Having been dubbed as ‘the Silicon Valley of AI’, Montreal is steaming ahead with startups, and tech giants with leading AI research coming out of the Canadian city. With pioneers such as Yoshua Bengio, Yann LeCun, and Geoffrey Hinton hailing from Canada, it was only a matter of time before Canada took the lead in the AI race.
At the centre of this hub is the Vector Institute, who ‘will drive excellence and leadership in Canada’s knowledge, creation, and use of artificial intelligence to foster economic growth and improve the lives of Canadians’. Having launched in March of this year, Vector Institute was a response to a significant opportunity to make Canada a global leader in artificial intelligence. Canada produces ‘some of the best and graduates in machine learning, reinforcement learning and deep learning’ under the supervision of Bengio, LeCun and Honton, and Vector Institute’s goal is to:
Increase awareness among companies in Canada, both large and small, of the transformative potential of AI. The Vector Institute are ‘here to drive excellence and leadership in Canada’s knowledge, creation, and use of machine learning to foster economic growth and improve the lives of Canadians.
The increasingly popular Deep Learning Summit will be taking place in Canada for the first time this October 10 & 11, where we will hear from Vector Institute. Richard Zemel, Co-Founder & Director of Research will be exploring the current success of deep neural networks and how they’ve largely come on classification based on datasets containing hundreds of examples from each category. He will explore the way in which humans can easily learn new words or classes of visual objects from very few examples. A fundamental question is how to adapt learning systems to accommodate new classes not seen in training, given only a few examples of each of these classes. I will discuss recent advances in this area, and present ongoing work by my group on various aspects of this problem.
With the recent explosion of AI research coming out of Canada, we were keen to hear from Richard first hand about his work and how he’s assisting Canada in its growth in AI. He explained that ‘an essential component of Vector’s mission is to attract global talent focused on research excellence; our researchers and academic partners will be part of a vibrant community of innovative problem-solvers, working across disciplines on both curiosity-driven and applied research. Vector is working with Canadian industry and public institutions, as well as innovation clusters and start-ups, to ensure that they have the people, skills and resources to harness the potential of deep learning and machine learning.'
Whilst research is advancing rapidly, talented minds are still drawn to Silicon Valley and Richard’s team are striving to build a foundation to make Canada a place where top talent and industry converge to create a vibrant and lasting ecosystem.
In creating the Vector Institute, it has been very important to me and the team to build flexibility into our operating model. The advantage of Vector being an independent, not-for-profit institution means that our researchers won’t necessarily be faced with having to decide between working in academia or working with industry. Vector researchers will be able to work on projects with the private sector – so long as their research and academic obligations are fulfilled. This flexibility, combined with growing the pool of top talent, lends itself to one of the Vector Institute’s core objectives, which is to enable companies in Canada to become best-in-class adopters of AI technology.
Recently, major companies such as Google Brain and Uber have expanded their research capabilities in Canada and Richard has had students and postdocs decide to stay in Canada because of the emerging AI ecosystem; while in the recent past these opportunities did not exist, they are very excited about the prospect of staying here to continue doing research and working with companies.
Upon asking Richard what developments we can expect to see in AI and deep learning in Canada over the next 5 years, he explained that many of the current successes have been in areas such as machine vision (automated driving), language (automated speech recognition in voice assistants, machine translation), and recommender systems. There are now other areas that are on the brink of making huge advances including healthcare and robotics. ‘All of these areas have a wealth of data, which makes them amenable to machine learning methods. Other areas where we may see progress in the near future include automated drug and manufacturing design, and education. Progress in each of these cases has been and will be fuelled by research advances, in learning algorithms, optimization, and hardware.’