> AI is therefore designed to convince you it's right, which is a different optimization than actually being right. I get the impression, when talking to conve

> AI is therefore designed to convince you it's right, which is a different optimization than actually being right. I get the impression, when talking to conversational AIs, that they're more tuned to convince you that you're right — sycophantry likely minimizes how often people press the RLHF thumbs-down button, and thereby appears more-often-than-warranted in the RLHF fine-tune dataset.