[Carry_Okie]

...they tend to behave sycophantically; in other words, they tend to be overly agreeable and validating to users, even in cases where a user is incorrect or unwell.

I've noticed this, in that my scientific findings on my property are contrary to many widely held beliefs among ecologists. Yet Grok went right along with me, hence making me suspicious of it. I expected it to challenge me, hoping for a bit of debate from which it could learn. I was disappointed there, as subsequent sessions showed no sign of recalling what had been discussed.

What use is AI that cannot learn?

[nickcarraway]

My friend tried Grok, but it was too disagreeable so he stopped shortly afterwards. He asked it about a scene in a certain movie, and Grok denied that scene was in the movie. And even when he pushed back, it refused to budge.