Mockingbirds: Modelling attention, memory and the texture of repair

Abstract

How do we show what we know? How do the models used to interpret, build understanding and sustain relationships with the world, work? Artificial intelligence models ‐ particularly those characterized as ‘deep’ learning models ‐ provoke a reframing of, and renewed attention to, these basic questions. Machines designed to learn through continuous, embedded use give rise to a form of automated intersubjectivity premised on normative notions of continuity, completeness and repair that are often opaque. A turn to poetic practice may revivify supple categories of human and non-human, with attentive connection across multiple worlds, discursively explaining these models even as they enfold us. A companion video to this text can be viewed at: https://vimeo.com/525096901.