Mila – Quebec AI Institute (originally Montreal Institute for Learning Algorithms) is a research institute in Montreal, Quebec, focusing mainly on Machine learning research. As of 2022, approximately 1000 students and researchers and 100 faculty members were part of Mila. Along with Alberta's Amii and Toronto's Vector Institute, Mila is part of the Pan-Canadian Artificial Intelligence Strategy.
Mila is both a research and educational institution, providing postdoc and internship positions and a research environment for students enrolled in their respective universities working towards graduate degrees such as PhD and Master's in machine learning. The institute is known for its contributions to Deep learning and Reinforcement learning, as well as its open-source software and partnerships with private companies.
History
Mila was founded in 1993 by Université de Montréal professor Yoshua Bengio, a Turing Award winner and one of the founders of the field of deep learning. It was originally named the Laboratoire d'informatique des systèmes adaptatifs (LISA). Mila also traces its history to the Reasoning and Learning Lab (RL-Lab) at McGill University, founded in 2001 and currently co-directed by Professors Prakash Panangaden, Doina Precup, Joëlle Pineau, and Jackie Chi Kit Cheung.
In 2017, Mila was newly established as a partnership between the Université de Montréal and McGill University, along with École Polytechnique de Montréal and HEC Montréal. In December 2020, Mila teamed up with IBM to accelerate AI and machine learning research using open-source technology, integrating Mila's open-source software Oríon with IBM's Watson Machine Learning Accelerator.
Notable Researchers
Besides Bengio, Mila's faculty members include Joëlle Pineau and Doina Precup. Some of its graduates include Marc Bellemare (MSc 2007 – RL-Lab), Hugo Larochelle (PhD 2009 – LISA), and Ian Goodfellow (PhD 2014 – LISA). These researchers have gone on to influential roles in academia and industry, contributing to advances in Neural network architectures and Generative AI.
Research Activities
Research at Mila is largely, but not exclusively, focused on deep learning and reinforcement learning. Specific research topics include generative models, natural language processing, meta learning, computer vision, reinforcement learning, and applications of AI such as climate change research and COVID-19 research. The institute also explores Artificial general intelligence, superintelligence, AI Alignment, AGI governance, and frontier models.
Mila's research has influenced the development of Transformer (architecture) models and Large language models, with connections to organizations like OpenAI and Google DeepMind. The institute's work on Residual Network (ResNet)s and Batch Normalization has been foundational to modern deep learning.
Board of Directors
The Board of Directors oversees Mila's strategic directions. Members include Alexandre Le Bouthillier (General Partner – Linearis), Anne-Marie Hubert (Managing Partner – Ernst & Young), Benoit Boulet (Associate Dean, Research & Innovation – McGill University), and Yoshua Bengio (Observer, Scientific Director, Mila & IVADO). Other members represent partner institutions and industry, ensuring diverse governance.
Other Activities
Mila members have contributed to open-source software. Theano, one of the early programming frameworks for deep learning, originated at Mila. In 2020, active projects included myia, a deep learning framework for Python, and baby-ai, a platform for simulating language learning with a human in the loop. Mila is also involved in numerous partnerships with private companies, fostering technology transfer and applied research.
Mila's educational mission extends to hosting postdoctoral fellows and interns, and its research environment supports students from various universities. The institute's collaborative culture and interdisciplinary approach have made it a global hub for machine learning research.