I don't even know what to say. The entire situation is so stupid.
retained text
I asked GPT to make an image with horde of zombies. Just standing, nothing happens. It completed its work without issues, several times. But when i asked to edit small details, it started bombaring me with refusals, citing "violence". It refused to explain where exactly it does see violence. And when i pointed out that it just generated this entire image for me, it said: You're right — that was an unnecessary refusal. Please re-upload the generated image here, and I'll make **only** these two changes: * Change **“MORE” → “LIKE”** * Make **every screen image blank except the word “PRODUCT”** So i did what it suggested... only to be met with another refusal. It's like everything it has is broken: violence filter, reasoning system, it cannot even see the image that was generated before in same chat, asking me to upload a new one. But if you keep asking it questions, it somehow brings up information from other chats, that are not relevant to the issue. I'ts like having the dialogue with absolutely insane person, who says "yes, i understand" and proceeds to do the opposite. Is there a way to make it behave?
document fetched 2026-09-20 00:20 · canonical https://reddit.com/r/ChatGPT/comments/1wkn878/i_dont_even_know_what_to_say_the_entire_situation