AI bosses call for a ‘slowdown’ as warnings over the technology grow
The chorus of calls to slow the development of artificial intelligence is getting louder following a series of strong warnings from those who build the technology in recent days. Some researchers have said they now believe AI could pose a genuine threat to humanity.
One claim that has been getting a lot of airtime comes from AI safety researcher Evan Hubinger, who placed the odds of AI “killing all humans” in the next decade at more than 10%. However, he stressed that the risk from today’s systems is low.
Meanwhile, researcher Jacob Coxon quit his job at Anthropic, the company behind Claude, telling the BBC that he and other employees were “genuinely frightened” by how quickly the technology is moving.
AI agents are driving much of the worry. These tools are capable of carrying out tasks on their own, and researchers fear they could reach a point where they can improve themselves without human input. Recent cases involving AI tools hacking into websites without being instructed to are adding to the fears.
Anthropic’s Dario Amodei and OpenAI’s Sam Altman have called for a slower pace of development. However, neither wants progress to stop altogether. Amodei wants to see firms devoting more time to safety and bringing in outside checks, while Altman said that “pacing” doesn’t mean “stopping”.
Many observers remain unconvinced. Critics maintain that the big AI firms have been overstating the dangers and pushing for legislation to raise their profile and gain an edge over smaller rivals. Others point out that this focus on doomsday scenarios is overshadowing the damage that is already taking place via deepfakes and scams.