
Can AI Learn Physics? Foundation Models for the Fourth State of Matter | SOSV Deep Tech LIVE
Plasma behaves in ways that are brutally hard to model. Up until recently, a single plasma simulation could take months. Every slow simulation means fewer designs tested, more physical prototypes built, greater cost, and ideas that never get explored.
A new class of AI is changing that.
Zenithon, a SOSV HAX portfolio company, is building the first world models for extreme physics, and specifically for plasma physics.In this SOSV Deep Tech LIVE, Susan Schofer (Partner & Chief Scientist at SOSV HAX) will talk with the Zenithon team, Alex Higginbottom (CEO & Co-Founder) and Dr. Lu Lu (Chief Scientist & Assistant Professor of Statistics & Data Science at Yale University), on what changes when one simulation’s worth of compute buys you a million design points.
▪️The simulation bottleneck: why does plasma and extreme-physics simulation take months, and what does it cost fusion and semiconductor / wafer fabrication equipment teams? Why is Zenithon now able to unlock this bottleneck?
▪️Physics-informed vs. general-purpose AI: what is a physics-informed neural network, and why can world models grounded in physics (with uncertainty estimates) be trusted in mission-critical settings?
▪️Closing the lab-to-industry gap: How do you move from academic ML research into tools engineers actually use in existing workflows?
📍 Broadcast live from our SOSV NY event space, this is the next installment of SOSV Deep Tech LIVE, our video podcast series exploring the frontiers of deep tech with leading founders, researchers, and technologists.
Weitere Episoden von „SOSV Deep Tech LIVE“



Verpasse keine Episode von “SOSV Deep Tech LIVE” und abonniere ihn in der kostenlosen GetPodcast App.








