Why Google is rearranging its top AI experts

Demis Hassabis, who previously ran the entire AI division, is transitioning into a role as Alphabet’s chief scientist and chair of Google DeepMind. He will also maintain his leadership of Isomorphic Labs, a separate project focused on using computers to discover new drugs. Filling his shoes as head of the main AI group is Koray Kavukcuoglu, who previously served as the group’s chief technology officer. Meanwhile, two legendary engineers at the company, Jeff Dean and Sanjay Ghemawat, are leaving their full-time roles to launch a new company called Discovery Loop, which is backed by an investment from Google. They intend to build automated systems that accelerate scientific and engineering research.

To understand this shift, think of how a company evolves. In the early days, the goal was building a model—a digital engine that learns to predict and create by studying massive amounts of examples. Think of this model like a blank-slate apprentice. Through a process called training, the apprentice reads billions of pages of data, learning patterns until it can answer questions or write code. Once the apprentice is smart enough, you need to decide what job they should do. Google has spent years building these smart, versatile apprentices. Now, the company is moving from the phase of simply training the apprentice to assigning them specialized, high-stakes roles. By moving leaders into specific niches like drug development, the company is treating AI less like a generic tool that can do anything and more like a scientist tasked with solving real-world biological puzzles.

When top architects leave or change roles, it is rarely just about personality; it is a signal of where the industry thinks the value actually lives. For years, the race was about who could build the smartest, most versatile apprentice. Now, the goal is proving that this technology can be useful beyond writing emails or creating images. By splitting their focus between internal research and new, external spin-off companies, Google is betting that the next decade of AI will be defined by results in health and science, rather than just technical benchmarks. For the rest of us, it marks a transition where we will likely see fewer general AI updates and more breakthroughs in how we approach medicine and industrial engineering.