Princeton AI Outpaces Human Operators in Fusion Plasma Control

Princeton University researchers demonstrated an artificial‑intelligence system that can monitor and adjust fusion plasma in milliseconds, far quicker than a human operator can react. In a test run, the AI predicted a damaging plasma instability roughly 200 milliseconds before it manifested and automatically altered the plasma conditions to prevent the instability from forming. The experiment shows that machine‑learning algorithms can detect and mitigate plasma disruptions in real time, a capability that could reduce the risk of damage in future fusion reactors and improve the stability of sustained fusion reactions.