On Wed, May 26, 2010 at 5:14 PM, Paul Davis <[email protected]> wrote: > > > On Wed, May 26, 2010 at 5:35 PM, Charles Henry <[email protected]> wrote: >> >> The degree to which computers can >> compose music depends on the success of modeling musical experience in >> humans. > > I'm willing to grant you the benefit of the doubt with regards this claim, > because I suspect you mean it in a different way than you wrote it. I don't > think that having a model of human musical experience is at all a necessary > component of computer composition, just as it isn't for human composition. I > feel that this can be said with some confidence given the fact that > *different* humans have wildly different musical experiences when presented > with the same material. There is no comprehensive model of human musical > experience, because there is no comprehensive human musical experience.
True, there's a lot going on and a lot of factors to consider. I should have said, "at some level of musical experience", because we all have a lot in common. Barring amusia or any other major hearing difficulty, people's sensory experience is much the same. Up to the primary auditory cortex, the auditory system is highly specialized. It's a neural architecture driven by evolution in pursuit of specific auditory functions, that precedes learning and exposure to music. For example, pitch perception is no longer considered a learned, template-matched response (as it has been debated since the 60s/70s), but is intrinsic to neurons themselves (see works by Julyan (JHE) Carwright and Dante Chialvo, for reference). At some level of conscious experience, we all hear the same things. However a powerful model should also be able to explain why people hear things differently. >> As musicians and composers, we approach the "tiling problem" >> with a set of techniques, instruments, and vocabulary. We are able to >> get direct, immediate feedback on the effectiveness of a giving >> tiling, which computers, at present, cannot. > > Not sure about this either. Its only been in the very recent past that human > composers could compose anything for an ensemble and get "direct immediate > feedback" on the effectiveness. In fact, a lot of the skill of the composers > of western classical music from the baroque era on seem to hinge on their > ability to *imagine* what the composition would sound like rather than have > any "direct, immediate feedback" on their ideas. I perhaps should have used the term "capable" rather than "able", b/c what I was really getting at was the complete lack of a computer's ability to say, "this sounds good" :) I would like a computer to be able to say, "This would sound good, if I were a human". Better yet, I'd like the computer to describe it to me in numbers that I myself could not calculate. There's certainly no point in having a computer tell me what I already know, because I was there, I heard it, and I know what I like. But I also can't listen to all the possibilities of music, though a sufficiently powerful computer could do so. _______________________________________________ Linux-audio-dev mailing list [email protected] http://lists.linuxaudio.org/listinfo/linux-audio-dev
