Probability and Mathematical Physics Seminar
Generative AI and Diffusion Models: a Statistical Physics Perspective
Speaker: Giulio Biroli, École normale supérieure (Paris)
Location: Warren Weaver Hall 1302
Date: Friday, April 24, 2026, 11:10 a.m.
Synopsis:
Generative AI represents a groundbreaking development within the broader “Machine Learning Revolution,” significantly influencing technology, science, and society. In this talk, I will focus on remarkable connections between generative AI, in particular diffusion models, and statistical physics. Diffusion models are the state of the art to generate images, videos, and sounds. They are very fascinating algorithms for physicists and mathematicians, as they are very much connected to concepts from stochastic processes, stochastic thermodynamics, particularly time-reversed Langevin dynamics, and transport of measures. These diffusion models start from a simple white noise input and make it evolve through a Langevin process to generate complex outputs such as images, videos, and sounds. I will show that statistical physics provides principles and methods to characterise this generation process. The emergence of features can be understood through the lens of symmetry breaking and dynamical phase transitions, whereas the emergence of creativity - or its absence (memorisation) - can be studied using theories of disordered systems, and their connections to stable laws for sums of random exponentials. I will conclude discussing current research lines and highlighting a deep connection between multi-scale generation of diffusion models and the renormalisation group analysis of physical systems.