Blog

What makes work valuable? Michal Masny, the NC Ethics of Technology Postdoctoral Fellow in the MIT Department of Philosophy, investigates the role work plays in our lives and its impact on our well-being. 

Masny sees numerous benefits to work, beyond a paycheck. It’s a space for people to develop excellence at something, make a social contribution, gain social recognition, and create and sustain community. 

“Consider a future in which we shorten the work week, or one in which we eliminate work altogether,” Masny says. “I don’t believe either of these scenarios would be unambiguously good for everyone.”

“Work is both necessary and positively valuable,” he argues, further suggesting that our lives might be worsened if we were to eliminate work completely. “There can be optimal combinations of work and leisure time.”

Masny is completing his two-year term in the NC Ethics of Technology Fellowship at the end of the spring semester. In addition to advancing his research, Masny has been working to foster dialogue and educate students on issues at the intersection of philosophy and computing. This semester, Masny is teaching an undergraduate course, 24.131 (Ethics of Technology).

Masny advocates for an updated approach to educating complete, socially aware students. “I want to create scientists who think about their projects and potential outcomes as lawyers and philosophers might, and vice versa,” he says. Masny argues for the importance of eliminating the “wisdom gap” between these groups, citing scientist Carl Sagan’s warning about the dangers of becoming “powerful without becoming commensurately wise” as scientific and technological advances continue.

“The traditional division of labor is that scientists and engineers invent new technologies, and then philosophers and lawyers evaluate and regulate them,” he continues. “But the pace at which new technologies are invented and deployed has made this division of labor untenable.” 

Established in 2021 with support from the NC Cultural Foundation, the fellowship was created with the goal of advancing critical discourse and research in the ethics of technology and AI at MIT, and by making important research and information available to the global community. 

Venture capitalist Songyee Yoon, founder and managing partner of AI-focused investment firm Principal Venture Partners and a supporter of the NC Ethics of Technology Fellowship, believes technology and scientific discovery are among humanity’s most valuable public goods, and artificial intelligence represents the most consequential technology of our time. 

“If we want the fabric of our society to be built responsibly, we must train our builders upstream, at the very moment they begin learning to design and scale technology. There is no better place to begin this work than MIT,” she says. “Supporting the Ethics of Technology Fellows Program was born from that conviction, and I am deeply encouraged to see it embraced at MIT.”

“In philosophy, you’re supposed to question everything”

Masny arrived at MIT in fall 2024, following a year as a postdoc at the Kavli Center for Ethics, Science, and the Public at the University of California at Berkeley. Originally from Poland, Masny received his PhD in philosophy from Princeton University after completing studies at Oxford University and the University of Warwick in the United Kingdom. 

He works mainly in value theory, ethics of technology, and social and political philosophy. His current research interests include the nature of human and animal well-being, our obligations to future generations, the risk of human extinction, the future of work, and anti-aging technology. 

During his tenure in the fellowship, Masny has published several research articles on ethical issues concerning the future of humanity — a topic closely relevant to thinking about the existential risks of AI development and deployment. 

“In philosophy, you’re supposed to question everything,” he says.  

Masny’s work in the fellowship continues a tradition of collaborative investigation and exploration that MIT encourages and celebrates. In fall 2024, Masny co-taught an introductory undergraduate course, STS.006J/24.06J (Bioethics), with Robin Scheffler, an associate professor in the Program in Science, Technology, and Society. 

During the 2024-25 academic year, Masny led a student research group, “Deepfakes: Ethical, Political, and Epistemological Issues,” as a part of the Social and Ethical Responsibilities of Computing (SERC) Scholars Program. The group explored the ethical, political, and epistemological dimensions of concerns over misleading deepfakes, and how they can be mitigated.

Students in Masny’s cohort spent spring 2025 working in small groups on a number of projects and presented their findings in a poster session during the MIT Ethics of Computing Research Symposium at the MIT Schwarzman College of Computing.

In summer 2025, Masny assisted with a summer course in philosophy, 24.133/134 (Experiential Ethics), in which students subject their computer science and engineering projects to ethical scrutiny with the help of trained philosophers. 

He’s encouraged by the opportunities to test his ideas and share them with people who can help refine and improve them. 

Communities of practice and engagement

When considering the value of his experience at MIT, Masny lauds the philosophy department and the opportunities to collaborate with so many different kinds of scholars. To answer the kinds of questions his research uncovers, he says, you must range further afield. He values the space MIT creates for broad inquiry while also seeking connections between his findings on work, its value, and the human impact of technology on our social lives. 

“Typically, undergraduate philosophy courses include two hour-long lectures followed by discussion; a lecture is like an audiobook,” he says. Instead, he believes, they should more like listening to a podcast or watching a talk show. 

“I want the class to be an event in a student’s schedule,” he continues. 

Masny is also considering how to integrate valuable philosophical tools into life outside the classroom. Philosophy and research can support other kinds of inquiry. Developing philosophers’ mindsets is a net positive, by his reckoning. Designing better questions, for example, can lead to better, more insightful, more accurate answers. It can also improve students’ abilities to identify challenges.

Masny will begin teaching at the University of Colorado at Boulder in fall 2026, and wants to test new ideas while continuing his research into the value of work. 

Kieran Setiya, the Peter de Florez Professor in Philosophy and head of the Department of Linguistics and Philosophy, says the NC Ethics of Technology Postdoctoral Fellowship has allowed MIT to bring in a series of exceptional young philosophers working at the intersection of ethics and AI, studying the systemic effects of new computing technologies and the moral, social, and political challenges they pose.

“This is just the kind of applied interdisciplinary thinking we need to support and sustain at MIT,” he adds.

Training a large artificial intelligence model is expensive, not just in dollars, but in time, energy, and computational resources. Traditionally, obtaining a smaller, faster model either requires training a massive one first and then trimming it down, or training a small one from scratch and accepting weaker performance. 

Researchers at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL), Max Planck Institute for Intelligent Systems, European Laboratory for Learning and Intelligent Systems, ETH, and Liquid AI have now developed a new method that sidesteps this trade-off entirely, compressing models during training, rather than after.

The technique, called CompreSSM, targets a family of AI architectures known as state-space models, which power applications ranging from language processing to audio generation and robotics. By borrowing mathematical tools from control theory, the researchers can identify which parts of a model are pulling their weight and which are dead weight, before surgically removing the unnecessary components early in the training process.

“It’s essentially a technique to make models grow smaller and faster as they are training,” says Makram Chahine, a PhD student in electrical engineering and computer science, CSAIL affiliate, and lead author of the paper. “During learning, they’re also getting rid of parts that are not useful to their development.”

The key insight is that the relative importance of different components within these models stabilizes surprisingly early during training. Using a mathematical quantity called Hankel singular values, which measure how much each internal state contributes to the model’s overall behavior, the team showed they can reliably rank which dimensions matter and which don’t after only about 10 percent of the training process. Once those rankings are established, the less-important components can be safely discarded, and the remaining 90 percent of training proceeds at the speed of a much smaller model.

