The Saltshaker and the Snail (silkscreen ink on acid-free paper). (2018)
Computer vision systems exceed human level performance, but it is easy to overlook the shallowness of their understanding. This print juxtaposes two abstract objects readily identified by neural classifiers trained on the commonly used ImageNet dataset. For most human viewers with experience and imagination these two objects have a suggested relationship to each other, but ImageNet trained classifiers will currently only resolve the scene into two labels with no connecting context.