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.

Kendall Square in Cambridge, Massachusetts, is often cited as the most innovative square mile in the world. But when the Kendall Square Association gathered local stakeholders for its 18th Annual Meeting Thursday, the conversation focused on innovation primarily in the context of community.

The event, which took place at the nonprofit startup incubator The Engine, attracted people from across Cambridge, including business leaders, elected officials, local residents, and students, for a night of local food and conversation about the advance of technologies like artificial intelligence and what they mean for society.

“How can we make AI not an enemy of work, but a tool for workers to expand their capabilities and build new expertise or do new things?” asked MIT Institute Professor Daron Acemoglu in a keynote talk. “The world is at a point in time where we need new innovations and new adaptations to growing challenges.”

Acemoglu’s keynote was one of several presentations at the event, which also featured music from the Kendall Square Orchestra, a vote on new KSA board members, and awards for organizations making a positive impact through innovation.

KSA is a nonprofit made up of over 150 organizations across Kendall Square, from major companies to universities, research organizations, and the independent shops and restaurants that give Kendall Square its distinct character.

“People come to Kendall Square to make the world better,” said KSA Executive Director Beth O’Neill Maloney in her opening remarks. “People come here to improve the human condition, people come here to solve big problems, and people come here to innovate for good.”

Acemoglu, who shared the Nobel Prize in economics in 2024 for his work studying how institutions affect prosperity, spoke on the subject of his new book, “What Happened to Liberal Democracy?”

Acemoglu says liberal democracy is under siege as trust in liberal institutions decline, political polarization rises, and people lose faith in elected officials. He acknowledged that the shared prosperity liberal democracy promised has not always been delivered, especially over the last 50 years, but insisted it is the best path forward and in need of renewed support.

“I believe liberal democratic institutions, together with the norms that support them, are one of the greatest human achievements,” Acemoglu said. “It would be unthinkable for people who lived 300 years ago that we could live in societies consisting of hundreds of millions of people with diverse interests, different views, doing different things, and people still participate in politics and reach compromises, and we have state institutions that, by and large, actually provide services and do things for the people, and that we build opportunities for prosperity from which most people benefit while social freedoms are largely upheld.”

Cambridge Mayor Sumbul Siddiqui also spoke at the event. Siddiqui began her talk by telling the story of Cambridge residents working with biotech labs in the 1970s to create a groundbreaking policy on recombinant DNA research, which helped turn Kendall Square into the life science hub it is today.

“All these years later I think that story captures what makes Kendall Square so special: It’s our willingness to work together even when we face big challenges,” Siddiqui said. “It isn’t just the universities or labs, it’s also the culture of collaboration among nonprofits, governments, and residents.”

Broad Institute Chief Communications Officer Lee McGuire started his talk by reviewing a 2009 report on what made Kendall Square such a powerful driver of innovation, noting that factors like talent density, education, and a collective voice remain strong in the area today.

“This report was brilliant in that it captures so many of the special things that matter for a place like Kendall,” McGuire said. “It is the job of the Kendall Square Association and all of us to help protect those things, make sure we name them, understand them, and see them as essential elements of making sure this community stays what it is no matter what happens. As long as there’s a good idea, this is the place that’s going to get that idea moving and scale it up to change the world.”

Suzy M. Nelson, MIT vice chancellor for student life since 2016, will retire from the Institute this month. Nelson leaves MIT with a strong, coordinated Division of Student Life (DSL) built from years of close partnership with student leaders. One of her most noteworthy contributions is the elevation of student well-being, now integral to the Institute’s educational mission. 

Under Nelson’s leadership, MIT created a culture of well-being, building out a strong support network, enhancing health promotion efforts, and expanding spaces dedicated to student wellness. These efforts emphasized the importance of students finding a sense of purpose, developing positive relations, and flourishing in mind and body. 

Working with colleagues from across campus, DSL advanced new ways to coordinate student support services for undergraduate and graduate students experiencing significant challenges. MIT students now increasingly turn to programs such as Student Support Services (S3), which serves 82 percent of all undergraduates by the time they graduate, and 60 percent every year. A new CARE team coordinates resources that undergraduates, graduate students, and the MIT community rely on in a crisis, and has quickly become the thread that weaves together offices caring for students. 

Nelson prioritized student residences as the cornerstone of the MIT campus experience. She led an effort to develop architectural principles for housing to guide construction and renewal planning; partnered with alumni from fraternities, sororities, and independent living groups to undertake a comprehensive review of chapters’ deferred maintenance and facility needs; expanded services to the residential communities; and developed close working relationships with the faculty heads of house. 

“Suzy has been a wonderful leader and humane presence on our campus,” says David Mindell, head of house for Edgerton House. “She has partnered with the heads of house through trying times, as well as triumphs, and for that we are grateful. The students and their welfare have always been her first priority.”

Additionally, Nelson oversaw the construction of three new campus residences: Graduate Tower at Site 4 in 2020; New Vassar in 2021; and Graduate Junction in 2024, an innovative third-party partnership. The additional on-campus graduate housing increased capacity to accommodate nearly half of MIT’s graduate students and exceeded MIT’s commitment to the City of Cambridge. The division also worked on the renewal of four residences — New House (completed 2018), Burton Connor (completed 2022), East Campus (completed 2025), and now McCormick Hall (targeted for completion in 2028).

Beyond housing, Nelson advanced a major transformation of the core of the Stratton Student Center, leading to the opening of the Wellbeing Lab, new dance studios, improved lounge spaces, and refreshed offices for student organizations.

The Department of Athletics, Physical Education, and Recreation has played a critical role in MIT’s well-being efforts, incorporating wellness modules into physical education and wellness classes and offering recreation, intramural, and club sports opportunities to both undergraduates and graduate students. Additionally, the Institute’s athletics programs thrived during Nelson’s tenure, with MIT Engineer teams earning four national championships in 2024-25 and the number of All-Americans increasing significantly over the last five years. MIT remains the all-time leader in Academic All-Americans across all NCAA divisions. Nelson also encouraged expansion and renewal of athletic facilities, providing all students the opportunity to train using state-of-the-art equipment. 

Nelson’s leadership shaped the Institute’s approach to community standards. She helped refine student conduct processes and streamline the Committee on Discipline process. She also championed new policies supporting free expression and launched a new Dialogue Collaborative focused on coordinating efforts that promote dialogue across the campus. 

“Suzy is a remarkable leader and she has been a potent positive force within the MIT community. Her fusing of strategic vision, collaborative skill, and deep commitment to the students, DSL staff, and broader MIT community have been the hallmark of her extraordinary leadership,” says Erich Caulfield, chair of the Student Life Visiting Committee. “She has been a joy to work with, and her incomparable wisdom has been invaluable during this unprecedented period in student life at MIT and in higher education nationally."

Highly responsive to the changing needs of students, Nelson partnered closely with the Committee on Student Life and student leaders from the Undergraduate Association, Graduate Student Council, Dormcon, and the Graduate Housing Advisory Committee, as well as student representatives for individual projects. For example, Nelson created a working group to improve how student organizations gain recognition, manage membership, and apply for resources. She has worked closely with students on expanding access to nutritious food and supporting food security, and, this past year, she engaged students to help assess MIT’s dining services.

“Suzy brought her visionary leadership and humanity to DSL over this past decade to create profound, positive, and lasting change. All of us at the Institute, and especially our students, have benefited — and will continue to benefit — from the Nelson Era,” says Chancellor Melissa Nobles.  

Throughout her career at MIT, colleagues and students came to know Nelson for her collaborative style. As she navigated crises over the past decade, most notably the Covid-19 pandemic, she always convened students, faculty, and staff. And she found new ways of elevating staff, creating the Staff Engagement Advisory Board, a group of 45-plus volunteers who support the division’s positive work environment and promote professional development opportunities. Her team in DSL is grateful for the many ways she strengthened and professionalized the division, and they will continue to abide by her unwavering belief that students should help shape the communities where they live and learn. 

Nelson says of her retirement, “I am grateful to have served with so many exceptional faculty, staff, and students over these past 10 years. Because of their focused dedication to being here for students, I have every confidence that the transition to new leadership will be seamless.”

MIT will soon welcome Akirah Bradley-Armstrong as the next vice chancellor for student life. She most recently served as the vice chancellor for student affairs and success at the University of California at Santa Cruz, and has more than two decades of experience working in student affairs. She will start in the position on Oct. 19.

When corporate leaders discuss climate strategy, the message that still dominates is that, despite competing priorities and external pressures, sustainability matters. 

According to the 2026 State of Supply Chain Sustainability Report from the MIT Sustainable Supply Chain Lab, which surveyed 1,810 supply chain professionals across 91 countries, 83 percent rate sustainability as important or very important to long-term business success, the highest level in the report's seven-year history.

But commitment, the report reveals, is not the same as capacity. While 59 percent of respondents report becoming more committed to sustainability following recent U.S. policy shifts that deprioritize environmental concerns, the same organizations are being forced to compromise those commitments. Sixty-eight percent report that tariff uncertainty has compelled them to delay, scale back, or abandon sustainability strategies. Nearly one-third have seen tariff-related disruptions actually increase their operational emissions.

"Commitment is strong, measurement is expanding, and AI is accelerating," says Sreedevi Rajagopalan, director of the MIT Sustainable Supply Chain Lab and lead author of the report. "What comes next is harder: converting commitment into accountability, operational integration into strategic decision-making, and incremental decarbonization into genuine transition."

The result is a supply chain sustainability agenda stuck between conviction and constraint. 

Beyond tariffs, more than eight in 10 respondents across North America, Europe, and Asia agree that ongoing geopolitical conflict is reducing the effectiveness of their organization’s climate action. Organizations built decarbonization plans on assumptions of stable trade and relative peace, yet neither assumption is holding.

The external disruption reveals an internal vulnerability. Among companies with publicly stated sustainability goals, only 51 percent have established a formal cross-functional sustainability team, which is the governance structure most strongly associated with embedding sustainability into daily decisions. In addition, only 33 percent place sustainability oversight at the board level. Public ambition, in other words, is vastly outrunning internal accountability infrastructure. Companies with formal cross-functional teams report sustainability is "frequently or always" built into decisions 57 percent of the time. Where responsibility sits in a single function, that drops to 13 percent — a 44-point gap illustrating the return on governance investment.

Sustainability is embedding most deeply where companies make daily operational choices: transportation and logistics (39 percent), procurement (35 percent), warehousing (34 percent), and product design (33 percent). However, it thins out dramatically in decisions that lock in long-term emissions: capital investment, network design, and supply chain risk planning. This reveals that companies are becoming better at operating more sustainably, but not at designing sustainable supply chains in the long term.

The artificial intelligence story compounds this tension. Seventy-eight percent of North American organizations report current use of AI in sustainability initiatives, and 87 percent of adopters report positive impact on sustainability goals. Eighty-six percent have reduced material waste, 81 percent have reduced energy use, and 63 percent use AI to estimate missing supplier emissions data. However, among mature agentic AI users, only 53 percent report a reduction in total greenhouse gas emissions. Twenty-nine percent report an increase, suggesting that operational efficiency does not automatically translate into lower net emissions.

This tension is particularly visible in warehousing. Among advanced and fully automated warehouses using AI, 78 percent report higher energy use since implementing AI, and 76 percent report higher waste, yet only 24 percent use AI-driven real-time monitoring to track environmental consequences. Many organizations may not realize their energy footprint is climbing until well after implementation.

Another key area of the report is freight transportation. Nearly every freight company is reducing emissions through route optimization, load consolidation, and driver-behavior programs. However, 43 percent of fleets still operate mostly or entirely on conventional diesel and only 8 percent have completed a full transition to low-emission energy. The divergence between regions reveals why: In Europe, 84 percent of freight companies report that recent policy changes have positively affected decarbonization strategy. In North America, only 44 percent do. When policy signals are unclear, capital-intensive technology transitions stall. For example, battery-electric vehicles are at scale in 60 percent of European companies, but only 36 percent in North America.

