Google AI Architect Jeff Dean Exits After 27 Years to Launch Discovery Loop
A team of Google's most influential AI scientists is departing to automate the scientific method via a new public benefit corporation.
Jeff Dean, Google’s Chief Scientist and a foundational architect of the company's AI efforts, has left the search giant after nearly 27 years to launch a new startup. The venture, called Discovery Loop, represents a significant exodus of institutional knowledge from Alphabet as the company battles for dominance in the generative AI era.
Dean is joined by a cohort of fellow Google veterans, including Sanjay Ghemawat, Quoc Le—both co-founders of Google Brain—and Oriol Vinyals, who previously served as a VP of Research at DeepMind. Structured as a public benefit corporation, Discovery Loop aims to move beyond using AI as a simple tool for researchers. Instead, the company intends to create an automated version of the scientific method, where AI systems autonomously propose, implement, and evaluate experiments to accelerate breakthroughs in biology, chip design, and machine learning architecture.
The Shift Toward Autonomous Research
This departure comes at a critical juncture for Google. Dean and Ghemawat were among the company's earliest employees, responsible for the core computing infrastructure that allowed Google to scale globally. Their exit, alongside other key leads from the Gemini and DeepMind projects, underscores a growing trend of "superstar" researchers leaving big tech for the agility of smaller ventures. According to Oriol Vinyals, the move was driven by a desire to escape the inertia inherent in large organizations to build something fundamentally different.
Industry Implications
The focus of Discovery Loop signals a strategic pivot in how AI is applied to hard sciences. While current models are often used to summarize data or suggest hypotheses, Discovery Loop is pursuing a model of recursive self-improvement. Vinod Khosla, whose firm provided funding, noted that the fundamental shift is moving from using AI to do research to having the AI actually be the researcher. If successful, this approach could disrupt the traditional timelines of drug discovery and materials science by removing human bottlenecks from the experimental loop.
Strategic Partnerships and Funding
Despite the high-profile exit, the relationship between the founders and their former employer remains collaborative. Alphabet is a founding investor in Discovery Loop and has committed to providing the startup with compute power for its first year of operation. Additional funding for the venture was provided by Radical Ventures and Khosla Ventures, among other firms.
Observers will now watch whether this autonomous approach to discovery can produce tangible scientific breakthroughs or if the complexity of physical-world experimentation proves too great for current AI architectures to manage without human intervention.