Paused for thought: how stereotyping and anomalies in generated images prompt reflection on the technicity of AI

Abstract

The gallery ‘Friend or Foe—the Rise of Artificial Intelligence’ at Questacon—Australia’s National Science and Technology Centre, encourages visitors to reflect on social and ethical issues accompanying machine-learning artificial intelligence (AI). Outputs of the text-to-image generator, Midjourney, are displayed as if they were artworks and framed with speculative and poetic text that leaves the ‘black box’ of text-to-image generation intact. Drawing on Bernard Stiegler’s notion of the discretisation of the image, this paper’s reading of the exhibition proposes that the images’ obvious racial and gender stereotypes, in addition to their lexical and spatial irregularities, provoke reflection on text-to image generators’ distinct technicity as an image-making technology. Displaying select outputs lifted out of the proliferation of text-to-image generation via chatbot prompt enables a contemplative viewing. Reflecting on the unfamiliar outputs of diffusion models of image generation engages the multilayered social, cultural and ethical issues surrounding this technicity. Most importantly, the gallery cultivates the stance that data-driven, networked AI is an interpretive and disputable practice and underscores how viewers are implicated in the present and future of AI.