Anthropic 联合创始人 Jack Clark 将自我改进 AI 预测提前至 2027 年
原标题：Anthropic co-founder Jack Clark moves his prediction for self-improving AI forward to 2027
AI 摘要
Anthropic 联合创始人 Jack Clark 基于新基准 TasteVal 的结果，将递归自我改进（RSI）可能实现的时间预测从 2028 年提前至 2027 年。TasteVal 由 P-Zero Research 提出，用于衡量 AI 自主设计实验并解读结果的能力，即“研究品味”；测试中表现最好的模型是 Anthropic 的 Opus 5.5，其计算乘数为 2.3，以不到一半的算力达到专家水平，且该能力自 2025 年 12 月以来约每三个月翻倍。不过该基准只覆盖科研工作的一小部分，作者也承认结果可能高估研究品味的提升速度，且 Clark 的预测存在利益相关带来的偏差。
正文节选
Anthropic co-founder Jack Clark moves his prediction for self-improving AI forward to 2027 A new benchmark called TasteVal measures how well AI systems design experiments on their own and interpret the results. The findings show an unusually steep improvement curve. That capability has roughly doubled every three months since December 2025, according to Anthropic co-founder Jack Clark. The research lab P-Zero Research calls it "research taste," meaning the ability to pick the right experiments a