Critical AI: A Field in Formation

At first glance the most striking aspect of Anna Ridler’s 2018 installation Myriad (Tulips) is the highly ordered array of tulips themselves—thousands of photographs taken over the course of three months in the Netherlands, their meticulous gridded arrangement presenting as geometric abstraction at a distance (fig. 1).1 Up close the colors, shapes, and textures of the individual flowers become apparent, this subjective perceptual frame underscored by the handwritten labels—not didactics with botanical metadata but, rather, a registering of attributes as processed by the human eye: dead, blooming, some stripes, no stripes. The digital photographs themselves comprise a training data set for Ridler’s subsequent artwork,Mosaic Virus, which uses a generative adversarial network (GAN) for an iterative production of “fake” tulips that reflect on speculative forms of value.2 The technical and conceptual complexity of Mosaic Virus might seem to overshadow the photographic installation, but of course that data set is its necessary precondition, and, taken together, the two works make visible the end-to-end apparatus of artificial intelligence (AI), from the human labor of image classification, data curation, and machine learning (ML) model architecture design to the material infrastructural support of GPUs (graphics processing units) and the management and manipulation of generated output. The rationale for drawing on Ridler’s mediated tulips as a frame for this special issue of American Literature on the emerging field of critical AI is perhaps intuitive— this is, after all, an aesthetic engagement with ML that delights and instructs, translating machinic instrumentalization (still the bête noire of the humanities) into the lexicon of cultural critique, situating AI within intertwined genealogies of capitalism