The analogies fail because the copyrighted material were not used for creating the copy machine, Illustration, or (maybe?) the keyboard suggestion engine. If LLMs were produced ethically, then the whole discussion is moot. But if the only way to produce copyrighted material requires being trained on copyrighted material, then...
Copyright law applies to distribution of output, not input.
An artist, writer, whoever, could read all the copyrighted material in the world, even pirated material, unless their output is a copy or copyrighted artifact, then there is no infringement.
The image the AI generates is not copyrighted (except maybe by OpenAI I guess) unless it ends up being an exact duplicate of an existing image. Copyright applies to a specific work. The character may be trademarked like Mickey Mouse, but that is a different IP protection.
You are responsible for the output, just like any other tool.
If I use a copy machine to reproduce your copyrighted work, I am responsible for that infringement not Xerox.
If I coax your novel out of my phones keyboard suggestion engine letter by letter, and publish it, it’s still me infringing on your copyright.
If I make a copy of your clip art in Illustratator, is Adobe responsible? Etc.
What if the ceo of xerox went on social media and promoted copy machines by showing how you could use them for infringement?
Is that what is happening in reality?
It seems that all of the big players in the industry are perfectly fine with disallowing output that infringes on copyright.
The analogies fail because the copyrighted material were not used for creating the copy machine, Illustration, or (maybe?) the keyboard suggestion engine. If LLMs were produced ethically, then the whole discussion is moot. But if the only way to produce copyrighted material requires being trained on copyrighted material, then...
Copyright law applies to distribution of output, not input.
An artist, writer, whoever, could read all the copyrighted material in the world, even pirated material, unless their output is a copy or copyrighted artifact, then there is no infringement.
3 replies →
The image the AI generates is not copyrighted (except maybe by OpenAI I guess) unless it ends up being an exact duplicate of an existing image. Copyright applies to a specific work. The character may be trademarked like Mickey Mouse, but that is a different IP protection.
"Substantially similar" is the standard, not "exact duplicate".
I would argue yes, since the output is what you are probably after, when buying access to ChatGPT.