New research debunking AI existential threats argues that social, physical, and regulatory constraints prevent an AI apocalypse, favoring sector-specific policy instead.