Canada has been the worst performing advanced economy in the Organisation for Economic Co-operation and Development since 1976. Governments of all partisan stripes have tried and failed to reverse the trend. If nothing changes, the OECD projects, our economic growth per capita will continue to stagnate for decades to come. This article is part of an occasional series called Per Capita, which examines how and why policy interventions have come up short – and how fresh approaches to economic growth are urgently needed.
Ask anyone involved with the Pan-Canadian Artificial Intelligence Strategy if the federal program has been a success and you’ll get a decisive “yes.”
“When we travel abroad other countries are envious of what Canada has done around AI,” said Alan Bernstein, recently retired chief executive of the private-public Canadian Institute for Advanced Research (CIFAR), which oversees the strategy.
The way he and Elissa Strome, executive director of the strategy and a CIFAR employee, put it, the talent the AI strategy has cultivated should speak for itself. It has funded more than 100 experts, including many foreigners who came here. The initiative has trained more than 1,000 AI students in Ontario alone.
More than 90 per cent of graduates from Toronto’s Vector Institute for Artificial Intelligence, funded by the program, have stayed in Ontario.
Many University of Alberta machine learning graduates used to move to London to work for Alphabet Inc.’s DeepMind Technologies AI unit. “Now they have a potential home here,” including a DeepMind office in Edmonton, says Cam Linke, CEO of the Alberta Machine Intelligence Institute (AMII).
But for many leaders in Canada’s innovation sector, the government’s strategy to build on the country’s early lead in AI is emblematic of how Ottawa routinely fails to leverage the country’s brains to produce economic growth. There are similar criticisms of other federal innovation programs: There’s too much focus on research, not enough on commercialization and economic development; an overemphasis on welcoming foreign players, not supporting domestic companies; and Canada now risks losing its early leadership in AI to other faster-moving countries.
The AI strategy “exemplifies a lot of issues” with the Trudeau government’s innovation agenda, according to Benjamin Bergen, president of the Council of Canadian Innovators, which represents domestic tech companies.
A strategy is born
That Canada has a well-regarded AI sector is a genuine accomplishment stemming from the country’s key advantages: It is an advanced economy with top research institutions and is often regarded as a kinder, gentler place than the United States.
By the 2010s, Canada was home to three renowned AI professors: Geoffrey Hinton at University of Toronto and Yoshua Bengio at Université de Montréal were both pioneers of deep learning, a branch of machine learning that simulates how the human brain works to learn from data. And Rich Sutton was an expert in another branch of AI called reinforcement learning at University of Alberta. None were born in Canada, though Dr. Bengio was raised here. They had migrated from the United States, aided by CIFAR, because they preferred to live here for political reasons.
Their genius spawned an academic ecosystem in a field that had faded into near obscurity. That changed owing to breakthroughs by the trio and others, many students they taught, which caught the eye of tech giants a decade ago.
Before Canadian policy makers realized what was going on, foreign companies were feasting on Canada’s AI brains. Dr. Hinton sold his startup DNNResearch Inc. to Google Inc. in 2013 (he still works for the search giant), and other big foreign players snapped up Canadian-trained AI researchers and startups. The buyers wanted them to help build innovative applications for content recommendation, speech and image recognition, driverless vehicles and other uses.
By 2016, Canadian corporate leaders were pressing Ottawa to protect the country’s shrinking AI legacy. A group that included AI entrepreneurs Jordan Jacobs and Tomi Poutanen, Prof. Hinton and former Toronto-Dominion Bank CEO Ed Clark pushed to create an AI institute in Toronto to expand the number of machine learning graduates and give them opportunities to stay.
“I realized AI would add more to the global economy than any technology since the microchip,” said Mr. Jacobs. Because foreign players siphoned it away, “not much economic value had been built in Canada.”
In February, 2017, AI leaders, CIFAR officials, institute proponents and federal officials met to hash out the AI strategy. The program was announced weeks later in the budget, with $125-million in funding for five years.
