Vocal Aesthetics, AI Imaginaries

Abstract

In 2019, a report commissioned by the United Nations Educational, Scientific and Cultural Organization (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, Apple’s Siri, and Google Assistant are increasingly ever-present in domestic spaces. Against this backdrop, this paper proposes that critical thinking about Artificial Intelligence (AI) might be advanced and elaborated by simultaneously thinking critically about voice, unpacking beliefs that spoken language operates as a seamless “natural user interface,” as well as broader ideological convictions about voice’s exceptional status as a marker of unmediated self-presence. I examine the entangled conceptual roots between theories of voice and apocryphal origin stories of artificial intelligence in two eighteenth-century inventions—Wolfgang von Kempelen’s Mechanical Turk and speaking machine. The aesthetic practices surrounding these and other so-called thinking and speaking machines engender shifting relations of visibility and invisibility, obfuscating the role of human labor in creating the seeming magic of AI. I also examine two artistic projects—Holly Herndon’s electropop album PROTO (2019) and a series of collaborative works by Sofian Audry and Erin Gee titled of the soone (2018), to the sooe (2018), and Machine Unlearning (2018–19)—that, in their varied deployments of machine learning in vocal performance, challenge the ideological foundations of the sociotechnical milieu from which they emerge. These artworks tune into the frictions and glitches inherent to vocal interfaces and thereby unravel the vocal imaginary that sustains the illusion of their neutrality, self-sufficiency, or, indeed, intelligence.