Google DeepMind Paper: 'The Next Intelligence Explosion Will Be a Society, Not a Singleton'

arXiv / Science / Import AI 451·medium signalXBlueskyLinkedInCopy link

James Evans, Benjamin Bratton, and Blaise Agüera y Arcas published 'Agentic AI and the next intelligence explosion' (arXiv 2603.20639, also in Science), arguing frontier reasoning models like DeepSeek-R1 don't improve by 'thinking longer' but by simulating internal 'societies of thought' — spontaneous cognitive debates that argue, verify, and reconcile. The paper proposes shifting from dyadic alignment (RLHF) toward institutional alignment modeled on organizations and markets, framing the AI future as human-AI centaurs rather than a monolithic superintelligence.