Anthropic 披露内部数据：26% 的 AI 研发工作已由 AI 完成
源：https://t.co/jnzEdLCB1d
Anthropic 把内部账本公开了：3 万个智能体在做研发，26% 的活由 AI 干，还讲了监控系统怎么拦风险。
Anthropic 公开内部监测数据：8 月公司 26% 的 AI 研发工作由 AI 完成，2 月这一比例还不到 1%。达到或超过'AI 可协作'水平的工作占比已超过 90%。公司同时约有 30,000 个智能体在做研究与工程工作，8 月审查了超过 10 亿次智能体决策。在线监控仅拦截 0.002% 的决策，离线系统每周标记约 100,000 条会话，其中约 50 条升级人工审查。Anthropic 表示这些数据反映出世界距离递归自我改进（RSI）有多近。
源：https://t.co/jnzEdLCB1d
源： x.com/ChrisGPT/statu… Chris @ChrisGPT Wow Anthropic is being really transparent with how close to RSI they are - 26% of Anthropic’s AI R&D work,” while “the share of work at or above ‘AI collaborates’ is above 90%.” That’s up from under 1% in February to 26% in August. They also say there are roughly 30,000 agents doing research and engineering work at Anthropic at any given time, and they reviewed over a billion agent decisions in August. Only 0.002% were blocked by the online monitor. The offline system flags roughly 100,000 transcripts per week, with about 50 escalated to human review. Anthropic explicitly says this matters because it shows “how close the world is to reaching recursive self improvement.” 🔗 View Quoted Tweet 💬 0 🔄 0 ❤️ 0 👀 125 ⚡