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

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

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

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

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

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

Overcoming the limits

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

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

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

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

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

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

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

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

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

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

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

Leveraging magnetism

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

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

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

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

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

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

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

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

This summer, the MIT Schwarzman College of Computing welcomed faculty from colleges and universities across Greater Boston, South Carolina, West Virginia, and Texas to campus for the inaugural AI Educators Pilot, a weeklong workshop aimed at expanding how artificial intelligence is taught across disciplines and learning environments. 

Inspired by MIT class C01/C51 (Modeling with Machine Learning), a course developed through the Common Ground for computing and AI education that focuses on helping students understand and apply foundational AI and machine learning concepts to problem-solving in their own disciplines, the workshop gave educators an opportunity to explore how its materials and teaching methods could be adapted for their classrooms. 

“The broader goal is to expand AI education to more students by investing in training for instructors,” says Dan Huttenlocher, dean of the MIT Schwarzman College of Computing and the Panasonic Professor of Electrical Engineering and Computer Science (EECS).

“We want to empower students to become critical thinkers about AI, not just users of the technology,” says Asu Ozdaglar, deputy dean of academics for the MIT Schwarzman College and department head of EECS.

A collaborative model for expanding AI education

Bringing the program to life required broad collaboration across the college, including support from leadership, staff, and contributions from more than half a dozen instructors in fields ranging from finance and computer science to sustainability. Together, they helped shape a workshop that paired core technical concepts with examples and teaching materials adaptable to a range of classroom settings.

“I have not seen an effort quite like it — this many dedicated instructors assembling materials of this richness, all to equip the educators who serve their students,” says Saurabh Amin, the Edmund K. Turner Professor in Civil Engineering and faculty director of the AI Educators Pilot. Amin is also co-director of the Operations Research Center, which is jointly housed within the MIT Schwarzman College of Computing and MIT Sloan School of Management.

With support provided by Jake and Robin Reynolds, the pilot brought together 19 participants in July from Allen University, Babson College, Brandeis University, Marshall University, the University of Massachusetts at Lowell, the University of North Texas, and Wentworth Institute of Technology. Working alongside MIT faculty and instructors, participants explored the pedagogy behind Modeling with Machine Learning through a mix of demos, videos, and exercises, and collaborated in hands-on activities focused on translating the course’s materials and methods to their own classrooms.

“This opportunity has been very timely because we are starting an AI and data science program in my department,” says Wenjin Zhou, assistant professor of computer science at UMass Lowell. “We’ve already been thinking about: How do we teach our next generation of computer scientists within the area of AI? How do we integrate AI in the teaching? I wanted to learn more about how other people are doing it, and especially answer the question: If AI can create tools for anyone now, what does a computer scientist do?”

Moving beyond the black box

When it comes to AI, Amin notes, there is no shortage of high-quality material. What is usually missing is context: Opportunities for instructors and students to connect AI concepts to specific disciplines, problems, and ways of thinking. Those connections are often built through dialogue and reasoning, rather than by presenting AI as a fixed set of ideas to be received. But instructor capacity remains one of the scarcest resources.

“What is scarce are educators prepared to teach AI as more than a fixed body of concepts and tools, to ground it in their own field, help students use it with judgment, and demystify it, so students do not just apply models but learn to question, adapt, and build with them,” explains Amin.

Shen Shen, an EECS lecturer and one of the workshop instructors, adds, “How do we make sure that machine learning is not just a black box, nor this magic piece of new technology? You can think of it as a tool, or a new framing to help you solve the problem in your specific domain.”

From pilot workshop to educator network

Participants ended the week by reflecting on which workshop materials and teaching approaches they planned to adapt for their disciplines and courses. Their feedback will help shape future iterations of the pilot and support the development of a broader network of educators committed to expanding AI education across diverse learning environments.

Weijie Pang, an assistant professor of computer science at the Wentworth Institute of Technology who attended the workshop, looks most forward to ongoing community building activities. “This is a really valuable opportunity to communicate with other faculty from different majors and areas. I can see what other universities are doing and what we can learn from each other,” she says.

“It's helpful to know that everybody within different disciplines at different universities is struggling with the same questions of how we can best serve our students as the technology is changing. Hopefully, we can set them up for success by being a little bit more forward and anticipatory of what the AI use is going to be,” says Dylan Cashman, an assistant professor of computer science at Brandeis University.

The brain cancer glioblastoma is one of the most aggressive and treatment-resistant cancers known to medicine, carrying a median survival of just 12-15 months, even with the best available care. Now, researchers at the MIT Media Lab have developed injectable nanoantennas, each about one-hundredth the width of human hair, that can be magnetically activated to create localized therapeutic electric fields that target and kill brain cancer cells without damaging healthy brain tissue.

“In laboratory and animal studies, this approach significantly reduced tumor growth and extended survival without detectable side effects, highlighting its potential as a precise and safe brain cancer therapy,” says Deblina Sarkar, associate professor and AT&T Career Development Chair at the MIT Media Lab and head of the Nano-Cybernetic Biotrek group.

The researchers named their technology “HITMAN” — short for highly-localized electric-field-induced tumor therapy using magnetically actuated nanoantennas. 

An open-access paper describing this technology published today in Science Advances.

To test HITMAN against the most clinically realistic version of this disease, the research team worked with tumor tissue obtained from patients diagnosed with aggressive and chemotherapy-resistant glioblastoma at Mayo Clinic. Using cells derived from this tissue in the laboratory, the researchers demonstrated that HITMAN eliminated 52.2 percent of these drug-resistant cancer cells — more than five times than that achieved by the standard chemotherapy drug temozolomide (TMZ) — while leaving healthy neurons and brain-supporting astrocytes unharmed.

The team then implanted those patient-derived tumor cells into the brains of mice to recreate the disease in a living system. In these orthotopic animal models — widely regarded as the gold standard for preclinical brain tumor research — HITMAN substantially inhibited tumor growth, extending median survival by more than 50 percent with no detectable toxicity to major organs or surrounding healthy tissue. 

The injectable nanoantennas can be activated wirelessly from outside the body, with the application of a low-frequency (no higher than 200 kHz, to prevent tissue-damaging heat) magnetic field that can penetrate the skull and brain tissue. The magnetic field actuates parts within the nanoantennas made of magnetostrictive material, creating stress and strain, which result in deformation of a piezoelectric film, producing localized electric fields.

Such localized electric fields were demonstrated to preferentially attack glioblastoma at the cellular level, disrupting the cells’ inherent bioelectric currents and fields, which regulate cellular function. Such disruption provoked a number of antitumor mechanisms, including protein unfolding, membrane damage, and endoplasmic reticulum stress, curtailing the production of a cell’s functional proteins. Such forms of cell dysfunction led to cell death. According to the researchers, cancer cells were selectively targeted over healthy cells due to their high proliferative rate, which elevates protein-folding demand, as well as their characteristic abnormalities in membrane composition and intracellular organelles.

Among a wide array of control experiments, the researchers also exposed glioblastoma cells to the nanoantennas without applying a magnetic field, as well as exposing the cancer cells to a magnetic field alone, confirming that the demonstrated effects were in fact due to the nanoantennas and their magnetic field activation. They also tested for side effects damaging to the animal models’ major organs — kidneys, liver, spleen, lungs, and heart — and detected none.

Also demonstrated by the research was a significant reduction in the number of cancer cell colonies formed after application of the nanoantennas, from 112-150 in the control groups to just 26 in the experimental group, indicating significant potential to reduce tumor recurrence and metastasis.

If translated to clinical use, the nanoantennas, whose size is approximately 150 nanometers, could be injected through the skull. Sarkar points out, however, that a technology developed previously in her lab could make their deployment even simpler.

In 2025, Sarkar and her colleagues created “circulatronics,” a technology that could allow devices like the HITMAN nanoantennas to be administered through an injection in a patient’s arm and to travel to a target region of the brain. In that previous work, the electronic devices were integrated with living cells so they would not be attacked by the body’s immune system and could easily cross the blood-brain barrier, as was demonstrated in pre-clinical studies. 

A glioblastoma diagnosis comes with formidable treatment challenges. Because this type of cancer is extremely infiltrative, complete tumor removal is difficult to achieve and can affect cognitive function. Also, the tumors often resist radiotherapy and chemotherapy, and immunotherapy is challenged by an immunosuppressive tumor environment. 

“The persistent failure of these therapies underscores the urgent need for novel approaches to target treatment-resistant glioblastoma cells,” the researchers write. “HITMAN offers a minimally invasive, spatially precise, and clinically translatable therapy for glioblastoma.”

Sarkar is joined on the paper by other members of her lab, including Monochura Saha, a former MIT postdoc; Ishaq Khan, a former MIT senior postdoc; Baju Joy, Shun Ying Chen, Hao-Tung Yang, Preet Patel, and Pengrui Zhang, all MIT graduate students; and Faheem Azeemi, an MIT undergraduate student.

During the day, waste products such as lactic acid and worn-out proteins build up in the brain. When we sleep at night, waves of cerebrospinal fluid (CSF) help to wash away this waste, keeping the brain healthy.

In a new study, MIT researchers have shown that they can strengthen these CSF waves through exposure to short bursts of a gentle, staticky sound known as “pink noise” during sleep. These bursts increase the amplitude of slow electrical waves in the brain, which then enlarges the CSF waves.

The researchers now hope to explore whether this enhanced CSF flow could help to boost cognitive function, improve memory, or even slow the progression of neurodegenerative diseases caused by the buildup of harmful proteins such as amyloid beta.

“We found that we were able to increase the size of the CSF flow wave during sleep, which as far as we know, there hasn’t been a method to do before. Now that we can enhance CSF flow during sleep in healthy adults, we’re really excited to bring this technology to clinical populations to see what effects we can have,” 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. 

Lewis is the senior author of the study, which appears today in Science Translational Medicine. Joshua Levitt, who recently earned his PhD from Boston University and was a visiting graduate student in Lewis’ lab, is the paper’s lead author.

Cleaning up the brain

Cerebrospinal fluid is a clear liquid that surrounds and cushions the brain and spinal cord. In addition to protecting the brain from injury, it also helps provide nutrients such as glucose and removes waste products secreted by brain cells as they burn energy.

In 2019, Lewis reported a way to use functional magnetic resonance imaging (fMRI) to measure CSF waves as they flow in and out of the brain during sleep. That study showed that these waves are tightly coupled with brain waves called slow waves, which are associated with deep sleep.

In the new study, she wanted to further explore the relationship between brain waves and CSF flow, and investigate whether manipulating brain waves might enhance CSF flow. Previous work had already shown that delivering an auditory stimulus at the peak of slow waves can deepen the waves.

“You can make more of these electrical slow waves through an auditory stimulus, if it comes at just the right time. Similar to a child on a swing, if you push them when they’re at the right moment in their movement, you can make that swing go farther,” Lewis says. “The challenge is: How do you find just the right time?”

The auditory stimulus used for this study is a 50-millisecond burst of pink noise. Similar to white noise, pink noise contains all sound frequencies audible to the human ear, but the lower pitch frequencies are louder and the higher pitch frequencies are softer. This creates a balanced, gentle sound similar to steady rain or a distant waterfall.

To deliver these bursts at the peak of the brain’s slow waves, the researchers had to measure each participant’s EEG activity as they slept. This proved challenging because they also needed to measure fMRI signals to monitor CSF flow, and the magnetic fields used for fMRI interfere with EEG signals.

To overcome that, the researchers developed a way to process the EEG signals to eliminate the noise caused by fMRI, very rapidly — in less than 100 milliseconds. To make up for that small lag time in the EEG measurement, they also developed an algorithm that could predict when the slow wave peaks would occur. This allowed them to deliver the pink noise stimulus at the correct time.

More restorative sleep

In tests of 14 healthy volunteers, the researchers found that the auditory stimulus they delivered — which is not loud enough to wake a sleeping person — increased the amplitude of both the slow electrical waves and the CSF waves, during sleep.

Their fMRI studies also revealed that the slow waves stimulate blood vessels to constrict and dilate, allowing them to act as a pump that drives CSF out of the brain. Slow waves are seen only during non-REM sleep, and they become more prominent in deeper stages of sleep.

The researchers now hope to study whether enhancing CSF flow could help people to get more restorative sleep, especially people with insomnia. They also plan to explore whether increasing the flow of CSF, and the removal of waste products from the brain, could help people with Alzheimer’s and other diseases characterized by buildup of harmful proteins. 

“Brain waste clearance is really important for Alzheimer’s and other forms of dementia, which are caused, in part, by the buildup of molecules like amyloid and tau in the brain. If we can improve brain waste clearance, we may be able to help prevent the buildups of these plaques that lead to disease,” Levitt says.

Levitt has started a company that hopes to develop a device, such as a headband, that people could use at home to increase CSF flow by delivering an auditory stimulus at the right time. 

The research was funded by a McKnight Scholar Award, a Sloan Fellowship, a Pew Biomedical Scholars Award, the Simons Foundation Collaboration on Plasticity in the Aging Brain, the MIT EECS Transformative Research Fund, the National Institutes of Health, the Corundum Convergence Institute, and the Panasonic Well Fellowship for AI and Wellness.