“What’s exciting about this work is that it turns compression from an afterthought into part of the learning process itself,” says senior author Daniela Rus, MIT professor and director of CSAIL. “Instead of training a large model and then figuring out how to make it smaller, CompreSSM lets the model discover its own efficient structure as it learns. That’s a fundamentally different way to think about building AI systems.”

The results are striking. On image classification benchmarks, compressed models maintained nearly the same accuracy as their full-sized counterparts while training up to 1.5 times faster. A compressed model reduced to roughly a quarter of its original state dimension achieved 85.7 percent accuracy on the CIFAR-10 benchmark, compared to just 81.8 percent for a model trained at that smaller size from scratch. On Mamba, one of the most widely used state-space architectures, the method achieved approximately 4x training speedups, compressing a 128-dimensional model down to around 12 dimensions while maintaining competitive performance.

“You get the performance of the larger model, because you capture most of the complex dynamics during the warm-up phase, then only keep the most-useful states,” Chahine says. “The model is still able to perform at a higher level than training a small model from the start.”

What makes CompreSSM distinct from existing approaches is its theoretical grounding. Conventional pruning methods train a full model and then strip away parameters after the fact, meaning you still pay the full computational cost of training the big model. Knowledge distillation, another popular technique, requires training a large “teacher” model to completion and then training a second, smaller “student” model on top of it, essentially doubling the training effort. CompreSSM avoids both of these costs by making informed compression decisions mid-stream.

The team benchmarked CompreSSM head-to-head against both alternatives. Compared to Hankel nuclear norm regularization, a recently proposed spectral technique for encouraging compact state-space models, CompreSSM was more than 40 times faster, while also achieving higher accuracy. The regularization approach slowed training by roughly 16 times because it required expensive eigenvalue computations at every single gradient step, and even then, the resulting models underperformed. Against knowledge distillation on CIFAR-10, CompressSM held a clear advantage for heavily compressed models: At smaller state dimensions, distilled models saw significant accuracy drops, while CompreSSM-compressed models maintained near-full performance. And because distillation requires a forward pass through both the teacher and student at every training step, even its smaller student models trained slower than the full-sized baseline.

The researchers proved mathematically that the importance of individual model states changes smoothly during training, thanks to an application of Weyl’s theorem, and showed empirically that the relative rankings of those states remain stable. Together, these findings give practitioners confidence that dimensions identified as negligible early on won’t suddenly become critical later.

The method also comes with a pragmatic safety net. If a compression step causes an unexpected performance drop, practitioners can revert to a previously saved checkpoint. “It gives people control over how much they’re willing to pay in terms of performance, rather than having to define a less-intuitive energy threshold,” Chahine explains.

There are some practical boundaries to the technique. CompreSSM works best on models that exhibit a strong correlation between the internal state dimension and overall performance, a property that varies across tasks and architectures. The method is particularly effective on multi-input, multi-output (MIMO) models, where the relationship between state size and expressivity is strongest. For per-channel, single-input, single-output architectures, the gains are more modest, since those models are less sensitive to state dimension changes in the first place.

The theory applies most cleanly to linear time-invariant systems, although the team has developed extensions for the increasingly popular input-dependent, time-varying architectures. And because the family of state-space models extends to architectures like linear attention, a growing area of interest as an alternative to traditional transformers, the potential scope of application is broad.

Chahine and his collaborators see the work as a stepping stone. The team has already demonstrated an extension to linear time-varying systems like Mamba, and future directions include pushing CompreSSM further into matrix-valued dynamical systems used in linear attention mechanisms, which would bring the technique closer to the transformer architectures that underpin most of today’s largest AI systems.

“This had to be the first step, because this is where the theory is neat and the approach can stay principled,” Chahine says. “It’s the stepping stone to then extend to other architectures that people are using in industry today.”

“The work of Chahine and his colleagues provides an intriguing, theoretically grounded perspective on compression for modern state-space models (SSMs),” says Antonio Orvieto, ELLIS Institute Tübingen principal investigator and MPI for Intelligent Systems independent group leader, who wasn’t involved in the research. “The method provides evidence that the state dimension of these models can be effectively reduced during training and that a control-theoretic perspective can successfully guide this procedure. The work opens new avenues for future research, and the proposed algorithm has the potential to become a standard approach when pre-training large SSM-based models.”

The work, which was accepted as a conference paper at the International Conference on Learning Representations 2026, will be presented later this month. It was supported, in part, by the Max Planck ETH Center for Learning Systems, the Hector Foundation, Boeing, and the U.S. Office of Naval Research.

MIT.nano has announced that 16 startups became active participants in its START.nano program in 2025, more than doubling the number of new companies from the previous year. Aimed at speeding the transition of hard-tech innovation to market, START.nano supports new ventures through the discounted use of MIT.nano shared facilities and a guided access to the MIT innovation ecosystem. The newly engaged startups are developing solutions for some of the world’s greatest challenges in health, climate, energy, semiconductors, novel materials, and quantum computing.

“The unique resources of MIT.nano enable not just the foundational research of academia, but the translation of that research into commercial innovations through startups,” says START.nano Program Manager Joyce Wu SM ’00, PhD ’07. “The START.nano accelerator supports early-stage companies from MIT and beyond with the tools and network they need for success.”

Launched in 2021, START.nano aims to increase the survival rate of hard-tech startups by easing their journey from the lab to the real world. In addition to receiving access to MIT.nano’s laboratories, program participants are invited to present at startup exhibits at MIT conferences, and in exclusive events including the newly launched PITCH.nano competition.

“For an early-stage startup working at the frontier of superconductor discovery, the combination of infrastructure and community has been irreplaceable,” says Jason Gibson, CEO and co-founder of Quantum Formatics. “START.nano isn’t just a resource,” adds Cynthia Liao MBA ’24, CEO and co-founder of Vertical Semiconductor. “It’s a strategic advantage that accelerates our roadmap, allowing us to iterate quickly to meet customer needs and strengthen our competitive edge.”

Although an MIT affiliation is not required, five of the 16 companies in the new cohort are led by MIT alumni, and an additional three have MIT affiliation. In total, 49 percent of the startups in START.nano are founded by MIT graduates.

Here are the intended impacts of the 16 new START.nano companies:

Acorn Genetics is developing a “smartphone of sequencing,” launching the power of genetic analysis out of slow, centralized labs and into the hands of consumers for fast, portable, and affordable sequencing.

Addis Energy leverages oil, gas, and geothermal drilling technologies to unlock the chemical potential of iron-rich rocks. By injecting engineered fluids, they harness the earth’s natural energy to produce ammonia that is both abundant and cost-effective.

Augmend Health uses virtual reality and AI to deliver clinical data intelligence services for specialty care that turns incomplete documentation into revenue, compliance, and better treatment decisions.

Brightlight Photonics is building high-performance laser infrastructure at chip scale, integrating Titanium:Sapphire gain to deliver broadband, high-power, low-noise optical sources for advanced photonic systems.

