After people reacted so positively to the banner I made I had the idea to do the same with c/anarchism. But what to draw? Well I took the existing image and traced over it, adding some more character.
License (as always): CC-0, no rights reserved. Especially since it’s traced over a generated image.
Those regular people aren’t here. And If they were what would you want them to see? People arguing over “stealing” and “copyright”, completely sidestepping the actual point of the conversation like Mystic was. Or people warning each other over using corporate models and replacing artists.
From now on whenever I make something AI related I’m going to say “Don’t use corporate AI. Hire actual artists.” As the first line to combat this. Because, as I hope you’ve seen, arguing on whether or not this is stealing will simply go nowhere.
These models exist, and normal people will see generated content as every other platform is full of it. Allow this one space to be about doing image generation right: “No destroying the planet or replacing humans”.
As with consent. I think training falls under fair use. You aren’t going to be able to get an exact replica from a model anyway so any output I consider transformative, similarly to how a human brain transforms content (although obviously the internals completely different). For me the argument that you need consent to train a model is the same as arguing you need consent to train yourself, or cover a song, or trace over an image. Yes the transformation is mechanical, but if a human was doing all those calculations by hand it would be considered art, and no-one would question if they had the right.
Slight tangent as you mentioned music. I actually like generated music. Like the sound of it. It’s got the vibe of melting a bunch of different songs together to make something completely unoriginal but still different. Real music is obviously superior in every way, but when you want something like that one song that’s unlike any other and you just wish it were longer, that’s what generated music is for. It can’t replace any actual musician but it can provide an interesting listening experience… For like an hour at a time before it becomes grating.
But the point I want to make is that generated music does not sound like music. It’s something different. So arguing that it’s somehow plagiarising just seems silly to me. The model transformed it. It’s not the same, audibly.