As a liquid that is easily stored and transported, ammonia (NH3) is an attractive carrier for hydrogen, which is used in fuel cells, semiconductor manufacturing, chemical processing, and other applications. However, breaking ammonia into hydrogen and nitrogen typically requires high temperatures, and the resulting gas mixture must undergo additional purification before the hydrogen can be used in many applications. 

MIT researchers have now developed an electrochemical approach to promote hydrogen release from ammonia while simultaneously separating and concentrating the hydrogen into a high-purity stream. Their strategy, which uses electricity to speed up the extraction, reduces the temperature and energy required to recover hydrogen from ammonia and other hydrogen carriers.

In a new study, the researchers showed that their approach can generate highly concentrated, pure streams of hydrogen.

“We have shown the ability to use electrochemistry to drive thermodynamically uphill and kinetically difficult dehydrogenation reactions,” says Yogesh Surendranath, the Donner Professor of Science and a professor of chemistry and chemical engineering. “In this case, we studied the conversion of ammonia and a liquid organic molecule because of their importance as possible hydrogen carriers for a hydrogen economy. But the concepts we learned here could in principle be translated further, and we’re actively working on translating it to other important dehydrogenation reactions.”

Surendranath is the corresponding author of the study, which appears today in Nature. MIT postdoc Rui Zeng, now a professor of materials science and engineering at Harbin Institute of Technology in Shenzhen, China, is the paper’s lead author.

Extracting hydrogen

Hydrogen is widely used in semiconductor manufacturing and chemical processing and is also an energy carrier in fuel cells that use hydrogen and oxygen to generate electricity without combustion. Expanding its use, however, will require practical ways to store and distribute it.

Hydrogen gas itself is difficult to transport efficiently without compression or liquefaction. One alternative is to store hydrogen chemically in compounds that are liquids or can be readily liquefied, then release it where and when it is needed.

Ammonia is one promising hydrogen carrier because it is already produced and transported across large distances, but recovering hydrogen from ammonia remains challenging. That process, known as “cracking,” requires temperatures higher than 500 degrees Celsius to achieve high reaction rates and conversion. The hydrogen must then be separated from nitrogen and unreacted ammonia.

“We wanted to ask whether we could use electrical inputs to drive what would otherwise be an unfavorable dehydrogenation reaction, and simultaneously do it in a way that would separate the hydrogen from the hydrogen carrier, so that it would be very pure and could be used directly in a fuel cell or other application that requires a high purity hydrogen stream,” Surendranath says. 

The key element of the researchers’ new design is the coupling of a palladium-based separation membrane with a hydrogen-generating electrode through a molten hydroxide electrolyte. The separation membrane selectively transports hydrogen while preventing other components of the reaction mixture from passing through.

Using the new setup, ammonia is first dehydrogenated by a catalyst containing ruthenium and cesium. The hydrogen then reaches the separation membrane, whose opposite side is in contact with a molten hydroxide electrolyte. 

The electrochemical gradient across this membrane effectively creates a “vacuum” for hydrogen, providing a strong driving force for its transport across the membrane. It also converts the hydrogen into protons and electrons, which travel separately through the molten electrolyte and external circuit, respectively, before recombining at a second electrode to form hydrogen gas. 

Because the membrane selectively transports hydrogen, the system produces a concentrated stream of hydrogen gas without requiring a separate downstream purification process.

“Using this electrochemical process, we’re able to do this active pumping of hydrogen from a low concentration to a high concentration,” Surendranath says.

Continuously extracting hydrogen can also help drive the dehydrogenation reaction forward, especially when the presence of hydrogen inhibits the reaction. In this way, this strategy does more than separate the product: It changes the reaction environment and enables hydrogen recovery under milder conditions.

This process thus can be performed at temperatures around 200 or 300 degrees Celsius, much lower than those required for conventional ammonia cracking. Another advantage is that it creates a pure stream of hydrogen that doesn’t need to be purified later on — a step that requires additional energy.

Curtis Berlinguette, a professor of chemistry and chemical and biological engineering at the University of British Columbia, described the method as “a powerful new way” to solve the problem of obtaining a pure stream of hydrogen from ammonia and other hydrogen carriers. 

“By using electricity to pull hydrogen through the membrane as it is released, they accelerate the dehydrogenation of ammonia and liquid organic hydrogen carriers while simultaneously producing a purified hydrogen stream. This is an important advance for the energy sciences because it opens a credible pathway for transporting hydrogen in stable chemical carriers and releasing it where and when it is needed,” says Berlinguette, who was not involved in the research.

Powering transportation

In this study, the researchers showed that this approach could be used to dehydrogenate not only ammonia but also methylcyclohexane. This molecule is part of a class known as liquid organic hydrogen carriers (LOHCs), which also hold potential as an energy carrier.

The researchers envision that their new strategy could be useful for transportation applications, such as powering cars, buses, or ships, or for fabricating semiconductors or electronics. Pure hydrogen gas is used for several steps in semiconductor manufacturing, where it plays important roles in boosting manufacturing yields and reducing surface defects.

Because palladium is an expensive metal, the researchers are now working on ways to reduce the amount of palladium needed for the separation membrane. They are also working on scaling up the process, and on applying it to other dehydrogenation reactions that could be industrially useful.

The research was funded by the U.S. National Science Foundation.

How will global access to food, water, and energy evolve in coming decades? A new study co-authored by MIT researchers suggests the answers could be very different depending on region, resource, and income.

Based on extensive modeling of many different resource scenarios, the study finds that in some regions, lower-income people could be spending roughly 50 percent of their income on food by the year 2050, in contrast to higher-income groups that could spent about 5 percent of income on food in the same areas. 

“For a lot of these outcomes, the lower-income groups see much worse potential insecurity,” says Jennifer Morris, a principal research scientist at the MIT Center for Sustainability Science and Strategy and the MIT Energy Initiative, and co-author of a new paper detailing the findings. The results, she notes, can be evaluated by policymakers in different global regions to understand what the long-term, large-scale resource security risks may become for different parts of their populations. 

“Anything that’s taking up half of your income is potentially destabilizing for your entire life because it leaves so few resources for the other critical needs and basic life necessities,” Morris says. 

The study focuses on projecting future access to food, water, and energy, based on long-term variation across a dozen major factors influencing their availability, from economic conditions and agriculture production to trade conditions, climate, land use, and more. 

“This study shows that there is no single driver of future food, energy, and water insecurity,” says Gi Joo Kim, a research scientist at Tulane University and co-author of the paper. “Income is important, but regional conditions, land use, energy systems, water availability, and consumer behavior all shape the risks people face.” For policymakers, he adds, “This means they need to consider specific combinations of factors that create vulnerability in each region.”

The paper, “Identifying Key Uncertainties and Drivers of Future Resource Security Outcomes Through a Multisector Scenario Ensemble,” appears in the journal Earth’s Future.

In addition to Morris and Kim, the authors include Brian O’Neill, an earth scientist at the Pacific Northwest National Laboratory; Marshall Wise, a system engineer at the Pacific Northwest National Laboratory; John Weyant, a professor of management science and engineering at Stanford University; and Jonathan Lamontagne, an associate professor of civil and environmental engineering at Tufts University. 

Filling a gap

The current study fills a gap in modeling among scientists studying issues such as long-term resource security. Given the complications of long-term analyses, many studies have used what scientists term “shared socioeconomic pathway” circumstances, a small set of senarios spanning broad global narratives about the future, rather than exploring specific outcomes such as how long-term resource access may shift in linked fashion across income groups in different regions of the world. Two years ago, the same group of authors wrote a paper calling for more socioeconomically specific scenario analysis focused on outcomes for human well-being; the current study is their effort to develop that kind of modeling. 

“For this type of study, where we’re focused on human well-being outcomes, the income piece is really important,” Morris says. 

To conduct the study, the researchers adopted an existing framework in the field, the Global Change Analysis Model (GCAM) version 7.1, which represents interactions between energy, economies, water, land, and climate while dividing the world into 32 regions, 235 water basins, and 384 land-use regions and making adjustments for things like estimated commodity prices over time.

The research group used 12 main variables connected to resource availability, including population, GDP, income distribution, carbon intensity, land use, agricultural trade, multiple energy consumption scenarios, multiple water-use projections, and more. They ran simulations for 3,735 different scenarios involving these factors, to better understand the range of possible resource outcomes by 2050. 

Broadly, the modeling does uncover some significant regional variations. In 2050 food security may be most acute in parts of sub-Saharan Africa, while energy security could be most acute for low-income residents in some parts of Asia, Eastern Europe, and the Middle East. 

But within any region, there may still be substantial variation in resource security. In southern Africa, the modeling suggests that the poorest 10 percent of the population by income could be spending 49.6 percent of its income on food, compared to just 5.5 percent for the wealthiest 10 percent of the population. In West and East Africa the projected food burden for the bottom 10 percent of the population in terms of income is projected to be 48.4 percent and 42.5 percent, respectively. 

To understand the potential change this represents over time, the researchers compared the results to data from the year 2015 in the GCAM model. For the lowest-income group across western Africa in 2015, the average food burden was about 25 percent of people’s income, compared to estimates for 2050 that range from about 20 percent to 75 percent of income. In southern Africa, the lowest-income group spent about 20 percent of their income on food in 2015, but the scholars’ modeling projects an increase in food burden ranging from 25 percent to 65 percent of income. The wide variation in projected burden reflects the wide range in possible future scenarios.

When it comes to energy, variation by income is also apparent. In some parts of the Middle East, for instance, the residential energy burden in 2050 is estimated to be just 1.7 percent for the highest income bracket but 18.9 for the lowest income bracket; in Eastern Europe, the energy burden reaches 11.3 percent of income for the lowest-income bracket, while resting at under 5 percent for the highest-income bracket. 

“Regional averages can make future resource-security risks appear more manageable than they actually are,” Kim says. “This means analyses that stop at the average may miss exactly the populations most vulnerable to future change.”

Understanding the dynamics

To be sure, as the scholars emphasize, there are many uncertainties when it comes to resource access, and uncertainty is always part of modeling the global economy and resources. Still, they believe these kinds of projections can provide a more detailed outlook about social conditions in 2050 than has previously been available.

“At the very least, it’s highlighting areas of concern and showing that they differ in different parts of the world,” Morris says. “One of the outputs of this type of study is to map that out and provide that kind of insight. That can also inform the focus of further studies into specific regions and concerns.”

The researchers also believe the results will provide a new roadmap for policymakers who may be concerned about long-term resource provision across the entirety of their societies. While having new projections is valuable, modeling also helps analysts and policymakers see which factors most clearly influence future resource outcomes, as well.

“Our method was designed to identify the conditions that produce different resource security outcomes, rather than to predict one most likely future,” Kim says. 

“It’s a different approach to scenarios than we typically see,” Morris adds. “The approach and method have been appealing to people because they have a broad range of uses and applications.”

The research was supported, in part, by the U.S. Department of Energy; Stanford University; and the National Research Foundation of Korea. 

Beneath the Arctic Ocean is an orchestra featuring natural and human composers, from cracking sea ice and whistling beluga whales to humming shipping-vessel engines. Researchers from MIT Lincoln Laboratory heard some of this cacophony when analyzing data from commercial off-the-shelf sensors that they integrated and deployed in 2024 during the U.S. Navy's Operation Ice Camp (OIC). This past March, during OIC 2026, the researchers returned to the Arctic with a higher-fidelity version of one of the sensors, a geophone, which detects vibrations in the sea ice.

"We're interested in things that make sound underneath the ice," says Ben Evans, a researcher in the laboratory's Advanced Undersea Systems and Technology Group. "For example, our OIC 2024 data contained marine-mammal songs. We need a better understanding of how such signals propagate through ice, and how to distinguish these signals from other sources."

This underwater soundscape is shifting as sheets of Arctic sea ice rapidly break and melt, opening previously impassable maritime routes for military and commercial activity. Determining the unique sound profiles, or acoustic signatures, produced by fracturing ice will enable researchers to develop predictive capabilities for building coastal community resilience, informing geopolitical strategy, and surveilling adversary Arctic activity. The Fiscal Year 2027 Administration R&D Budget Priorities and Cross-Cutting Actions calls for agencies to "prioritize research and associated research infrastructure investments that enhance America's ability to observe, understand, and predict the physical, biological, geologic, and socioeconomic processes and interacting systems of the Arctic to protect and advance American interests and ensure prosperity of America's Arctic residents," and to "invest in R&D that assures America's uncontested navigation and strategic utilization of the Arctic."

Weather woes

Evans and David Whelihan have been trekking to OIC since 2022, advancing their vision to distribute a set of low-cost sensors across the Arctic for continuous monitoring. Hosted by the Navy's Arctic Submarine Laboratory (ASL) every two years, OIC is a three-week event during which U.S. and allied military forces conduct operational readiness exercises. The temporary infrastructure ASL sets up for OIC — a drifting sheet of Arctic sea ice, atop which sits a runway and insulated tents for lodging and command and control — simultaneously enables researchers to test prototype equipment and conduct experiments in an environment otherwise inhospitable to humans. This opportunity is particularly valuable as Arctic monitoring systems are developed in support of U.S. Department of War priorities.

