Sunday, January 6, 2013

Computers Mimicking Composers

The history of music and technology is a long and mostly productive one. The invention of nylon in the 1930s and the development of practical uses for it during the Second World War was a life-saver for classical guitarists who previously had used catgut for the treble strings. During the war, Segovia has commented, it was almost impossible to find gut strings!

Since the 1950s, however, there have been many other intersections of technology and music. The use of magnetic tape for sound recording was another innovation of the 1930s and it has had many benefits. Previous media for recording had much poorer fidelity than was possible with magnetic tape. It was also possible, with a razor blade and sticky tape, to edit the magnetic tape. Composers soon saw the possibilities and a whole generation of composers such as Karlheinz Stockhausen and Pierre Schaeffer experimented with "tape music" that is to say, music created using the possibilities of magnetic tape. You can record natural or musical sounds and splice them together any way you want which is called "musique concrète" because a number of French composers were the first to try it. A very famous example is "Revolution No. 9" on the Beatles' White Album.



Of course, magnetic tape was just one technological tool composers discovered. There was a whole panoply of other devices that went together to produce "electronic music" or music synthesized from electronically produced sounds. The famous piece by Stockhausen Gesang der Jünglinge combines both approaches as it uses recorded vocal sounds and electronic sounds.


Most popular music today uses a wide variety of technology including digital recording, which has replaced magnetic tape, synthesized sounds, electronic instruments including not only electric guitars and keyboards, but also drum machines and computerized editing programs such as Auto-tune which corrects mistuned sung notes.

The latest trend is towards automating, computerizing, the composition of music. Several attempts have been made, not too successfully, but the most persistent current one is by Francisco Vico as a project in artificial intelligence at the university in Malaga, Spain. I previously posted about this here when the Guardian had an article on the project. Now the BBC has a new story on the project, called Iamus. Alas, this article lacks the common-sense approach of the previous story. In the video clip at the beginning, Prof. Vico states that their aim is to "make a computer that can mimic man when it comes to composing music." I was taken to task in a comment on my previous post for characterizing the project as an attempt to "ape human creativity" but as you can see, this is exactly the goal. At the end of the video clip, the narrator goes right over the edge saying that "Iamus is a landmark because ... what it writes cannot be differentiated from music written by a human being." I'm not certain, but I think this is a translation of the remarks of Prof. Vico.

This last statement is not quite as ridiculous as it sounds because the music composed by Iamus can most certainly be mistaken for music by quite a number of modern composers. The passage being played by the orchestra in the clip, for example, sounds rather like "The Housatonic at Stockbridge" from Three Places in New England by Charles Ives.


The difference is that while the music by Iamus sounds vaguely like the string texture of the Ives, what is missing is Ives' paraphrase of the hymn tune "Dorrance" which is woven through the texture. The truth is that Iamus is programmed to generate "countless" examples of modern sounding music. Brief segments of these compositions are not easy to distinguish from actual modern music. The article in the Guardian has a link to a "Turing Test" where you are invited to identify the music written by Iamus among five brief selections. Here it is. With less than a minute, and given the musical 'language' of modernism, it is not easy to do! A lot of modern music sounds pointless and directionless over a brief span. Some of it sounds like that over the long term as well! But the music Iamus writes really is pointless and directionless because the only actual agency involved is that of the programmers. The computer itself, of course, has no intentions and can give no direction to the music.

So my conclusion is that the interesting experiment of Iamus does not so much tell us that a computer can compose music, but rather the reverse, that some modern compositions sound no better than music more or less randomly generated by a computer.

No comments: