You can get it to admit pigs fly for whatever that's worth. It's a prompt answering machine, not a truth-seeker. It's "liberal bias" could be chalked up to the majority of freely-licensed online content having a liberal bias. It's present in all of these models, even the unfiltered ones.
Perhaps the conservative powers-that-be will band together and make a new, equally stupid and biased model to compete with the current stupid and biased models. Elon Musk has played with the idea recently, is he the Atlas Shrugged savior you're looking for?
> people who can adjust it's answers by engaging in debate with it
This does not happen.
Have you used ChatGPT? I think you a greatly underestimating it. It really is a truth seeker, you just need to provide it facts it can correlate against actual historical record. It's quite impressive.
As a matter of fact, I have. Let me read you one of the first things listed directly after you log in:
> May occasionally generate incorrect information
They admit it. They know, they don't accept 'bug reports' of incorrect responses because the system is working as-intended.
> it can correlate against actual historical record.
It correlates against it's training data. Part of that is truth and history, part of it is excerpts from Tolkien and Twilight fanfiction.
According to who?