The Cargo Cult : (Real Time AI Live Coding using Evolutionary Algorithms)

In this paper, we present the piece and the underlying soft-ware system of "The Cargo Cult" by Damian T. Dziwis. The piece is anaudio-visual live coding performance using an autonomous programmingmachine learning (ML) agent in the context of contemporary music. TheML-agent is a genetic programmer system from the eld of evolutionaryalgorithms. It is designed to live code in a proprietary and specially de-veloped esoteric-like programming language, resulting in real-time audio-visual events presented to the audience. In this case, the usually used tness function of the executing algorithm is replaced by a text-based,chat-like input interface. This creates a genuine human-computer inter-action between the autonomous agent and the artist on stage, re ectingthe theme of "cargo cults" emerging in society in the context of interac-tion with arti cial intelligence systems. The piece was developed for the\die digitale" festival (Düsseldorf/Germany) in 2018.