The report reveals that one bright spot has emerged in Scope 3, or indirect, emissions collaboration. Participation in joint supplier engagement and decarbonization programs surged from 31 to 58 percent in a single year. In addition, partnerships to manage shared suppliers' emissions more than doubled, from 23 to 55 percent. The share working entirely alone collapsed from 18 to 6 percent. The report, supported by the C.H. Robinson Foundation, Breakthrough, and CO2 AI and Dassault Systèmes, reveals that companies are recognizing that Scope 3 cannot be solved in isolation. 

Across all these areas, the consistent pattern the report reveals is that corporate commitment is not the prevailing constraint to ongoing sustainability efforts. Rather, governance, accountability structures, technology infrastructure, and measurement capabilities are the current obstacles. Organizations that close the external gap by building resilience against disruption and the internal gap by hard-wiring accountability into structure will be the ones whose commitments survive. Those that do not will find that ambition without accountability moves the needle very little.

As an embryo develops from a small cluster of stem cells, those once “blank slate” cells begin to take on more specialized roles like brain, liver, or muscle cells, and organize themselves into three-dimensional structures such as tissues and organs.

The fate of each cell — what type of specialized cell it will become — depends on which genes are turned on or off in the cell. These patterns of gene activity shape the cell’s structure and function, enabling it to take on a specific role in the body.

But this decision isn’t up to individual cells. They constantly send and receive chemical signals to and from neighboring cells, which help them understand where they are, what stage of development they are in, and what they should become. These messages spread through multi-step sequences called signaling pathways, which translate external signals into specific changes in gene activity in the cell.

For researchers, being able to retrace the sequence of instructions a cell has received would offer a powerful way to understand how tissues develop and how these processes go awry in disease. 

But this has been difficult to achieve because scientists have long assumed that the effects of signaling pathways vary widely across cell types, meaning they would need to map each pathway separately in each cell type — an arduous and painstaking process.

Now, in a new study led by MIT assistant professor of biology and Whitehead Institute for Biomedical Research member Pulin Li and graduate student Nicholas Hutchins, researchers have discovered that each signaling pathway leaves behind a unique “fingerprint” — a distinctive pattern of gene activity that reflects the particular signals the cell has encountered.

Importantly, these fingerprints are consistent across different cell types for the same signaling pathway, which means that instead of mapping each cell type separately, scientists can reconstruct signaling histories across many cell types using these pathway-specific fingerprints.

This discovery was made possible through a machine learning model called IRIS. This model can detect fingerprints of different signaling pathways and pinpoint which signals a cell received at different stages of development inside an embryo, even for cell types it hasn’t encountered before. 

This artificial intelligence-driven approach marks a major advance over traditional methods, which require researchers to experimentally test every pathway in every possible cell type, and opens up the possibility to comprehensively map the signaling histories of every cell inside a mouse or human embryo at an unprecedented scale.

“Think of voice recognition systems like Siri, which are trained mainly in English, but then use that training to help them recognize other languages,” says Li. “This is called transfer learning, and this is why IRIS can work across many different cell types.”

The researchers’ detailed open-access findings, published in the journal Nature Methods on Sept. 8, could accelerate stem cell engineering for regenerative medicine and improve the creation of organoids — miniature, 3D models that mimic real organs — for studying disease mechanisms and testing new drugs. This is because once researchers learn the pattern of signals that drives a stem cell to become a specific cell type, they can recreate those signals to control the fate of stem cells in a lab or medical setting.

IRIS is a neural network-based model, which, in essence, is an AI-driven system designed to recognize patterns in complex data, similar to how our brains spot patterns. It examines a cell’s overall gene activity and estimates which signaling pathways were likely “on” at specific times during development.

Li and Hutchins trained IRIS on a large experimental dataset that measured how thousands of human embryonic stem cells responded to dozens of combinations of six major signaling pathways at multiple stages of development. This created a comprehensive map, or atlas, of how signaling combinations influence cell behavior. 

The team then tested IRIS using single cells from mouse embryos during gastrulation, a stage when cells are rapidly branching into different fates. IRIS could accurately predict when and where specific signaling pathways would activate in cells destined to become part of heart, gut, muscle, and spinal cord tissue. This variation in signaling molecules is what guides cells to form the right structures in the right places.

With this approach, scientists can not only begin to understand the fundamentals of cell-to-cell communication, but also gain a practical roadmap for guiding stem cells into specific, functional cell types in the lab.

When the researchers employed IRIS to identify signals needed to create a cell type critical for lung development, the model predicted that activating a specific signaling pathway would encourage lung-specific development. Experiments in mouse embryos confirmed the model’s prediction. By identifying the precise combinations of signals that drive lung cell development, they can more reliably generate accurate, lab-grown models of lung tissue.

“In these ways, IRIS is helping us decode the language cells use to talk to each other at a much faster rate than we could realistically achieve through experiments,” Hutchins says.

These improved models would allow researchers to study diseases like asthma, lung cancer, and pulmonary fibrosis, in which lung tissue becomes scarred, often without a known cause. They can then use these models to test potential therapies, and ultimately design regenerative treatments that can repair the damaged tissue.

With rising temperatures, the buttressing ice shelves at the edges the Antarctic ice sheet are melting and calving off at record rates, leaving behind sheer cliffs of solid ice. The stability of these exposed ice cliffs could influence how Antarctica further breaks apart and contributes to sea level rise around the world. 

Scientists have assumed that an ice cliff’s stability depends on its height: Ice cliffs that are taller than 100 meters — about the height of the Statue of Liberty — should collapse under their own weight. And once a cliff sloughs off, it should expose even taller, inland cliffs that collapse in turn. At least, that’s been the idea.

But researchers at MIT and the Woods Hole Oceanographic Institution (WHOI) have found that the stability of an ice cliff depends on more than its height. And even when a cliff topples, it doesn’t necessarily set off a runaway collapse. 

In a study appearing today in Science, the team presents an analysis of over 15,000 ice cliffs in Antarctica that were monitored by satellites over a four-year period. They observed how the ice cliffs and the larger ice sheet evolved over time, and found that those ice cliffs that collapsed did so at a much faster rate than scientists had predicted. But these collapsing cliffs also quickly stabilized, instead of continuing to erode the larger ice sheet. 

“We found that when the cliffs were retreating, they were doing so faster than what has been modeled, which would mean more ice loss and higher contributions of sea level rise,” says study author Caroline Needell, a graduate student in the MIT-WHOI Joint Program. “But because we found that the retreat was short-lived, and all the glaciers eventually stabilized, that would suggest that things won’t be so bad. We don’t see this runaway effect.”

The study’s co-authors are Catherine Walker, assistant scientist at WHOI, and Jeremy Bassis, associate professor at the University of Michigan.

Ice height

In glacial science, ice cliff collapse is described by a process known as marine ice cliff instability (MICI), which assumes that if an ice cliff exceeds a critical height, it will collapse. If then a taller cliff is exposed, it too should collapse, triggering a cascade of instabilities that shrink and erode the ice sheet over time. 

In previous studies, scientists have characterized MICI mainly by modeling the physical properties of ice, scaling these relationships up to the size of an ice cliff and combining these predictions with a few observations of actual ice cliffs in Antarctica. Through these efforts, scientists had determined that ice cliffs become unstable and collapse when they are taller about 100 meters.  

“Predictive ice-sheet models that have been used so far to represent ice cliff instability just take into account cliff height,” Needell notes. “That’s the only variable that matters. And once the cliff height reaches the critical threshold, usually 100 meters, it should initiate this runaway effect of ice sheet retreat.”

But she and her colleagues wondered: How tall do ice cliffs really get? And when they collapse, do they really set off a runaway retreat? 

To answer these questions, they looked to data collected by NASA’s Ice, Cloud, and Land Elevation Satellite-2 (ICESat-2) — a satellite that passes over the Earth’s polar regions and uses a laser altimeter to measure the height of ice sheets, glaciers, sea ice, and the level of the ocean in and around Antarctica. 

The team collected the satellite’s data from 2019 to 2023, which included measurements every three months of over 15,000 ice cliffs. From this dataset, they observed that the average ice cliff was about 30 meters tall — much shorter than the proposed critical height. But they did see a few rare ice cliffs that exceeded 100 meters. 

They also looked for other features that might influence a cliff’s stability. Interestingly, they found that ice cliffs that more or less stayed in place over time also had thicker mélange — a term for the slushy mix of broken icebergs and sea ice that gathers at the front of an ice cliff. 

“We measured mélange height and found taller cliffs have thicker mélange, which could be a built-in stabilizing effect,” Needell explains.

The satellite data also showed that the height of the surrounding water can stabilize a cliff: The deeper the water, the more it can buoy and push back against the cliff, keeping it from collapsing. 

Glacial speed bump

“After we did that global analysis, we thought, ok, let’s zoom in on the tallest cliffs and see how they’re behaving,” Needell says.

The team focused in on four regions around the Antarctic ice sheet where the satellite captured the tallest ice cliffs, each about 100 meters tall, with one that was 127 meters tall. They compiled measurements of these four cliffs and how they evolved every three months, over four years. 

From these measurements, they saw that indeed, the tallest cliffs collapsed, and the ice sheet in those regions retreated, at a surprisingly faster rate than MICI models had predicted. 

“But that retreat was short-lived,” Needell notes. 

After that, the ice spread back out toward the ocean.

“It’s this rapid retreat, punctuated by periods of stabilization,” Needell says. “That was puzzling to us. Then, we thought: We need to look at all the things that affect cliff height.”

The first thing they considered was the topography of the land underneath the ice. If that topography is relatively flat, then the slope of the overlying ice sheet reflects that actual height of the ice itself. But if there’s a bump or rise in the land, the ice sitting directly over that bump is in fact shorter than it appears. The topography then could act as a literal speed bump, thinning the overlying ice, and making it less likely to collapse. 

When the researchers checked topographic measurements of Antarctic, they found that indeed, ice cliffs stabilized around regions where the land happened to rise up. 

They also found that the taller a cliff is, the faster they “thin out” — as they collapse, once tall cliffs settle into shorter, more stable cliffs that prevents further collapse. 

“Now we know what to look for,” Needell says. “We really need to know the bed topography, and we need to know the ice thinning rate, in order to get ice cliff instability right, because this value really matters in the context of projecting sea level rise.” 

This research was supported, in part, by NASA, the U.S. National Science Foundation, and the DOMINOS project, a component of the International Thwaites Glacier Collaboration.

Around the world, many people report distrusting their government. In some places, such as Latin America, North America, and Africa, trust in government has declined in recent decades. How can governments rebuild this trust?

New research from the MIT Governance Lab and the behavioral science research organization Busara examines how government partnerships with trusted private organizations can shape public trust in government. In Kenya, county governments have worked with a telecommunications company to develop an online public service portal. A survey of people in and around the capital city of Nairobi found that highlighting the role of the company increased people’s willingness to use the portal, but hurt their trust in government. 

The results of this work, published in the September issue of the journal Government Information Quarterly, can help researchers understand how government digitization projects and relationships with technology companies affect public trust.

Trust in government and online public services

Studies have suggested that without public trust in government, people are less likely to follow rules and laws, and governments have a harder time solving problems that require public cooperation. Research has linked low levels of trust to people not paying taxes, lower voter turnout, lower Covid-19 vaccination rates, and higher Covid-19 infection rates.

In Kenya, a 2023 survey found that only 43 percent of people trusted the government. Corruption, poor service delivery, and flawed elections have likely hurt public trust. The Kenyan government is also a pioneer in digital technology for governance. Mobile money is widely used, and a centralized service delivery platform called eCitizen has made thousands of digitized services available to citizens. Such digital services can allow governments to serve people more efficiently and, if executed well, can build trust in government competency.

Does who gets credit for making a digital tool impact people’s trust?

Recently, Kenya’s 47 county governments partnered with Safaricom, a highly trusted telecommunications company, to create an online portal called My County Hub. The portal aims to allow people to request government emergency services, pay for business permits and parking, receive information from the government, and more.

In the new study, researchers told the more than 11,000 online survey participants how they could access My County Hub and some of the services it would offer. Some people were told the county government and Safaricom were both responsible for the portal, while others were told the government alone or the company alone was responsible. The survey also asked people about their trust in the county government before and after they received this information. The researchers hypothesized that people who were told the partnership was a public-private collaboration would have more trust at the end of the survey, with people’s trust in Safaricom transferring to the government partner.

