AI Futures Project moves superhuman AI researcher forecast to 2027
AI Futures Project updated its forecast after METR found that Anthropic’s defined-task model scores were improving at roughly 1.9 times their March 2024–October 2025 rate. The group now says a superhuman AI researcher could appear in 2027 if frontier AI development is not deliberately slowed, although its methods produced median estimates ranging from February 2027 to mid-2028. METR’s estimate that about 80% of the extra acceleration came from AI was explicitly uncertain and based partly on evidence Anthropic withheld, while Anthropic said Claude Mythos 5.2 had not doubled research speed or approached replacing researchers.
The update materially advances a prominent forecast based on an independent assessment of accelerated Anthropic R&D, but its near-term timeline remains conditional and depends on disputed estimates of AI-driven acceleration.
编者按:本文来自微信公众号 “新智元”(ID:AI_era),作者:元宇,创业邦经授权转载。
我们的时间表提前了!
11月11日,前OpenAI研究员Daniel Kokotajlo在X上转发团队的新预测时说,「我感到恐惧、悲伤和愤怒」。
他认为,智能爆炸似乎终究会在2027年发生:如果METR的判断成立,AI驱动的研发提速,就比《AI 2027》当初推演的提前了约五个月。

当AI开始帮人类研发下一代AI,连最激进的预测者,也可能低估AI模型进步的速度。
#AI Futures Project#METR#Anthropic#AI 2027#Claude Mythos 5.2