Hey everyone! I've been experimenting with ChatGPT-4o for a visual design project, and I've run into a pretty interesting issue that I thought might be worth discussing, especially for those of you working with detailed prompts in creative workflows. I'm creating a series of AI-generated posters that either play with or reflect the essence of individual words using bold vector designs. For instance, I worked on an ironic 'SADNESS' poster that used bright colors to subvert the theme. Later, when trying to create a poster that truly conveyed the essence of a different word, I noticed something strange: despite my precise new prompt, the AI produced another subverted output instead of sticking to the new instructions. It seems like the AI was pulling context from my previous project instead. Has anyone else run into this issue where the projects seem to bleed into each other or the AI seems to ghost previous interactions?
1 Answer
I think I’ve faced something similar! Sometimes it feels like even though you've given clear prompts, the previous context lingers and affects the new outputs. I’d suggest checking your settings and try temporarily turning off 'Reference Chat History.' It might help with clarity, but I get that you might prefer keeping that feature on for personalization! Just seems like the AI needs to prioritize the specific project context more effectively.
Thanks for the tip, but I prefer having that feature on! It feels like the AI should be capable of handling distinct prompts without mixing them up. It could be leading to some confusion or even hallucinations when previous chats bleed into the current one.