Canada’s AI Future: The Case for a Nationalized, Public Model

Canada stands at a critical juncture in its artificial intelligence journey. With the Carney administration’s $2-billion, five-year Sovereign AI Compute Strategy now underway, the nation must decide whether the benefits of “sovereign AI” will truly accrue to Canadians, or if the investment will primarily bolster American Substantial Tech. The debate is intensifying, fueled by initiatives like OpenAI’s “OpenAI for Countries” and concerns about the influence of foreign corporations on a technology poised to reshape society.

The OpenAI Dilemma and Concerns Over Transparency

OpenAI, the creator of ChatGPT, is actively pursuing opportunities within Canada’s AI strategy. The company’s top lobbyist has engaged with Ottawa officials, including Artificial Intelligence Minister Evan Solomon [Security Boulevard]. However, questions surrounding OpenAI’s transparency have emerged. Following a shooting in Tumbler Ridge, British Columbia, where the perpetrator had engaged in gun-violence chats on ChatGPT, OpenAI initially resisted alerting law enforcement. The information only became public through reporting by the Wall Street Journal [Security Boulevard].

This incident highlights a fundamental issue: when AI development is driven by private interests, the technology may not align with public safety or ethical considerations. OpenAI’s initiative is as well explicitly coordinated with the U.S. Government [Security Boulevard], raising concerns about dependence on U.S. Laws and priorities, particularly given the increasingly protectionist stance of the Trump administration [Security Boulevard].

The Case for Publicly Funded AI

Critics argue that simply attracting foreign companies to establish data centers in Canada, as proposed by OpenAI, does not address the underlying issue of control and benefit. The only viable alternative, they contend, is a bold investment in a wholly Canadian, publicly funded AI model – an AI built and operated as public infrastructure for the benefit of all Canadians [Security Boulevard].

Such a model would prioritize practical applications tailored to Canadian needs, such as:

• Improving healthcare through AI-assisted radiology and early cancer risk detection.
• Providing personalized education through AI tutors aligned with provincial curriculums.
• Matching job seekers with relevant government programs and analyzing labor market trends.
• Optimizing transit schedules, energy grids, and urban planning.
• Streamlining court processes and customer service.

The Swiss Apertus Model: A Potential Blueprint

Switzerland offers a compelling example with Apertus, a publicly funded AI model developed by ETH Zurich, EPFL, and the Swiss National Supercomputing Centre [Security Boulevard]. Apertus, leveraging renewable hydropower and existing infrastructure, demonstrates that a powerful AI can be built without the massive investments and potentially unethical practices often associated with Big Tech. While its performance currently lags behind leading corporate models by a year or two, it is more than sufficient for a wide range of applications and is freely available for use and further development [Security Boulevard].

Apertus also highlights a more sustainable economic framework for AI, demonstrating that frequent, astronomically expensive training runs are not essential for practical AI development. The model was trained with 70 billion parameters, significantly less than the largest offerings from tech giants [Security Boulevard].

Democratic Oversight and Ethical Considerations

A public AI system would allow for democratic input and oversight on critical ethical questions, such as data privacy, bias mitigation, and access control. These decisions, which will profoundly shape society as AI becomes more pervasive, are currently made in secret by corporate entities [Security Boulevard].

Canada’s Existing Strengths

Canada possesses significant strengths to build upon, including world-class AI research institutions like the Vector Institute, Mila, and CIFAR, and the existing $2-billion Sovereign AI Compute Strategy [Security Boulevard]. The key now is to reorient the strategy towards a fully open, public AI model.

OpenAI is building a global network of data centres to store the massive amounts of information collected by its services, including ChatGPT, and Canada’s cheap energy could support with that [CBC News].