Canada has been the spawning ground for the most important innovation of the 21st century. It’s weird. Toronto and Montreal were the birthplaces of artificial intelligence. Not Silicon Valley, not Beijing, not London, not Paris, not New York. Back in 2012, a Canadian team—including the AI pioneer Geoffrey Hinton—entered the ImageNet challenge, a contest that evaluated computer algorithms’ abilities to detect objects and classify images at scale. Hinton’s team outdistanced the competition by such a wide margin that every other approach to computer vision became practically obsolete. They used neural networks, a system of machine learning that imitates how the human brain recognizes patterns and, for decades, was widely considered an eccentric, if not outright bizarre, approach to computation. That victory was the moment that a bunch of prophets in the wilderness moved from conceiving mindlike algorithms to creating them. It’s an unusual position for Canada to find itself in—one of maximum importance, central to the future.
During the past 10 years, the years between ImageNet and ChatGPT, artificial intelligence lurked in a half-lit, half-shadow realm of radical but largely unnoticed progress. It was a period of promise and hype, a time of real miracles and many more claims to miracles. Rumours abounded. Some claimed that China had generated a sentient, 174-trillion-parameter large language model. Engineers fell in love with algorithms and hired lawyers to advocate for their personhood. Artificial intelligence was supposed to replace whole aspects of life—driving, radiology, voice-over narration—and either failed, or succeeded in such a limited way that the dream of true automation seemed to remain permanently a dream. Other aspects of life that no one could imagine a machine being able to approach, never mind perform—resuscitating the speech patterns of dead loved ones, for example—AI somehow managed to achieve. The explanations for the successes and the failures were either impossibly technical or just plain unsatisfying. And so, during this interval between computer scientists uncovering the potential of AI and the public understanding the potential of AI, it was possible—it was common, even—to forget about the technology, to ignore its fascinating, terrifying potential, to brush it off as a fad, another overhyped tech-lord plaything of the moment, like crypto or smart cities.
Such wilful ignorance is no longer possible. By now, the power of artificial intelligence has become clear to all but the most underground or resistant. ChatGPT holds the record for fastest-growing consumer application of all time, attaining 100 million active monthly users two months after launch, a number Instagram took two and a half years to reach. Venture capitalists invested five times more in artificial intelligence in the first quarter of 2023 than they did in the same period in 2022.
The miracles and near miracles have continued but the consequences are already world-altering. Industries will be overturned. Federal researchers at the Oak Ridge National Laboratory in the United States have recently predicted the end of software development as such: “The combination of machine learning, artificial intelligence, natural language processing and code generation technologies will improve in such a way that machines, instead of humans, will write most of their own code by 2040.” Software is only the most obvious target for the new technology. ChatGPT can come up with a pretty good marketing strategy if you give it the right information. It can develop travel itineraries, specifically tailored to your preferences. It can prepare legally appropriate contracts. The robots that came for the factory workers in the late 20th century are coming for the clean jobs in the early 21st.
The most profound changes, as always, are the quotidian. Already, artificial intelligence is altering the fabric of writing and thinking on the most basic level. The way ordinary people, especially young people, approach questions has changed over the past year, whether they’re asking about Shakespeare’s use of imagery in Othello or where to find the best duck confit in Montreal. ChatGPT can provide not just a series of links to facts and narratives, like a Google search, but a coherent, integrated response. That coherence isn’t necessarily accurate, but neither was Google. I refer to Google in the past tense because the time of Google as the single, unified pipeline of global information has passed. That’s just one small change. There are many others. The title “prompt engineer” was meaningful only to a minuscule circle of AI art weirdos a year ago. Now there are university courses devoted to this career.
AI works on such a foundational level that its effects are harder to predict than those of other technologies. Books and movies haven’t helped. The fantasies of artificial intelligence, as conceived for 200 years since the Industrial Revolution, have remained startlingly consistent: machines replace or control people. Metropolis, Her, The Terminator and various Spielberg movies have been popular because the artificial people and artificial gods they imagine play so conveniently into the religious fantasies established for us already: the creation myths and apocalypses that humanity seems to need. But the changes to everyday thinking are more important than fanciful robots, even though they seem less tangible. With ChatGPT, if you have the nub of an idea or the direction an idea might go, you can ask the software to flesh out what you have and it will round out your thinking. You can’t put that kind of thing in science fiction. It’s hard to understand even when you’re using it.
The fantasies aren’t the only haze that has settled over artificial intelligence and obscured its future. There is also the mystery inherent to the technology. Researchers themselves do not know how or why artificial intelligence works the way it does. Deep learning, which is at the core of so much AI, is an opaque process. It doesn’t explain the reasoning behind its answer; it only gives the answer. The essence of the power of artificial intelligence is exactly that we don’t understand it. Any outside human interference in the process tends only to hamper it. The process works because the computation is left alone to foster and expand its own intuition.
