Google uses AI to understand search queries that have been written as spoken
A British technology company, set up two years ago by a former child chess prodigy who became a groundbreaking neuroscientist, has become Google's largest European acquisition.
The search giant is spending $625-million on DeepMind Technologies, a London-based firm which recently developed a computer system capable of understanding and playing an Atari computer game simply by looking at it on a screen as a human would.
The artificial intelligence (AI) firm was created by Demis Hassabis (37). Described as "very brilliant" by his peers, he was a chess master by 13, completed his A-levels two years early and at 17 was lead programmer on the classic game Theme Park at the videogames company Bullfrog.
In 1999, aged 23, he won the Mind Sports Olympiad – an annual international multi-disciplined competition for games of mental skill. He won it a record five times before retiring in 2003 from competitive play.
Born in north London, Hassabis also carried out research on brain-damaged patients which established that being able to imagine experiences is key to being able to remember past events.
DeepMind reportedly competed with Google and other AI companies for talent, and Google's chief executive, Larry Page, is said to have led the deal himself after an earlier approach from Facebook was turned down.
Deep learning
Sources close to the purchase indicated that the technology would be built into Google's search systems, rather than becoming part of its fast-expanding robotics division. Google has bought eight robotics companies, including Bot & Dolly which made the computer-controlled cameras used in the film Gravity.
"DeepMind was generally interested in reinforcement learning, and in deep learning, which is very useful in mining so called 'big data', something Google has a lot of and is interested in processing," said Murray Shanahan, a professor of cognitive robotics at Imperial College in London.
Google uses AI to understand search queries that have been written as spoken, as well as pattern recognition for image search. Its translation service also relies heavily on AI to understand the context of words and their meaning in different situations and sentences.
The broader question of whether AI technology could be misused or pose a threat to humans has led to the creation of the Cambridge Centre for the Study of Existential Risk, which notes that "many scientists are concerned that developments in human technology may soon pose new, extinction-level risks to our species as a whole".
The DeepMind acquisition is conditional on Google setting up an internal ethics board, sources told the Information.
Google confirmed the deal but would not supply any details.
Google's other recent acquisitions have included the $3.2bn purchase of the smart fire alarm company Nest, described by Google's Eric Schmidt as "an important bet" which will lead to products that are "infinitely more intelligent".
Hassabis
Hassabis got a double first in computer science at Cambridge University in 1997, and returned to games as lead AI programmer on the landmark game Black & White. He set up his own games business, Elixir Studios, in 1998, but in 2005 he left for academia, working on cognitive neuroscience and artificial intelligence, and publishing influential papers on memory and amnesia.
After attaining a doctorate in cognitive neuroscience from University College London in 2009, Hassabis returned to business in 2012 to found DeepMind Technologies, alongside Shane Legg and Mustafa Suleyman.
Google acquired a string of robotics firms in 2013, culminating in the purchase of Boston Dynamics in December, the most high-profile purchase at the time and a company holding contracts with the US military.
Google's robotics division was put under the leadership of the father of Android, Andy Rubin, in December, combining seven technology companies to foster a self-described "moonshot" robotics vision.
Google's executive chairperson, Eric Schmidt, told the Guardian last week that the company was aiming to be the world's best personal assistant, and said the only limitation was the capability of the technology itself.
"We haven't held back because people aren't ready – we have held back because the technology doesn't work yet," he said. "It's very hard to do. But we want it to be the best it can be – with opt in, full permission – to help me get through the day, figure out my questions and suggest questions I should ask people." He added: "People are doing research into how computers discover knowledge instead of reporting what they figured out – many people are on the edge of that, so it's maybe five years away." – guardian.co.uk © Guardian News and Media 2014