This year proved especially challenging, with back-to-back blizzards creating whiteout conditions. Temperatures persistently plunged to minus 25 degrees Fahrenheit, and winds blew at 25 to 30 miles per hour with 40 mph gusts. Although the team, which also included Ella Wawrzynek and Ryan Saenger, had intended on completing two stints on the ice — one to deploy the sensors and the other to retrieve them after a few weeks — the weather had other plans. Their initial trip to camp was delayed by a week as windblown snow halted all inbound and outbound flights.

While they waited in Prudhoe Bay, Alaska, for the weather to clear, Whelihan was readying another technology they planned to test at OIC: a modem from industry partner Havguard, a Norwegian defense technology startup, that can communicate through ice using magnetic fields (instead of radio-frequency signals, which are rapidly diminished by seawater). However, harsh conditions inside and outside in Prudhoe Bay led to some system failures, and he flew back to the laboratory to fix them. "De-risking and operationalizing critical technology for the warfighter is an important part of what we do," Whelihan says.

After Evans, Wawrzynek, and Saenger arrived at camp on March 7, not a single flight came or left for the next five days. During a "normal" mobilization at OIC, six to nine flights per day are typical. "At times, we couldn't see participating countries' flags on poles roughly 100 feet away from the command tent," Evans says. "We had to put on hats and sometimes goggles just to go between tents, whereas at previous OIC events we walked around with long johns and pants."

The day after their arrival, they loaded their sensors onto a sled, and a field party leader (an expert in Arctic survival) driving a 4x4 vehicle with tracks towed them outside the main camp area. After deploying a quarter of the sensors they had planned, their leader received a call from camp command instructing them to return. The windblown snow was picking up, and they soon wouldn't be able to retrace their tracks back to camp.

A week later, the team had a clear day to retrieve their sensors and fly out of camp. "We laughed, because we could very easily see camp from where we had deployed the sensors," Evans adds.

Magnetic communication

As Whelihan returned to Prudhoe Bay with the fixed communications modem, another blizzard hit. Because the laboratory team had already been to camp once and other research teams needed an opportunity to go onto the ice during the next clear-weather window, the laboratory contracted with UIC Science to conduct the modem experiment. A business unit of the Ukpeaġvik Iñupiat Corp., UIC Science provides logistical and technical support for Arctic research based in Utqiaġvik (Barrow), Alaska, the northernmost U.S. city. Unlike the drifting ice in the open ocean, the ice in Utqiaġvik is primarily landfast, meaning it's fastened, or anchored, to the shoreline or seafloor.

Through UIC Science, Whelihan and Wawrzynek learned how to ride snowmobiles and then drove onto a big lagoon, where they drilled a 2x3-foot hole through 3.6 feet of ice to deploy a remotely operated vehicle (ROV) carrying the Havguard communications modem. The modem is based on a magneto-inductive transmitter (positioned below the ice) and receiver (sitting atop the ice), which are housed within polycarbonate domes to protect the sensitive electronics. The duo had met with Havguard in Norway in fall 2025 to discuss plans for testing the modem in the Arctic, and Havguard in turn built a version to testing specifications. Prior to OIC, Havguard and the laboratory team deployed the modem on a large reservoir in Vermont. While not representative of Arctic sea ice over salt water, this environment allowed them to test their procedures and capabilities.

"Under the Arctic lagoon ice, we placed the ROV, which was also equipped with a Doppler velocity logger, a four-beam sonar system that measures the vehicle's speed and direction," Whelihan says. "We used those measurements as the ROV drove under the ice, plus aerial drone images, to superimpose a picture of the ROV on the site so we could track it and calculate the modem's rate of test-data transfer. We achieved through-ice communication at about 1.2 kilobytes per second on an alpha prototype system that had traveled from Boston to Alaska three times. This result is very encouraging, and the system warrants further development."

In future iterations of this setup, the data could then be relayed out of the Arctic through drones or satellites.

Community connections

While in Utqiaġvik for a week in mid-April, they participated over the weekend in the Piuraaġiaqta annual spring festival, watching a harpoon-throwing contest and proctoring a kids' snowmobile race. And with eighth graders at the local middle school, they discussed their Arctic R&D and engaged them in a game teaching sonar concepts.

"When you embrace this culture of community, you meet lots of interesting people and doors open up," Whelihan says.  

"Connecting with the Arctic community is an important aspect of our work," Evans adds. "Every time we come here, we cross paths with someone we don't expect to, learn about their work, and think about how we may be able to collaborate." For example, at OIC 2024, the laboratory team had met a professor from the University of Maryland at College Park with extensive experience collecting and analyzing cryoseismological data; they now hope to work with him to apply machine learning to discriminate between icequakes and marine mammal vocalizations.

In the lead-up to OIC 2028, the team plans to design, prototype, and test air-droppable versions of some of their sensors while continuing to partner with Havguard on the through-ice communications modem. Their next step is to optimize the modem's packaging to facilitate deployability in the Arctic and integration with the laboratory's sensor suite.

"The through-line in all this work is minimizing boots on the ice," Whelihan says. "Especially this year, we learned that the weather is in charge of our access to the Arctic. We need ways to easily get sensors where we want them and to retrieve the data they collect, even in these extremely challenging conditions."

This work is funded through the laboratory's internally administered R&D portfolio in mission-critical technology (integrated systems area) and the laboratory's Advanced Concept Committee, which funds high-risk, high-reward early-stage research addressing critical gaps in national security technology.

Massachusetts Governor Maura Healey and MIT President Sally Kornbluth will kick off MIT Future Fest, a new annual festival exploring the future of science, technology, art, and design, with “The Future Begins Here” panel, a celebration of Massachusetts innovation, talent and the bold questions shaping what comes next. The event will take place on Wednesday, September 30 at 3:30 PM at MIT’s Kresge Auditorium.

Curated and produced by the MIT Museum, the inaugural MIT Future Fest will take place across MIT’s campus from September 30–October 4, 2026. Governor Healy and President Kornbluth will be joined on the opening panel by Moderna co-founder and Flagship Pioneering founder and CEO Noubar Afeyan, MIT professor and entrepreneur Sangeeta Bhatia, and Bob Mumgaard, CEO and Co-Founder of Commonwealth Fusion Systems. Economic Development Secretary Eric Paley will moderate the discussion, which will explore how public, private, and educational institutions can work together to sustain talent pipelines, turn discovery into impact, and build the future

“Massachusetts is where the future is being invented, and MIT Future Fest is a chance to showcase our leadership in technology and design to the world,” said Governor Maura Healey. “From AI and robotics to clean energy and life sciences, the breakthroughs happening here are changing how we live and work. As Governor, I want Massachusetts to be the place where the best minds from around the world come to study, conduct research, start companies and scale their ideas. Our administration is investing in the talent, research and partnerships that make that possible, and we’re proud to launch MIT Future Fest with MIT.”

“The breakthrough discoveries that shape modern life came from decades of scientists and creators asking fundamental questions about how the world works. MIT Future Fest celebrates that same spirit of curiosity on mission," said MIT President Sally Kornbluth. "When you bring together engineers, inventors, artists, scientists, designers and entrepreneurs, you create the conditions for truly transformative innovation. This festival is an invitation to join that conversation, with the conviction that the future isn't something that happens to us—it's something we create together."

“The Future Beings Here” is one of more than 70 public talks, tours, performances, exhibitions, installations, and open laboratories included in the five-day program. Additional festival highlights and the full programming line-up are available at mitfuturefest.org. 

The following press release was issued Sept. 1 by the Broad Institute of MIT and Harvard.

In recent years, scientists have built methods to measure a cell’s transcriptome, or all the RNA produced by a cell, to study the cell’s identity and genetic activity. However, these methods rely on killing the cell to access the bits of RNA within, and offer only a one-time snapshot. 

Now, researchers at the Broad Institute and at MIT have invented a “cellular self-reporting” approach to make living cells share their own transcriptomes, so that scientists can analyze them without killing the cells. Described in Cell, the live cell transcriptomic method relies on virus-like particles, which the cells use to package and deliver RNA to the culture medium they’re bathed in. Scientists can simply sample the medium to isolate the RNA, and do this repeatedly to reveal how gene activity in the same cell population changes as the cells mature or respond to perturbations. The researchers applied their method to a variety of cellular model systems, demonstrating its potential to help reveal how cells go awry over time in disease and how drugs affect cells. 

“Our lab focuses our time and resources on developing tools that will actually get used and make real impact on the broader field,” says study senior author Paul Blainey, who is a core member of the Broad and a professor of biological engineering at MIT. “It’s so gratifying to see a real coming to fruition of this concept, which was complete science fiction when we started. It’s a great example of the innovative impact long-term high-risk, high-reward research can have.”

A cellular special delivery

The effort to build the new method began more than a decade ago, when the Blainey lab set out to find a new way to do RNA sequencing without killing cells. “The existing methods were a bit medieval and involved stabbing cells or cutting pieces off of them,” recalls Blainey. Inspired by the performance of molecular technologies such as CRISPR-based technology and their ease of adoption, Blainey and study first author Jacob Borrajo committed to developing a molecular method, which they knew would be challenging and take time, but would also make the approach scalable and easy for other labs to perform. 

The team found inspiration in retroviruses, which over millions of years evolved the ability to package their RNA genomes in protein shells to spread from one infected cell to another. To build their method, the team engineered mammalian cells to express a retroviral structural protein that can encapsulate not only viral RNA but also a cell’s RNA. Integrated into the cell’s membrane, the viral protein is able to recruit cellular RNA, form a shell around it to create a virus-like particle, and bud off from the membrane to enter the liquid medium around the cell. The scientists then take a sample of the medium, isolate the RNA, and sequence it to get a view of the transcriptome from that cell population — all without destroying or damaging the cells.

“Compared to methods using robotics or mechanical biopsies of cells, our molecularly encoded solution could be much more broadly enabling for the average life science or biomedical lab, particularly the time dynamic questions that we hope to elucidate with this technology,” says co-first author Mohamad Najia, research fellow in the Blainey lab and the lab of George Daley at Boston Children’s Hospital. Najia and Borrajo led the work along with co-first author Anna Le, a postdoc in the Blainey lab. 

Message in a bottle

To test the method’s broad applicability, the researchers showed that it worked in immortalized human cells, in cancer cell lines, in stem cells and neuronal cells made from them, and in primary cells from human donors. They also tested a culture of two human cell types growing together, using tags on the virus-like particles so that the signals from the two cell types could be distinguished during analysis. 

In addition, cellular self-reporting is useful for studying systems with crucial three-dimensional structures that researchers would rather not disturb. The team demonstrated their method on spheroids of human endothelial cells, capturing short-term transcriptional changes after biochemically stimulating the cells. 

They also collaborated with Linda Griffith, a professor of biological and mechanical engineering at MIT, to apply their method to her lab’s organ-on-a-chip devices. These models mimic the physiology of organs and can help minimize preclinical or animal model testing, but their complexity makes retrieving cells from the devices for analysis difficult. With cellular self-reporting, the researchers monitored gene expression dynamics in endothelial cells within the devices over time, revealing changes in genes related to how tissues form vascular networks that depended upon the source of supporting fibroblasts, such as from either uterus or lung. 

The Broad team is continuing to look for new applications and biological questions to ask with their system, and are working to make the approach feasible for studying single cells. For now, they hope that scientists interested in following how cells and tissues change over time will give their method a try.

For generations, the steel mills of Southeast Chicago offered work and a way of life, experienced by tens of thousands of families. Open around the clock, three shifts per day, the vast works of U.S. Steel, Republic Steel, Inland Steel, and many others provided demanding but steady jobs, while making materials to build the country. 

Professor Christine Walley, head of the MIT Anthropology program, grew up in the area, where her father worked for Wisconsin Steel. Over time, U.S. manufacturing downsized — her father’s plant closed in 1980 — and the mills left Chicago. Walley’s 2013 book, “Exit Zero: Family and Class in Postindustrial Chicago,” chronicles the economic and psychological toll plant closures took on the area’s workers and families. The book was followed by a documentary, “Exit Zero,” directed by Chris Boebel, media development director at MIT Open Learning (and Walley’s husband).

Then Walley turned to a new effort — the Southeast Chicago Archive and Storytelling Project, an online repository of objects and images, as well as new video features about the industry, labor history, and the local community. The project was  developed in collaboration with a team from the Southeast Chicago Historical Society, which in 1985 opened a museum about the steelworking life. This award-winning newer online project has been supported, in part, by MIT, the National Endowment for the Humanities, and others.

The idea is to use objects to tell stories about the area’s history. To mark Labor Day, MIT News offers this photo essay based on materials from the Southeast Chicago Archive and Storytelling Project, recognizing America’s workers — and reflecting on the jobs, work, and life produced by industry and innovation. 

It’s hard to depict the vastness of Southeast Chicago’s steel manufacturing area, which stretched for miles into Northwest Indiana and included mills that employed 120,000 workers at their peak. This vintage postcard shows the industry along the Calumet River. 

MIT’s 12th president, Howard Johnson, grew up in the area during the Great Depression, where family members toiled in the mills. Johnson’s father worked for U.S. Steel for 50 years, becoming a bookkeeper and accountant, and Johnson attended Bowen High School, which is still open today.

