Tuesday, July 10, 2012

The Computer as Composer

Just a week ago there was an interesting premiere from the Centro de Tecnología de la Imagen - Universidad de Málaga. The composer of the music performed is a computer program named Iamus. Here is an article in the Guardian on this and on an upcoming CD to be released in September. Here is a link to a video of the recent concert. For your amusement, you can take a "Turing Test" in which you have to identify which of five musical clips was composed by Iamus. Here it is. Don't worry if you guess wrong--I did!

The article in the Guardian has a brief history of attempts at computer composition in the past and the meager results, accurately described as "Michael Nyman on a bad day." Of course, I know some music by student composers that also sounds like Michael Nyman on a bad day! One of the things we can say is that the task of creating a computer program that can compose music is rendered much easier if, instead of trying to imitate Bach or Beethoven, you use as your model Stockhausen or Ligeti. Due to the kinds of procedures some composers developed in the 20th century, which include aleatoric ones (follow the link for the details, but it boils down to using chance procedures), just writing a program that generates random notes and rhythms could easily sound like something by Cage or Xenakis. Why? Because they used random procedures.

Now obviously the computer program Iamus is doing something quite different. The interviews with the people involved (which come before the concert in the clip) indicate that they are using 'biomimetic' procedures. Here is quote from the Guardian article that reveals a bit about how it works:
Iamus – named after the son of Apollo who could understand the language of birds – composes by mutating simple starting material in a manner analogous to biological evolution. The compositions each have a musical core, a "genome", that gradually becomes more complex.
"Iamus generates an initial population of compositions automatically," Vico says, "but their genomes are so simple that they barely develop into a handful of notes, lasting just a few seconds. As evolution proceeds, mutations alter the content and size of this primordial genetic material, and we get longer and more elaborated pieces." All the researchers specify at the outset is the rough length of the piece and the instruments it will use.
"A single genome can encode many melodies," explains composer Gustavo Díaz-Jerez of Musikene, the Higher School of Music of the Basque Country in San Sebastian, who has collaborated with the Malaga team since the outset and is the pianist on the new recordings. "You find this same idea of a genome in the western musical canon – that's why the music makes sense."
Does the music "make sense"? The program is certainly generating themes that return and are varied, such as the trill figure in the piano piece Colossus, the first one in the concert clip. But though recurring themes are a characteristic of musical composition, I think we can demand a bit more. The harmony, especially at the end of the piece, seems to make little sense. Mind you, I can think of a number of 20th century pieces where the harmony is just as nonsensical. It is remarked in the Guardian article that the program can generate these pieces "endlessly". Well, in that case, we need to have some way of evaluating the music to decide what is worth performing and what not. Right now, the reason to perform and listen to this is "historical" value. If indeed this is music worth hearing, and it is the first music worth hearing to be composed by a computer then this is of historic interest.

But we can dig a bit deeper. The question of agency is always important. Who is the responsible agent for this music? Obviously the technological center at the University of Malaga, specifically the leader of this project, Dr. Francisco Vico, a professor of computer science. His interest is obvious: what a coup for a computer scientist to devise a computer program that can ape human creativity! The performers, who include the London Symphony Orchestra under the direction of Lennox Mackenzie, have probably accepted the engagement because it seems like a cool project. Mackenzie offers some aesthetic criticism of the music:
 "I felt it was like a wall of sound," says Lennox Mackenzie, the LSO's chairman. "If you put a colour to it, this music was grey. It went nowhere. It was too dense and massive, no instrument stuck out at any point. But at the end of it, I thought it was quite epic." 
"The other thing that struck me," Mackenzie adds, "was that it was festooned with expression marks, which just seemed arbitrary and meaningless. My normal inclination is to delve into music and find out what it's all about. But here I don't think I'd find anything." 
"It went nowhere."  Well, yeah! That's the problem with the piano piece Colossus: it also goes nowhere. Presumably because the programmers couldn't figure out how to program in the idea of direction in music. "Festooned with expression marks, which just seemed arbitrary and meaningless." Didn't just "seem", they are arbitrary and meaningless! See, here is the obvious truth: art only has meaning and direction--intention--when someone means something by it, expresses something with it. If a piece of music has direction and expression, then it is because of human agency. What we have here is an elaborate fake of music to show off the programming skills of Francisco Vico and his team. The only human agency comes either from the programmers or from the performers. If you put a reasonably plausible score in front of a pianist, he will do his darnedest to try and squeeze some expression out of it. As we hear in the concert video.

Just for fun, let's listen to one of those previous attempts at computer composition, programmed by composer David Cope.

Yep, Michael Nyman on a bad day...

I think that these sorts of experiments are actually very useful because they underline for us what is actually important about a musical composition: it is not so much the rules or procedures as it is the human agency. And if some avant-garde compositions cannot be distinguished from music composed by a computer program, then so much the worse for the avant-garde!

UPDATE: For those interested, here is a brilliant discussion of how computers actually work, with lots of detail. I think reading this rather lengthy essay will show pretty clearly why a computer cannot compose music, though it can deliver an interesting fake.


Dan Fair said...

We, computer scientists, are engineers and also scientists. As engineers we make systems that perform in a given way. As scientists our quest is to dig into the main questions of nature (humankind included). I would never think that a serious researcher intends to ape humans. In the sense that C. Darwin did not. At the end of your post you go to the point: "they underline for us what is actually important about a musical composition", and this is probably what this research is intended to do. But your comment seemed to me rather insulting.

Bryan Townsend said...

Hi Dan,

Thanks for your comment. I am delighted to elicit a comment from a computer scientist. I think I insulted computer people less than I did music people in this post! I think that what prompted my approach was the way the whole project was presented. Was not the whole thrust of the coverage the fact that a computer is composing music as humans do? The fact that what is really happening bears very little resemblance to what humans actually do means, to me at least, that the program is aping, i.e. copying the surface without engaging with the real nature of a musical composition.

Federico said...

The link in your update has gone [here](https://nplusonemag.com/issue-13/essays/stupidity-of-computers/).

Bryan Townsend said...

Link rot is often a problem with older posts.