Canada is a force in AI research. So why can’t we commercialize it?

“We’re proud of Canada’s early role in developing AI,” Trudeau said on Hard Fork, noting that many breakthroughs have happened because Canadian scientists are well-funded.

In 2017, Canada became the first country to have a national AI strategy. It launched a second phase five years later, allocating $443 million to connect research capacity with programs aimed at enabling commercialization.

This year’s federal budget included an additional $2.4-billion investment in AI. And the government has boasted that Canada has 10 per cent of “the world’s top-tier AI researchers, the second most in the world.”

Among them are two so-called godfathers of AI.

But Ottawa is “fighting to make sure we keep our skin in the game,” Trudeau told the podcast hosts.

He made the pitch, saying Canada has many of the ingredients it needs: among other things, clean energy, a good quality of life for workers and government programs to encourage the sector.

In spite of that, Canada hasn’t always been “great at commercializing,” Trudeau conceded.

More than that, Canadians have “fallen far behind,” argued Benjamin Bergen, president of the Council of Canadian Innovators, which represents the tech sector.

The government spent “a tremendous amount on the talent side of the equation,” he said recently, but not on converting it “into building companies.”

Bergen said the government has “institutionalized the transfer of our AI intellectual property to foreign firms.”

The government’s 2022 strategy update promised that the country’s three AI institutes are “helping to translate research in artificial intelligence into commercial applications and growing the capacity of businesses to adopt these new technologies.”

But Bergen argued an AI strategy focused on commercialization must start with Canada owning its own IP. “You cannot commercialize what you don’t own.”

Intellectual property lawyer Jim Hinton has been trying to quantify that problem.

And the numbers show “a train wreck I’ve been watching happen in slow motion,” he said.

About three-quarters of patents produced by researchers who work for Toronto’s Vector Institute and Montreal’s Mila leave the country, and most of these are in the hands of Big Tech, Hinton’s research has found.

Another 18 per cent of the 244 patents he tracked — 198 from Vector and 46 from Mila — are now owned by North American academic institutions.

Just seven per cent are held in the Canadian private sector.

Of the foreign-owned patents, the largest number, 65, went to Uber, while 35 landed with the Walt Disney Company. Nvidia, which recently displaced Microsoft as the world’s most valuable company, got 34.

IBM ended up with 15 and Google with 12. A handful of the patents were co-owned.

Foreign companies benefit from Canada’s public funding, Hinton argued, and there are “no guardrails put on the ability for these foreign companies to basically pillage Canada’s really good AI invention.”

Researchers can work at the AI institutes and foreign tech companies at the same time, Hinton said, charging that this is what allows the tech giants to take advantage.

The Canadian Institute for Advanced Research, which co-ordinates the government’s AI strategy, pushed back strongly on that assertion.

Executive director Elissa Strome said a “small number of our researchers” have part-time employment in the private sector.

“Those private-sector organizations own the rights to the IP that is generated by those researchers,” she said, but only when they’re on the clock for those companies.

Strome said it’s long-standing practice in Canadian research “that there are relationships around contract research with industry,” and “a really strong firewall” is in place between IP generated via public funds at the AI institutes and that which is generated through private funds.

She said Hinton’s statistic on patents was inaccurate, but did not provide data to refute his findings.

She also argued that patents are not a good measure of commercialization, and “it’s the people that we’re training in the AI ecosystem that actually hold the greatest value in AI, not patents.”

When it comes to sponsorship agreements at Toronto’s Vector, any IP created at the institute “belongs to Vector,” a spokesperson said, adding it is not the primary employer for most of its researchers.

If academics don’t have an opportunity to work for companies, they’re more likely to leave altogether, Montreal’s Mila said in a statement. It said the three institutes have turned around a “massive brain drain in AI in Canada” that existed prior to 2017.

The multi-billion-dollar investment in this year’s budget seeks to further protect against that brain drain by beefing up Canadian infrastructure and computing power.

The envelope includes a “relatively small” amount of money to help Canadian companies scale up, noted Paul Samson, president of the Centre for International Governance Innovation.

Overall, the government is “doing the right thing” by ensuring that’s part of the equation, he said.

But people in the tech sector are skeptical. Bergen said companies were given little time to provide input.

“The government already had a top-down strategy that it wanted to implement … and didn’t really care what CEOs and leaders of domestic firms were actually needing in order to be successful,” he said.

Nicole Janssen, co-CEO of AI company AltaML, raised the concern that the Canadian government might end up simply throwing money at American firms to move north.

“What I’m trying to figure out is how the government thinks they’re going to spend $2 billion on building computers without just handing that $2 billion to Microsoft,” Janssen said.

The budget said the money would go towards both access to computational power and developing AI infrastructure that is Canadian-owned and located in Canada.

A spokesperson for Industry Minister François-Philippe Champagne said more details would be provided in the coming weeks.

Companies like Microsoft and Nvidia are already looking to Canada as a place to build computing infrastructure, Janssen said, due to factors like climate and relative political stability.

“We don’t need to do anything to attract them.”

A better approach, Janssen said, would see the government helping Canadian firms adopt AI more quickly — a gap her company has been trying to help fill.

It takes AltaML an average of 18 months to start building an AI product in Canada, she said, compared to four months in the United States.

“We definitely do not have the ecosystem of companies that you would expect for the amount of talent that we have,” she said.

There’s real clout at Canada’s AI institutes, with veterans Yoshua Bengio and Geoffrey Hinton heading up Mila and Vector, respectively.

They and other elite researchers have “attracted students from all over the world to come study under them,” said Janssen, and that’s a big advantage for Canada, especially if it wants, as Trudeau said on the podcast, to lead in developing a more democratic AI.

The prime minister said one of his biggest preoccupations is maximizing “the chance that it actually leads to better outcomes and better lives for everyone” instead of only benefiting those “with the deepest pockets.”

Canada could be a leader in responsible AI, Janssen said.

“That is a title that is up for grabs,” she said. “And no one has grabbed it yet.”

This report by The Canadian Press was first published June 26, 2024.

Anja Karadeglija, The Canadian Press

by Canadian Press

Canada is a force in AI research. So why can’t we commercialize it?

Pro Tips