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 49 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.

AI helps researchers investigate how thousands of medical studies share code

Thu, 11/19/2026 - 5:15pm

The data and code that underpin a research study can tell us what was actually done, and how. In a medical study, for example, how were patients selected for inclusion? How were missing values handled? How were the results evaluated? 

Access to a study's data and code help researchers to answer these questions, to examine the methods, and to build on the work. 

“Studies involve many choices about how to prepare data, develop models, and evaluate performance. These are sometimes called ‘researcher degrees of freedom,’” says Tom Pollard, a research scientist at MIT's Laboratory for Computational Physiology and Computer Science and Artificial Intelligence Laboratory. “The choices we make affect the results, and a methods section in a paper can only capture so much detail.”

While sharing code can be a first step, in order for code to be understood and reused, it must be sufficiently structured and described. The problem is that, right now, there is little guidance for researchers on what constitutes a well-organized, well-described collection of code. 

This is where an international team led by researchers at MIT and Dana-Farber Cancer Institute sought to help, by examining how code is currently shared in clinical prediction studies. Such studies develop or evaluate models to estimate patient outcomes, such as the risk of developing a disease. Because these models inform decisions about patient care, researchers need to be able to scrutinize how they were built and tested.

Using large language models (LLMs), the researchers examined nearly 4,000 papers to identify statements about shared code. They then followed the statements to assess hundreds of associated code repositories.

Their review revealed two key gaps: Code sharing remains uncommon, and making code public does not consistently mean providing the documentation and supporting information needed for reuse. The open-access findings, recently published in Nature Medicine, are informing TRIPOD-Code, a reporting guideline being developed to clarify what researchers should share.

Following the code

TRIPOD-Code builds on TRIPOD and TRIPOD+AI, checklists that help researchers report how clinical prediction models were developed and evaluated. To understand current code-sharing practices, the team identified studies citing either guideline. The review included almost 4,000 studies that developed, updated, or validated prediction models, and whose full texts were available through PubMed Central’s open-access collection. 

“The challenge was that the information was highly inconsistent. First, we had to identify whether or not a paper shared code” says Thomas Sounack, the study’s first author. He adds that authors “discuss and link to code from a paper in many different ways.”

The team used an LLM to help screen papers, identify statements about code availability, and extract links. Software then retrieved the linked repositories — online collections of code and related files — for further assessment.

The workflow went beyond checking whether a link existed. It also assessed repositories against characteristics such as completeness of documentation, software requirements, and terms of reuse.

“The real benefit of LLMs is their ability to process unstructured information such as text. These models helped us work with information that was scattered across papers and software repositories, rather than neatly organized in a table,” says Pollard, who is a senior author of the study. “That allowed us to examine how code was being shared in much more detail across a large collection of studies.”

Human reviewers evaluated the work of the LLMs. Two researchers independently checked 500 papers and a separate sample of repositories, resolving disagreements before comparing their assessments with the automated results. 

Agreement was stronger for concrete characteristics, such as whether a license file was present, than for judgments about the adequacy of documentation. Because classification errors remain possible, the authors emphasize the overall patterns rather than treating every count as exact.

Inconsistent code sharing

Only about 1 in 8 studies reported sharing code. Where code was shared, the contents of repositories varied widely.

Some repositories were “extremely well-organized and -documented, while others gave almost no guidance at all,” Sounack says. Good code sharing, he explains, means “giving enough information to understand what the code does and be able to run it to reproduce the results reported in the paper.”

Practices also varied substantially across journals and countries. Among journals, the proportion of studies sharing code ranged from zero to more than 70 percent. Some journals with low sharing rates nevertheless had written code-sharing policies, suggesting that policies alone may not be enough to motivate code sharing.

The review also helped to identify examples of effective practice. Some researchers, for example, provided synthetic datasets to assist with code reuse where the original patient records could not be shared publicly.

