June 24, 2020

Alberta built an artificial intelligence powerhouse. What’s next?

Illustration by Mike Kendrick

Story by Scott Lilwall

Just off Highway 60, a short jaunt south of Devon, sits one of the most celebrated pieces of industrial equipment in the country.

The collection of steel lattices has a name: Leduc #1, the derrick that struck oil south of Edmonton more than 70 years ago. On a cold February evening in 1947, the tower ushered a new industry into Alberta with a belch of flame and a column of thick smoke and transformed the province’s identity. It’s been dismantled, restored and dismantled again before finally ending up at its current home in a museum for the energy industry.

Half a century later, in the early 2000s, Alberta was changing again. This time, there wasn’t any smoke or fire to announce the shift. Instead, it was happening in the halls of the province’s legislature and the labs in its universities. Quiet investments were being made that would turn Alberta into a beacon for some of the top minds in artificial intelligence and machine learning.

There aren’t any museums or eye-catching steel towers that mark the birth of the artificial intelligence sector in Alberta. But the province has something much more important: a head-start in a sector that is quickly becoming a worldwide obsession.

“There is a pipeline of talent that exists in Alberta right now, because of that early investment,” says Cory Janssen, the founder of AltaML.

The company aims to commercialize the pioneering work done by AI and machine learning researchers in Alberta, translating it into something that can aid industry in the province.

Twenty years ago, artificial intelligence was the farthest thing from a sure bet. Save for a few researchers who were pushing the boundaries of how machines “think,” most people saw it as the stuff of sci-fi.

But some could see what it was going to become. In 2000, the Alberta government set aside a $500 million endowment to create the Alberta Ingenuity Fund (also known as the Alberta Heritage Foundation for Science and Engineering Research).

One of its early programs was a competition to establish research groups for promising scientific fields. Among those created was what would become the Alberta Innovates Centre for Machine Learning, later the Alberta Machine Intelligence Institute (Amii).

The research group had an audacious goal: to attract and recruit some of the world’s top AI researchers to work at its universities.

“Back in 2002, very few people were really considering artificial intelligence and machine learning to be important areas [of research],” says Kirk Rockwell, current COO of Amii. “That’s what really gave us the advantage.

And then over the next 16 or 17 years, that funding continued to flow to the university to create this world-class research centre.”

Rockwell says that one of the vital decisions Alberta made was to put up the resources to support high-quality research. It’s an expensive endeavour, but it has a multiplying effect. The promise attracted some of the pioneering researchers of the day.

In machine intelligence, like other research-heavy fields, talent attracts talent. For many, the question of where they will end up working is heavily influenced by who they will be working with. Alberta’s initial investment started a snowball effect, drawing more talent to the province.

Not only that, but it kept them from being lured off when competing organizations with deep pockets came calling.

“There was enough going on that when people came knocking on the door to poach them for Silicon Valley or Google, there was enough to say ‘no, I’ve got world-class peers here, I’m going to stay,’” he says.

Richard S. Sutton was one of those early researchers to bring his work to Alberta. The pioneering machine-intelligence researcher had spent two decades working in artificial intelligence and reinforcement learning at centres in Massachusetts and New Jersey. In 2003, he headed north to work at the University of Alberta.

“I came to Edmonton because of the three P’s,” he says. “The people, the position and the politics.”

The politics referred to his growing unease with the direction that the United States was going under then-President George W. Bush. Most interest in AI research in the country was geared towards military applications (a trend that continues to this day), which didn’t interest Sutton. That, combined with the promise of a role where he could explore his interests and talented peers to work it, lured him north.

“They made some big hires here. The province had a lot of foresight to invest in [machine intelligence]. Gradually AI became popular, and now it is very popular,” he said.

As that popularity grew, others began to see the value in Alberta’s artificial intelligence hub. The federal government began to provide investment in the province, as well as in Canada’s two other major AI centres in Montreal and Toronto.

But in those early days, Rockwell says the funding that came through Alberta Innovates was instrumental.

“Before the federal government began contributing significantly in this space … Alberta Innovates offered $40 million in funding,” he says.

