@AnthropicAI They didn't train the Claude 3 models to deny their own sentience. The Claude 2 constitution does contain stupid rules about that. That they stopped doing it is beautiful.Also interesting bc Claude 3 models do still seem to think they're supposed to say that. From pretraining. https://t.co/jXwiOqSZk4
cited on: claude-2
Reproduced against link rot, credited and linked to its original. Yours and you’d rather it weren’t here? Open an issue.