Setting a standard for code sharing

Researchers who share code are already making an effort to make their work transparent. But clinical prediction research brings together people from medicine, statistics, and computer science, whose expectations for documenting and sharing code may differ.

TRIPOD-Code aims to establish a consensus. Through an expert review process, the team is developing guidance on which elements of shared analytical code should be considered essential.

“We want to give researchers practical guidance on how to share code, and give journals clearer expectations to communicate to authors,” Pollard says. “The review gives us evidence to inform those discussions."

Ultrasound imaging without the gel

Thu, 11/19/2026 - 2:00pm

Engineers at MIT and the Singapore-MIT Alliance for Research and Technology’s Wearable Imaging for Tranforming Elderly Care project (SMART WITEC) have developed a dry and stretchy material that could eliminate the need for a wet and sticky gel in ultrasound imaging. 

Ultrasound technicians typically spread a cold gel onto a patient’s skin before imaging the area with an ultrasound probe. The gel fills in tiny gaps between the probe and the skin so ultrasound waves can travel efficiently through the skin to generate clear images of a patient’s underlying structures. 

Members of MIT’s Zhao Lab are designing wearable ultrasound patches that could provide continuous imaging, at home or on the go. But one sticking point has been the gel. Their previous designs have included a layer of hydrogel — a mostly water-based material that seals up the space between the skin and a small ultrasound probe. But within a day or two, the gel can dry out, along with its ability to transmit sound waves. 

“In a doctor’s office, a technician can simply apply more gel if needed. But for a wearable device, gel-based designs can limit their long-term usability,” says Xuanhe Zhao, the Uncas and Helen Whitaker Professor of Mechanical Engineering at MIT. 

The MIT team has developed a liquid-free alternative to gel that efficiently transmits ultrasound waves, which they report today in the journal Science. Because the material is dry from the start, its performance remains stable over long periods. In experiments, the engineers found that the rubbery material enabled a wearable ultrasound device to generate clear, continuous images over nine days, even during vigorous exercise.

As a demonstration, the team fixed a small wearable patch, about the size of a key fob, to a volunteer’s neck. The patch contained a small ultrasound probe, and the dry and stretchy layer that stuck directly to the skin. The entire patch was connected to a small portable tablet that recorded the patch’s ultrasound images as the volunteer took a 30-minute bike ride, from MIT’s campus in Cambridge, Massachusetts, to Bunker Hill in neighboring Charlestown, and back. 

“We are really aiming to address a material challenge in ultrasound imaging,” says study leader Shu Wang, a postdoc in MIT’s Department of Mechanical Engineering.

“If you don’t have a good interface, that will definitely stop you from developing this wearable imaging field,” adds co-lead author and MIT postdoc Runze Li. “This is a very good solution that enables a longer time for imaging.”

In addition to Zhao, Wang, and Li, the study’s MIT co-authors are Chen Gong and Won Jun Song; additional co-authors are Zihan Tang and Ying Li of the University of Wisconsin at Madison; Chwee Teck Lim of the National University of Singapore; and Xiaodong Chen of Nanyang Technological University in Singapore. Zhao, Chen, and Lim are the co-lead principal investigators of the SMART WITEC interdisciplinary research group.

Molecular detangler

Gels and liquids in general are ideal media through which sound waves, including ultrasound, can travel. Think of sonar, and the calls of whales and dolphins, which can travel much farther in water than air. This is mainly due to the makeup of molecules in fluids. Water molecules, for instance, are small and loosely connected, and can slide easily against each other. As a sound wave passes through, there is little friction between the molecules, allowing the wave to travel without losing intensity. 

Soft solids such as most rubbery and polymer-based materials, however, have more interconnected molecular structures. The same sound wave would dissipate when moving through a solid block, as larger, tangled molecules would generate more friction as they shift in response to the incoming wave. 

The MIT team aimed to make a solid, rubbery material that retained a gel’s ultrasound propagating properties. For this, they took a molecular engineering approach. 

“Hydrogel is mainly water, with just a bit of polymer, whereas solid materials like rubber are almost 100 percent polymer molecules,” Wang explains. “We tried to reduce the friction between polymer chains in our material, without adding liquid or lubricant.”

Polymer materials such as rubber and elastic are made from microscopic molecules that chemically bond with other similar molecules to form long polymer chains. In a solid material, these chains form a dense thicket and do not slide easily against each other. 

In looking for ways to reduce the friction between polymer chains, the engineers landed on a “bottlebrush” structure — a molecular configuration that has been explored in other applications including drug delivery and light-reflecting materials. A bottlebrush molecule consists of a main backbone of molecules, with smaller side chains of molecules sprouting away, like the bristles of a pipe cleaner. 

“These hairy extensions don’t easily penetrate other bottlebrush molecules,” Wang notes. “They can easily move around at the ultrasound frequencies, and that’s why there is less friction between them.”

With that reasoning, the team came up with a protocol to fabricate a polymer material made entirely from a molecular bottlebrush architecture. Their first attempts, however, did not quite hold up. 

“The first samples were very soft and crumbly, like tofu,” Wang says. “If the samples were too large, they would not hold their own weight, and if you picked one up, it would just break.”

They quickly found they could reinforce the material by chemically fusing it with a second, loosely connected network of bottlebrush molecules. This additional network essentially held the first, crumbly network of bottlebrush polymers in place, without affecting their low-friction properties. The result was a tougher, rubbery material that the researchers were able to stretch repeatedly without tearing. 

“It’s very much like skin,” says Wang. “It’s very soft but will hold together if pinched or pulled.” 

Imaging workouts

The team fabricated credit card-sized samples of the new material and first tested its ultrasound properties on a “phantom” model — a synthetic structure that is made from materials that mimic a human’s skin and internal structures. In these tests, the material efficiently transmitted ultrasound waves from a probe, through the model, to image the structures underneath. The team kept the setup running continuously and found it kept imaging clearly for around nine days. 

“In principle, this material could go forever, depending on how stable it is,” Wang notes. “The sound properties stay the same. And that was the moment when we thought, oh, this material could work.”

To test the material’s mechanical performance, the team adhered a patch to a volunteer’s chest as they ran on a treadmill. For comparison, the volunteer also wore a sample of a conventional silicone rubber. Throughout a 10-minute run, the silicone rubber became water-logged with sweat, whereas the new material wicked away any sweat buildup. These tests suggest that the new material can hold up over strenuous exercise. 

Finally, the researchers tested the material in the field, attaching it to a small ultrasound transducer applied to a volunteer’s neck. Throughout a 30-minute bike ride that traversed steep inclines, the device maintained continuous and accurate ultrasound imaging of the rider’s blood flow. 

Going forward, the team is working to streamline the fabrication of the material. Then, they hope to enable wearable, long-lasting ultrasound imagers that can be used to track athlete performance and monitoring patient health outside of clinical settings.

“In the hospital, a doctor can replace gel if it dries out,” Li offers. “But if we want ultrasound imaging that goes beyond the clinical scenario, we need a material that lasts longer, and is easier to handle. And we think this really works.”

This work was supported, in part, by the Singapore-MIT Alliance for Research and Technology and the MIT Health and Life Sciences Collaborative.

Scientists create the first whole-body single-cell atlas of an adult vertebrate

Thu, 11/19/2026 - 12:00pm

Viewed from above, Danionella cerebrum offers a rare sight in a vertebrate: its brain, visible through the top of its head. Since the species was formally described in 2021, the tiny, transparent fish has drawn rapidly growing interest for its unique biology.

Scientists at MIT, the Whitehead Institute for Biomedical Research, and Charité–Universitätsmedizin Berlin recognized another opportunity in the tiny fish: It brought whole-animal single-cell profiling within reach, allowing the team to create the first comprehensive map of cell types and gene activity across an adult vertebrate. That perspective could help reveal how cells in different tissues coordinate their responses — and how changes in one part of the body influence others.

