Diffusion, Probability, and the Age of Creative AI

Explore the cutting-edge world of denoising diffusion models and their impact on AI art generation in this 58-minute lecture presented by Greg Ver Steeg from UC Riverside. Gain a high-level overview of recent developments in the field and delve into a new information-theoretic foundation for diffusion models. Discover how this innovative approach establishes an exact equivalence between probability and optimal denoising, leading to improvements in density estimation, interpretability through information decomposition, and faster sampling. Learn from Ver Steeg's expertise in incorporating physics and information theory concepts into machine learning to understand complex systems like human behavior and biology.