Anima Anandkumar on Trillion-Param Universal Models & Physical AI

AnimaAnandkumar · x · 2026-08-27

Anima Anandkumar appeared on the Latent Space podcast to discuss AI for physical simulation and understanding. She hinted at scaling these methods to trillion-parameter model sizes with 4D context lengths in the trillions. The conversation covered building high-resolution AI-based weather models that are tens of thousands of times faster and as accurate as existing forecasts, predicting plasma behavior in fusion for real-time corrective action, and how physical understanding helps replace experiments.

Related event: Anima Anandkumar on neural operators and trillion-parameter physics AI(2 posts)→

Original post →

More from Research

Research channel →