Princeton and PPPL test PACMAN, an AI that controls fusion plasma in real time with ~20ms decisions
Dr_Singularity · x · 2026-09-12
Researchers at Princeton and PPPL successfully tested PACMAN, an AI framework for real-time control of fusion plasma.
- Makes control decisions in about 20 milliseconds, fast enough for live tokamak operation
- Goes beyond reactive control: it can predict and prevent instabilities before they occur
- A notable AI-for-Science result in fusion energy control
Related event: Princeton's PACMAN Controls Fusion Plasma in Real Time(2 posts)→
More from Research
- LiteReality-Agent turns real-world scans into simulation-ready MuJoCo environments — elliottszwu · 2026-09-12
- Study argues self-replicating AI agents may soon be within reach of low-resource actors — rohanpaul_ai · 2026-09-12
- Auto-derived FlashAttention with SMEM and tensor core assignment shown off — vtabbott_ · 2026-09-12
- 100-Question Forecasting Benchmark: Arise PiT Search Pits Deep Research Agents vs Prediction Markets — CShorten30 · 2026-09-12
- Penn launches CIS 6280 World Models course with all lectures and slides public — thoma_gu · 2026-09-12
- Waymo's self-driving cars were roughly 900% safer than humans across 25.3M miles — Polymarket · 2026-09-12