News Summary

Yoshua Bengio, a leading figure in artificial intelligence and founder of the Montréal-based AI research institute Mila, has issued a stark warning about the potential catastrophic risks posed by agentic AI systems. Speaking at the World Summit AI conference in Montréal, Bengio emphasized the dangers of creating superintelligent AI agents without addressing self-preservation issues, which could lead to loss of human control and even human extinction. Bengio, a Turing Award laureate, has been vocal in calling for caution in AI development, signing open letters that stress the existential risks AI poses. He advocates for non-agentic AI models, which focus on assisting humans in areas like medicine and climate change without having goals or intentions of their own. Bengio’s concerns extend to the economic implications of powerful AI being concentrated in a few countries, potentially disadvantaging nations like Canada. He urges prioritizing safety and beneficial scientific advances over job replacement in AI research.