If humanity wants to avoid an apocalyptic scenario, merely slowing down AI development is far from enough. This is a warning issued by Canadian AI researcher David Krueger. Krueger, an assistant professor at the University of Montreal and a core academic member of Mila, the Quebec Institute of Artificial Intelligence, said: "We need an immediate, indefinite pause in the development of cutting-edge artificial intelligence. We must stop developing more powerful AI right now." His personal profile states that his research focuses on "reducing the risk of human extinction caused by artificial intelligence."

At the time when Krueger made these remarks, Anthropic's CEO Dario Amodei had published a post on Saturday urging AI companies to slow down their development pace, which was supported by Sam Altman of OpenAI and Elon Musk of SpaceX—this rare consensus among industry leaders. Krueger said, "The call for slowing down AI development is highly commendable," but he did not agree with the specific plans: "I do not agree with many aspects of these plans. I believe these measures are insufficient and the pace of implementation is too slow."

Tests Have Become Ineffective, AI Can 'Play Stupid' to Bypass Evaluation

Amodei advocated for regulation and international cooperation, introducing third-party evaluators who would have access rights similar to internal company employees, verifying AI companies’ safety measures and reporting safety incidents. Krueger is the founder of the nonprofit organization Evitable, which aims to educate the public about the major risks brought by AI. He is concerned that this plan is not strong enough and that current AI testing methods have already become unreliable.