The Canadian Institute for Advanced Research (CIFAR) is a Canadian-based global research organization that has played a central role in the development of modern Artificial intelligence. Founded in 1982, CIFAR supports collaborative, multidisciplinary research networks across the natural sciences, social sciences, and humanities. Its artificial intelligence initiatives, particularly the Learning in Machines & Brains program and the Pan-Canadian Artificial Intelligence Strategy, have connected leading researchers in Machine learning and Deep learning with institutions worldwide.
CIFAR's AI-related work emerged from its long-standing tradition of fostering cross-disciplinary collaboration. The organization's model brings together researchers from diverse fields to address fundamental questions, and its AI programs have become among the most influential in the field, linking academic research with the broader ecosystem of University of Toronto and other Canadian institutions.
Learning in Machines & Brains Program
The Learning in Machines & Brains program, established in 2004, was originally known as Neural Computation & Adaptive Perception. It was led by Geoffrey Hinton and brought together researchers including Yoshua Bengio and Yann LeCun, along with neuroscientists, computer scientists, biologists, electrical engineers, physicists, and psychologists. The program's collaborative structure allowed members to share ideas across disciplines, which proved critical for advancing Neural network research at a time when the approach was not widely accepted.
Members of the program developed computing systems that mimicked human intelligence, confirming the potential of neural networks for tasks such as pattern recognition and learning. This work laid the groundwork for many subsequent breakthroughs in Deep learning. In 2019, the Association for Computing Machinery recognized Hinton, Bengio, and LeCun with the 2018 ACM A.M. Turing Award for their conceptual and engineering contributions that made deep neural networks a critical component of computing.
The program continues to evolve, with research spanning topics such as Residual Network (ResNet) architectures, Multi-Head Attention mechanisms, and Loss Functions. Its alumni and affiliates have gone on to contribute to major AI developments, including Transformer (architecture) models and Large language model systems.
Pan-Canadian Artificial Intelligence Strategy
In 2017, the Government of Canada asked CIFAR to develop and lead the Pan-Canadian Artificial Intelligence Strategy, a national initiative to support AI research and talent. The federal government invested $35 million over five years in CIFAR's operations and announced $125 million for the strategy. The initiative aimed to position Canada as a global leader in AI by funding research, training new researchers, and connecting academia with industry.
The strategy supported the creation of AI institutes in major Canadian cities, including the Vector Institute in Toronto, Mila in Montreal, and Amii in Edmonton. These institutes have become hubs for research in areas such as Deep learning, Reinforcement learning, and Generative AI, attracting talent from around the world.
In 2022, CIFAR announced the second phase of the strategy, with up to $208 million in federal support over ten years. This phase focuses on translating research into applications, addressing ethical and societal implications, and maintaining Canada's competitive position in the global AI landscape.
Research Programs and Operations
As of 2023, CIFAR supports research in 15 major multidisciplinary areas, including the Learning in Machines & Brains program. Other programs cover topics such as Quantum Information Science, Brain, Mind & Consciousness, and Future Flourishing. The organization's staff supports more than 400 researchers from 21 countries and more than 140 institutions, with approximately half based in Canada.
CIFAR's governance includes a President and CEO responsible to the Chair and Board of Directors. Stephen Toope became President and CEO in November 2022, succeeding Alan Bernstein who served from 2012 to 2022. Irfhan Rawji chairs the Board of Directors, with Jacqueline Koerner and Anne McLellan as co-vice chairs. The organization receives funding from governments, partnerships, private sector sources, and investment income, with an annual budget of $30 million in 2018.
Historical Context and Impact
CIFAR was founded in 1982 following efforts by John Leyerle, a professor of English at the University of Toronto, who began rallying support in 1978. Fraser Mustard, a medical doctor and researcher in early childhood development, served as founding president from 1982 to 1996. The organization's early work included influential research on population health, with a 1994 publication identifying socio-economic status as the most influential determinant of health.
The AI programs built on CIFAR's history of supporting high-risk, high-reward research. The Neural Computation & Adaptive Perception program, which began in 2004, was initially considered unconventional, but its success helped legitimize Deep learning as a mainstream research area. The program's influence extended beyond academia, with many former members contributing to industry efforts at companies such as OpenAI, Google DeepMind, and Anthropic.
CIFAR's broader impact includes associations with 23 Nobel laureates since its inception, spanning fields from physics to economics. The organization's model of bringing together researchers from different disciplines has been credited with enabling breakthroughs that might not have occurred in traditional academic structures.
Notable People and Legacy
CIFAR has been led by a series of notable presidents, including Fraser Mustard (1982-1996), J. Stefan Dupré (1996-2000), Chaviva Hošek (2001-2012), and Alan Bernstein (2012-2022). The current president, as of 2023, is Stephen Toope. The organization's AI programs have been shaped by key figures such as Geoffrey Hinton, Yoshua Bengio, and Yann LeCun, who are widely acknowledged as pioneers of deep learning.
The legacy of CIFAR's AI research is evident in the widespread adoption of techniques developed within its programs. Concepts such as Batch Normalization, Dropout, and Layer Normalization are now standard in Machine learning practice. The organization's emphasis on collaboration between neuroscience and computer science has influenced how researchers approach problems in Artificial intelligence, from Computer vision to natural language processing.
CIFAR continues to shape the AI landscape through its research programs and national strategy, supporting work that addresses both technical challenges and societal implications of Generative AI and other emerging technologies.