AI-Torture-Chamber

1 points | by rozumbrada an hour ago

1 comments

  • rozumbrada an hour ago

    Steering language models into strong negative and positive valence states, and measuring what they say and what they're willing to do about it.