Canada’s big AI threat isn’t robots — it’s people
There’s a certain irony to all the talk about how artificial intelligence is going to take our jobs. Particularly if what keeps you up at night is your ability to hire enough humans with AI skills.
Jordan Jacobs is the co-founder of Layer 6 AI, a Toronto company that deploys its AI prediction and personalization engine for the financial services industry. Jacobs has found financial institutions are an ideal customer for the machine learning (ML) technology Layer 6 is developing. “They have more [customer] data than almost anybody,” explains Jacobs “and they can apply AI to that data to dramatically improve customer experience”.
Read: Artificial Intelligence, Machine Learning, Deep Learning: What’s the Difference?
“Banks ask themselves: ‘so how do I break down the data silos between the wealth management unit, the mortgage unit, the car loan unit and the retail bank so I get one picture of my customer instead of having the same customer overlapping four different units without understanding that that’s the same person?’”
Despite being located in the heart of Toronto and having deep connections to University of Toronto, which has an impressive track record for turning out AI talent, Jacobs quickly realized that to grow Layer 6 into a global company, the competition for hiring talent was going to be a huge challenge.
“Tomi (Layer 6 cofounder Tomi Poutanen) and I made a list of a bunch of companies we knew,” says Jacobs “and these were just companies in the Toronto-Waterloo corridor. We started asking ‘How many PhDs and Masters in machine learning do you hope to hire in the next five years?’ I think we only got halfway through the list. We stopped counting when we got to 5,000 (new hires), because that was already ten times more than the whole country was going to produce.”
Jacobs and Poutanen knew this was a problem they weren’t going to be able to solve themselves. So they pitched an idea to University of Toronto professors Richard Zemel and Geoffrey Hinton, and the idea for the Vector Institute was born.
U of T was a natural home base for the Vector Institute. Not only is Toronto home to dozens of established AI companies, but the university is also the longtime home to professor Hinton, widely recognized as ‘the Godfather’ of AI research. Hinton’s work at U of T developing artificial neural networks that mimic the learning processes of the human brain goes back more than 25 years, well before the computing power and vast quantities of training data were available to prove Hinton’s theories actually worked.
“Canada invents a lot of this technology,” says Jacobs “but to date, a lot of the value has accrued to US tech companies, partly because the big consumer tech companies had massive [amounts of] data and they recognized the value of AI first. But we felt that if you want to build an AI ecosystem locally, you need to build around the absolute best-in-class that we have here in the research world, then plug companies into that to make them more proficient in AI, create an AI startup ecosystem, and then build in parallel all the different pieces, so that 30 years from now you look back and say ‘oh that’s how we became one of the two or three leaders in the world in AI’ ”.
“So we see it as a training ground, and as a magnet for talent — both for keeping talent here, and for getting talent to come to Toronto, and as an engine for building an economic ecosystem.”
Read: For Google, the AI Talent Race Leads Straight to Canada
Vector’s long game of creating a world-leading AI ecosystem in Toronto is inspired in part by the opportunities already lost. On the list of AI superstars that did their post-doctoral research under Geoffrey Hinton at U of T are: Ruslan Salakhutdinov, currently the director of AI research at Apple; Ilya Sutskever, director of the Elon Musk-backed Open AI institute; Yann LeCun is the director of Facebook AI Research; and Zoubin Ghahramani, Chief Scientist at Uber’s A.I. Lab. The fact that all of these people now work south of the border speaks directly to the same issues that drove Jacobs and his co-founders to create the Vector Institute.
Read: Why one of Facebook’s top executives left Silicon Valley for an AI startup in Toronto
The most recent high-profile example is Raquel Urtasun, an associate professor at U of T and a world-leader in the field of computer vision and deep learning. Urtasun did her undergraduate degree in Pamplona, Spain, her Ph.D. in Lausanne, Switzerland and her postdoc at MIT and UC Berkeley before landing at U of T. Urtasun was aggressively pursued by many companies in the autonomous vehicle space — companies that offered her astronomical salaries and impressive research facilities. Fortunately for Canada, Urtasun’s strong desire to remain in her adopted city with her U of T team won out. Ultimately, Uber offered to set up a third Advanced Technology Group research lab in Toronto with Urtasun in charge, to join their existing centres in Pittsburgh and San Francisco.
“It’s a critical factor that people tend not to consider,” Jacobs suggests. “But to give you an example, when Geoff Hinton ends up starting a company with Google with two of his grad students, and one of them (Alex Krizhevsky) is now a very senior person at Google Brain and the other (Sutskever) is now running Open AI. So imagine those two guys were back here in this ecosystem. I mean they’re just invaluable people to keep.
“And so the case of Raquel — she, frankly, was one of the reasons that Tomi and I felt an urgency to create Vector. We knew she was being pursued by almost every self-driving car company that existed, and they all wanted her to leave.”
“And we knew it wouldn’t just be her leaving. She has 20+ students — including the ones that she shares with others. And if those people stay, you could literally build the core of a self-driving car industry, or an aspect of of it — in this case computer vision — and keep growing in Toronto. Some of those people will stay with her, and some of them will eventually go start their own startups. Some of them will join other companies. But that’s the core around which you build. If they left, they’re gone. And it’s almost impossible to get them back. People don’t realize they wouldn’t just be losing her, they would be losing the possibility of one of the most important industries of the future.”
When the Vector Institute launched in March of this year Raquel Urtasun was named as a co-founder and faculty member.
The Institute will be funded with approximately $100M over five years from both federal and provincial governments, and an additional $100M over ten years from the private sector (as this article was being written the province announced an additional $30M).
Vector’s near-term goal is much more tangible, and directly addresses the skills gap identified by Jacobs’ informal survey. Graduating more Masters and Ph.Ds in AI specialties like machine learning and computer vision is critical, but Vector aims to provide broad support for education across a much wider spectrum.
“The key thing is just training many, many, many more people,”says Jacobs “both by expanding the size of the graduate school and also undergraduate AI streams, and then creating programs for high school or younger levels. So really, the first goal is just trying to massively scale-up the proficiency at every level.”
Jacobs adds that Vector plans to directly support the private sector as well by “creating training programs that companies can take back and train-up people internally who are software engineers to become more proficient in ML.
“We need to provide tools so that companies with software engineers who went to school before ML was relevant — or just aren’t familiar with it — have the ability to get trained up. Maybe they’re not doing the core research that is the purview of a Ph.D. or maybe a master’s degree, but they are proficient enough to be using the tools. So a goal of the Vector Institute to create some of those programs a company can take back and use to train their software engineers. Google does that already. And lots of other companies should be doing that too.”
—
Part of TWG’s #ConnectTheBots series: An exploration of the human side of AI. To find out more about how TWG helps companies leverage practical AI solutions, please contact rswitzer@twg.io or follow him on twitter @rswitzer