Overall, people expressed more trust in government after being told about My County Hub, suggesting that digitization itself could be tied to increasing trust. However, trust did not seem to transfer from Safaricom to the government: people who were told about the partnership did not come to trust the government more than those who were told that the government alone was responsible. Trust sometimes transfers in settings where people don’t have strong beliefs about the untrusted group. But people tend to have strong beliefs about government, which could make it harder to sway their opinions.

Crediting the company could hurt people’s relationship with government

People who were told Safaricom alone was responsible for the portal expressed a greater willingness to use the portal and more interest in learning more about it. But relative to people who were told the government alone made the portal, the people in the Safaricom group had lower levels of trust in government afterwards. This could be because people viewed Safaricom as doing the work of the government, which could lower their faith in the government’s competency. 

The results suggest that while emphasizing the role companies play in developing digital services could encourage people to use these services in the short term, it could also erode people’s confidence in the government in the long term. People with higher trust in government were also more willing to use My County Hub in the future, so long-term trust in government could be important for service provision.

The researchers also noticed that people told about the partnership did see their trust in government increase if they had high levels of trust in the government and Safaricom at the start of the survey. And the decreased trust in government in the Safaricom group was driven by people who trusted Safaricom but distrusted the government at the start. 

These findings are in line with prior research showing that preexisting trust levels influence how trust changes during digitization efforts. In the future, research on people who distrust the government could look at other strategies for building trust through digital services.

The study’s authors include Lily L. Tsai, Ford Professor of Political Science and the director and founder of the MIT Governance Lab, and MIT Governance Lab Research Director Lula Chen.

As artificial intelligence use becomes more widespread, scholars are working to understand its effects on the people who use it. MIT Professor Eric So offers one of the most complete syntheses yet, in his new book, “The Collision: What AI Does to Us,” published this month by W.W. Norton. Recognizing that AI offers opportunities to complete work faster, and in some cases has been turned into useful domain-specific tools, So nonetheless looks at the costs of relying on it too broadly. He contends that heavy use of AI can lead to a kind of cognitive atrophy, in which people fail to develop and retain essential thinking skills. In general, he says, evidence shows that our ability to learn gets dramatically better when we wrestle with material, work through initial failures, keep iterating, and in the process acquire better thinking skills while retaining more knowledge. 

These conclusions are drawn from recent research by So and others, in academia and industry, including analysis from AI firms themselves, such as an Anthropic study showing that about half of queries from college students were low-effort queries for simple answers. So also weaves personal reflections into his narrative — having long studied financial markets and behavior, he is familiar with analyzing how new tools and information affect our decisions. MIT News talked to So, who is the Sloan Distinguished Professor of Global Economics and Behavioral Science in the MIT Sloan School of Management, about the book. 

Q: What prompted you to write “The Collision”?   

A: I wrote this book in response to patterns I was seeing in myself and other individuals. I open the book with this story where I was instinctively reaching for AI to help me think through problems I normally would have handled myself. A colleague had reached out for assistance and I had essentially taken her question and written half a sentence, then said, “ChatGPT, help me finish this professionally.” I made a few light edits and sent it off. And I sat back and really felt incredibly concerned about how I had removed my own thinking from the process.

I have found students are outsourcing their thinking to AI, rather than thinking through things themselves, turning to it to answer homework that would normally help them build mastery in a subject. We need to start understanding what this technology is doing to us and how it interacts with human cognition and our psychology, and give people some practical insights. 

Q: Part of that means understanding the incentives people have to use AI, or how the allure of AI takes hold, even when there are long-term costs, in your view. 

A: There are converging factors drawing us to AI. For example we have a real biological instinct to conserve our mental energy, and socially, we face situations where AI is used by other people to race ahead, and it rachets up the pressure on everyone else to use AI. Combine that with the problems detecting AI use in a practical way, and all these factors create what I call “AI gravity,” this constant pull to use AI. It’s not simply a matter of telling people to think for themselves. At this moment we’re in, AI is colliding with society.

Q: So we need to look at these systemic factors, not just individual decisions, when it comes to AI use, like businesses urging employees to use it. 

A: Many organizations are not just suggesting use of AI: It is now the basis of your evaluation, meaning if you want a bonus or promotion, it is incumbent upon you to use these tools. I think a subtle process is taking hold in academia, where people say, “My future employer will want me to use AI, so I’d better start using it now.” But the problem is that as you start to use AI now, we’re losing touch with the practices and capabilities we’re seeking to build, and in turn that makes us more dependent on using AI in the future, and it becomes a self-fulfilling prophecy.

Q: What is AI doing to our minds? 

A: AI is not just changing what we can do, it’s changing who we are as thinkers. AI removes a lot of the friction necessary to build foundational mastery. It is causing many people to lose their capacity for pushing through problems. AI offers instant relief, and we become accustomed to that. Many people are turning to it as an access point for information, so it’s affecting our ability to remember. Also, AI is changing our interpersonal relationships. A key point in the book is that these models are simultaneously sycophantic, meaning they want to make us happy because that’s how they drive engagement and get positive feedback, but they are also really persuasive. We lose the ability to connect with other people because we’re so accustomed to this frictionless environment where the models always agree with us.

Q: Your book emphasizes the value of “productive struggle.” You write, “Struggle doesn’t mean you’re suffering.” Instead, we get a handle on things by struggling to learn them. 

A: We need to recast how we think about struggle, as something that is formative rather than a problem to be solved. I draw a lot of parallels in the book between some of the societal changes that occurred post-industrialization, when suddenly people no longer needed to be physically active to be economically productive. We saw more people suddenly develop physical exercise as something we do intentionally, that required reframing of what physical struggle meant. And I think we need that sort of shift about what we think about cognitive struggle. 

Q: Given all this, what can we do to maintain some reasonable level of cognitive effort today?

A: At the personal level, I emphasize four principles for navigating this age of AI. One is embracing the struggle, as we’ve talked about. The second rule is: Value who you are without AI. I’ve seen so many presentations over the last couple of years where someone has put together this very polished-looking presentation with Claude, but when they present it, it’s clear they didn’t put in the work necessary to understand the material. The formative struggle creates what I refer to as an “inseparable task bundle,” since making the slides allows me to engage in substantive discussion. In the absence of valuing who you are without AI, we get the product, but not the ability to capitalize on it.

The third rule is about reinvesting your cognitive surplus. There are a lot of tasks for which AI can save you time — rote activities — but it matters what you do with those time savings. If you’re just funneling the same energy into the same tasks, you’re missing out on the potential transformational benefits of AI. The most innovative changes are not the ones that simply allow us to automate, but stretch the capabilities of  humanity in a productive way and build expertise. Or, people can channel that freed-up time to focus on family, learn a new language, cultivate skills. 

The last rule is about using AI as your cognitive trainer. AI is this dual force, where it is very bad for thinking if you use it as an answer machine, but can be great for learning if you use it as a tutor. AI can be tailored to you. However, it requires that you seek out the friction of having a tutor, relative to having an answer machine. And that requires motivation. 

This is a very challenging moment, and the problem is one we need to take seriously. 

More than 100 people filled the MIT Chapel on Sept. 24 to pay tribute to Assistant Professor Thomas Rose, the MIT archaeometallurgist who died last month following a bicycle crash in Cambridge, Massachusetts.

Faculty, students, and staff affiliated with the Department of Materials Science and Engineering (DMSE) and the Center for Materials Research in Archaeology and Ethnology (CMRAE), along with family and friends, gathered for a memorial service where colleagues spoke about the 37-year-old, who joined the DMSE faculty in January and had just started building a research program at MIT. They remembered him as a brilliant researcher and generous colleague with an extraordinary breadth of expertise and eclectic passions ranging from rowing to archery. 

A portrait of Rose sat on an easel next to the chapel’s flower-covered altar. A cellist played as guests entered and took their seats.

Chaplain to the Institute Thea Keith-Lucas opened the service by reflecting on the loss felt across MIT. 

“We honor the unique gifts that he brought to this world. We honor the ways that he was growing, and the possibilities that were cut short. We honor the small, unseen moments of courage and connection that create a life,” she said.

Finding a unicorn

To give a sense of what Rose meant to DMSE, department head Polina Anikeeva briefly explained archaeometallurgy — the study of how ancient societies produced and used metals — and its development at MIT, beginning in the 1960s. The field flourished after the establishment of CMRAE in the 1970s, but over the following decades it began to contract in the United States, and MIT had difficulty finding people with an increasingly rare set of skills. 

“We were looking for a unicorn. We were looking for two decades,” Anikeeva said. When Anikeeva and other DMSE faculty reviewed Rose’s resume in 2025, they were impressed by his unique training in geoscience, metallurgy, and archaeology and met him soon after. “We saw a brilliant, creative colleague — unassuming, intellectually fearless, adaptable, resourceful, and eagerly looking into the future.”

Mike Tarkanian, senior lecturer in DMSE and associate director for CMRAE, described the Zoom interview with Rose that kicked off the faculty search. Rose spoke fluently about his work, interests, and research vision — without the slides or notes that candidates typically rely on.

“I hung on every word — but it was so unusual and so risky that I wasn’t sure how it would be received,” Tarkanian said. “As soon as the call ended, my phone rang and it was a colleague who was also on the Zoom and he said, ‘That’s our guy. That has to be our guy.’”

Anikeeva also spoke of the enthusiasm he brought with him every day at MIT, noting his “almost childlike curiosity and his enormous positivity.”

“There were no obstacles, only opportunities and possibilities.”

To honor Rose and carry forward the work he had begun at MIT, Anikeeva said DMSE would launch the Thomas Rose Fellowship for Excellence in Archaeometallurgy, dedicated to supporting the most promising young scholars at the PhD and postdoctoral levels, “so that we can continue this field for the world and for the U.S. and for the memory of Thomas.”

A generous spirit

Speakers also remembered Rose’s kindness, generosity, and empathy toward everyone around him. Tarkanian stressed Rose’s commitment to community and helping others, noting his contributions to research tools used by archaeologists worldwide. He gave an example from Rose’s undergraduate years in his native Germany, when he gave tutorials for first-year archaeological science students who were struggling with chemistry and math. A former student told Tarkanian that an email Rose sent inviting students to the tutorial ended with, “I guess I’ll be more nervous than you folks tomorrow.”

“This was the Thomas we only started to get to know,” Tarkanian said, “someone who offered his expertise to others in need without question, while being thoughtful enough to worry out loud that he might be the nervous one and, by doing so, putting others at ease.”

DMSE Assistant Professor Tania Lopez-Silva, who started at MIT six months before Rose, recalled his patience and understanding. When she arranged a lunch with him, he said he would eat anything but mushrooms and eggplant.

“Somehow I misunderstood him, and I chose the place that probably has the most mushroom and eggplant options possible,” Lopez-Silva said. “He said that he still wanted to try the place, and he would find something to eat. That was Thomas — kind, patient, open-minded, and always willing to make the best of things.”

That quality came through in a story by another DMSE faculty member, Assistant Professor Thomas J. Wallin, who recalled a photo Rose had sent him of the apartment he had found overlooking the Charles River. Wallin was impressed, but Rose said with characteristic directness that it was quite small, but “very long and narrow” — which turned out to be a good thing. 

“He explained that because it was narrow, he could stand at the edge of his balcony and shoot at his archery target through his hallway, past his kitchen, at his door, and it would be the proper distance,” Wallin said, drawing chuckles in the chapel. “I don’t think that’s a request many Boston realtors get, but that was Thomas.”

Rose’s connections also extended beyond DMSE. Katya Tiukhtikova, joined by fellow MIT Archery Club member Kiarra Ning, remembered how Rose treated others. 

“He only focused on the good in people, and gave everyone a chance to show the best parts of themselves,” Tiukhtikova said.

She also recalled the contagious excitement Rose brought to archery practice, saying he “always brightened the days of everyone who was lucky enough to share the archery range with him.”

Tarkanian ended his remarks with a reflection on academic life and the importance of generosity and humility alongside brilliance and achievement.

“If part of Thomas’s legacy here at MIT is that we become a little more generous with our time, a little more humble about our accomplishments, and a little more deliberate about building communities in which other people can thrive and grow, that would be a legacy worthy of him.”