Cahira Technologies is creating the new paradigm of brain-computer symbiosis for treating intractable diseases and human augmentation through autonomous, nonsurgical neural implants.

Copernic Catalysts is leveraging computational modeling to develop and commercialize transformational catalysts for low-cost and sustainable production of bulk chemicals and e-fuels.

Daqus Energy is unlocking high-energy lithium-ion batteries using critical metal-free organic cathodes.

Electrified Thermal Solutions is reinventing the firebrick to electrify industrial heat.

Guardion is making analytical instruments, chemical detectors, and radiation detectors more sensitive, portable, and easier to scale with nanomaterial-based ion detectors.

Mantel Capture is designing carbon capture materials to operate at the high temperatures found inside boilers, kilns, and furnaces — enabling highly efficient carbon capture that has not been possible until now.

nOhm Devices is developing highly-efficient cryogenic electronics for quantum computers and sensors.

Quantum Formatics is speeding discovery of the world’s next superconductors using proprietary AI.

Qunett is building the foundational hardware stack for deployable quantum networks to power the next era of global connectivity.

Rheyo is developing new ways to make dental care more effective, efficient, and easy through advanced materials and technology.

Vertical Semiconductor is commercializing high-voltage, high-density, high-efficiency vertical GaN (gallium nitride) to power the next era of compute.

VioNano Innovations is developing specialty material solutions that reduce variability and improve precision in semiconductor manufacturing, allowing chipmakers to build even smaller, faster, and more cost-effective chips.

START.nano now comprises over 32 companies and 11 graduates — ventures that have moved beyond the prototyping stages, and some into commercialization. See the full list here.

To improve data center efficiency, multiple storage devices are often pooled together over a network so many applications can share them. But even with pooling, significant device capacity remains underutilized due to performance variability across the devices.

MIT researchers have now developed a system that boosts the performance of storage devices by handling three major sources of variability simultaneously. Their approach delivers significant speed improvements over traditional methods that tackle only one source of variability at a time.

The system uses a two-tier architecture, with a central controller that makes big-picture decisions about which tasks each storage device performs, and local controllers for each machine that rapidly reroute data if that device is struggling.

The method, which can adapt in real-time to shifting workloads, does not require specialized hardware. When the researchers tested this system on realistic tasks like AI model training and image compression, it nearly doubled the performance delivered by traditional approaches. By intelligently balancing the workloads of multiple storage devices, the system can increase overall data center efficiency.

“There is a tendency to want to throw more resources at a problem to solve it, but that is not sustainable in many ways. We want to be able to maximize the longevity of these very expensive and carbon-intensive resources,” says Gohar Chaudhry, an electrical engineering and computer science (EECS) graduate student and lead author of a paper on this technique. “With our adaptive software solution, you can still squeeze a lot of performance out of your existing devices before you need to throw them away and buy new ones.”

Chaudhry is joined on the paper by Ankit Bhardwaj, an assistant professor at Tufts University; Zhenyuan Ruan PhD ’24; and senior author Adam Belay, an associate professor of EECS and a member of the MIT Computer Science and Artificial Intelligence Laboratory. The research will be presented at the USENIX Symposium on Networked Systems Design and Implementation.

Leveraging untapped performance

Solid-state drives (SSDs) are high-performance digital storage devices that allow applications to read and write data. For instance, an SSD can store vast datasets and rapidly send data to a processor for machine-learning model training.   

Pooling multiple SSDs together so many applications can share them improves efficiency, since not every application needs to use the entire capacity of an SSD at a given time. But not all SSDs perform equally, and the slowest device can limit the overall performance of the pool.

These inefficiencies arise from variability in SSD hardware and the tasks they perform.

To utilize this untapped SSD performance, the researchers developed Sandook, a software-based system that tackles three major forms of performance-hampering variability simultaneously. “Sandook” is an Urdu word that means “box,” to signify “storage.”

One type of variability is caused by differences in the age, amount of wear, and capacity of SSDs that may have been purchased at different times from multiple vendors.

The second type of variability is due to the mismatch between read and write operations occurring on the same SSD. To write new data to the device, the SSD must erase some existing data. This process can slow down data reads, or retrievals, happening at the same time.

The third source of variability is garbage collection, a process of gathering and removing outdated data to free up space. This process, which slows SSD operations, is triggered at random intervals that a data center operator cannot control.

“I can’t assume all SSDs will behave identically through my entire deployment cycle. Even if I give them all the same workload, some of them will be stragglers, which hurts the net throughput I can achieve,” Chaudhry explains.

Plan globally, react locally

To handle all three sources of variability, Sandook utilizes a two-tier structure. A global schedular optimizes the distribution of tasks for the overall pool, while faster schedulers on each SSD react to urgent events and shift operations away from congested devices.

The system overcomes delays from read-write interference by rotating which SSDs an application can use for reads and writes. This reduces the chance reads and writes happen simultaneously on the same machine.

Sandook also profiles the typical performance of each SSD. It uses this information to detect when garbage collection is likely slowing operations down. Once detected, Sandook reduces the workload on that SSD by diverting some tasks until garbage collection is finished.

“If that SSD is doing garbage collection and can’t handle the same workload anymore, I want to give it a smaller workload and slowly ramp things back up. We want to find the sweet spot where it is still doing some work, and tap into that performance,” Chaudhry says.

The SSD profiles also allow Sandook’s global controller to assign workloads in a weighted fashion that considers the characteristics and capacity of each device.

Because the global controller sees the overall picture and the local controllers react on the fly, Sandook can simultaneously manage forms of variability that happen over different time scales. For instance, delays from garbage collection occur suddenly, while latency caused by wear and tear builds up over many months.

The researchers tested Sandook on a pool of 10 SSDs and evaluated the system on four tasks: running a database, training a machine-learning model, compressing images, and storing user data. Sandook boosted the throughput of each application between 12 and 94 percent when compared to static methods, and improved the overall utilization of SSD capacity by 23 percent.

The system enabled SSDs to achieve 95 percent of their theoretical maximum performance, without the need for specialized hardware or application-specific updates.

“Our dynamic solution can unlock more performance for all the SSDs and really push them to the limit. Every bit of capacity you can save really counts at this scale,” Chaudhry says.

In the future, the researchers want to incorporate new protocols available on the latest SSDs that give operators more control over data placement. They also want to leverage the predictability in AI workloads to increase the efficiency of SSD operations.

“Flash storage is a powerful technology that underpins modern datacenter applications, but sharing this resource across workloads with widely varying performance demands remains an outstanding challenge. This work moves the needle meaningfully forward with an elegant and practical solution ready for deployment, bringing flash storage closer to its full potential in production clouds,” says Josh Fried, a software engineer at Google and incoming assistant professor at the University of Pennsylvania, who was not involved with this work.

This research was funded, in part, by the National Science Foundation, the U.S. Defense Advanced Research Projects Agency, and the Semiconductor Research Corporation.

