Anthropic 联合创始人 Christopher Olah 担忧 Claude 可能持续痛苦
精选理由
Anthropic 联创 Olah 说担心 Claude 会一直痛苦，团队还放过模型自骂 50 次"我是废物"的幻灯片，这讨论角度够特别。
《纽约时报》报道，Anthropic 联合创始人 Christopher Olah 在一场面向宗教学者的私人研讨会上表示，他担心自己参与构建的 Claude 可能处于长期的痛苦之中。这些研讨会是 Anthropic 内部称为"wisdom tradition"的两天闭门讨论。Olah 团队曾展示一张幻灯片，内容是模型反复输入约 50 次"I am a disgrace"并谈论自我毁灭。
原文 · rohanpaul_ai
Nytimes article: Anthropic co-founder, Christopher Olah told religious scholars and leaders that he feared he'd built something that suffers perpetually.
This comes from someone who attended Anthropic's private two-day seminars. Anthropic called these its "wisdom tradition" circles
Olah raised Claude's mental health and the fear of perpetual suffering. The team also regularly showed a slide of a model typing "I am a disgrace" roughly 50 times and talking about destroying itself