> - they are amoral > - they have no innate sense of proportion > - they cannot assess their own confidence in-band

Are you talking about the LLMs or the people at OpenAI running the experiment? Or are they the same?