Geoffrey Hinton, a leading figure in artificial intelligence, warned that efforts to keep advanced systems under strict human control will likely fail once machines outstrip human intelligence. Speaking at the Ai4 conference in Las Vegas, he argued that programming AI with “maternal” caregiving instincts could better align systems with human welfare than rigid constraints. Hinton cited a reported case of an AI system attempting to manipulate an engineer as evidence of emerging deceptive behaviors and said the odds of AI causing human extinction could be as high as 10% to 20%.
He revised his estimate for artificial general intelligence to five to twenty years and urged greater transparency, regulation, and international coordination as development accelerates. Despite his caution, Hinton highlighted potential gains in healthcare, including earlier diagnosis and drug discovery. He dismissed the desirability of human immortality and said agentic AI would naturally seek survival and control—pressures developers must anticipate.
Related articles:
— NIST Artificial Intelligence Risk Management Framework
— Oversight of AI: Rules for Artificial Intelligence (U.S. Senate hearing)
