3 questions: On the future of AI and mathematical and physical sciences | MIT News

Curiosity-driven research has long fueled technological innovation. A century ago, curiosity about atoms led to quantum mechanics, and eventually the transistor at the heart of the modern computer. On the other hand, the steam engine was a success, but it took fundamental research in thermodynamics to fully exploit its potential.

Today, artificial intelligence and science find themselves in the same situation. The current AI revolution has been inspired by decades of research in the mathematical and physical sciences (MPS), which have provided the challenging problems, data sets, and insights that make modern AI possible. The 2024 Nobel Prizes in physics and chemistry, recognizing fundamental physics-based AI methods and AI applications to protein design, have made this connection impossible to miss.

By 2025, MIT will host ia Workshop on the future of AI+MPSfunded by the National Science Foundation with support from the MIT School of Science and the MIT Departments of Physics, Chemistry, and Mathematics. The workshop brought together leading AI researchers and scientists to plan how MPS domains can best monetize – and contribute to – the future of AI. Now a white paper, with recommendations for funding agencies, institutions, and researchers, has become published in Machine Learning: Science and Technology. In this interview, Jesse Thaler, MIT professor of physics and chair of the workshop, explains the key themes and how MIT is positioning itself to lead in AI and science.

Question: What are the main themes of this report about the collection of leaders last year in all the mathematical and physical sciences?

A: Gathering so many researchers at the forefront of AI and science in one room was enlightening. Although the workshop participants came from five different scientific communities – astronomy, chemistry, materials science, mathematics, and physics – we found many similarities in how we each engage with AI. A real consensus emerged from our animated discussions: Coordinated investments in computing and data infrastructure, diverse research strategies, and rigorous training can advance both AI and science.

One of the central ideas was that this should be a two-way street. It’s not just about using AI to do better science; science can also make AI better. Scientists are at the forefront of dissecting information from complex systems, including neural networks, by uncovering the underlying principles and behaviors that emerge. We call this “the science of AI,” and it comes in three forms: driving scientific AI, where scientific thinking informs the basic mechanisms of AI; science that inspires AI, where scientific challenges push the development of new algorithms; and the science that explains AI, where scientific tools help illuminate how machine intelligence actually works.

In my field of particle physics, for example, researchers are developing real-time AI algorithms to handle the flood of data from collider experiments. This work has direct implications for discovering new physics, but the algorithms themselves end up being very important beyond our field. The workshop made it clear that the science of AI should be a public priority – it has the potential to change the way we understand, develop, and manage AI systems.

Of course, combining science and AI requires people who can work in both worlds. Attendees consistently emphasized the need for “hundred scientists” – researchers with real interdisciplinary knowledge. Supporting these polymaths in all phases of work, from integrated undergraduate courses to interdisciplinary PhD programs to joint faculty recruitment, has emerged as essential.

Question: How do MIT’s AI and science efforts align with the workshop’s recommendations?

A: The workshop built its recommendations on three pillars: research, talent, and community. As director of the NSF Institute for Artificial Intelligence and Fundamental Interactions (IAIFI) — a collaborative effort in AI and physics between MIT and Harvard, Northeastern, and Tufts universities — I’ve seen firsthand how this framework can work. Measuring this up to MIT, we can see where progress is being made and where opportunities lie.

On the research front, MIT is already allowing AI-science work in both directions. Even faster scrolling MIT News shows how individual researchers across the School of Science are pursuing AI-driven projects, building a pipeline of knowledge and uncovering new opportunities. At the same time, collaborative efforts such as IAIFI and the Accelerated AI Algorithms for Data-Driven Discovery (A3D3) Institute are focusing interdisciplinary efforts for greater impact. The MIT Generative AI Impact Consortium also supports application-driven AI work at the university scale.

In order to develop the talent of AI-science for the first career, several initiatives are training the next generation of centaur scientists. The MIT Schwarzman College of Computing’s Common Ground for Computing Education program helps students become “bilingual” in computing and in their home practice. Interdisciplinary PhD approaches also benefit; IAIFI worked with the MIT Institute for Data, Systems, and Society to create one in physics, mathematics, and data science, and about 10 percent of physics PhD students now choose it — a number that is likely to grow. Dedicated postdoctoral roles such as the IAIFI Fellowship and the Tayebati Fellowship give early career researchers the freedom to pursue interdisciplinary work. Funding centaur scientists and giving them space to build connections across fields, universities, and careers has been revolutionary.

Finally, community building brings it all together. From focused workshops to large-scale symposia, organizing interdisciplinary events shows that AI and science is not a confined activity – it’s an emerging field. MIT has the talent and resources to make a big impact, and holding these gatherings at multiple scales helps establish that leadership.

Question: What lessons can MIT learn about advancing its AI-science efforts?

A: The workshop revealed something important: The leading institutions in AI and science will be those that think systematically, not narrowly. Resources are limited, so priorities are important. Workshop attendees were clear about what could happen if the institution linked employment, research, and training around a unified strategy.

MIT is well-positioned to build on what’s already going on with systematic initiatives — collaborative lines of inquiry across computing and science domains, expanded interdisciplinary degree pathways, and intentional funding for “AI science”. We are already seeing movement in this case; this year, the MIT Schwarzman College of Computing and the Department of Physics are conducting their first joint faculty research, which is exciting to see.

A virtuous cycle of AI-science has the potential to be truly transformative – providing deeper insights into AI, accelerating scientific discovery, and producing robust tools for both. By building a deliberate strategy, MIT will be well positioned to lead, and benefit from, the next wave of AI.