Most of the $1.1 billion in federal funding for artificial intelligence over the last 13 years has gone to the private sector and to academic projects closely linked to industry, according to a new paper from a McGill University researcher. Government support for the technology has also been unevenly distributed, flowing to a handful of startups, Big Tech firms and universities based in a few provinces.

“So much money is being poured into this industry … and we do not have any clear regulatory frameworks,” said author Ana Brandusescu, a professor of practice at the Centre for Interdisciplinary Research on Montreal. She said citizens deserve more information about and say in which AI companies and projects governments back, because—unlike some other publicly funded fields—the technology may invade their privacy or produce biased outcomes that hurt them.

Ottawa has leaned into the domestic AI sector’s potential as an economic driver, providing funding, applying the technology to government operations and championing global discussion forums about its ethical use. The 2017 federal budget set aside $125 million for the Canadian Institute of Advanced Research (CIFAR) to run a new Pan-Canadian Artificial Intelligence Strategy (PCAIS) of hiring academics and supporting development. Ottawa has also provided loans and grants to startups employing AI, like auditing-software maker MindBridge Analytics, warehouse-robotics firm Attabotics and biotech company AbCellera. And in December 2018, it allocated almost $230 million to Scale AI, the Montreal-based supercluster focused on applying the technology to supply chains.

Governments around the world are backing AI because “it promises economic growth, military advantage, and streamlining labour functions through automation,” the report states. In Canada, the federal government awarded 71.7 per cent of disclosed AI funding to for-profit organizations between September 2007 and June 2020—almost $795 million—according to Brandusescu’s analysis; non-profits and charities (15.7 per cent) and academia (12.6 per cent) received significantly less.

That’s in line with Ottawa’s goal of growing the economy and creating jobs. But Brandusescu sees a need for more accountability on how public funds are spent on AI, because of its potential to harm people, particularly those from marginalized communities. Researchers and companies have found that machine learning systems replicate existing racial bias in fields like hiring and policing, and expand it to new technologies like smart devices and online search. “So much of who we are [is] these data points,” said Brandusescu, noting that the results of government-backed AI projects are more likely to directly impact citizens’ lives than, say, a new publicly subsidized manufacturing facility. 

“We don’t really have regulation of AI” in Canada, said Brandusescu. “The government invests in firms [and relies] on private regulation and internal ethics frameworks.” Federal departments launching automated decision-making systems are now required to assess their potential negative consequences. But there’s currently no such directive for how AI can be used in the private sector, the report notes. Ottawa’s proposed new consumer-privacy law would require companies to disclose whether and how they’re using algorithms to make decisions. 

Brandusescu called for greater transparency requirements for AI companies and projects that receive significant government funding, such as releasing annual reports or quarterly budgets, while respecting trade secrets. “We as the public are supporting you through taxes,” she said. “Can we see a little bit about your progress … so we understand what you’re making?” 

Innovation, Science and Economic Development Canada did not directly answer questions about whether its grants and contributions to AI companies include any ethical requirements. “AI innovation and growth need to reflect Canadian values such as human rights, inclusion, diversity, innovation and economic growth,” said department spokesperson Riyadh Nazerally, citing Ottawa’s advisory council on the technology and participation in the Global Partnership on Artificial Intelligence as examples of its efforts.

Brandusescu’s report also states that public AI dollars have mostly flown to a small number of provinces, startups and the engineering and computer-science departments of universities. 

Making big bets on a handful firms is risky, since they may be lost to foreign takeovers. In June 2020, Google acquired Kitchener, Ont.-based North, a smart-glasses company that was awarded $24 million in federal funding. In November 2020, ServiceNow bought Element AI, a Montreal firm in which the provincial government and pension fund had invested significantly. By spreading public money more equitably, “we’ll have other ones that will flourish” when firms are acquired, said Brandusescu. (Governments recouped their funding in both cases).     

According to the analysis, organizations in Quebec received just over two-fifths of the $1.1 billion in federal funding, while those in British Columbia (32.7 per cent) and Ontario (17.7 per cent per cent) were also awarded significant sums. “That’s where you have these long-term relationships, investments [and] people,” Brandusescu acknowledged, but allocating funding on that basis risks replicating existing inequalities. “What about the territories—what are they doing with AI? There’s zero money going to them.” She also sees a need to fund social-science research into the economic, social and political effects of AI, in addition to the labs creating it.

The report cites “a concentration of research and activity” at three CIFAR-backed centres, each built around a leading scientist: the Université de Montréal and McGill University’s Mila partnership; the University of Toronto’s Vector Institute; and the University of Alberta-linked Amii in Edmonton. 

Those three cities “already had a very long history and a critical mass of AI research expertise,” said Elissa Strome, executive director for the PCAIS at CIFAR, citing academic strength, provincial and local government support, venture capital and companies. While all of the 100-plus research chairs named under the strategy are affiliated with one of the three institutes, some are based elsewhere, backed by a dedicated pool of funding. Strome also noted that CIFAR has “a dedicated program on AI and society.”

***

Methodology

The paper’s analysis of federal government AI funding is based on a data set drawn from Ottawa’s database of proactively disclosed grants and contributions decisions between September 2007 and June 2020; keywords included “artificial intelligence,” “AI,” “ADM,” “machine learning” and “deep learning.” Brandusescu also conducted 53 interviews with government officials, company executives, policy experts and academics.