Some of Canada’s artificial intelligence (AI) leaders shared their vision for the future—and how we might get there—at the Creative Destruction Lab (CDL) Super Session yesterday during Toronto Tech Week.
Speakers at CDL’s annual event included reinforcement learning (RL) pioneer and recent Turing Award winner Richard Sutton, fellow Canadian Institute for Advanced Research (CIFAR) AI chair Patrick Pilarski, and Kindred founder, Sanctuary AI co-founder and Nirvanic founder and CEO Suzanne Gildert, among others.
“We’re entering the era of experience.”
Richard Sutton
Sutton kicked things off by explaining to a packed room at the University of Toronto’s Rotman School of Management why he believes the world is nearing “the end of the era of human data.”
The University of Alberta computer science professor, who also works as fellow and chief scientific advisor at the Alberta Machine Intelligence Institute (Amii), noted that AI is beginning to exhaust the amount of human-generated text and images available to it as training data, echoing his recent remarks elsewhere.
As companies have already scraped most of the high-quality data on the internet for AI training, Sutton argued that to advance further, the tech requires a new source of data that grows and improves as agents—or AI systems that are empowered to take actions—become more powerful. He thinks RL could hold the answer: Sutton believes that such data could be generated from an agent’s first-person interaction with the world, or its “experience.”
“We’re entering the era of experience, I think,” Sutton said. “In that era, we have a single goal—reward—but we need data, we need curiosity.”
Pilarski, who also works at Amii and the University of Alberta, but as a professor of medicine, expanded on the theme of experience with some concrete examples. He and some of his colleagues want to enable “a new class of bionic technologies” to create limb prosthetics that more deeply integrate humans and machines.
“Our mythical beast we’ve been pursuing … is to try to figure out how to really transform the science [and] the art of prosthetic restoration [and] all kinds of neuroprosthetic technologies,” Pilarski said.
RELATED: Richard Sutton warns against “centralized control” of AI regulation based on fear
Pilarski noted that the world has entered an era where a variety of technologies are “interacting with and connecting tightly” with human bodies, from cellphones to smart watches, fitness wearables, bionic body parts, pacemakers, and cochlear implants. He argued that “we’re beginning to have the capacity to truly build what we might consider as an exocortex,” or an external machine processing unit for the human body and mind that extends, amplifies, and augments human abilities to perceive the world, make decisions, and act upon it.
Pilarski showcased some videos of the work his lab has been doing in areas such as bone anchoring, revising the nerves of the body to better connect with robotics, and agent as interface.
This included a clip of a man who lost his lower leg, and had connected a robotic replacement with an implant bolted directly into his bone that he could easily use to attach and remove his prosthetic. Pilarski claimed this approach allows “direct, immediate coupling,” makes a wider range of physical activities possible, and allows for the subjects to begin to feel by way of their skeleton.
He also shared a video of a man who had lost part of his arm competing in a 2024 competition using a robotic, AI-powered replacement.
Gildert’s remarks were more speculative. She argued that the human mind has two main modes of operation—autonomous decision-making and conscious decision-making—whereas AI only makes decisions autonomously. Gildert argued that in order for AI to handle the real world, a new theoretical framework is required.
She used her remarks to walk attendees through one possible option: quantum consciousness theory—the idea that there may be quantum effects going on inside our brains, and that ”these, in some way, may contribute to our consciousness”—a notion she is exploring through her work with Nirvanic.
Gildert acknowledged that this sounds like a “really weird, new, wacky” idea, but claimed it is actually an old premise that dates back to when the theory of quantum mechanics was developed in the 1920s.
Until recently, Gildert said quantum consciousness was merely the subject of philosophical speculation. What excites Gildert is the idea that this theory can finally be put to the test now that quantum computers exist.
“It’s early days, but this is a really exciting area of research, because if we can understand conscious decision-making—the second mode of operation that AI is completely ignoring for now—I think we’ll be able to enable, finally enable, AI in the physical world.”
BetaKit is the official media partner of Toronto Tech Week. Feature image courtesy Creative Destruction Lab via X.