Paper short abstract:

Audio compression algorithms operate by identifying & eliminating excess sound, thereby reproducing political partitions between voice/noise. This paper situates these partitions in the history of political thought as products of struggles over definitions of logos and the political subject.

Paper long abstract:

This paper explores the political and economic rationalities that are (re)produced by the compression algorithms involved in audio data transmission and formatting. While economic logics of compression are largely expressed through the terms of an empiricist paradigm—e.g., where a message is said to be compressed when all “excess” has been eliminated and its “essence” has been preserved (thereby producing surplus bandwidth)—the operation of identifying and eliminating excess sound exists within a pre-existing realm of the sensible, one that is entangled with political partitions of silence/sound and voice/noise that have already been drawn in advance. For instance, a number of political assumptions underwrite the construction of so-called “redundant” and “unnecessary” audio data—these are conceptions of legibility and intelligibility that are situated in a broader history of political thought, one that begins with Plato and Aristotle’s distinction between logos and phoné, or ‘reasoned speech’ and ‘noise’, which functioned to delineate the political subject from the non-subject. This paper seeks to surface the ways that audio data compression algorithms participate in this legacy through the construction of models that trace these distributions of sense and subjects. Then, drawing from a rich critical tradition invested in the deconstruction of such notions of the subject and of logos (in particular, from the work of Jacques Derrida and Jacques Rancière), AI models can be understood to always leave behind an residue that cannot be contained by orders of the sensible.