The strategy established three AI institutes: the Montreal Institute for Learning Algorithm, Toronto’s Vector Institute and AMII in Edmonton. Most of the funding went to top-flight AI researchers who would be attached to the three bodies and hold chairs at Canadian universities.
The plan was to reverse the AI brain drain. “Ottawa agreed, and we felt strongly, that if you really wanted an innovation strategy, the first and foremost thing you need is people,” Mr. Bernstein said.
Many AI experts stayed, but the foreign opportunities also came here. About 50 companies, including Facebook Inc., Google and Samsung Group, opened AI labs in Canada. (Several Canadian companies, such as Royal Bank of Canada and Thomson Reuters Corp., opened labs as well.) Researchers no longer had to emigrate to do cutting-edge work. Mr. Bernstein welcomed them and maintains foreign giants are an essential ingredient of a “rich, confident ecosystem.”
But the work those AI researchers employed by foreign companies here do and the intellectual property they create is owned and commercialized abroad. It’s not quite a brain drain, but the result isn’t much different in terms of employers from other countries generating economic benefits thanks to Canadian talent. Is this just another branch-plant strategy, with Canada funding growth in other countries?
‘I want more patents’
Jim Hinton thinks so. According to research by the Waterloo IP lawyer, AI experts funded by Ottawa through the strategy, primarily those with Vector, have been granted 232 patents since the AI strategy launched. Of those, 75 per cent are owned by foreign entities, including Uber Technologies Inc., Nvidia Corp. and Google and IBM Corp. Just 18 of those patents, or 7.8 per cent, are assigned to Canadian companies. It amounts to a “philanthropic contribution by Canada to some of the world’s most valuable companies’ AI research,” Mr. Hinton told The Globe and Mail.
Another issue is that there is little AI patenting activity in Canada. The U.S. Patent and Trademark Office says there were nearly 80,000 AI-related patent applications in 2020. Vector estimates there were just 66 new AI patent filings in Canada in 2021-22, although that was up 61 per cent, year over year.
But AI patenting has spiked in other countries, too, notably China and Britain. “Our companies own very few AI patents,” Ottawa patent lawyer Natalie Raffoul said. “That means long term they will have trouble with AI-enabled technology in the global marketplace. Other countries are holding troves of valuable IP rights in AI.”
Mr. Hinton said that while other countries’ research institutions are busy patenting AI inventions, “Canada’s taxpayer-funded institutions are having virtually zero impact on commercialization benefits for Canada’s economy.”
Surprisingly, some people overseeing Canada’s AI strategy contend IP doesn’t matter. “I don’t think IP is a measure of anything other than how many patents you have. It’s not a measure of how many jobs you’ve created or added wealth to the economy,” Mr. Bernstein said.
AMII’s Cam Linke echoed that: “Patents aren’t really the big driver. As a specific measure of success, it isn’t necessarily the right [one] in my mind.”
Ms. Strome, the head of the AI strategy, agreed. “It hasn’t been our objective” to generate patents, she said. “Patenting is not a major course of action in this sector” nor “actually relevant. I don’t think those patents that exist are generating value for the people who do own them.”
The key to generating value, Ms. Strome said, is training people to get jobs, launch startups and work “at Canadian companies and others in the ecosystem.”
It’s true some researchers hired by foreign companies here have later started companies locally, including driverless-vehicle maker Waabi Innovations Inc. and language generator Cohere Inc. In the past, they might have moved to Silicon Valley.
But the view that IP doesn’t matter in AI beggars belief. AI patent applications in the U.S. in 2020 were nearly double the level in 2010. And many transformative inventions have coincided with spikes in patent filings and court battles over who owns the ideas. It happened with the sewing machine, lightbulb, telephone, car, airplane, TV, semi-conductor and smartphone.
While mathematical formulas can’t be patented, the mounting paperwork in patent offices shows there is perceived value in AI inventions. And with those patents come rights such as freedom to operate and to sue for infringement or extract costly licenses. “It’s no surprise we don’t have an IP strategy: Our AI ‘leaders’ are naïve and unsophisticated about IP,” Mr. Hinton, the IP lawyer, said. “In the U.S., IP is a way of life.”
