Public-private partnerships can help reverse the AI brain drain

Last month, the Alberta government committed $100 million over 5 years to attracting artificial intelligence (AI) companies to the province, following similar investments by Ontario and Quebec. The Alberta Machine Intelligence Institute (Amii), an Edmonton-based non-profit research institute, will receive $27 million of that funding to build industry collaborations, while the Vector Institute in Toronto is similarly forging partnerships between machine learning researchers and industry. Vector counts 27 large enterprises as partners, all of which have committed funds to the organization.

The collaborations led by Amii, Vector and the Quebec Artificial Intelligence Institute (Mila) — the three institutes underpinning the Pan-Canadian AI Strategy — are designed to help industry access the experts they need to create new AI-driven products and services. But they are also a kind of national defence strategy in the battle to retain Canada's AI talent.

“In terms of AI research, Canada already leads the world. In terms of employment, we need to do more,” said Cameron Schuler, Chief Commercialization Officer & Vice President, Industry Innovation at the Vector Institute, in an interview with RE$EARCH MONEY.

[rs_related_article slug="quebec-ai-institute-mila-opens-giant-new-facility-in-montreal"]

By creating a dynamic AI ecosystem in Canada that prioritizes industry collaboration alongside research and training, Vector and the other AI hubs are betting homegrown job opportunities will follow. “Vector was really built on the principle that supply creates demand — sort of a build-it-and-they-will-come approach,” said Paige Dickie, Senior Engagement Manager at Vector, where she leads a multi-stakeholder project focused on using AI to combat financial crimes and improve financial services.

In return for their sponsorship, Vector's partners are looking to address their most urgent need: accessing AI talent. As a result, "Vector is trying to create a more efficient market between graduates and companies. It's helping companies understand what opportunities need to look like to be able to attract the talent we produce," said Schuler. His industry innovation team also works with sponsors to identify business opportunities and challenges that could be addressed using machine learning.

[rs_related_article slug="cifar-unveils-first-cohort-of-ai-chairs-at-pan-canadian-ai-strategy-annual-meeting"]

Randy Goebel, a machine learning pioneer at the University of Alberta who co-founded Amii in 2002, has supervised approximately 80 graduate students over his career. Only two still work in Alberta, he said in an interview with RE$EARCH MONEY. However, the arrival in Edmonton of companies such as DeepMind, which is owned by Google's parent company Alphabet, and Borealis AI, the Royal Bank of Canada's research and development lab, is beginning to change that. 

DeepMind is actively collaborating with Amii researchers. In fact, DeepMind Alberta, which launched in 2017, was the company’s first research lab outside of London, United Kingdom. It currently employs five of Goebel’s university colleagues on a part-time basis. They work for DeepMind Alberta 25 to 75 percent of the time, he said, retaining their positions as university professors, where they continue to teach and supervise graduate students.

In contrast, Goebel said the University of British Columbia was "devastated" by the hiring away of three of their top AI researchers. "From my point of view, just hiring them and moving them out of the academic system is short-range thinking," he added.

Borealis AI has labs in Toronto, Montreal, Waterloo, Vancouver and Edmonton, where it has 10 full-time research staff and 10 positions for internships for students. Unlike DeepMind, it doesn't directly employ professors from the University of Alberta, but is poised to take advantage of the talent pool of graduating students. In 2018, it also launched a Graduate Fellowship Program to support graduate students in the fields of machine learning or artificial intelligence who are studying at Canadian universities.

"The mature industries recognize that they don't want to hire away an old professor like me to work for them until I grind to a halt. They want to have access and build relations that make them the most attractive employer for my graduate students," said Goebel.

Bridging curiosity-driven research and value creation

However, establishing partnerships that are attractive to both researchers and industry can be a challenge because their incentives are starkly different. Curiosity-driven research is the real draw for academics, whereas businesses want to create value. So tackling problems that would dramatically reduce costs might be a priority for industry; in contrast, academic researchers may be drawn to solving fundamental scientific problems they can publish. Timelines and the element of ambiguity in fundamental research can also be an impediment to successful partnerships. Industrial timelines are often short and definite, in contrast to academic ones. 

Schuler summed up these differences: “Vector really does focus on building the future of AI, whereas companies tend to focus more on the application of AI.”

In spite of these challenges, new academic-industry partnerships are continually emerging. For example, Amii's new Innovation Affiliated program, which is designed to help companies incorporate artificial intelligence into their enterprise, signed its first agreement last week with Korea's MINDs LAB.

Canada’s AI hubs are just a few years old, so it is still difficult to assess whether the approach is working. But Dickie said some companies have directly attributed the choice to open AI labs in the Toronto area to the Vector Institute. According to Goebel, academic-industry partnerships are changing the employment dynamic for AI experts in the Edmonton area: "Our politicians aren't going to be able to say in one year, 'Oh yeah, last year we had zero hiring away of our AI graduate students and now they're all working in Edmonton.' But it is changing."

R$