Composing for Performers and Electronics Using a Dynamically Trained Neuron as an Effects Modulator

This article presents an alternative approach for creating compositions for performers plus electronics, in contrast to the standard AI methods developed for music over the past 30 years. Despite their robust construction, the la>er seem less useful to contemporary composers who pursue aesthetic directions that have shaped experimental music since the 1950s and that focus on texture and timbre. This study begins with an overview of the categories of AI-based autonomous and interactive composition, examining selected projects considered notable by scholars, with an aim to assess their usefulness for contemporary music. An alternative method for experimental composition is then introduced, employing a Max for Live patch in which a simple neural network model controls an audio plug-in. Consisting of a single neuron, the model is not designed to emulate human intelligence but generates complex modulation functions. The patch's use is examined through results from two performances of the composition “Outside-Time Sketch.” The article describes artistic and technical implications, the project's relation to those discussed in the overview, and other methods for shaping and expanding the neuron's input. Future directions include the elaboration of this approach to use recorded sound samples in electronic music composition and spatial music.