No Priors: AI for Atoms — Periodic Labs with Liam Fedus

Summary

Liam Fedus (ex-OpenAI, co-founder of Periodic Labs) discusses applying scaling laws of large language models to the physical work of atoms — materials science. The approach uses language models as orchestration layers alongside specialized neural networks for molecular simulation. Topics include transitioning from dark matter physics to materials engineering, and how the same scaling principles driving LLMs can accelerate discovery of new materials. A frontier example of AI moving beyond software into physical-world impact.

Source: No Priors Podcast Date: 2026-04-03

Listen to episode