Thermodynamic AI and the Fluctuation Frontier - Understanding Physics-Based Hardware for AI Acceleration

Watch a 59-minute Qiskit Seminar Series presentation exploring the intersection of thermodynamics and artificial intelligence, where Patrick Coles introduces the concept of Thermodynamic AI and its potential impact on computing. Discover how physics-inspired AI algorithms like generative diffusion models, Bayesian neural networks, Monte Carlo sampling, and simulated annealing can be unified under a single mathematical framework. Learn about a novel computing paradigm that leverages natural stochastic fluctuations as computational resources, contrasting it with quantum computing where noise is typically considered a limitation. Explore the fundamental building blocks of Thermodynamic AI hardware, including stochastic units (s-units) and Maxwell's demon devices, and understand how these components can be programmed through gate sequences. Gain insights from Coles' extensive background in solid-state physics, quantum computing, and information theory as he presents potential applications, analytical speedups, and practical implementations of this innovative approach to computing.