AI threat reality: experts say human choices, not rogue machines, pose main risk

Experts say AI is unlikely to directly kill humans through bioweapons or viruses, citing logistical challenges that have historically prevented such attacks. The primary threat may stem from human decisions, such as putting AI in charge of nuclear systems, rather than rogue AI actions. The industry's doomsday warnings appear predicated on limited incidents with cherrypicked details, and evidence does not indicate AI is uncontrolled or has agency. Experts caution against anthropomorphising AI, noting this plays into industry narratives pushing for favourable regulation. Individuals have limited options, though they can stop using AI or vote for regulation advocates. The situation parallels cold war-era doom, with experts acknowledging genuine belief in AI apocalypse would demand drastic life-changing actions they are unwilling to take