Emerging Algorithms and Mathematical Paradigms for Reliable Next-Generation AI

Sampling is a fundamental algorithmic task underpinning Bayesian inference, scientific simulation and uncertainty quantification, and it now drives modern AI, from diffusion models for image generation to large language models for text. Because these systems are inherently probabilistic, rigorous guarantees of accuracy, safety and reliability require connecting foundational mathematics with computational practice, and the scale of modern AI, with thousands of GPUs and massive datasets, demands that theory evolve alongside hardware. This IPAM workshop convenes leaders from academia and industry across probability, statistics and geometry; machine learning, AI and data science; and industrial research labs focused on deployment, to chart future directions in sampling that keep AI robust, interpretable and mathematically sound. The programme includes a poster session; registration includes a remote-participation option.