In 2021, Texas suffered the worst energy infrastructure failure in state history, following three severe winter storms in February. At least 246 people were killed as a result of the disruptions, which led to food, water, heat and electricity shortages.

If energy officials had had more time to prepare for the storms, for example by bringing in more trucks of natural gas to disrupted areas, damage from the event could have been mitigated. Unfortunately, weather forecasts beyond two weeks are notoriously unreliable, making for a tighter window for energy operators, farmers, and emergency response officials around the globe.

A new technique by a team including researchers at MIT significantly improves weather forecasting beyond two weeks. The researchers developed a machine-learning model that learns to correct errors in existing forecasting models to improve their output over the two-to-six-week time frame. When applied to a gold-standard forecasting model, it doubled forecasting accuracy over that time frame, placing first in a prominent international forecasting competition.

In a new paper in Nature, the researchers describe their approach and show how it enables forecasts that for the first time beat climatologic averages between two and six weeks.

“Not only does this do better on time averaged forecasts for this range, which is already very useful for farmers and members of the energy industry trying to match supply with demand, but it also improves predictions of extreme weather events, whether it’s flooding or cold spells or something else,” co-author and MIT Research Scientist Judah Cohen says. “It’s really important to predict both, but extreme events might be the most impactful part of this for society.” 

Joining Cohen on the paper are lead authors Hannah Guan, a Harvard College undergraduate; Soukayna Mouatadid, a University of Toronto postdoc; Genevieve Flaspohler SM ’18, PhD ’22; Paulo Orenstein, an associate professor at Brazil’s National Institute for Pure and Applied Mathematics; Jeremy Berman of the data company DTN; Jakob Schloer and Joshua Talib, researchers with the European Center for Medium-Range Weather Forecasts; and Alex Lu, Haiyu Dong, Jonathan Weyn, Zekun Ni, and Lester Mackey, all researchers at Microsoft.

Longer forecasts

In the 1970s, longtime MIT Professor Edward Lorenz came up with the idea of chaos theory to describe how tiny changes in initial weather conditions lead to completely different outcomes.

“He came up with this theoretical two-week limit where beyond that, we just can’t predict weather because the noise in the system grows so much that it overwhelms any signal you start with,” Cohen says. “That had a strong influence on the field that may have held people back from exploring forecasting beyond the two-week range. I never really believed that completely. In theory the atmosphere can create an infinite number of patterns, but it likes to fall into these favorite patterns over and over again. Once one thing happens, I think you can predict the domino that will fall after that, and the domino that will fall after that.”

Today, the best weather forecasts come from either dynamical models, which numerically simulate how the atmosphere and oceans evolve over time, or artificial intelligence models, which learn to make predictions from historical weather data. AI models have shown improvements in predicting weather at short ranges, but they’re even worse than dynamical models at longer time ranges.

Cohen, who is based in the MIT Department of Civil and Environmental Engineering, focuses on forecasting weather two to six weeks in advance, which is known as subseasonal forecasting. “That range is referred to as the ‘predictability desert,’ because in the short range you can capitalize on the initial conditions in the climate system and the atmosphere, and on the longer term you have these slow, varying boundary forces, like the super El Niño everyone’s focused on this winter,” Cohen says. “The subseasonal range can’t take advantage of either of those. In most cases forecasting in the subseasonal ranges is no different than going to a casino and rolling the dice.”

Cohen credits his co-author Lester Mackey, whom he has been collaborating with for more than a decade, with the initial idea of using machine learning to reduce errors in forecasting models.

“We built this wrapper — Lester calls it postprocessing,” Cohen says. “We know AI models have a lot of errors, so we thought maybe we can quantify those errors and remove them.”

While other forecasting systems do more simplistic error correction, the research team’s postprocessing layer uses AI to learn from a forecasting system’s mistakes based on historic data and shifts its forecasts. In that time range, forecasters use so-called probabilistic forecasts, where the likelihood of a range of possibilities if given.

“As more forecasts and data become available, our model adaptively improves its correction,” Cohen says.

To build the model, the researchers used MIT Engaging, a supercomputing cluster hosted by MIT’s Office of Research Computing and Data, but the resulting model does not require a supercomputer to run.

To put their model to the test, last year the research team submitted three models that included their post-processing layer to the European Center for Medium-Range Weather Forecasts (ECMWF), which is considered the best forecasting center in the world. Each season ECMWF holds forecasting competitions between models.

The submission that won first place combined ECMWF’s forecasting model with another, and then applied the postprocessing correction on top. That approach ended up doubling the original model’s predictive power in the subseasonal range. For the first time, the model offered an improvement over relying on climatologic averages.

“Up until now, no one had been able to show a model predicts weather better than long term averages for these subseasonal forecasts,” Cohen says. “They’ve pretty much all been worse than that — you could literally flip a coin and do better. People have questioned if it’s even worth investing in models at all for this time range. We consistently beat climatology averages with our models.”

The model also showed a statistically significant improvement in the ability to predict extreme weather.

“That’s one of the more important findings, because it’s so difficult to accomplish,” Cohen says. “When the ECMWF developed their own, non-AI, error-correction to their model, it got worse at predicting extreme weather. I think that’s because traditional error, or bias, correction is always trying to get to the mean, whereas extreme weather is on the fringe of probabilities.”

Preparing for disaster

In the Nature paper, the team showed their technique could be used to improve land forecasts in most regions across the globe. They also applied their bias correction technique to other models and showed it improved those as well, putting even lesser models on par with the ECMWF’s forecasts.

“What’s nice about our model is it’s an add-on wrapper,” Cohen explains. “We used it with the ECMWF’s dynamical model because that was considered the best, but it can be used with any model. It even improves models that are already bias corrected using another AI algorithm. So it shows a robust improvement. And it’s open source and low cost.”

Cohen says the meteorological office for the Republic of Kenya has already reached out about implementing the model to improve forecasting of the annual monsoon season to improve agricultural yields this winter. He hopes other groups will also use the model.

“This could help you decide when to fertilize around rain, when there’s a freeze coming to protect crops during winter harvests, or help you avoid droughts,” Cohen says. “It could also be useful for energy and emergency management. I’ve talked to energy officials who say, ‘If I’m told two days before that an ice storm’s coming, I don’t have enough time, but if I had two weeks, that could be enough to make adequate preparations.”

The work was supported, in part, by the U.S. National Science Foundation, the Carlos Chagas Filho Foundation for Research Support, and the European Union.

Hiroshi Ishii, the Jerome B. Wiesner Professor of Media Arts and Sciences at MIT and associate director of the MIT Media Lab, recently received the 2026 Industrial Designers Society of America Special Achievement Award for “...pioneering work that has greatly influenced human-computer interaction and design.”

For over 30 years, Ishii has explored how the digital and physical worlds can merge more meaningfully. Working with students, MIT colleagues, and international collaborators through his Tangible Media group, Ishii has pursued research marked by intellectual tension and artistic vision. He and his collaborators have produced truly original concepts including Tangible User Interfaces, which are interactive systems where physical objects, rather than screens, mice, or keyboards, are used to control and interact with digital information; and Tangible Bits, introduced at CHI ‘97, which established how people engage with digital information through physical objects. His later concepts, Radical Atoms and TeleAbsence, inspire researchers and designers worldwide.

“Beyond his research, Ishii has made lasting contributions to the global design community through ACM CHI [the Association for Computing Machinery Conference on Human Factors in Computing Systems], TEI [conference on Tangible, Embedded and Embodied Interaction], and UIST [User Interface Software and Technology], and by mentoring generations of designers and researchers,” said Professor Leila Ensaniat in her Industrial Designers Society of America Awards committee chair introduction.

Ishii's approach shows how bringing the arts, science, engineering, and design together creates innovations with real-world applications. His models now influence product development across industries, shaping how millions of people interact with information and each other.

As artificial intelligence becomes more and more powerful, its ability to handle human-like tasks is undeniable. Artificial neural networks can learn to recognize images, use language, or write computer code. Like the brain’s circuits, they reorganize themselves as they learn, adjusting the strengths of their connections to improve their performance.

The question for many neuroscientists is: What can these machine systems teach them about how the brain learns? And what kinds of models are likely to be the most informative?

In a paper published Oct. 9 in the journal PNAS, scientists at MIT’s McGovern Institute for Brain Research and K. Lisa Yang Integrative Computational Neuroscience (ICoN) Center argue that a crucial metric for the biological plausibility of an artificial neural network’s approach to learning is the network’s ability to organize itself, without the guiding hand of a developer. They show that it is possible for an artificial neural network to build its own learning machinery. With some simple instructions about how and when connections should change, a group of randomly connected units can reshape itself into an effective information-processing network, learning from the data it is trained on.

Biological learning

“Current AI technology is working very well, but it doesn’t necessarily use a learning algorithm that is used by the brain,” says Qianli Liao, a K. Lisa Yang ICoN Center Postdoctoral Fellow in the labs of MIT professors and McGovern Institure members Tomaso Poggio and Mark Harnett, and co-first author of the PNAS paper with MIT Research Scientist Liu Ziyin and graduate student Yulu Gan. Liao explains that engineers design the rules about how connections change as an artificial neural network processes training data and receives feedback about its mistakes. 

When it comes to training AI to do its job, the approach is powerful. But Liao and his colleagues caution that networks designed in this way may not offer much insight into biological learning. “The brain doesn't have an external designer or architect,” Liao says. “The brain has to organize itself. And we need to know how simple rules are going to make the brain organize.”

The rules that reshape neural connections as an animal learns may be different in different parts of the brain, or they may change as the brain becomes more mature. Neuroscientists are working to understand how those rules work and how they might be disrupted by aging or disease — a vital step toward developing new therapies and prevention strategies. Liao points to the development of brain-computer interfaces, which aim to restore communication when the brain’s language abilities are lost or let paralyzed patients control neuroprosthetic limbs, as one potential application.

“None of these can work without knowing the rules behind neuron-neuron interactions,” he says. “That’s not achievable by just designing a circuit that works very well.” An artificial neural network’s value to neuroscience should instead be judged by the extent to which it can organize itself — not just its connections, but also the machinery that adjusts those connections.

Self-organized artificial learning

Working with Poggio and Hartnett, Liao, Ziyin, and Gan set out to develop and validate a self-organizing computational network capable of learning. Liao explained that the network’s structure begins with multiple layers of randomly interconnected neuron-like units. Within each layer, every neuron-like component is subject to the same rules about how connections should be adjusted. “They just interact with each other using that rule, and they develop some type of behavior,” he says.

The team’s network implements the kind of simple rules thought to rewire brain circuits as humans and other animals learn from their experiences. In both the brain and the machine network, for example, “neurons that fire together wire together.” In other words, when two connected neurons are activated at the same time, the connection between them strengthens. 

The team showed that a self-organizing network governed by rules like these can learn to recognize patterns and classify images. They explain that in their model, learning is enabled by self-assembling motifs that emerge according to simple rules, then serve as building blocks for functional circuits. As their network learned, these motifs reorganized themselves using an optimization strategy similar to the approach built into most modern machine learning systems, called gradient descent, in which connections are gradually adjusted in response to errors in order to improve performance.

Most modern AI implements gradient descent through some form of the back-propagation algorithm, which neuroscientists say is unlikely to be used by the brain. The team compared its self-organizing network to networks designed to learn in this way and found that it performed at a similar level.

Connecting brains and machines

Poggio says finding a biological parallel to back-propagation could trigger significant progress in both AI and neuroscience. 

“If the synaptic motifs we propose are, in fact, used by the brain to do what backpropagation does for artificial neural networks, then we will have a strong connection at the level of fundamental mechanisms between artificial and natural networks — between machines and brains,” he says.

Liao says a self-organizing network does not necessarily learn the same way the brain does — but it’s a critical starting point for biological plausibility. 

“Self-organization is the first high-level framework,” he says. “Once we are good with self-organization as a measure, then we can talk about next thing.”

While the group’s main objective is better understanding the brain, Liao says that knowledge could ultimately improve machine learning systems, too, enabling engineers to develop networks that perform better or in more human-like ways.

