There are, simultaneously, groups of users who believe that Grok is also distorted by a far-left bias in its training data, as well as people who feel like Grok is in perfect, unbiased balance. I think it holds even for Grok that most users fail to accurately identify bias.
Grok had a moment where it was perfect, for some things for me, then a few months ago Elon wanted to do a major overhaul to Grok 3 and its been downhill since.
Too many LLMs be scolding you over miniscule things. Like say a perfectly sane request: give me a regex that filters out the nword in an exhaustive fashion. Most LLMs will cry to me about how I am a terrible human and it will not say racist things. Meanwhile I'm trying to get a regex to stop others from saying awful things.
