CIFAR unveils first cohort of AI chairs at Pan-Canadian AI Strategy annual meeting

The first Pan-Canadian AI Strategy annual meeting (AICan) convened in Montreal on December 3 to announce its initial cohort of 29 Canada CIFAR Artificial Intelligence (CCAI) Chairs.

The CCAI Chairs are part of a talent recruitment and retention plan worth $86.5 million of the total $125 million that the the Government of Canada gave to the Canadian Institute for Advanced Research (CIFAR) to lead the Pan-Canadian Artificial Intelligence Strategy. CIFAR will recruit 50 CCAI Chairs in total, the rest of whom will be announced in 2019.

Beyond the CCAI program, CIFAR will divide the remaining funding between three other strategic pillars: establishing and supporting three AI research centres across Canada; building a national program of activities to foster collaboration between AI researchers; and leading an AI and society program to reflect on the ethical, economic and social implications of AI.

The first 29 CCAI Chairs will receive $30 million between them, to support the direct costs of their research programs, such as paying salaries to graduate students or buying equipment, as well as a personal stipend. Each chair is associated with one of the three national AI institutes supported by CIFAR: Alberta Machine Intelligence Institute (AMII) in Edmonton, the Montreal Institute for Learning Algorithms (MILA) and the Vector Institute in Toronto

The CCAI Chairs were nominated by the three AI institutes and the nominations were reviewed by an International Scientific Advisory Committee made of scientific leaders from major institutions like Google, DeepMind, Microsoft, Facebook, Stanford, and Princeton. At the AICan event, each of the 29 chairs appeared on stage over the course of five panel discussions, addressing key themes in contemporary AI research, such as machine learning, reinforcement learning and deep learning.

The initial cohort is a diverse group that includes established experts and early-career researchers. Half are taking their first faculty positions in Canada, and nine are women, far exceeding the usual ratio of female professors in computer science, said Elissa Strome, executive director of the Pan-Canadian AI Strategy at CIFAR, in an interview with RE$EARCH MONEY.

The right place at the right time

When industry and academic leaders convened in Montreal two years ago to begin formulating the Pan-Canadian AI Strategy, they wanted to confront the brain drain of Canadian AI researchers, who are frequently poached by large American tech companies. “We were worried about losing that critical mass of research expertise in Canada, and that’s why talent retention and recruitment is such an important part of the strategy,” said Strome.

Strome believes that Canada is now in a better place to draw top AI talent from abroad.  “Canada is a really attractive place for people from all over the world to come,” she said. “[Researchers] see Canada as a safe haven.”

CCAI chair Marzyeh Ghassemi, 33, said that social and political factors played an important role in her decision to come to Canada. Ghassemi works on machine learning for health at the University of Toronto and the Vector Institute. She completed her PhD at MIT and was named one of MIT Technology Review’s 35 Innovators under 35.

“I do think the political climate in which you conduct your research affects you, and I don’t think that’s insignificant for a lot of people,” she told RE$EARCH MONEY. Ghassemi pointed to Canada’s inclusive health system and said that when she hires doctoral students from abroad, she tells them that Canada has stricter gun controls than America. “If you are an international student and you have darker skin, that might resonate with you,” she said.

Other CCAI chairs made references to Canada’s socio-political advantages during the daylong conference. “Politically, [Canada] is a more open society than other North American societies,” observed University of Montréal professor and former Stanford postdoctoral scholar Ionnis Mitliagkis, drawing laughter from the crowd at the thinly veiled reference to the United States.

But even if this is the right place and time to implement a talent recruitment strategy in Canada, Strome recognizes that geopolitical circumstances can change quickly. “We have to make sure that Canada continues to be a really attractive place for people to stay and for new researchers to come to,” she said.

Building a collaborative research community

Canada’s appeal for foreign researchers goes well beyond socio-political factors. Panelists at the AICan event emphasized the country’s strong support of fundamental research. “The government really has the courage to fund fundamental machine learning research,” said CCAI chair Roger Grosse, a professor at University of Toronto who completed his PhD at MIT. Grosse expressed gratitude for the “freedom of thought to go in new directions with these ideas.”

Several chairs also highlighted favorable circumstances for partnerships with industry as an important factor, pointing to many opportunities for collaboration with Canadian AI startups. “Half my projects are with companies in the local ecosystems,” said Aaron Courville, a CIFAR Fellow and Associate Professor at University of Montreal.

Other chairs pointed to Canada’s long history of supporting machine learning and advanced AI research. “Canada is the spiritual home of the type of research I do,” said Blake Richards, a faculty associate at Toronto’s Vector Institute.

“Twenty years ago, this was already an interesting place to do research on artificial neural networks inspired by the brain,” said Pascal Vincent, a professor at University of Montreal and research scientist at Facebook AI Research. That deeper history contributes to strong connections between researchers, he said: “We all know each other, which makes a lively, collaborative, spirited community, and that’s what’s best for advancing research.”

Addressing the hype

AI researchers must contend with the problem of hype, which feeds popular fears of human obsolescence, noted Yoshua Bengio, a homegrown AI superstar, now the co-founder of global software provider Element AI. But we shouldn’t worry too much about competing with artificial general intelligence. “There’s lots of basic research in front of us, and we have no clue whether it will take decades or centuries to reach anything close to human-level intelligence,” said Bengio.

Bengio ascribed the hype in part to the amount of money that’s being invested in this space. “As academics, we have to be careful to keep our ground,” he said. However, AI is so intrinsically complex that advancements in research necessarily tests the boundaries of research. “If we would limit ourselves to what we can understand analytically, the field would stop moving completely,” Bengio said.

Google Brain researcher Hugo Larochelle echoed this sentiment. Since the field is so experimental, some over-rating of results is inevitable, he observed. Larochelle emphasized that AI researchers need to take steps to ensure their experiments can be reproduced by others in their field, and answer the question, How do we collectively share the credit for what we do? “We need to shift the way we think of research, not as individual ideas but as goals that we set collectively,” he said.

In his closing remarks, CIFAR president and CEO Alan Bernstein also emphasized the need for shared effort. “In a country like Canada, it’s critical that we’re all on the same page,” he said. “We’re too small to pull in different directions.”

R$