Okay, so I was just talking with a friend and mentioned Alice wasn't an official Disney Princessβ’. Then started questioning if I was right about that or not. So I asked ChatGPT ΒΉ and it said Alice is totally a Disney Princess. Then it started going off on all the Disney Princess boxes that Alice checks.
About halfway through, I realized it was talking about me, and not the Wonderland Alice when it mentioned they were queer, neurodivergent, and had lockpicks π€£
I expect LLMs to be confidently incorrect, but that was unexpectedly so.
ΒΉI have no problem with folx wasting OpenAI's free-tier resources, so long as they don't pay them money or expect it to tell them the truth. Reminder: upvote the bad replies and downvote the good ones, that way you train it to be a little worse every time you use it, while also costing them money.