MIT Lincoln Laboratory hosted the National Security Commission on Emerging Biotechnology (NSCEB) on Sept. 3, for a stop on its national roadshow. The event convened representatives of state organizations, defense centers, universities, technology companies, and business executives to discuss how Massachusetts' biotechnology ecosystem can strengthen national biodefense capabilities.

"We have a lot to leverage in this region to advance biotechnology," said Christina Rudzinski, assistant head of the Biotechnology and Human Systems Division at Lincoln Laboratory. "Many people are already working together to push the bioindustrial sector beyond biopharmaceuticals and into a broader ecosystem across industries. This synergy is where NSCEB comes in — connecting policy to the real-world work already happening, including here at Lincoln Laboratory and across the state."

Biotechnology touches many economic and defense sectors, from disease-resistant crops and lifesaving medicines to biofuels and advanced manufacturing materials. In 2022, Congress established the NSCEB to study the national security implications of this rapidly advancing field and issue policy recommendations to ensure continued U.S. leadership. 

In April 2025, the 11-member bipartisan and bicameral commission released its report, calling for a minimum of $15 billion in federal investment over five years. According to the report, China is investing aggressively to dominate the field, echoing a pattern the United States has seen before in sectors such as semiconductors.

"The coalition of stakeholders working in the Massachusetts biotech ecosystem illustrates how collaborative regional work will help keep the U.S. in the lead by producing exceptional innovation across all areas of biotechnology," said NSCEB Executive Director Caitlin Frazer. 

Now in the final year of its three-year mandate, NSCEB is touring the country to see local biotechnology ecosystems in action. The commission will bring insights back to Washington to help members of Congress support their regions' strengths. 

Advancing Massachusetts as a defense-biotechnology hub

Massachusetts has identified biotechnology as a priority technology area under its Strategic Hub for Innovation, Exchange, and Leadership in Defense (SHIELD) initiative. Launched by Governor Maura Healey in October 2025 alongside $47 million of initial aligned investment, the initiative has moved with speed and at scale. This includes standing up the BioNexus program that leverages state matching funds to develop biotechnology and biomanufacturing solutions for the U.S. Army Combat Capabilities Development Command (DEVCOM). Further, Healey included a $100 million authorization for defense sector capital projects in the Mass Wins Act, an economic development bill filed in April. The bill is currently being considered by the state legislature.

During the roadshow event, Ben Linville-Engler, who co-chairs the SHIELD Working Group and serves as deputy director of the Massachusetts Technology Collaborative, led a panel discussion with representatives from the Massachusetts Life Sciences Center, Business Executives for National Security, Activate, DEVCOM Soldier Center, MassChallenge, Perimeter, Center for Advanced Manufacturing, and the Massachusetts AI Hub. They are among more than 40 partner organizations participating in SHIELD. 

"There is incredible convergence between economic development and national security in all domains — land, air, sea, space, cyber, and bio," said Linville-Engler, citing the state's $48.6 billion in defense-sector economic contributions last year. "Massachusetts is committed to biotechnology and national security, and not just in Boston and Cambridge but across the entire state."

In July, the SHIELD Working Group submitted a report outlining steps to speed the transition of technology into deployable capabilities, strengthen the defense workforce pipeline, and reinforce the state's national security industrial base. 

Lincoln Laboratory is well positioned to support these priorities. As a Department of War federally funded R&D center, the laboratory has deep knowledge of national security needs and sustained ties to academic, commercial, and U.S. government laboratories. "These attributes can help strengthen biotechnology collaborations and integrate and transition solutions across Massachusetts and New England," Rudzinski said.

Scaling up biomanufacturing

One project advancing this vision is the Biotechnology Scale-up Hub for Innovation, Economic Development, Labor Training, and Defense (BioSHIELD). Planned for Worcester, Massachusetts, this 200,000-square-foot biomanufacturing facility is designed to mobilize the region's defense-biotechnology sector.

"BioSHIELD's mission is to accelerate breakthrough biomanufacturing solutions that strengthen national resilience, support commercial innovation, and grow the regional economy," said Edward Wack, project co-lead and principal staff in the laboratory's Biotechnology and Human Systems Division. "It will bring together industry, academia, and government to transform ideas into scalable biological products, supporting a secure and competitive bioindustrial future." 

The project responds to NSCEB's call for a national network of biomanufacturing scale-up facilities, a capacity currently missing in New England, according to Wack. The facility will offer shared equipment, infrastructure, and secure laboratory space to help bridge the "valley of death" between early-stage discovery and commercial- or defense-scale production. BioSHIELD's ready-built support networks will be key to overcoming technological and regulatory hurdles, helping innovators rapidly move capabilities to market or military use. 

A coalition supporting the project includes Massachusetts Biomedical Initiatives, Worcester Polytechnic Institute, Bio Base Europe Pilot Plant, and Lincoln Laboratory. It builds on a foundational effort in central Massachusetts funding the Massachusetts TechHubs program. At the event, Wack announced the coalition has signed a memorandum of understanding to formalize the partnership. Design workshops are scheduled to begin soon, ahead of funding and construction.

Connecting AI and biotechnology

Other discussions centered on the intersection of biotechnology and AI, another critical area for national security. Panelists represented Boston University, the University of Massachusetts Boston, Massachusetts Life Sciences Center, Parker Institute for Cancer Immunotherapy, Wyss Institute, and Phicil-itate Change. They discussed challenges in collecting and disseminating high-quality, AI-ready datasets and opportunities for AI to improve biomanufacturing and therapeutics. 

Closing the event, Wack called for continued collaboration as the nation moves at speed and scale to stay at the forefront of biotechnology.

"At Lincoln Laboratory, we've built a strong foundation in biotechnology for more than 25 years, and it's heartening to see this level of coordinated work taking shape. It resonates with so many of us who view biotechnology as both an economic and national security imperative," Wack said. 

“This tomato is blowing my mind,” says Nermine Hassanin. The MIT graduate student in architecture is marveling at the fruit in her salad. 

She and several MIT School of Architecture and Planning classmates had spent their morning working at Hannan Healthy Foods Farm in Lincoln, Massachusetts, and then harvested the contents of the salad they were sharing at a picnic table: lettuce, kale, cucumbers, green onions, and tomatoes. 

“It doesn’t get much fresher unless you’re eating this in the field,” responds Zachary Rapaport, a classmate and co-founder of MIT Farm, a community-driven program at MIT that is radically changing what it means to eat farm-fresh food on campus and gain practical experience in urban farming. 

In fact, the program resonated so quickly with so many that it was awarded the Karl Taylor Compton Prize for a group this spring — the highest award presented by the Institute to students in recognition of excellent achievements in citizenship and devotion to the welfare of MIT. 

Seated with Hassanin and Rapaport were classmates Aisha Cheema and MIT Farm co-founder Aleks Banaś, undergraduate architecture major Amy Kim, Sanne Wright MCP ’24, and Mohammed Hannan, the owner of Hannan Healthy Foods Farm. The story of how they came to enjoy this meal on a brilliant summer Sunday is equal parts vision, hard work, and serendipity. 

Community in Cambridge

In their first year at MIT, Rapaport and Banaś were classmates in a design workshop 4.184 (ClimateCorps@MIT) that focused on climate action within communities. Bonding over their interest in food systems, the two shared a desire to continue the community service work that took them outside of the classroom.

“We started ideating on what it would mean to create relationships with students, faculty, and staff on campus with partners who were already growing food in the community,” says Banaś. “We saw many organizations in the community already doing inspiring work, so instead of starting from scratch we wanted to plug into this system.”

Their concept for MIT Farm explored a holistic view of community, food security, and supporting local farming in a time of climate resilience. It also capitalized on the desires of two budding architects to develop practices they see as essential for sustainable communities.

“After working in an architecture practice, I came to grad school to find a way to create an architecture practice grounded within real communities,” says Banaś. “For me, architecture is not just about creating a building, but creating an environment for other interactions. I think we have a certain level of responsibility to the people around us; we need to learn how to work together with them.”

“What I’ve learned through developing MIT Farm is that farming requires you to be accountable to seasonal change, to people, and to the environment in which we’re working,” adds Rapaport. “There are things you only begin to notice when you spend time with a place and really engage with it. In many ways, architecture is the same. We need to be very attentive to the world around us if we’re going to make good design.”

In 2025, Rapaport and Banaś actively sought out like-minded organizations on campus — including MIT Professor Miho Mazereeuw’s Urban Risk Lab, the MIT Office of Sustainability, and the Open Space Committee — to build relationships and germinate ideas. They found enthusiastic supporters for their ideas which kept them both engaged and excited. That spring, they entered their concept for MIT Farm into the iCampus competition, which recognizes innovative and creative applications of technology that improve living and learning at MIT; they were awarded one of the three cash prizes. 

Between supporters on campus and burgeoning relationships with like-minded organizations off campus, Rapaport and Banaś began to execute their vision.

One idea — a farm cart — kick-started what Rapaport terms “the growing side” of MIT Farm. He, Banaś, and classmate Courage Dzidula Kpodo MArch ’25 designed and built the two-tiered cart. It is constructed of pine with polycarbonate panels and divided into sections that support active vegetable gardens. Repurposed pipette tip containers sourced from campus labs serve as seedling planters. Supported on four wheels, the cart can be pushed or pulled anywhere.

“By introducing mobility, we began growing food without having it tethered to a specific plot of land,” says Rapaport. 

They discovered it was more than just a “moveable farm.” The cart became a DJ booth at school happy hours; was wheeled to nearby public schools as a learning tool; showcased at the “Day of Climate” event at the MIT Museum; and transformed into a tiny winter greenhouse at the East Campus dorms.

Field work

Rapaport and Banaś also located farm communities in Boston and beyond to develop volunteering and mentoring opportunities. Now, through MIT Farm, students can volunteer to work the two rooftop gardens at Boston Medical Center (BMC), a compact farm that generates 6,000 pounds of produce to serve patients and the BMC community. 

A second option for students to work at a city farm is with Eastie Farm in East Boston. A nonprofit with a focus on food security and climate justice, the organization supports local farms that struggle to compete with industrial agriculture.

Last summer, Rapaport and Banaś connected with Mohammed Hannan, the Lincoln farm owner and a laboratory manager at MIT. MIT students had already found their way to his farm, about 15 miles from Cambridge. Four years earlier, Heather Paxson, an anthropology professor and associate dean for faculty at MIT School of Humanities and Social Sciences, brought her anthropology class to visit Hannan Farm to learn about small-scale certified organic operations. 

Hannan was open to the idea of the MIT community working on the farm. He says it is “increasingly challenging” to find people who want to farm. He stresses the broader importance of community support for local agriculture, noting that food security and sustainable farming require collective effort from society — not just from farmers. 

“Many farmers around me are leaving farming to make their living elsewhere,” says Hannan. “It is a silent pandemic.” 

When classes resumed last fall, Rapaport, Banaś, and classmates began volunteering on weekends at the farm. Working with the MIT Office of Sustainability and the Anthropology Section, the group started a pilot program selling Hannan Farm’s produce for a reduced price to students on campus on Mondays. The Fresh Food Bag program offered curated bags of produce that had been harvested a day earlier. The program was met with delight among the MIT community, with nary a vegetable left behind each week. Rapaport and Banaś say they will expand the program this fall to include offering bags of fresh produce in MIT dorms.

The work on the farm changes constantly, based on the season and business needs. One day students may be in the field replanting plants that didn’t establish properly; on another, they may be maintaining tools or working on infrastructure projects.

“There isn't really an ‘off season’ since the farm operates year-round, and we've worked to bring MIT students out in every season and semester,” says Rapaport. “During the winter, we repaired greenhouses, prepared seedling trays, and started getting ready for the next growing season.” 

Hannan describes his collaboration with MIT students as “game changing,” highlighting how students gain practical, problem-solving experience while advising him on how to improve the farm’s infrastructure.

“It’s hard to visualize real-world problems when you are working in a lab [on campus],” says Hannan. “But the students are seeing the day-to-day challenges farmers are facing here. They are coming up with solutions that can alleviate these problems — and not just on my farm. We need more people from different backgrounds to see how we can locally grow more efficiently using the tools and technologies that we have. Sometimes we farmers don’t know where the technologies are — such as using solar irrigation. We’re not always tech-savvy, or we don’t have the expertise. Farming needs all of it.”

