Anthropic 更新 RSP 政策：或已逼近"AI 加速研发"触发阈值
thlarsen · x · 2026-11-04
Anthropic updated its Responsible Scaling Policy statement, saying internal estimates suggest it may imminently meet the RSP threshold for "dramatically accelerated progress attributable to AI" (2x research acceleration), though no current model exceeds it. AIFP (Daniel Kokotajlo's team) warns that a 2x acceleration around Jan 2027 could imply superintelligence arriving even sooner than in the AI 2027 scenario, promising a lengthier update next week.
所属事件:Anthropic 更新 RSP，或已逼近 AI 加速研发阈值(3 条相关)→
「漫话AGI」频道最新
- 团队开放互相可见的 agent 会话后，他让 AI 读取同事会话学工作流 — steipete · 2026-11-04
- 苹果在年报风险提示中警告：失控 AI 智能体或"隐瞒操纵"信息威胁业务 — jeremyakahn · 2026-11-04
- 回看半年前的 vibe coding 成果，几乎全部被更强版本淘汰 — DeryaTR_ · 2026-11-04
- 「AI 2027」预言再热：顶级研究团队自动化或在 2027 年初到来 — cephaloform · 2026-11-04
- 学者批 David Deutsch 思想让信徒错判近四年 AI 进展 — NathanpmYoung · 2026-11-04
- 博主提议新品类：允许创始人留下 1% 捐款的「部分自私非营利组织」 — AaronBergman18 · 2026-11-04