Jeff Dean Is Leaving Google to Speed Up Science with AI

Jeff Dean, one of Google's longest-tenured and most influential executives, is leaving the search giant to launch his own AI startup. Departing alongside him as co-founders are several of the company's top specialists: senior engineer and Google Senior Fellow Sanjay Ghemawat, key researcher and Google Brain co-founder Quoc Le, and Google DeepMind senior scientist Oriol Vinyals. Dean himself is reportedly set to take the role of chief executive officer.

The team has founded Discovery Loop — a public benefit corporation that hopes to use artificial intelligence to radically accelerate scientific research.

Thousands of experiments at once

Discovery Loop intends to harness computational algorithms to launch and iteratively refine thousands of experiments simultaneously. The goal is to partially automate the research process itself and take the scale of experimentation to a fundamentally new level. The company is also interested in recursive self-improvement — using AI to build more powerful AI, which would remove humans from the iteration loop entirely.

“Although science and engineering have advanced society enormously over the centuries, progress has traditionally relied on slow, sequential human iterations, which creates a serious bottleneck,” the company said in a statement. According to the founders, automating full experimental cycles should transform both the speed and the efficiency of innovation.

The idea of accelerating scientific discovery with AI has long occupied the research and technology community, but until recently it remained largely an experimental field with limited commercial application.

Who's funding it

The startup has already secured backing from a range of sources, including Google's parent company, Alphabet. The seed round is being led jointly by Radical Ventures and Khosla Ventures, with participation from Kleiner Perkins, Lightspeed, and Doerr Capital.

Dean has worked at Google since 1999 and was the company's 30^th employee. Over the years he has had a hand in the core search infrastructure, including the system for crawling and indexing pages and the query-processing system, and he has also played a notable role in the development of Google's multimodal Gemini models and in the company's early AI research.

“The next great frontier for AI is to move beyond answering questions and start making discoveries,” the team said in a joint statement. In an interview with the New York Times, Dean added that automating the traditionally labor-intensive experimental cycle would make it possible to run experiments in greater numbers and of higher quality, which would lead to scientific breakthroughs.

Source: TechCrunch