“Damage to a computer” is legal logorrhoea, possible interpretations range from not even crashing a program to STUXNET, completely under-defined so it’s up to the courts to give it meaning. I’m not at all acquainted with US precedent but I very much doubt they’ll put the boundary at the very extreme of the space of interpretation, which “causes a program to expose a bug in itself without further affecting functioning in any way” indeed is.
Which is fun because the people stealing from you face absolutely no retribution at all for their theft,
Learning from an image, studying it, is absolutely not theft. Otherwise I shall sue you for reading this comment of mine.
So any art done in a style of another artist is theft? Of course not. Learning from looking at others is what all of us do. It’s far more complicated than you’re making it sound.
IMO, If the derivative that the model makes is too close to someone else’s, the person distributing such work would be at fault. Not the model itself.
But again, it’s very nuanced. It’ll be interesting to see how it plays out in the courts.
Of course not, but what does this have to do with generative models? Deep learning has as much to do with learning as democratic people’s north Korea does with democracy.
It does not get damaged, it stays as it is. Also it’s a bunch of floats, not a computer.
But making works derivative from someone else’s copyrighted image is a violation of their rights.
“Derivative work” doesn’t mean “inspired by”. For a work to be derivative it needs to include major copyrightable elements of the original work(s). Things such as style aren’t even copyrightable. Character design is, but then you should wonder whether you actually want to enforce that in non-commercial settings like fanart, even commissioned fanart, if e.g. Marvel doesn’t care as long as you’re not making movies or actual comics. They gain nothing from there not being, say, a Deadpool version of the Drake meme.
“Damage to a computer” is legal logorrhoea, possible interpretations range from not even crashing a program to STUXNET, completely under-defined so it’s up to the courts to give it meaning. I’m not at all acquainted with US precedent but I very much doubt they’ll put the boundary at the very extreme of the space of interpretation, which “causes a program to expose a bug in itself without further affecting functioning in any way” indeed is.
Learning from an image, studying it, is absolutely not theft. Otherwise I shall sue you for reading this comment of mine.
The model is the thing of value that is damaged.
But making works derivative from someone else’s copyrighted image is a violation of their rights.
So any art done in a style of another artist is theft? Of course not. Learning from looking at others is what all of us do. It’s far more complicated than you’re making it sound.
IMO, If the derivative that the model makes is too close to someone else’s, the person distributing such work would be at fault. Not the model itself.
But again, it’s very nuanced. It’ll be interesting to see how it plays out in the courts.
Of course not, but what does this have to do with generative models? Deep learning has as much to do with learning as democratic people’s north Korea does with democracy.
It does not get damaged, it stays as it is. Also it’s a bunch of floats, not a computer.
“Derivative work” doesn’t mean “inspired by”. For a work to be derivative it needs to include major copyrightable elements of the original work(s). Things such as style aren’t even copyrightable. Character design is, but then you should wonder whether you actually want to enforce that in non-commercial settings like fanart, even commissioned fanart, if e.g. Marvel doesn’t care as long as you’re not making movies or actual comics. They gain nothing from there not being, say, a Deadpool version of the Drake meme.