The Quebec Artificial Intelligence Institute in Montreal, known as Mila, is one of the world’s largest academic research centres for machine learning. Its more than 80 faculty members, 600 students and 70 industry partners are focused on building algorithms that can learn autonomously without being explicitly programmed.
In 2020, Mila’s members attracted approximately Can$2 million (US$1.6 million) in industry sponsorship from partners such as Google, in addition to Can$11 million in government grants. Its strength is in deep learning, a machine-learning technique in which algorithms trawl through unstructured data, such as images, and learn to interpret them using a layered system that mimics processes in the human brain.
The deep-learning approach has become prevalent in AI, and is used in applications such as voice recognition and Internet searches. In 2018, it won Mila founder and scientific director, Yoshua Bengio, the most prestigious prize in computer science, the A. M. Turing award, shared with Geoffrey Hinton, chief scientific adviser of the Vector Institute in Toronto and an engineering fellow at Google, and Yann LeCun, who holds dual positions at Facebook and New York University.
“Mila has always been one of the places with the greatest critical mass of deep-learning people, and it has snowballed,” says Bengio, a pioneer of the field.
Mila is run as a non-profit partnership between McGill University and the University of Montreal, ranked second and seventh, respectively, among Canadian institutions in the Nature Index in 2020, based on output in 82 selected natural-sciences journals.
Researchers in Canada have benefited from intensive support for fundamental AI research, and its Pan-Canadian Artificial Intelligence Strategy. In 2017, the federal government invested Can$125 million in this strategy over five years to cultivate networks of AI researchers across the country, establishing three hubs: Mila in Montreal, the Vector Institute in Toronto, and the Alberta Machine Intelligence Institute in Edmonton. The government renewed the commitment in April 2021 with an additional Can$443.8 million in funding over ten years.
With a strong focus on the ethical use of AI, many of Mila’s projects and subsequent spin-off companies serve as examples for how such technologies can be used for good. These include Korbit AI, which provides a personalized tutor to teach data science, and Ubenwa, which analyses the cries of babies to help detect signs of potentially dangerous low oxygen levels. Many Mila researchers switched their focus during the pandemic to see how AI might help in the fight against COVID-19, says Bengio, by seeking novel drug candidates, for example.
Key challenges remain in building algorithms that rival the human intelligence systems they mimic, says Bengio, such as giving AI the ability to form abstract imagination and teaching it to understand the difference between causation and coincidence. “Some think it’s just a matter of scaling up to meet the intelligence of humans,” says Bengio. “But I think we’re missing something. That’s what keeps us busy.”
The institute continues to attract top-class researchers, including Irina Rish, who in 2019 left her position as a research scientist at IBM’s T. J. Watson Research Center in New York for an appointment at the University of Montreal and its affiliation with Mila. She came for the energy and imagination of its students and researchers, she says, along with a position as a Canadian Excellence Research Chair at the University of Montreal, which brings funding of Can$10 million over seven years.
“It allows people to focus more on research and less on chasing grant money,” says Rish. Managing a team more than 30 students and plugging into the Mila ‘hive mind’ has felt like “acceleration from zero to a hundred miles per hour in a few seconds”, she says. “It’s quite a ride.”