> It's meaningless to argue about 'levels of similarity', in a legal context. The question is whether the second author based himself on the artistic values that the original author put in the work.
That's an excellent point, and I understand that. But as a practical matter, it seems like "levels of similarity" would be at the heart of an argument that document B was "based on" document A. Having direct evidence that B was based on A seems a less likely scenario, although a legally much more straightforward one.
To take your play example, I may assert your script was based on my book, but if you deny it I'm stuck with trying to demonstrate that the similarities between the two are above and beyond what we'd expect from mere coincidence, right?
> So if someone blends two works into one (without permission), he's violating 2 author's copyrights.
This is more in the realm of philosophy than legality, but what I was getting at is this: I think there's a pretty strong and conventional argument to be made that much of what we call "creativity" is really a synthesis of pre-existing works and pre-existing ideas. At some point we start to consider this synthesis more inspiration than plagiarism. Intellectually I think it is interesting to consider where that line lies. (And, taking us back to the world of algorithms, whether you could claim, for instance, that a sophisticated enough markov-chain, given a large and diverse enough body of input, is eventually creating new works rather than simply synthesizing existing ones.)