Mila – Quebec AI Institute

What it is

Mila, the Quebec AI Institute, is a deep-learning research institute in Montreal, Quebec, Canada. It was founded in 1993 by Université de Montréal professor Yoshua Bengio, and it was originally called the Laboratoire d'informatique des systèmes adaptatifs (LISA).^[1] Valérie Pisano is listed as its president and CEO.^[1]

Origins

Mila traces its history to LISA, founded in 1993, and also to the Reasoning and Learning Lab (RL-Lab) at McGill University, founded in 2001.^[1] In 2017 the institute was established as a partnership between the Université de Montréal and McGill University, with École Polytechnique de Montréal and HEC Montréal also involved.^[1][3]

Bengio was its scientific director until 2025, and he is now its founder and scientific advisor.^[2]

Known for

The lab's early work included Theano, an early deep-learning framework, which originated at Mila.^[1] The lab has also trained researchers who went on to well-known work. Hugo Larochelle received his PhD in 2009 from LISA, Ian Goodfellow received his PhD in 2014 from LISA, and Marc Bellemare earned an MSc in 2007 from the RL-Lab.^[1]

Two 2014 papers show the range of the research done in Montreal. The paper “Generative Adversarial Networks” by Ian Goodfellow, Yoshua Bengio and colleagues was written at the Université de Montréal.^[4] A generative adversarial network trains two neural networks against each other: one produces samples, and the other tries to tell those samples apart from real data. The paper “Neural Machine Translation by Jointly Learning to Align and Translate”, by Dzmitry Bahdanau, Kyunghyun Cho and Yoshua Bengio, introduced the attention mechanism for neural machine translation, and it came from the Université de Montréal.^[5] Attention lets a model weigh different parts of its input when it produces each part of its output.

Both papers came from the university where Mila's founder works, which is why they sit naturally beside the institute's history.

Place in AI history

Mila is one of the three national AI institutes of the Pan-Canadian AI Strategy, with Toronto's Vector Institute and Alberta's Amii.^[6] In 2022 it had approximately 1,000 students and researchers and 100 faculty members.^[1]

Mila therefore links early neural-network research in Montreal, the lab's origins in 1993 and its partnership with McGill, to the 2014 papers on generative adversarial networks and attention. For readers, it is a clear case of a regional research centre that became part of a national AI strategy and kept its own research identity.