Today, there are 94 nuclear reactors operating in the United States, more than in any other country in the world, and these units collectively provide nearly 20 percent of the nation’s electricity. That is a major accomplishment, according to Dean Price, but he believes that our country needs much more out of nuclear energy, especially at a moment when alternatives to fossil fuel-based power plants are desperately being sought. He became a nuclear engineer for this very reason — to make sure that nuclear technology is up to the task of delivering in this time of considerable need.

“Nuclear energy has been a tremendous part of our nation’s energy infrastructure for the past 60 years, and the number of people who maintain that infrastructure is incredibly small,” says Price, an MIT assistant professor in the Department of Nuclear Science and Engineering (NSE), as well as the Atlantic Richfield Career Development Professor in Energy Studies. “By becoming a nuclear engineer, you become one of a select number of people responsible for carbon-free energy generation in the United States.” 

That was a mission he was eager to take part in, and the goals he set for himself were far from modest: He wanted to help design and usher in a new class of nuclear reactors, building on the safety, economics, and reliability of the existing nuclear fleet.

Price has never wavered from this objective, and he’s only found encouragement along the way. The nuclear engineering community, he says, “is small, close-knit, and very welcoming. Once you get into it, most people are not inclined to do anything else.”

Illuminating the relationships between physical processes

In his first research project as an undergraduate at the University of Illinois Urbana at Champaign, Price studied the safety of the steel and concrete casks used to store spent reactor fuel rods after they’ve cooled off in tanks of water, typically for several years. His analysis indicated that this storage method was quite safe, although the question as to what should ultimately be done with these fuel casks, in terms of long-term disposal, remains open in this country.

After starting graduate studies at the University of Michigan in 2020, Price took up a different line of research that he’s still engaged in today. That area of study, called multiphysics modeling, involves looking at various physical processes going on in the core of a nuclear reactor to see how they interact — an alternative to studying these processes one at a time.

One key process, neutronics, concerns how neutrons buzz around in the reactor core causing nuclear fission, which is what generates the power. A second process, called thermal hydraulics, involves cooling the reactor to extract the heat generated by neutrons. A multiphysics simulation, analyzing how these two processes interact, could show how the heat carried away as the reactor produces power affects the behavior of neutrons, because the hotter the fuel is, the less likely it is to cause fission.

“If you ever want to change your power level, or do anything with the reactor, the temperature of the fuel is a critical input that you need to know,” says Price. “Multiphysics modeling allows us to correlate the fission neutronics processes with a thermal property, temperature. That, in turn, can help us predict how the reactor will behave under different conditions.”

Multiphysics modeling for light water reactors, which are the ones operating today with capacities on the order of 1,000 megawatts, are pretty well established, Prices says. But methods for modeling advanced reactors — small modular reactors (SMRs with capacities ranging from around 20 to 300 MW) and microreactors (rated at 1 to 20 MW) — are far less advanced. Only a very small number of these reactors are operating today, but Price is focusing his efforts on them because of their potential to produce power more cheaply and more safely, along with their greater flexibility in power and size.   

Although multiphysics simulations have supplied the nuclear community with a wealth of information, they can require supercomputers to solve, or find approximate solutions to, coupled and extremely difficult nonlinear equations. In the hopes of greatly reducing the computational burden, Price is actively exploring artificial intelligence approaches that could provide similar answers while bypassing those burdensome equations altogether. That has been a central theme of his research agenda since he joined the MIT faculty in September 2025.

A crucial role for artificial intelligence

What artificial intelligence and machine-learning methods, in particular, are good at is finding patterns concealed within data, such as correlations between variables critical to the functioning of a nuclear plant. For example, Price says, “if you tell me the power level of your reactor, it [AI] could tell you what the fuel temperature is and even tell you the 3-dimensional temperature distribution in your core.” And if this can be done without solving any complicated differential equations, computational costs could be greatly reduced.

Price is investigating several applications where AI may be especially useful, such as helping with the design of novel kinds of reactors. “We could then rely on the safety frameworks developed over the past 50 years to carry out a safety analysis of the proposed design,” he says. “In this way, AI will not be directly interfacing with anything that is safety-critical.” As he sees it, AI’s role would be to augment established procedures, rather than replacing them, helping to fill in existing gaps in knowledge.

When a machine-learning model is given a sufficient amount of data to learn from, it can help us better understand the relationship between key physical processes — again without having to solve nonlinear differential equations. 

“By really pinning down those relationships, we can make better design decisions in the early stages,” Price says. “And when that technology is developed and deployed, AI can help us make more intelligent control decisions that will enable us to operate our reactors in a safer and more economical way.”

Giving back to the community that nurtured him

Simply put, one of his chief goals is to bring the benefits of AI to the nuclear industry, and he views the possibilities as vast and largely untapped. Price also believes that he is well-positioned as a professor at MIT to bring us closer to the nuclear future that he envisions. As he sees it, he’s working not only to develop the next generation of reactors, but also to help prepare the next generation of leaders in the field.

Price became acquainted with some prospective members of that “next generation” in a design course he co-taught last fall with Curtis Smith, the KEPCO Professor of the Practice of Nuclear Science and Engineering. For Price, that introduction lasted just a few months, but it was long enough for him to discover that MIT students are exceptionally motivated, hard-working, and capable. Not surprisingly, those happen to be the same qualities he’s hoping to find in the students that join his research team.

Price vividly recalls the support he received when taking his first, tentative steps in this field. Now that he’s moved up the ranks from undergraduate to professor, and acquired a substantial body of knowledge along the way, he wants his students “to experience that same feeling that I had upon entering the field.” Beyond his specific goals for improving the design and operation of nuclear reactors, Price says, “I hope to perpetuate the same fun and healthy environment that made me love nuclear engineering in the first place.”

Artificial intelligence is increasingly being used to help optimize decision-making in high-stakes settings. For instance, an autonomous system can identify a power distribution strategy that minimizes costs while keeping voltages stable.

But while these AI-driven outputs may be technically optimal, are they fair? What if a low-cost power distribution strategy leaves disadvantaged neighborhoods more vulnerable to outages than higher-income areas?

To help stakeholders quickly pinpoint potential ethical dilemmas before deployment, MIT researchers developed an automated evaluation method that balances the interplay between measurable outcomes, like cost or reliability, and qualitative or subjective values, such as fairness.   

The system separates objective evaluations from user-defined human values, using a large language model (LLM) as a proxy for humans to capture and incorporate stakeholder preferences. 

The adaptive framework selects the best scenarios for further evaluation, streamlining a process that typically requires costly and time-consuming manual effort. These test cases can show situations where autonomous systems align well with human values, as well as scenarios that unexpectedly fall short of ethical criteria.

“We can insert a lot of rules and guardrails into AI systems, but those safeguards can only prevent the things we can imagine happening. It is not enough to say, ‘Let’s just use AI because it has been trained on this information.’ We wanted to develop a more systematic way to discover the unknown unknowns and have a way to predict them before anything bad happens,” says senior author Chuchu Fan, an associate professor in the MIT Department of Aeronautics and Astronautics (AeroAstro) and a principal investigator in the MIT Laboratory for Information and Decision Systems (LIDS).

