Theoretical Physics for Artificial Intelligence

Organizers:

Adam Brown, Google DeepMind & Stanford
Ethan Dyer, Anthropic
Dmitry Krotov, IBM Research & Massachusetts Institute of Technology
Eva Silverstein, Stanford University

Some of the world’s most ambitious and consequential experiments are taking place not in particle accelerators or space telescopes but in silicon, training ginormous neural networks. The results have been transformative, yet much of the progress has been driven by empirical advances, with theoretical understanding struggling to keep pace. This meeting will explore how the tools and insights of theoretical physics can deepen our understanding of modern artificial intelligence. We will bring together physicists and computer scientists, with a shared goal of illuminating the principles underlying successful machine learning methods—and ultimately guiding the development of better architectures and algorithms.