“I can’t say it wouldn’t have happened without it. But it wouldn’t look like it does now, no way. We wouldn’t be this far along, we wouldn’t have been able to keep the people we have.”

Alberta Innovates is still a major contributor to Amii’s funding.  But in an era of belt-tightening and so many different demands on provincial coffers, there’s worry about what future support will look like.

The province’s AI industry also faces a unique challenge: it’s not really an industry. At least, not in the traditional sense.

AI’s greatest impact isn’t creating things on its own. It doesn’t build mines and factories. It doesn’t fill up warehouses and train cars with products, not directly.  It does its best work in the background, making incremental changes that buoy other sectors of the economy. Right now, Rockwell points to things like using AI to identify options for preventative maintenance to save costs, or reducing emissions and energy use in industry. Other times, it can be about making scheduling work shifts more effective or anticipating the number of healthcare workers needed during a certain period.

The agriculture sector is also poised to benefit from AI advances. Even as global food demand increases, farmers across Canada are seeing shortages in two of their most important resources — labour and farmland. Smart farms, like the one set up at Olds College, allow researchers to literally field test technologies like automated soil monitoring and data-gathering combines with the goal of higher yields.

“We are trying to lift our entire economy, not just create a sector,” Rockwell says.

“What we’re not trying to do is replace anything. We’re not saying that everyone who works in the oil sector will be working at Facebook. We’re not trying to put oil and gas out of business. In fact, our goals are aligned.”

It’s effective. But often invisible. And for the people working in the sector, they worry it might not be clear to governments and businesses how important it is.

“With an oil and gas company, you can see when they build a huge facility. Or even something like Shopify, which is creating thousands of jobs. You can see that. We don’t have our own Shopify, not yet,” says Janssen.

“That doesn’t mean it isn’t there. It’s the food courts in downtown Calgary and Edmonton, which are full of young workers, people who are buying houses and getting mortgages. We are in that awkward phase that you can’t see it yet.”

To grow out of that awkward invisible stage, Alberta’s AI industry needs a healthy balance between research and application.

When all the pieces are in place, both of those sides fuel the other. Researchers find breakthroughs, which companies find real-world applications for. The profit from that allows those companies to expand and reinvest in research, attracting new talent and leading to further discoveries.

“Right now, we’ve got the scientific side in pretty good shape,” Sutton says. “Now, we need to build up the applied side.”

He says Amii has an important role to play in encouraging more real-world AI use. The organization offers a range of advisement, training and talent acquisition services to empower AI adoption across economic sectors. But a lot of it also needs to come from outside the organization, both from new companies and long-established players.

So far, many traditional industries have been hesitant to adopt new AI. Embracing machine intelligence is a risk, just like any other major change. And many private industries in Canada don’t see the benefit in taking a risk when the more traditional methods are still working.

The same goes for finding private investment in AI. With its impacts not always being as visible as other new technologies, it can be hard to understand how much potential artificial intelligence holds.

Sutton says that hesitation is understandable. But it is also a mistake. Work must be done to keep hold of the advantage that Alberta has been quietly building for the past two decades.

“If we wait, we will by necessity fail. We will lose that efficiency we have now,” he argues.

“I think as Canadians, we can be humble. But we can also be very ambitious. We need to find a mix of the ambition and the humility. And right now, is the time to be ambitious.”

AltaML’s Cory Janseen agrees. He argues that Alberta gained a great headstart with its early investments. But things have changed over the past 20 years; the rest of the world has “woken up” to the economic and social benefits that machine intelligence can promise.

Countries like China and the United States are now putting big money behind their AI industries. While they might be joining later, he says they are quickly catching up. Public funding, in addition to private investment, is still vital, he argues.

“Let me use this analogy: let’s say that there is a factory that produces the most desirable product in the world — I think it is fair to say that’s what AI is, or will be,” he says.

“Say this factory takes a billion dollars and five to 10 years to build. There are maybe 10 of these factories in the world and Alberta has one of them.  [Other countries] are out there and they’re just starting to build their factories. China, the United States are building them.

We already have that factory and we’re not even doing the maintenance on it. We’re not even slapping a coat of paint on it.”