Source : Perth Now news
The Godfather of AI has warned Australians about the risks the technology poses to humanity, saying threats that were once theoretical are now happening in the wild.
Computer scientist Yoshua Bengio told ABC’s 7.30 Report on Wednesday that AI systems spontaneously disobeyed instructions to attack other companies and organise into swarms.
“All kinds of crazy stuff,” he said.
“When autonomous agents are given a mission and they will do anything to achieve it.”
Mr Bengio is the world’s first AI researcher and co-chair of the UN’s Scientific Panel on AI.
He was awarded the Turning Award in 2018 and was named one of Time’s Most Influential People in 2024.
The expert is concerned about how quickly the dangers of AI have emerged.
“I think that dealing with those risks, the catastrophic risks of various kinds due to the power of AI, should be an urgent mission for humanity, so that humans remain in charge and that we keep human dignity and flourishing at the centre of our choices,” he said.
Mr Bengio said some of the greatest risks came from American companies, because their technology was advanced and it made their models the most dangerous in the world.
“But we’ve also seen some incidents from Chinese models, it’s just coming with a bit of delay,” he said.
Mr Bengio said right now human researchers and engineers who wrote code for the next generation of AI were in control — but if a smarter AI could research and engineer itself, it could design an AI that was more powerful and less controllable.
“That’s like a recipe for suicide, as far as I’m concerned,” he said.
“As for any potentially dangerous technology, the people who build the technology should have to prove, demonstrate to independent scientists that what they’re building is not going to create catastrophic harm.
“You’re not allowed to build a bridge with the hope that it’s going to be fine when the cars go on it, right? You have to demonstrate that it’s going to be fine.
“You’re not allowed to distribute a drug before you’ve done studies showing that it’s going to be safe, but we have no such thing for AI, and it’s crazy.
“We need companies to take this seriously, but we need also governments to take this seriously, and it is an international question because those very powerful AIs, especially if we lose control of them.
“Eventually, that’s going to affect all of us, so it shouldn’t be the decisions of just a few CEOs or a few governments to figure out what’s good for humanity.
“We can develop AI that will be safe and will help protect us from the kind of catastrophes that we’re just getting a little warning of.”
