Jeff Dean, one of Google‘s longest-serving executives, is leaving the company to launch a new artificial intelligence startup. The Jeff Dean startup, called Discovery Loop, aims to use AI to speed up scientific research, and Dean plans to serve as chief executive.
Three other senior Google researchers are joining him as co-founders: Sanjay Ghemawat, a senior fellow at Google; Quoc Le, a founding member of Google Brain; and Oriol Vinyals, a senior research scientist at Google DeepMind. Discovery Loop is registered as a public benefit corporation, a structure that lets it pursue a stated mission alongside profit.
What the Jeff Dean startup plans to build
Discovery Loop wants to run thousands of scientific experiments at once, using algorithms to design and iterate on them with less human involvement. The goal is to widen the scale at which researchers can test ideas, instead of moving through them one at a time.
The company is also looking into recursive self-improvement, where AI systems help design better versions of themselves without waiting on each human-led revision. Discovery Loop said in a press release that scientific progress has long depended on “slow, sequential human iterations, creating a significant bottleneck,” and that its systems are meant to automate full experimental loops instead.
Backers include Alphabet, Radical Ventures and Khosla Ventures
Discovery Loop has already raised money from several investors, including Google’s parent company Alphabet. Radical Ventures and Khosla Ventures are co-leading the funding round, the company said Wednesday. Kleiner Perkins, Lightspeed and Doerr Capital also put money in.
The founding team described the startup’s goals in a joint statement: “The next great frontier for AI is to go beyond answering questions and to begin making discoveries.” They added that speeding up engineering and scientific discovery would let the world see the benefits of new technologies sooner.
A departure after 27 years at Google
Dean joined Google in 1999 as the company’s 30th employee and became one of its most influential engineers. He helped build core pieces of Google’s search infrastructure, including its crawling, indexing and query-serving systems, and later worked on the multimodal models behind Google Gemini.
Dean told The New York Times that Discovery Loop is meant to remove some of the manual work from experimentation. “We think there is opportunity for AI to more fully automate what has traditionally been a very human-intensive experimental loop. You will get both a higher quantity and a higher quality of experiments, and that will lead to scientific breakthroughs and advances,” he said.
Using AI to speed up scientific discovery has interested researchers for years, though the field has mostly stayed experimental with little commercial use so far. Discovery Loop is betting that enough computing power and the right algorithms can change that.