IAIFI + CSAIL Colloquium: Symbiosis of Physics and Artificial Intelligence, Konstantin Rusch

Speaker: T. Konstantin Rusch, Assistant Professor, Max Planck Institute for Intelligent Systems

Title: Symbiosis of Physics and Artificial Intelligence

Abstract: Artificial Intelligence (AI) is transforming how we advance science and engineering, yet current approaches remain limited by computational inefficiencies, insufficient theoretical grounding, and a focus on commercial rather than scientific applications. To address these, we propose an interdisciplinary research agenda that bridges physics and AI along two complementary directions: physics for AI and AI for physics. In the former, we develop physics-inspired AI models that exploit structural principles from physical systems to design architectures that are more efficient, expressive, and mathematically tractable. We further enhance scalability of our proposed models through control-theoretic in-training compression methods that enable efficient training of large models. In the latter direction, we apply AI, including our physics-inspired architectures, to challenging problems in the physical sciences, such as gravitational-wave analysis and scientific computing. One highlighted contribution in this context is Message-Passing Monte Carlo (MPMC), the first machine-learning approach for generating low-discrepancy point sets, which combines geometric deep learning with discrepancy theory and achieves 4- to 25-fold performance improvements over prior methods in physics, scientific machine learning, and robotics applications. Together, these efforts highlight the potential of uniting physics and AI to achieve models that are both powerful and principled, paving the way for the next generation of scientific AI systems.

Attend in Person: MIT Kolker Room (26-414)

Join on Zoom: https://mit.zoom.us/j/91200832411

Watch on YouTube: https://www.youtube.com/channel/UCueoFcGm_15kSB-wDd4CBZA