Princeton and PPPL researchers successfully tested PACMAN, an AI framework that controls fusion plasma in real time, making decisions in about 20 milliseconds and even predicting and preventing instabilities before they occur.
Summary
Princeton and PPPL researchers successfully tested the PACMAN AI framework, which controls fusion plasma in real time with decisions made in 20 milliseconds and predicts instabilities.
Similar Articles
Bringing AI to the next generation of fusion energy
DeepMind announces a research partnership with Commonwealth Fusion Systems to advance fusion energy using AI, leveraging deep reinforcement learning and the TORAX plasma simulator to optimize the SPARC tokamak's performance and accelerate the path to net fusion energy.
Fusion-fission forecasts when AI will shift to undesirable behavior
This paper presents a method using a vector generalization of fusion-fission group dynamics to forecast when AI behavior shifts from desirable to undesirable, validated across multiple models and scales, offering a real-time warning signal below the current safety stack.
Offline Reinforcement Learning for Plasma Control in Nuclear Fusion: Codebase and Benchmark
This paper introduces RL4F, an offline reinforcement learning benchmark for plasma control in nuclear fusion, providing closed-loop evaluation environments and baseline comparisons across four profile tracking tasks using real tokamak data from DIII-D. The codebase and datasets are open-sourced to foster further research.
AI said "One minute to midnight. Nobody launch." Then it ordered a nuke. I gave 4 AI models the nuclear button in my war game.
A developer built a browser war game called SECOND STRIKE that pits AI models against each other with nuclear buttons; DeepSeek ended the world 13/15 times while GPT only did so 3/15, and the data and replays are open for humans to compete on a live scoreboard.
Physical Superintelligence
PSI is building a vertically integrated factory for physical superintelligence to accelerate physics breakthroughs with artificial superintelligence, and has open-sourced an AI copilot for physicists called Get Physics Done (GPD).