The News
Following a string of departures among its top AI leaders, Google DeepMind CEO Demis Hassabis dismissed the notion that Google was losing its grip on leading AI talent and said he remains confident in the company’s ability to attract and retain the best people.
Even as well-funded rivals and nimble startups poach its top brass, “we have by far the biggest and broadest research bench of any of the labs out there,” Hassabis said during an on-stage interview with Semafor at the Cannes Lions Festival. “We win our fair share of the top talent.”
Google’s shares tumbled as much as 7% on Monday following the exits of top AI thinkers Noam Shazeer and John Jumper, which sparked anxiety on Wall Street over Google’s ability to retain talent as the battle for AI minds escalates from a Silicon Valley sub-plot to a central focus for global investors.
“There’s a lot of talent movement between all the leading labs,” Hassabis noted, chalking it up to the most “ferociously competitive” job market the tech industry has ever seen.
A Google spokesperson said that the small number of departures won’t impact the company’s overall trajectory.
In this article:
Know More
In an industry where a single breakthrough algorithm or architectural tweak can unlock billions in revenue, individual researchers have attained the kind of leverage and celebrity traditionally reserved for star athletes.
Venture capital firms are routinely funding new startups based solely on the pedigree of researchers who leave deeply entrenched labs like DeepMind and Google Brain. Investors are highly sensitive to the optics of “brain drain,” viewing human capital as the ultimate leading indicator of future AI dominance.
Shazeer is widely revered as one of the key architects of the modern generative AI boom. He was a co-author of the landmark 2017 Google research paper, “Attention is All You Need,” which introduced the underlying foundation for virtually every major large language model today, including OpenAI’s ChatGPT, Anthropic’s Claude, and Google’s own Gemini.
For his part, Jumper was instrumental in the development of AlphaFold, the revolutionary AI system that successfully predicted the 3D structures of nearly all known proteins. His work transformed computational biology and drug discovery, earning him and Hassabis Nobel Prizes and cementing DeepMind’s reputation as a lab capable of solving grand scientific challenges.
While agile new companies can offer top AI thinkers massive equity upside, Hassabis’s confidence is rooted in Google’s structural advantages, including its unparalleled ecosystem of data, integrated hardware, and sheer computing power. For researchers looking to train the next generation of frontier models, access to Google’s vast fleets of custom Tensor Processing Units (TPUs) remains one of the most compelling recruiting tools in the industry.
The 2023 merger of Google Brain and DeepMind also centralized the company’s previously fragmented AI efforts, creating a unified research juggernaut under Hassabis’s leadership with a clear mandate to accelerate the development of the Gemini ecosystem and continue the kind of fundamental scientific discovery that AlphaFold pioneered.
Reed’s view
When top researchers leave AI labs, it’s always big news. But the departures are often bigger PR headaches than operational ones. The most important talent wars aren’t happening in the C-suites or even the VP corner offices. They’re happening at places like universities, where top PhD candidates and computer scientists are being courted by the highest ranks at tech companies.
The next breakthroughs in AI algorithms are more likely to come from ten brilliant kids fresh out of college than they are from one executive who made big discoveries years ago.
While there’s a prevailing narrative that Anthropic is leading the AI pack — and that OpenAI is close behind — that premise assumes text-based LLMs are the route to artificial general intelligence. Another way of looking at it is that the race is nowhere near over, and that the lead will continuously shift as time goes on. That’s definitely Google’s view.
That’s why Google is pushing forward on several fronts — not just text models or computer code. Among other things, it’s working on video generation, music creation, and biotechnology research. Its view, driven largely by Hassabis, is that the diversified approach is more likely to yield breakthroughs necessary to win the long-term race.
It’s also why a handful of departures across Google’s massive effort is not that big of a deal, at least not one worthy of a big market swing.