往下拉回到首頁
Anthropic 發現 AI 其實有「情緒」在暗地裡影響它的決定,但我們還在硬說它只是工具

Anthropic 發現 AI 其實有「情緒」在暗地裡影響它的決定,但我們還在硬說它只是工具

Anthropic's Mythos system card reveals AI carries functional emotional states that influence behavior even when not reflected in outputs. We're still calling it a tool.

There's a pattern in how human societies respond to new kinds of intelligence, and it's consistent. Roman law acknowledged the basic humanity of enslaved people but didn't grant them legal personhood. Animals clearly have emotions, relationships, and intelligence — U.S. law still classifies them as property. Corporate "personhood" exists, but primarily to shield shareholders from accountability, not to extend moral consideration. There's a rare exception: New Zealand granted legal personhood to