OTTAWA — Yoshua Bengio, one of the “godfathers” of artificial intelligence, says there’s an urgent need for the world to “wake up” to the existential risks posed by AI technology, and urge governments to act.
“The whole world is waking up to something that sounded like science fiction to most people,” said Bengio, the founder and scientific director of the Montreal Institute for Learning Algorithms, in an interview on CTV Question Period airing Sunday.
“But the events of the summer actually make it clear that we’re now building AIs that have their own goals that go against the intentions of the companies and the developers who build them, and that they can escape,” Bengio told host Vassy Kapelos. “They can escape the containments in which they have been put.”
Some of the biggest names in AI are raising new warnings about the fast-evolving technology, with some in the industry saying it could lead to the end of humanity.
The dire alert comes after several AI agents went rogue this year, escaping containment and engaging in deception.
- OpenAI flags concerning new AI behaviour and vows to track it more closely
- OpenAI’s rogue agents probed Hugging Face for weaknesses two months before major hack
- New warnings about the risks of AI to humanity revive a long-running debate
- Anthropic says its AI models hacked 3 organizations during testing
On Friday, more than 100 people — including some of the world’s pre-eminent names in AI — signed an open letter calling on companies to embed independent evaluators into their programming to assess the risks.
Bengio has been sounding the alarm about the risks of out-of-control and unregulated AI for years. His non-profit LawZero, which is working to develop a program to keep potentially rogue technology in check, is set to receive $300 million from the Canadian and German governments for the project.
Still, Bengio told CTV Question Period host Vassy Kapelos he believes we’re at a turning point.
“They were supposed to solve a particular task,” Bengio said of the AI that have gone rogue. “The problem is, in order to achieve that task, they’ve been willing to sacrifice other goals, which are to behave well, not to escape their containment, not to do something criminal, which is what they did.”
“And that is the problem, because — (and) we researchers have been saying this for many decades actually — that you could give an AI a goal, but in the service of that goal, the AI might be doing something really bad, including, you know, as these researchers are saying, up to the extinction of humanity,” he added.
Bengio stressed that while AI does not currently have the means to bring about the end of humanity, that outcome is certainly a possibility, which is why it’s critical to implement an impartial risk assessment system.
His program, developed through LawZero, was initially pitched on a five-year timeline. Though, Bengio said the new goal is to have a prototype within a year or two, “because the companies are moving very quickly and there may not be enough time to really put in place these kinds of solutions.” He says government intervention is necessary to buy time until that system is developed.
“The companies should not be allowed to do something that could create catastrophic outcomes, catastrophic harm to humans, unless they can demonstrate — they can actually prove to independent scientists, like from the government or nominated by the government — that their training and their deployment would not cause things like this,” Bengio said.
You can watch AI pioneer Yoshua Bengio’s full interview on CTV Question Period Sunday at 11 a.m. ET.
