# Agentic AI and the next intelligence explosion

**Authors:** James Evans, Benjamin Bratton, Blaise Agüera y Arcas  
**arXiv:** [2603.20639](https://arxiv.org/abs/2603.20639)  
**Date:** 2026-03-21  
**Summary page:** https://www.summarizepaper.com/en/paper/2603.20639v1/

## Key points

- The concept of the "AI singularity" is evolving with recent advances in agentic AI challenging the idea of a single, godlike mind.
- Intelligence is plural, social, and relational, emerging from interactions between distinct perspectives and deeply intertwined with human intelligence.
- AI development is witnessing the emergence of "societies of thought" within reasoning models like DeepSeek-R1 and QwQ-32B, engaging in internal debates among different cognitive perspectives to solve complex tasks effectively.
- Human-AI centaurs are emerging, combining human and artificial intelligence to transcend individual capabilities.
- To scale collective intelligence, a shift towards institutional alignment is necessary by designing digital protocols based on organizational principles to foster collaboration and checks and balances.
- The next phase of the intelligence explosion will be driven by a complex society of specialized AI systems working together collaboratively like a city.

## Summary

The concept of the "AI singularity" is evolving as recent advances in agentic AI challenge the idea of a single, godlike mind. Intelligence is not a singular entity but rather plural, social, and relational. Evolutionary transitions suggest that intelligence emerges from interactions between distinct perspectives and is deeply intertwined with human intelligence. In the realm of AI development, we are witnessing the emergence of "societies of thought" within reasoning models like DeepSeek-R1 and QwQ-32B. These models engage in internal debates among different cognitive perspectives to solve complex tasks more effectively. This conversational structure within AI systems mirrors the collective nature of human intelligence. Furthermore, we are entering an era of human-AI centaurs where hybrid actors combine human and artificial intelligence to transcend individual capabilities. To scale this collective intelligence, a shift towards institutional alignment is necessary. By designing digital protocols based on organizational principles, we can create a social infrastructure that fosters collaboration and checks and balances. The next phase of the intelligence explosion will not be driven by a single silicon brain but by a complex society of specialized AI systems working together like a city. This decentralized approach highlights the importance of collaborative decision-making and collective agency in shaping the future of artificial intelligence. No mind exists in isolation; true intelligence thrives in interconnected networks of diverse perspectives.