This should be a moment of triumph for University of Toronto computer science professor Geoffrey Hinton.
The British-born 69-year-old is known as the godfather of the strain of artificial intelligence called neural networks or “neural nets,” which involves setting up computer systems to mimic the human brain, allowing them to learn. It is, some experts say, going to radically transform our lives – already is, actually – the way electricity did in the 20th century.
For years, Prof. Hinton worked not just in relative obscurity but on the losing side of a decades-long battle within the academic cloister of computer science. His neural nets were considered “weak-minded nonsense” by better-funded adherents of more conventional methods of creating artificial intelligence, which involve more hands-on programming. Academic journals used to reject papers on neural nets out of hand, Prof. Hinton says.
But in just the past five years or so, after a series of stunning breakthroughs by his graduate students, neural nets are all the rage and he is being hailed as a guru of a new era of computing. Neural nets are already powering most of the voice recognition software in your cellphone. They can recognize pictures, down to different breeds of dog, almost as accurately as humans can.
And the world’s biggest tech companies are now throwing millions of dollars into neural net research, hiring many of Prof. Hinton’s former students, who now run or conduct AI research at Apple, Twitter, Google and Facebook.
He says the possibilities for AI range from driverless cars to smartphones that can diagnose skin cancer better than any dermatologist. The new wave of technology is expected to disrupt industries and make those who develop it and control it a lot of money.
Prof. Hinton splits his time between U of T and Google’s Toronto office, where he is an engineering fellow and will help direct a new AI lab. He has just been named the chief scientific adviser of the newly announced Vector Institute, which will fund research into artificial intelligence and is aimed at turning Toronto into a global AI hub.
Despite all these recent victories, Prof. Hinton sounds as if he is still waging a rearguard action. In an interview in his narrow, spartan office at Google – there are no chairs, and a wall-sized whiteboard behind him is covered in equations – he wastes almost no time on small talk before launching into a blow-by-blow account of the battle that’s still raging between advocates of neural nets and those who back more traditional forms of AI.
His own university, he says, dragged its feet on hiring a new neural-net professor, despite receiving $1-million from Google to do just that. Now, many people are jumping on the neural-net bandwagon, he says, as research funding flows more freely.
“Now that neural nets work, industry and government have started calling neural nets AI. And the people in AI who spent all their life mocking neural nets and saying they’d never do anything are now happy to call them AI and try and get some of the money,” Prof. Hinton said.
(It soon became apparent that Prof. Hinton enjoys tweaking noses. During the interview, he made a couple of snarky asides about U.S. President Donald Trump before halfheartedly apologizing to his Google PR handler with a grin.)
The traditional concept of AI relies on logic and rules to program computers to think. In the 1960s, when much of this work was more theoretical and not yet available in the palm of your hand, the neural-net alternative was “destroyed” and discredited, Prof. Hinton says. The traditional model was accepted as an article of faith.
But breakthroughs in the past few years, made possible partly by dramatic increases in computing power, have changed all that. In 2009, two of Prof. Hinton’s graduate students won a speech-recognition competition, besting more established methods by using a neural net that was then upgraded and incorporated into Google’s Android phones. In 2012, two of his other students handily won an image-recognition competition. That technology, which involves training a system by using a database of one million images, can recognize and describe an image with a 5-per-cent error rate – about the same as humans.
To explain just how neural nets work, Prof. Hinton uses the example of a translation program. Using a neural net as a translator involves feeding a computer network a mountain of words and word fragments, he explains. The system figures out the meaning of a sentence, then feeds that into another neural net to spit out the sentence in another language, without the use of programming or linguistic rules. It even learns the difference between the active voice and the passive – by itself.
“Nobody ever told it this thing about actives and passives. Just like your little kid, you don’t say: ‘Look, Johnny, there’s actives and there’s passives.’ … No, after a while, they just get it,” Prof. Hinton said. “And the point about these neural nets is they just get it.”
He credits two factors for his decision to come to U of T in 1987, after bouncing between a handful of universities in the United States. One was funding for his brand of AI from the Canadian Institute for Advanced Research. The other was more political: “I didn’t want to take money from the U.S. military. And most of the AI funding in the States came from the military.”
Born in Wimbledon and raised in Bristol, England, Prof. Hinton’s mother was a math teacher and his father was an entomologist with a fondness for beetles. His great-great-grandfather was 19th-century logician George Boole, the inventor of Boolean algebra, a foundation of modern computing. Prof. Hinton attended what he described as a second-tier private school (called a public school in Britain): “I wasn’t particularly good at math at school. I liked physics. And soccer.”
He went to the University of Cambridge for physics and chemistry but only lasted a month, dropping out and switching to architecture, where he said he only lasted a day. He re-enrolled in physics and physiology but found the math in physics too tough and so switched to philosophy, cramming two years into one.
“That was a very useful year, because I developed very strong antibodies against philosophy,” Prof. Hinton said. “I wanted to understand how the mind worked.”
To that end, he switched to psychology, only to decide “that psychologists didn’t have a clue.” He spent a year as a carpenter before heading to graduate school at the University of Edinburgh in 1973 to study artificial intelligence under Christopher Longuet-Higgins, whose students included Nobel Prize winners John Polanyi, the U of T chemist, and theoretical physicist Peter Higgs.
Even then Prof. Hinton was convinced that the discredited neural-net concept was the way forward. But his supervisor had recently converted to the traditional AI camp.
“I had a stormy graduate career, where every week we would have a shouting match,” Prof. Hinton said. “I kept doing deals where I would say, ‘Okay let me do neural nets for another six months and I will prove to you they work.’ At the end of the six months, I would say, ‘Yeah, but I am almost there, give me another six months.’ And since then I have been saying, ‘Give me another five years,’ and people have been saying, ‘You have been doing it these five years, this never worked.’ And finally, it worked.”
He denies ever doubting that neural nets would one day be proven superior: “I never had doubts, because the brain’s got to work somehow. The brain sure as hell doesn’t work by somebody programming in rules.”
Confronted with the typical “Are robots going to take over the world?” question, he agrees that limits must be placed on AI. He recently signed a petition asking the United Nations to ban artificially intelligent lethal weapons, a cause championed by the aptly named Campaign to Stop Killer Robots: “I think that’s the scariest bit. And that’s not the distant future … That’s now.”
The more benign future for AI, he predicts, will see neural nets used in doctor’s offices to diagnose diseases or skin cancers. They will also be refined into personal assistants that not only remind you of a lunch appointment but use “common sense” to observe your behaviour and decide to interrupt you if you have forgotten your appointment.
Big banks, cable companies and many others are looking to use AI to analyze things such as sales data and to better interact with their customers, says Steve Irvine, who left Facebook to return to Toronto and launch a startup called Integrate.ai to help firms do just that.
“I don’t think he can get enough praise,” Mr. Irvine said of Prof. Hinton. “Because he’s been in AI in the dark days, when he looked like a mad scientist and people never thought this was going to happen. … Now all these things that were talked about for 20, 30 years are happening. I think it’s a nice reward for him … and now it is this worldwide hysteria, and he’s the godfather. It was definitely not an overnight success.”Report Typo/Error