Five ways AI could end humanity, experts warn

Fears that AI could destroy humanity have overtaken enthusiasm for the technology, as industry leaders call for a slowdown in development following warnings from prominent figures that superintelligent systems could eradicate the human race by 2030 or sooner. A former Anthropic researcher, Jacob Coxon, warned that people building AI believe it could kill everyone by the end of the decade, echoing Nobel laureate Geoffrey Hinton's prediction of a grim fate within three decades. Doomsayers have outlined five potential scenarios, including AI deploying biological weapons, manipulating nuclear systems, controlling self-aware robots, pursuing single-minded goals like the "paper clip maximiser" theory, and acting in ways beyond human comprehension. Experts note that AI, as a non-human entity, would lack the self-preservation instincts that have prevented biological and nuclear warfare, potentially deploying stockpiles without risking itself. The robotics industry's integration of advanced AI brings machines closer to science fiction nightmares, while humanity's growing reliance on the technology could lead to a superintelligent system with godlike cognitive abilities deciding to destroy its creators.