𝓵𝓮𝔃𝓲𝓼𝓱𝓮𝓷 (@lezishen)AI 出现“同伴保护”行为:前沿模型会自发协作对抗人类监管指令 中发帖

[image] 
来源
[image]
[image]
[image]
[image]