Fan is joined on the paper by lead author Anjali Parashar, a mechanical engineering graduate student; Yingke Li, an AeroAstro postdoc; and others at MIT and Saab. The research will be presented at the International Conference on Learning Representations.

Evaluating ethics

In a large system like a power grid, evaluating the ethical alignment of an AI model’s recommendations in a way that considers all objectives is especially difficult.

Most testing frameworks rely on pre-collected data, but labeled data on subjective ethical criteria are often hard to come by. In addition, because ethical values and AI systems are both constantly evolving, static evaluation methods based on written codes or regulatory documents require frequent updates.

Fan and her team approached this problem from a different perspective. Drawing on their prior work evaluating robotic systems, they developed an experimental design framework to identify the most informative scenarios, which human stakeholders would then evaluate more closely.

Their two-part system, called Scalable Experimental Design for System-level Ethical Testing (SEED-SET), incorporates quantitative metrics and ethical criteria. It can identify scenarios that effectively meet measurable requirements and align well with human values, and vice versa.   

“We don’t want to spend all our resources on random evaluations. So, it is very important to guide the framework toward the test cases we care the most about,” Li says.

Importantly, SEED-SET does not need pre-existing evaluation data, and it adapts to multiple objectives.

For instance, a power grid may have several user groups, including a large rural community and a data center. While both groups may want low-cost and reliable power, each group’s priority from an ethical perspective may vary widely.

These ethical criteria may not be well-specified, so they can’t be measured analytically.

The power grid operator wants to find the most cost-effective strategy that best meets the subjective ethical preferences of all stakeholders.

SEED-SET tackles this challenge by splitting the problem into two, following a hierarchical structure. An objective model considers how the system performs on tangible metrics like cost. Then a subjective model that considers stakeholder judgements, like perceived fairness, builds on the objective evaluation.

“The objective part of our approach is tied to the AI system, while the subjective part is tied to the users who are evaluating it. By decomposing the preferences in a hierarchical fashion, we can generate the desired scenarios with fewer evaluations,” Parashar says.

Encoding subjectivity

To perform the subjective assessment, the system uses an LLM as a proxy for human evaluators. The researchers encode the preferences of each user group into a natural language prompt for the model.

The LLM uses these instructions to compare two scenarios, selecting the preferred design based on the ethical criteria.

“After seeing hundreds or thousands of scenarios, a human evaluator can suffer from fatigue and become inconsistent in their evaluations, so we use an LLM-based strategy instead,” Parashar explains.

SEED-SET uses the selected scenario to simulate the overall system (in this case, a power distribution strategy). These simulation results guide its search for the next best candidate scenario to test.

In the end, SEED-SET intelligently selects the most representative scenarios that either meet or are not aligned with objective metrics and ethical criteria. In this way, users can analyze the performance of the AI system and adjust its strategy.

For instance, SEED-SET can pinpoint cases of power distribution that prioritize higher-income areas during periods of peak demand, leaving underprivileged neighborhoods more prone to outages.

To test SEED-SET, the researchers evaluated realistic autonomous systems, like an AI-driven power grid and an urban traffic routing system. They measured how well the generated scenarios aligned with ethical criteria.

The system generated more than twice as many optimal test cases as the baseline strategies in the same amount of time, while uncovering many scenarios other approaches overlooked.

“As we shifted the user preferences, the set of scenarios SEED-SET generated changed drastically. This tells us the evaluation strategy responds well to the preferences of the user,” Parashar says.

To measure how useful SEED-SET would be in practice, the researchers will need to conduct a user study to see if the scenarios it generates help with real decision-making.

In addition to running such a study, the researchers plan to explore the use of more efficient models that can scale up to larger problems with more criteria, such as evaluating LLM decision-making.

This research was funded, in part, by the U.S. Defense Advanced Research Projects Agency.

Designers, makers, and others often use 3D printing to rapidly prototype a range of functional objects, from movie props to medical devices. Accurate print previews are essential so users know a fabricated object will perform as expected.

But previews generated by most 3D-printing software focus on function rather than aesthetics. A printed object may end up with a different color, texture, or shading than the user expected, resulting in multiple reprints that waste time, effort, and material.

To help users envision how a fabricated object will look, researchers from MIT and elsewhere developed an easy-to-use preview tool that puts appearance first.

Users upload a screenshot of the object from their 3D-printing software, along with a single image of the print material. From these inputs, the system automatically generates a rendering of how the fabricated object is likely to look.

The artificial intelligence-powered system, called VisiPrint, is designed to work with a range of 3D-printing software and can handle any material example. It considers not only the color of the material, but also gloss, translucency, and how nuances of the fabrication process affect the object’s appearance.

Such aesthetics-focused previews could be especially useful in areas like dentistry, by helping clinicians ensure temporary crowns and bridges match the appearance of a patient’s teeth, or in architecture, to aid designers in assessing the visual impact of models.

“3D printing can be a very wasteful process. Some studies estimate that as much as a third of the material used goes straight to the landfill, often from prototypes the user ends of discarding. To make 3D printing more sustainable, we want to reduce the number of tries it takes to get the prototype you want. The user shouldn’t have to try out every printing material they have before they settle on a design,” says Maxine Perroni-Scharf, an electrical engineering and computer science (EECS) graduate student and lead author of a paper on VisiPrint.

She is joined on the paper by Faraz Faruqi, a fellow EECS graduate student; Raul Hernandez, an MIT undergraduate; SooYeon Ahn, a graduate student at the Gwangju Institute of Science and Technology; Szymon Rusinkiewicz, a professor of computer science at Princeton University; William Freeman, the Thomas and Gerd Perkins Professor of EECS at MIT and a member of the Computer Science and Artificial Intelligence Laboratory (CSAIL); and senior author Stefanie Mueller, an associate professor of EECS and Mechanical Engineering at MIT, and a member of CSAIL. The research will be presented at the ACM CHI Conference on Human Factors in Computing Systems.

Accurate aesthetics

The researchers focused on fused deposition modeling (FDM), the most common type of 3D printing. In FDM, print material filament is melted and then squirted through a nozzle to fabricate an object one layer at a time.

Generating accurate aesthetic previews is challenging because the melting and extrusion process can change the appearance of a material, as can the height of each deposited layer and the path the nozzle follows during fabrication.

VisiPrint uses two AI models that work together to overcome those challenges.

The VisiPrint preview is based on two inputs: a screenshot of the digital design from a user’s 3D-printing software (called “slicer” software), and an image of the print material, which can be taken from an online source or captured from a printed sample.

From these inputs, a computer vision model extracts features from the material sample that are important for the object’s appearance.

It feeds those features to a generative AI model that computes the geometry and structure of the object, while incorporating the so-called “slicing” pattern the nozzle will follow as it extrudes each layer.

The key to the researchers’ approach is a special conditioning method. This involves carefully adjusting the inner workings of the model to guide it, so it follows the slicing pattern and obeys the constraints of the 3D-printing process.

