Founding and Vision of DeepMind

Key Insight

DeepMind originated from a chance encounter in 2010 between Mustafa Suleyman and Demis Hassabis at London's Victoria Casino after both were eliminated from a poker tournament. Having grown up in North London and known each other through Suleyman's friendship with Hassabis's younger brother, they commiserated over desserts before discussing robots and the future. Hassabis, holding a PhD in cognitive neuroscience, articulated the tantalizing proximity of machines that could truly learn, with Suleyman noting, 'surely a machine could learn to play poker, a machine could learn a set of heuristics and then produce those patterns.'

Not many months later, Suleyman, Hassabis, and Shane Legg formed DeepMind. Their audacious goal was to achieve Artificial General Intelligence (AGI), a superintelligence capable of everything a human brain could do, only better. This vision was met with widespread skepticism; Suleyman remarked it 'felt pretty far out at the time,' and Hassabis noted, 'Most people thought we were completely mad to be embarking on this journey.' Hassabis's philosophy, derived from neuroscience, proposed focusing on the algorithmic level of the brain, aiming to 'Step one, solve intelligence. Step two, use it to solve everything else.' Legg, a researcher known for openly discussing AGI, believed it was inevitable by 2028 with a 50-50 chance, fueled by exponential growth in compute power and data, despite peers struggling with basic tasks like image recognition.

The trio founded DeepMind in September 2010, explicitly stating their AGI goal in their pitch deck, which elicited 'an astonishing amount of eye-rolling at conferences.' Hassabis assumed the CEO role, Legg became chief scientist, and Suleyman took the chief product officer title, despite the startup being far from having a product. They raised 350000 pounds (approximately $540000) from British angel investors and set up in a cramped attic office in central London. Though none of the founders were strong computer programmers—Hassabis was a neuroscientist, and Legg a mathematician—Suleyman realized their 'very different skills' were complementary in their unique dynamic, allowing them to collectively 'wing it' in their quest to solve for intelligence.