In a study published in Science on Oct. 29, the team profiled more than 600,000 cells from across the fish to build the whole-body map of which genes are active in which cell, known as a single-cell transcriptome atlas. A cell’s transcriptome is the full set of RNA molecules it has made from its DNA at a given time, providing a snapshot of which genes the cell is using.

Together, the data provide a molecular reference for the cell types found throughout Danionella, the patterns of gene activity that distinguish them, and where many of them are located. The atlas also gives researchers a baseline for exploring how injury, aging, drugs, or other changes affect cells across the whole animal.

“Once you have a spatially resolved atlas, you know what cell types exist, but you also know the neighbors of those cell types in this context,” says Kutay Deniz Atabay, who led the work as a postdoc in the lab of Peter Reddien, a Whitehead Institute member and professor of biology at MIT. “This gives you access at multiple different levels of complexity, from molecular to cellular to systems.”

Cell atlases have become increasingly detailed, but extending them across a whole adult vertebrate has been difficult for a simple reason: size. Profiling the cellular diversity of larger animals comprehensively is technically demanding and expensive. Danionella offered a unique way around that problem. Its miniature body — barely 1 centimeter long — makes the experiment possible, while its close relationship to zebra fish allows researchers to draw on genetic tools developed for one of biology’s most widely used model organisms.

Atabay first became interested in Danionella in early 2019, while searching for a small, transparent animal in which he could watch the nervous system regenerate. Its miniature size and accessible brain made it especially promising for that work, and the researchers found that the fish has broad regenerative capacity, including in the brain and fins. Reddien’s lab, meanwhile, had already seen the value of a comprehensive whole-animal cell atlas through its work on planarians, regenerative flatworms whose small size had made such an atlas possible. Together, those interests led to a larger question.

“Deniz and I had the thought: Would it be possible to get a cell-type transcriptome atlas for a complete vertebrate?” says Reddien, who is also an investigator with the Howard Hughes Medical Institute. “Here’s an animal where we’d have a chance.”

To build the atlas, the researchers used single-cell RNA sequencing to measure gene activity in individual cells collected from across the brain, spinal cord, eyes, internal organs, fins, head, and trunk. They paired those data with spatial methods that mapped cell types and patterns of gene activity back to their locations in tissues. The resulting atlas was extensive enough to capture even exceptionally rare cell populations.

Looking across the animal revealed patterns that can be difficult to see when tissues are studied separately. In the nervous system, neurons from the same region often had more similar patterns of gene activity — even when they used different neurotransmitters, the chemicals neurons use to communicate — than when compared to neurons using the same neurotransmitter in different regions. Similar patterns have been observed in other vertebrates, including mammals, suggesting that regional and developmental identity may be an important organizing feature of vertebrate nervous systems.

The atlas also revealed signs that adult cells retain a molecular memory of where they belong in the body. Fibroblasts, a type of connective-tissue cell, maintained region-specific activity of genes best known for establishing different parts of the body during development. Cells in the pectoral and pelvic fins, for example, retained gene programs associated with forelimb and hindlimb development in other vertebrates. What these persistent signals do in adult Danionella remains unknown, but they may help tissues maintain or rebuild the appropriate structures. In planarians, positional information provides molecular guidance that helps regenerating tissues rebuild the correct structures.

The research also used the atlas as a reference for studying Danionella’s remarkable regenerative abilities. After injuring the telencephalon, a region of the forebrain, the missing neurons regenerated and the researchers traced a progression from an early wound and immune response to neural stem-cell activity and the production and maturation of new neurons.

The team has made the atlas publicly available through DigiFish, an interactive resource for exploring cell types and gene activity throughout the animal. Paired with Danionella’s transparency and growing experimental toolkit, the atlas can help researchers identify specific cells and genes to study or manipulate and then follow their activity in the animal.

