22位AI专家联合撰文探讨AI自我进化风险

Anthropic的R&D工作已有26%由AI完成，远高于3月份的1%。论文指出，一旦AI达到专家水平，一个实验室的计算能力可运行相当于'数百万顶尖人类研究员'的工作量。

在完全自动化场景下，一年AI进步可能仅需五周完成。

22 researchers, including OpenAI's chief scientist, Anthropic's co-founder, and AI 'godfathers' Geoffrey Hinton and Yoshua Bengio, just co-wrote a paper on how AI building AI could set off an "intelligence explosion." Some of the wildest parts: - AI now completes 26% of Anthropic's R&D work with only high-level human oversight, up from 1% in March - Once AI hits expert level, one lab's current computing power could run the equivalent of "at least millions of top human researchers" - In one scenario after full automation, a year of today's AI progress would take about five weeks - AI labs are spending more (as a share of U.S. GDP) than the Manhattan Project and Apollo Program combined At the extreme end of the range of outcomes, according to the paper: "the marginalization or extinction of humanity."

本条情报归属 AITOP「论文」核心栏目，重点聚焦北美前沿实验室与硅谷科技巨头。在跨区域产业观察坐标中，该动态直接关联「中美大模型技术代差与商业化对照」研究主线，反映出当前技术演进与商业落地的最新风向。涉及核心实体与技术议题包括：Geoffrey Hinton、Yoshua Bengio、Anthropic、OpenAI，读者可通过下方标签进入对应专题页查阅历史脉络与关联对比。

信源核验说明：本文原始线索采编自「The Rundown AI」（@therundownai），经 AITOP 多维评分管线结构化提炼。截至目前，全网已有 11 家独立信源对该事件进行交叉印证与跟进报道（含 Decoder、IT之家、The Information 等），属于近期高关注度行业信号。