Physical Sciences & Engineering track brings together researchers and engineers working to close the gap between AI and physical-world systems — and in each case, that gap has already narrowed further than most people expect. Recovering compact, interpretable laws directly from data, instead of treating a system as a black box, has moved out of toy examples and into real scientific discovery. A single model trained across dozens of distinct physical systems is starting to act as a genuine surrogate for simulation, generalizing to dynamics it was never directly trained on. Large-scale learned forecasting has already outperformed the numerical systems that have run our weather and ocean predictions for decades, on real storms and real events, not just benchmarks. Physics-informed AI has left the lab entirely, delivering measured, order-of-magnitude speedups in live industrial deployments. And AI-driven screening has identified genuinely new, stable material candidates at a scale no manual process could match, feeding directly into the next generation of materials research. Each speaker will share their own account of what got them there — before a closing panel that asks the harder question: how far does this actually go, and where does the evidence run out