Len writes about computer generated music:

>my sentiments as well.  Until we understand exactly/objectively what it
>is that causes us as human beings to perceive (i.e., subjectively) a work
>of art as "moving" or "meaningful" or "transcendant", we have only random
>choice within a set of structural principles to guide algorithmic
>composition.  And though that set of structural principles may have grown
>considerably in scope and complexity over the past fifty years, our
>understanding of what choices within that space make music "great"
>remains primitive.

I don't think that the great composers understood this either, yet they
wrote great music.  I assume they thought "this sounds great to me so
others will like it too" but they did not know why from a scientific "how
the mind works" point of view.

My software knows the rules about melody construction, what works and what
doesn't in the same way that the deaf Beethoven knew what would sound good
and what wouldn't.  Same for harmony.  There are a great many rules about
form, harmony/melody, orchestration that work even if we don't know why.
By randomly combining these things, the chances of getting a good work are
a lot better than the chances of winning the lottery.

My generated music works on this principle:

It uses the approach taken by those who photograph models for magazine
covers.  The model does not pose, she (or he) begins a series of movements
with her body, throwing her hair back, smiling, twisting, throwing her arms
up in the air.  The photographer is busy rapidly taking photographs of her.
Then, the photographs (usually 36) are developed and laid out on a table.
The photographer then looks for that great picture that could have never
been posed for.  The camera caught her head going back, with her hair in
the air, and that certain look in her eyes and that certain smile at just
the right time with her body and arms in just the right position.

Bill Pirkle