Other participants in the AI strategy side with him. “IP is an important tool in building a competitive, defensive, profitable company,” a Vector spokesman said in an e-mail. In a blog last October, outgoing Vector CEO Garth Gibson wrote: “For countries, succeeding in the knowledge-based economy is about more than education and innovation. Increasingly it’s about securing and utilizing a nation’s intellectual property. That’s especially true in the era of AI.”
And Innovation Minister François-Philippe Champagne said when it comes to AI, “I want more patents.” He said the government includes IP generation and retention in Canada as a condition of grants. “I’m a big believer that in the 21st century economy, IP is key. Whatever AI we finance we want that to benefit Canadian companies.”
From leader to follower
Canada had the first national AI strategy, but dozens of countries followed. Their approaches were more than talent plays. Many treated AI as a key strategic area, plotting multibillion-dollar plans to harness its economic potential.
China is investing tens of billions of dollars in AI. The White House established a National AI Initiative Office in 2020 and the U.S. government made AI a focus in spending bills last year.
Britain’s AI strategy involves an Office for AI, the country’s defence department, an Intellectual Property Office and other agencies and departments. They want to ensure Britain invests broadly to adopt and use AI, that it benefits all sectors and regions, and that the technology is governed effectively.
Those and other national strategies have been explicit about something lacking from Canada’s initial AI strategy: commercialization, to ensure their economies benefit from adoption and exploitation of AI.
Canada’s second AI strategy, unveiled in the 2021 budget, is no groundbreaker, but a distant follower. Ottawa committed $443.5-million: $208-million to CIFAR to fund institutes, researchers and training programs for 10 years, $40-million for computing resources for the institutes and $8.6-million to develop and adopt AI standards.
The government did earmark $185-million for commercialization: $125-million to superclusters to invest in AI projects and $60-million for the institutes over five years. Much of the institutes’ efforts involve helping existing companies learn how to use AI.
But the AI strategy is still largely a talent-retention plan and the dollars are relatively modest. To the extent there is a strategy to wring economic benefits out of Canadian AI, it is downshifted to arm’s-length bodies. It lacks a central agency responsible for policy, planning and delivery to make AI a national priority.
In 2019, Ottawa did establish an Advisory Council on AI composed of public and private sector figures. Part of its mandate is to ensure Canadians benefit through job creation and economic growth.
But while the council tasked a separate working group to make recommendations on commercialization, summaries of the council’s meetings in 2021 and 2022 show it mostly discussed responsible AI and Ottawa’s still-unadopted digital charter that will govern how the personal data of Canadians is used. The summaries make little mention of IP and none of commercialization.
There also seems to be no overarching plan in government for how AI fits into the big picture. Jim Balsillie, the former co-chief executive of BlackBerry and a leading proponent for more effective innovation programs, said he’s asked “repeatedly” to see the document that guides the strategy, “but no one in government can provide a document or confirm it even exists.”
In fact, there isn’t one, a senior government official told The Globe and Mail in October. The Globe is not identifying the source, as they are not authorized to be named publicly.
While Mr. Champagne insists the AI strategy “is run by my office and me personally,” Mr. Bernstein said: “I think the strength of our strategy is it’s not being run centrally by a bunch of bureaucrats in Ottawa whose strength is not running programs.”
The first AI strategy succeeded at retaining people, but “the problem is we have not gone to the next step, creating firms that are scalable and will remain in Canada,” said Robert Asselin, who was budget director for then finance minister Bill Morneau. ”If we don’t take these next steps, it’s not a strategy, it’s just an R&D lab for foreign firms.”
Raquel Urtasun, a University of Toronto professor who led Uber’s driverless-car research group before starting Waabi in 2021, said helping established companies adopt AI, as the institutes do, “is a big miss. Having more Shopifys will make more things happen now,” she said. “I would definitely focus on scale-up financing.”
Coming next: The Liberals have yet another innovation plan, and industry experts offer suggestions