The recent exodus of top AI researchers from Google's DeepMind to rivals Anthropic and OpenAI has sparked intrigue and raised questions about the underlying factors. In this article, we'll delve into the 'London Rule' and explore the intriguing dynamics at play.
The London Connection
What's particularly fascinating about this story is the geographical thread that ties it all together. It's not just about talent leaving; it's about a specific city, London, and the unique legal environment it presents. DeepMind's research leadership is based in London, and the UK's enforcement of non-compete agreements creates a distinct advantage for companies like Anthropic and OpenAI.
The legal gap between the UK and California, where such clauses are largely void, means researchers in London are bound by longer non-compete periods. This detail becomes crucial when examining the timing of these exits. For instance, Nobel laureate John Jumper, who co-led AlphaFold, will join Anthropic next year, suggesting a contractual cooling-off period.
The Migration Theory
One theory that emerges is the migration of important AI work from London to California, specifically Mountain View. Researcher Lucas Beyer noted that recent departures are mostly long-time London-based researchers, indicating a potential shift in the center of gravity for pre-training, a foundational phase in AI model development. If the most critical work is moving to California, it provides a clear incentive for London-based researchers to follow, especially if non-compete agreements allow it.
This theory gains traction when considering Google's internal decisions. Shortly before Noam Shazeer's departure, computing power was reallocated to a London-based DeepMind team for pre-training collaboration. Compute is the lifeblood of AI research, and its movement often signals where the action is.
The Timing Factor
The timing of these exits is intriguing, as both Anthropic and OpenAI are gearing up for public listings. A pre-IPO offer carries the allure of equity, which could translate into significant paydays once the stock starts trading. For senior researchers, the choice between a stable Google role and a stake in a soon-to-be-public company is a tough one. Especially when considering Anthropic's recent valuation near $965 billion, overtaking OpenAI, and its potential public listing this fall.
Implications and Reflections
The implications of this talent migration are significant. Alphabet's stock drop after Jumper's exit highlights the market's perception of Google's talent war performance. DeepMind CEO Demis Hassabis acknowledges the fierce competition, but the recent departures suggest it's not just about intensity; it's about geography. The map, for now, seems to be pointing away from London.
In my opinion, this story underscores the complex interplay of talent, geography, and legal environments in the AI industry. It raises questions about the future of AI research hubs and the strategies companies employ to attract and retain top talent. As the AI landscape evolves, keeping an eye on these dynamics will be crucial for understanding the industry's direction.