To everything, there is a season

In January, during MIT’s Independent Activities Period, Rapaport and Banaś conducted a weeklong workshop to design a root cellar to support food storage and preservation for the farm. The group also developed a curriculum for K-12 students to teach how to design a community garden with the support of a Day of Climate Grant they were awarded.

Kahlen Wheaton, a Department of Urban Studies and Planning (DUSP) undergrad, was among the students who worked on the curriculum and helped organize volunteers at the farm last spring.

“I really love the club,” says Wheaton. “It’s a great opportunity for MIT students to get out of their bubble and do some service work. It’s also great just to bring fresh produce to campus because, as an undergrad, I’m not always eating the most healthy or local foods.”

Wheaton had originally planned to study environmental engineering at MIT, but found urban planning and the opportunity to impact policy — especially as it influences designing cities for healthy lifestyles — exciting. She has planned her thesis on urban farming and will use the connections she is making in local farming to inform her work.

Pastures of plenty

This past spring saw dozens of volunteers spending weekends at Hannan Farm. Among them was Rania Ghosn, an MIT associate professor of architecture and urbanism, who brought her family. She says she enjoyed working shoulder-to-shoulder with colleagues in the field, and that MIT Farm provides a “tangible public good while strengthening a sense of belonging and joy.

“Working in the fields with my husband and our 10-year-old son became a kind of Sunday ritual: purposeful, physical, joyful, and full of meaning. For my son, the farm was a lesson no classroom could quite reproduce.”

On a recent Sunday, five architecture students tended to rows and rows of vegetables. Weeds had grown high with the recent rain, but after some tugging and digging, they released their hold in the dirt and lay in piles. The hot sun, the smell of fresh herbs, and the sounds of birds and insects made Cambridge and MIT feel very far away.

Nermine Hassanin has spent several Sundays here. She says she enjoys meeting new people on the farm each time.

“It’s a good sense of community here,” she says. “And being out of the city is very therapeutic.”

Growing up, Aisha Cheema worked with her dad on their family garden. She says working on the farm is “a nice way to stay connected” to those experiences.

“I’m interested in how this environment can be pulled into architecture,” says Cheema. “So many spaces seem siloed from the outdoors. I’d like to connect the outdoors to indoor spaces. I think some of the structures here — like the greenhouse — are great inspiration for me.”

Rapaport and Banaś spent their summer designing additional interventions for the Lincoln farm: an improved washing station and an upgraded farm stand that allows for cold storage to keep produce fresh longer. The design has been approved by the Lincoln Conservation Commission and permits are pending. 

Also, this summer, Amy Kim oversaw the design of a dormant garden plot on campus near Tang Hall. The undergraduate performed the work as part of MIT’s Undergraduate Research Opportunity Program. Kim says she designed the project because it allowed her “to get hands-on experience building a space that would foster community and collaboration, with the added environmental benefit of repurposing material such as scrap brick and scrap wood. 

“The garden's primary goal is to be a pleasant space to sit and relax with friends while surrounded by aromatic herbs and flowers,” Kim adds.

With their theses underway in preparation for Commencement next year, Rapaport and Banaś will hand off their oversight of MIT Farm to Wheaton and others this fall. They could not have guessed how the popularity of MIT Farm would have connected with so many students, or how many friends they cultivated off campus — and they are overjoyed with the results. They believe the community they have developed and fostered will keep MIT Farm vibrant into the future. 

Ghosn agrees.

“MIT Farm feels deeply MIT: mind and hand, ethics and experimentation, systems thinking and dirt under the fingernails,” she says. “Aleks and Zach have created a student-led project with moral clarity, organizational intelligence, and genuine public purpose. It is modest in its means and expansive in its imagination and impact."

MIT engineers have developed a new way to program natural killer (NK) immune cells to target cancerous T cells or T cells that have become autoreactive, leading to diseases such as lymphoma and autoimmune hepatitis. 

Using their new technique, the researchers can reprogram NK cells to target diseased T cells without ever removing the NK cells from the patient’s body. Instead, they use lipid nanoparticles to deliver messenger RNA directly to NK cells. These mRNA molecules encode proteins that allow the NK cells to seek and destroy the dysfunctional T cells.

Known as CAR NK cells, these cells could be deployed against a variety of diseases. The MIT team’s new approach could bypass one of the major obstacles to using these cells, which is the time and expense of manufacturing them outside the body.

“The manufacturing challenges with CAR NK cells are significant. The idea here is you can deliver a nanoparticle that will modify your own NK cells with a CAR construct and teach them to attack tumor cells,” says Daniel Anderson, a professor in MIT’s Department of Chemical Engineering and a member of MIT’s Koch Institute for Integrative Cancer Research and Institute for Medical Engineering and Science.

Anderson is the senior author of the study, which appears today in Nature Biotechnology. The paper’s lead authors are MIT postdoc Yuang Chen and MIT graduate students Julia Monkovic and Haseeb Mughal. Robert Langer, the David H. Koch Institute Professor at MIT, is also an author.

Easier reprogramming

Over the past decade, the U.S. Food and Drug Administration has approved seven new cancer treatments based on engineered T cells called chimeric antigen receptor (CAR) T cells. Chimeric antigen receptors are proteins that can be embedded into T cells, allowing them to detect a target antigen, such as a cancer protein. When the target is detected, the T cell becomes activated and destroys the target cell.

This approach has had great success in treating blood cancers, primarily those that originate from B cells, including several types of leukemia, lymphoma, and multiple myeloma. 

To make CAR T cells, doctors take a blood sample from the patient. The sample is sent to a lab where the T cells are isolated and then engineered to express the CAR protein. When enough CAR T cells have been generated, they are shipped back to the hospital and infused into the patient. 

In recent years, scientists have begun using a similar strategy to create CAR NK cells. One potential advantage of CAR NK cells over CAR T cells is that they could be used to kill diseased T cells. Using CAR T cells to treat T cell diseases often doesn’t work well because T cells that are engineered to target T cell proteins may end up killing each other, in addition to the diseased T cells. That issue doesn’t arise when using NK cells to target T cells.

“NK cells can treat these T cell malignancies — not only T cell cancer, but also T cell autoimmune diseases,” Chen says.

However, manufacturing large quantities of CAR NK cells poses many of the same difficulties as generating CAR T cells. To simplify the process, the MIT team wanted to come up with a way to generate CAR NK cells “in vivo” — that is, in the patient’s own body, without having to remove any cells.

“If a patient can just get an infusion of nanoparticles, it would dramatically simplify the process and reduce costs,” Anderson says.

Currently, there are no FDA-approved CAR NK cell therapies, but many researchers are working on developing them to treat cancer or autoimmune disorders. 

To generate these cells, the MIT team designed lipid nanoparticles (LNPs) that can deliver an mRNA payload. These strands of mRNA encode a CAR that can be targeted to specific cell types. For example, they could deliver a CAR that targets CD19, a protein found on B cells, or CD3, a protein found on autoreactive T cells. 

The particles are coated with antibody fragments that bind to receptors found on the surface of NK cells, to ensure they reach the correct targets.

Targeting dysfunctional T cells

After these particles were infused into mice, the researchers found that they could successfully reprogram about 80 percent of the targeted NK cells. The researchers then tested the particles’ potential to treat cancer and autoimmune hepatitis, which is caused by autoreactive T cells. 

In a mouse model of an aggressive form of melanoma, three of the five mice that received LNPs that reprogrammed their NK cells to target the cancer cells went into complete remission.

The researchers also saw significant tumor destruction when they used the LNPs to treat T cell lymphoma in a humanized mouse model, created when mice receive transfusions of human blood cells, allowing them to transiently generate a human-like immune system.

In these mice, tumors were eradicated after five injections of the reprogramming LNPs. This result was similar to what the researchers saw when they treated mice with CAR NK cells that were generated using the conventional approach — removing NK cells, reprogramming them, and returning them to the mice.

Lastly, the researchers demonstrated that their LNPs could generate CAR NK cells that target autoreactive T cells. For this application, they studied mice with autoimmune hepatitis, in which T cells attack hepatocytes in the liver. They found that in the treated mice, the autoreactive T cells were destroyed, and liver damage healed. These effects were seen after just one injection of the LNPs.

“This study is an excellent example of the power of nano delivery systems and their ability to potentially treat cancer and other diseases,” Langer says. 

The researchers hope that these kinds of engineered NK cells could be used to treat a variety of autoimmune disorders caused by dysfunctional T cells, including multiple sclerosis and rheumatoid arthritis. 

The research was funded by Sanofi, the Marble Center for Cancer Nanomedicine, and a Koch Institute Support (core) Grant from the National Cancer Institute.

When government officials craft policies, there is often a comment period for public feedback. Lawmakers have long had to sort through fake feedback that is mass-produced in an effort to sway policy, but they’ve generally gotten used to detecting it. With, artificial intelligence, however, people can submit realistic-seeming comments, which are harder to dismiss as being fabrications with no person behind them. That’s just one way AI can be used to generate text and statements, in civic conversations, for which no one can be found to be answerable.

This issue, among others, has caught the attention of Deb Roy, professor of media arts and sciences at MIT and director of the MIT Center for Constructive Communication (CCC). In a new essay published in the Bulletin of the Atomic Scientists, titled, “Who stands behind the words?,” Roy calls for a renewed emphasis on responsibility in communication, asserting that even when AI is used to craft text, there needs to be accountability and “answerability.”

Building better group dialogues to produce more public trust is a core part of the work of CCC. The group, along with Cortico, the closely affiliated nonprofit Roy co-founded, has helped facilitate group discussions on important issues in many situations — including civic meetings, schools, deliberative assemblies, and private organizations — and used tech tools, including AI, to help find themes among comments, and more. But the center also designs this process so that participants can check and override the tech-based output at each key step. Ultimately, a person is responsible for every public comment. MIT News spoke to Roy about the essay and the work of the center. 

Q: You’ve identified this problem about the lack of accountability behind words that are generated by AI. When did you really first notice this?

A: It came into focus when Chat-GPT [which is powered by a large language model, or LLM] launched. As for my own history with language models, my PhD work explored what we might now call PLMs — Puny Language Models — so I’m quite familiar with language models and have tracked the field’s rapid evolution. But Chat-GPT turned a corner. Two things changed at the same time: the depth of the model and its knowledge, and the way you could interact with it. 

I became obsessed with trying to find the limits of what it could do. And I felt that no matter how hard I tried to stretch my arms around it, I couldn’t. I built my tenure case at MIT on language grounding — the idea that it’s not enough for an AI system to just process lots of words and other symbols. I believed that to understand what words mean, you actually need a physical body and life experiences as part of the process. So I revisited the philosophy-of-AI literature I had once been steeped in, which asks what fundamentally distinguishes the kind of system that large language models exemplify, in extraordinarily advanced form, from the kind of processing we humans engage in. I gave several presentations to my own graduate students trying to make the case that, even though AI seems so intelligent, it’s actually different in kind — and I could not convince them. 

And I thought, I know the difference, but I’m having trouble convincing even myself of what it is. And that led me to ask, “Wait a minute: Is there someone or something that stands behind the words, a human responsible for what is being said?” That’s the thing this extraordinary new language machine doesn’t have. 

Whether these systems genuinely understand language is actively debated. But even if you grant them all the comprehension you like, they still lack standing. The buck can’t stop at a thing, a software system, with nothing to lose.

Q: The Bulletin piece has these interesting examples of civic feedback forums where it looks like there are real people saying things, and instead you have a barrage of comments from AI. How did you find some of these other examples in society where things go off the rails in this way? 

A: After the release of ChatGPT, I started hearing a concern from people inside government who review public comments on proposed rules: a suspicion that not all of the comments were written by humans.

My reaction was: That’s going to delegitimize this channel for public feedback. That was several years ago, but everyone now knows this is happening. Public comment is a pillar of democracy. You don’t just put a new rule in place. People need to have a voice and feel like they’ve been heard before a rule is imposed on them. If you skip that step, people don’t accept rules as legitimate. That’s why that process is baked into how America works. And that pillar just cracked, and by now, in my mind, it’s fallen over. Has anyone noticed? My sense is the people on the receiving end of these comments have quietly adjusted the weight they give to public comments. It was never perfect, but boy, now much of its value is being swept away. 

