Conference Agenda

We live in a world where machines “listen,” translating low-level audio data into high-level representations of all kinds, whether it be voices in health-monitoring platforms like Sonde (Bagwan 2022), songs in audio-fingerprinting apps like Shazam (Wang 2003), or coral reefs in ecology surveillance projects like EARs (Lammers et al. 2008). But how can we trust what they hear? And what are the societal implications of delegating such a wide array of critical judgements to these nascent technologies?

The question of machine listening is at the center of a developing discourse in media and cultural studies, where authors debate the epistemological and ontological status of this kind of audition (Napolitano and Grieco 2021, Sterne 2022), as well as investigate its embeddedness in different social, economic, and geo-political networks (Li and Mills 2019, Sterne and Sawhney 2022, Parker and Dockray 2023, Ma et al. 2023). Building on this research, my presentation focuses on what machine listening means in the context of contemporary music practices, where it has been harnessed in commercial applications for automatic transcription, audio classification, and similarity searches, as well as in generative AI music-making applications like Open AI’s Jukebox (Dariwhal 2020) or Google’s MusicFX (Agostinelli et al. 2023). Drawing on Nina Sun Eidsheim’s reflexive “listening-to-listening” framework (2018), I highlight a general tendency towards the reification of sound and identity in machine listening tools, which too often rely on a problematic reduction of relational phenomena to audio-descriptive essences. I show how this tendency operates in voice analysis-synthesis tools developed at IRCAM in Paris (Benaroya et al. 2023), which purport to “disentangle” voices into independent parameters for age, gender, emotion, and other aspects of identity. I then read these claims against the grain of critical mis-uses and de-scriptings of such technologies (Ackrich 1992) found in recent music by experimental artists Jennifer Walshe, Alexander Schubert, and Jessica Feldman. Through a series of analytical vignettes, I show how this music offers a mode of resistance to the prevailing techno-optimism that characterizes commercial rhetoric around AI and machine listening, thus opening space to interrogate its ideological effects and rethink some of its underlying assumptions.

What if artificial intelligence is Chinese? This paper argues that generative AI today occupies the same philosophical space to which the European Enlightenment consigned the Chinese: a hyperrational figure threatening humanity with its mastery of imitation but inability to create or relate. I begin with the sonic undertones of John Searle’s “Chinese Room Argument,” which critiques the Turing Test by distinguishing human from artificial intelligence. Foundational to this distinction is a misunderstanding of Chinese writing as relying silent and rational symbols rather than socially conventional speech sounds. An Enlightenment legacy, such phonocentrism positioned the Chinese as incapable of freedom or unalienated sociality—qualities necessitating speech and phonetic writing. Thinkers like Herder and Hegel framed the Chinese as pathological embodiments of pure reason, foreshadowing industrial-age anxieties about automation, AI, and Chinese laborers alike with their inhuman rationality and efficiency.

If such racist stereotypes of the Chinese as machine-like hinged on (mis)understandings of their relationship to sound, what happens to this construct in sonic encounters? One unfolded in Jean-Joseph Marie Amiot’s 1776 account of playing European music to the Chinese to judge their taste. I read this as a Turing Test avant la lettre, underscored by the metahistorical irony where the pieces played were precisely the kind of Baroque music later appropriated for Muzak and AI-generated compositions for their mechanistic regularity. The Chinese reversed the test, dismissing not just European music but also modern Chinese music for lacking the soul-moving depth only ancient Chinese music once yielded. Amiot credited this for inspiring his study of Chinese music not because he benevolently extended humanity to the Chinese but because he recognized their shared failure in being human. It was not by expanding but by traversing Enlightenment ideals of humanity as a fetish—a placeholder concealing the unutterable truth of its own fictitiousness—that Amiot and his interlocutors shared a moment of understanding.

If humanity is a fetish, conjured at encounters with its own insurmountable lack, what might the legacies of sonic Sinophobia teach us about building relations with AI—not by negotiating humanity’s boundaries but by questioning, or abandoning, this fetish altogether?