[Smt-talk] Aesthetics of Computer-Generated Music

[Marcel de Velde]

Hello Nicolas,

> It seems to me that the case of computer-generated music is much akin 
> to that of fakes in art. This is all the more interesting that, 
> according to Nelson Goodman, fakes cannot exist in an allographic art 
> such as music. 

I feel more along the line with Michael Morse and Nelson Goodman here I 
think.


> What would be faked, in the case of computer-generated music, is its 
> intentionality. The extent to which such music could produce an 
> aesthetic effect strongly depends on the intentionality of the 
> listener. For this reason, I doubt that an experimental study would 
> produce any interesting result: the listener's answers would not 
> depend on the music itself, rather (a) of their awareness of it being 
> computer-generated; (b) of their opinion about this. I would think, 
> therefore, that the question must be approached from a semiotic and/or 
> philosophical point of view, not from a merely empirical/experimental one.
>
> Consider these cases:
> – At a not too recent conference in the Sorbonne, one of the papers 
> was read by a guy who had written a piece of software generating what 
> he thought was (good) tonal music. He explained that he had come to 
> love this music so much that he couldn't hear any other any more. The 
> examples he made us listen to where awful – or so thought several of 
> us. Obviously, neither this guy nor any of us judged the music on its 
> inherent aesthetic value: we were guided by our personal convictions.
> – Suppose that a real piece of music, written by a real composer, is 
> presented as computer-generated. Many a listener might dislike it 
> merely on the basis that it is (erroneously) thought to be a fake.
> – Inversely, suppose that a computer-generated work is made to pass 
> for the work of an interesting forgotten composer: many critics will 
> praise it and praise the rediscovery of an unjustly neglected master.
> – Etc.

I personally hope the music itself would matter most of all.
But indeed knowledge of how it was composed will be influential to at 
least some degree in many people (perhaps strongly sometimes indeed).
But I'd find the effects in a blind listening study as Sandeep Bhagwhati 
requested much more valuable.
Sadly I'm not aware of any such study.



>
> I do believe that common-practice tonality can be modelized with much 
> more satisfying (and more precise results than what Marcel de Velde 
> believes. Examples do exist (e.g. Mario Baroni, Rossana Dalmonte and 
> Carlo Jacoboni's Legrense software described in their /Regole della 
> musica/, successfully modelizing arie by Legrense).

Thank you, I was unaware of Legrense. I will look it up.
If it is indeed done truly successful, depending on how it's done and 
written, that may weaken my point that one needs just intonation to do 
successful algorithmic composition :)
Perhaps this music is suitable for the kind of study Sandeep Bhagwhati 
is looking for.
But the first thing I'm thinking about is how did they write the 
algorithm, is it still to be seen as a computer composition should they 
do something more along the lines of "remixing" existing arie by Legrese?


>
> I don't think that intonation has any important share in this matter. 
> After all, there exist recordings of, say, /The Art of Fugue/, on 
> early synthetizers (Moog) which played in ET: I did not feel that the 
> music suffered so much. The problem remains that just intonation isn't 
> really usable in tonal music because the directionality of tonal 
> harmony produces an unavoidable shift in pitch in just intonation. I 
> suppose that one might construct a harmonic functionality that would 
> balance the shifts in pitch, and that just intonation in that case 
> might produce some sort of consonant effect that might seem of 
> aesthetic value; but that would not be tonal music.
>     Marcel, your experiment in just intonation seems to me to sound 
> much more like "modal" polyphony of the 16th century than like tonal 
> music, precisely because your algorithm probably makes no provision 
> for a tonal directionality of the harmony – and because limiting the 
> range to the 5-limit-harmonic did not confront you to the problem of 
> having to prepare and resolve 7ths, which one of the main causes of 
> tonal directionality.
>
>
> Yours,
>
> Nicolas Meeùs
> Université Paris-Sorbonne

Aah Wendy Carlos and Bach, lovely music :)
It was later redone by Wendy Carlos in several more historically correct 
(Bach most likely did not use ET contrary to popular belief) and nicer 
sounding temperaments.

I agree that intonation does not have a very important share (though 
some will disagree here) in the rendering of common practice music.
At least 12 tone equal temperament comes close enough for the ear / 
brain to interpret it correctly (it just sounds less good, colourful and 
emotional)

The potential importance of just intonation is not about how it sounds 
(though how it sounds starts becoming more important for certain things 
that 12tet does not approach closely enough, though one could use 24tet, 
31tet or 53tet etc in these cases, for instance for some arabic tones).
But the potential importance of just intonation is for understanding how 
the tonal side of music functions. Therefore being of importance to 
composition in the first place.

Btw the experiment in just intonation which I posted is not in correct 
just intonation as I've written in the original post.
The example goes out of tune very often.
It was merely a small experiment and a wrong one at that, though it did 
get some parts right in hindsight.


About an unavoidable shift in pitch in just intonation, this is not true.
There are many many systems which claim to be just intonation. And if I 
didn't miss any then I've researched almost all of them and re-tuned 
common practice music to almost all of them.
There are systems in which for instance every major chord is said to be 
1/1 5/4 3/2 and every minor 1/1 6/5 3/2 (most commonly known ones) and 
they produce both comma shifts in held notes, and produce drifts on the 
5*5 axis. It also sounds absolutely horrible and completely 
unacceptable. And makes no sense musically or mathematically.
You can find a variant of this in some automatic tuning software, 
sometimes called adaptive JI.

Then there's a combination of 5-limit and Pythagorean in which a major 
triad can be 1/1 5/4 3/2 and 1/1 81/64 3/2 depending on how it connects 
to other chords, and no 5*5 plane (dominant 7th is most often 1/1 5/4 
3/2 16/9 here). This variant will actually work without wolves or comma 
shifts or drifting and sounds pretty good (much better than 12tet or 
Pythagorean when done with some care)

Then there's extended JI in which for instance the dominant 7th is 1/1 
5/4 3/2 7/4 (correct for some jazz or blues, but horrible in common 
practice classical, it's not even dissonant) and even worse when ratios 
like 11 and 13 are used, making common practice music sound like some 
drunken arabic polyphonic comic haha. (however, the 19th harmonic for 
the minor third is good in some functions)

The above systems are flexible, that is they are not fixed scales but 
scales of potentially infinite size in which the music moves according 
to a certain system for instance lead by the movements of the 
fundamental bass.
Then there are countless variations which assign a fixed scale to the 
the tonic or key. They produce wolves in unacceptable places (with the 
exception of Pythagorean since our notation was based on it).

There's a popular belief that just intonation is a mathematical 
impossibility.
But this is only true if one starts with very rigid and artificial rules 
as a starting point.
Once you allow the music itself to indicate it's tuning and the function 
of it's tones, then an almost infinite world of possibilities opens up.
And doesn't just intonation make so much more sense than for instance 
equal temperament?
An octave is 2/1 in 12tet, but a fifth is not 3/2, and a major third in 
an ending chord not 5/4 but irrational numbers (infinite length). They 
have no relation to their root in any way in 12tet.
Surely music must be just intonation at it's heart, it's even how we 
naturally sing and play on instruments that allow it (with some degree 
of error).
I do not think that music demands the impossible of us. I think that in 
all these thousands of years we still have not uncovered it's greatest 
secret of all.


Kindest regards,
Marcel de Velde
marcel at justintonation.com
Zwolle, Netherlands

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.societymusictheory.org/pipermail/smt-talk-societymusictheory.org/attachments/20110409/84d6b0db/attachment-0003.htm>