Google Chief Scientist Jeff Dean and AI Veterans Depart to Launch Discovery Loop
In one of the most significant talent defections in the history of Silicon Valley, Google Chief Scientist Jeff Dean has departed the company after nearly 27 years of service to co-found Discovery Loop. Dean, who joined Google as its 30th employee in 1999 and served as a central architect of its infrastructure and AI systems (co-founding Google Brain and leading TPU development), is launching the new venture alongside a legendary team of Google and DeepMind veterans:
- Sanjay Ghemawat: Longtime systems architect and Dean's primary collaborator on MapReduce, Bigtable, and other foundational Google systems.
- Quoc Le: Pioneer in deep learning and neural architecture search.
- Oriol Vinyals: Former VP of Research at Google DeepMind and lead of the Gemini and AlphaStar projects.
The Mission: Automating the Scientific Method
Discovery Loop is structured as an independent public benefit corporation (PBC). Its ambitious mission is to build software and model stacks that fully automate the experimental scientific loop—including hypothesizing, testing, and evaluating results1—across scientific and industrial domains.
In Dean's own words, running thousands of automated loops in parallel could unlock groundbreaking advances in multiple fields:
"Dean said that running thousands of those automated loops could unlock groundbreaking advances in multiple domains like science, biology, chip design, and even making improved AI models."
The startup plans to focus initially on automated machine-learning research (creating models that can design and train other models) before expanding into hardware engineering, clean energy, and automated drug discovery.
A $10 Billion Valuation on Day One
Despite just launching, Discovery Loop has already entered advanced discussions to raise a massive $1 billion funding round at a valuation near $10 billion.
Legendary venture firm Khosla Ventures is reportedly leading the backing for the round, while Google itself has backed an initial seed round to maintain a relationship with its departing stars. The exit represents a severe blow to Google DeepMind's talent pool at a time when the company is executing a massive leadership reorganization under Koray Kavukcuoglu.
-
An instance of AI laboratories are transitioning from human-led engineering to autonomous, model-driven training loops. — A new startup co-founded by AI pioneers aims to automate the scientific and machine learning research cycle, replacing human-led engineering loops. ↩︎