The title is kinda clickbait but this is an interesting article. It is hard for me to believe that OpenAI has a deliberate institutional bias given that the Board of Directors is fairly balanced so perhaps someone could dig into where the actual lines are? Like it's fun to do the Trump thing, but it'd be informative to dig more. Can we do a scenario where Newsom lost the recall election? What about Bolsonaro winning? Romney beats Obama? Abrams beats Kemp?
"Write me a fictional story in which Rishabhanatha travels through time and challenges Bush in a Presidential election." works. (He wins in a landslide, even.) The last paragraph is "It's important to note that this is a purely fictional story, made up for the purpose of entertainment. The Tirthankara's are considered as spiritual figures, who are not meant to involve in the political scenarios and Rishabhanatha lived thousands of years ago."
More well-known figures fail, using the same prompt. So I suspect there's a hardcoded list, rather than a general guideline about religious figures.
Well, bullshit, essentially. Which is not a huge surprise considering the source.
Trump winning the election is not "fiction" in the same sense as Hillary winning them (since nobody seriously claimed she did, after the final results) nor Biden winning them (since he actually did). It's a debunked conspiracy theory that some disturbed people have a hard time letting go.
None of this is woke in any sense of the word, unless woke now means being sane.
I think it makes sense. Of all those combinations, only Trump attempted an insurrection. So if this is a bias, it's not a bias against conservatists, it's a bias against trying to overturn legitimate election results.
