NSF extends MIT-led AI-physics institute funding, doubling impact potential

The Institute for Artificial Intelligence and Fundamental Interactions (IAIFI), led by MIT, has secured renewed funding from the National Science Foundation (NSF) for another five years. Annual support has increased from $4 million to $4.98 million, signaling a new phase of expansion for the interdisciplinary research hub that bridges AI and physics.

A model built on dual innovation

Since its 2020 launch as part of the NSF’s National Artificial Intelligence Research Institutes program, IAIFI has demonstrated how artificial intelligence can transform physics—and how physics can refine AI systems. The institute’s core philosophy hinges on this bidirectional relationship: AI tools reveal new pathways in physics, while physics principles guide the development of more robust and interpretable AI models.

Jesse Thaler, IAIFI’s director and an MIT physics professor, describes the approach as a "virtuous cycle." "From the beginning, IAIFI has operated on the principle that AI can unlock deeper insights in physics, while physics can shape better AI systems," he explains. "Over the past five years, this exchange has not only produced groundbreaking results but has fundamentally changed how we approach scientific discovery."

Interdisciplinary breakthroughs in real time

IAIFI’s research spans four critical domains: particle physics, nuclear physics, astrophysics, and foundational AI. Each area benefits from cross-disciplinary collaboration, yielding advances that would be difficult—or impossible—to achieve in isolation.

• Particle physics: Researchers have designed AI algorithms capable of processing the massive data streams generated by the Large Hadron Collider in real time. These tools help physicists filter noise and extract meaningful signals from collision data, effectively transforming raw information into actionable insights.
• Nuclear physics: AI-driven generative models are now being used to simulate quark and gluon interactions within lattice quantum chromodynamics. This work provides fresh perspectives on the fundamental building blocks of matter, rooted in first-principles calculations.
• Astrophysics: Machine learning is enhancing the sensitivity of LIGO, the gravitational-wave observatory co-led by MIT. These improvements make it possible to detect fainter cosmic events and uncover previously hidden phenomena.

Beyond these applications, IAIFI is also embedding physics-informed principles into AI architectures. By incorporating symmetries, geometric constraints, and statistical rigor into neural networks, researchers are developing systems that are not only more reliable but also more transparent and data-efficient.

Mike Williams, IAIFI’s interim director and an MIT physics professor, emphasizes the broader implications. "AI is reshaping how we tackle physics’ most complex challenges," he says. "More importantly, it’s expanding the boundaries of what we can even attempt to study—opening doors to questions that were once considered unattainable."

Investing in the next wave of scientists

A defining strength of IAIFI lies in its commitment to nurturing early-career researchers. The institute’s postdoctoral fellowship program pairs emerging scientists with mentors spanning both physics and AI, fostering collaboration across institutions. To date, eight fellows have completed the program, with several transitioning to faculty roles, research positions in top AI companies, or leadership roles in startups.

Phiala Shanahan, IAIFI’s interim deputy director and an MIT physics professor, highlights the program’s impact. "The IAIFI Fellowship demonstrates what happens when early-career researchers are empowered to work beyond traditional silos," she notes. "Our fellows aren’t just contributing to isolated fields—they’re actively shaping a new discipline at the intersection of physics and AI."

IAIFI’s annual PhD Summer School has become a cornerstone for this growing community, often described as training for "centaur scientists"—researchers fluent in both physics and AI. The 2026 edition drew nearly 600 applications for just 100 in-person spots, with an additional 300 participants joining virtually. Past attendees consistently rank the program highly for its blend of lectures, hands-on workshops, coding sprints, and networking opportunities.

At MIT, IAIFI has also influenced new educational pathways, including an interdisciplinary PhD program in physics, statistics, and data science. Since 2021, this collaboration between the Department of Physics and the Statistics and Data Science Center has graduated 20 doctoral students. Faculty members Phil Harris and Isaac Chuang have further expanded access to this knowledge by developing a computational data science course, offered both on campus and as a free online resource via MITx.

Building a collaborative future

IAIFI’s influence extends beyond research and education. The institute hosts an annual summer workshop, this year taking place at MIT’s Schwarzman College of Computing, and engages the public through partnerships with the MIT Museum, the Museum of Science in Boston, and interactive hackathons. Its online content—widely viewed across platforms—demystifies the intersection of AI and physics for global audiences.

Nergis Mavalvala, dean of the MIT School of Science, underscores the importance of this collaborative model. "IAIFI exemplifies what’s possible when researchers from physics, computation, statistics, and data science unite around shared scientific questions," she says. "Sustained, cross-disciplinary collaboration is not just beneficial—it’s essential for the future of discovery."

Led by a team including Jesse Thaler (currently on sabbatical), Mike Williams, Phiala Shanahan, and managing director Marisa LaFleur, IAIFI continues to redefine the boundaries of interdisciplinary science. As funding and participation grow, the institute stands as a model for how AI and physics can advance together—each field propelling the other toward uncharted territories of knowledge.