Recursion as Disruption

In Ionesco’s 1950 Bald Soprano, based on the author’s own experience of language learning, characters endlessly repeat elementary forms of linguistic communication making absurd statements about who they are, where they are, and what they do. In Paglen and Crawford’s 2019 ImageNet Roulette and the ensuing 2019-20Training Humans exhibition, ImageNet facial recognition labels such as ‘pipe smoker’, ‘first offender’, ‘slut’ or ‘trollop’ are absurdly assigned to human faces.

At first sight, in both of these examples, disruptions to/derailments of the expected logic of sense-making seem to be caused by rule misapplication or misinterpretation. But this is only if we assume that there is such a thing as axiomatic logic which maps (universal) rules to (concrete, particular) instances. Most algorithmic processes are of course axiomatic. Despite this, their micro-operation is recursive (Chaitin 2007), which is to say that it’s causing micro-disruptions to the axiomatic course of action all the time. Using historical and contemporary examples from Cage and Xu to Jaromil and the Dirty New Media movement, I discuss the performative working of algorithmic micro-change arguing that AI disruptions have as much to do with Leibnitz and Turing’s axioms as with the history of indeterminate artistic procedures.