Synthesis Superintelligence: from Semiconductors to Superconductors — Periodic Labs’ Liam Fedus and Ekin Dogus Cubuk
About
Liam Fedus and Ekin Dogus Cubuk of Periodic Labs explain why scientific discovery differs from mathematics and coding, and what it takes to build AI scientists that discover new materials. They describe “synthesis superintelligence”: reinforcement learning grounded in physical experiments, AI-powered materials characterisation, simulation and density functional theory, and high-throughput labs whose systems learn from the whole process of doing science, including failed experiments, rather than from published results alone, along with how prediction, synthesis and characterisation fit together in the materials discovery loop.