Synthesis Superintelligence: from Semiconductors to Superconductors — Periodic Labs’ Liam Fedus and Ekin Dogus Cubuk
Periodic Labs bets AI's next frontier comes from real-world physical experimentation, not more internet data.
“the next frontier of AI won't come from simply training on more internet data, it will come from letting models experiment with the real world”
Periodic Labs, founded just one year ago, is building 'synthesis superintelligence' by grounding reinforcement learning in physical laboratory experiments rather than internet-scale text. Founders Liam Fedus and Ekin Dogus Cubuk argue that scientific discovery requires reasoning under uncertainty and noise in ways that pure language model scaling cannot address. Their autonomous lab vision — giving every instrument '140 IQ' and training on failed experiments — could compress decades of materials science into months, with implications for superconductors, batteries, and compute hardware.