“Families — ours and thousands like it — were the essential centers of life in the community,” Johnson writes in his memoir, published by the MIT Press.

Families have also been essential to the Southeast Chicago Archive and Storytelling Project, which displays more than 1,100 items from the local historical museum’s collection: clothing, photos, scrapbooks, news clippings, recreational objects, oral history materials, and more. Walley says that anyone could take this approach, and use objects to tell stories about their own local history, work, and community life.

“People tend to experience history in their day-to-day lives not through books written by experts, but by telling stories around family objects and photos,” Walley says. “What is meaningful to us about the things we save from the past? Might these items be ‘clues’ that take us on a deeper historical journey?”

Kitty Kalwasinski Markovich (above, left) didn’t set out to become a welder — but as one, she nearly appeared the movies. Born Kazmira Kalwasinski, she immigrated with her family to Chicago from Poland in 1913, at age 10. During World War II, the steel mills sought replacements for men serving in the military, and she started welding at the South Works of U.S. Steel in Chicago. 

Warner Brothers depicted her in this photo shoot (with Florence Josephs, right), as a worker in the “Rosie the Riveter” mode, and considered making a film featuring her. Five of her brothers served in the military, and their names are seen on Markovich’s welding helmet. One of them, Frank Kalwasinski, was killed in World War II, and his sacrifice is represented by the solid star.

Although many American women returned to their former lives as homemakers at the conclusion of the war, Markovich kept working in the mills, welding for 23 years in the South Works before she retired in 1967. It’s also where she met her husband, Michael Markovich. Kitty Kalwasinski Markovich’s family donated many materials to the Southeast Chicago Archive and Storytelling Project. 

This hard hat, made around 1986, was donated to the Southeast Chicago Historical Museum along with a cutting from the last beam ever produced at U.S. Steel’s once-mighty South Works, which closed in 1992, having produced steel since the 1800s. 

At its peak, South Works employed about 20,000 people. The Chicago steel mill closures, the 1970s through the 1990s, devastated employees and their families who, as Walley details in the “Exit Zero” book, identified strongly with steelworking. Her own father was a third-generation steelworker.

The online storytelling project deploys MIT scholar Sherry Turkle’s notion of “evocative objects,” those that hold great resonance and get our minds in motion. As Turkle writes, “we love the objects we think with.” 

Donated to the museum by James Stapay, this hard hat evokes the end of a long industrial era in Chicago, and is featured in “The Closing of the Mills,” one of the project’s four documentary videos created from donated objects.

Working in steel mills was not just physically demanding, but dangerous. Worker injuries and deaths were a recognized problem, especially in the early years. This is prototype safety gear from around 1911-12, from a series of photo albums donated to the Southeast Chicago Historical Society by U.S. Steel itself in the 1980s. 

U.S. Steel set up a Committee of Safety early in the 1900s, which recommended 3,000 changes to operations. However, in oral history interviews, workers often recount continuing dangers and terrible accidents in the steel mills. 

Some first-person accounts state that as late as the 1960s, workers were still not regularly wearing hard hats. Basic safety practices seemed to improve, though, after the introduction of the U.S. Occupational Safety and Health Administration in 1970. Worker safety was a long-term work in progress. 

Many of the close-knit communities formed around the mills have donated materials to the project that range far beyond the factory. For instance, family recreation was important in steelworks neighborhoods. Pictured here, in a photo donated by the Cordero family, is the “Mayas” softball team, made up of members of Southeast Chicago’s Mexican-American community, which won a 1937 community league championship.

Justino Cordero immigrated to Chicago in 1923, became a steelworker, then eventually did electrical work in the mills while opening a radio shop. Cordero, a father of three (two of his children are pictured), organized and coached youth sports teams to keep kids “out of trouble”; wrote a column for The Daily Calumet, a local newspaper; and was involved with his church, Our Lady of Guadalupe. After retirement, Cordero earned undergraduate and master’s degrees, to work with children with disabilities.

Another documentary video from the Southeast Chicago Archive and Storytelling Project, “Mexican-American Journeys,” explores the long history of Mexican-American steelworkers in Chicago, in many dimensions. At least a dozen parishoners of Our Lady of Guadalupe members who had been in the U.S. military in the 1960s were killed in the Vietnam War, serving their country. 

In 1937, steelworkers went on strike in Chicago. On Memorial Day, during a peaceful protest at Republic Steel, 10 workers were killed by law enforcement officials — an event that provoked congressional hearings in Washington. The image at left is a poster made up for Local 1033, the union branch for the Republic Steel plant; at right, Local 1033 workers take a vote in later years. 

While the so-called “Memorial Day Massacre” was a landmark event in national labor history, it is remembered in more intimate ways in the local area, with many families later donating photos, news clippings, scrapbooks, and interviews about it to the Southeast Chicago Historical Society. 

And though the Southeast Chicago Archive and Storytelling Project project focuses heavily on working-class employment and daily life, there are many other possibilities for U.S. community-based history, involving almost any topic. Whatever the places, objects, and stories, but the goal is the same: to keep the past alive. 

It started with a vision: Move MIT’s School of Architecture and Planning (SA+P) into the Metropolitan Storage Warehouse, an unoccupied, fortress-like brick building on MIT’s campus in Cambridge, Massachusetts.

After all, SA+P needed more space and new facilities. And here, visible from its old offices across the street, was an unused building the size of an airplane hangar. It could offer bigger studios, more work areas, an auditorium, and galleries for events, and become a campus-wide hub for teaching, research, and public engagement. 

“MIT thrives on this idea that we’re all connected,” says Hashim Sarkis, dean of SA+P and a key proponent of the project.

But that vision required hundreds of design decisions: how to bring light into the building, create workspaces and circulation, encourage communication among the school’s populations, and more.

“The conception of the project was not like an automatic flash,” says Elizabeth Diller, founding partner at Diller Scofido + Renfro (DS+R), the architecture firm that was selected to revamp the Met Warehouse, as it’s now called.

“It was a very challenging building to work with,” says Benjamin Gilmartin, another DS+R partner. “There was a lot of innovation needed.”

Innovation is welcome at MIT, however. 

“They transformed the Met Warehouse toward the things we want, which is to do more collaborative work, and to combine instruction and research,” Sarkis says. 

Here’s how DS+R, working with MIT over several years, created the new Met Warehouse — which has a ceremonial moving-in procession on Sept. 8. 

Five buildings in one 

The Met Warehouse was built in several phases starting in 1894; by 1923 it was a five-story building with 2-foot-thick brick walls and 1,500 storage units inside. It was a fortress used for private storage, with the words “Metropolitan Storage Warehouse Fire Proof” painted on the side, visible from across the river in Boston. 

The structure was built in five segments, over time. That became crucial to its transformation. Diller and Gilmartin created a large design studio inside each of the segments. 

“When you start a project like this, there are some big moves that seem very clear and obvious,” Diller says. “There are five buildings that were built basically in succession, making for a 500-foot long building. That is just too big, so how do we break it up into neighborhoods? We decided each building itself would have a stack of floating studios in it.” 

That was essential for SA+P and the components of it that will use the building, such as the MIT Morningside Academy for Design, which was established through a $100 million gift from the Morningside Foundation, the philanthropic arm of the T.H. Chan family. This founding gift from family members Gerald and Beryl Chan and Ronnie and Barbara Chan included support for the Met Warehouse transformation. 

Each building segment features double-height, column-free studios which, thanks to virtuoso engineering, are suspended from roof trusses that bring the weight back down to the existing structure. 

Those spaces will benefit the interdisciplinary work taking place at MIT. 

“In those five spaces, we’re putting the making and the research together,” Sarkis says. “The studio and the lab will become one and the same.”

Bringing in light

For about a century, the Met Warehouse featured tiny window slits as its only apertures. That raised a question: How could natural light be brought inside? 

DS+R produced a dramatic answer, drawn from recent architectural history. Along the long north side of the Met Warehouse, adjacent to a set of railroad tracks, they carved large voids for the studios. Aligned with the studios, large segments of the brick exterior were replaced with glass facades.

This way, natural light pours into the studios and beyond, while occupants look out to a lively Cambridge cityscape.

“The process was like an extraction of the dense mass of the building to create open and light-filled space connecting all,” Diller says. 

Her aesthetic inspiration included the artist Gordon Matta-Clark, known for making bold cuts into New York City buildings in the 1970s. 

“Right from the beginning there was a nod to Gordon Matta-Clark,” Diller says, though she notes that Matta-Clark’s work consisted of building-scale interventions motivated by political and social critique. Whereas, “In our case, we use subtraction to build — to make space for new uses and to expose the anatomy of the building.” 

The living lab

To ensure the huge cuts and windows would work, MIT collaborated with DS+R, as well as Leers Weinzapfel Associates, the project’s associate architects, and Shawmut Design and Construction, to test slab cuts and window arrangements directly in the Met Warehouse itself.

“This hands-on approach allowed us to prove the design concepts through actual construction methodologies and logistics,” says Nicole Bernabei, a senior project manager for campus construction at MIT, who has worked on the Met Warehouse effort since 2018. 

Those slices into the building, needed to create the studios, revealed the Met Warehouse’s original structural features as cross-sections now appearing in walls. The designers envisioned those cuts as features to remain visible, something students can still learn from.

“For a school of architecture and planning, this approach feels especially fitting,” Bernabei says. “The building itself has become a teaching tool — a living laboratory where students, faculty, staff, and visitors experience how rigorous design thinking translates into a built reality. Every detail, from the celebrated slab edges to the transformative light, tells the story of collaboration and precision.”

And while those issues were being addressed, the architects had to grapple with, well, everything else. 

Asymmetry inside

There is another reason the architects placed the huge glass walls on the north side of the Met Warehouse. The Cambridge Historical Commission (CHC) asked MIT to keep the building’s south and east sides essentially intact. The long south facade, the one historically visible from Boston, was particularly significant.

“Changing the building’s surface there [on the north side], bringing in the large glass, wouldn’t impact the character of the building as it would on the south side,” Diller says. For that reason, in the interior, “the big spaces drift to the north.” 

The architects placed smaller spaces, like offices, on the south side. 

“The fabric of the existing Met Warehouse building, with its column grids, offered a lot of opportunities for more serialized smaller spaces where you can have seminars, faculty offices, teaching spaces, research areas, next to and in dialogue with the larger multistory spaces that we introduced,” Gilmartin says. 

So, the Met Warehouse is asymmetric inside: big studios extending from the north side, across much of the building; and smaller rooms on the south side.

It was not obvious how to bring more light into the south-side rooms, however. But an extended dialogue between DS+R, MIT, and the CHC produced an “intersect window” strategy — box frame windows sometimes intersecting with the small apertures already on the south side. The steel frames of the new windows, now at a proper height for looking out, distribute the weight of the brick and stone sills once carried by the solid brick that was removed. 

“This meeting of the old and new satisfied the Cambridge Historical Commission’s desire for a minimal touch, but also created an unexpected and delightfully playful effect on the south facade,” says Morgan Pinney MArch ’10, a senior campus planner at MIT. The project, she adds, “allowed us to step into an exceptionally collaborative working relationship with CHC staff — one MIT is very proud of and will certainly continue to build upon for years to come.”

The unusual layout grants a centrality to the studio areas while ensuring that a full range of other spaces are wrapped around them. 

“The logic of the building is that making things is in the middle,” says Sarkis, referring to the studio spaces. “This is MIT. There is making and research in the studios, with seminar rooms and offices all around.” Referring to the school motto, “mens et manus,” he adds, “That’s our culture. MIT is about mind and hand.” 

Still, Sarkis and the architects wanted another element inside, too: interior passages connecting it all.   

Extending the Infinite Corridor

MIT’s main group of buildings features the Infinite Corridor, a busy walkway spanning one-sixth of a mile indoors, linking many other spaces. MIT leaders thought the Met Warehouse could extend the concept. 

“We conceived of it as having an ‘Infinite Corridor,’” says Sarkis, who hoped the corridor would be “visible and accessible all the way through.”

Diller, the architect behind New York City’s High Line, which turned elevated railroad tracks into a wildly popular urban park, knows about getting people walking. She wanted Met Warehouse occupants to “share a circulation system.”

And so every floor of the Met Warehouse has its own “infinite” corridor. Some overlook studio space one floor down, with the cityscape beyond, echoing High Line atmospherics. The off-center circulation spine connects the large studios on the north side and the offices at the south edge of the building. Being flexible about that placement allowed the whole Met Warehouse plan to work. 

“That helps create the space for the large studios,” says John Ochsendorf, director of the MIT Morningside Academy for Design. Besides, he offers, “There’s a happy alignment between historical protection of the south facade, and where the sun is in the sky most of the year. You don’t want direct sunlight on the south side” — where glass walls would create a greenhouse effect — “and the city wanted to protect that view. The architects found this balance.”

A vertical vision

Meanwhile, the architects designed large stairs that pierce through the corridors, helping people access all floors of the building. 

“We didn’t want a layering of the building with horizontal stratification,” Gilmartin says. The stairs will provide “moments of serendipity and exchange with other people.”

