A reminder:
even if training AI models on copyrighted works without permission of the author turns out to be legally ok, and under the "fair use" exceptions, you are still not allowed to create and disseminate copyright protected works with it.
So, even if Open AI is in the clear for using Calvin and Hobbes in their training data, the user prompting this
comic strip is not. The "it is so easy" defense wouldn't stand in court.
I don't think that if the author would want to sue the person making the fake Calvin and Hobbes comic, the "but your honor, I gave a prompt to the tool that explicitly orders it to make content that infringes on copyright, so it is not my fault but OpenAI's!"'
Just because a tool makes it easy for you to make copyright infringing stuff, it doesn't absolve you from making it.
+bonifartius 𒂼𒄄
@nicemicro yet, if i were to base64 encode a movie and upload it, i would be in jail even when the decoding is done by others.
either copyright is bullshit or AI is infringing.
@bonifartius A key difference:
a base64 encoding is only made from that specific work and it can only be decoded to that specific work. The AI model itself does not have the work itself encoded in it, rather, descriptions of millions of works.
If, for example, you made a book which contains a description of how Calvin and Hobbes look like, among 1000 other characters, that is legal. Using that book to create infringing new images is not.