Lately I’ve been reading a book called “Computer Models of Musical Creativity” by David Cope. Cope is a musician and programmer who has created software which composes classical music, usually within the style of existing historical composers. The method by which the software does this is complex – you basically have to read the book to understand it – but it does create “human sounding” music that is good, if not great.

One question I’ve had while reading the book is why Cope limits himself to (western) classical music. He explains…

Popular music, for the most part, relies on lyrics, particular timbres, performance context, and many other factors my program cannot control. The mere fact that we know most popular music by its performer, rather than its composer, should confirm the problem.

(Italics are Cope’s.)

He makes a good case, especially in regards to timbre. A song originally played on electric guitar but transferred to zither will not have the same impact. The electric guitar has a certain beefy, manly machismo that gets lost with other instruments. Cope’s point is that it’s not the notes themselves that drive, say, “Black Dog” by Zeppelin, but the notes combined with the guitar tone and various other factors and nuances of performance. On the flip side, Bach’s first invention is largely driven by the notes on the page (combined, one hopes, with a good performance.)

Nonetheless, I don’t see why Cope or some other programmer couldn’t create music that takes timbre into account. I could envision a music creation program that tracks trends in instrument timber and then predicts what will be next and generates some very hip music!

Here’s an example of some of Cope’s music. It’s a bit stiff as it is being rendered by a computer (as opposed to being played by a human (which it could be)), but gives you the picture.