Demis Hassabis: Why Google’s AI CEO said no to Mark Zuckerberg and yes to Larry Page despite Facebook offering more money |


Why Google's AI CEO Demis Hassabis said no to Mark Zuckerberg and yes to Larry Page despite Facebook offering more money
Before OpenAI or Anthropic existed, Google and Facebook fought over a small London AI startup called DeepMind. Its CEO Demis Hassabis turned down Mark Zuckerberg’s richer offer and sold to Larry Page’s Google for $650 million in 2014. A new book by Sebastian Mallaby reveals how a Palo Alto dinner, a poker-style bluff, and AI safety concerns decided everything.

Google DeepMind CEO Demis Hassabis turned down a richer offer from Mark Zuckerberg’s Facebook to sell his AI startup to Larry Page‘s Google back in 2013—and a dinner at Zuckerberg’s Palo Alto home was the turning point. According to a new Wall Street Journal excerpt from Sebastian Mallaby’s upcoming book The Infinity Machine, Hassabis deliberately tested Zuckerberg during that dinner. He steered the conversation away from AI and into virtual reality, augmented reality, and 3D printing. Zuckerberg sounded equally excited about all of them. For Hassabis, that was the tell. “Facebook offered more money, but I wanted somebody who really understood why AI would be bigger than all these other things,” Hassabis said, per the WSJ excerpt.Page, on the other hand, had made his pitch months earlier at Elon Musk‘s birthday party. He argued that Hassabis could spend the best part of his career trying to build a company—or he could use Google’s existing infrastructure to go straight at artificial general intelligence. Hassabis found the logic hard to argue with.

Hassabis used Facebook as leverage—and Zuckerberg knew it

The Facebook courtship, it turns out, was largely a negotiating tactic. Hassabis and co-founder Mustafa Suleyman used Zuckerberg’s interest to pressure Google into committing faster. Suleyman, a poker player by instinct, talked up DeepMind’s billionaire backers—Elon Musk, Peter Thiel, Solina Chau—even though those investors didn’t exactly have their backs in any binding sense.Zuckerberg later acknowledged the play. In an interview with South Park Commons, the Meta CEO said Hassabis did a “very good job” of playing Facebook off Google—and that he respected the move.

Facebook’s dismissal of AI safety concerns sealed the deal

Beyond vision, there was another dealbreaker. When Suleyman raised the need for an independent AI safety oversight board, Facebook brushed it off. Google engaged seriously—its then-CFO Patrick Pichette compared AI to atomic energy, capable of both catastrophic harm and transformative good.In January 2014, Google acquired DeepMind for $650 million—a figure that looks almost absurdly cheap today. And Zuckerberg, spurned, went and hired deep learning pioneer Yann LeCun to build Facebook’s AI research lab from scratch instead.



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *