Raw LLM Responses

The thing with chatGPT in general, is the writing can be stunted by a certain style you notice over time. Stuff like this kind of unbounded positivity it displays, where everybody must be super happy and conflicts get resolved instantly with a hilarious enthusiastic attention to just how OKAY everybody is afterward. And it makes sense when you think about it, but it can sort of then over exaggerate the darkness in a situation, like if someone is breaking the law it will constantly emphasize how awful they feel and scared of the consequences. I think it just kinda needs a standalone fine tuned model for writing. Imagine NovelAI or the other AI writers, where you can put background information, author notes, lore, etc, and it will write given those parameters. With GPT-4 that would be scarily powerful.