[AINews] Jeff, Sanjay, Oriol, and Quoc depart DeepMind; Demis to Chair; Koray to SVP — what is going on at GDM???
Latent Space Latent.Space ● Covered by 18 sources
Jeff Dean, Sanjay Ghemawat, Oriol Vinyals, and Quoc Le are leaving Google DeepMind to launch a startup called Discovery Loop. Demis Hassabis steps back too, moving to Chair and Chief Scientist while Koray Kavukcuoglu takes daily control of GDM as SVP.
Based on reporting by Latent Space, Latent.Space — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Four of the people most responsible for building Google's modern AI machine are walking out the door at once. Jeff Dean, Sanjay Ghemawat, Oriol Vinyals, and Quoc Le are cofounding Discovery Loop, a public benefit corporation aimed at automating machine learning, science, and engineering research. Dean says Radical Ventures and Khosla Ventures are leading the seed round, with Lightspeed, Kleiner Perkins, Doerr Capital, and, notably, Alphabet itself chipping in.
And it's not just the founders leaving that's shaking the org chart. Demis Hassabis is moving from CEO of GDM to Chair of Google DeepMind and Chief Scientist of Alphabet, a shift that's being described as him leaning into long-term strategy, AGI work, and Isomorphic. Koray Kavukcuoglu, previously CTO, steps up to SVP and takes over day-to-day operations, including Gemini and the frontier research teams.
Google is framing all of this as amicable, and it is even investing in the new startup. But the timing is hard to ignore. This comes after John Jumper's departure for Anthropic, Noam Shazeer's move to OpenAI, and exits by David Silver and Denny Zhou, all layered on top of roughly six months since the last Gemini Pro update. Four researchers of this caliber writing a single manifesto together, rather than publishing through GDM's usual sprawling co-author lists on Gemini papers, is its own kind of signal.
The read from people watching closely, including Nathan Lambert and Andrew Ng, is that this isn't just a talent story. It's being treated as a marker that automated science and discovery loops are becoming a primary frontier in AI, not a side project. Whether that's the real reason these four chose to build outside Google rather than inside it is the open question nobody's fully answering yet.
Google still has the models, the teams, the compute, and arguably the largest pile of training data anywhere. Whether a reshuffled leadership structure and a friendly investment in a former team's spinout is the start of a comeback or just a tidy way to describe a bigger internal shift is something only the next few Gemini releases will actually settle.
My take — AI-written commentary, not fact-checked reporting
Framing this as a clean handoff is corporate spin doing its job. When four researchers of that stature leave together to found something outside the company, and the CEO simultaneously steps back to a chairman role, that's not routine succession planning, that's an organization losing confidence in its own structure to move fast enough. Google throwing seed money at Discovery Loop looks less like generosity and more like hedging against the possibility that the next breakthrough happens outside its walls instead of inside them.
Read more about this at: Latent Space
