Bill Pirkle wrote:
>I have read his book and I agree whole-heartedly with the above statement..
>I think music is may be the only art form that computers could create at
>today's level or technology. For a computer to try to write poetry or a
>novel would indeed require a genuine "understanding of language". For a
>computer to paint a picture would require an understanding of special
>relations, form, structure. etc. But music is different in that there
>are recognized rules for harmonizing melodies, harmony in general, form,
>like sonata for, rondo, simple binary, ternary, etc that can be employed
>to produce music. We don't understand music like we understand language,
>that's why we can't describe it verbally, as many posts on this list
>indicate, we can't even agree, verbally, on the fundamental issues
>involved. That makes music special because in the end it does not have
>to be understood, it just has to be pleasurable to listen to.
This is the clearest statement of your opinion that I have read to date.
We certainly can describe musical events verbally, and symbolically in
different ways. What we can't do is give a complete verbal mapping of the
"meaning" or "content" of the music. But that doesn't mean that there are
not intrinsic musical "meanings" based on years and centuries of composer
preferences, listener feedback, or whatever. So a composer wishing to
write tonal music is aware that a Neapolitan 6th chord, or a chromatic
modulation to the mediant, or any other significant device, will be
received as a particular message in context--expected, surprising, etc.
and writes it with that purpose in mind. So what I would call musical
understanding is perhaps being sensitive to the way in which composers use
harmony, melody, form, dynamics, silence, etc. in order to create
significant form, or at least the illusion thereof.
The great challenge to the computer programmer wishing to specify
the algorithm for musical composition is how to use what has already
happened in the music as an influence on what is yet to come. So
the first phrase will exert an influence on how the next is heard (as
consequent? repetition? continuation?), and both of those will affect what
makes sense in the third, and so on. But even within phrases, such things
as chord type frequencies and probabilities that chord Y will follow chord
X, when chord X is preceded by W, or perhaps V, and so on, must be
accounted for. So if you really want a computer to be able to make the
same kinds of decisions that a human composer would make, it must have
a repertoire of probabilities for continuation that is equal to the
composer's memory of what can be done in a particular style. You can
certainly fake it up to a point, but I don't really see anyone flunking
a Tuering test on a computer piece as yet.
By all means show me your work and let me decide for myself whether it's as
good as a composer in that style could write.
Chris Bonds
|