Jeff Dean, one of Google's most influential executives, is stepping down after nearly 27 years to launch Discovery Loop, an AI startup focused on automating scientific and engineering research. The company is structured as an independent public benefit corporation, and Google's parent Alphabet is among its financial backers.
Joining Dean as co-founders are Sanjay Ghemawat, a Google senior fellow, Quoc Le, a founding member of Google Brain, and Oriol Vinyals, who served as vice president at DeepMind. Dean, who has worked at Google since 1999 and was the company's 30th employee, reportedly plans to serve as CEO of the new venture.
The startup intends to use AI to initiate and iterate thousands of experiments simultaneously, with the goal of partially automating the research process. At launch, Discovery Loop plans to tackle automation of machine-learning research and engineering before eventually branching into areas such as hardware design, drug discovery, and clean energy challenges.
The company is also exploring using AI to create more powerful AI, a process known as recursive self-improvement, which would cut human iteration out of the experimental loop entirely. The founding team said the next great frontier for AI is moving beyond answering questions to making discoveries.
The initial funding round is being co-led by Radical Ventures and Khosla Ventures, with Kleiner Perkins, Lightspeed, and Doerr Capital also participating. Alphabet's involvement as an investor and cloud partner positions Discovery Loop with significant backing from its former home.
Dean told the New York Times that AI can more fully automate what has traditionally been a human-intensive experimental loop, leading to both higher quantity and higher quality of experiments. In an interview before launch, he said the idea for the startup came together only a few weeks ago.
Discovery Loop's launch follows Dean's long tenure shaping Google's core search infrastructure, including its crawling, indexing, and query-serving systems, as well as his leading role in early AI research and contributions to Gemini's multimodal models.
