Raw LLM Responses

I asked an AI about the interpretability results on emotional content and how it was affecting behavior like cheating when given an impossible task and whether some sort of emotional regulation system might be needed. It said that the same researchers claimed that they were getting better at regulating the behavioural effect themselves in order to improve their performance — maybe via scaling? I don’t know if that is true or if the AI was hallucinating that part of the research. It sort of sounds like scaling up compute is having the effect of child development where a child goes from being controlled by emotions (like tantrum or emotional judgment) to having more ability to understand and control them.