Demis Hassabis: "A real one-off".
Tony Corfe still remembers the time he first saw Demis Hassabis play chess. He was in charge of the primary schools team in Barnet, north London, and looking for new recruits when one week a slight six-year-old boy turned up.
"He was very small," recalls Corfe, "so we had to sit him on a telephone book and a couple of chairs just to get his head up to table height so he could see the board."
The DeepMind offices on Fenchurch Street, London. Photo: Bloomberg
Once Hassabis was settled at the table, however, he needed little else. "He was sparkling," Corfe says. "He was determined, and he definitely wanted to win. Of all the schools that I had contact with, he was the best player. He was top of the infants."
The boy quickly developed into a chess prodigy, winning dozens of tournaments before representing England in competitions. By the age of 13, he had reached the standard of chess master, and he went on to win the World Games Championship a record five times.
These memories acquired new significance this week when Hassabis sold his company to Google. The firm, DeepMind, specialises in Artificial Intelligence, and is apparently attempting to develop computers that can think spontaneously like humans, rather than having to be pre-programmed.
Multinational firms are in a race to be first to produce this new era of sophisticated technology, explaining why Google – which has already established its own robotics division – reportedly paid $757 million. At a stroke, the deal made Hassabis one of Britain's most successful technology entrepreneurs.
What is it, then, in this multimillion-dollar mind that Google values so highly? Friends describe Hassabis, now 37 and married with two children, as shy but determined. "He is not a showman, and he keeps his head down," says Professor Geraint Rees, director of the Institute of Cognitive Neuroscience at University College London, where Hassabis studied for a PhD. "Nerdy is the wrong word, but he is definitely of a technical bent. His determination and drive are quite striking."
A glance at his CV will vouch for that. He left school at 16, having taken his A-levels. He spent his gap year gaining experience in computer games programming, which was to become his first career, by co-writing Theme Park, one of the most successful games of the nineties.
After graduating with a double first in computer science from Cambridge, he set up his own games business, Elixir Studios, which was responsible for such hits as Evil Genius and Republic, the latter being nominated for a BAFTA award.
"He is extraordinary, a real one-off," says Joe McDonagh, who co-founded Elixir with Hassabis. "He was only 21 when we set it up [McDonagh was 25], but he had an incredibly old head on such young shoulders. When I saw the news this week, I wasn't remotely surprised. I've always expected he would do something like this."
Although he credits Hassabis' intelligence, McDonagh says his more important ability was to inspire a team. "He truly believes the thing you are working on will change the world and will be remembered for ever. That is incredibly inspiring. Most of us are held back by fear, but Demis takes the attitude that everything is possible."
He was not reclusive at Elixir, and took an active role in organising social events. "We'd be working on all this brainiac stuff during the day," says McDonagh. "But at night, Demis would lead the five-a-side team in a [football] league in Tottenham where we'd get kicked around the park. He'd be just as happy to argue with you about Liverpool versus Spurs as he was about Artificial Intelligence."
But he could also be competitive and would challenge colleagues to bout after bout of video game competitions. McDonagh recalls one occasion when Hassabis was transfixed by a colleague's ability to win repeatedly in the science fiction game Starcraft.
"Demis wanted to beat this guy," he explains. "He would lock himself in a room with the guy night after night. He'd handicap him, by getting the guy to play without a mouse or one-handed so he could analyse exactly what he was doing to be brilliant. It was a bit like going into the boxing ring and getting beaten up, and then returning every night. It showed his incredible will to win."
This competitive edge also drove Hassabis to the Mind Sports Olympiad, to prove his worth alongside other "mental athletes". He spent a week at the competition in 1997, and has returned every year since. He was champion of the elite "Pentamind" contest – where competitors challenge each other in five different disciplines, including chess – in five of the first seven years.
Despite his business interests, he still makes time for the Olympiad each August. "It's for the competition," explains Corfe, who is one of the organisers. "He has had his name on the trophy more times than anybody else."
These days, however, it is Hassabis' seven-year-old son, Alexander, who is touted as a prodigy. After he accompanied his father to last year's championships – and collected the junior gold prize – the competition's website speculated "whether we might be seeing a dynasty in the making".
Hassabis' path has not been completely smooth. According to friends, he grew frustrated with the more mundane aspects of managing Elixir. "The direction of the business became too much dictated by accountants and people other than himself," says David Levy, who founded the Mind Sports Olympiad.
Instead, Hassabis returned to academia, completing his PhD in cognitive neuroscience at University College London in 2009. In three years, he wrote a dozen research papers and succeeded in systematically linking memory with imagination for the first time. The journal Science trumpeted his achievement as one of the breakthroughs of the year. His former computer programming colleagues were more succinct, telling him he had "a brain larger than a planet".
He was soon awarded a prestigious Henry Wellcome postdoctoral fellowship. "Those kind of fellowships are given to people not on the basis of their entrepreneurial ability or because they are chess prodigies, but because of their scientific achievement," says Professor Rees. Even so, he adds, "Demis always had an eye on the bigger picture."
For the moment, however, that picture remains blurred. Despite the glut of publicity surrounding the sale of DeepMind, nobody seems sure exactly what it does – and those who know won't tell. Hassabis politely declined to comment for this article, saying he has been "instructed by Google not to respond to interview requests at this time".
Frank Meehan, who helped secure much of the initial investment, is equally circumspect. "[DeepMind's function] is the one thing I can't tell you. I'm completely locked down," he says. He does, however, explain that computer scientists such as Hassabis are trying "to get to the point where machines can learn rather than being taught".
"If they can learn from their surroundings and their actions, that's a massive step forward because you don't have to spend huge amounts of time programming everything."
This may seem a daunting mission, but it is unlikely to faze Hassabis. "I'm actually more worried about not taking risks and playing safe [than taking chances]," he once said. "I've always been prepared to jump in at the deep end and see if I can swim or not."