AI-Generated Musical Works and their Authorship

Nemesio G. C. Puy (Complutense University of Madrid)

Some AI systems are artifacts designed to perform the role typically played by composers in Western musical practice. Paradigmatic examples include the IAMUS Computer, which follows an evo-devo approach based on evolutionary algorithms, and Playform, a machine learning AI system constituted by a generative adversarial network. Those cases present an apparent problem for the authorship of their musical outcomes insofar as the notion of authorship has been traditionally defined in psychologist terms implying human agency (e.g., intentionality, responsibility, endorsement et al.) and associated to exclusive authorial rights, both economic and moral rights. This puzzling situation leads some scholars to look for authorship beyond the causal source of those AI-generated works, locating it in some human agent like the system programmer, the user of the AI system or the human authors of works used in training the system. Here, I will present what I call the simple view on the authorship of AI-generated musical works, according to which AI systems are the authors of those works in a fullest sense, assimilating this to the case of human works-for-hire. The simple view is sustained by the identification of a notion of minimal authorship in music that needs not to adopt the psychologist stance to explain the standard role of composers, in which deflationary notions of intention consistent with AI can do the job. In addition, the authorial status attributed by the simple view to AI systems detachs them from authorial rights, but it will be shown that this is a non-revisionary unproblematic consequence since those rights are only contingently attached to authorship.