Silicon listening. MEMS, near-ultrasound, and machine listening beyond AI

This article introduces silicon listening as a materially grounded and threshold-oriented mode of machine listening that emerges at the blurry edges of the audible. Drawing on media archaeology, writings in sound studies, and patent analysis, this contribution explores how near-ultrasonic communication and microelectromechanical systems (MEMS)-based audio technologies organise listening in terms of what machines hear and how listening itself is enacted, operationalised and valued. The first part traces the emergence of near-ultrasound as a medium of acoustic data exchange, showing how applications like "SmartTones" and "Things that talk" encode specific assumptions about attention, perception, and the sonic. These systems use the upper range of human hearing to enable machine listening, often in covert or extractive ways. The second part turns to the micro-scale materiality of MEMS microphones – devices that render silicon into a resonating, listening surface. By focussing on their auditory thresholds, the article shows how such components reshape the relation between sound, data, and power, foregrounding sonic connectivity over computation. Ultimately, silicon listening offers a conceptual lens for thinking machine listening beyond AI, centring material, sonic thresholds, and the inaudible as active sites where listening is shaped.