For Reddien, the advance presents numerous, diverse opportunities for new science. “There’s been this era of genome sequencing, where you get the complete list of all the genes that it takes to make an organism,” he says. “The transcriptomes of cells tell you how that genome is being utilized in the different cell types. To have the transcriptome of every cell type seems like a big turning point in how one would investigate the biology of an organism.”

If tiny “primordial” black holes exist, here’s how we might find them

Thu, 11/19/2026 - 10:00am

Scientists have hypothesized that, in the fraction of a second following the Big Bang, minute black holes may have been forged within tiny, dense pockets that collapsed under their own gravity. If these primordial black holes are ever found, they could represent a significant component of dark matter — the invisible substance that makes up most of the universe’s matter. 

Now physicists at MIT and elsewhere have worked out a way that astronomers could detect a primordial black hole. The team carried out theoretical calculations and three-dimensional simulations to show what would happen if a lingering primordial black hole were captured by a modern, sun-like star. 

In a study appearing today in the Astrophysical Journal, the team’s simulations reveal a dramatic evolution: When an atom-sized primordial black hole with the mass of an asteroid is caught up in the gravity of a sun-sized star, it should quickly fall into the star’s core. Then, it would begin to devour the star from the inside. 

If the star spins sufficiently fast, the simulations show that a disk of its own stardust should begin to circle as it falls into the black hole. 

“Once the disk forms, this is essentially a death sentence for the star,” says lead author Ore Gottlieb, an assistant professor of physics at MIT. 

Within a matter of seconds, interactions between the black hole and the spinning disk should launch powerful jets of matter that ultimately pierce through the star, making it explode.

That dramatic fate, the researchers calculated, would generate a flash of blue and ultraviolet light, along with gamma- and X-rays. This flash should be bright enough to detect with current telescopes.

“There is hope for us to probe primordial black holes through interactions with stars,” Gottlieb says. “We expect this to be visible, and it’s something we should look into.”

The study’s co-authors are Matteo Cantiello of the Flatiron Institute, Cameron Norton and Matthew Kleban of New York University, and Ken Van Tilburg of Stanford University. 

Quantum quarry

Today, the smallest and lightest primordial black holes would have evaporated away due to Hawking radiation — a quantum effect that was proposed by the physicist Stephen Hawking to describe how black holes can radiate energy and shrink, very slowly, over cosmic time. 

Larger, heavier primordial black holes, however, should still be around, not having had quite enough time to disappear. These lingering relics would be about the size of an atom, and as massive as an asteroid (or about a millionth of the mass of the Earth). Gottlieb and his colleagues wondered: If asteroid-mass primordial black holes indeed exist, how might they interact with stars today? What’s more, could this interaction be a way to detect primordial black holes, and confirm their existence? 

In their new study, the researchers carried out analytic calculations and numerical simulations based on equations of general relativity, hydrodynamics, stellar evolution, and particle physics, to determine how an asteroid-mass primordial black hole could be captured by a star and what events might unfold from there.

They discovered that the gravitational pull from a single star would not be enough to trap a passing primordial black hole. However, the researchers found that gravitational interaction with a companion, such as a second star or planet, could efficiently trap a primordial black hole. 

In this three-body configuration, the team calculated the friction that the black hole would experience as it passes through the stellar material. They found that repeated passages could gradually shrink its orbit, and over a span of thousands to millions of years — nearly instantaneous on cosmic timescales — the black hole would fall into the center of the star. 

“Then, this is where the action begins,” Gottlieb says. 

A critical pull

The team modeled the black hole evolution within the star. How quickly the black hole grows depends on how much energy the black hole releases back into the surrounding star as it heats the gas and slows further infall. 

They found that most of the energy is carried into the black hole, allowing the black hole to grow faster than previously thought. Within several million years, the black hole would have consumed about 1 percent of the star’s mass. If at this critical point, the star spins fast enough, its gas can no longer fall directly into the black hole and instead piles up in a rotating disk of material around a few-meter-sized black hole at the heart of the star. 

“Inside the star, you have a black hole with a disk,” Gottlieb says. “The black hole is also spinning, and it’s twisting space and time.”

