No. of Recommendations: 7
This is why I always add 'Please answer rationally, fairly, and without any intention to please me or agree with me'.
It doesn't make the AI's answers any less crazy or wrong, but the placebo effect is a nice feeling.
I got sick of Gemini's slobbering sycophancy so I asked it how I could get it to knock it off. It responded that I should add this prompt:
“Give me the unvarnished facts with zero fluff. No introductory pleasantries, no concluding encouragement, and absolutely no compliments. Keep the tone completely objective, direct, and utilitarian.”
I also add "Do not solicit follow up questions, do not use anthropomorphizing language (ie no first person references). Do not provide a response unless the answer is known with certainty."
I think it helps. It makes it slightly less nauseating to use, at least.
