You should read every single story about adverserial filters that break picture-bots with extreme scepticism. Remember dazzle facepaint that totes broke facial recognition? You know, that story that's still occasionally recycled, despite the fact it only ever worked against specific versions of specific facial recognition software, and was inevitably obsoleted by the fact of its own publicity?
Magic DRM is not going to save the world from shite generators.
(Also: whatever you do to artwork you're uploading now can obviously only affect affect some hypothetical future bot trained on a hypothetical future data set that includes your poisoned work. Y'know, assuming the adverserial filter is any more than a placebo against whatever specific version of whatever specific software that is.)
while I agree, I also think it's better to incorporate as many of these things as possible. Take a look at the US dollar bill:
it incorporates a fuckton of security features, each of which is publicly known, and each of which on their own is easy enough to break. A cloth printer isn't that expensive, neither is holographic film, nor even microprinting these days. But the net sum of these features still prevents the vast majority of fakes from coming to market. If we flood our art and our posts with all these filters, and we always wear our facial-recognition defeating wear, and so ona nd so forth, we may not make it impossible to break, but we will make it uneconomical to break.
Just a thought.
