When AI don’t sound like AI: Negotiating aesthetic expectations in technology-mediated musical practice

Abstract This paper examines how aesthetics are constructed in technology-mediated musical practice, focusing on the interplay between cultural expectations of AI-generated sounds and the technical structures determining the behaviour of AI algorithms. Through a reconstruction of events in the Surfing Hyperparameters project, we capture how the sonic aesthetics of the system were constructed by negotiating between our sonic expectations (informed by cultural narratives of ghosts in machines) and the sound produced by the system. We argue that the aesthetics of AI-generated sound are often inspired rather than directly caused by the technology itself. While existing research has identified how tools embed ‘paths of least resistance’ towards certain sonic aesthetics, our work reveals a complementary force: how aesthetic expectations rooted in cultural narratives – from science fiction’s stories of autonomous machines to sonic hauntology’s spectral presences – actively shape design decisions and sonic outcomes. Through a radically transparent approach to documenting mismatches between expectation and reality, we show that the stories practitioners tell while building and making music with technology are performative, constructing rather than merely describing aesthetic realities. Addressing these interplays between imagination, expectation and material reality constitutes an important step towards addressing the complex sociotechnical assemblages in which technology-mediated musical practices come into being.