Event ·Multi-source

字节跳动等机构发布75页论文，将AI自优化分为5级，当前多数模型处于低级阶段

这篇由字节跳动、清华大学等机构联合发布的75页论文，将AI自优化能力分为五个等级。在Level 1阶段，人类仍决定AI的改进方向；而Level 5则意味着AI能自主重写构建其后续版本的过程。论文对491篇现有论文进行分类，结果显示75%的模型处于Level 1或2，仅有不到6%达到Level 5。OpenAI、Anthropic和Google DeepMind等西方实验室已将AI自动化研发列为安全框架中的风险。

Current read

朋友，字节跳动和清华等团队搞了个75页的论文，把AI自己优化升级的能力分了五级，现在看主流模型都还处在比较初级的阶段，挺有意思的。

11 sources78° AI heatLast updated 2026/9/15 16:32:06

Evidence

11 sources

Primary sourceThe Rundown AI

2026/9/15 16:32:06View source

Related source 1AI Will

2026/9/16 09:07:37View source

Related source 2arXiv: Google DeepMind

2026/9/13 21:32:35View source

Related source 3theverge

2026/9/14 22:59:41View source

Related source 4techcrunch

2026/9/15 15:47:03View source

Related source 5Detik Inet

2026/9/16 12:00:30View source

Related source 6Decoder

2026/9/14 09:39:59View source

Related source 7IT之家

2026/9/15 02:18:23View source

Related source 8Latent Space (swyx)

2026/9/15 04:50:36View source

Related source 9Blognone

2026/9/15 12:15:21View source

Related source 10Gary Marcus

2026/9/15 18:36:49View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.

Open the story