It's pretty tricky to play guitar while laying facedown on the floor with your arms stretched out in front of you, fingertips reaching out to the fretboard like it was a stray plank drifting from a sinking ship. The notes splutter out cumbrously from James Creed's guitar, prone before the feet of the audience in the front row. Peter Nagle fares a little better, eking out a relatively spry melody from the cello he seems to be in the process of wrestling to the ground like a bear, one leg wrapped around the bout in a quasi-erotic entanglement. Erratic bursts of mariachi band music leap jerkily from a single turntable, behind which only Megan Steinberg's legs are visible. James McIlwrath is peeling an orange, hands projected into the air from the rear of a table splayed with kazoos, Coke bottles and a small child's tambourine. The peel projects a delightfully zesty olfactory note in counterpoint to the tumble of sounds all around. Projected on to the wall behind the stage, the score reads: ‘Performers are asked to lie down on the floor and do the best they can.’
I was at the Darmstadt Ferienkurse in July 2018 when composer Jennifer Walshe delivered her lecture, titled ‘Ghosts of the Hidden Layer’. She spoke then, among many other things, of her love for text scores, the most ‘democratic, efficient, powerful form of notation’, and her frustration that their grammar and syntax seemed perennially stuck in a Fluxus-shaped rut. Over the previous year, she said, she'd been building up a vast corpus of such scores in order to train up a neural network to produce new scores using Deep Learning. The initial results, shared by Walshe with the class that day, seemed promising. ‘Get a girlfriend,’ she read. ‘Look after your daddy.’
A little over four years later, Walshe's neural net now holds a dataset of over 3,000 scores and is generating new verbal notations at a rate of knots. It is these new, AI-powered partitions that are being performed tonight by Steinberg, McIlwrath, Nagle and Creed with occasional contributions from Neil Luck.
Luck's part in the proceedings required the removal of both shoes and socks. The score, an epic four-page prose marathon entitled ‘Vernacular Stories of the Rural Family’, called for such non-standard musical materials as ‘regional accents’ and ‘back issues of Jane's Defense Weekly’. McIlwrath recited fragments of conversation and computer-shop sales patter at breakneck speed, occasionally holding a Walkman up to the microphone (the score says something about ‘real-time recordings’) while Luck sprawled, his bare toes deep in a pile of dead leaves and twigs, rustling frantically into a stereo pair of directional microphones. The results were deeply odd but strangely compelling, like spying on some compulsive private ritual or flicking through channels late at night and stumbling upon a previously unknown and especially sadistic gameshow.
Throughout the evening, we saw scraps of Erik Satie scratched and stretched and recorded to tape then played back and against themselves; we saw McIlwrath tied naked to a tree; we saw an unseen videographer throwing punches at a double robe hook screwed to a door (the score to that one began ‘You get to fight an octopus’). Even the audience found themselves embroiled in the performance, when a score simply requesting ‘Be quiet when you go to the bathroom’ flashed up on the screen immediately before the start of the interval. Walshe, clearly, is right: text scores are democratic and powerful. In a sense, anyone could have played these scores (and thinking to yourself about how you might have interpreted the instructions differently almost becomes an integral part of the piece). But perhaps few performers could have played them with such verve and good humour.
