Good, hopefully it irreparably taints all models
Ok? The datasets already exist. Models have been trained. “poisoning” your images isn’t going to change the fact that anyone can use these still.
I wonder how one would work for LLMs that are also being trained off intellectual property. Flooding it with nonsensical references maybe?
Why do Lemmy users suddenly care about copyright and intellectual property when it’s bad for AI? What do you have against AI? It’s a really useful tool that democratises creativity.
I don’t give a shit about IP. My computer has three hard drives full of pirated material and it’s only because I don’t need a fourth yet. However, I obviously oppose AI because that “dEmOcRaTiSaTiOn Of CrEaTiViTy” is an assault on creative labour in the same way the powered loom was on artisanal. If you aren’t standing up for yourself as a worker you’re a fucking dog who’s distracted by the toys you’re given by your betters.
edit: But to clarify approaching it from the standpoint of IP is just the most useful weapon against it. There’s the broadest base of support for the broadest level of regulation short of outright banning AI models, and I’m also for that.
Why do Lemmy users suddenly care about copyright and intellectual property when it’s bad for AI? What do you have against AI? It’s a really useful tool that democratises creativity.
Creativity was already democratized, you just have to practice.
I think it is more about the corporate interest looking to squeeze the ever additional dollar from everything, and in this case in particular, from an already (mostly) struggling community.
