This text is part of the special section 100 years of Acfas
The invention of the Internet at the end of the XXe century has transformed our lives. It also allowed another technological revolution to accelerate in the 2010s: artificial intelligence (AI). In Montreal, a rich ecosystem of researchers is forming around one of the fathers of deep learning, Yoshua Bengio. The AI story is just beginning.
Artificial intelligence (AI) as an autonomous field of research was born in 1956, at the Dartmouth Conference. But the craze that followed in the 1960s gradually gave way to a lean period. “The promises of the 1970s and 1980s were not kept, and the funding for artificial intelligence was greatly reduced,” says Maxime Colleret, science historian and doctoral student at the University of Quebec in Montreal. Researchers interested in these technologies therefore do not have the data necessary to make the algorithms work.
A new craze
A favorable wind was blowing again on AI in the early 2010s. “The computing power of computers has improved a lot, says Maxime Colleret. In addition, the Internet and social networks generate large amounts of data on which algorithms can be trained. »
At Université de Montréal, Yoshua Bengio, a professor in the Department of Computer Science and Operations Research, is working with a team of researchers on deep learning, which takes inspiration from the human brain to teach machines how to learn. “Around 2011-2012, these methods began to see performance breakthroughs for voice recognition and object recognition, which aroused great interest in the world of research, but also from industry,” explains Simon Lacoste. -Julien, professor in the same department.
Artificial intelligence is starting to attract investors again. “We promise that it will have major socio-economic benefits. Some even predict that it will transform society as a whole and automate intellectual work,” says Maxime Colleret. Predictions “hyperbolic, as often during major technological breakthroughs”, according to the historian of science. The Canadian and Quebec governments reacted by injecting significant capital.
Mila: a crossroads of talents
In 2017, $100 million was released to support the development of AI in Quebec and the Steering Committee for the Artificial Intelligence Cluster (COGIA) was set up to guide investments. In 2018, he recommended that the government invest significant public funds to support the creation and operation of a pivotal organization between academia and industry: Mila (the Quebec Institute of Artificial Intelligence). This institute founded by Yoshua Bengio at the University of Montreal in 1993 previously bore the name of Lisa (Laboratory of Computer Science for Adaptive Systems). “Yoshua Bengio wanted to create a kind of artificial intelligence ‘Silicon Mountain’ in Montreal, in order to continue to build a critical mass of researchers and attract talent,” says Simon Lacoste-Julien, who joined his team in 2016. Other organizations created since 2015 (IVADO, IVADO Labs and Scale AI) also receive public funds.
Funding and the Montreal ecosystem are attracting giants like Google, Facebook, Samsung and Microsoft, which are opening research laboratories in the metropolis. Former researchers from Yoshua Bengio’s team, such as Hugo Larochelle (at Google Brain), are returning to Montreal after working abroad. Students too. “We are in the second or third generation formed by Mila, who is starting to train other people,” rejoices Sasha Luccioni, researcher in artificial intelligence.
Ethics at the forefront
Several great researchers who worked in the field of artificial intelligence joined the industry during the 2010s, such as the British Geoffrey Hinton (Google) or the French Yann LeCun (Facebook). “Yoshua Bengio chose not to follow the siren song and to stay at the Université de Montréal to continue to grow a university ecosystem,” said Simon Lacoste-Julien. He is also interested in the ethical aspects of AI. Playing a central role in the development and dissemination of the Montreal Declaration for the Responsible Development of Artificial Intelligence, the preliminary version of which was released in 2017, he “surrounds himself with people who have the same vision, and that creates a critical mass around ethics”, explains Sasha Luccioni. A concept engraved in the values of Mila. “It helps us remember to always include it in our research,” says Simon Lacoste-Julien.
The institute has also just integrated a new training in ethics, the TRAIL program, into the course of its students. This innovation will fill an important gap, according to Sasha Luccioni. “Computing is generally thought of as computer-centric, not society-centric. Until now, master’s or doctoral students in AI were not trained in it,” says the researcher.
However, problems abound, such as the biases created or amplified by algorithms or virtual assistants (chatbots in English) not declared on the Web. “These have become so efficient that users can confuse them with humans,” warns Simon Lacoste-Julien. However, too often, researchers from computer science “are not really aware of the problems posed by these tools”, he observes. There is nothing artificial about training them for these challenges.