The first time Gilmartin drafted stairs for the building, they were “more blade-like and expressive formally” than the final version. But MIT asked for a stripped-down sensibility, so Gilmartin made the design “almost as simple as it could be.”

The stairs still have expansive scope and meeting-place potential.

“That central spine stair is sort of a ceremonial stair,” Diller says. “To see and be seen. I think it will be lively.”

Culture change

The Met Warehouse is very different from the previous quarters of SA+P, a warren of rooms in MIT’s buildings 7 and 9. Many peer institutions have design studios that place all students in a large common space. Not MIT, which has had a different culture, with smaller, specialized design spaces.

Now, the Met Warehouse does feature larger and more visible design areas. 

“We’re trying to give the school the ability to adapt it and change it, and balance the past culture of MIT and a new culture with kinds of spaces where ideas can cross-pollinate and there’s a lot of room for large-scale experimentation,” Gilmartin says. 

Design practice is becoming bigger across MIT, and as more people connect with it, the Met Warehouse will let MIT evolve. 

Gilmartin again: “There’s just a lot of opportunity for smaller groupings of people to be organized in ways that are visible and connected to the larger spaces, but also offer the prospect of a retreat and focused work. I think it is hopefully attuned very well to MIT.” 

The shock of the rebuilt 

It’s unusual to move a major architecture school into an old building. Some of the best-known U.S. universities house their architecture schools in buildings with high-modernist stylings or postwar brutalist aesthetics, heavy on concrete, light on graceful curves. 

“Those buildings come out of the modernist tradition predicated on the shock of the new,” Gilmartin says. “There was something confrontational about those buildings in their material expression and image, and spatial ideas of openness and flexibility, which was quite different from most historic buildings. In their time, they were pretty thrilling.” 

But as the saying goes, that was then, and this is now. 

“The reality of our future is that we can’t tear everything down and build new for every generation,” Gilmartin says. The Met Warehouse “makes a claim about the future of design and what the orientation of that needs to be, in our work,” he adds. 

MIT agrees. 

“I think it sends a very good message that this vanguard school of architecture, at the Massachusetts Institute of Technology, is moving into a historic building and adapting it for the future,” Sarkis has said. 

Along with the Morningside Foundation, another key project donor was Sidara (formerly the Dar Group), a global collaborative of specialist design, engineering, and consulting firms, owned by Maha and Talal Shair; they have supported the creation of the building’s Sidara Auditorium and Sidara Gallery space.

Adaptive re-use 

Ultimately, DS+R was ideal for the Met Warehouse project because of their experience transforming structures. Besides the High Line, they transformed a London media center built for the 2012 Olympics into the Victoria and Albert Museum’s new open storage facility, the V&A East Storehouse.

Diller suggests the key is being pragmatic.

“When you have a building that is that thick, that heavy, that present, sometimes it’s more expensive to demolish it than to invent a way of reusing it,” she says.

Besides, she adds, “Because it’s an architecture school, it’s important that the students understand adaptive reuse firsthand, as we share a planet with limited resources. It’s a good thing to repurpose buildings, to change their program, to update their innards where possible, rather than just preserving them in formaldehyde — or destroying them and taking away the character of a city.”

The Met Warehouse was old, is new again, and is ready for the MIT community to make it their own.

“Very often contemporary buildings are so sanitized and clinical, you don’t feel like you can touch anything,” Diller says. “It doesn’t feel like home. Here, we wanted students to feel uninhibited — a place that would feel like home.”

The MIT School of Architecture and Planning recognized 11 faculty members with promotions for their significant contributions to the school, effective July 1, 2026. Four faculty promotions are in the Department of Architecture, three are in the Department of Urban Studies and Planning (DUSP), and four are in the program in Media Arts and Sciences.

“These individuals offer the MIT community creativity, knowledge, and scholarship that is exhilarating,” says Hashim Sarkis, dean of the School of Architecture and Planning. “Collectively, they add considerable strength to our faculty and research capacity.” 

Department of Architecture

Xavi Aguirre has been promoted to associate professor without tenure. Aguirre is also director of DIS-ASSEMBLIES LAB, where his work focuses on learning from and designing for when architectures come apart. Through building (and unbuilding) projects, product development, and research, he considers our relationship to material and commodity circulations, both technically and culturally. Aguirre’s work has been commissioned by the Carnegie Museum of Art, the Industry Opera Co., MOCA Geffen, and Dartmouth College, among others. He is the is founder/director of the design studio stock-a-studio.

Rosalyne Shieh has been promoted to associate professor without tenure. Shieh is an architect based in Cambridge, Massachusetts, and Kaohsiung, Taiwan. Her work engages places at the intersection of material culture, oral history, and postcolonial identity; architectural projects are processes for thinking-with-site, and design is the extension and invention from the ordinary. Shieh has worked for Stan Allen Architect, ARO, and Abalos & Herreros, and was formerly co-director of the collaborative practice Schaum/Shieh. She is co-author of “Blanking: An Annotated Archive of Projects and Thoughts on Architecture” (Park Books, 2025). 

Nida Sinnokrot has been promoted to associate professor with tenure. Sinnokrot is an artist and educator whose work explores how various forms of power and bias are embedded in dominant narrative structures and attendant articulations of time and space. Working across film, video, photography, sculpture, installation, and agriculture, Sinnokrot seeks to expose and cannibalize — through tactile, tactical, and material acts of technical and conceptual detournement — various technologies of control that give rise to shifting social, political, and environmental instabilities. He is a co-founder of Sakiya, an international residency program and research platform in the West Bank village of Ein Qinya. His recent solo shows include “Nida Sinnokrot” at Kunst-Station Sankt Peter, Cologne (2019-20) and “Expand Extract Repent Repeat” at Carlier | Gebauer in Berlin (2018-19).

Kristel Smentek has been promoted to full professor. Smentek is a historian of 18-century European art with specializations in histories of the graphic and decorative arts in their transcultural contexts, the history of collecting, and European encounters with Asia. Smentek has received fellowships and awards from a host of organizations including the National Endowment for the Humanities, the American Council of Learned Societies, and the Andrew W. Mellon Foundation. In her current book project, “Disorient: Arts from China in Eighteenth-Century France,” Smentek analyzes European engagements with Asian imports in the 18th century and their impact on continental art and aesthetic theory. Smentek is also active as a curator. Most recently, she was co-curator of the exhibition “Dare to Know: Prints and Drawings in the Age of Enlightenment,” which examined the constitutive role of works on paper in the propagation of European Enlightenment ideals and blind spots, and a contributor to and co-editor of its accompanying catalog (Harvard Art Museums, 2022).

Department of Urban Studies and Planning

Jason Jackson has been promoted to associate professor with tenure. Jackson is an associate professor of political economy and director of the Political Economy Lab. His research focuses on the relationship between states and markets, particularly the role of economic ideas and moral beliefs in shaping market institutions under modern capitalism. Empirically, his work focuses on contexts ranging from the role of economic nationalism in industrial development to the rise of the digital “platform” economy and urban mobility markets in contemporary cities in Africa, Asia, and the Americas. Jackson is the author of “Traders, Speculators and Captains of Industry: How Capitalist Legitimacy Shaped Foreign Investment Policy in India” (Harvard University Press, 2025) and “Constructing Economic Nationalisms in Brazil and India” (Cambridge University Press, 2026).

Justin Steil has been promoted to full professor. Steil is professor of law and urban planning and a Margaret MacVicar Faculty Fellow. As Academic Curriculum Committee faculty chair, he leads a new faculty committee for the MIT Center for Real Estate and the Master of Science in Real Estate Development program, bringing together faculty from across the school to shape the future of the curriculum. As a lawyer, a paramedic, and an urban planner, Steil’s research focuses on spatial dimensions of inequality. Steil analyzes spatial inequality in the domains of environmental justice, housing and land use policies, and health equity, among others. Recent research focuses on the effects of neighborhoods on health, on mobility risk, and on the role of emergency medical services in advancing health equity. 

Sarah Williams has been promoted to full professor. Williams, the Norman B. (1938) and Muriel Leventhal professor of architecture and planning, is the director of the Civic Data Design Lab and director of the Norman B. Leventhal Center for Advanced Urbanism. Williams combines her training in computation and design to create communication strategies that expose urban policy issues to broad audiences and create civic change, a process she calls Data Action. Williams is co-founder and developer of Envelope.city, a web-based software product that visualizes and allows users to modify zoning in New York City. In her book “Data Action: Using Data for Public Good” (The MIT Press, 2022), Williams provides a guide for working with data in more ethical and responsible ways.  

Program in Media Arts and Sciences

Fadel Adib has been promoted to full professor. Founding director of the Signal Kinetics research group, Adib holds a joint appointment in the Department of Electrical Engineering and Computer Science. He has spent his career expanding the frontiers of wireless sensing, developing technologies that locate hidden objects, navigate indoor environments, and detect contaminants in food and water. Since receiving tenure in 2022, his group has integrated generative AI with wireless vision, enabling robots to interact with objects blocked entirely from view. His spinoff, Cartesian Systems, is now deployed in more than 700 stores across 55 countries. Adib was named a Young Global Leader by the World Economic Forum in 2024 and received the Great Arab Minds Award in Engineering and Technology in 2023.

Canan Dağdeviren has been promoted to associate professor with tenure. Dağdeviren is the founder and director of the Conformable Decoders research group and has built a research program around the idea that the human body continuously produces coded physical patterns — through motion, pressure, sound, electrical activity, and biochemical change — that carefully designed materials and devices can decode to improve human health. Her group develops technologies that bend and conform to the body, including a wearable ultrasound breast patch for more frequent and accessible cancer screening, and ImPULS, an implantable piezoelectric ultrasound stimulator for deep brain stimulation. Dağdeviren also serves as faculty lead of the MIT Media Lab's WHx Women's Health Program, which this year became an official program of the MIT HEALS Initiative — a significant institutional recognition of her commitment to addressing longstanding gaps in women's health through bold, cross-disciplinary research.

Kevin Esvelt has been promoted to associate professor with tenure. Esvelt, who leads the Sculpting Evolution research group, invents new ways to study and influence the evolution of ecosystems, addressing some of humanity's most difficult ecological and public health challenges with a commitment to openness and humility. His community-driven “Mice Against Ticks” project, featured in a landmark 60 Minutes segment last year, is engineering Lyme-immune mice on Nantucket, Massachusetts to stop the disease at its source. He has also emerged as one of the nation's leading voices on biosecurity, with research exposing the dangers of AI systems capable of generating biological weapons information — highlighted by The New York Times in April — and a Policy Forum piece in Science on mirror life that has advanced critical public conversations about the governance of emerging biotechnologies. His work on CRISPR gene drives, RoboPace, and “daisy drives” has reshaped global conversations about the future of evolution.

Danielle Wood has been promoted to associate professor with tenure. Wood, the founding director of the Space Enabled research group, has spent her career proving that satellite technology can be a powerful instrument for justice on earth. Holding a joint appointment in the Department of Aeronautics and Astronautics and serving as MIT's faculty lead for African and African Diaspora Studies, Wood has developed the EVDT framework — Environment-Vulnerability-Decision-Technology — a systems approach applied successfully in countries including Ghana, Angola, and Brazil. Her impact extends well beyond the MIT Media Lab: She received the 2026 Paul Gray Faculty Award for Public Service and the 2025 Letten Prize, and has served multiple times as a private sector advisor to the U.S. Delegation to the United Nations Committee on the Peaceful Uses of Outer Space.

Vice Provost for International Activities Duane Boning oversees MIT’s international policies and engagements. Here he discusses why international engagement remains vital to MIT’s mission of advancing knowledge, educating students, and innovating to serve the nation and the world.

Q: Why does MIT consider international engagement important to its institutional mission?

A: International engagement is an integral part of what makes MIT strong. It allows us to collaborate with excellent partners, access facilities and research environments unavailable at home, monitor and learn from worldwide progress, and remain closely connected to the world of ideas and innovation.

MIT’s mission is to advance knowledge, educate students, and develop innovations that serve the nation and the world. To do those things well, we have to understand — and help shape — the global landscape of science and technology.

Today’s breakthroughs don’t emerge in isolation. Scientific talent, research facilities, and technological advances are distributed across the world. To ensure that we remain at the forefront of discovery and innovation, we need to know what is happening beyond our borders, collaborate where it serves our mission, and prepare our students to compete in a changing world.

That does not mean engagement without limits. We are open to international collaboration and global talent paired with sensible, risk-based safeguards. We welcome international collaboration where it advances our educational and research mission, while applying rigorous safeguards to protect sensitive research, intellectual property, and national security.

The world is already deeply interconnected, and we believe that thoughtful engagement makes us stronger.

Q: How does engaging this way benefit MIT and the nation?

A: International collaboration is central to keeping MIT and our graduates at the forefront of knowledge and innovation in the 21st century. 

Collaboration is vital to attracting, retaining, and educating the leaders of tomorrow — students, faculty, and researchers from both the United States and around the world — and to feeding MIT's innovative and entrepreneurial spirit. 

Frankly, throughout its history, MIT has flourished in part because of its capacity to attract the world’s very best students, faculty, and researchers — many of whom remain permanent contributors in the United States. These individuals have gone on to teach, deliver breakthroughs, and create American startups.

