At one point does AI recreating patterns it has seen from reading source code count as a derived work? What if a human learns to code by reading only GPLed code, does all the code they write fall under GPL as a derived work now?
This has some interesting implications – for example, it means I can't mirror somebody else's (open source) code on GitHub without their explicit agreement.
So any code uploaded by someone other than the copyright holder renders someone liable to be sued for copyright infringement, AFAICS. The only question is whom it makes liable -- the uploader, GitHub (=Microsoft!), or both?
I can see arguments either way: The uploader is clearly infringing by giving away a right that isn't theirs to give. But so is GitHub / Microsoft, for using a "right" they haven't been properly given. So I'm provisionally leaning towards "both".
> I can't mirror somebody else's (open source) code on GitHub without their explicit agreement.
Who is doing the "mirroring" -- you, in uploading the code, or GitHub / Microsoft in actually hosting it, keeping it available for download from their "mirror"[1] site?
___
[1]: Is that even the correct terminology nowadays, when AIUI for lots of projects GitHub is their primary code repository?
Also, there's no way for anyone to know what portion of code that I commit was hand written vs. generated, so you kind of have to treat it all as written by the committer anyway.
Though this does bring up interesting questions about what happens with things like automated PRs that fix bugs / update dependencies... are those then non-copyrightable? ¯\_(ツ)_/¯
Just as much as a hero riding off into the sunset is not copyrightable in a movie script. However, a hero riding off into the sunset with bananas in the pistol holsters would be.
This is what I would want to hear more about when discussing if Copilot violates copyright.