Former OpenAI and DeepMind researchers raise whopping $300M seed to automate science 

Periodic Labs came retired of stealth connected Tuesday with a warfare thorax of $300 cardinal arsenic a effect round, backed by a tech industry’s who’s who: Andreessen Horowitz, DST, Nvidia, Accel, Elad Gil, Jeff Dean, Eric Schmidt, and Jeff Bezos.

Period Labs was founded by Ekin Dogus Cubuk and Liam Fedus. Cubuk led the materials and chemistry squad astatine Google Brain and DeepMind, wherever 1 of his projects was, for instance, an AI instrumentality called GNoME. That instrumentality discovered implicit 2 cardinal caller crystals successful 2023, materials that could 1 time beryllium utilized to powerfulness caller generations of technology, researchers say.

Fedus is simply a erstwhile VP of Research astatine OpenAI, and 1 of the researchers who helped make ChatGPT. He besides led the squad that created the first trillion-parameter neural network.

Its tiny squad is likewise filled with researchers who person worked connected different large AI and materials subject projects, from gathering OpenAI’s cause Operator to moving connected Microsoft’s MatterGen, an LLM materials subject find AI.

The extremity of Periodic Labs is thing little than to automate technological discovery, creating AI scientists, the institution says. This means gathering labs wherever robots behaviour carnal experiments, cod data, iterate, and effort again, learning and improving arsenic they go.

The lab’s archetypal extremity is to invent caller superconductors that it hopes execute amended and perchance necessitate little vigor than existing superconducting materials. But the well-funded startup besides hopes to find different caller materials.

Another extremity is to cod each the carnal satellite information that its AI scientists nutrient arsenic they premix and vigor and different manipulate assorted powers and earthy materials successful their hunt for thing new.

“Until now, technological AI advances person travel from models trained connected the internet” and LLMs person “exhausted” the net arsenic a root that tin beryllium consumed, the institution says successful an introductory blog post. “At Periodic, we are gathering AI scientists and the autonomous laboratories for them to operate.”

The anticipation is that, not lone volition the labs invent next-generation materials, but they volition nutrient invaluable caller information that AI models tin devour to proceed their evolution.

While this mightiness beryllium 1 of the astir awesome groups of researchers to assemble a startup for this purpose, it’s not the lone 1 moving connected AI scientists. AI arsenic a instrumentality to automate chemistry discoveries has been a topic of world probe since astatine slightest 2023. It is the pursuit of tiny startups similar Tetsuwan Scientific, arsenic good arsenic non profits similar Future House and the University of Toronto’s Acceleration Consortium.