AI/ML in the Sonic Arts - Pitfalls and Pathways

This commentary considers the application of Artificial Intelligence (AI) and Machine Learning (ML) technologies to music and the sonic arts. It critiques the classical computational theory of mind (CCTM), a doctrine deriving from functionalism, which codifies “mind” as a mathematical function symbolic representations from one dimension (mind) can be directly mapped onto another (world) in accordance with a given transfer function. Such a function is thought to be computable on either biological or mechanical hardware, thereby rendering the internal workings of thought irrelevant. This technocratic impulse has been used to sell AI & ML products as “magical” solutions, capable of ushering in Utopian futures. This viewpoint began with the foundation of computer science itself, as metaphors for computational processes were adopted without adequate grounding in the philosophy of mind. Computers were given attributes of human cognition as a teleological basis for investment in these technologies. Our current situation sees the world’s accumulated media scraped for so-called “knowledge bases,” in many cases reinforcing cultural biases, ignoring creator rights, and consuming energy resources... all to create pastiches of existing art. As critical consumers, we should evaluate each novel technology for the social, cultural, and political assumptions that underlie their functioning. This paper will take steps towards that goal by analyzing Google’s MusicLM as a test case. The musical artists Delia Beatriz, Nao Tokui (interviewed elsewhere in this issue) and Moisés Horta Valenzuela will be used as exemplars of creative engagement with ML.