Edmonton could lead the way in AI: expert
U of A professor hails Edmonton’s investment in machine learning
Edmonton could become a major centre for developing ways to use the artificial intelligence (AI) that will probably drive the world’s next industrial revolution, a top scientists in the field says.
“AI could be well thought of as the leading edge in the second industrial revolution; Edmonton is a leader in the science,” University of Alberta computer science professor Richard Sutton said Tuesday.
“Why can’t we be a player in AI applications?”
Sutton, who moved to the U of A in 2003 from AT&T Labs in New Jersey, is considered a pioneer in AI and reinforcement learning, which involves letting computers learn from experience and interacting with their environment.
He wrote a major textbook on the subject and was picked last year to be part of Google DeepMind in Edmonton, the company’s first research lab outside its London headquarters.
The city is considered a top location for machine learning research partly because the province and the federal government started investing tens of millions of dollars in the area years ago, and the bet paid off, Sutton told the AccelerateAB technology conference.
“Edmonton invested in machine learning, and a decade-and-a-half later, it was important.”
Advances in the last seven years include a U of A program that beat professional players in no-limit poker and DeepMind’s AlphaZero found how to defeat other top programs in games such as Go and chess for which it was given only rules, not strategies, he said.
One big reason for these improvements is the continuing drop in the cost of computing power, which Sutton said in 10 years will be 100 times less expensive than it is today.
“As it gets cheaper, we will just use much more of it and it will play a much bigger role in all our endeavours. This is the driver of everything,” he said.
“It’s like a slow-motion explosion … We’re in the midst of an explosion, a doubling of computer power every 18 months.”
AI has applications for a vast array of technology, from mobile cheque deposit to social media to online shopping.
The big development down the road will be systems that learn increasingly sophisticated skills by themselves, as they come to understand the world without having all the rules laid out for them, he said.
This might initially be valuable in such areas as speech recognition software for writing documents, or operating driverless vehicles, he said.
While Sutton doesn’t see why people can’t make artificial intelligence that’s greater than what humans possess, he thinks as this happens, our intelligence will probably also grow.
“It will be disruptive, it will require us to re-examine things … but AI will bring more diversity.”