>>104193480Sadly, yes. We're still not there yet especially with Chinese models. Deepseek is notorious for its inability to follow instructions so being hyper specific and expansive will actually get you nowhere. I've done experiments where I showed it paragraph blocks from novels that I liked and the effects were mixed/negligible. You could try doing more extensive experiments with future iterations of deepseek with say 8-10 paragraphs of example text. I only did 3-4.
What's bad is that major developers have abandoned creative writing as a frontier. They're all racing for coding and math. They all looked at their respective bots' writing styles and decided it was good enough. Besides, pushing this field apparently invites us, the rp crowd which they decided is market not worth prioritizing. Freezing the bot's writing style also gives it a consistent, recognizable "voice." If you would recall, being able to recognize AI outputs was in the AI ethics agenda a year ago.
>SEX WITH COLPHNE SEX WITH COLPHNE SEX WITH COLPHNE 