69���˵���

The University of Toronto’s Geoffrey Hinton took to the stage at the Swedish Academy in Stockholm this weekend – two days before he officially accepts his Nobel Prize in Physics for fundamental work in AI – to deliver a lecture on the inventions and discoveries that led to him being given the prestigious award.

“Today, I’m going to do something very foolish – I’m going to try and describe a complicated technical idea for a general audience, without using any equations,” said Hinton, a U of T emeritus of computer science, prompting laughter from the audience. 

The “godfather of AI” then proceeded to outline how decades of his fundamental research, and that of his co-laureate John J. Hopfield of Princeton University, enabled the development of artificial neural networks and machine learning – technologies that underpin today’s AI revolution.

The Nobel lectures are among the highlights of , which runs from Dec. 6-12 in Stockholm and Oslo and includes award ceremonies, banquets, media engagements and commemorations at the Nobel Museum. There is also a series of Nobel Week events taking place at U of T, including watch parties on all three campuses for the livestream of the Dec. 10 award ceremony. 

Sunday’s first Nobel Prize lecture in physics was delivered by Hopfield, who shared how his fascination with the workings of the human brain inspired his development of the Hopfield network – an associative memory that can store and reconstruct patterns in data. 

“How mind emerges from brain is, to me, the deepest question posed by our humanity,” Hopfield said.

When it was Hinton’s turn to take the stage, he described how he and Terry Sejnowski – one of Hopfield’s students – came upon a novel use of Hopfield nets: “Instead of using them to store memories, we could use them to construct interpretations of sensory input,” Hinton said.

He then went on to discuss the resulting Boltzmann machine, a type of neural network that is capable of recognizing elements within data. 

Yet, despite its promise, the original Boltzmann machine was too slow, Hinton said, and it wasn’t until several years later that he came up with “restricted Boltzmann machines,” which impose limitations on connections between neurons in order to increase system efficiency – a development that would prove pivotal in training deep neural networks (Hinton donated an early Boltzmann chip, about the size of a postage stamp, to the Nobel Prize Museum). 

Following the lecture, Hinton was joined on stage by Hopfield, with the pair sharing a vigorous handshake and posing for photos.

Separately, U of T convened an expert panel in Stockholm on Monday about the direction of AI research and development.

Moderated by Leah Cowen, U of T’s vice-president, research and innovation, and strategic initiatives, the panel included: Eyal de Lara, a professor and chair of the department of computer science in U of T’s Faculty of Arts & Science; David Lie, director of the  and a professor in the Edward S. Rogers Sr. department of electrical and computer engineering in the Faculty of Applied Science & Engineering; Tony Gaffney, president and CEO of the ; and Amy Loutfi, professor of computer science and pro-vice-chancellor for AI at Örebro University in Sweden.

The conversation touched on areas including the promising applications of AI, how responsible deployment of AI can mitigate the technology’s potential pitfalls and implications of AI’s rise on education.

U of T President Meric Gertler said that the AI breakthroughs fostered by Hinton’s research were made possible by Canada’s longstanding support of basic research.

 “Geoff was interested in the novel but unproven concept of artificial neural networks, an area that was sometimes described as the ‘unpromising backwater’ of AI research,” President Gertler said in his remarks introducing the panel, noting that Hinton joined U of T in 1987 and was one of the first scholars to receive support from the  (CIFAR).

“Canada was investing in brilliant people, their ideas and their students – and those investments have paid off many years later.”

Canada was also the first country to launch a national AI roadmap, President Gertler said, in the form of the Pan-Canadian Artificial Intelligence Strategy – which funds three national AI institutes including the Vector Institute for Artificial Intelligence, which is now housed in U of T’s new Schwartz Reisman Innovation Campus. The state-of-the-art building also hosts the Schwartz Reisman Institute for Technology and Society, which is at the forefront of research and thought leadership on AI safety and responsible development, with Hinton one of its advisory board members.

“In short, Canada has played a key role in launching and driving the AI revolution and we’re a world leader in understanding and promoting safe, human-centred AI,” President Gertler said. 

The theme of responsible AI was also brought up during a Q-and-A with Hinton, who revealed that the remarkable information-sharing abilities of large language models played a big role in sparking his now oft-repeated concerns around the current pace of the technology’s development. 

“That’s when I came to realize that the fact that they’re so much better at sharing probably means that digital intelligence is just a better form of intelligence than us – and that’s what got me so worried,” he said. 

Asked what responsible AI regulation looks like, Hinton said there must first be a consensus around solving the problem from a scientific standpoint – not unlike the scientific consensus that has emerged around climate change.

“Like the early days of climate change, the first thing to do is figure out what’s causing it and get scientific agreement on how you can fix it. Then, the second thing to do is get the politicians to do something about it … but here, we haven’t finished the first thing yet.”