Holly Herndon and Mat Dryhurst have been working on new protocols and materials for training artificial intelligence models. The Call, their exhibition at the Serpentine, uses multiple AI models to turn a few inputs, including The Sacred Harp, a religious songbook that spawned a tradition of choral folk music, and the voices of fifteen choirs around the UK, to a sound that they think is quite special. When I speak with them days before the show’s opening, Herndon describes it as a new polyphonic language: “I would never just sit down and write music like this. The iterative capability of the model allows it to mutate on a seed and create variations that are cut and combined using a multichannel system.”
What exactly is AI? What does it do? What’s choral folk music? What’s a polyphonic language? Armed with the description I have just offered, you might wonder whether The Call is worth visiting as a civilian without the answer to these questions. You might assume it’s one of those exhibitions that’s brainier than it is beautiful; one that’s only really for those who get it. You’d be wrong: “even though we’re dealing with very contemporary, cutting-edge technology and in-the-weeds data issues, the show is still something that’s viscerally beautiful,” Herndon tells me. “Mat and I were baptised in music festivals. We had to learn how to listen to an audience and deliver.”
The exhibition begins with The Hearth, a chorus of GPU fans that Herndon describes as “the soundtrack of our studio and of our home.” Hanging in the centre of the gallery is The Wheel, a large candelabra-like structure suspending eight songbooks at eye-height. “The music isn’t in this songbook, it’s inside you. This is a map to help you find it,” reads the opening page. The book contains songs written using an AI model trained on selected Sacred Harp songs. Speakers around The Wheel play recordings of these songs, sung by the real human voices of the fifteen choirs. Around the perimeter of the space, these recordings, along with a few other inputs including Herndon and Dryhurst’s own back catalogues, are transmuted into a series of multichannel compositions with the help of another suite of AI models. Here voices are spectral, joining together in songs without words, sometimes overlapping, other times forming a chorus. If I didn’t know what Herndon meant by “new polyphonic language,” I understand after hearing this.
To the uninitiated, myself included, the term “AI” signals an ill-defined and vaguely threatening field. It brings to mind dubiously-acquired intellectual property and lost employment opportunities. Some such cases are frighteningly close to home: at the time of writing, London newspaper the Evening Standard is publishing AI-generated content that emulates the voice of the late art critic Brian Sewell. Such cases of businesses using AI trained with human outputs to cut us out of the deal are enough to turn someone off of the whole field.
As is often the case, the technology itself is neutral. In the hands of dubious people it can be used to dubious ends, but the inverse is also true. Herndon and Dryhurst are more interested in AI’s applications for collective empowerment, bringing opportunities for creative collaboration and cross-pollination. “A mega-company appropriating or interpolating data for their purposes, for profit; that’s a separate discussion to matters of art,” Dryhurst explains. They have collaborated with more people on this project than any other; “that’s part of the point of the show.”
From the Sacred Harp songs and voices of the fifteen choirs, their AI models allow something new and communal to emerge. It’s no less authored by any of the human individuals involved, but is distributed between them: a sort of collective voice.
Folk music feels like the perfect site for such an application of AI. The history of folk is one of knowledge and music shared within communities, passed down and built upon, tunes and phrases continually interpolated. It is an iterative tradition, a perpetual call and response. The kind of polyphony that we hear in choral folk music is necessarily the product of many voices. “It’s not like it was invented by one person; it was created over time and by many people’s contributions,” Herndon explains. This isn’t so different, in fact, to the way that human culture more generally develops. “Culture is literally growing stuff and seeing it mutate and spread,” says Dryhurst.
In a small, curtained room at the centre of the gallery is a relief sculpture depicting a procession of figures climbing a mountain towards a cloudy, gilded sky. Herndon and Dryhurst tell me that the work is an homage to the Lebanese-American poet and artist Kahlil Gibran. Gibran was interested in how we might transcend our finite day-to-day existences and tap into something universal. In some sense, contributing to AI models like the ones used in The Call make such transcendence possible. Dryhurst explains “contributing your work to the public and having it be plastic and available for people to work with is, in a sense, a way of becoming infinite; it’s a way of not constraining the potential of where your work might land.”
The aural possibilities of The Call are beyond any individual’s capabilities, but emerge from the voices of precisely those individuals. As Gibran put it in his 1923 poem On Beauty, “beauty is eternity gazing at itself in a mirror. / But you are eternity and you are the mirror.”
