I agree that after a download is complete, a copy has come into existence, and it is located on the downloader’s computer. But, the downloader did not have the work prior to downloading. How can he make a copy of something he does not yet possess? What is the “original” from which this copy came to exist? Who had any obligations under copyright law regarding that original?
Unless you can point where the law says you have to make the copy from a copy you posses it is irrelevant.
But we do actually have precedent where there was creation of copies out of thin air. VHS recordings of broadcast, Sony Corp. of America v. Universal City Studios. It was actually settled on time-shifted of free-aired material being fair use. Nobody argued that the VCR owners having no copy before recording did not make a copy.
No, Silverman’s argument is not that the mere possession of the work by ChatGPT violates copyright, because what question has long since been answered: the artist controls the work, not the audience. The artist cannot decide who is and is not allowed to consume the work. Regardless of how someone came to consume the work, they are fully entitled to speak about it.
I will concede that there are situations where you can just consume copyrighted material without copying them (which downloading is). That would be if you I downloaded a movie and invited you to watch it, or a sports bar showing illegal streams.
My whole point is that it does not matter if you have committed copyright infringement, you can always make fair use derivative works such as reviews. I could get DVDs from a friend in the 00s and copy them to a my own disc before watching the copy. That would mean I infringed even in your wrong understanding of copyright. If it was worth it and there was evidence of it the copyright owner would be able to successfully sue me for copying them.
He could correctly argue that me copying the disc, infringing on his copyright, was necessary for me to write a review of his movie, a derivative work. It would not matter.
I could later make another film that is inspired by the movie whose copyright I infringed upon. If the movie is not too similar it would not be itself infringing. If it too similar it could be infringing but so would a movie made by someone who committed no copyright infringement to be able to watch the original.
This is what the discussions was about. AI opponents push the idea that if there was copyright infringement on the training process, any output of AI must be infringing or derivative of the original work. Which is bullshit.
I suppose you are not pro-copyright, same as me, but you are not helping any argument by making claims that are besides the point and wrong.