My ability to execute my creative ideas is far outstripped by my volume of thoughts of possible art to come. Therefore I present my first speculative essay. The topic is the use of computers to advance the practice of music composition and performance:
Due to it’s mathematical roots, music lends itself well to use in computing. Western classical music notation is discrete in nature with pitch, duration, dynamics, and phrasing meticulously spelled out for the performer. The framework of music theory and counterpoint both can be reduced to more or less mathematical relationships. 20th century twelve tone and tone clusters are, of course, mostly completely mathematical in nature to the point of being arguably musical. Even instrumental idiosyncrasies and ensemble approaches and relationships used to balance music between individual instruments, octaves, and sections can be modeled and represented in numeric and abstract terms. The compositional rules which govern the writing for a single instrument or group of instruments closely resembles an “interface” in object-oriented programming. Notation could be said to trigger “methods” and set “properties” of the instrumental treatment.
All this to say that we can teach a computer a great deal about music without losing much in the translation. Design patterns for many aspects of music can be derived and boiled down to algorithms. If this were accomplished, the next step would be to impress upon the machine higher-level considerations of musical form and style, rhythmic and melodic fashions of different ages. Then provide a tremendous amount of raw material, musical data for the machine to digest and learn from, comparing it’s models to reality and allowing it to adjust the models to it’s own satisfaction.
The most challenging consideration would be the introduction of the idea of emotion in music. I wonder if this has been attempted with much seriousness or success. First we might codify the body of emotional attitudes expressed through music, most of which are defined by their Italian or German names. Then the delicate procedure of performance approaches would be needed and here the water becomes deep indeed. It is unclear what defines the human concept of musical shape in emotional terms. Perhaps we might start with patterns of phrase and multi-phrase shapes, cuing on dynamics in relation to contrapuntal and harmonic context. Here we move into intangibles which must be demystified. What, exactly, is a turn of phrase? How, in precise terms, is a single note held without becoming static? In a piece with thousands of notes, what holds it together from moment to moment and from beginning to end? At this point the many simplified models and algorithms regarding melody, harmony, dynamics, phrase, and all the other variables available in music begin to work together in enormously complex ways to result in effective musical phrases, movements, and pieces. Most critical is the idea that emotion is a feedback of information. A composer or musician responds emotionally to their own output, adjusting their output accordingly to moderate this response, this “feeling”. It could be said that great compositions or great performances are not producing audio information as their end result, they are producing a response in the listener. I contend that machines will only be able to stimulate us intellectually through music when they have learned to stimulate us emotionally. In order to achieve this they must be taught to feel. Additionally, they must be taught to learn from experience to develop their ability to differentiate feelings and place them at their disposal for their use in music. I realize this is getting more than a bit sci-fi but this is precisely where this line of inquiry leads us and we have the means at our disposal to experiment with this.
So in order to build musically effective machines, we must first teach them the mechanics of music then give them a way to respond emotionally to their own sounds.
Image from Jaime E Oliver.
