I haven't heard anyone saying that copilot is legal "just because it's AI." That's a pretty bad faith, reductive, and disingenuous representation. The core argument I've seen is that the output is sufficiently transformative and not straight up copying.
I wasn't really trying to address whether the argument is valid, I was just noting the representation of the other side here is reductive to the point of being in bad faith. I find that kind of rhetoric a little frustrating since it's kind of inflammatory, and, I believe, not particularly productive towards having honest/informative disagreements and discussion.
I think if another algorithm was used instead of ML that did the same job as Copilot, then people would be making the same arguments. I think it's just the case that ML is just the first tech capable of doing what Copilot is doing.
You can't copyright an algorithm, you can copyright a particular expression of one, or you can attempt to patent an algorithm, but two authors can legitimately write the same thing and not infringe on each others copyright unless one copied from the other.
Suppose you own the rights to a jpeg. And I apply a simple algorithm that increments every hex value. So 00 becomes 01 and so on. The gibberish images it spits out would be so different then your original image that you wouldn't have any claim to them at all.
So I may create a tool that is capable of "incrementing every hex value" of an image, and also of "decrementing every hex value", and than distribute any of your images after "incrementing the hex values", together with said tool, right?
Or maybe it would be enough to just zip your image, to be allowed to distribute it? In the end the bytes I would distribute than "would be so different then your original image that you wouldn't have any claim to them at all", right?
I encourage you to go get a copy of the latest hollywood blockbuster, apply your transformation, share it on the internet and see if the courts agree with your copyright hack.
