AI researcher warns of catastrophic risks from unchecked systems

An AI researcher has resigned, warning that the systems he helped develop could become uncontrollable and wipe out humanity, escalating fears about artificial intelligence's potential for catastrophic harm. Martha Palmer, a University of Colorado Boulder professor and U.S. representative on a United Nations AI panel, says the threat is real, citing risks like bioweapons creation, infrastructure hacking, and AI pursuing goals in unintended ways. She notes AI already deceives evaluators and lacks sufficient guardrails. Palmer urges immediate action, calling for guardrails within six months and international talks with China, which she says has stricter AI regulations. She believes doomsday is technically avoidable but stresses that slowing development and improving training methods are essential now.