AI for Atoms: How Periodic Labs is Revolutionizing Materials Engineering with Co-Founder Liam Fedus
- •
Scientific AI requires a closed-loop interface with the physical world - progress in material science is bottlenecked by the inability of purely digital models to conduct experiments and learn from real-world feedback loops.
“Science ultimately isn't sitting in a room thinking really hard. You have to conduct experiments. You have to learn from them. You have to interface with reality.”
- •
Domain-specific labs should leverage existing LLM priors - Periodic Labs focuses its R&D exclusively on material science while utilizing third-party foundation models for coding and general reasoning to accelerate development.
“Periodic spends zero effort on improving coding models. We're incredibly impressed by Codex, Cloud Code and so that's been a huge accelerator for the company.”
- •
Physical experimentation provides the ground truth missing from literature - because scientific papers often contain noisy or contradictory data spanning multiple orders of magnitude, physical labs are required to ground ML models in reality.
“One of the engineers on our team was looking at a reported material property. And it was just sort of extracted values from literature. It was really interesting to see the reported value spanned many orders of magnitude. And so you train an ML system on that and it's like, well, the best you can do is model this distribution, but you're no closer to like a ground truth.”