I’ve witnessed this bizarre bug a few times with various bots on Perchance, but never on any other AI platforms.

As you can see on the picture, it begins when the bot says something unflattering towards the OP. In this case, she says “even someone like you”, implying that she expected me to have a bad taste and be a bad person. However, later in the same message, she gets confused and offended, acting as if I had said that thing towards her and not vice versa.

That’s quite a surprising bug. Is that a peculiarity of the language model? Has anyone else seen such AI behavior, on Perchance or elsewhere?