tekhnē

Dubwise Machine Listening

For Pedro Oliveira's residency at Q-O2, he would like to continue his investigation on the limits and failures of machine listening when it comes to what is it that a classification and identification algorithm understands as “voice” or “language.” Pedro's approach moves completely away from the idea of misusing machine listening at its limits, or of providing a corrective for its workings. Rather, he approaches machine listening in a non-utilitarian way, that is, in breaking apart the “hows” of machine listening systems as they try to come up with answers to questions such as “who is speaking?” (voice) or “where are you from?” (language). In approaching the inner workings of voice and language recognition systems with the mentality of dub music – of distorting, delaying, re-purposing, and versioning a given input – he intends to research and apply a methodology to uncover different paths for the flow of sonic information inside a given system which disregards linear correlations between input and output in favor of a poetic, dubbed out, ambiguous listening experience. With that, questions regarding any “intelligence” of machines return to the listener, prompting a reorganization of bodies in space as a disposition and dispositive of listening. In other words, a dubwise machine listening redistributes agency in machine listening and emphasizes ambiguity, subjectivity, and the accountabilities they take. In experimenting with the possible juxtapositions of dub (as a sonic process and material approach) and data (that which accumulates in time within a system), questions such as agency, probability, classification, or even recognition become open and at the limits of listening.