OTTAWA — British-Canadian computer scientist Geoffrey Hinton’s Nobel Prize is a marker of how far artificial intelligence has come, from a backwater to today’s booming sector.

“He is rightfully credited as the creator of the modern field of AI,” said Nick Frosst, co-founder of Cohere, a Toronto-based startup whose AI models build on the neural-networks approach Hinton advanced. 

On Tuesday, the Royal Swedish Academy of Sciences awarded the Nobel Prize in physics to Hinton and Princeton University professor John Hopfield for “foundational discoveries” in machine learning. 

For more than two decades at the University of Toronto, Hinton and his collaborators made crucial breakthroughs in the technology’s capabilities, while training a generation of influential researchers and corporate leaders. He’s now one of the most prominent voices warning of AI’s potential dangers, including what he sees as existential risks. 

Hinton won the Nobel Prize for work he began at Pittsburgh’s Carnegie Mellon University in the 1980s on a type of neural network called Boltzmann machines. While earlier AI systems could recognize information on which they had already been trained—a handwritten “3,” for example—his new version could handle new examples that were similar but not the same—a new kind of “3.” 

But Boltzmann machines required a lot of processing power to train. “The algorithm was ahead of its time,” said Graham Taylor, who got his PhD under Hinton and now leads the University of Guelph’s machine learning group. 

Hinton moved to U of T in July 1987 and worked with other researchers to tweak and refine the technology. At the time, other computer scientists scoffed at their approach—but the arrival of new chips in the late 2000s delivered a neural network breakthrough.

The models for which the Nobel committee recognized Hinton and Hopfield are no longer in widespread use, Taylor said. But some of the advancements they made—like Boltzmann machines’ ability to store knowledge that they haven’t explicitly been taught—are crucial to modern AI systems such as the ones that power ChatGPT.

Canadian AI leaders say the Nobel Prize win is testament to Hinton’s persistent belief in neural networks. “It took him and his students 25 years to demonstrate that the theory could do something useful” while “fighting the doubters and the naysayers,” said Tomi Poutanen, CEO of Signal1, a Toronto-based health-care AI startup. Poutanen studied under Hinton at U of T, and the two were among the co-founders of the Vector Institute in June 2016.

In Canada, Hinton found “an environment where he could continue to advance research that was really ahead of its time,” said Elissa Strome, executive director for the federal AI strategy at research non-profit CIFAR. Beyond technological advancements, she noted his work training students.

Frosst took Hinton’s undergraduate neural networks course at U of T, then went on to work for him at Google’s AI lab. “I pretty much learned how to do science from him,” said Frosst, noting that Hinton has also been willing to work with researchers who don’t have graduate degrees, or come from fields other than computer science.

The Nobel isn’t Hinton’s first major recognition. In March 2019, he won the Turing Award—which computer scientists often consider their version of a Nobel—along with long-standing collaborators Yoshua Bengio of Université de Montréal and Yann LeCun of New York University. 

While Hinton and Hopfield don’t work directly on physics research, Taylor said the models they developed use principles also found in areas like thermodynamics. Hinton has long drawn from different fields in his AI work. “Physical systems are hugely influential on his thinking about neural nets,” said Frosst, recalling that at Google the two would discuss algorithms using analogies to the physics of elastic bands and balls suspended on fabric.

Following the launch of ChatGPT, Hinton left Google to focus on pushing the technology industry and policymakers to address AI’s bad side effects. Along with Bengio, he’s among a number of prominent researchers warning it might imperil humanity. “He has a different platform now around the AI safety issues,” Taylor noted. The Nobel Prize “will give him an opportunity to talk about what he feels is very important.” 

Other researchers have called Hinton’s focus on existential risk a distraction on the ways AI is already being used to harm people, including discriminating against members of marginalized communities. On Tuesday, former Google researcher Timnit Gebru criticized the Nobel award.

Hinton has also said AI progress could provide benefits in fields like health care. “While attuned to the vast potential of AI, [Hinton] recognizes that we must be responsible stewards of this powerful technology,” said Aaron Brindle, a partner at Radical Ventures, a Toronto-based capital firm that Hinton has backed.

No one expects the Nobel Prize to slow Hinton down. While “there aren’t more awards to win,” Frosst predicted Hinton will continue to “think about the long-term future [and] to add his intellect and experience to that conversation.”