says Melissa Choi, Lincoln Laboratory director. 

Since 2010, Lincoln Laboratory has received 111 R&D 100 Awards recognizing the transfer of its unclassified technologies to industry and government. Each year, many technology transitions also occur for classified projects. This transfer of technology is central to the laboratory's role as a Department of War (DoW) federally funded research and development center.

Secure-by-design operating system

Computers embedded in deployed systems — such as satellites, drones, or radars — keep missions intact and people safe. The operating systems (OS) inside these embedded computers, however, are prime targets for cyberattacks. They are large, complex, and built on decades of legacy designs predating modern cyber threats. Increasingly, artificial intelligence is helping adversaries exploit OS vulnerabilities. 

To protect embedded systems, Lincoln Laboratory developed an inherently more secure OS called Magnetite. "If Windows and Linux are the familiar workhorses of general-purpose computing, Magnetite is built for the different realities of embedded systems, for which real-time performance matters, resources are tight, and system failure has physical consequences," says Juliana Furgala, principal investigator in the Secure Resilient Systems and Technology Group. 

A few components assure Magnetite's security. It is built on top of seL4, a formally verified microkernel (i.e., software core). Formal verification means that mathematical proofs have guaranteed the microkernel's code behaves exactly as intended, eliminating a source of security flaws. On top of this foundation, Magnetite’s OS services — such networking, storage, and communication — are written in Rust. This programming language prevents common memory-corruption bugs that hackers often exploit. Finally, Magnetite configures its services as isolated processes under the principle of "least privilege." This configuration prevents different parts of an embedded system from accessing unnecessary data. If a cyberattack compromises one service, that component can be locked down while critical functions of the system continue running.

Free for government use and approved partners, Magnetite is now being integrated into space and flight platforms for the DoW. Looking ahead, the team sees potential to extend the platform to other embedded domains, such as control systems for critical infrastructure. Because Magnetite offers a common, reusable foundation for building specialized mission capabilities, the team is also collaborating with other technical groups across the laboratory to offer Magnetite as an alternative to costly, proprietary operating systems currently in use.

In addition to earning an R&D 100 Award, Magnetite was selected as MIT Lincoln Laboratory's Best Invention of 2025.

Wafer-scale platform for quantum systems

Modern computing systems are increasingly limited not by transistor performance, but by interconnection bottlenecks. Electrical pathways linking densely packed chips are taking up space, slowing data transfer, and generating heat. The problem is especially pronounced in quantum systems, which require bulky coaxial cables to bridge superconducting chips (which must be cryogenically cooled to achieve zero electrical resistance) with room-temperature control electronics.

The Quantum Microsystems platform addresses this interconnection challenge, integrating hundreds of superconductor and semiconductor chips into a single, wafer-scale system. The platform combines several innovations. First, an AI decision-tree approach determines the optimal way to route connections between chips across a wafer. After the wafer is tested for defects, a digital lithography process "draws" custom wiring layers between chip locations and around defective circuit components to enable the highest yield possible from the wafer. Chips are then connected using microbump (tiny "bumps" of conductive material used to connect stacked chips) or bumpless techniques that place them extremely close together; a superconducting fabric extends these connections across the wafer to create a unified computing platform, carrying signals at near light speed. Coaxial cables are replaced with integrated, lithographically defined flexible interconnects, significantly reducing cryogenic wiring overhead and system complexity. 

The Quantum Microsystems platform is designed to be reworkable, so chiplets can be replaced or upgraded without rebuilding the entire system. The platform enables the design and fabrication of systems ranging from individual and stitched‑reticle architectures to mix‑and‑match multireticle configurations and fully integrated wafer‑scale platforms. The platform's AI engine continuously adapts to process variations and design changes to enable efficient, data-driven microsystem manufacturing. 

"By shifting from chip-level scaling to AI-optimized system-level integration, the platform establishes a new paradigm for building ultra-efficient, large-scale computing infrastructure for applications in quantum computing, AI, and high-performance computing," says Rabindra Das, principal investigator in the Quantum-Enabled Computation Group. The technology is designed for transition to government and commercial foundries and may be available for licensing and collaborative development through Lincoln Laboratory’s technology transfer process.

Algorithm to cancel 5G interference 

As 5G cellular networks have expanded, so has an unintended source of interference. In certain environments, powerful signals emitting from 5G base stations can encroach into frequency bands used by nearby non-5G RF antennas, burying signals for such systems as military radars, aircraft altimeters, satellite ground stations, or radio telescopes. 

To cancel this interference, Lincoln Laboratory researchers developed an algorithm that isolates and removes the 5G signal from the affected antenna's signal. Because 5G networks adhere to a published international standard, the canceler can decode the structure of a nearby base station's transmission — and estimate distortion caused by multipath scattering in the environment — well enough to reconstruct a copy of the interfering signal. That reconstructed copy is then subtracted from the incoming data, leaving a clean signal.

Unlike traditional interference cancellation approaches like beamforming, the technique does not depend on spatial separation of interfering and desired signals. Moreover, because it removes interference on each individual antenna channel of the receiver, it preserves the antenna resources that may be needed to handle other types of interference, such as jamming.

The team validated the algorithm using signals captured over the air from a Nokia 5G base station at Hill Air Force Base. The tests recovered a test radar signal buried by the cellular interference. The signal was restored to within two decibels of its original signal-to-noise ratio, translating to 89 percent preservation of the radar's interference-free detection range. 

"While 5G waveforms exhibit complex structure, their generation is governed by global standards that allow our cancellation methodology to generalize across base station radio heads made by a diverse set of vendors," says Binoy Kurien, principal investigator in the Tactical Edge Communications Group. 

The DoW is now evaluating the technology for use on multiple platforms. Because the core structure of the 5G waveform is expected to carry into future cellular generations, this approach may also extend to interference from 6G networks as they emerge.

Pages