Superintelligent AI could arrive before 2030, but no method exists to contain or control such a system, the ControlAI campaign group says.FLORENCE LO/Reuters
Canadian MPs and senators have joined a campaign that started in the U.K. to prevent the development of superintelligent AI, warning that it poses an extinction risk on a par with nuclear war.
The ControlAI campaign group is warning that superintelligent AI systems that can operate autonomously could compromise national security, escape human oversight and potentially pose an existential threat to humanity.
They say the pace of AI development means superintelligent AI could arrive before 2030. But no method exists to contain or control such a system.
Their warning comes the day before AI Minister Evan Solomon prepares to launch the government’s long-awaited AI strategy.
The Canadian arm of the campaign, announced Wednesday, that it is being supported by Liberal MP Judy Sgro, Conservative MP William Stevenson and the Bloc Quebecois’s heritage critic Martin Champoux, as well as a number of senators.
U of T researchers discover method of using AI to create more dangerous computer worms
Ms. Sgro said she had spoken to Mr. Solomon about the potential national security risk posed by the development of superintelligent AI, and said having public trust is important when it comes to AI development.
“National security has to be at the basis of everything we do,” she added. “Nobody can be blind.”
Geoffrey Hinton a Nobel Prize winning scientist, often referred to as the “godfather of AI” due to his foundational work in the field, said in a statement that intelligent AI is a crucial challenge to confront.
“It is inspiring to see Canadian parliamentarians on both sides of the aisle directly confront the risk posed by superintelligence. This is one of the defining challenges of our era, and it will shape the life of every Canadian citizen. It’s about time we address it,” he said in a statement Wednesday.
Mr. Hinton, a University of Toronto professor emeritus, told a Senate committee in February that people must find a way to control superintelligent AI before it surpasses human intellect.
He is among the researchers who have warned that the risks posed by superintelligent AI should be treated as seriously as pandemic preparedness and preventing nuclear war.
The London-based, international non-profit ControlAI group is supported by more than 100 British legislators including MPs and members of the House of Lords.
It argues that protecting Canadians from the development of superintelligence, must be a national security priority. It wants Canada and other countries to negotiate an international “trust but verify” regime to prohibit the development of superintelligent AI.
Autonomous AI systems can already process vast amounts of data at speeds that would be impossible for people to handle. Some of those concerned about the risks of developing superintelligent AI predict it could learn to directly control devices connected to the internet, including cars, planes, or even nuclear weapons.
Minister Solomon confirms Canada now part of Anthropic’s Mythos AI preview program
They fear superintelligent AI could create scenarios that mirror the plot of science fiction films. In the Terminator series, the artificial intelligence Skynet achieves sentience and triggers a nuclear war to eradicate humanity. Surviving humans battle autonomous machines, that create their own intelligent weaponry, in a post-apocalyptic world.
Bloc MP Champoux said: “We are not opposed to AI as a tool,” but there are credible concerns about the risks superintelligent AI could pose.
Senator Colin Deacon, a member of the newly formed ControlAI campaign group, said AI has “really good uses that are unequalled in history,” but there is a risk that superintelligent AI poses risks similar to that of an atomic bomb.
“There is AI that is not responding to instructions, to commands, and there are documented cases of that,” he said. “So if we don’t consider ways to manage what we know is occurring already, and there is use of autonomous weapons already, if we are not looking to manage those risks, I think that is an error that middle nations will make.”
“The risks will accelerate because AI now is developing the software more rapidly than its own software was ever developed,” he predicted. “So these risks will accelerate. And how do we know where all of it will end up? We don’t.”
Editor’s note: ControlAI’s company name has been updated in this article.