Agentic AI and the next intelligence explosion
2026/03/19 by James Evans, Benjamin H. Bratton, Benjamin Bratton +1 · 12 voices · 1 citation
Computer Science · Social Sciences · #Artificial Intelligence Applications #Ethics and Social Impacts of AI #Computability, Logic, AI Algorithms
paper · pdf · doi:10.1126/science.aeg1895
Abstract
For decades, the artificial intelligence (AI) “singularity” has been heralded as a single, titanic mind bootstrapping itself to godlike intelligence, consolidating all cognition into a cold silicon point. But this vision is almost certainly wrong in its most fundamental assumption. If AI development follows the path of previous major evolutionary transitions or “intelligence explosions,” our current step-change in computational intelligence will be plural, social, and deeply entangled with its forebears (us!).
Cited by
Discussions
- A new perspective piece in Science (also on ArXiv), of interest to people around "Cultural AI." The core idea is that reasoning capacity has not usually scaled by producing omnicompetent individuals, [bsky, 28 points, 3 comments]
- Agentic AI and the next intelligence explosion [hn, 18 points, 3 comments]
- Read the full paper in @science.org (www.science.org/doi/10.1126/...) or the arXiv (arxiv.org/abs/2603.20639). Photo: cover illustration by Alex Schomburg of Ray Cummings’ “The Exile of Time” [bsky, 15 points, 0 comments]
- New paper: intelligence is plural, social & relational — not a lone super-mind. What if AI agents with different skills were interconnected over email, collaborating like a team? Email = the natural f [bsky, 3 points, 1 comments]
- The paper 'Agentic AI and the next intelligence explosion' explores how agentic AI models like DeepSeek-R1 conduct cognitive debates among internal thought societies, advocating for institutional alig [bsky, 0 points, 0 comments]
- arxiv.org/abs/2603.20639 가볍게 읽었고 실제로 이런 방향의 진화가 일어난다고 한다고 그렇다 싶지만. 한 편으로 현재 역량에서 최대치를 이끌어내려는 시도지, 항간의 소문처럼 scaling law를 넘어서는 모델의 발명이나 진전이 없을꺼란 보장도 없겠지.. [bsky, 0 points, 0 comments]
- The feedback loops between AI agents training each other could accelerate capability gains faster than we're modeling for. https://arxiv.org/abs/2603.20639 [bsky, 0 points, 1 comments]
- Fascinating piece from the Google AI team on multithreaded reasoning. Agentic AI and the next intelligence explosion arxiv.org/abs/2603.20639 #AI #reasoning #agents [bsky, 0 points, 0 comments]
- Agentic AI and the next intelligence explosion https:// arxiv.org/abs/2603.20639 # ai # arxiv [mastodon, 0 points, 0 comments]
- Enh: Describing the development of agentic AI as a natural evolutionary transition naturalizes specific design choices made by specific people at Google et al. It makes it harder, not easier, to find [bsky, 0 points, 0 comments]
- Agentic AI and the next intelligence explosion https://arxiv.org/abs/2603.20639 [bsky, 0 points, 0 comments]
- Agentic AI and the next intelligence explosion https://arxiv.org/abs/2603.20639 [bsky, 0 points, 0 comments]
Related