What happened
A roundtable of fusion scientists and AI developers documented which problems in fusion research AI can genuinely solve and which ones it cannot—a shift from treating AI as a universal tool to thinking about where it actually belongs. This matters because fusion labs have been experimenting with AI for years without clear guidance on when it's the right choice, so having experts agree on what works and what doesn't means smaller labs and companies can stop wasting money on AI solutions that don't fit their problems.
Why it matters
For the first time, people actually building fusion reactors have sat down with AI specialists and documented what AI is good for (controlling plasma, analyzing sensor data) versus what it isn't (designing novel reactor geometries from scratch, making physics breakthroughs). This prevents the next 5 years of fusion R&D from being derailed by hype or misapplied machine learning.