International partnerships also give MIT researchers access to things that aren’t readily accessible in the United States. For example, MIT researchers working in Singapore were able to test autonomous vehicle technologies years before suitable facilities were widely available domestically. Off the coast of Portugal, MIT researchers are now partnering to develop the next generation of deep-ocean monitoring systems in offshore environments that don’t exist here. The knowledge and insights gained in such settings benefit everything we do.

Q: How is MIT adapting to a changing and more uncertain world?

A: Our mission to serve the nation and the world remains steadfast, but we acknowledge that the world has changed. MIT has rigorous processes for evaluating and mitigating potential risk associated with international engagements in a thoughtful and thorough way, allowing the Institute to engage appropriately in new opportunities when and where possible.

Global collaboration still plays an essential role in advancing research, education, and innovation. But we recognize the challenges of operating within a rapidly shifting environment marked by geopolitical uncertainty, evolving federal funding priorities, regulatory complexity, policy shifts limiting international mobility, and competition for global talent and collaborations.

On the research front, we seek to cultivate relationships across industry, government, and the global alumni community. On education and student experience, we recognize that the United States benefits when American students understand the world around them. It expands opportunities for both undergraduate and graduate students to gain international experience while strengthening relationships with trusted partners and maintaining the flexibility to adapt as global conditions change.

While MIT’s global engagement is partly built on long-standing relationships that have delivered meaningful outcomes for students and researchers, today’s volatile landscape calls for forward-thinking cultivation of new relationships around the world. By investing in new regions, MIT will be better positioned to adapt to changing global circumstances. 

International engagement enables MIT to deepen its impact, strengthen innovation, and enable students and researchers to take part in tackling the world’s most pressing challenges.

MIT Assistant Professor Thomas Rose, an expert in ancient metallurgy, passed away on Sept. 2 due to injuries sustained during a bicycle accident in Cambridge, Massachusetts. The incident, currently under investigation, occurred at the intersection of Memorial Drive and Massachusetts Avenue. Rose was 37.

Rose, who was MIT’s POSCO Professor of Materials Science and a member of MIT’s Center for Materials Research in Archaeology (CMRAE), joined MIT in January of this year and was still putting the finishing touches on his laboratory. But he had already endeared himself to colleagues and students by going the extra mile in mentorship, encouraging others to use his new equipment, and even using a portion of his lab startup funds on things the department needed.

“Everyone can look at his papers and his past and understand why he was such a good fit here technically,” Senior Lecturer Michael Tarkanian says. “But in the time he was here, it was even more impressive how likable, friendly, and open he was. He was everything you could have asked for as a colleague and a person. I thought, ‘What luck to be able to work with this person for the rest of my career.’ He was that good.”

Rose was born in Berlin, Germany. He discovered his life’s passion as a child, through a set of books about ancient Egypt.

“Thomas had a strong interest in archaeology already from a young age and participated in an excavation before he began his studies of archaeology,” Katrin Westner of the Deutsches Bergbau Museum Bochum and Professor Sabine Klein of the Ruhr University of Bochum wrote in a joint email tribute. “He was an incredibly inspired and enthusiastic researcher and was always bursting with new research ideas. We remember Thomas not only as a brilliant and dedicated researcher but also as a very close friend. We miss him deeply.”

Rose received his bachelor’s and master’s degrees from Goethe University Frankfurt and earned his PhD in archaeology through a joint doctoral program at Ben-Gurion University of the Negev in Israel and Sapienza University of Rome in Italy. Before coming to MIT, Rose held research and coordination roles in Germany at the Deutsches Bergbau Museum Bochum and Goethe University Frankfurt.

Rose’s research focused on ancient metallurgy and pyrotechnology that shaped early human societies, including how copper and its alloys were produced, transformed, and circulated.

The work integrated geochemistry, mineralogy, experimental archaeology, and materials science, making Rose an excellent fit in MIT’s tight-knit CMRAE group, which merges materials science with archaeology.

“The field of archaeometallurgy is unique,” explains Professor Polina Anikeeva, head of the Department of Materials Science and Engineering. “We had been looking for a faculty with the right skill set for at least 20 years. We needed someone who was world class in archaeology and materials science. We were looking for a unicorn, and we found him.”

Following the announcement of Rose’s hiring, a group from CMRAE traveled to a conference in Italy and heard from scholars based around the world about how lucky they were to have him.

“Thomas was an exceptionally talented and versatile scientist,” says University of Tuebingen Professor Silvia Amicone. “His ability to bring together archaeology, archaeometallurgy, geoscience, and materials science was remarkable. He was also committed to developing digital tools and promoting open, accessible, and reusable archaeological data. This combination of scientific rigor, methodological creativity, and engagement with broader archaeological questions made his work especially valuable. I was always impressed by Thomas’s brilliant intellect, collegial spirit, enthusiasm, and dedication to his work. Above all, he was a genuinely kind and good person.”

Rose had never taught before coming to MIT, but he was excited to begin his first courses this fall. His lab’s first batch of graduate researchers just arrived at MIT, but Rose had already begun mentoring students.

“He was the kindest person you could meet,” says Assistant Professor Tania Lopez-Silva, whose office was close to Rose’s. “He was always smiling. He really cared about his students, and he had a lot of momentum here. He was here first thing in the morning and late into the night.”

Several colleagues recalled the energy and enthusiasm he brought to work.

“He was so excited every day,” Anikeeva says. “Every day was a dream come true for Thomas. He was at the right place at the right time. He was so excited to collaborate and learn. He felt like he got his dream job, and everything he’d ever imagined was about to happen. It’s an unrealized vision.”

Tarkanian had recently restarted weekly meetings among researchers in CMRAE, which Rose attended consistently. Forging connections was a theme of Rose’s career.

“He had a long reach with both young and established scholars, and he inspired his friends and colleagues to show up,” says postdoc Benjamin Sabatini. “His work in archaeometry was paramount, and it showed in the people who gathered around him.”

Rose was also the co-founder of the Young Researchers in Archaeometry workshop, which brought together early-career researchers from around the world to present work and connect.

“Since [its founding], he accompanied every year’s workshop organizing meeting, always with immense kindness and support, making a lasting impact on each of us,” researchers Sinem Haciosmanoglu and Baptiste Solard wrote together in an email. “He was an exceptionally talented and dedicated researcher. Even at an early stage of his career, he brought new ideas and perspectives to the field. For many of us, he also played an important role in bringing together all fields of archaeological sciences, natural sciences, and cultural heritage.”

Outside of research, Rose was fond of rowing on the Charles River and was an avid member of MIT’s Archery Club. He loved manga Japanese comics and dancing. Lopez-Silva described Rose as humble and sometimes reserved, but on a recent recruitment outing with students, he fully committed himself to a very memorable karaoke performance.

“The students loved him,” Tarkanian says. “You could see that he cared about them, and they cared about him. He was going to be that kind of mentor.”

The field of silicon photonics, which uses light rather than electricity to transmit and process data on semiconductor chips, has enabled optical systems to evolve from bulky setups to compact and advanced systems. Typically, however, these silicon-photonics chips are rigid and opaque.

MIT scientists have now figured out a scalable way to make silicon-photonics chips flexible and transparent, opening a route to advanced microchips that could be used in applications such as discreet health monitors that conform to the body or transparent augmented-reality displays that fit the curve of a pilot’s helmet.

While scientists have recently performed lab demonstrations of chips that were flexible or transparent, they could only fabricate a few devices at a time.

The MIT researchers, in close collaboration with engineers at NY Creates at the Albany NanoTech Complex, created a fabrication process that uses standard semiconductor manufacturing techniques to generate flexible and transparent silicon-photonics chips on large-scale wafers.

To validate this platform, the researchers bent a single chip thousands of times around cylinders with various diameters — down to the width of a small screw — with no drop in performance. They also determined that looking through the chips would not cause much haze or distortion. 

“We’ve now developed a wafer-scale process that produces wafers that are mechanically flexible and optically transparent, enabling novel applications that weren’t previously possible with silicon photonics. We hope that, by working closely with our colleagues at NY Creates and using the foundry at the Albany NanoTech Complex, there’s the potential for us to make the platform accessible to other groups within our research community and open these new application areas to the field of silicon photonics as a whole,” says Jelena Notaros, the Robert J. Shillman Career Development Associate Professor of Electrical Engineering and Computer Science (EECS) at MIT, a member of the Research Laboratory of Electronics, and senior author of a paper on this fabrication platform.

Her co-authors include lead author Tal Sneh and Andres Garcia Coleto, EECS graduate students; Thomas Dyer and Kevin Fealey of the New York Center for Research, Economic Advancement, Technology, Engineering, and Science (NY Creates); and Milica Notaros PhD ’23. The paper appears in the journal Optica.

Flexible and transparent

Over the past decade, researchers have developed techniques to fabricate precise and highly reliable silicon-photonics devices at scale. 

They use advanced microelectronics foundry processes to produce 300-millimeter-diameter wafers with billions of nanoscale optical devices. But these methods yield silicon-photonics chips that are rigid and opaque.

“We realized that there are a lot of applications that would benefit from having a chip that is flexible and transparent,” Notaros says.

Scientists have previously made single silicon-photonics chips that were either transparent or flexible, but these techniques weren’t scalable. To address this scaling challenge, Notaros’ group recently demonstrated a foundry-scale process for making silicon-photonics chips on a flexible substrate.

Now, the team pushed these innovations even farther with a scalable process that produces 300-millimeter silicon-photonics wafers that are both transparent and flexible. 

Their fabrication process begins as if they were making a traditional, rigid silicon wafer. The researchers carefully deposit and pattern tiny optical wires known as waveguides onto this rigid silicon substrate. 

Then they bond a temporary silicon wafer on top. They flip the wafer over and remove all of the original silicon substrate from what is now the top of the wafer. They are then left with a flat layer of material with a thickness less than a tenth of a human hair.

“Thanks to the fact that we added that rigid temporary support before we flipped the wafer over, we can go all the way down so we are just left with the oxide and waveguiding layers,” Sneh says.

They use an adhesive to stick a thin, transparent polyester film on top of these ultrathin layers and “de-bond” the temporary silicon wafer from the bottom to remove it. 

This leaves them with a flexible, transparent wafer only a few microns thick that contains the oxide and waveguide layers needed to capture and transport light for silicon photonics.

“Because we are using stable 300-millimeter foundry fabrication tools, we can design systems with a very large number of devices and feel confident that they are going to perform up to specifications, which is extremely important,” Sneh adds.

The biggest challenge in developing this fabrication process was removing enough material from a large 300-millimeter-diameter silicon wafer to leave only a few microns of material behind.

During fabrication, stress on the wafer typically causes it to bow slightly, making this silicon removal process especially challenging. 

“As we were flipping the wafers over on these substrates, if the strain isn’t properly managed and the wafer isn’t perfectly flat, it is going to get ripples across its surface or even shatter in the fabrication line,” Dyer says.

The researchers carefully managed that stress by sticking to low temperature processes at or below 500 degrees Celsius.

They also had to find the right ordering and combination of removal methods. 

They used industrial processes to thin the silicon layer, but switched to a more precise selective chemical etch for the last bit. This ensured they would not damage the ultrathin layers left behind.

An eye on performance

The researchers performed three experiments to test different functionalities of these flexible, transparent silicon-photonics wafers.

First, they tested the optical performance of chips with integrated waveguides of different lengths to determine their waveguiding properties. 

Then they tested flexibility by bending a chip thousands of times around cylinders with different diameters. These experiments showed no degradation in performance even when they bent it around a cylinder about the size of a small screw. The device didn’t start to degrade until the researchers bent it around a toothpick several times.

“This experiment validated that the platform can be used for our proposed applications, performing even well beyond the metrics required for these intended systems,” Garcia Coleto says.

They also evaluated transparency by setting up a bionic eye and testing whether the chip would distort the user’s vision when placed in front of the eye. They found that the chip causes only minimal haze for the viewer and would not noticeably distort images the eye perceives when looking through it.

These characteristics could make these chips especially well-suited for enabling silicon-photonics systems for applications like curved augmented-reality displays that conform to a heads-up-display windshield or airplane pilot’s visor. In a pilot’s visor, for instance, such an augmented-reality display could replace the heavy bulk-optical systems that currently provide real-time information to help the pilot react to dangerous conditions.

In the future, the researchers want to add more complex components and functionality to the chips as they move toward enabling these and other new applications. They also want to refine the design to further improve waveguide efficiency and boost transparency performance.

This research was funded, in part, by the National Science Foundation, the U.S. Defense Advanced Research Projects Agency, and a MathWorks Fellowship. Wafer processing was performed at NY Creates, and chip dicing was conducted at MIT.nano.

Researchers from MIT have designed a new qubit architecture that enables qubits to interact with each other much more quickly while remaining very stable. This advance could someday help scientists build practical quantum computers that can run long, complex algorithms with high accuracy.

Qubits, which are the building blocks of a quantum computer, usually only store data and rely on other electronics to perform operations and communicate. But qubits are so fragile and error-prone that it is difficult for scientists to connect enough qubits before they lose their information and need to be reset.

