Sound piece commissioned for the exhibition Lava Lines at Biblioteka Kyiv in London.
The piece situates the listener inside a stream of sensor data, rising to a machinic breakdown. The piece was played for an audience in the exhibition space in an installation by Naimé Perette. Listeners received a printed, double-sided card with two texts (see photos below).
All samples are sourced from AI training data sets that contain audio recordings of broken industrial-grade machine parts. These sounds are meant to be abstracted in feature space. They are not meant to be heard by human ears but are intended for the development of AI applications that can recognize broken machines in factory settings, automating the capacity to detect malfunctioning and further removing the worker from the factory floor.
The piece asks, what would it mean to re-qualify sensory data, rather than submit it to further classification and automated processing? What event on a scale of unpredictability could initiate such a turn? As historian of science Deborah Coen writes, citing Georges Canguilhem:
"Beginning in the 1870s, explanations of natural disasters turned increasingly reductive. Theories of climate-related catastrophes like drought came to focus on sunspot cycles and global atmospheric oscillations. Simultaneously, the hunt for microbes replaced the early nineteenth century’s more multifaceted, socio-environmental explanations of disease. And seismologists gradually turned to the ‘hard’ evidence of seismographs and accelerometers, rejecting data filtered by human bodies. By the 1950s, Georges Canguilhem could argue that ‘the essential function of science is to devalue the qualities of objects that make up the milieu proper, by offering itself as a general theory of the real, that is to say nonhuman, milieu. Sensory data are disqualified, quantified, and identified.’"
Lava Lines was curated by Leïla Arenou and Naíme Perette.
Audio on request
Photography by Jordan Ramone