Alex Borkowski | Voicing Neural Networks: AI in/as Performance
In 2019, a report commissioned by UNESCO claimed that the way users interface with technology is in the midst of a paradigm shift from text input and output to voice input and output. Indeed, voice assistants such as Amazon’s Alexa and Apple’s Siri, are increasingly central to daily life, embedded in a proliferating number of smart technologies including speakers, cell phones, televisions, cars, and security systems. Recent scholarship has crucially examined the way in which the auditory aesthetics of femininity and whiteness serve to conceal the racialized power dynamics in domestic labour (Phan, 2019; Schiller and McMahon, 2019), as well as assuage users’ anxieties regarding surveillance and intimate data extraction (Woods, 2018). However, the specifically sonic terms of this paradigm shift—what it means for neural networks to vocalize—remain under-explored. What insights might be gleaned regarding human/AI interactions when considered as vocal performances in an artistic sense? What possibilities might emerge through thinking algorithms alongside the formal conventions of music? This paper therefore looks to the work of two contemporary artists and composers—Erin Gee and Holly Herndon—that engage with the complex, evolving, and increasingly ubiquitous relationships between voices, both human and nonhuman, and algorithms. Both Gee and Herndon adopt a playful, tinkering approach to training sets, creating multi-agential sonic assemblages that challenge conventions of authorship and creative expression. In to the sooe (2018), Gee performs an affecting ASMR interpretation of a surreal and stuttering text generated by a deep learning neural net as it learns to speak using Emily Brontë’s Wuthering Heights. Herndon also sonifies intimate interactions with algorithms on her album PROTO (2019), wherein she “feeds” vocal inputs to Spawn, a singing AI that she refers to as her “inhuman child.” Whereas generative music typically employs scores as training sets so as to emulate the conventions of a musical genre, Spawn acts as one among several performers rather than a composer. Herndon and Gee thus initiate entangled and reciprocal relationships between human and nonhuman composers, performers and listeners that exceed the capitalist incentives that propel recent advances in algorithmic music (Goodman, 2009; Dean and McLean, 2018). These artistic practices further provide points of entry to consider posthumanist and new materialist approaches to voice. Thinking with the disarticulation of liberal humanist assumptions regarding the alignment of voices with singular cohesive subjects (Chadwick, 2021; Mazzei, 2016), I suggest that Gee and Herndon crucially drain, revise and reorient the dissembling capacities weaponized in voice assistant technologies. These artistic practices therefore offer subversive models for vocalizing algorithms otherwise.
BIO | Alex Borkowski (she/her) is a writer, artist and PhD student in Communication & Culture at York University, Toronto. She holds a BA in Art History from McGill University and an MA in Aural and Visual Cultures from Goldsmiths College, University of London. Her critical and creative writing has appeared in openwork (forthcoming), Journal of Radioand Audio Media, Prefix Photo, Canadian Art, KAPSULA, The Quietus, this is tomorrow, and The Happy Hypocrite. Her sound artworks have been presented at the plumb (Toronto), Well Projects (Margate) and Artsadmin (London), and online at ilyd.nu. Her research interests include sound studies, performance studies, feminist theory and STS; and her proposed dissertation project seeks to excavate points of resonance between the disembodiment/re-embodiment of voices in 19th c. spiritualism and contemporary female-coded vocal interfaces.
FEATURE IMAGE: PUBLIC DOMAIN