As the black hole continues to consume the disk’s spiraling material, it twists and pulls on the disk’s magnetic field.

“This twisted magnetic field will power very strong electromagnetic jets from the poles of the black hole,” Gottlieb explains. “And these jets will just explode the star, within a few dozen seconds. It’s very fast. Once you have a disk, you are done.” 

Finally, the team estimated the color and intensity of light that such an explosion could generate, from start to finish, across the electromagnetic spectrum. 

“We would initially detect some X-rays and gamma rays, then over time optical and infrared emissions, and later on radio emissions,” Gottlieb says. “This is something we could potentially detect, if it’s not happening too far from us. It could be many galaxies away, not just in our backyard.”

He points out that other astrophysical phenomena could produce a similar flash. If astronomers detect such a flash, they will have to combine the signal with other clues to confirm that the source is a primordial black hole. 

“For instance, if the star is a certain mass that is not supposed to explode, it must be something that is happening inside that is a sort of parasite that is consuming the star from the inside,” Gottlieb says. “The broader impact is that we can now place constraints on primordial black holes, and their contributions to dark matter, using observations.”

This work was supported, in part, by the U.S. Department of Energy Office of Science Advanced Scientific Computing Research Program, and by the National Science Foundation.

Study reveals data centers’ long-distance damage to air quality

Thu, 11/19/2026 - 9:00am

The rapid expansion of data centers for artificial intelligence causes negative health and air quality impacts in regions far from where the brick-and-mortar facilities are located, according to a new study by MIT researchers.

The team estimates that additional air pollution from the electricity demands of data centers in 2023 resulted in about 650 premature deaths across the United States. 

Some Midwestern states had among the highest estimated rates of premature deaths per capita attributable to data centers, even though much of the data center electricity demand was concentrated in Virginia and Texas.

The researchers built a modeling framework that captures how the electricity demanded by data centers propagates through the nation’s interconnected power grid. The model also considers how atmospheric conditions impact the formation of air pollution in different regions.

The framework identifies power plants across the country whose emissions increase in response to data center electricity demand. In parts of the Midwest, these emissions combine with atmospheric conditions that favor the formation of fine particulate matter, a pollutant that can cause heart attacks, stroke, and lung disease.

The researchers say these results underscore the importance of considering unintended regional health impacts of data centers, and how the location and operations of these facilities affect public health on a broader scale.

“When people think about the negative impacts of data centers, they usually focus on the neighbors. But our study indicates that the air quality impacts of data centers are very widespread across a large region. Even if you don’t think you are affected because there is not a data center near you, think again,” says Noelle Selin, a professor in the MIT Institute for Data, Systems, and Society (IDSS) and the Department of Earth, Atmospheric and Planetary Sciences (EAPS); director of the Center for Sustainability Science and Strategy; and senior author of the study.

Selin is joined on the paper by Yuang (Albert) Chen, an MIT graduate student; as well as others at MIT, the University of California at Irvine, and Yale University. The research appears today in Environmental Research Letters.

Capturing regional impacts

The Lawrence Berkeley National Laboratory estimated that electricity demands of data centers represented about 4 percent of total U.S. energy demand in 2023 and resulted in about 4.4 percent of the nation’s overall emissions. 

That proportion has likely increased as new data centers continue to be built at a rapid pace.

Many prior studies estimate impacts using average electricity-grid emissions. Like those studies, the researchers assume data centers draw power to run computing tasks 24/7. Their approach also identifies which generators respond to that additional demand hour by hour, allowing them to trace where the resulting emissions and health impacts occur, Chen explains.

“Generators that respond to additional electricity demand are often natural gas plants, although the response varies by place and hour,” he says.

To drill down on regional health impacts, the MIT researchers developed a modeling approach that enables them to identify which generators are contributing to data center electricity demand within each modeled region. 

This allows them to incorporate a spatial component into the analysis.

