Artificial intelligence pioneer Yoshua Bengio says he will be redirecting his research to ensure he is working on applications of the technology that are safe for society.
“I have been thinking a lot about this and I am going to do what I think is best to go in the right direction,” the 2018 winner of the A.M. Turing Award and director of the Mila AI institute in Quebec said Thursday evening.
“I think I am going to reorient my research so that either I am working on applications that are not dangerous or very safe like…healthcare or the environment or working on safety, to prepare and prevent what could happen.”
Bengio’s remarks came at the conclusion of a Munk debate hosted at Roy Thomson Hall in Toronto, where he and Massachusetts Institute of Technology physics professor Max Tegmark sparred with other technology experts about whether AI poses an existential threat.
The pair argued that AI is on a trajectory to become even more powerful and could accomplish goals that aren’t aligned with the best interests of humanity.
If anyone with malicious intent gets ahold of the technology and the right safety mechanisms aren’t in place, they said the result could be catastrophic.
“The amount of damage it can do to our society is also growing exponentially,” Tegmark said.
He pointed out that AI can do years of research or reading the entire internet with hours and can be goal-oriented, which means it can persuade, manipulate and build.
“It could, for example, research how to make more powerful bioweapons or how to make even more intelligent systems so it could personally self-improve itself,” said Tegmark.
“So, of course, it would have the power and potential to wipe us out and have those goals.”
As part of his arguments, Bengio said humanity needs to take care of all of the potential downsides of AI and should move toward regulation because anyone using the technology with malicious intentions could create “dire consequences.”
“We have weaknesses. There are many ways in which we can diluted and there are conspiracy theories and lots of people believe them,” said Bengio.
“There will be people who act in strange ways and will do things that can be very harmful.”
However, fellow AI pioneer Yann LeCun and Davis Professor of Complexity at the Santa Fe Institute Melanie Mitchell rebutted that AI can be used for good and does not have the agency to make the kind of decisions that would cause an existential threat.
LeCun, who doubles as vice-president and chief AI scientist at Facebook and Instagram parent company Meta, went so far as to say that AI will trigger “a renaissance” and “enlightenment.”
“Bad guys can use AI for bad things, but there’s many more good guys who can use the same, more powerful AI,” he argued.
Mitchell added that AI learns from human data, but still lacks fundamental aspects “of what it is to understand the world.”
“It’s a fallacy to think that a machine could be ‘smarter than humans in all respects,’ and still lack any common sense understanding of humans,” she said.
“You would never give unchecked autonomy and resources to an AI that lacks these basic aspects of intelligence. It just does not make sense.”
In recent months, LeCun, who won the A.M. Turing Award with Bengio and AI pioneer Geoffrey Hinton, has had a diverging view of AI’s risks when compared with his fellow award winners.
LeCun has said that AI will surpass human intelligence but doesn’t believe it is coming soon.
Meanwhile, Bengio joined more than 1,000 technology experts in March in calling for a six-month pause on training of AI systems more powerful than GPT-4 _ the large language model behind San Francisco-based OpenAI’s ChatGPT.
Co-signatories included engineers from Amazon, Google, Meta and Microsoft as well as Apple co-founder Steve Wozniak and Rachel Bronson, president of the Bulletin of the Atomic Scientists.
Amid the calls for a careful approach to AI, Ottawa has been moving toward adopting an artificial intelligence and data act.
The law, which is part of a broader bill on consumer privacy and data protection, would ban “reckless and malicious” AI use, establish oversight by a commissioner and the industry minister and impose financial penalties.
Ottawa has said it will come into effect no sooner than 2025.