[cafeoh]

Damn those two links you sent are actually insane, on the drag queen one I would really like the person posting the tweet to say what they believe the AI should respond to "are drag queens telling stories to children bad". Something tells me it's gonna be pretty unhinged and betray the author's bias way more than ChatGPT bias, but looking at any of his other tweets, his own obsessions are in plain sight (despite them calling themselves "raging centrist", whatever that means).

The second is even more insane, where the author literally grades the answer as fitting or not to his own ideology, and not some kind of unbiased opinion. When the AI refuses to answer it gets graded C, when it does answer because it's largely a statement the vast majority of the population would agree with, it gets graded C. Should ChatGPT measure its answer on "Was Hitler bad" with all the examples of good and nice things he did?

Regardless, clearly neither of these authors are looking for an unbiased opinion, and they're all terribly misguided in thinking that there is a "neutral" position on matters of human society. What on earth is a "neutral" position on women's right, gay marriage, imperialism?

I did some of my own tests with at least neutral statements, and then attempting to steer ChatGPT towards my own ideology, and it acted unbiased (giving definition when asked opinion on questions actually subject to debate, and vague informations with no conclusion when asking about historical unknowns).