Although it’s easy to feel alienated by the opaque processes behind artificial intelligence and fearful that the technology isn’t regulated, the artists Holly Herndon and Matt Dryhurst want you to know that A.I. can be beautiful.
Their exhibition “The Call,” at the Serpentine Galleries in London through Feb. 2, is the first large-scale solo museum show for the artist duo, who have long been at the forefront of A.I.’s creative possibilities.
Herndon — who was born in Tennessee, grew up singing in church choirs and later received a Ph.D. in music composition from Stanford — has made cutting-edge, A.I.-inflected pop music for over a decade. With Dryhurst, a British artist who is also her husband, she has branched out to make tools that help creatives monitor the use of their data online, and recently, into the visual arts.
The couple’s work “xhairymutantx,” commissioned for this year’s Whitney Biennial, uses A.I. text prompts to produce an infinite series of Herndon portraits that highlight the playful nature of digital identities.
The Serpentine show combines musical and visual elements. With the varied a cappella choral traditions of Britain in mind, Herndon and Dryhurst worked with diverse choirs across the country, from classical to contemporary groups of assorted sizes, to produce training data for an A.I. model. In a wall text, the artists explain that “The Call” consists of more than just the A.I.’s output. They also consider the collection of the data and the training of the machine as works of art.
“We’re offering a beautiful way to make A.I.,” the artists’ statement adds. Their utopian take is that A.I. is collectively made: It learns from whatever it is exposed to and can therefore be shaped for good.
An open call for singers took Herndon and Dryhurst from Bristol to Leeds, London to Belfast, Salford to Penarth and beyond, with 15 choirs taking part. At each location, the singers were presented with a special songbook of hymns and musical exercises devised by the artists to cover the widest range of musical and phonetic samples to train the A.I. model.
Where some might see sinister portents or “2001 Space Odyssey” or “The Terminator,” the artists see creative possibility and beauty. They recorded each choir in its traditional home, including community halls and churches, preserving the acoustics of each space. The singers were arranged in a circle around a central microphone so that the sound captured is diffuse and multidirectional. Far from being engaged in a rote or rigid process, participants were encouraged to interpret the scores as they wished.
The result, which echoes from speakers throughout the Serpentine’s cavernous brick spaces, is uncanny, exquisite, unpredictable. Fed through the A.I. model, the voices of the choirs come back ghostly and ethereal. Individual voices weave in and out, harmonies overlap and tunes that seem recognizable, but just out of reach, loop and swell. It makes you feel like you are the microphone at the center of a vast chorus.
But the magic ends here.
With such a conceptually fascinating and sonically gorgeous project, it’s hard (for me, impossible) to see why the “output” also included a series of cumbersome and artless sculptural installations inexplicably farmed out to the Berlin-based architecture studio sub, known for its collaborations with the fashion house Balenciaga. The exhibition describes its three spaces as visual representations of the A.I. training process (the accompanying booklet also offers a series of schematic diagrams), but they are ungainly and hard to grasp.
The first, “The Hearth,” opens the show with a kind of altarpiece of white cables, whirring fans and panels decorated with white and gold patterns and a gold panel at its center depicting an infant blowing into an ancient wind instrument. In the next room, “The Wheel,” a giant chandelier-like object is printed with gilded images of mysterious cloaked figures in procession, and rigid, plasticized pages of Herndon and Dryhurst’s songbook hang from the sculpture with chains.
The last section “The Oratory,” consists of two private, curtained rooms where you can engage with the choral A.I. model yourself by singing into a microphone. (Unfortunately, I could mostly hear my own voice trying out Laurie Anderson’s “O Superman.”)
For all Herndon and Dryhurst’s engagement with ideas of accessibility, pedagogy and collaboration, and their stated aim of making something beautiful, the exhibition’s physical installation serves only to mystify and, to some extent, confuse.
The careful, complex process undertaken by the two artists with their 15 choirs is crudely materialized in these objects that vaguely and uncritically gesture at religion and ritual with muddled images that, for many tech-skeptics, will gesture at unthinking veneration.
More than neural networks, the whirring fans of “the “Hearth” installation reminded me of A.I.’s huge carbon footprint and the massive amounts of electricity it sucks up: a pressing collective concern of another variety. In the end, I closed my eyes and concentrated on hearing something beautiful made with A.I. Maybe, someday, we’ll be able to see it, too.
The post They Taught A.I. to Sing, and It Was Beautiful appeared first on New York Times.
