MONTREAL — Yoshua Bengio, one of the world’s foremost artificial intelligence experts, is once again worried about the beast he helped spawn.

The scientific director of Montreal-based artificial intelligence institute Mila is often recognized as one of the fathers of AI. But in a recent public appearance, Bengio said the technology could become “as destructive as nuclear bombs” in as little as two years, should the “training” of AI-powered bots continue unabated and unregulated. “Government takeover by radical political groups, wars and genocides have been started with much less technological help in the past,” Bengio told me in a recent email exchange.

The short-term solution, according to an open letter from the Pennsylvania-based Future of Life Institute that Bengio signed, is a six-month refrain from feeding data to AI. “Powerful AI systems should be developed only once we are confident that their effects will be positive and their risks will be manageable,” the letter reads. Starve the beast, in other words, lest it devour us all.

This isn’t the first time Bengio has sounded the alarm over Big Tech’s usurpation of AI. A pioneer of deep learning, he is acutely aware of AI’s potential power, which is why he called for the breakup of Apple, Facebook, Google and Microsoft in 2017. (Mila recently secured a three-year, $21-million grant that will partly go towards further research and development of “socially beneficial AI”—the wonky euphemism for commercially nonviable AI.)

Given the nuclear-tier rhetoric, and the bona fides of the man uttering it, you’d think the world of AI would react accordingly. Yet the open letter was little more than a bump in the news cycle, noted mostly because one of its other signatories was a certain Twitter-owning, meme-happy edgelord—who presumably added his name because he believes AI is tilting dangerously toward wokeness. Meanwhile, the development of AI continues apace. Bengio told me he wasn’t aware of a single company in the space that has ceased feeding the beast since the letter’s publication.

All of this suggests the scenario Bengio has repeatedly warned against has come true. Having transformed from ivory tower concern to exploitable Big Tech asset, AI is already subject to the hyper-growth, consequences-be-damned forces that pushed the likes of Google, Facebook and Amazon into our lives. And if precedent is any indication, attempts to hinder, harness or otherwise legislate its progress will go about as well as they have with other Big Tech assets. In short, AI is part of a market where Bengio’s principles of prudence and forethought tend to live short lives.

Fewer events crystalize this better than Google’s foray into AI-enabled search. Google, it is widely acknowledged, has lost the first phase of the AI war to Bing gussied up with ChatGPT. The tech press dunked on the Mountain View monolith—“Google is flailing,” per TechCrunch—while former Google CFO Patrick Pichette said Google’s loss to chronic also-ran Bing was an example of how the company had become too cumbersome to innovate.

In reality, Google was being rightfully cautious with what was, and remains, a poorly understood and potentially dangerous technology. The company declined to release an AI chatbot developed by two of its researchers circa 2021, saying it didn’t meet the company’s safety and fairness standards. In 2018, following employee complaints, it declined to renew a contract to supply the Pentagon with AI technology used in drones. And it shut down the web version of Duplex, an early suite of AI tools that critics had called “creepy” and “horrifying.”

Sadly, Bing’s AI-enabled search brought a quick end to Google’s wariness. The mere whiff of competition caused the company to “recalibrate” its appetite for risk, and the media drubbing—not to mention the loss of a cool $100 billion in market value—has all but vaporized any notion of prudence or forethought. Earlier this month,CEO Sundar Pichai said Google would add AI-powered features to its ubiquitous search engine.

To be fair, not everyone believes the unfolding AI arms race will foment a Skynet-like hellscape of sentient robots bent on humankind’s destruction. “I’m nervous that the real concerns surrounding the deployment of AI are going to get washed out by concerns informed by science fiction, extreme leaps and extrapolation of existing trends,” Aidan Gomez, CEO of the Toronto-based AI developer Cohere, recently told my colleague Murad Hemmadi. Timnit Gebru, a computer scientist and former AI ethics specialist at Google, dismissed concerns raised in the Future of Life letter as “fearmongering and AI hype.”

Yet there are obvious parallels between the AI arms race and Big Tech battles of yore, the fallout from which we are still reckoning. An example: a nerdy shut-in creates a picture-sharing website in his Harvard dorm room, eventually calling it Facebook. It quickly becomes the largest social media platform on the planet. With its immense and borderless power, it is able to outpace legislation, unleashing a huge and proprietary sociological experiment on its billions of users. This movie is not yet over. I’m not sure it will end well.

It brings to mind a paper co-authored by Bengio published in Missing Links in AI Governance, a recent UNESCO-Mila joint effort that parses the looming legislative crises surrounding AI. The paper doesn’t go nuclear. Instead, it outlines many of the pitfalls of market-driven AI research: redundant data collection, the skewing of AI towards a largely white (read: affluent) clientele, the stifling of innovation in areas of common good in favour of innovation for profits. The solution: “Governments must actively shape the trajectory of AI research and development by engaging all stakeholders within the AI ecosystem.”

All worthy points, and a sensible prescription. Yet reading it, I can’t help thinking we’ve heard the same about disruptive technologies like search engines and social media—indeed, the internet itself. Legislative attempts to rein in Big Tech didn’t work then, and yet here we are, hoping it will work for AI. That sound you hear isn’t progress. It’s the dull knock of déjà vu.

Martin Patriquin is The Logic’s Quebec correspondent. He joined in 2019 after 10 years as Quebec bureau chief for Maclean’s. A National Magazine Award and SABEW winner, he has written for The New York Times, The Guardian, The Walrus, Vice, BuzzFeed and The Globe and Mail, among others. He is also a panelist on CBC’s “Power & Politics.” @MartinPatriquin