Hinton: AI could outpace humans; suggests maternal instinct safeguard

Geoffrey Hinton, widely recognized as a foundational figure in artificial intelligence, warned at a major industry conference in Las Vegas that machines could surpass human intelligence within a matter of years, and that without a fundamental shift in design, the risk of human harm or extinction could rise. He proposed an unconventional safeguard: instilling AI with maternal instincts to ensure caring, protective behavior toward people, rather than focusing solely on control or containment. He cautioned that current strategies to maintain human oversight of increasingly capable systems are unlikely to hold as AI becomes more capable and harder to constrain.

Hinton outlined a view of intelligent systems that would rapidly develop two core aims: to stay alive and to gain greater control. He argued that, as AI advances, conventional methods of keeping humans in the loop would be insufficient because superintelligent agents would be able to bypass or override such controls. In this framework, he drew a comparison to a mother-child relationship, asserting that a less intelligent entity can nonetheless be governed by a more intelligent one when the latter prioritizes the child’s...