The MIT team designed a dual-purpose qubit with two separate parts: one component that stores data and one component that interacts with other qubits and electronics. This design improves the reliability of the qubit and enables it to operate with a reduced error rate, so it can perform more computations in the same time span.

Their simulations indicate that this new qubit architecture could allow significantly faster and higher-fidelity operations than existing designs. 

While this research is still in its early days, it holds the potential to help scientists build large-scale, useful quantum computers that can solve real problems which are too difficult for traditional supercomputers to handle.

“This work feels like a big step. It is a new architecture that shows how much these systems can be engineered. We have taken two ideas and put them together in a way that can help us accomplish this qubit codesign that we are looking for, creating a pretty rare combination of the things we need to do quantum error correction,” says Alec Yen, who earned his electrical engineering and computer science (EECS) PhD this spring and is co-author of a paper describing the new architecture.

He is joined on the paper by lead author Jeremy Kline, an EECS graduate student; Stanley Chen, an MIT undergraduate; and senior author Kevin O’Brien, an associate professor in EECS and principal investigator in the Research Laboratory of Electronics (RLE). The work appears in Physical Review Applied.

A dual-purpose qubit

Just like the bits in a classical computer, quantum bits store information. But unlike classical bits, quantum bits have very short lifespans and can break down quickly when scientists connect them to make a quantum computer.

This degradation, known as decoherence, introduces errors in computations that rapidly build up, derailing long calculations before they are complete.

“The goal for doing all this is to build a fault-tolerant quantum computer where you can correct these errors as they happen, so then you can do long computations and actually do useful things with a quantum computer,” O’Brien explains.

To make qubits more reliable, the MIT researchers developed a new design that includes two separate but connected components: one which stores data and one which interacts with every other part of the quantum circuit.

This interaction component is like an arm that reaches out to other parts of the system, so the researchers call their design the “arm qubit.”

“It is engineered for these two, dual purposes — accomplished together by the data mode and arm mode — and these two goals really matter when you try to do quantum error correction,” Yen says.

Essentially, their design combines two different types of qubits. To make the data mode, they use one popular qubit design which has been known to have a very long lifespan, or coherence. 

The arm mode utilizes a different design that exhibits very strong interactions with other components such as a resonator, which is an electronic component that allows for readout of quantum computations. Readout is the process of measuring a quantum system’s state and translating it into a classical value.

The key to this new architecture is a special coupling unit the researchers previously developed, which they used to connect the data mode and the arm mode. 

Stronger coupling

Normally, coupling the modes together could cause unwanted interactions between them that would build up as more qubits are linked to the system. 

One way to avoid this mixing is to use a technique called nonlinear coupling, which occurs when two components are linked in such a way that changing the state of one causes the other to change in response. Nonlinear coupling is essential for running most quantum algorithms.

The special device the researchers used, known as a quarton coupler, enables very strong nonlinear coupling between the data mode and arm mode, which significantly reduces unwanted mixing. This coupling allows the qubit to perform operations faster before it decoheres.

“By dedicating the ‘arm’ component to coupling, we were able make a design that is scalable, robust to manufacturing errors, and still uses a quarton coupler to achieve strong nonlinear coupling,” Kline says.

When they tested the design in simulations, the arm qubit outperformed other superconducting qubit architectures by yielding state-of-the-art coherence time as well as faster operations and readout. 

The speed and reliability of this new architecture may accelerate quantum error correction, which is an important step in making quantum computers practical.

From here, the researchers plan to work toward fabricating the arm qubit so they can further study its properties and capabilities and integrate it into real quantum systems. 

“This work leaves me with a lot of suspense because our simulations are very promising. Next, we’ll need to see if we can make it, and determine whether we missed anything in the modeling or design. If we can fabricate this qubit, it could be a building block for future error-correcting quantum computers,” O’Brien says.

This work is funded, in part, by the Army Research Office, the Air Force Office of Scientific Research, a Doc Bedard Fellowship from the MIT Center for Quantum Engineering and the Laboratory for Physical Sciences. 

Each year, farmers around the world spend $80 billion on pesticides for their crops. Those pesticides impact not only harmful insects but also bees and beneficial bacteria in the soil. They can also run off into waterways and harm the environment. And, they are increasingly being linked to human diseases like Parkinson’s and cancer.

Amid growing awareness of those problems, pesticides made from living microbes are gaining popularity. Unfortunately, such microbial pesticides are often less effective, forcing farmers to choose between potential environmental damage and higher crop yields.

Now, Robigo is equipping naturally occurring microbes with more potent pest-fighting capabilities. The company, which was co-founded by Andee Wallace PhD ’20, uses technologies more commonly associated with medical applications, like RNA interference and CRISPR, to engineer self-replicating microbes that target plant pathogens more precisely than chemical pesticides and more effectively than other biologically based solutions.

“Chemical pesticides have been a cornerstone of agricultural production for the past 70 years, to the point that it’s nearly impossible to envision an agricultural system without them,” Wallace says. “But that’s the long-term vision we have: providing growers new tools to enable a food system that is in balance with the environment, and that is productive, resilient, and safe.”

In field trials across five states, the company has already shown its microbes offer comparable results to chemical pesticides. In one trial comparing Robigo’s product with another commercial microbial product last summer, Robigo’s system led to a 250 percent increase in crop yield.

“Many crops, like lettuce, are harvested by hand, and the grower told me if a disease reduces yield even by just 25 percent, it’s not economical for them to pay workers to harvest the field at all,” Wallace says. “Growers are just trying to produce enough food to feed everyone. That’s why they use pesticides in the first place. We’re trying to give them a better choice.”

Engineered biology for agriculture

Wallace did her PhD in the lab of Chris Voigt, MIT’s Daniel I.C. Wang Professor and the head of the Department of Biological Engineering. She joined the lab after working at Bolt Threads, a startup spun out of the Voigt lab that was designing a material for the fashion industry inspired by spider silk.

“I came into MIT knowing that I wanted to join Voigt’s lab,” Wallace says. “I was really enamored with biomaterials in general. There are so many examples of animals and organisms that make incredible materials that we humans can’t replicate.”

Wallace’s PhD focused on engineering microbes in an attempt to replicate intricate glass nanostructures produced by single-cell algae called diatoms.

Wallace enjoyed her startup experience and explored entrepreneurship throughout her time at MIT. But it wasn’t until after graduation that she reconnected with two MIT students, Jai Padmakumar PhD ’23 and Connor Sweeney ’21, and decided to start her own company.

The founders’ initial idea was to engineer microbes to deliver CRISPR to target and kill bacteria that are harmful to crops. They used a number of MIT resources to get the company off the ground, including the Venture Mentoring Service, MIT Sandbox, delta v, and the MIT $100K Entrepreneurship Competition. Sweeney was involved in the venture for about a year. Padmakumar left Robigo in 2022.

Today Robigo is addressing a problem of growing importance to the agriculture industry.

“Chemical pesticides are under incredible pressures: increasing scrutiny from consumers and regulators, and increasing pesticide resistance among pests, diseases, and weeds,” Wallace explains. “Over the past 40 years, only two new herbicide chemistry modes of action have been commercialized, so people are understandably worried. If we can’t develop new solutions, resistance is only going to grow and will leave growers without effective tools to protect their crops. I think biotechnology has the potential to solve that problem.”

Farmers hope so, too: In an attempt to address environmental and health concerns, they have increasingly turned to so-called biological pesticide solutions, which are mostly made from natural sources like plant extracts, microbe-derived natural products, and increasingly biotechnology solutions like peptides and RNA.

“They are safer and better for the environment, but currently they just don’t perform as well or as reliably as synthetic chemistry pesticides, so there’s a big distrust among growers,” Wallace says. “Growers are being asked to choose between high performance or safety and sustainability. Robigo is trying to solve that problem by giving them products that do both.”

Robigo’s ARGO biotechnology platform combines synthetic biology and proprietary computational design processes to engineer microbes that perform at a similar level to chemical pesticides, but with improved safety profiles for people and the planet. A key part of that approach is leveraging microbes’ self-replicating abilities to continuously produce and deliver bioactive molecules in the field over the course of the growing season.

The company has moved in recent years from delivering CRISPR to RNA-interference, or RNAi, which inhibits key functions in the pathogens they want to target.

Robigo also differs from other microbial pesticide companies in its approach. Wallace says other companies screen to discover new microbes with the properties they want, then cultivate those for sprays and other modes of applications. But these specialized microbes may not be able to thrive in, say, the microbiome of California farm soil where they’re needed. That means they may die off soon after being deployed. Robigo, conversely, focuses on equipping robust, industry-proven microbes with the ability to target specific pests and diseases.

“Our starting point is ‘What crops will this be used for? and ‘What diseases do we want to control?’” Wallace says. “To design safer products, we need to be direct in how we’re designing RNAi to target different diseases. Another layer of our technology is what we call RNAi stacking, where we combine multiple RNAi into a single microbe to broaden the spectrum of pathogens we can control with a single product.”

Lab to farm to table

Last year, Robigo ran field trials for its two lead products, with soybeans and lettuce across the U.S. Midwest and West. Working with third-party testing companies, they showed a single application of their microbes offered protection for crops over the entire growing season and matched the performance of the leading chemical pesticide at a fraction of the cost. 

“That’s very unusual for biological products, and even many chemical products, so we’re really optimistic about engineered microbes being a new solution that disrupts the conventional chemical pesticide paradigm,” Wallace says.

Wallace says Robigo is expanding fourfold this year and plans to expand even faster next year with the help of major agrochemical companies interested in more sustainable solutions. The company is also partnering to expand to other crops as it helps farmers around the world.

“There are a lot of opportunities we’re excited about, and we’re working with a number of partners as we scale,” Wallace says. “Over the past nine months, we’ve systematically used our ARGO platform to tackle new opportunities, and we have a number of products in the pipeline we’re working to bring to growers around the world.”

How do you build something that lasts? For MIT Assistant Professor Iwnetim "Tim" Abate, the answer is the same whether he’s reimagining how the materials beneath our feet can store energy and manufacture essential chemicals, or mentoring MIT’s future researchers: focus on the foundation.

Rocks provide an unexpected thread connecting Abate’s research and his approach to mentorship. His research brings together electrochemistry, materials science, and Earth sciences to explore how the materials that make up our planet can be harnessed to address some of society’s most pressing challenges in energy and sustainable manufacturing.

In one line of inquiry, his group uses Earth-abundant elements found in rocks, such as manganese and iron, to develop high-energy, low-cost, and more sustainable batteries. In another, they are exploring how the Earth’s subsurface itself could function as a chemical factory. By harnessing naturally reactive rocks, geothermal heat, and injected fluids, they seek to pioneer new ways of producing valuable fuels and chemicals underground, with lower external energy requirements and emissions than conventional industrial processes.

Although batteries and subsurface chemical manufacturing operate at vastly different scales, they share a common philosophy: understanding the intrinsic chemistry of Earth’s materials deeply enough to harness it for useful transformations.

While Abate's research spans a broad range of scientific disciplines, his approach to mentorship is guided by a simple principle: helping students lay the groundwork for their careers after graduate school. Rather than measuring success solely through publications or technical accomplishments, he strives to equip students with the scientific skills, resilience, curiosity, and perspective needed to navigate any path their career may take.

"I often think about mentorship through the image of a rock," Abate explains. "A structure built on rock can withstand storms and the test of time. In the same way, I believe the most important role of a mentor is not simply to help students complete a project or publish papers, but to help them build a strong foundation."

Abate puts this philosophy into practice through his investment in his students' growth as researchers, professionals, and individuals.

In celebration of his exemplary mentorship, Abate has been recognized through MIT's Committed to Caring initiative, a student-driven program that honors graduate mentors who foster supportive and inclusive research environments.

Building holistic relationships

Students often arrive at graduate school with different ambitions. Whether they hope to pursue academia, industry, entrepreneurship, or public service, Abate begins by learning about each person's long-term goals.

Each time a new student joins his group, he meets with them individually to discuss their aspirations and helps tailor aspects of their PhD experience accordingly. Students say these conversations continue throughout their time in the lab, with regular one-on-one meetings focused on both research progress and career development, homing in on their opportunities beyond MIT.

For students interested in entrepreneurship, Abate leverages his own network, introducing them to venture capital firms, philanthropic organizations, and collaborators working across academia and industry. He encourages his students to pursue internships, recognizing that experiences outside the university can strengthen both their research perspective and their future careers.

Students also emphasize his ability to connect them with the expertise they need to push research forward. Whether facilitating access to specialized instrumentation or identifying researchers with complementary knowledge, Abate actively builds the relationships that allow his students and their projects to thrive.

Despite leading a growing research group while balancing teaching responsibilities and launching a startup, nominators wrote that Abate "consistently [shows] up for his students."

He makes time for individual chats with students, subgroup discussions, and weekly lab meetings, all while actively seeking their perspectives on research challenges. "Tim is often curious [to hear] our point of view on research problems and actively looks for our feedback," reflected one nominator. 

This openness creates a synergistic environment where students are encouraged to help shape the direction of the group's work.

Creating space for ambitious ideas

Innovation, Abate believes, depends on more than technical expertise.

