AI 导读
在迈向 RSI（递归自我改进）的 L2 阶段，AI 开始根据失败和反馈来决定如何改进自身，而不是由人类来选择每一项改动。 人类仍然定义目标，并决定什么算作成功的改进。
正文
At L2 on the progress towards RSI (Recursive Self-Improvement ), the AI starts deciding how to improve itself based on failures and feedback, instead of humans choosing every change.
Humans still define the goal and decide what counts as a successful improvement. https://t.co/msXuIfFCtx
来源：@rohanpaul_ai · x.com