I'm not disagreeing with "let them do it", but the comparison with computer music isn't really fair.

Computer music, as it existed a couple decades ago, still played exactly what you asked it to, and it wasn't filling areas where you underspecified the music with a statistical model of trillions of existing songs. And that's the difference, for me: the ability to underspecify, and have the details be filled in and added in a way that to the audience will be perceived as intentful, but which is not.