Who Owns AI When the Chief Scientist Isn't the One Running It
On August 5, Google split the top job in AI research into two titles. Demis Hassabis moved from CEO of Google DeepMind to chair of Google DeepMind and chief scientist of Alphabet. Koray Kavukcuoglu, DeepMind's chief technology officer, took over day-to-day operations with the title of senior vice president, reporting directly to Sundar Pichai. Hassabis had held the CEO title. Some DeepMind staff called it a demotion with a better business card.
Hassabis kept the bigger stage. Chief scientist of Alphabet sounds like the top of the pyramid, the name attached to whatever research defines the company for the next decade. Kavukcuoglu got the smaller title and the actual machine: the budget and the roadmap, plus a reporting line that now runs to him instead of to Hassabis. That split is worth sitting with. Prestige moved one way. Authority moved the other.
This is what happens when a research organization outgrows the point where one person can hold both jobs credibly. Early DeepMind ran on the assumption that the best scientist and the best operator were the same person, and at a few hundred people that assumption mostly held. At the scale Google DeepMind runs now, with frontier models shipping across a company juggling dozens of competing product priorities, the operating job requires something else: someone who can decide fast and own the roadmap, the same thing any large engineering org needs.
Something real gets lost in that split. A chief scientist with no budget and no roadmap authority can shape where research points, but shaping and shipping are different verbs now. If Hassabis wants a project prioritized, he has to persuade Kavukcuoglu and Pichai. Deciding alone stopped being an option the day the org chart changed. Research authority becomes advisory, and that is a different job than the one Hassabis held for a decade, whatever the new title implies.
Jeff Dean and Sanjay Ghemawat leaving Discovery Loop within days of the restructuring looks like part of the same pattern. Both built their standing over twenty years on being the people whose research judgment set direction inside Google. A structure that pays that judgment in advisory currency instead of operational currency is a worse deal for exactly the people who held the most operational power under the old one.
There are two ways to read the departures. One is that talent follows wherever the org chart authority actually sits, and senior technical leaders with no reporting line into the research chair have less reason to stay attached to it. The other reading is closer to a values mismatch: people who spent careers earning influence through research leadership have little appetite for a structure that just told them, formally, that research leadership doesn't run anything anymore. Both readings land on the same fact: the title carried power once, now it carries credibility, and the people who understood that difference best are the ones leaving first.
What lingers is whether any AI lab can keep its most credible research voice and its most decisive operator inside the same reporting line once the company outgrows one person doing both jobs well. Google just answered that question, and called the answer a promotion.