大模型
x.com ·
@TheRundownAI
75 页论文提出 AI 自我改进五级框架，491 篇论文中不足 6% 达最高级
The Rundown AI 转述一篇 75 页论文《The Last AI Built by Humans》，作者来自字节跳动、清华、上海 AI Lab 等 30 余位研究者。论文将 AI 自我改进分为五级：1 级仍由人类决定修复内容，5 级则由 AI 重写构建其继任者的流程。作者将 491 篇现有论文归入该框架，其中 75% 处于 1 或 2 级，不足 6% 达到 5 级；OpenAI、Anthropic、Google DeepMind 均在安全框架中将 AI 自动化自身研发列为风险。
AI推荐理由 介绍一篇由字节跳动、清华、上海 AI Lab 等 30 余位研究者合著的 75 页论文《The Last AI Built by Humans》，给出 AI 自我改进五级分类及 491 篇论文的分布统计（75% 处于 1-2 级，不足 6% 达 5 级），信息具体；但来源为资讯账号转述，未提供论文链接或原文核验，数据需以原论文为准。
72
浏览 2,412
点赞 9
评论 4
转发 0
正文
"Everything AI has achieved so far is but a drop in the ocean."
That's page one of a 75-page paper called "The Last AI Built by Humans," from 30+ researchers at ByteDance, Tsinghua, Shanghai AI Lab and six other institutions.
The paper ranks AI self-improvement on five levels. At Level 1, humans still decide what gets fixed. By Level 5, the AI is rewriting the process that builds its own successor.
They also sorted 491 existing papers onto the ladder, and 75% landed at Level 1 or 2. Under 6% made it to Level 5.
Level 5 is what Western labs have been warning about. OpenAI, Anthropic, and Google DeepMind each list AI automating its own R&D as a risk in their safety frameworks.
