Agentic AI and the next intelligence explosion.

The article focuses on the evolving nature of artificial intelligence (AI) development, arguing that intelligence is inherently social and plural rather than a singular, monolithic phenomenon. It highlights recent findings that advanced AI models internally simulate multi-agent interactions, resembling societies of thought, which enhance reasoning accuracy. The authors propose that future AI progress will depend on designing complex, hybrid human-AI social systems and institutions that enable cooperation, conflict, and governance at scale, rather than solely increasing individual AI capacity. This perspective emphasizes the importance of institutional alignment and oversight in AI deployment, especially in high-stakes governance contexts, envisioning intelligence growth as a collective, socially embedded process involving both humans and AI agents. [Extracted from the article]