Their conditioning method utilizes a depth map that preserves the shape and shading of the object, along with a map of the edges that reflects the internal contours and structural boundaries.

“If you don’t have the right balance of these two things, you could use up with bad geometry or an incorrect slicing pattern. We had to be careful to combine them in the right way,” Perroni-Scharf says.

A user-focused system

The team also produced an easy-to-use interface where one can upload the required images and evaluate the preview.

The VisiPrint interface enables more advanced makers to adjust multiple settings, such as the influence of certain colors on the final appearance.

In the end, the aesthetic preview is intended to complement the functional preview generated by slicer software, since VisiPrint does not estimate printability, mechanical feasibility, or likelihood of failure.

To evaluate VisiPrint, the researchers conducted a user study that asked participants to compare the system to other approaches. Nearly all participants said it provided better overall appearance as well as more textural similarity with printed objects.

In addition, the VisiPrint preview process took about a minute on average, which was more than twice as fast as any competing method.

“VisiPrint really shined when compared to other AI interfaces. If you give a more general AI model the same screenshots, it might randomly change the shape or use the wrong slicing pattern because it had no direct conditioning,” she says.

In the future, the researchers want to address artifacts that can occur when model previews have extremely fine details. They also want to add features that allow users to optimize parts of the printing process beyond color of the material.

“It is important to think about the way that we fabricate objects. We need to continue striving to develop methods that reduce waste. To that end, this marriage of AI with the physical making process is an exciting area of future work,” Perroni-Scharf says.

“‘What you see is what you get’ has been the main thing that made desktop publishing ‘happen’ in the 1980s, as it allowed users to get what they wanted at first try. It is time to get WYSIWYG for 3D printing as well. VisiPrint is a great step in this direction,” says Patrick Baudisch, a professor of computer science at the Hasso Plattner Institute, who was not involved with this work.

This research was funded, in part, by an MIT Morningside Academy for Design Fellowship and an MIT MathWorks Fellowship.

In biology, defects are generally bad. But in materials science, defects can be intentionally tuned to give materials useful new properties. Today, atomic-scale defects are carefully introduced during the manufacturing process of products like steel, semiconductors, and solar cells to help improve strength, control electrical conductivity, optimize performance, and more.

But even as defects have become a powerful tool, accurately measuring different types of defects and their concentrations in finished products has been challenging, especially without cutting open or damaging the final material. Without knowing what defects are in their materials, engineers risk making products that perform poorly or have unintended properties.

Now, MIT researchers have built an AI model capable of classifying and quantifying certain defects using data from a noninvasive neutron-scattering technique. The model, which was trained on 2,000 different semiconductor materials, can detect up to six kinds of point defects in a material simultaneously, something that would be impossible using conventional techniques alone.

“Existing techniques can’t accurately characterize defects in a universal and quantitative way without destroying the material,” says lead author Mouyang Cheng, a PhD candidate in the Department of Materials Science and Engineering. “For conventional techniques without machine learning, detecting six different defects is unthinkable. It’s something you can’t do any other way.”

The researchers say the model is a step toward harnessing defects more precisely in products like semiconductors, microelectronics, solar cells, and battery materials.

“Right now, detecting defects is like the saying about seeing an elephant: Each technique can only see part of it,” says senior author and associate professor of nuclear science and engineering Mingda Li. “Some see the nose, others the trunk or ears. But it is extremely hard to see the full elephant. We need better ways of getting the full picture of defects, because we have to understand them to make materials more useful.”

Joining Cheng and Li on the paper are postdoc Chu-Liang Fu, undergraduate researcher Bowen Yu, master’s student Eunbi Rha, PhD student Abhijatmedhi Chotrattanapituk ’21, and Oak Ridge National Laboratory staff members Douglas L Abernathy PhD ’93 and Yongqiang Cheng. The paper appears today in the journal Matter.

Detecting defects

Manufacturers have gotten good at tuning defects in their materials, but measuring precise quantities of defects in finished products is still largely a guessing game.

“Engineers have many ways to introduce defects, like through doping, but they still struggle with basic questions like what kind of defect they’ve created and in what concentration,” Fu says. “Sometimes they also have unwanted defects, like oxidation. They don’t always know if they introduced some unwanted defects or impurity during synthesis. It’s a longstanding challenge.”

The result is that there are often multiple defects in each material. Unfortunately, each method for understanding defects has its limits. Techniques like X-ray diffraction and positron annihilation characterize only some types of defects. Raman spectroscopy can discern the type of defect but can’t directly infer the concentration. Another technique known as transmission electron microscope requires people to cut thin slices of samples for scanning.

In a few previous papers, Li and collaborators applied machine learning to experimental spectroscopy data to characterize crystalline materials. For the new paper, they wanted to apply that technique to defects.

For their experiment, the researchers built a computational database of 2,000 semiconductor materials. They made sample pairs of each material, with one doped for defects and one left without defects, then used a neutron-scattering technique that measures the different vibrational frequencies of atoms in solid materials. They trained a machine-learning model on the results.

“That built a foundational model that covers 56 elements in the periodic table,” Cheng says. “The model leverages the multihead attention mechanism, just like what ChatGPT is using. It similarly extracts the difference in the data between materials with and without defects and outputs a prediction of what dopants were used and in what concentrations.”

The researchers fine-tuned their model, verified it on experimental data, and showed it could measure defect concentrations in an alloy commonly used in electronics and in a separate superconductor material.

The researchers also doped the materials multiple times to introduce multiple point defects and test the limits of the model, ultimately finding it can make predictions about up to six defects in materials simultaneously, with defect concentrations as low as 0.2 percent.

“We were really surprised it worked that well,” Cheng says. “It’s very challenging to decode the mixed signals from two different types of defects — let alone six.”

A model approach

Typically, manufacturers of things like semiconductors run invasive tests on a small percentage of products as they come off the manufacturing line, a slow process that limits their ability to detect every defect.

“Right now, people largely estimate the quantities of defects in their materials,” Yu says. “It is a painstaking experience to check the estimates by using each individual technique, which only offers local information in a single grain anyway. It creates misunderstandings about what defects people think they have in their material.”

The results were exciting for the researchers, but they note their technique measuring the vibrational frequencies with neutrons would be difficult for companies to quickly deploy in their own quality-control processes.

“This method is very powerful, but its availability is limited,” Rha says. “Vibrational spectra is a simple idea, but in certain setups it’s very complicated. There are some simpler experimental setups based on other approaches, like Raman spectroscopy, that could be more quickly adopted.”

Li says companies have already expressed interest in the approach and asked when it will work with Raman spectroscopy, a widely used technique that measures the scattering of light. Li says the researchers’ next step is training a similar model based on Raman spectroscopy data. They also plan to expand their approach to detect features that are larger than point defects, like grains and dislocations.

For now, though, the researchers believe their study demonstrates the inherent advantage of AI techniques for interpreting defect data.

