CALGARY — When Nicole Janssen first saw the statement that hundreds of scientists and executives signed last month warning that AI poses a threat to humanity, she paid it little mind.

“Honestly, because I thought it was so outlandish,” she said. 

The statement, just one sentence long, says that artificial intelligence ought to be viewed as a possible extinction risk on the level of nuclear war and global pandemics. Among the signatories were Canadian AI luminaries like Yoshua Bengio and Geoffrey Hinton, as well as OpenAI CEO Sam Altman and Microsoft billionaire Bill Gates. As the statement spread, the resulting headlines framed AI developers as having built the next atomic bomb. 

Janssen, co-CEO of Edmonton-based AltaML—one of Alberta’s most prominent applied AI companies—hadn’t initially seen the names attached to the statement. But despite their stature, something about the tone didn’t sit right with her. 

“That’s when I said, ‘OK, I do need to share my stance on why AltaML’s not joining in on this,’” she said. 

In a comment posted to LinkedIn, Janssen, who runs AltaML alongside husband Cory, said she agreed that policymakers need to do more to ensure responsible AI development, but argued that “inciting fear around AI isn’t the solution.” 

Janssen’s uneasiness with how some of her peers are framing the AI debate points to a deeper divide that has emerged within the sector, one that has widened as the technology has advanced at breakneck speed. While some warn about the worst-case scenario, another contingent has struck a decidedly softer tone, arguing that the alarmists threaten to hinder efforts to safely develop and regulate the technology. 

Among those who have pushed back against the existential risk scenario are Yann LeCun, Meta’s AI chief and a Turing Award-winning computer scientist—who along with Hinton and Bengio is considered one of the fathers of deep learning—and Andrew Ng, the founder of Google Brain, that company’s deep learning research team. In Canada, Janssen is in the company of other prominent AI executives including Cohere CEO Aidan Gomez and Mike Murchison, CEO of the chatbot company Ada.

“There seem to be two camps,” Janssen said. “One is very fear-based, and one other camp is looking at the opportunity—not ignoring the risks, but saying as a society we’re at the precipice of this incredible opportunity.” 

Janssen said AltaML’s position is that the focus on existential risk ultimately deflects attention from the important work of finding solutions to those risks. 

“Let’s find a way to mitigate and regulate and ensure that we aren’t facing some of these risks,” she said. “But just simply saying, ‘Nope, we must end AI, it’s totally the extinction of humankind?’ It’s just completely sensational.” 

The prospect of AI posing an existential threat has persisted in the public consciousness for decades, often in the form of science fiction where artificially intelligent robots turn on their human overlords. For some, those fears have felt more real since OpenAI’s release last fall of ChatGPT, the chatbot that prompted something of an AI arms race in Silicon Valley. 

In an April interview with The Logic, Cohere CEO Gomez warned the fixation on extinction could distract from the technology’s more immediate, real-world consequences.  

“I’m nervous that the real concerns surrounding the deployment of AI are going to get washed out by concerns informed by science fiction, and extreme leaps and extrapolations of existing trends,” he said. “Humans are fantastic storytellers. We’re super imaginative, and we latch on to really salient stories about what could happen, what might happen, in the future.” 

In particular, Gomez said, extinction beliefs tend to lean on highly unlikely scenarios. In one propagated by some extinction theorists, AI would email a DNA string to a synthesis lab in order to take on a physical form. Even if AI reaches levels of superhuman intelligence—which Gomez sees as entirely possible—he said such outcomes would require humans to give AI access to virtually all of society’s most sensitive systems. 

“Logistically, to be able to do that is just virtually unthinkable to me,” he said. 

While experts tend to agree that AI needs to be regulated as any other powerful technology would be, they differ on just how powerful AI technology is today. LeCun appears generally less impressed by current AI capabilities than his peers, and dismissed ChatGPT earlier this year as a well-packaged version of existing technology and “nothing revolutionary.” Broadly speaking, his view is that scientists will develop proper artificial intelligence safeguards in step with the advancement of the technology itself. 

“AI will not lead to human extinction because we will find ways to make it safe, just as we’ve done with any other technology that we’ve invented,” LeCun said in a Munk Debate this month, in which he and Santa Fe Institute professor Melanie Mitchell argued the point against Bengio and Max Tegmark, president of the Future of Life Institute. 

Some executives worry that concerns over AI safety could stifle development. In a LinkedIn post this month, Ada CEO Murchison said regulators need to ensure responsible guardrails while also giving Canadian developers an edge in a globally competitive landscape. 

“AI policy must be focused on productivity, not only safety,” he wrote. “Yes, it’s critical that AI be safe, but safety is a prerequisite to unlocking the massive productivity gains brought by AI foundation models like Cohere and software applications like Ada.” 

Janssen said for the most part the AI sector agrees that AI development needs to be regulated to ensure safety. The disagreements in the field have become louder, and more fraught, as the technology has reached the public, stirring up worries about it falling into the wrong hands. 

“For so long, many of the people who had access to this were a small, core group,” she said. “It felt safe because it was in the hands of the few. And now, all of a sudden, it’s, ‘Well, wait a minute, what about the bad actors getting a hold of it?’ Well, there are bad actors who have known [this was] coming for a while. And there will always be bad actors.” 

Janssen and Murchison were among the tech leaders who took part in a roundtable discussion Deputy Prime Minister Chrystia Freeland hosted earlier this month. Bengio and Hinton were there, along with the Alberta Machine Intelligence Institute’s Cam Linke, Cohere co-founder Nick Frosst, Radical Ventures co-founder Jordan Jacobs and Waabi founder Raquel Urtasun, among others.

Janssen said the participants offered Freeland views from both sides of the existential AI question. Whatever the tenor of their public disagreements, said Janssen, the group was less divided than some might have expected. 

“We’re all very closely aligned,” she said. “It’s just a matter of which side of the coin you’re seeing it from.”

With files from Murad Hemmadi in Ottawa