So are the robots going to kill us all?
In the first of a three-part series, Joe Shute asks the British ‘Godfather of AI’ about advances that will change our lives
Deep within the inner sanctum of Google’s downtown HQ in Toronto, past the rooftop crazy golf putting greens, foosball tables and ergonomic furniture sporting the bold primary colours of the company logo – stands a scruffy figure so incongruous, he might have been drawn by Quentin Blake.
In person, Professor Geoffrey Hinton bears all the hallmarks of the quintessential British academic: tousled hair; crumpled shirt with a barrage of biros in the top pocket and flanked by a vast, mucky whiteboard scrawled with impregnable equations. There are no chairs in his office. The 69-year-old prefers to stand.
Gleefully eccentric he may be, but to the bright young things outside his office, Hinton is akin to a deity: the so-called “Godfather of Artificial Intelligence (AI)” and the brilliant mind behind the technology that has sparked a global revolution.
In this, his first British newspaper interview, Professor Hinton admits to being bemused by the nickname that has accompanied his late career surge.
His students have been poached by Silicon Valley to lead AI research at the likes of Apple, Facebook and Google (which has also appointed him a vice-president engineering fellow). In the coming months, he will take the helm of Toronto’s new $180million (£140million) Vector Institute, which it is hoped will cement the city’s status as a world leader in AI.
“I feel slightly embarrassed by being called the godfather,” he says in a cut-glass English accent that has resisted all North American overtures.
What brought Geoffrey Hinton from years of relative academic obscurity to leading the cutting edge of AI is an unshakeable faith in his work. “I have a Reagan-like ability to believe in my own data,” he grins.
Hinton is a pioneer of something called machine learning, which enables computers to come up with programmes to solve problems themselves. In particular, he has devised a subset of machine learning called “deep learning”, whereby neural networks modelled on those that form the human brain enable machines to learn in the same way a toddler does.
This means computers can autonomously build layers of intelligence. Such systems have been supercharged in recent years by the advent of hugely powerful processing technology and are now becoming mainstream, infusing everything from speech recognition patterns in our smartphones to image detection software and Amazon telling you which book to buy next.
Through the work of Hinton and his colleagues – dubbed by their rivals the “Canadian Mafia” – the potential of machine learning has become limitless. The Brave New World of AI is upon us and already permanently changing our lives – for good and ill.
Hinton was born in Wimbledon in post-war Britain. His father, Howard, was an entomologist with a fondness for beetles. His mother, Margaret, a schoolteacher. A streak of brilliance runs through the family DNA. His uncle was the economist Colin Clark, who invented the term “gross
‘With artificial intelligence it is difficult to predict beyond five years’
national product”. His great-greatgrandfather was the logician George Boole, who invented Boolean algebra, a foundation of modern computing. The family moved to Bristol, where Hinton attended Clifton College. It was there that a school friend first introduced him to the wonders of AI by talking to him about holograms and how the brain stores memories.
After school, he was awarded a place at King’s College, Cambridge, to read physics and chemistry but dropped out after a month. “I was 18, it was the first time I had lived apart from home, it was awfully hard work, there weren’t any girls and I got depressed,” he says. The following
year he re-applied to read architecture but again dropped out – this time after just a day – and switched instead to physics and physiology. He changed again to philosophy, then fell out with his tutors, quit his studies and became a jobbing carpenter in Islington, north London. “I made shelves, hung doors, nothing fancy. The sort of stuff people get paid for.”
After a few years of toil, he returned to academia and in 1973 started a PHD in artificial intelligence at the University of Edinburgh. His tutors told him he was wasting his time on neural networks, but Hinton plugged on regardless.
He continued his research at the Carnegie Mellon University in Pittsburgh, but soon realised the Department of Defence was funding work on AI in his department and across the US. He quit in protest to move to Canada, where military funding was less pernicious.
“When I left, I took an American penny and blew it up with a Xerox machine and put it up on my office door,” he says. “But I changed the G to a D so it read: ‘In DOD we trust’.”
According to Hinton, rather than fearing the growing intelligence of machines, a far more pressing threat to humanity is the development of killer robots (underlined this week by a petition signed by the founders of 116 AI companies to the UN calling for a ban on lethal autonomous weapons).
Hinton has signed a similar petition himself and previously wrote to express his concerns to Britain’s Ministry of Defence. “The reply said there is no need to do anything about this now because the technology is a long way away, and anyway, it might be quite useful,” he says.
He also fears the use of AI in surveillance of the civilian population and reveals he once declined a job to sit on the board of the Canadian equivalent of the National Security Agency because he feared his research could be abused by security services.
Still, even while discussing the terror of weaponised “drone swarms”, Hinton remains evangelical about the benefits of AI – particularly in healthcare and education. He lost his
‘I have a Reagan-like ability to believe my own data’
first wife, Ros, to ovarian cancer in 1994, leaving him to look after their two young adopted children. His current wife, Jackie, has been diagnosed with pancreatic cancer.
Medicine, he believes, will become far more efficient as a result of AI. Soon he envisages anybody being able to pay $100 (£78) to have their genome mapped (the current cost is 10 times that, at $1,000). Unpopular as it makes him with radiologists, Hinton also believes X-ray detection could soon be largely robot work. Work will be lost, but he insists it is the job of governments and business to ensure that automation does not leave people behind.
“In a sensibly organised society, if you improve productivity there is room for everybody to benefit,” he says. “The problem is not the technology, but the way the benefits are shared out.”
Even the visionary admits he does not know where the AI revolution will take us. “It is very hard to predict beyond five years in this area and things always turn out differently to what you expect,” he says.
Suffice to say, the world as we know it is about to be turned on its head.