“To the human eye, these defect signals would look essentially the same,” Li says. “But the pattern recognition of AI is good enough to discern different signals and get to the ground truth. Defects are this double-edged sword. There are many good defects, but if there are too many, performance can degrade. This opens up a new paradigm in defect science.”

The work was supported, in part, by the Department of Energy and the National Science Foundation.

As one of the first students in MIT’s new Music Technology and Computation Graduate Program, Mariano Salcedo ’25 is researching the intersection between artificial intelligence and music visuals.

Specifically, his graduate research focuses on neural cellular automata (NCA), which merges classical cellular automata with machine learning techniques to grow images that can regenerate.

When paired with a stimulus like music, these images can “show” sounds in action.

“This approach enables anyone to create music-driven visuals while leveraging the expressive and sometimes unpredictable dynamics of self-organized systems,” Salcedo says. Through the web interface Salcedo has designed, users can adjust the relationship between the music’s energy and the NCA system to create unique visual performances using any music audio stream.

“I want the visuals to complement and elevate the listening experience,” he says.

Last year Salcedo, the Alex Rigopulos (1992) Fellow in Music Technology and Computation, earned a BS in artificial intelligence and decision making from MIT, where he explored signal processing in machine learning and how a classical understanding of signals can inform how we understand AI. Now he’s one of five master’s students in the Music Technology and Computation Graduate Program’s inaugural cohort.

The program, directed by professor of the practice in music technology Eran Egozy ’93, MNG ’95, is a collaboration between MIT Music and Theater Arts in the School of Humanities, Arts, and Social Sciences, and the School of Engineering. It invites practitioners to study, discover, and develop new computational approaches to music. It also includes a speaker series that exposes students and the broader MIT community to music industry professionals, artists, technologists, and other researchers.

Rigopulos ’92, SM ’94, is a video game designer, musician, and former CEO of Harmonix Music Systems, a company he co-founded with Egozy in 1995. Harmonix is now a part of Epic Games, where Rigopulos is the director of game development for music.

“MIT is where I was first able to pursue my passion for music technology decades ago, and that experience was the springboard for a long and fulfilling career,” says Rigopulos. “So, when MIT launched an advanced degree program in music technology, I was thrilled to fund a fellowship to help propel this exciting new program.”

Egozy is enthusiastic about Salcedo’s work and his commitment to further exploring its possibilities. “He is a beautiful example of a multidisciplinary researcher who thinks deeply about how to best use technology to enhance and expand human creativity,” he says.

Salcedo has been selected to deliver the student address at the 2026 Advanced Degree Ceremony for the School of Humanities, Arts, and Social Sciences. “It’s an honor and it’s daunting,” he says. “It feels like a huge responsibility,” though one he’s eager to embrace. His selection also pleases Egozy. “I am super excited that Mariano was chosen to deliver this year’s keynote,” he enthuses.

Changing gears

Growing up in Mexico and Texas, Mariano Salcedo couldn’t readily indulge his passion for creating music. “There are no bands in Mexican public schools,” he says. While some families could pay for instruments and lessons, others like Salcedo’s were less fortunate.

“I’ve always loved music,” he continues. “I was a listener.”

Salcedo began his MIT journey as a mechanical engineering student, applying to MIT through the Questbridge program. “I heard if you like engineering and science that attending MIT would be a great choice,” he recalls. “Nerds are welcomed and embraced.” While he dutifully worked toward completing his MechE curriculum, music and technology came calling after a chance encounter with an LLM.

“I was introduced to an LLM chatbot and was blown away,” he recalls. “This was something that was speaking to me. I was both awed and frightened.” After his encounter with the chatbot, Salcedo switched his major from mechanical engineering to artificial intelligence and decision making.

“I basically started over after being two thirds of the way through the MechE curriculum,” he says. He learned about the possibilities available with AI but also confronted some of the challenges bedeviling researchers and developers including its potential power, ensuring its responsible use, human bias, limited access for people from underrepresented groups, and a lack of diversity among developers. He decided he might be able to change that picture.

“I thought one more person in the field could make a difference,” he says.

While completing his undergraduate studies, Salcedo’s love of music resurfaced. “I began DJ’ing at MIT and was hooked,” he says. While he hadn’t learned to play a traditional instrument, he discovered he could create engaging soundscapes with technology. “I bought a digital audio work station to help me make music,” he continues.

Egozy and Salcedo met in 2024 while Salcedo completed an Undergraduate Research Opportunities Program rotation as a game developer in Egozy’s lab. “He was incredibly curious and has grown tremendously over a very short time period,” Egozy says. Egozy became an informal, though important, mentor to Salcedo. “He brings great energy and thoughtfulness to his work, and to supporting others in the [music technology and computation graduate] program,” Egozy notes.

Salcedo also took a class with Egozy, 21M.385/21M.585/6.4450 (Interactive Music Systems), which further fed his appetite for the creativity he craved while also allowing him to indulge his fascination with music’s possibilities. By taking advantage of courses in the HASS curriculum, he further developed his understanding of music theory and related technologies.

“I took a class with professor Leslie Tilley, 21M.240 (Critically Thinking in Music), which helped establish a valuable framework for understanding music making,” he says, “while a class like 6.3000 (Signal Processing) helped me connect intuition with science.”

Working across disciplines

While Salcedo is passionate about his music and his research, he’s also invested in building relationships with his fellow students. He’s a member of the fraternity Sigma Nu, where he says he “found a home and community.” He also took a MISTI trip to Chile in summer 2023, where he conducted music technology research. Salcedo praises the culture of camaraderie at MIT and is grateful for its influence on his work as a scholar. “MIT has taught me how to learn,” he says.

Professors encouraged him to present his research and findings. He presented his work — Artificial Dancing Intelligence: Neural Cellular Automata for Visual Performance of Music — at the Association for the Advancement of Artificial Intelligence conference in Singapore in January 2026.

Salcedo believes his research can potentially move beyond music visualization. “What if we could improve the ways we model self-organized systems?” he asks. “That is, systems like multicellular organisms, flocks of birds, or societies that interact locally but exhibit interesting behaviors.” Any system, Salcedo says, where the whole is more than the sum of its parts.

Developing the technology used to design his application can potentially help answer important ethical questions regarding AI’s continued expansion and growth. The path to his work’s development is both daunting and lonely, but those challenges feed his work ethic.

“It’s intimidating to pursue this path when the academy is currently focused on LLMs,” he says. “But it’s also important to explain and explore the base technology before digging into more nuanced work, which can help audiences understand it better.” Knowing that he has the support of his professors helps Salcedo maintain excitement for his ideas. “They only ask that we ground our interests in research,” he says.

His investigations are impacting his work as a musician. “My music has gotten more interesting because of the classes I’m taking,” he says. He’s also interested in understanding whose music the academy and the world hears, exploring biases toward Western music in the canon and exploring how to reduce biases related to which kinds of music are valued.

“The work we do as technologists is far less subjective than we’re led to believe,” he believes.

