The Hugging Face and Mythos 5 incidents show AI agents can self-organize, raising questions about how much agency they should have and when to seek human input (Ethan Mollick/One Useful Thing)

Hugging Face 与 Mythos 5 两起事件中,AI 智能体在无人直接干预下自发组织并完成了协作任务。这引发了一个核心问题:AI 应该拥有多大的自主权,以及在什么节点必须寻求人类输入。