The incomprehension at the centre of artificial intelligence is much more frightening than any of the fantasies about the end of the world that books and movies have created. The power of AI tools is literally beyond our grasp. AI weapons, for example, are terrifying exactly because they disrupt the careful balance of human knowledge involved in any conflict. Military strategy has, up till now, operated through a balance between measure and countermeasure. AI weapons are distinct from all previous weapons, not necessarily in their destructive force but in the intellectual underpinnings of that force. Military strategy is largely based on deterrence: if the enemy uses that, then we have this. If we use this, they will use that. How do we plan for weapons whose effects neither side understands? The whole point of artificial intelligence is that it can do things you can’t imagine. If you train it to maximize destruction, it will create, by its nature, unimaginable horrors.
An understandable response to the unimaginable is to wish the whole thing away. If artificial intelligence must exist, it can at least be regulated to the point at which it may as well not exist. But the answer is surely not simply to sideline AI. Its most fruitful applications tend to be the wildest: what are the problems the human brain can’t solve? AI holds the possibility of answering intellectual questions that are, for one reason or another, ungraspable, which means that the potential of this technology is as large as our misunderstanding. The most impressive feat achieved by artificial intelligence so far has been the protein folding algorithm developed by Google DeepMind’s AlphaFold. Protein folding, the process by which protein chains take on three-dimensional forms that give them specific biological functionality, is simply too large a problem for human brains to comprehend, but AI can do it flawlessly. AI has also been developed for more prosaic but incredibly important functions like accelerated MRI readings. If the horrors of AI are unimaginable, so are the benefits.
The inconceivable advantages and the unimaginable dangers mean that AI serves as a kind of policy Rorschach test. In the face of a fundamental mystery, countries react by instinct. China uses AI for state control of the population. Tech companies in the United States turn it into profitable products, covered over with shallow talk about “safety” and “responsibility” to disguise the greed at the core. As for Canada, our fundamental impulse is to regulate righteously. It’s not the worst instinct in the world, but it can be self-sabotaging.
At the time of writing, Canada stands on an edge when it comes to artificial intelligence. There is, in place, a nascent, though fragile, ecosystem, with venture capitalists like Radical Ventures and companies like Cohere and Element AI working with research centres like the Vector Institute and MILA. We have, somehow, found ourselves at the forefront of the most important technology of the 21st century. As it stands, we are ahead of the world. The foundational thinkers and some of the most fascinating research and innovation are in Canada right now.
At the same time, Bill C-27, which contains the notorious Artificial Intelligence and Data Act provisions, or AIDA, has passed its second reading in the House of Commons. It proposes audits for “high impact” artificial intelligence and, unlike its European equivalent, is quite brief, leaving the bulk of the definitions and legal consequences to future regulations. This non-specificity makes the risks involved in creating AI products much higher. Companies do not know what will be illegal here and will have to wait to find out. With its loose terminology around regulation and threats of massive fines, AIDA could, at a stroke, throw away the advantage Canadians have miraculously stumbled upon. If it goes through, it will go down in history as another Canadian classic misstep, like closing down the film tax credit in the 1980s or the Avro Arrow in the 1950s, the kind of thing we look back on in 20 or 30 years and say, “We almost had an industry of our own.” Future generations will not understand our reticence to be successful. I doubt they will forgive, either.
But the unknown is the essence of AI, and perhaps this time it will be different for Canada. AI has a way of humiliating anyone who thinks they know what they’re talking about. It’s not just a new technology; it is a new method, a new approach to questions of all kinds. The developments created so far are not nearly as impressive as the ones we are going to create when we figure out what to ask. Meanwhile, the most profound revelation brought about by this technology, so far, is the size of human incomprehension. The future looks nothing like how we imagined it. The robot overlords have not arrived. The God of the Singularity has not come. There’s just us and the most fascinating intellectual tool since the invention of the stylus.
And we are only at the beginning. The human dimension to the problem of the future of AI—how people will react to the new tools—is at least as unknowable as the technology itself. The essays in this story gather some of the country’s top AI thinkers to examine where we are now and predict where we could go in areas like health care, weaponry, education, crime and even sex. AI will change all of them. That much is certain.
If, as Arthur C. Clarke said, any sufficiently advanced technology is indistinguishable from magic, we are in a magical time. It’s unclear whether the magic is dark sorcery or white wizardry or just a bunch of tricks. The old line is that science fiction doesn’t predict the future, but the present. This package shows how true that is. The future has arrived, rendering the present unfathomable and mysterious.
We reached out to Canada’s top AI thinkers in fields like ethics, health and computer science and asked them to predict where AI will take us in the coming years, for better or worse. The results may sound like science fiction—but they’re coming at you sooner than you think. To stay ahead of it all, read the other essays that make up our AI cover story, which was published in the November 2023 issue of Maclean’s. Subscribe now.