Automating Signal Processing, Automating Aesthetic Labour and Decision-Making
Jonathan Sterne, James McGill Professor of Culture and Technology, Department of Art History and Communications, McGill University
This project builds on Sterne’s prior research on signal processing (2003, 2008, 2012, 2015a, 2015b, 2016, 2018) to analyze how certain companies are developing machine learning (ML) to mediate aesthetic labour and decision-making. It explores attempts to automate decisions around the signal processing of audio. Every sound that comes out of loudspeakers is manipulated to sound a particular way. This work is called signal processing (Sterne and Rodgers 2011, Sterne 2016). Increasingly, ML is becoming an integral part of signal processing, shaping the way that audio and media cultures sound and feel. In a pilot study of LANDR, an online mastering platform (Sterne and Razlogova 2019), it became apparent that attempts to automate decisions about how music should sound are often shrouded in secrecy shaped by a mixture of technical affordances and corporate practices (Burrell 2016, Crawford 2016b). Although ML refers to specific computational approaches and not others, in practice these lines are often blurred – for users who experience the software as a black box, and for commercial companies that may overstate the role of ML in their software, branding their musical output as ‘AI’ in order to seem up to date. The ethnography will be complemented by a cultural-technical history of automation in audio signal processing.