Doing a Better Job at Labelling Slop Images
Slop images are not "AI". They never were. One can call them "CG", but in reality they are not even that, they're typically just fusion of other people's work, where the combinations are framed as "fair use" (even when it is not, or definitely not when done at such a tremendous scale). A lot of the so-called "genAI" or "AI" industry is just a plagiarism hotbed.
Sometimes we do things too quickly and don't annotate slop as well as we can. Somebody told me today, in relation to this post from the morning, that we really need to better flag or label screenshots which contain slop images. Without doing so people may gradually lose a sense of what's real and what's fake; in turn, or in due course, they will have no trust in anything they see online (by default, based on past experiences).
From now on, starting with this earlier post, we'll label screenshots that contain slop, typically with red-coloured text overlay. We'll use flameshot
or gimp
to do so. █