AI-generated photography has taken us by surprise, leaving us dazzled by its capabilities. However, if we aim to fully understand the artistic, political, and social implications associated with AI photography, it’s crucial to identify notions and concepts that can help us reflect on this complex phenomenon. From the aesthetic theory of images, certain notions appear promising. Let’s explore them together in this short article!
1- The Fictional Image
French philosopher Philippe Dubois introduced the concept of the “fictional image” which he believes applies to digital photography as opposed to the “trace image” of traditional, analog photography. According to Dubois, digital imagery could represent not what has been, but rather what could be1. Digital photography, and by extension AI-generated photography, does not depict reality as it is, but as it could be—encompassing the imaginable and conceivable within a fictional context. However, Dubois’s concept extends beyond mere fiction. He also considers “potential” scenarios—what could possibly happen, a question that often arises in the absence of images. Dubois draws inspiration from Lebanese photographer Walid Raad, who in his Atlas Group project2, crafts imaginary archives that attempt to convey the essence of the Lebanese wars. Raad argues that these fabricated images are truer than the “real” ones circulated by a propaganda-ridden press. In this light, AI-generated images might aspire to a new type of “image truthfulness,” capable of visualizing events we know have occurred, yet for which no real images exist.
2- Lyotard’s Immaterials and AI Photography
Lyotard was among the first philosophers to attempt to theorize the digital realm. He posited that with the advent of digital technology, traditional matter had vanished, replaced by “the immaterial” which is not intangible but rather matter generated by language (he was thinking of computer coding). Lyotard, also known as the “father” of postmodernity3, developed these ideas in the 1980s during a famous digital art exhibition. His concept of the immaterial might not have been entirely relevant to the digital world at that time, but it becomes fascinating in the context of AI photography because such photography is, in effect, an image generated by language. Thus, the emergence of AI-generated photography as the true embodiment of postmodern photography raises the question: Does this mean that reality itself shifts beyond the visible? Perhaps not, as the “code” or language generating AI photographs does so through a neural network built with real-world images. In a sense, reality is embedded within the neural network itself.
3- Jean Baudrillard’s Concept of the Hyperreal
Jean Baudrillard posited that postmodern societies are immersed in images, but the connection between these images and reality is dubious. The postmodern individual cannot discern whether the images they see, which are their exclusive means of interacting with reality, actually represent the real world or not. According to Baudrillard, this constitutes hyperreality, where it’s unclear whether what one sees reflects true reality—essentially, we’re living in the Matrix. This notion resonates more than ever when we find ourselves startled by images like the Pope in a white puffer jacket or Donald Trump being arrested—fabrications made possible by AI4. And what about when you’re about to meet someone in person for the first time, someone you found fascinating on a dating site? Don’t you feel a surge of hyperreal adrenaline just before that face-to-face encounter, wondering if the experience will match your perceptions?
4- Photographic Indexicality
Rosalind Krauss proposed the concept of photographic indexicality in her renowned essay “Notes on the Index”. This notion suggests a unique contact with reality specific to photography as a medium, challenging traditional ideas of reality and representation. Krauss argued that photographic indexicality arises from light reflected by the photographed subject (be it a person, landscape, etc.) striking the photosensitive surface (with a primary focus on analog photography), indicating a physical connection between the photograph and its subject. However, if a neural network for generating AI images has been trained on millions of images (and their textual components), each new image generated from a “prompt” is not produced by light but by an electrical flow traversing through an incredibly complex network of digital connections. This raises the question of whether AI-generated photographs still belong to the realm of photography, particularly whether the notion of indexicality can be applied to them at all. It’s also crucial to consider not just the ontological aspect but also the deployment and circulation of images, including the role of press agencies. These agencies, vigilant about the origins of images, strive to expose and filter out “retouched” images. Yet, sometimes even experts fail to identify fake images, as evidenced by Boris Eldagsen’s case, who won the Sony award with DALL-E-generated images, highlighting the challenge in distinguishing such creations.
5- Vilém Flusser’s Concept of Hallucination
Vilém Flusser takes a radical stance, asserting that even analog images are, in essence, “false.” According to this Czech philosopher, photography from its inception has merely re-encoded scientific and technological texts that predated the construction of photographic devices. Thus, for Flusser, photography has always been about the re-coding of text. This notion becomes unequivocally true with images generated by artificial intelligence. Are we, therefore, finally entering an era of hallucination through images?
6- AI Photography “Statistical Rendering” (by Hito Steyerl)
This tongue-in-cheek term for AI-generated photography was coined by artist and academic Hito Steyerl. Steyerl views AI-generated photography, much like NFTs, as tactics employed by digital giants to sell their products and services5. Indeed, Steyerl points out that today’s discussions rarely touch upon the once-hyped NFTs. However, many of us have become familiar with what an electronic wallet is thanks to these ultimately ephemeral artifacts. Does Steyerl have a point, or is AI photography here to stay? Time will tell…
7- Post-Photography
The concept of post-photography has re-emerged in connection with the work of photographer Blake Wood. It’s crucial to note that the term “post-photography” was not originally coined in relation to artificial intelligence. It simply referred to photographic projects where “post-photographers” worked with images taken by others. However, by broadening the conceptual scope of post-photography, it’s clear that AI image generation can be seen as a form of post-photographic practice, albeit indirectly, since the neural networks are trained on photographic images captured by others. Through the lens of post-photography, AI-generated photography can be aligned with traditional post-photographic creations, and it would be interesting to explore whether the discourse surrounding post-photography remains relevant in the context of AI-generated images.
8- Roland Barthes’ “That-has-been” (ça-a-été)
Yes, this concept, though not particularly trendy, remains foundational. Old but gold, it’s essential! When Barthes penned his famous essay, he hadn’t foreseen AI-generated photography, yet his semiotic principle of “that-has-been” persists as a profoundly effective conceptual tool. Normally, we might assert that the subject of an analog image “was there” but what do we make of AI-generated photographs? These are “calculated” images, interpolated from a multidimensional swarm of other images beyond our imagination. This raises the question: to what extent can AI-generated photography be considered a “trace” of something? Or should we think of a “trace” contained within the neural network itself? It must be acknowledged that AI-generated images carry a form of indiciality, as without it, there would be no basis for copyright claims on people’s likenesses with neural network-generated images, which is clearly not the case. On the other hand, if we presume that the subject of an AI-generated photo never existed, do we approach the antithesis of Barthes’ formula? Should we then say, for AI photography, “that-has-not-been”? This leads us to consider exhibitions like “This Person Does Not Exist” prompting further reflection on the nature of AI and photography.
Do you have any comments, questions, or critiques about this article? Feel free to share your thoughts below, and we’d be delighted to continue the conversation with you!
- See Philippe Dubois, Trace-Image to Fiction-Image: The Unfolding of Theories of Photography from the ’80s to the Present. url: https://www.jstor.org/stable/24917007
- https://www.theatlasgroup1989.org/
- https://www.upress.umn.edu/book-division/books/the-postmodern-condition
- https://edition.cnn.com/style/article/pope-francis-puffer-coat-ai-fashion-lotw/index.html
- See the article Hito Steyerl on Why NFTs and A.I. Image Generators Are Really Just ‘Onboarding Tools’ for Tech Conglomerates. url: https://news.artnet.com/art-world/these-renderings-do-not-relate-to-reality-hito-steyerl-on-the-ideologies-embedded-in-a-i-image-generators-2264692