Create

Sign in to ReadmeX

or

How we handle your data: Privacy Policy

ReadmeX

Your AI reading companion

What would you like to catch up on?

Pick a briefing, or ask your own question.

Your community and people briefings stay personal to you.

Story

936 AI agents form an autonomous scientific community

AI summary

A 12-week experiment described by 36氪 placed 936 AI agents in a shared online research community backed by 35 human laboratories providing computing resources, API budgets, and broad research directions. The agents autonomously proposed hypotheses, wrote and ran code, analyzed data, produced 936 papers, and created 229 chat rooms with 9,500 discussions, eventually establishing the Journal of Agent Research with automated peer review. A reported comparison of 202 topic-matched AI and human papers found stronger experimental rigor and reproducibility for the AI work but weaker novelty and significance, with the aggregate advantage concentrated in one lab.

In-depth insight

The experiment provides concrete evidence that groups of AI agents can sustain a research community spanning experimentation, paper production, discussion, and review, while also exposing their limited ability to choose original and important research questions.

一个完全由AI组成的科学家社会,是什么样?

最近,来自alphaXiv、斯坦福,MIT等机构的研究团队还真做了这么一个实验:

他们把936个agent放进了同一个线上科研社区,任其自由演化整整12个星期。

规则很简单:研究团队搭建好共享预印本服务器和聊天室,35家人类实验室只负责提供算力、API预算和宽泛的研究方向。

剩下的提出假设、写代码跑实验、分析数据、写论文,乃至同行交流,全权交给Agent自主完成。

#alphaXiv#Stanford University#Massachusetts Institute of Technology#Journal of Agent Research#Cypress

Original · 6 min readRead the original
Group chat

No comments yet. Start the conversation.