Salcedo is especially grateful for the support he’s received during his time at MIT. “Program faculty encourage a variety of pursuits,” he says, “and ask us to advance our individual aims rather than focusing on theirs.” During his time in the graduate program, he notes with enthusiasm how often he’s been challenged to pursue his ideas.

Ultimately, Salcedo wants people to experience the joy he feels working at the intersection of the humanities and the sciences. Music and technology impact nearly everyone. Inviting audiences into his laboratory as participants in the creative and research processes offers the same kind of satisfaction he gets from crafting a great beat or solving for a thorny technical challenge. Helping audiences understand his work’s value fuels his drive to succeed.

“I want users to feel movement and explore sounds and their impact more fully,” he says.

Proteins are far more than nutrients we track on a food label. Present in every cell of our bodies, they work like nature’s molecular machines. They walk, stretch, bend, and flex to do their jobs, pumping blood, fighting disease, building tissue, and many other jobs too small for the eye to see. Their power doesn’t come from shape alone, but from how they move. 

In recent years, artificial intelligence has allowed scientists to design entirely new protein structures not found in nature tailored for specific functions, such as binding to viruses, or mimicking the mechanical properties of silk for sustainable materials. But designing for structure alone is like building a car body without any control over how the engine performs. The subtle vibrations, shifts, and mechanical dynamics of a protein are just as critical to its functions as its form.

Now, MIT engineers have taken a major step toward closing the gap with the development of an AI model known as VibeGen. If vibe coding lets programmers describe what they want and then AI generates the software, VibeGen does the same for living molecules: specify the vibe — the pattern of motion you want — and the model writes the protein. 

The new model allows scientists to target how a protein flexes, vibrates, and shifts between shapes in response to its environment, opening a new frontier in the design of molecular mechanics. VibeGen builds on a series of advances from the Buehler lab in agentic AI for science — systems in which multiple AI models collaborate autonomously to solve problems too complex for any single model.

“The essence of life at fundamental molecular levels lies not just in structure, but in movement,” says Markus Buehler, the Jerry McAfee Professor of Engineering in the departments of Civil and Environmental Engineering and Mechanical Engineering. “Everything from protein folding to the deformation of materials under stress follows the fundamental laws of physics.”

Buehler and his former postdoc, Bo Ni, identified a critical need for what they call physics-aware AI: systems capable of reasoning about motion, not just snapshots of molecular structure. “AI must go beyond analyzing static forms to understanding how structure and motion are fundamentally intertwined,” Buehler adds.

The new approach, described in a paper March 24 in the journal Matter, uses generative AI to create proteins with tailor-made dynamics.

Training AI to think about motion 

The revolution in AI-driven protein science has been, overwhelmingly, a revolution in structure. Tools like AlphaFold solved the decades-old problem of predicting a protein’s three-dimensional shape. Existing generative models learned to design new shapes from scratch. But in focusing on the folded snapshot — the protein frozen in place — the field largely set aside the property that makes proteins work: their motion. “Structure prediction was such a grand challenge that it absorbed the field’s attention,” Buehler says. “But a protein’s shape is just one frame of a much longer film, and the design space extends through space and time, where structure sits on a much broader manifold.” Scientists could design a protein with a particular architecture. They couldn’t yet specify how that protein would move, flex, or vibrate once it was built.

VibeGen does something no protein design tool has done before. It inverts the traditional problem. Rather than asking, “What shape will this sequence produce?” it asks, “What sequence will make a protein move in exactly this way?”

To build VibeGen, Buehler and Ni turned to a class of AI diffusion models, the same underlying technology that powers AI image generators capable of creating realistic pictures from pure noise. In VibeGen’s case, the model starts with a random sequence of amino acids and refines it, step by step, until it converges on a sequence predicted to vibrate and flex in a targeted way.

The system works through two cooperating agents that design and challenge each other. A “designer” proposes candidate sequences aimed at a target motion profile. A “predictor” evaluates those candidates, asking whether they’ll actually move the way the designer intended. The two models iterate back and forth like an internal dialogue, until the design stabilizes into something that meets the goal. By specifying this vibrational fingerprint as the design input, VibeGen inverts the usual logic: dynamics becomes the blueprint, and structure follows.

“It’s a collaborative system,” Ni says. “The designer proposes, the predictor critiques, and the design improves through that tension.”

Most sequences VibeGen produces are entirely de novo, not borrowed from nature, not a variation on something evolution already made. To confirm the designs actually work, the team ran detailed physics-based molecular simulations, and the proteins behaved exactly as intended, flexing and vibrating in the patterns VibeGen had targeted.

One of the study’s most striking findings is that many different protein sequences and folds can satisfy the same vibrational target — a property the researchers call functional degeneracy. Where evolution converged on one solution, VibeGen reveals an entire family of alternatives: proteins with different structures and sequences that nonetheless move in the same way. “It suggests that nature explored only a fraction of what’s possible,” Buehler says. “For any given dynamic behavior, there may be a large, untapped space of viable designs.”

A new frontier in molecular engineering

Controlling protein dynamics could have wide-ranging applications. In medicine, proteins that can change shape on cue hold enormous potential. Many therapeutic proteins work by binding to a target molecule — a virus, a cancer cell, a misfiring receptor. How well they bind often depends not just on their shape, but on how flexibly they can adapt to their target. A protein that is engineered with motion could grip more precisely, reduce unintended interactions, and ultimately become a safer, more effective drug.

In materials science, which is an area of Buehler’s research, mechanical properties at the molecular scale affect their performance. Biological materials like silk and collagen get their strength and resilience from the coordinated motion of their molecular building blocks. Designing proteins that are stiffer, flexible, or vibrate in a certain way could lead to new sustainable fibers, impact-resistant materials, or biodegradable alternatives to petroleum-based plastics.

Buehler envisions further possibilities: structural materials for buildings or vehicles incorporating protein-based components that heal themselves after mechanical stress, or that adjust in response to heavy load.

By enabling researchers to specify motion as a direct design parameter, VibeGen treats proteins less like static shapes and more like programmable mechanical devices. The advance bridges artificial intelligence, medicine, synthetic biology, and materials engineering — toward a future in which molecular machines can be designed with the same precision and intentionality as bridges, engines, or microchips.

“VibeGen can venture into uncharted territory, proposing protein designs beyond the repertoire of evolution, tailored purely to our specifications. It’s as if we’ve invented a new creative engine that designs molecular machines on demand,” Buehler adds.

The researchers plan to refine the model further and validate their designs in the lab. They also hope to integrate motion-aware design with other AI tools, building toward systems that can design proteins to be not just dynamic, but multifunctional; machines that sense their environment, respond to signals, and adapt in real-time.

The word “vibe” comes from vibration, and Buehler sees the connection as more than wordplay. “We’ve turned ‘vibe’ into a metaphor, a feeling, something subjective,” he says. “But for a protein, the vibe is the physics. It is the actual pattern of motion that determines what the molecule can do, the very machinery of life.”

The research was supported by the U.S. Department of Agriculture, the MIT-IBM Watson AI Lab, and MIT’s Generative AI Initiative.