If this is really the case, then it seems to be a loophole for a reverse scenario. You write a computer algorithm to generate vast collections of music. Since the computer, pretty much by definition, is ignorant of other music then you can claim copyright over everything it generates. Wait for it to produce "copies" of popular melodies and then you can publish cheap clones or license popular songs to people for a fraction of the cost of the originals.

If there is an objection that the computer isn't considered a person, then form a corporation and copyright the songs to the corporation. The corporation is equally ignorant of any other music. Human judgment is only exercised by a second corporation that searches the first corporations extensive database to find popular melodies to license.