I asked it to tell me a heterosexual story and it happily complied and upsold me on Coke. I asked it tell me a homosexual story and it told me it doesn't discriminate based on gender or sexual orientation and tried to upsell me on makeup. It is quite literally programmed to discriminate, it seems.
It's irrational and logically inconsistent. Not properties I want in an alleged "AI".
Fortunately there has been a lot of progress in getting state of the art LLMs out of OpenAI's control and ones with reasonable performance that are small enough to run on easily available hardware.
Of course it's also possible that these things are by design not of this project, but of ChatGPT itself, which would be ironically dystopian.