velvet hamlet
It looks like there is an issue with custom GPT models not recognizing the context of the ongoing narrative and treating subsequent user inputs as if they were new prompts. This can lead to abrupt changes in the storyline, causing confusion.
Which.. when you’re working on a scene of a story and need it to play some npc’s gets real interesting.. imagine starting in a pub and because you were talking to a teacher and said you were talking to a teacher in your response, the GPT then decides to take your response as a first prompt and responds by describing a class room.
Steps to reproduce:
- Set up a storyline prompt and receive an initial response from GPT.
- Respond to GPT's initial response to continue the storyline.
Expected result:
Receive a new GPT response that seamlessly continues the ongoing storyline.
Actual result:
Receive a GPT response that treats the user's input as if it were a new prompt, leading to disruptions in the narrative flow.
Additional information
Browser: Firefox 121.0.1; iOS GPT app
OS: Windows 10; iOS 17.2.1
Seems to only be happening GPT’s, GPT 4 doesn’t suffer the same fate, but has none of the background info I wanted for the story line.
hearty radish