Why the Next AI Breakthrough May Come from Physics with Max Welling - #774

Max Welling, a researcher and entrepreneur, applies geometric deep learning and equivariance to materials science through his startup, CuspAI. By utilizing generative models and molecular dynamics, the company accelerates the discovery of materials for carbon capture, semiconductors, and energy storage, effectively replacing expensive quantum mechanical calculations with efficient neural network surrogates. Beyond industrial applications, Welling explores the mathematical convergence between generative AI and non-equilibrium statistical mechanics, detailing how concepts like entropy and information theory bridge these fields. His research further investigates using spontaneous symmetry breaking to introduce wave-like propagation in neural networks, addressing challenges like oversmoothing and long-range memory in deep architectures. This cross-disciplinary approach leverages physics-based design principles to enhance model stability and performance, demonstrating a fundamental synergy between physical systems and artificial intelligence.