In 2004 the Canadian Institute for Advanced Research (CIFAR) launched a research program called Neural Computation and Adaptive Perception (NCAP), with Geoffrey Hinton as its director. CIFAR’s own materials record that Hinton directed the program from 2004 until 2013 and that the program, now renamed Learning in Machines & Brains, was founded in 2004 and has been renewed repeatedly since. This is the institutional piece of the deep-learning origin story that is often left out of the usual narrative.
The timing is what makes the program important. In the early 2000s neural networks were deeply unfashionable. Most of the machine-learning field had moved toward other methods such as support vector machines, and funding and prestige for neural-network research were scarce, a period later described as a neural-network winter. CIFAR’s model was unusual: rather than funding individual grants for specific deliverables, it funded a small, hand-picked network of researchers to think and collaborate over the long term, with regular workshops that let them share unfinished ideas.
The program convened the people who would later define the field. It brought together the groups led by Hinton at the University of Toronto, Yoshua Bengio at the University of Montreal, and Yann LeCun at New York University, the three researchers who would jointly receive the 2018 Turing Award for deep learning. The collaboration and the program’s summer schools, the first held in 2005, helped train a generation of students and kept the research community coherent during the years when few others were paying attention.
The payoff came soon after. The 2006 work on deep belief networks, the breakthrough usually credited with reviving deep neural networks, emerged from this same circle of CIFAR-supported researchers. The entry is a reminder that the deep-learning revival was not only a matter of better algorithms, more data, and faster GPUs; it also depended on patient institutional support that kept an unfashionable idea alive long enough to succeed. CIFAR’s role was to fund and connect people through a lean period, not to build the technology itself.