Holly Herndon's 'PROTO' Thoughtfully Challenges Our Conceptions of AI, Tradition, and Progression
On her third full-length PROTO, electronic experimentalist Holly Herndon wonders how AI can contribute to the longstanding human tradition of shape-note hymnals, exploring the possibilities and ethics of present and especially future AI protocols in art.
10 May 2019
Spawn is a two-year-old choir singer, that is, she is an AI agent that has learned to mimic human vocals. For her third full-length PROTO, Holly Herndon features what she refers to as her "AI baby" as a part of her ensemble of vocalists, developers, and guest contributors. The result of this human-inhuman collaboration is a project that not only humanizes the discourse of AI but also uses AI to understand humanity better. PROTO wonders how AI can contribute to the longstanding human tradition of shape-note hymnals, exploring the possibilities and ethics of present and especially future AI protocols in art.
For the past two years, Herndon, the digital artist Mat Dryhurst, and the developer Jules LaPlace have trained Spawn to replicate human voices, but their results are not exactly replications. While AI music is typically trained upon fixed, existing datasets, Herndon experimented with what she refers to as "live training ceremonies". That is, Spawn's dataset was composed of varying performances from hundreds of volunteers in Berlin. Moreover, the album's given live training sessions exemplify data that is far from the typical MIDI inputs that train most AI composers. As Herndon tells Vice, "We wanted to approach Spawn as a performer rather than a composer, and use sound as material rather than midi data scores."
As such, Herndon inundated Spawn with live shape-note congregations. Shape-note singing uses an alternative notational system that places shapes on note heads to simplify pitch indications. This system was first developed for communal, sacred singing in 18th-century New England, but the tradition became especially popular in the American South. Perhaps, for Herndon, who grew up singing for a church choir in East Tennessee, shape-note singing is a very familiar practice.
On "Canaan (Live Training)", Evelyn Saylor and Annie Gårlid sing a rendition of the Irish shape-note hymnal "Parting Friends". For Spawn, the powerful duet rings raw, sonorous. Their live session expresses what was first etched in the song's early 19th-century shape-note tune books. The AI receives the traditional hymnal as it should, maintaining its origins as an intimate, communal practice. PROTO, then, develops Spawn as a new participant of such longstanding rituals, introducing new forms of inhuman intelligence to old human traditions.
Yet, Spawn is still an "AI baby", as Herndon explains. And, "Evening Shades (Live Training)" demonstrates her limited capabilities for hymnal mimicry. On this live session, a choir and Spawn engage in a call-and-response to produce pulsating, clipping echoes. Even more, "Godmother" shares Spawn's inability to truly replicate track stems from the electronic experimentalist Jlin. While Spawn can capture the heavily percussive form of Jlin's compositions aesthetically, she sonically fails to reproduce any drum sounds.
However, true mimicry is not Spawn's purpose. Herndon explains, "I don't want to live in a world in which humans are automated off stage. I want an AI to be raised to appreciate and interact with that beauty." Thus, Spawn's mimicry entails imperfections, and that is Herndon's desired effect. These imperfections can even be understood as acts of creativity and improvisation. It is her variable nature, then, that allows one to explore every possibility of human-inhuman collaboration. For, Spawn's contribution to PROTO and shape-note singing is not her ability to replicate human vocals but rather her uncertain potentials for adding to them.
As such, the division between human and inhuman contributions becomes quite indistinguishable for the rest of PROTO. While the aforementioned pieces clearly demonstrate live training sessions or attempted replications, the remaining songs inconspicuously feature Spawn. After all, PROTO explores the possibilities of AI in art, but it is certainly not a product of AI. Rather, Spawn is just another agent of Herndon's ensemble, simply contributing to and by the composer's vision.
PROTO truly finds human-inhuman harmony as intersected and guided by Herndon. For instance, "Eternal" opens with the ensemble's folk wailings, but only to underlie Herndon's leading, gliding voice. Thrusting along with anthemic drums and symphonic stabs, there is no moment to discern the human vocal processing from the inhuman replications. Rather, the ensemble becomes just that, a communal voice that is uninterested in its individual sources, faithfully following Herndon's conceptual direction.
Even more, on "Frontier", Herndon better develops the human-inhuman harmony, not only sonically but in practice. This piece is Herndon's reinterpretation of Appalachian Sacred Harp music, that is, the traditionally scared, a cappella music is glitched to Herndon's composition of AI mimicry, live vocal processing, and digital percussions. Of course, her vastly different mode of production does not misconstrue the tradition of shape note singing. Rather, it reimagines this centuries-old practice of congregation and singing to adapt to the presently emerging forms of intelligence. Herndon tells 4AD that her experience of a Sacred Harp meetup "felt like a rare union", and it is exactly this feeling of community that she recreates by gathering human and AI, thoughtfully challenging our conceptions of tradition and progression.
The spoken word piece "Extreme Love", which features the multimedia artist Jenna Sutela and Herndon's niece Lilly Anna Haynes recites, "We are not a collection of individuals / But a macro-organism, living as an ecosystem." PROTO is a project that deeply cares for this developing ecosystem of varying species and intelligences. Certainly, it is an ecosystem that is being twisted to naturalize the evolution of feudalism, to the present racial capitalism, and finally to a centralized AI power of surveillance capitalism. Yet, it is also an ecosystem that holds the imagination for something better. Perhaps then, the discourse and development of AI should not only project the oppressive systems of our past and present but rather, it should be open to a decentralized future. For, PROTO demonstrates that small data and machine learning can be used to evolve our practices of art, community, and tradition, using AI to enhance our most human practices.