Many scholars have written about truth and indexicality in photography, but recent breakthroughs in generative artificial intelligence call for a reassessment of the existing literature. The quality of photographic images produced by text-to-image generative models is rapidly approaching that of the real thing (images created and distributed by a camera). Users have no need of any prior understanding of the complex scientific and technical principles behind text-to-image systems. Word pictures are accurately translated into images. Text-to-image systems are not mechanical; physical. As a consequence, the images produced do not relate directly to the object they represent by the movement of actual photons of light; they are not indexical in the usual sense. We could, therefore, hypothesise that the event we are currently experiencing, what we might call ‘the invention of stochastic images’, will be a turning point in human culture. However, such images are, like the real thing, created by an apparatus. Thus, they are technical images and can be decoded as such. 

The focus of my essay is not the meaning of technical images/photographs. Rather, I am interested in the relationship between a photograph and the object it represents (its referent). What is the nature of the referent's being? That is, does it exist in fact or merely in appearance? I ask: is the existence (the facticity) of a referent absolute and necessary? Is a photograph contingent on photons of light striking the surface of a sensor or photosensitive emulsion? My findings steer me away from semiotics and towards the phenomenological approach. My argument builds on those of the philosophers Vilém Flusser and Jacques Derrida, while my principle conclusion represents a step towards an understanding of photographs/technical images which does not pivot on a category mistake; thus an understanding more likely to be of benefit when dealing with stochastic images.