MIT Schwarzman College of Computing

Congratulations to Paul Liang, a shared faculty member between the MIT Media Lab and MIT Schwarzman College of Computing, on being named to Forbes’ 30 Under 30 list in the category of science for 2025. Liang, an assistant professor who joined MIT in September 2024, builds AI systems that learn and interact with the world in new ways. His work has laid the theoretical foundations which enables AI to understand and retrieve multimedia content, infer human intents and navigate social interactions, and even identify humor and sarcasm in conversations. Forbes’ 30 Under 30 is an annual list that recognizes young innovators, disruptors, and visionaries building the future in 20 industries. https://lnkd.in/gnrPZa-Q

A new method developed by MIT researchers uses generative AI to produce sharp, lifelike 3D shapes that are closer in quality to the best model-generated 2D images. “In the long run, our work can help facilitate the process to be a co-pilot for designers, making it easier to create more realistic 3D shapes,” says Artem Lukoianov, an electrical engineering and computer science graduate student who is lead author of a paper on this technique. https://lnkd.in/eBdQ9srv

Marzyeh Ghassemi, associate professor in MIT EECS and MIT Institute for Medical Engineering and Science (IMES) and principal investigator at the MIT Laboratory for Information and Decision Systems (LIDS), works on the deep study of how machine learning can be made more robust and subsequently applied to improve safety and equity in health. Captivated as a child by video games and puzzles, Ghassemi was also fascinated at an early age in health. Luckily, she found a path where she could combine the two interests. “Although I had considered a career in health care, the pull of computer science and engineering was stronger,” says Ghassemi,. “When I found that computer science broadly, and AI/ML specifically, could be applied to health care, it was a convergence of interests.” https://bit.ly/MGhassemi

AI systems are being trained to make decisions across fields like robotics, medicine, and political science. But unfortunately, teaching an AI system to make good decisions is no easy task. Reinforcement learning models, which underlie these AI decision-making systems, still often fail when faced with even small variations in the tasks they are trained to perform. To address this, MIT researchers developed a more efficient algorithm that strategically selects key tasks for training. This approach improves performance on related tasks while reducing training costs. “We were able to see incredible performance improvements, with a very simple algorithm, by thinking outside the box. An algorithm that is not very complicated stands a better chance of being adopted by the community because it is easier to implement and easier for others to understand,” says senior author Cathy Wu, the Thomas D. and Virginia W. Cabot Career Development Associate Professor in Civil and Environmental Engineering and the MIT Institute for Data, Systems, and Society (IDSS), and a member of the Laboratory for Information and Decision Systems. https://lnkd.in/ehKStA-P

Sara Beery, an assistant professor of artificial intelligence and decision-making in MIT EECS, combines machine learning with human expertise to better understand a rapidly changing planet. According to the World Wide Fund for Nature, it is estimated that nearly 70% of wild animals have vanished since the 1970’s. With advancements in cameras, satellite imaging, acoustic sensors, drone-based surveys, animal tracking devices and other ecosystem sensing equipment, scientists are collecting a huge explosion of wildlife data that can help to understand this global crisis. The vast majority of that data, however, is sitting in hard drives under someone’s desk untouched. Beery’s research has shown that AI can help in filtering through this enormous flood of ecological data to discover ecosystem trends and biodiversity losses. “We have to figure out how to make use of a combination of expert human intelligence and large-scale, hopefully very robust machine learning systems,” she says. https://bit.ly/SBeery Story: Eric Bender for MIT Industrial Liaison Program (ILP) Photo: David Sella, MIT ILP

As part of the MIT Institute for Data, Systems, and Society (IDSS) Initiative on Combatting Systemic Racism (ICSR), researchers are building an open data repository to advance research on racial inequity in domains like policing, housing, and health care. “There’s extensive research showing racial discrimination and systemic inequity in essentially all sectors of American society,” explains Fotini Christia, the Ford International Professor of Social Sciences in the Department of Political Science, director of IDSS, and co-lead of ICSR. “Newer research demonstrates how computational technologies, typically trained or reliant on historical data, can further entrench racial bias. But these same tools can also help to identify racially inequitable outcomes, to understand their causes and impacts, and even contribute to proposing solutions.” https://lnkd.in/eJVFHS5Z

Applications are now open for MEnTorEd Opportunities in Research (METEOR), a postdoctoral fellowship program to support exceptional scholars in computer science and artificial intelligence while broadening participation in the field. Established in 2020 by the MIT Computer Science and Artificial Intelligence Laboratory (CSAIL), METEOR is now spanning MIT through the Schwarzman College of Computing. Learn and apply at https://bit.ly/MITMETEOR

Researchers from MIT's Laboratory for Information and Decision Systems (LIDS) and beyond, show that even the best-performing large language models don’t form a true model of the world and its rules, and can thus fail unexpectedly on similar tasks. “One hope is that, because LLMs can accomplish all these amazing things in language, maybe we could use these same tools in other parts of science, as well. But the question of whether LLMs are learning coherent world models is very important if we want to use these techniques to make new discoveries,” says Ashesh Rambachan, assistant professor of economics, a principal investigator in LIDS, and senior author of the paper on the study. https://lnkd.in/ey5PRd5j

The MIT Schwarzman College of Computing has launched the Tayebati Postdoctoral Fellowship Program, which will support leading postdocs to bring cutting-edge AI to bear on research in scientific discovery or music. https://lnkd.in/eVUsdBqm

MIT researchers developed a training technique inspired by large language models that combines diverse, multimodal data from simulations, real robots, and various sensors into a shared “language” for generative AI. This approach enables robots to learn multiple tasks without retraining from scratch. “In robotics, people often claim that we don’t have enough training data. But in my view, another big problem is that the data come from so many different domains, modalities, and robot hardware. Our work shows how you’d be able to train a robot with all of them put together,” says Lirui Wang, an electrical engineering and computer science (EECS) graduate student and lead author of a paper on this technique. Wang’s co-authors include fellow EECS graduate student Jialiang Zhao; Xinlei Chen, a research scientist at Meta; and senior author Kaiming He, an associate professor in EECS and a member of the Computer Science and Artificial Intelligence Laboratory (CSAIL). https://lnkd.in/eVWif6iz