Q&A with Doina Precup

DeepMind , the AI company famous for its Go-playing AlphaGo program, is setting up an office in Montreal, making it the latest high-profile company to join the city’s artificial-intelligence research community. Doina Precup, an Associate Professor in McGill’s School of Computer Science, will lead DeepMind’s office here, while continuing to teach and do research at the University. We spoke with Prof. Precup about her new role at DeepMind, her ongoing work at McGill, and the rapidly evolving AI scene in Montreal.

My portfolio within DeepMind will be under fundamental research. It’s going to be research in reinforcement learning, deep learning and the intersection of the two. What I work on are algorithms that take inspiration from natural intelligence - so that automated programs can learn not just under supervision, but by doing experiments and receiving feedback from their environment on those experiments.

DeepMind’s mission is to solve intelligence and use that to tackle some of the world’s most pressing and complex problems, and the hope is that these general purpose learning techniques could, in the future, be used by scientists and researchers to generate insights in a multitude of areas, from superconductor material design to drug discovery.

I’ve known them for a very long time, since they were a very small startup. Very early on in their development as a company they came to one of the workshops we organized at the Bellairs Institute, and it was a really fun experience. And so I followed the research that they do, and I went there and visited once. There is a strong "Canadian connection" as well, in the sense that many of their researchers were trained in Canada, including several graduates from our lab, and they also recently opened a research lab in Edmonton with UAlberta. It’s great that DeepMind’s research accomplishments have helped focus the world’s attention on AI and propelled new scientific discoveries, and the company’s mission is perfectly aligned with my own research work and goals. I am really excited to join forces with DeepMind and to help build the new Montreal team.

This should create some really exciting opportunities for students. It is very likely that some of my students will do internships at DeepMind, then come back to McGill. And some DeepMind employees may want to have some kind of adjunct teaching position at McGill. We already have some people from Google Brain, another research group that’s here in Montreal, who have just gone through the process at the department level for becoming adjuncts. So it sort of depends on who we hire at DeepMind, but I can imagine that for some people this would be an attractive option, because it gives them the opportunity to teach the next generation of researchers, collaborate with graduate students and supervise some students in the academic context.

The ones that I’ve talked to are very excited and they see the internship opportunities. DeepMind in the community is perceived as the flagship industry research lab, because in our field - in reinforcement learning - they have many top-notch researchers. I think DeepMind Montreal will help to keep more brilliant researchers in the Canadian research ecosystem and also attract others from elsewhere.

Some people say Canada, to benefit from AI, really needs to be developing its own companies - not just attracting investments by international companies. What do you think?

I think we need to do our best to have both. A healthy local ecosystem relies on large companies, research startups like Element AI or RBC Research, incubators and academia coming together, so there’s a place and a role for everyone. I think we have an obligation, from the side of the university, to support this ecosystem, partly by training our very talented students. I view the investments by international companies in Canada as recognition of the value of the talent and quality of the research here, and it’s great that researchers who would otherwise have left are now going to stay in Canada, in this really vibrant AI community. For me, this can only be a good thing.

Exactly. There are going to be people who will come and work for these companies and then want to make their own ventures. There is so much choice and mobility within Montreal that people will come and won’t want to move anywhere else. This is one of the big playing cards of Silicon Valley: all the companies are there, right? So if you’re a computer scientist, if you go to the Valley, you know there are plenty of choices. So I think that’s the kind of ecosystem we’d be aiming to have here, around AI and machine learning.