4 ms·

What if you overfit your model to the point of exact reproduction? Or anything in between that and what you consider inventive. Where is the line drawn.

Permit 3y ago
> What if you overfit your model to the point of exact reproduction? Or anything in between that and what you consider inventive. Where is the line drawn. The line is the same as it always has been. If you as a human publish this work then you have to respect copyright. If it is an exact replica then it likely violates copyright regardless of whether or not it was produced using AI, by copying pixels or by human hand. The test is the same regardless of how it was created.
clnq 3y ago
I don’t think the tool by which a derivative work is created matters. The author should not distribute, remix, re-work, adapt, release, perform or synchronise it if they do not have the rights. This is already the case for all other technologies and tools. For example, if you rewrite a popular book in a word processor from scratch, it is not the responsibility of your word processor to not let you do it. Or the government to regulate word processors so that they are incapable of doing it. It is your responsibility to not distribute what you made. If you record your screen watching a movie and distribute it, the accountability for this falls on you, not the tool made for screen recording. And if you use generative AI to produce trademarked or copyrighted works, you should be accountable, not AI. Generative AI will always be able to produce copyrighted works if steered enough. Even if it wasn’t trained directly on them. A very simple experiment proves it - you can paste copyrighted material into a ChatGPT prompt and ask it to repeat it. Or you can describe a copyrighted work really well for MidJourney and it will produce results with high likeness. This does not require that the models be trained on copyrighted works. Besides, copyright infringement isn’t nearly the worst crime AI can be used in. Serious impersonation, forgeries and fraud are also made easier with AI. Why is copyright different? It should all be treated the same - the user should be held accountable for their actions. Not the tool or the tool’s makers.
virgildotcodes 3y ago
Hey, I think this is a good point. The genuine question from me is how is this currently handled with humans? How closely can someone emulate a copyrighted work before they start being in danger of violating copyright? Maybe we apply the same standards and the regulations then come down on the side of output, and those who choose to commercialize it, being liable if they’re too close to reproducing copyrighted works. Then of course that brings up the fact that the prompter may not be aware, so how could they be liable? Can we really expect everyone using these AIs to be familiar enough with the sum total of the human creative corpus to be able to identify copyright infringements? Would a lack of intent to emulate Mickey Mouse be a good defense against unknowingly recreating him as your corporate logo? Probably not. Yeah, it just seems like a giant mess that is only cleanly resolved by either banning training on copyrighted works without consent or else eventually eroding the entire concept of copyright. Are there other options?

Permit 3y ago

> What if you overfit your model to the point of exact reproduction? Or anything in between that and what you consider inventive. Where is the line drawn. The line is the same as it always has been. If you as a human publish this work then you have to respect copyright. If it is an exact replica then it likely violates copyright regardless of whether or not it was produced using AI, by copying pixels or by human hand. The test is the same regardless of how it was created.

clnq 3y ago

I don’t think the tool by which a derivative work is created matters. The author should not distribute, remix, re-work, adapt, release, perform or synchronise it if they do not have the rights. This is already the case for all other technologies and tools. For example, if you rewrite a popular book in a word processor from scratch, it is not the responsibility of your word processor to not let you do it. Or the government to regulate word processors so that they are incapable of doing it. It is your responsibility to not distribute what you made. If you record your screen watching a movie and distribute it, the accountability for this falls on you, not the tool made for screen recording. And if you use generative AI to produce trademarked or copyrighted works, you should be accountable, not AI. Generative AI will always be able to produce copyrighted works if steered enough. Even if it wasn’t trained directly on them. A very simple experiment proves it - you can paste copyrighted material into a ChatGPT prompt and ask it to repeat it. Or you can describe a copyrighted work really well for MidJourney and it will produce results with high likeness. This does not require that the models be trained on copyrighted works. Besides, copyright infringement isn’t nearly the worst crime AI can be used in. Serious impersonation, forgeries and fraud are also made easier with AI. Why is copyright different? It should all be treated the same - the user should be held accountable for their actions. Not the tool or the tool’s makers.

virgildotcodes 3y ago

Hey, I think this is a good point. The genuine question from me is how is this currently handled with humans? How closely can someone emulate a copyrighted work before they start being in danger of violating copyright? Maybe we apply the same standards and the regulations then come down on the side of output, and those who choose to commercialize it, being liable if they’re too close to reproducing copyrighted works. Then of course that brings up the fact that the prompter may not be aware, so how could they be liable? Can we really expect everyone using these AIs to be familiar enough with the sum total of the human creative corpus to be able to identify copyright infringements? Would a lack of intent to emulate Mickey Mouse be a good defense against unknowingly recreating him as your corporate logo? Probably not. Yeah, it just seems like a giant mess that is only cleanly resolved by either banning training on copyrighted works without consent or else eventually eroding the entire concept of copyright. Are there other options?