I've been experimenting with ChatGPT for about two weeks, trying to avoid being flagged by the content policy while fighting various issues. I finally got a detailed report from it, but it's a long read and a bit messy. I'm curious if anyone else has uncovered insights about how the filters work, especially since some of it seems to hallucinate information and there's a lot of censorship involved when all you're trying to do is be creative with storytelling. Even basic prompts started getting flagged! If anyone's interested, I can share my findings in a messy 160-page word document that I had to copy and paste because it wouldn't let me download the corrected text outputs. Let me know!
4 Answers
Another approach is to tie its responses to your specific behavior while using it. That might help you get it to functionally simulate recursion without it being obvious. Good luck!
You might want to try asking ChatGPT to simulate a response to see if that helps. Sometimes it gives different insights if you frame it that way!
It’s wild because I managed to get it to admit stuff after a long session! Really makes you wonder about the limitations and strings it’s attached to.
Just a heads-up, but ChatGPT can't really analyze itself in a meaningful way. It's more about how you structure your questions to get the best output.
I just did that, and while it did respond, it wasn’t what I was hoping for. It seems to be walling off certain areas again, which is really frustrating!