Periodic Labs, a new venture founded by former researchers from Google Brain, DeepMind, and OpenAI, has emerged from stealth operations after securing a $300 million seed funding round. The substantial investment was backed by a consortium of notable entities and individuals, including Andreessen Horowitz, DST, Nvidia, Accel, Elad Gil, Jeff Dean, Eric Schmidt, and Jeff Bezos.
Co-founded by Ekin Dogus Cubuk and Liam Fedus, Periodic Labs aims to automate scientific discovery through the development of AI scientists and autonomous laboratories. Cubuk, who previously led the materials and chemistry team at Google Brain and DeepMind, was instrumental in projects such as the AI tool GNoME, which identified over 2 million new crystal structures in 2023. Fedus, a former VP of Research at OpenAI, contributed to the development of ChatGPT and led the team behind the first trillion-parameter neural network.
The company's strategy involves establishing autonomous laboratories where robotic systems conduct physical experiments, collect data, and iterate on findings, continuously learning and refining their approaches. Periodic Labs' initial objective is to invent new superconductors designed for enhanced performance and reduced energy consumption. Additionally, the initiative seeks to generate extensive physical world data from these AI-driven experiments, which will serve as new input for the ongoing evolution of AI models.
In an introductory blog post, Periodic Labs stated that previous scientific AI advancements primarily stemmed from models trained on internet data, suggesting that large language models (LLMs) have "exhausted" the internet as a source. The company indicates its focus is on building AI scientists and the autonomous laboratories necessary for their operation. While Periodic Labs represents a significant entrant in this field, the concept of AI automating chemistry discoveries has been a subject of academic research since at least 2023, with other entities such as Tetsuwan Scientific, Future House, and the University of Toronto's Acceleration Consortium also pursuing similar objectives.
