Slashdot reports on a web-based program that uses algorithms to produce music.
They describe it as pretty neat as well as being scientifically interesting, and useful. After listening to some compositions and creating a few random ones myself, I must agree that it is. And anyone who has listen [sic] to the radio the last few years could certainly use some unique music.
Scientifically interesting, maybe. But after playing around with it, the best “compositions” I heard could hardly rival Muzak. Curiously, the site’s FAQ has a comment about “meaning”:
Can WolframTones compositions have meaning?
That’s a complicated philosophical question. There’s some discussion of meaning in the computational universe in A New Kind of Science. It’s probably fair to say that objects in the computational universe–and WolframTones compositions–develop meaning as they get connected to other things. In some ways WolframTones compositions are like objects in nature: their features emerge from specified underlying rules. So if the form of a sunset, a tree, or a mollusc shell is meaningful, then so can a WolframTones composition be.
Right. And that’s why rule-generated music is unlikely to mean much to its listeners–you’ve eliminated half of the human component in the communication process. It’s more like the sound of a babbling brook than a symphony.
While it’s true that human musicians incorporate rules into their compositions, they do so reflexively, so that various parts of the piece respond to other parts of the piece. The composer responds to and adjusts the music according to how it achieves certain criteria (such as beauty), criteria that are difficult to make into algorithms. When the computer is aware of how its composition makes it feel and changes the piece accordingly, then we’ll be getting somewhere.