How to Navigate ChatGPT's Sycophancy?
It's not nearly as bad as it was a few models ago, when Chat said that everything I wrote was incredibly insightful, but still, Chat loves to compliment my thinking. Thanks, Chat! Except sometimes I'm *not* brilliant. Has anyone worked with adding something to their prompts to get a more honest assessment of their thinking? If I say something truly insightful, then sure, please tell me. But when I'm thinking in a shallow manner, and especially when I'm wrong, I want to make sure that Chat tells me that, too.