A cohort of more than 20 top A.I. researchers has decamped from OpenAI, Google DeepMind and Meta to join Periodic Labs, a San Francisco startup co-founded by ChatGPT co-creator Liam Fedus and former DeepMind scientist Ekin Dogus Cubuk. Backed with over $300 million in seed funding led by a16z, Periodic aims to build A.I. systems that pair large-scale models with real-world, robot-run experiments to accelerate discoveries in physics, chemistry and related fields. The company’s pitch departs from Big Tech’s push for “superintelligence” and AGI, arguing that breakthrough science requires data from physical experimentation, not text alone. The talent shift includes researchers turning down major stock packages at incumbent labs. Periodic plans a Menlo Park facility where robots will iterate through thousands of experiments, while A.I. learns from the results to guide new trials. The bet echoes past industrial research powerhouses like Bell Labs and IBM Research. It’s an ambitious, capital-intensive approach that could take years to pay off, but proponents say the potential to speed materials and drug discovery is worth the risk.
Related article:
AlphaFold: A solution to a 50-year-old grand challenge in biology
