2026-09-25 04:40 UTC
@freya@social.highenergymagic.net you deserve tools that don't randomly fail in difficult to detect ways due to inherent features of the design, and that aren't built with the intention of sanitising and standardising human expression
Replies (1)
-
@freya@social.highenergymagic.net 2026-09-25 04:41
@Dangerous_beans@aus.social that would be nice, and I'd love to use tools that have all the capabilities of LLMs but none of the drawbacks, but, y'know, those don't exist. and no, pure classifier image describers are useless: "Image may contain: tree and sky and house and text" is not, in fact, helpful