IP lawyer here: This reads like a bunch of technical arguments that have no interesting legal meaning. You already can't avoid copyright infringement by using

IP lawyer here: This reads like a bunch of technical arguments that have no interesting legal meaning. You already can't avoid copyright infringement by using statistical distributions whose output is substantially similar to the original. IE Lossy compression does not avoid infringement This is already law. The only question is how far does that reach and where fair use provides a defense to something that would otherwise be infringement. How transformative must something be, etc. The basic test for infringement is access to the work and substantially similar result. Intermediate means do not matter. You could have monkeys whose kids have genes that got passed along that enable them to redraw your work, so that they have no literal knowledge and have never seen it. If the output is the original work, its still infringement. The article also gets wrong that court cases also have been a (reasonable) continuation of plenty of conversations, from civil rights to you name it. I personally am not a fan of this litigation, but this author doesn't make good arguments about it