Beyond Machine Vision: How to Build a Non-Trivial Perception Machine

Abstract

Approaching the problem of artificial creativity through the lens of machine vision, this article examines the impact of computer science’s model of vision on our socio-political values and institutions. It also proposes a creative experiment in “conceptual engineering,” which entails an attempt to build a non-trivial perception machine. This idea references two science papers on the relationship between humans and machines: Heinz von Foerster’s “Perception of the Future and the Future of Perception,” in which the concept of a “non-trivial machine” was first introduced, and Gerald M. Edelman and George N. Reeke Jr.’s “Is It Possible to Construct a Perception Machine?”. Critically engaging with those papers, the author ends by constructing a conceptual scaffolding for the theory and praxis of machine perception, while addressing the wider problem of epistemic and racial (in)justice in the industry focused on getting machines to “see.”