Paper short abstract:

This presentation, based on a digital ethnography on the reception of AI-generated images, discusses the deep uncertainty surrounding such cyborgian images, and suggests that the public’s strategy for coping with this haunting uncertainty is a (doomed) attempt to establish the pictures’ facts.

Paper long abstract:

What is the truth of an AI-generated image? How is such images’ factuality established? In analyzing the results of a Twitter-based ethnography inquiring about the public reception of images generated by artificial intelligence software such as CRAIYON and DALLE, this presentation argues that the deep uncertainty which haunts visual artefacts (Taylor 1996) extends to, and is especially pregnant in, the uncanny realm of cyborgian, AI-generated pictures. It further suggests that the public’s strategy for coping with this uncertainty is an attempt to establish the pictures’ facts; an attempt to institute semantic, scientific, and factual control over imagistic uncertainty. It is proposed that observable attempts to institute such control notably take the form of the ubiquitous identification, by ideators, of the semantic prompts used by the software to generate these images, as well as the widespread user practice of responding to the cyborgian images with newly-generated ones from the same prompts, thus putting their reproducibility, and hence scientific factuality, to the test. It is argued, however, that such attempts at fact-making are doomed, as the factuality of these images’ context – their grounding in a ‘real’, empirical world, i.e., that of their input data – is irretrievable to both the images’ human ideators and their public, and that their having an idiosyncratic ‘aura’ is truer than their being reproductible (Benjamin 1935). Finally, by pointing at the necessity to look for such images’ subjective and Barthesian ‘truth’ (1981) rather than for their factuality, this presentation argues for an epistemology of the affective.