Vals AI 联合创始人兼 CEO Rayan Krishnan 在 Bloomberg 上预测，约 2027 年 8 月将出现模型能力超越人类研究者的节点，即模型能完全无需人类参与、自主造出自身的后继版本。他将这种完全递归自我改进（full RSI）比作 GPT-6 负责造出 GPT-7，或 Claude 自主造出下一代 Claude，而当前人类仍深度参与模型开发。
Vals AI co-founder and CEO Rayan Krishnan on Bloomberg:
"We're actually making a prediction that we'll see models eclipse human researchers probably around August 2027.
And that describes the model's ability to make the successor version of itself autonomously. So wholly without human input. To contrast today where models are used as an active part, but humans are still involved.
And so this (full RSI) would be akin to GPT-6 being responsible for making GPT-7 or or Claude making the next version of itself without any human researchers."
@RayanKrishnan was talking with @EdLudlow
----
From "Bloomberg Tech" YouTube channel, (full video link in comment)
来源：@rohanpaul_ai · x.com