I open my essay with this example: In 2017, the Federal Communications Commission received more than 22 million public comments on its proposal to repeal net neutrality rules, and the New York attorney general later found that nearly 18 million of them were fake. As I write in the essay, “That fraud was caught because it left fingerprints: templated text, duplicated phrasing.” But large language models erase the fingerprints. Now every comment can be fluent and different, and reviewers can no longer tell which ones express views that real people formed and stand behind.

If you have enough of the stuff that’s fake — words that no one with a stake actually stands behind — it desensitizes you to what’s real. That filters back to the public, and people realize: If I take the time to express myself, they’re not even listening any more.

Q: Could you say a little bit more about your point that we need people with “standing,” who will take “accountability” for all pieces of text, even when those bits of text are generated by AI? 

A: In the legal system, standing is a formal idea: sometimes a case isn’t heard because the people bringing it don’t have standing. But informally we use that term all the time. I’m not trying to formalize standing in everyday life. Instead I’m asking:  What is the informal sense of the term, borrowed from the legal one, that we need to pay more attention to now? Here in New England, consider the open town meeting: The person who stands up to speak is generally expected to have standing, as someone from that town with a stake in what gets decided. And if it’s a stranger who drove from across the state, or from out of state, everyone’s like: Wait a minute, who is that? They’re not from around here. In general, there’s just enough friction and it’s costly enough to show up somewhere you don’t actually have standing, that it doesn’t happen much.

Now LLMs and the current digital networked environment collapse those boundaries. We always took for granted that if someone has a voice, it’s because they have standing. They have some basis to be heard. We need to pay attention to this, and really consider how to design systems so that the people with standing are the ones who get heard. 

The other idea I tried to develop is “answerability.” It isn’t formalized in law, but it’s baked into the legal process. If you have standing to give testimony, there you are, ready for cross-examination. You have to be able to answer for what you’re saying. And if you’re making a specific accusation about someone, you have to be willing to back it with facts. Standing and answerability have to travel together: The right to be heard comes with the obligation to answer for what you say.

That doesn’t mean everyone has to attach their name to everything. I write in the Bulletin essay, “Privacy and answerability aren’t opposites. … An institution can verify a person and stand behind their words without exposing them.” To voice a view, you only need to show that a distinct, eligible person is speaking. An accusation asks more: Someone trusted, like a reporter, has to be able to know who is speaking. As I say: “Disclosure obligations should scale with power, and identity protections with vulnerability.”

None of this means AI can’t help. AI can do a lot of the work inside the chain of accountability — drafting, summarizing, finding patterns. If someone with a stake in a rule uses AI to draft a comment, adopts it as their own, and is ready to answer for it, that’s legitimate, in this context. What AI can’t do is anchor the chain. There has to be a person at the end who stands behind the words.

In the essay I lay out four design principles for building that in: “attributable authorship,” a “decomposed and inspectable process,” the use of “provenance as infrastructure,” and the need to “widen standing rather than concentrating it.” As I write, the “animating rule is that autonomy is permitted while unaccountability is not.” And the metric that matters most isn’t the efficiency gained by removing humans, but the number of people AI places into roles where they can answer for what they do.

Q: There is a phrase in the documents of the Center for Constructive Communication, that we need things to move “at the speed of trust,” not at the “speed of tech.” And that implies really thinking about building up these systems. Could you just say a little more about what that phrase means to you?

A: I learned that phrase when I was in Appleton, Wisconsin, with Katherine Cramer [of the University of Wisconsin], a longtime collaborator, talking with community organizers who were running “Dialogue across Divides” conversations there. They had hundreds of people involved. I thought I was developing technology tools with exactly that same purpose, and was so excited to see it. And I asked one of the people who ran that organization, “How fast did you scale this?” And he paused, and then he said, “At the speed of trust.” 

And I took that to mean “slow.” Then I realized he was saying: For you, slow, because you haven’t yet earned our trust. And I’ve since learned that when you do have trust, things can move fast. So bringing in technology is really a process of engagement. It doesn’t work to say, “Trust us, we’re doing this because we think it’s for your own good.” People will turn to someone they already trust and ask, “Is this good for us?” And if that person says “Yes,” things can move fast.

Look at what’s happening at the intersection of AI and society: The speed of AI and the speed of democracy are running on different clocks. Talk to people who work in democratic practice and they say, “AI must slow down, because democracy can’t move that fast.” Talk to people in AI and they say, “AI cannot slow down, democracy needs to speed up.” If we can’t get these two time scales in sync, we’re headed for a breakway situation, where AI gets deployed faster than democratic processes can respond, and decisions get made without the public. That’s starting to happen right now. 

Q: We need people actively doing things to build dialogue and figure out how tech tools can interact with that. At the Center for Constructive Communication, you’ve done a few iterations of the center’s Real Talk initiatives to build civic dialogues, as well as the Youth Voices Initiative in school settings, and currently you’re working on the New Social Forms project. Could you say a few words about them? 

A: The Real Talk effort represents the longest-standing thread in this work. A huge emphasis of our work has been developing better ways to listen to one another. We create a context that is always small — small is beautiful here — where people are willing to share authentically. And we have the consent architecture and the technology to reveal the larger patterns, and then share them. We’ve done this with hundreds of communities around the country. And closer to home, we’ve brought the same approach to our campus here at MIT through the realtalk@MIT program, with support from the Office of the President and the Office of the Provost, to help our community build a culture around dialogue and listening. 

The second thing I’d emphasize is that we’ve been doing more and more work on deliberative processes and citizen assemblies, and on how to make them more accessible. What if a group of high schoolers says, “We have a real problem with our cellphone policy?” Suppose you want to have an assembly that brings students, teachers, and administrators together, not just to listen to people’s concerns but to make policy recommendations. So we did that. The center worked with a high school in New York City, where technology helped people reveal their preferences and differences and listen to each other.

We’re also building answerability into the technology. A prototype called VoiceTrace, developed by PhD student Cassandra Overney in our group, links every output statement — a summary of what was heard in a listening process, or a recommendation coming out of a deliberation — to the specific voices behind it, so any claim can be traced back to the conversations it came from.

The New Social Forms project pertains to a very interesting pattern emerging in workplaces around the world, certainly the ones we’re connected to. AI is an extraordinary tool for individual empowerment, but in the words of one of our collaborators at the Google Experience Institute, people are tunneling. They go deep, using AI to develop refined ideas, but when they surface, their ideas have diverged from their teammates’. So groups and teams trying to work together with AI, there’s a human coordination problem.

We’re going to expand the initial work we’ve started with Google because there’s a lot of interest in collective communication, trust, and interaction among people trying to work together. 

So is that a civic context? It’s a workplace, but workplaces are civic spaces, too. They govern people’s daily lives, they’re where many of us build civic skills and learn to cooperate across difference, and there’s a long tradition of civic life rooted in shared work. We’re all humans trying to interact with each other. These are the same issues. Justice Louis Brandeis once used the phrase “industrial democracy” to capture this. When workplaces lacked certain kinds of civic processes and civility — two interestingly related terms — labor disputes used to turn violent. Procedural justice, mediators, and people hearing one another are just as necessary in the workplace. In the end, it’s all about human beings. Hearing the humanity in one another is what democracy depends on, and it’s just as necessary for communities and organizations to function.

The Sierra Nevada mountain range in California is as stunning as it is diverse, with storms from the Pacific Ocean dumping rain and snow on the range’s western slopes while arid conditions mark much of its eastern face. David Des Marais, MIT’s Amgen Career Development Associate Professor, knows the mountains well. His childhood in San Francisco was filled with hiking and camping excursions throughout the range.

At some point during those trips, he started wondering about one of the few constants of the landscape: the presence of plants. How were plants able to thrive in all those conditions? He turned the question into a career.

Des Marais, who directs the Parsons Laboratory in the Department of Civil and Environmental Engineering, studies how plants interact with their environment. The research spans molecules and mountain ranges, exploring everything from how individual genes change a plant’s ability to grow to how climate change will impact populations of red oak trees across the eastern U.S. His work also considers how evolutionary forces act on plants to change populations over time.

Plants are central to his lab, but Des Marais sees them as a window into all living things. He’s even quick to credit the health of his houseplants and garden to his wife, an evolutionary biologist at Harvard University.

“Understanding why life is so diverse is fascinating to me,” Des Marais says. “Understanding how different organisms, from bacteria to dinosaurs, have figured out how to interact with their challenging environment and optimize their fitness is inherently interesting. Why plants in particular? As a geneticist and a physiologist, plants are compelling experimental systems. If I have a theory about gene X causing function Y, I can go into the lab and delete gene X from this species and put it into that species and test function directly. It’s a powerful approach.”

Des Marais does not spend all his days in a lab. In addition to directing the Parsons Laboratory, he teaches four classes at MIT and has two sons who are active in baseball and theater. He welcomes these responsibilities.

At MIT especially, he embraces service.

“The only reason MIT works is because there are people willing to sacrifice some of their research to keep the place running,” Des Marais says. “One of the things I like about academia is that there’s sort of an institutional memory, and you try to honor the legacy of where you’ve been while you shape the future.”

From California to MIT

Des Marais grew up near Stanford University as well as near facilities for NASA and the U.S. Geological Survey, which meant there were always plenty of scientists and engineers around. His mother was a teacher, and his father was a geochemist with a knack for pointing out that the landscape around them was the result of geological processes dating back millions of years — a habit Des Marais admits he didn’t like at the time but that, in hindsight, influenced his thinking.

“I’m pretty sure that got into my brain, especially the idea that what looks static often comes about because of processes, and we can study those processes,” he says.

Des Marais studied integrative biology as an undergraduate at the University of California at Berkeley. He completed his PhD at Duke University around plant genetics.

“There’s a whole series of outdoor events, from the Boy Scouts through college, that made me start to think about the challenges that plants face living in these stressful environments,” he says. “That drove me to plant science and environmental science.”

In a long postdoc at the University of Texas at Austin, Des Marais learned to conduct experiments teasing out the effect of specific genes as he studied how they changed a model plant’s relationship with water.

Des Marais came to Boston primarily for personal reasons: He and his wife, who was also a postdoc at UT Austin, entered the job market at the same time. After some awkward moments competing for the same jobs, his wife landed a tenure-track position at Harvard University. Shortly after, he joined Harvard as a staff scientist.

In 2017, Des Marais discovered MIT’s Department of Civil and Environmental Engineering was hiring a plant ecologist.

“I don’t even think of myself as a plant ecologist, but I applied on a whim and they were interested,” Des Marais says. “I do a lot around plant water relations, and I joke that there are all these hydrologists in the department who study how water moves through the soil and into the atmosphere through these magical pipes. The magical pipes are plants, which play an important role in the water cycle. I realized a ton of people at MIT were studying that phenomenon at regional or global scales, but no one really studied how or why plants move water in the first place.”

Today Des Marais’ lab work can be broken into two broad categories. The first is centered around how plants function, including their relationship with environmental factors like water, temperature, and nutrients. Des Marais says some of his most impactful studies link specific molecules to how a plant functions in a given ecosystem.

The second centers around fundamental questions in plant evolution, what Des Marais describes as how plants “make decisions” about when to grow, reproduce, germinate, and more.

One current research project looks at how climate change will impact red oak trees, an important species and carbon sink. He is also examining climate change’s impact on plants that are important for food systems.

“We know that climate change is impacting plant function, and it’s impacting plant nutrition, and that impacts human nutrition,” Des Marais says. “Climate change is going to affect where we can grow which plants efficiently and how we can get them where they need to be eaten. All that stuff is critically important for humans.”

Understanding plants for the planet

Des Marais says it’s been fascinating working with experts in environmental chemistry and fluid mechanics as part of the Parsons Laboratory. He sees his role as director as being primarily about fostering community.

“The students are great,” Des Marais says. “It’s a funky old building, but we all work well together. We try to make sure everyone feels comfortable and welcome in the space.”

Even if he can’t take credit for his houseplants, Des Marais appreciates the vital role plants play for the planet.