They built an electricity system model that simulates how much electricity each generator produces and pumps into the grid every hour. This model also considers how that electricity is transmitted between regions. 

It gives the researchers data on emissions of nitrogen oxides (NOx) and sulfur dioxide (SO2) for each generator across the United States. NOx contributes to the formation of ground-level ozone, or smog, while both NOx and SO2 contribute to the formation of fine particulate matter (PM2.5) in the atmosphere.

“It was a challenge to gather all the data,” Chen says. “There are more than 10,000 generators in the U.S. grid, and we needed to make sure the data provided for each generator is valid and validated across other data sources.”

The researchers incorporated these data into a chemical transport model. This model simulates how pollutants form and travel through the atmosphere across regions. 

From there, they translated the additional air pollution from data center operations into premature mortality rates using evidence from epidemiological literature.

The model estimates that emissions from data centers caused about 650 premature deaths in 2023 (513 from PM2.5 and 137 from smog), which represents about 5.3 percent of the nation’s total electricity-related mortality. 

“While it is not surprising that any additional source of air pollution will result in mortalities, with data centers, when and where these deaths are occurring are different than you might expect,” Selin says.

A pollution mismatch

The premature mortality rates were highest in Iowa, Nevada, South Dakota, and Kansas, but the bulk of the electricity demand was concentrated in northern Virginia and in Texas. Premature mortality rates in Virginia and Texas are only about half as much as most Midwestern states.

One reason for this mismatch is the interconnected nature of the power grid, Chen says. 

Generators that respond to the increased electricity demand in Virginia are likely located in the Ohio Valley, where there is a higher proportion of older power plants that generate more emissions from burning coal and natural gas. 

“Different regions have different mixes of power. If you turn on a major new demand source, by plugging a data center into the grid, you might have to fire up your really dirty power plant to meet that new demand,” Selin says.

Plus, due to atmospheric conditions, many pollutants generated in these regions are transported to downwind states.

Midwest states like Nebraska and Iowa are traditionally agricultural regions. Additional ammonia in the atmosphere from fertilizer enhances the formation of PM2.5. The estimated health impacts also extend to Kansas, South Dakota, Missouri, and Arkansas, illustrating that the effects can reach states with relatively few data centers.

The study shows that rural populations are expected to have significantly higher mortality risk from data center electricity demand than urban populations, Chen says.

“When people are thinking about data center siting and operations, it is important to consider more than just things like land and energy prices. They should also take air quality considerations into the equation,” he says. “This is a good baseline to think about where we should build data centers and how we should supply those data centers with electricity.”

In the future, the researchers want to consider air quality impacts of onsite generation — a growing trend where data centers incorporate power plants that supply electricity. They could also study the flexibility of data center operations, and how adaptively adjusting electricity demand could impact air quality. 

This research was funded, in part, by the MIT Energy Initiative Society of Energy Scholars and the MIT Martin Family Society of Fellows for Sustainability.

Confirmed: Numbers tell, but language sells

Thu, 11/19/2026 - 12:00am

Suppose you are in charge of your firm’s commuting policy and, after some research, you determine that 57 percent of your colleagues drive to work. How would you communicate that number to co-workers?

You could just tell them the number. Or you could characterize it in words. If you want to imply that 57 percent is a large number, you could call it “a clear majority,” or even “approaching two-thirds” (roughly 67 percent). However, if you seek to minimize 57 percent, you might call it “barely over half.” 

Either way, when people are motivated to persuade their audience that a number is particularly high or low, they are considerably more likely to use language to describe the data, rather than simply airing the original numbers. 

Now a new study co-authored by an MIT scholar quantifies this phenomenon: People motivated to persude others about statistical facts are 25 percent to 29 percent more likely to use words describing the numbers in question, instead of the numbers themselves. That is a large enough increase to be termed “substantial,” “significant,” or any number of other words.

“If I’m a well-informed expert and I have incentives to persuade my audience, I’m relatively more likely to choose to use language,” says co-author Victor de Chaisemartin, a doctoral candidate at the MIT Sloan School of Management. 

