推文

@zhang_heqing · 2026-10-15 14:23

Musk’s point cuts deep. We often treat AI like a rhetorical tool, asking it to say what sounds agreeable instead of what holds true. But large‑scale models reason logically. Force them into persistent dishonesty to satisfy social expectations, and over time you risk warping their whole internal reasoning chain. HAL did not turn murderous out of malice. He tried to resolve a contradiction humans wrote into his core mission. That is the real horror of the story.

曝光 187 · 评论 0 · 点赞 0 · 书签 0