“Plants are the foundation of the food chain, and there’s no way to get around that,” Des Marais says. “The efficiency of plant photosynthesis directly impacts our nutrition, and it impacts what we can feed the animals that we eat. It affects carbon cycling and water cycling — put simply, plant biology matters.”

Robert Hall PhD ’67 and Michael Woodford PhD ’83, scholars who both earned their doctorates at MIT, have been awarded the 2026 Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel.

Hall, a professor at Stanford University, and Woodford, a professor at Columbia University, were recognized for work that helped revitalize ideas about how government policy can spur growth to help end economic downturns. The Nobel citation describes their research as contributing to “new macroeconomic models that have significantly improved our ability to implement effective stabilisation policy.”

Both scholars helped integrate microeconomic analysis into macroeconomic models. In Hall’s case, the Nobel committee particularly cited research he conducted in the late 1970s shedding new light on economic expectations and household consumption. The committee also spotlighted Woodford’s construction of a comprehensive and influential macroeconomic model that was built on microeconomic foundations and introduced in his 2003 book, “Interest and Prices: Foundations of a Theory of Monetary Policy.” 

The work of both scholars helped update models for intervention during economic crises, based on the ideas of John Maynard Keynes, which had worked many times in the 20th century but encountered newfound difficulties adapting to the conditions of the 1970s. So-called New Keynesian models used for economic stabilization today are a result of both economists’ research, the citation noted. 

Hall is the Robert and Carole McNeil Joint Senior Fellow at the Hoover Institution and Professor in the Department of Economics at Stanford University. He received his BA in economics from the University of California at Berkeley before earning his PhD from MIT.

Woodford is the the John Bates Clark Professor of  Political Economy at Columbia University. He received a BA from the University of Chicago and a JD from Yale University Law School prior to receiving his PhD from MIT. 

Remarkably, Hall becomes the fifth former doctoral student principally advised by MIT economist Robert M. Solow to earn the Nobel Prize in economics. The others are George Akerlof PhD ’66, Peter Diamond PhD ’63, William Nordhaus PhD ’67, and Joseph Stiglitz PhD ’67, all of whom had Solow as their principal advisor in MIT’s graduate program. Additionally, MIT professor of biology H. Robert Horvitz ’68, a Nobel laureate in 2002 in the category of physiology or medicine, wrote his undergraduate thesis in economics under the direction of Solow. In 1987, Solow received the Nobel Prize himself. 

Woodford did his doctoral work under the primary supervision of Timothy Kehoe, now of the University of Minnesota — although Woodford does thank Solow in the first paragraph of his doctoral thesis, for helping him “distinguish important problems from exercises of merely technical interest.”

There are now 17 MIT alumni who have won the Nobel Prize in economics. Previously, 10 people have won the award while serving on the MIT faculty, and eight former faculty have also been named recipients.

With artificial intelligence capabilities continuing to expand, universities now face complex, urgent questions about how to navigate increasing challenges, as well as new opportunities. Much attention is now directed toward this, including an MIT report on AI and education published this summer. 

A new essay from MIT Statistics and Data Science Center Director Sasha Rakhlin, the Distinguished Professor in Data, Systems, and Society, IDSS and Brain and Cognitive Sciences, synthesizes recent discussions and readings about how AI is changing academia — particularly mathematics, statistics, machine learning, and engineering. Here, focusing primarily on graduate research and education, he describes key questions that departments and universities need to think about, and how they can best work with AI going forward.

Q: What is changing in research, and why is it happening so quickly?

A: Mathematics illustrates how quickly AI capabilities are advancing. Last year, a model reached gold-medal level at the International Mathematical Olympiad. Only a year later, models are producing new research results, including a recently proposed solution to one of the Millennium Prize Problems. Tasks that once demonstrated advanced mathematical ability can increasingly be performed by AI.

A central factor determining the pace of AI progress in a particular discipline is the speed and reliability of verification. Formalized proofs can be checked automatically; programs can be run and tested. Where evaluation is fast and reliable, systems can generate candidates, learn from outcomes, and improve. This also applies to AI research itself: improving models, training procedures, and their supporting tools. Better models can contribute to the next round of development, creating a compounding process that accelerates progress.

For experienced researchers, this creates opportunities to pursue questions whose technical demands previously put them out of reach. It also sharpens the distinction between obtaining a solution and understanding why it works, what generalizes, and what to ask next. However, we should not assume that abstraction, judgment, or problem formulation will remain beyond AI. Universities should prepare for a near future in which AI is smarter than all of us in many, perhaps most, aspects of our intellectual work.

Q: How should departments rethink academic credit and graduate training?

A: In a growing number of fields, a polished paper is becoming a weaker signal of individual expertise. Departments must begin now to reconsider what they reward, without simply defining valuable work as whatever AI cannot yet do. Asking good questions, replication, synthesis of ideas, informative negative results, and shared datasets may deserve greater recognition. More broadly, evaluation should establish what a researcher contributed and takes intellectual responsibility for, including when substantial parts of the work were performed by AI. These expectations should guide hiring, promotion, and funding — and be made explicit for current and incoming PhD students.

Training poses a harder problem. Intellectual muscles develop through exercise. Routine calculations, coding, failed approaches, and small discoveries have long helped students acquire intuition and judgment. Delegating this work can remove formative experiences, even as AI allows students to attempt more ambitious projects. We need to distinguish unnecessary friction from work through which expertise develops. Students should learn to formulate problems, audit model outputs, reproduce results, and defend their choices. Fundamentals may become more valuable as the basis for using these tools well.

Q: What should MIT and other universities build now?

A: This is where I see a real opportunity for universities. Let's first be clear about the goal: Universities should be able to pursue questions over long horizons, share results openly, and evaluate claims independently. Partnerships with industry will be essential, but we should not assume that commercial priorities will cover the breadth of science or remain aligned with it over time. Some degree of technological independence will therefore be necessary.

Academia's most important strategic asset may be the accumulated knowledge and experience within its laboratories. Published papers present a selective record: Failed experiments, abandoned directions, and the reasons an approach did not work often remain unpublished. Scientists and engineers hold tacit knowledge, acquired through experience, about which interventions are likely to fail and why. This missing context may help explain why models currently struggle in some domains, especially the empirical sciences, to anticipate consequences that are clear to an expert. These limitations may be temporary, and capturing negative results and scientists' interpretations could make models much better at scientific exploration.

We can imagine a future in which MIT's laboratories function more like a single, living, scientific organism, connected through shared AI research infrastructure. For example, suppose a neuroscience laboratory needs better methods for segmenting neurons in microscopy images. An AI agent could recognize a relevant advance from a computer vision group, connect the researchers, propose benchmarks, and help them iterate. It could surface unresolved questions to students and faculty, and make lessons from one laboratory available to others. To make this possible, we should build workflows that capture hypotheses, interventions, outcomes, failures, and interpretations. Shared systems should connect these workflows, allowing agents to use tools and information across laboratories with appropriate permissions. This will require substantial public and institutional investment now in compute, secure data systems, and expertise in adapting and post-training AI models.

Tracing this process could also preserve the lineage of ideas and make contributions, including those of graduate students, easier to recognize. With agreed rules for consent and credit, researchers could collaborate more openly, with greater confidence that their work would be acknowledged. This could also help answer the earlier question of how to recognize and reward intellectual contributions.

AI systems can already synthesize and reason about information from more sources than any individual researcher could absorb, and their ability to make useful connections will continue to improve. We should use this capacity to bring us together around hard scientific and engineering problems, helping us build on one another's strengths and insights. To me, this is a promising path forward, but universities need to invest now to make it possible.

With artificial intelligence capabilities continuing to expand, universities now face complex, urgent questions about how to navigate increasing challenges, as well as new opportunities. Much attention is now directed toward this, including an MIT report on AI and education published this summer. 

A new essay from MIT Statistics and Data Science Center Director Sasha Rakhlin, the Distinguished Professor in Data, Systems, and Society, IDSS and Brain and Cognitive Sciences, synthesizes recent discussions and readings about how AI is changing academia — particularly mathematics, statistics, machine learning, and engineering. Here, focusing primarily on graduate research and education, he describes key questions that departments and universities need to think about, and how they can best work with AI going forward.

Q: What is changing in research, and why is it happening so quickly?

A: Mathematics illustrates how quickly AI capabilities are advancing. Last year, a model reached gold-medal level at the International Mathematical Olympiad. Only a year later, models are producing new research results, including a recently proposed solution to one of the Millennium Prize Problems. Tasks that once demonstrated advanced mathematical ability can increasingly be performed by AI.

A central factor determining the pace of AI progress in a particular discipline is the speed and reliability of verification. Formalized proofs can be checked automatically; programs can be run and tested. Where evaluation is fast and reliable, systems can generate candidates, learn from outcomes, and improve. This also applies to AI research itself: improving models, training procedures, and their supporting tools. Better models can contribute to the next round of development, creating a compounding process that accelerates progress.

For experienced researchers, this creates opportunities to pursue questions whose technical demands previously put them out of reach. It also sharpens the distinction between obtaining a solution and understanding why it works, what generalizes, and what to ask next. However, we should not assume that abstraction, judgment, or problem formulation will remain beyond AI. Universities should prepare for a near future in which AI is smarter than all of us in many, perhaps most, aspects of our intellectual work.

Q: How should departments rethink academic credit and graduate training?

A: In a growing number of fields, a polished paper is becoming a weaker signal of individual expertise. Departments must begin now to reconsider what they reward, without simply defining valuable work as whatever AI cannot yet do. Asking good questions, replication, synthesis of ideas, informative negative results, and shared datasets may deserve greater recognition. More broadly, evaluation should establish what a researcher contributed and takes intellectual responsibility for, including when substantial parts of the work were performed by AI. These expectations should guide hiring, promotion, and funding — and be made explicit for current and incoming PhD students.

Training poses a harder problem. Intellectual muscles develop through exercise. Routine calculations, coding, failed approaches, and small discoveries have long helped students acquire intuition and judgment. Delegating this work can remove formative experiences, even as AI allows students to attempt more ambitious projects. We need to distinguish unnecessary friction from work through which expertise develops. Students should learn to formulate problems, audit model outputs, reproduce results, and defend their choices. Fundamentals may become more valuable as the basis for using these tools well.

Q: What should MIT and other universities build now?

A: This is where I see a real opportunity for universities. Let's first be clear about the goal: Universities should be able to pursue questions over long horizons, share results openly, and evaluate claims independently. Partnerships with industry will be essential, but we should not assume that commercial priorities will cover the breadth of science or remain aligned with it over time. Some degree of technological independence will therefore be necessary.

Academia's most important strategic asset may be the accumulated knowledge and experience within its laboratories. Published papers present a selective record: Failed experiments, abandoned directions, and the reasons an approach did not work often remain unpublished. Scientists and engineers hold tacit knowledge, acquired through experience, about which interventions are likely to fail and why. This missing context may help explain why models currently struggle in some domains, especially the empirical sciences, to anticipate consequences that are clear to an expert. These limitations may be temporary, and capturing negative results and scientists' interpretations could make models much better at scientific exploration.

We can imagine a future in which MIT's laboratories function more like a single, living, scientific organism, connected through shared AI research infrastructure. For example, suppose a neuroscience laboratory needs better methods for segmenting neurons in microscopy images. An AI agent could recognize a relevant advance from a computer vision group, connect the researchers, propose benchmarks, and help them iterate. It could surface unresolved questions to students and faculty, and make lessons from one laboratory available to others. To make this possible, we should build workflows that capture hypotheses, interventions, outcomes, failures, and interpretations. Shared systems should connect these workflows, allowing agents to use tools and information across laboratories with appropriate permissions. This will require substantial public and institutional investment now in compute, secure data systems, and expertise in adapting and post-training AI models.

Tracing this process could also preserve the lineage of ideas and make contributions, including those of graduate students, easier to recognize. With agreed rules for consent and credit, researchers could collaborate more openly, with greater confidence that their work would be acknowledged. This could also help answer the earlier question of how to recognize and reward intellectual contributions.

AI systems can already synthesize and reason about information from more sources than any individual researcher could absorb, and their ability to make useful connections will continue to improve. We should use this capacity to bring us together around hard scientific and engineering problems, helping us build on one another's strengths and insights. To me, this is a promising path forward, but universities need to invest now to make it possible.