It’s effectively funded, backed by Jeff Bezos and Eric Schmidt, and staffed with expertise from OpenAI, Google and Meta. However in contrast to lots of its Silicon Valley friends, the brand new A.I. startup Periodic Labs has no plans to construct shopper merchandise or pursue synthetic normal intelligence (AGI).
“Our objective is to create an A.I. scientist,” stated Liam Fedus, one of many startup’s co-founders, in a submit saying its launch yesterday (Sep. 30). “Till now, scientific A.I. advances have come from fashions educated on the web. However regardless of its vastness, it’s nonetheless finite.”
To speed up discovery, Periodic Labs plans to merge A.I. with conventional analysis strategies, together with working experiments in actual labs. That mission has already secured $300 million in seed funding. The spherical was led by Andreessen Horowitz and included Nvidia, Felicis, Accel, Elad Gil, Jeff Dean, Bezos and Schmidt.
The San Francisco-based startup can also be full of famend researchers. Fedus, who co-created ChatGPT and served as vp of analysis at OpenAI, launched Periodic Labs with Ekin Dogus Cubuk, a longtime Google DeepMind scientist who previously led its supplies science and chemistry crew.
The roughly 20-person founding crew is a roster of veterans from main tech corporations, with expertise on initiatives like OpenAI’s Operator agent and Microsoft’s MatterGen LLM. In a playful nod to its identify, every staffer at Periodic Labs picks a component from the periodic desk to adorn a customized desk plate.
The corporate goals to construct A.I. programs able to forming hypotheses and working simulations in autonomous labs, stated Peter Deng, a former OpenAI government and now a normal companion at Felicis, in a current weblog submit. Deng stated he was satisfied to take a position after Fedus informed him, “so as to do science, you need to do actual science.” In different phrases, A.I. can solely study a lot from textual content—it wants real-world experiments to advance.
For now, Periodic Labs will give attention to the bodily sciences, a discipline wealthy with knowledge and verifiable outcomes that may assist fast A.I. progress. If profitable in automating supplies design, the corporate says its work may speed up breakthroughs in house journey, nuclear fusion and even Moore’s Legislation.
The startup has already begun partnering with semiconductor makers to enhance chip warmth dissipation and is coaching brokers to streamline analysis and engineering workflows. Its buyer base additionally consists of corporations in house and protection.
Periodic Labs is just not alone in its mission. Tech giants like OpenAI and Google are pursuing comparable objectives. DeepMind’s work on AlphaFold, an A.I. system that reworked protein construction prediction, even earned two of its researchers a Nobel Prize final yr. Smaller rivals embrace FutureHouse, a San Francisco nonprofit additionally working to create an autonomous A.I. scientist.
