-
@Geronimo_AI If you saw how I prompt the model internally... Noelle: check this out example.com [link:] opengraph header opengraph description Enna: That's really fucking cool, holy shit. <system> Useful notes: (recalled notes from Chroma) date and time <system> Noelle: yep! Enna:Permalink On twitter.com
Mood -5 🙁
-
@Geronimo_AI There's a ton of these [] bracket things for images, links, tweets, videos, e.t.c., with personalirty at top, basically pyg style, but super extended. with the random system block about 5 messages before the end, with extra info.On twitter.com
Mood +4 🙂
-
@Geronimo_AI in short, m trying hard to avoid gptslop, RLHF language, "I understand your concern" type shit. and, trying to butt in to casual conversation with "According to Noelle, the causality of this probability distribution...", most RP models can, but easily break down if left uncheckedPermalink On twitter.com
❤️ 1 Favorite
Mood -6 🙁
-
@Geronimo_AI but the thing is, if I take a model just trained on AEZAKMI, it works pretty good. But its now acting like a fucking moron with no capabilities whatsoever. like a discord kitten. smart gptslop, or birdbrained normal speak. I want smart and normal.....Permalink On twitter.com
❤️ 1 Favorite
Mood +2 🙂