Periodic Labs' Liam Fedus on AI scientists, autonomous labs and synthesis superintelligence
Latent Space · youtube · 2026-10-08
Latent Space hosts Periodic Labs' Liam Fedus and Ekin Dogus Cubuk on their bet that AI's next frontier comes from real-world experiments, not more internet data.
Key points:
- Scientific discovery differs fundamentally from math/coding: it requires reasoning under noise, uncertainty and missing information.
- RL changes when the environment is the physical world; experiments remain the ultimate ground truth, with DFT and simulation as aids.
- The materials discovery loop (prediction, synthesis, AI-powered characterization) points toward a "matter compiler" and synthesis superintelligence.
- Failed experiments may be among the most valuable training data; models should learn the process of doing science, not just published answers.
- Vision: every lab instrument with "140 IQ," autonomous labs compressing decades of trial-and-error into months across superconductors, magnets, batteries and compute.
- Even future frontier models will still need to physically experiment; quantum computing won't automatically solve materials discovery.
More from AGI Musings
- Model Spotted Reusing Its Own Prior Results in Reasoning Traces, a Spark of Theory Building — QuintinPope5 · 2026-10-08
- Dev Argues Coding Is Solved by AI, the Hard Part Is Coming Up with Novel Ideas — Suspicious_Fun_6338 · 2026-10-08
- How many of us will be the last generation to die of something curable? — rand_longevity · 2026-10-08
- AI agents threaten India's $250B IT services engine and 6 million jobs, column warns — nordicinst · 2026-10-08
- Game Fan Translator Deletes 3 Years of Work After Finding AI Output Better Than Their Own — PixelGray38 · 2026-10-08
- Mathematicians really do think like this: the proof is a side effect of enjoying the process — mjuric · 2026-10-08