Valerie Pisano, president and chief executive of the Montreal Institute for Learning Algorithms (MILA), in the company’s Montreal office in July, 2018.Dario Ayala
Valérie Pisano is president and chief executive at Mila. Mark Surman is president at Mozilla.
The deep learning breakthroughs that now power nearly every major artificial intelligence (AI) system in the world were first developed in Canada. As Canadians – one of whom leads a world-renowned AI lab and the other who runs a global open-source project – we’re proud of what our country contributed to this moment. And we believe the most important chapter of that story hasn’t been written yet.
Over the next decade, Canadian companies, hospitals, banks and governments will spend tens of billions of dollars on AI. Right now, almost every one of those dollars flows to a handful of hyperscalers from across the border. As Canada prepares to unveil its renewed national AI strategy, the central question is not whether we will adopt AI. It is how much of it we will own.
We won’t outspend Silicon Valley. We won’t retreat into protectionism. But we can do something neither hyperscalers nor trade barriers can: help build the open and trustworthy AI ecosystem the world actually needs.
OpenAI has shown it cannot be trusted. Canada needs nationalized, public AI
The history of the web is instructive here. It wasn’t built by a single company or in a single country. Rather, it was built on shared standards such as HTML and open infrastructure such as Linux. No one owns HTML. No one owns Linux. They are designed so that anyone anywhere can use them, build on them, improve them. It is this openness that made the web the foundation of the modern economy. Open source AI works on the same principle. The models, the tools, the building blocks are developed in the open, available to all, owned by no one exclusively.
Over the last year, we’ve reached the point where open-source AI models can go toe-to-toe with their closed competitors on performance, and can be run at a fraction of the cost. But having a powerful AI model is only the beginning. What determines whether it actually works for a hospital, a school, or a small business is everything built around it, including the tools, the safeguards, the human expertise to put it to use. Most of that infrastructure is currently controlled by a handful of U.S. companies.
This is the gap our two organizations, Mila and Mozilla, are working to close, and where Canada’s opportunity sits. But no single lab, and frankly no single country, can close this gap alone.
That is why middle-power co-operation may be the most important idea in Canadian AI policy right now. Germany has created a Sovereign Tech Agency to fund the open source infrastructure its economy depends on. Britain has set out to become the home of global open-source AI talent. France, Japan and the Nordics are making parallel moves. Together, middle powers can co-fund the shared infrastructure, from deployment tools to safety and cybersecurity systems, that all of us need and none of us should have to rebuild five times over. Open source is the mechanism that makes that co-operation real. Without it, middle-power AI coalitions are goodwill without enough backbone.
Ottawa plans major investment in AI pioneer’s non-profit
Canada is better positioned than most to lead that coalition. We have the research credibility. We have relationships across the democracies that are making these moves. And we have a rare opportunity, in this moment of strategic realignment, to turn those assets into something concrete.
That means formalizing a middle-power AI partnership with like-minded democracies, built around joint projects and shared procurement, such as the Sovereign Technology Alliance launched by Canada and Germany earlier this year. It means investing deliberately in Canadian open-source AI research and startups, with a focus on the sectors where the returns are most tangible: health care, education, agriculture. And it means taking the lead in building a global hub that pulls in ideas, talent and resources from countries and researchers around the world – making Canada the connective tissue, over and above being a participant in it.
None of this means walking away from the big AI platforms. It means building alongside them: using open source to ensure that Canadian institutions, Canadian researchers and Canadian companies own a piece of what is being built, rather than renting it indefinitely. Canada helped spark this revolution. The question now is how we’ll show up to own what comes next.