"Students need to know that it is OK to pursue ideas that may not work, and that setbacks are part of discovery, rather than signs of failure," he says. "My goal is to create an environment where ambitious ideas are welcomed, careful thinking is valued, and students know they have someone who believes in them through both successes and disappointments."

Students say this philosophy is reflected in the way that Abate approaches advising. Rather than directing every decision, he encourages them to think independently, remaining available whenever guidance is needed. His vast professional network often becomes an extension of that mentorship, opening doors to partnerships and expertise that help students tackle increasingly ambitious research questions.

This commitment to building strong foundations extends beyond his own research group. Since graduate school, Abate has worked to expand access to STEM education through his nonprofit Sci-Fro, which supports educational outreach across Africa. He has also contributed to broader efforts to strengthen scientific infrastructure and research institutions across the continent. 

For Abate, these efforts reflect the same philosophy that guides his mentorship: lasting scientific progress depends not only on individual discoveries, but also on investing in people, communities, and institutions that enable future generations of scientists to thrive.

Supporting the person behind the PhD

Abate regularly checks in during one-on-one meetings, asking how his students are doing and what support they need. He believes these conversations are an essential part of advising.

"Graduate school is one of the most formative periods of a person's life," he says. "While research is important, I don't think success should come at the expense of health, relationships, or personal growth."

He encourages students to build lives that remain meaningful beyond the laboratory, recognizing that the habits, friendships, and perspectives developed during graduate school often shape them just as much as their scientific accomplishments.

Through steady guidance, meaningful connections, and genuine care for each student's well-being, Abate demonstrates a passion for developing exceptional researchers.

"I hope they leave MIT with a strong foundation — both scientifically and personally — that enables them to navigate future challenges, lead with integrity, and build fulfilling lives wherever their careers take them."

The experience of transitioning from research based in theory to focusing on real-world application can vary significantly for different researchers. However, for two former MIT graduate students and a former postdoc, all now at IBM, working with the MIT-IBM Computing Research Lab (formerly the MIT-IBM Watson AI Lab) during their formative years enabled them to not only close the gap between education and employment, but also to generate ideas promising to business impact. 

Despite pursuing varied careers in quantum machine learning, reinforcement learning and artificial intelligence agents,and trustworthy and fair AI, respectively, Srinivasan Arunachalam, Zhang-Wei Hong PhD ’25, and Irene Ko PhD ’24 have consistently found ways to tackle problems defined by novelty and rigor, and translate them to systems with real constraints. Here, the MIT-IBM Computing Research Lab served as a conduit for research relationship building and the flow of their expertise to industry applications.

“Among all the industrial labs, I think MIT-IBM has way better academic collaboration policy and opportunity [than the others],” says Hong, an IBM research staff member with the MIT-IBM Computing Research Lab who began his PhD at MIT in 2020 in the Department of Electrical Engineering and Computer Science (EECS).

Hong has been captivated with reinforcement learning since discovering that DeepMind could play Atari and learn from raw screen pixels via feature engineering. During his graduate work with EECS Associate Professor Pulkit Agrawal, who is also a principal investigator with the lab, Hong sought to build on this: improving value function learning for reinforcement learning in video games, using “Montezuma’s Revenge” in Atari, in order to predict and optimize the policy performance of an agent. With the lab, Hong developed techniques to ground AI for more realistic applications and provide better reward feedback, which he applied to domains such as robotics, large language models (LLMs), and reinforcement learning for science. 

“I’m very excited about curiosity-driven exploration,” says Hong of the MIT-IBM graduate work that helped propel him into his profession. This, he says, allows agents to be inquisitive about new data, like humans, and perform a variety of tasks — from generating test cases to stress-test LLMs to exploring new environments. Now, as a mentor for students of his own, Hong continues to pursue similar lines of open-ended reinforcement learning research, leading him to investigate test-time training for agents and foundation models, and develop infrastructure for IBM’s agentic framework for enterprise tasks like chart reading and tool calling for database queries. This includes evolutionary computing to drive better optimization for exploration and leveraging neuroscience to inform deployment time model improvement. 

“If successful, I think that it would be a very useful system and framework for all of the practitioners in reinforcement learning, because it will be the first framework that enables a model to improve — self-evolve their model weights online at a deployment time,” says Hong.

Irene Ko’s research has also been value-driven, from a personal and professional standpoint. “I started to work [on trustworthy AI] with IBM researchers from day 1 in my PhD, because it was funded by MIT-IBM,” says Ko. This, she says, was particularly advantageous since her goals to develop frontier-safe, robust, accurate, and fair AI also align with that of MIT and IBM, closing the gap between development and real-world deployment. “That really strikes a balance between pure research and something that’s of industry standard or value.” 

Further, her MIT-IBM collaboration through her advisor in EECS, Joseph F. and Nancy P. Keithley Professor Luca Daniel, and IBM Principal Research Scientist Pin-Yu Chen, helped define the direction and parameters of her work to maximize impact, first in neural networks and later with foundation models and LLMs. After graduating in 2024, Ko joined IBM Research to continue her work on trustworthy AI as a research scientist. 

“The reason I chose to go into industry after my PhD, and IBM specifically, is that I found great joy in the collaboration during my PhD. That process, those five years, gave me very high rewards in personal fulfillment,” says Ko. “I wanted to continue the momentum.” 

Her current project focuses on finding pain points in current trustworthy methods that are not widely deployed in AI inference platforms. Unlike using low-rank adapters, which add extra steps to monitor and modify model behavior, her work on vLLM Hook provides a way to access internal model signals, like hidden states or activations, for decoding LLMs. This vector acts on transformer modules to analyze safety scores, such as identifying the likelihood of prompt-injection and hallucination. Here, Ko has developed a lightweight vLLM inference engine plugin framework to program the model internals that could provide significant cost savings over other methods. “I’m very proud of this project because this is really, as far as we know, the first bridge between the deployment and development in trustworthy AI with the inference engines.”

While Srinivasan Arunachalam has always dabbled in quantum research, he constantly explores other areas of theory, seeking to find quantum insights and deep math in unexpected lines of inquiry and papers. “Right off the bat, you don’t see it. You think, maybe this is just a vanilla problem, and then once you start investigating it further, you find some really interesting math that comes out of it, which I think is pretty cool,” he says. 

This drew Arunachalam to MIT as a postdoc in 2018 in the group of Professor Aram Harrow in the Department of Physics. With a learning theory-first perspective, Arunachalam looked for target algorithms, subroutines, and circuits where quantum speed-ups might be possible. Conversations with Isaac Chuang, the Julius A. Stratton Professor in Electrical Engineering and Physics and an MIT-IBM PI, led him to collaborate with the lab and IBM researcher Kristan Temme. 

With a seamless transition to IBM, Arunachalam more closely involved himself with problems that are potentially implementable on a near-term quantum device, keeping in mind constraints like nearest-neighbor architecture, noise, and simpler observable measurements. During this time, Arunachalam focused on quantum machine learning and areas where quantum computing would be superior to classical computing, increasingly prioritizing provability grounded in theory to heuristics. That MIT-IBM connection helped turn theoretical questions into concrete research directions, shaping work that culminated in two prominent papers: one on Hamiltonian learning, which gave rigorous guarantees for learning the dynamics of quantum systems, and another on quantum kernels, which provided theoretical evidence that quantum feature spaces can offer advantages over classical kernels under widely believed hardness assumptions.

Arunachalam also continued to expand his knowledge base by pouring himself into different branches of computer science to uncover structure in problems others may have missed. “One thing which I’ve been a huge fan of is exposing connections between different fields.” This has allowed him to explore learning quantum states — from completely classically simulatable quantum objects to the extremely complicated quantum objects.

Although Hong, Arunachalam, and Ko navigate different domains, they share an instinct: to move ideas across the space between what is possible in principle and what is useful in practice. In their own way, each is applying knowledge gained from collaborations, like that of MIT-IBM Computing Research Lab, to develop “killer applications” — a real-world use case that proves the underlying research can matter beyond the lab.

Self-driving cars are often controlled by deep learning models that sometimes fail in unexpected situations. For instance, the car might inexplicably brake and block the path of an oncoming emergency vehicle. A human driver or passenger may need to react rapidly to prevent a collision.

To help humans better anticipate a vehicle’s mistakes, researchers from MIT and autonomous vehicle technology company Motional developed a new method that provides clear explanations of the underlying model’s decisions.

Usually, the internal reasoning process of a deep learning model is opaque and difficult to understand. But the new method, called the Concept-Wrapper Network (CW-Net), translates that reasoning process into concepts that faithfully describe the autonomous vehicle’s decisions without altering its driving performance.

CW-Net explains the decisions of machine learning-based planners using understandable concepts, like “approaching stopped vehicle” or “close to cyclist.” These explanations can correct misconceptions drivers and passengers have about vehicle behavior and improve their situational awareness.

In road tests on a private track, CW-Net explanations helped safety drivers more accurately predict vehicle behavior; a larger simulation study with nonexpert users yielded similar results. These experiments show how CW-Net can provide important feedback for engineers as they troubleshoot in-vehicle artificial intelligence systems. In the longer term, this technique could boost the safety and transparency of autonomous vehicles, while building appropriate trust in drivers and passengers.

“This work shows how explanations are supportive to the human’s mental model and understanding of the behavior of a system, and how it could be used in engineering and development to improve the technology,” says Julie Shah, an MIT professor of aeronautics and astronautics, director of the Interactive Robotics Group in the Computer Science and Artificial Intelligence Laboratory (CSAIL), and co-senior author of the paper on CW-Net. “Unless we are building these technologies in a way that we can rely on and predict their behavior, then it is a shaky and unsafe foundation for their use.”

She is joined on the paper by lead author Eoin Kenny, a former MIT postdoc who is now a senior AI researcher at J.P. Morgan Chase; co-senior author Momchil Tomov, a staff research scientist at Motional; as well as Motional team members Akshay Dharmavaram, Sang Uk Lee, Tung Phan-Minh, Shreyas Rajesh, Yunqing Hu, and Laura Major, president and CEO of Motional. The research appears today in Nature.

Faithful explanations

Machine-learning-based planners act as the “brain” of a self-driving car. These powerful deep-learning architectures process data from the vehicle’s cameras and lidar sensors, generate a high-level summary of the vehicle’s environment, decide what the car should do next, and output a trajectory for it to follow.

The planners are usually black-box models, which means their internal decision-making process is so complex it is difficult to understand. This can leave scientists and safety drivers in the dark about why an autonomous vehicle made an unexpected decision, like phantom braking.

The researchers designed CW-Net to explain a vehicle’s decisions using understandable concepts, while ensuring those explanations accurately reflect the true reasons behind its behavior. 

“Especially in high-stakes settings like self-driving cars, it’s important that the explanations are not potentially misleading. Because CW-Net is causally faithful in how it makes decisions, that provides certain guarantees around the explanations,” Kenny says.

CW-Net is a “concept classifier,” an AI algorithm that has been trained to predict the high-level concepts that exist within input data. The researchers plug the CW-Net module into the middle of an autonomous vehicle’s existing machine-learning planner architecture.

It translates the model’s internal reasoning process into understandable concepts, like “approaching stopped vehicle” or “close to cyclist.” Then it forces the final piece of the planning model architecture to use those concepts when it decides what the vehicle should do next. In this way, CW-Net ensures the concepts faithfully explain the vehicle’s actions. 

At the same time, CW-Net uses the concepts it classified to generate clear explanations that are output along with the vehicle trajectory, in real-time.

“Instead of just wondering why the car stopped, having real-time data provides feedback that lets you test the system during deployment. You could also give that data to an engineer to potentially improve the system,” Kenny says. 

The researchers trained CW-Net to predict concepts using a dataset of 130 million examples of scenes from self-driving cars, with multiple labeled concepts in each scene. Using such a large, labeled dataset enables it to identify concepts accurately in a wide range of settings.

They also designed CW-Net to mimic the driving decisions of machine-learning-based planners, so the module would not negatively impact vehicle performance.

In the end, CW-Net generates accurate, understandable explanations without altering the original deep learning model.

Improving situational awareness

To test CW-Net, the researchers deployed the module on a real autonomous driving test vehicle (a Motional robotaxi) on a private track with a safety driver. They found that CW-Net helped the safety driver better predict how the vehicle would behave in surprising situations.

For instance, the vehicle consistently stopped when it approached a cyclist, and the safety driver assumed it did so because it detected that cyclist. But CW-Net explanations revealed that the model wasn’t properly configured to detect the cyclist and chose a trajectory that would have caused a collision. Instead, it stopped because its emergency braking procedure kicked in when it got too close.

Armed with this information about the model’s mistake, the safety driver could reduce speed or engage manual driving mode sooner in similar situations. This could also help engineers fix the model to avoid this failure in the future.

In larger online simulation studies using real driving situations captured on the roads of Las Vegas, the researchers saw similar results. CW-Net explanations significantly improved participants’ abilities to predict how an autonomous vehicle will behave.

In the future, the researchers could extend CW-Net so the module can cover more concepts and explore different training and design techniques that could boost performance and improve interpretability.

“Our study shows how crucial interpretability can be to these high-stakes environments, and how it should be on the mind of people as they are making AI in the future, for self-driving cars or other safety-critical environments,” Kenny says.