An Evo-Devo System for Algorithmic Composition that Actually Works

In algorithmic composition there have been countless works, using many diverse approaches, to achieve fully-automatic computer-composers. However, most of them just seem to be scattered and with no mainstream awareness or widespread use. The compositional system described here has produced a number of remarkable and original pieces, that have been premiered in live concerts, recorded by first-shelf musicians, and published as the first album made by a computer in its own, genuine style. We describe its evo-devo core, fundamented on evolutionary search of Lindenmayer systems. While the latter provides an indirect way of encoding musical pieces that imitates embryological development, the former browses the search space to find genomes that fit with requirements. The fitness function integrates formal principles of music writing, as well as general aesthetic criteria, in order to assess the quality of the final compositions. We also present some applications that derived from the particular functioning of this system, and discuss how these new fields of application might be disruptive in the music field.