Bill Pirkle replies to me:

>>I truly doubt that given one's knowledge of Brahms and Wagner, that one
>>could have predicted Mahler or Sibelius.
>
>I don't really know, not being acquainted with all CM, but I think that we
>can see trends like
>
>Melodies are containing more and more foreign notes,

Foreign in what sense?  To the "key?"  What if there is no key?

>Harmonies are tending to be less and less tonal center oriented

If they're harmonies, as opposed to simultaneously-sounding non-tonal
sonorities, then they're tonal-center oriented.

>Rhythm is getting more complex Form is getting more obscure

I wouldn't say "more obscure" but "more organic." While we have
sonata-allegro and the traditional forms based on dance, a lot of pieces
(not just vocal ones whose form is determined by text) whose form is sui
generis.  Of course, we've had this since the Renaissance at least.

>Non-traditional sounds are more widely in use

My original question had to do with the computer being able to surprise
a listener - that is, could it come up with something non-determinable?
I used the following example:  Given Wagner and Brahms, Mahler or Sibelius
could not have been predicted.  That is, although Mahler, for example,
derives to a certain extent from Wagner, one can't say that Mahler's idiom
sounds much like Wagner's.  In short, you can easily distinguish the music
of both (except for the second movement of the 8th symphony).  In that
sense, Mahler surprises you if you only knew Wagner before.  The computer
can be programmed to write the programmer's idea of any composer you like,
but can it create something new? That is, if you programmed the computer
to yield your idea of Wagner, would the computer be able to come up with
a "Mahler" or some other sport? That's the "Turing test" I'd give the
software - its ability to mimic the surprise.

Steve Schwartz