Entangled interaction with minimal algorithms: Is interactive process, and not model size, the key to more creative machines?

Book chapter, 2025

Most AI generative systems are aimed at imitating the output of creative humans, using pretrained models. As black-box mimetic systems, they merely deal with the symptoms of creativity, while most of the depth and meaning in human creative output emerges from the creative process, extended in time and involving thousands of decisions, interactions with others, and iterated bidirectional translations between conceptual and material representations of the work. Departing from this process-based model of artistic creativity, current investigations around the concept of entangled musicianship (where each action has multiple binding consequences for the continued process) and results from interactions with minimal algorithms that still render very complex output and a sense of agency (of the algorithm), I argue that it is this interactive and entangled situation that gives rise to meaning and complexity. Reproduction of superficial statistical features will never be enough. Based on these thoughts, I will discuss potential future paths of AI, where process and interaction are given a more prominent place.