How Montreal became the world’s leading AI and deep learning hub

Share this post:

The artificial intelligence industry is projected to be worth $36 billion by 2025. And the Canadian city of Montreal has somehow found itself in the middle of it. Earlier this year, Microsoft acquired a natural-language-processing startup by the name of Maluuba and simultaneously contributed $7 million to the AI labs at McGill University and University of Montreal. Maluuba had garnered much attention — and, most recently, $9 million in investment — for its deep and reinforcement learning research centre, established at the tail end of 2015 and advised by computer scientist Yoshua Bengio. Microsoft also announced that it would double the size of its Montreal AI research group from 40 to 80 people.

Prior to Microsoft’s announcement, Google — which already has an office in Montreal — made inroads of its own. In November of 2016, the technology company simultaneously announced that a group of seven Montreal researchers had won its Focused Research Award and that it was giving another $3.33 million over three years to the Montreal Institute for Learning Algorithms (MILA), an inter-university lab led by McGill University and the University of Montreal and headed by Yoshua Bengio. As Wired reported in November, ”This isn’t the first time the company has funneled money into the program. Over the past [10] years, Google had donated about $13 million CAD to academic research in the country and about half was earmarked for AI research.”

Just two months before Google’s announcement came one from the Canadian government. It would invest CAD$213 million into four Montreal academic institutions — McGill, University of Montreal, Polytechnique Montréal and HEC Montréal (the latter two of which are affiliated with the University of Montreal). As the Montreal Gazette city newspaper reported, some of the money will go to creating an AI and big-data lab called IVADO. Yoshua Bengio told the Gazette that the money would help create “a mini Silicon Valley in Montreal.”

But how did Montreal become a hub for AI in the first place? Evan Prodromou — the founder of Wikitravel, Identi.ca, StatusNet, founding CTO of Breather and now the founder of Montreal-based Fuzzy AI — has a simple reply: “I think the easy answer is Yoshua Bengio.”

Yoshua Bengio is considered one of the three fathers, along with Geoffrey Hinton (Toronto) and Yann LeCun (NYC), of an advanced subset of AI and machine learning called deep learning. Originally from France, Bengio earned three degrees at McGill University, including a PhD in computer science. He’s been a professor and researcher at the University of Montreal since 1993, where he’s established the Montreal Institute For Learning Algorithms. He is also a co-founder of Element AI, a Montreal company working to bring AI to every industry that will have it.

Bengio’s ascent to AI stardom began somewhere between 2010–2012, a time marked by the rise of big data — that is, the biggest datasets we’d seen, combined with the massive growth of available computing power. Suddenly the techniques Bengio had been inventing and refining for more than 20 years became extremely relevant to business. “I think these things coming together allowed for the evolution we saw in artificial intelligence,” says Philippe Beaudoin, the VP of Element AI’s research group.

Bengio, however, is just one person, and there are reasons beyond his presence in Montreal to explain why this city became a world-class hub for AI.

“Being an academic hub for deep learning was probably the first trigger,” says Narjès Boufaden, founder of Keatext, a cloud-based AI company that processes large volumes of unstructured data in the form of customer feedback. She also studied natural-language processing under Bengio while working on her PhD in computational linguistics between 1998 and 2004.

Montreal has four major public universities on the island alone (Concordia, McGill, University of Montreal, University of Quebec at Montreal, which collectively boast more than 160,000 enrollees), as well as a number of private and public academic institutions in the greater Montreal region. This year, Montreal was named the top city in the world for students. The honor was mostly predicated on Montreal’s affordability, desirability and people’s positive feelings about having studied there. And, according to Montreal AI lab IVADO, there are more than 150 deep-learning researchers at Montreal’s universities — the “biggest concentration in the world.”

Quebec as a province also has some of the lowest tuition in North America, and some of the lowest rents in Canada thanks largely to rent control. The city also has a large and vibrant cultural scene and a legacy of creativity, and Montreal’s bilingual nature makes it feel more European than North American. And finally, technology businesses in Quebec and Canada benefit from a variety of extremely generous tax credits and grants, which helps explain why so many technology companies start or move here, and why so many foreign investors are now speculating here. (In 2015, Quebec’s tech sector accounted for 6.4 percent of jobs in the province — the highest rate in all of Canada. California’s rate, by comparison, is 8.2 percent.) “People know about Canadian tax credits. Americans know their money will go further because Canada invests in tech,” says Prodromou of Fuzzy AI.

All of these factors have conspired to create a city that is cheap to live in and inspiring to create in. As Prodromou says, if you’re young and cool you want to spend a few years in Montreal.

Replicating Montreal’s momentum elsewhere would be a monumental task. Because the growth of the city’s technology industry is dependent on so many factors — many of which are decades in the making, like education and housing policies — it would likely be impossible to recreate what is happening in Montreal anywhere else.

That isn’t to say AI can’t, and isn’t, thriving anywhere else; it’s just that Montreal, at least for now, has an unfair advantage. “This is one of the very few moments where you feel like there’s an alignment,” says Boufaden of Keatext. “I think there is something happening here.”