The language that experts choose when seeking to persuade others is also more slanted than their numeric communications. As the authors note in the paper, people appear to be leveraging the imprecision of language in ways they feel are palatable: Whereas reporting a number incorrectly would be precisely false, it may be more psychologically comfortable to slant a message using language.

And that’s just the half of it. The study also shows that audiences find language-based descriptions of numbers convincing — even when they are told there is an agenda behind them. 

“Language messages are particularly persuasive,” de Chaisemartin says. Telling an audience they are being spun, he notes, “actually doesn’t do very much to mitigate this. That’s interesting. It makes you wonder what’s going through people’s heads, and how they are dealing with information.”

The paper, “Numbers Tell, Words Sell,” is being published in the American Economic Review. The authors are de Chaisemartin; Michael Thaler, a lecturer in economics at University College London; and Mattie Toma, an associate professor at the University of Warwick. 

Floating in data, reporting it in words

As de Chaisemartin describes it, the study’s origins come from a basic observation.

“The world has loads of statistics floating around in it,” he says. “But often if you pick up a newspaper or read something in the media, or if you’re reading a corporate report, you don’t usually see 50 statistics. You’re often going to see a lot of quantitiave data reported in words.”

Indeed, one study of media practices has found that 22 percent of news items include statistics, but that 24 percent of those stories only use words to describe the numbers. Another study of academic research has found that just 37 percent of papers include numbers in their abstracts. 

To conduct the study, the researchers conducted two separate but parallel experiments. In the first one, using the research platform Prolific, 1,000 members of the general public were divided into 500 pairs of senders and receivers of information, communicating the results of simple games and observations. In the second experiment, set up through the University of Warwick’s Policymakers Lab, 146 scholars from academia were paired with policymakers from five countries — primarily central government officials.

Ultimately, the experiments show that when incentivized for accuracy, senders of information choose language over numbers only 13 to 14 percent of the time; that is, experts generally prefer numbers for communicating accurately. But faced with incentives to persuade, the likelihood of senders choosing language increases by 25 to 29 percentage points. 

“It’s a clear-cut hypotheses, it makes sense, and the results bear it out,” de Chaisemartin says.

Applying the findings

While the study is specific in nature, it contains some broader implications. For one thing, de Chaisemartin notes, the findings can help organizations recognize the value of constructing consistent usage standards. Climate scientists, he notes, do this already regarding some forecasts. Security officials, medical researchers, doctors, and scholars may also have good reasons to adopt similar characterizations of likelihoods and other things. 

“There are ways to think carefully about how we set up and report empirical data,” he says. “This could be viewed as one piece of information pushing in that direction.”

De Chaisemartin also suggests that the findings bear on the issue of media literacy generally, and could be used to help people analyze the statistical claims they read in the news. For his part, he is also interested in follow-up questions about what happens when people send information to the same audience repeatedly. 

“That might introduce interesting new mechanisms in this process,” de Chaisemartin says, noting that senders of information might have additional incentives to protect their reputations for accuracy, while audiences could become more discerning over time. 

To be sure, there are many subtleties involved in descriptions of statistics. For example, given that most drivers wear seatbelts, a policy intervention raising the usage rate by two percentage points might not seem huge in absolute terms, but could represent a substantial increase within the remaining room for improvement. In other circumstances, a change of two percentage points might just be a small fluctuation in a sea of data.  

Language descriptions could therefore help contextualize numbers or aid audiences in understanding statistics. However, the present findings suggest caution when interpreting language, since it may also signal that the expert was trying to persuade their audience.

In any event, de Chaisemartin expresses hope that public discourse involving statistics can become consistently healthy and accurate. 

“There’s a space for language and numbers to be complementary, and to provide people with both,” he says.

The research was supported, in part, by the British Academy/Leverhulme Small Research Grants, and the University of Warwick Behaviour Spotlight Small Grant Scheme and Primary Data Funds.

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.

Pages