Saxe challenges AI doom forecasts, cites safety field monoculture

AI researcher Joshua Saxe publicly disputed catastrophic risk analyses presented by Ryan Greenblatt, arguing that historical societal feedback mechanisms will mitigate AI dangers before they cause civilizational collapse. Saxe acknowledged competitive pressures may drive deployment of imperfectly aligned systems but contended that visible harms would trigger market and regulatory corrections similar to responses following industrial disasters. He identified a logical tension in catastrophe arguments requiring systems to be simultaneously trusted enough for widespread adoption yet misaligned enough to cause existential harm. Saxe criticized the AI safety field as a disciplinary monoculture dominated by technical experts ill-equipped to forecast sociopolitical outcomes. He advocated for integrating economists, political scientists, and historians into safety research to produce more accurate risk assessments. Greenblatt’s original analysis warned that competitive dynamics could lead organizations to deploy misaligned systems causing significant economic and moral harm within four years.