Better Computer Music
“AI” does not stand for “automatically interesting”
For the last year I’ve been building generative music systems — little programs that produce music that doesn’t end or repeat. This software-based generative music is a form of “computer music,” which is a broader term that refers to a variety of ways computers, software, and algorithms can be used to make music. Computer music takes many forms, including programmers who code live performances of music in front of an audience, recordings of original compositions generated by deep-learning algorithms, and “collaborative” albums between human artists and AI.
Computer-first vs Music-first
When creating a work of computer music, there’s a temptation for the artist to emphasize the role of the computer, the algorithm, or the software, and treat the music as secondary. We might call this approach “computer-first.”
In a computer-first work, the audience is primarily meant to appreciate the computer’s capabilities. Imagine a machine learning algorithm trained with all of Beethoven’s works is used to produce similar music. Even if the music it creates isn’t as enjoyable as Beethoven’s music, we might still be impressed by how “Beethoven-ish” it sounds. We’re not enjoying the music so much as we’re appreciating that an algorithm…