AI for Science

Catalog description

AI is transforming scientific discovery - from protein structures prediction, to new materials discovery, to plasma control. This graduate-level seminar surveys the emerging discipline of AI for Science, emphasizing the recurring ML-theoretic themes most salient to scientific regimes: e.g., multi-scale modeling, multi-modal architectures, differentiable simulation, and inference - time reasoning. Weekly meetings pair a classical milestone (e.g., AlphaFold, AlphaGo) with a recent breakthrough (e.g., GNoME, GraphCast, TAE's Optometrist algorithm). Students read, present, and critique papers, replicate key results in lightweight coding labs, and pursue an open - ended research project targeting new science or tooling.