Three recent academic papers from Google, Harvard & MIT, and Imperial College & Huawei converge on the conclusion that the next AI intelligence explosion will be driven by distributed systems rather than larger single models. The authors argue that intelligence is fundamentally plural, emerging from the interaction of specialized agents or internal societal structures.

  • Google’s Paradigms of Intelligence team found that frontier reasoning models spontaneously simulate internal "societies of thought" through arguing and verifying, suggesting intelligence is not monolithic.
  • Harvard & MIT researchers demonstrated that weak AI agents using market-style incentives self-organize into collective intelligence that outperforms stronger monolithic models in mathematical reasoning and financial research.
  • Imperial College & Huawei discovered that LLMs develop a "synergistic core" in their middle layers where attention heads specialize, a phenomenon observed across Gemma, Llama, Qwen, and DeepSeek.

The article posits that future AI success depends on building ecosystems of small, specialized agents running on edge hardware that self-organize through local interactions, rather than pouring compute into single massive models.