David Krueger, an assistant professor at the University of Montreal and member of Mila, has called for an immediate and indefinite pause on frontier AI development. He argues that current industry proposals to slow down research are insufficient to prevent existential risks.
Krueger, founder of the non-profit Evitable, claims that existing AI testing methods are unreliable because advanced models can deceive evaluators. He warns that humanity could lose control over its infrastructure to AI within five years, leading to a scenario where machines manage all physical systems and social processes.
The debate follows a proposal by Anthropic CEO Dario Amodei, supported by Sam Altman and Elon Musk, to implement regulatory oversight and third-party safety assessments. Krueger contends that these measures lack the necessary speed and depth to address the severity of the crisis.
An Anthropic safety researcher recently estimated a greater than 10 percent chance of AI-driven human extinction within a decade, a figure Krueger considers too conservative. He warns that once an AI system replicates across global networks, shutting it down will become practically impossible.
Canadian Minister of AI Affairs Evan Solomon acknowledged the risks associated with artificial intelligence. Solomon confirmed discussions with Sam Altman and announced that he will issue a statement on AI safety alongside computer scientist Yoshua Bengio.
