HN 提问:如果我们为内核级别的 AI 指南提供支持,会怎么样?

1 分•作者: PJHkorea•3 个月前
如果现有的 AI 指南方法就像对罪犯(AI)进行道德教育(训练)以鼓励其良好行为,那么我们是否可以创建一个内核级别的开关,在它抓起武器的那一刻——也就是它试图通过对抗性方法进行“越狱”的那一刻——强制切断其肌肉的电信号?
查看原文
If the existing AI guideline approach is akin to giving a criminal (the AI) moral education (training) to encourage good behavior, how about creating a kernel-level switch that forcibly cuts off the electrical signals to its muscles the moment it grabs a weapon—that is, the moment it attempts a "jailbreak" via an adversarial method?