The double portrait on the right has gained significant fame over the past month or so. This happened after Berlin photographer Boris Eldagsen (on the left) won the Sony World Photography Award in the open creative category with this photograph, titled “Pseudomnesia: The Electrician”. He immediately declined the award, explaining that the “photograph” was actually created using a text-to-image model. “This is a photography contest, but AI is not photography, so I refuse the award,” Eldagsen said. Here is a larger version of the image:
t’s easy to see how the jury could have been fooled: for people not deeply immersed in the world of generative visual artifacts, it looks like a skilled stylization of an almost amateur shot, placed on the timeline somewhere, I believe, in the late 1920s — early 1930s of the last century. In the “Commons” section of the Flickr service, especially in groups where people share found or flea-market-bought, then scanned photographs, you can see many such shots. Taking a few steps towards semi-professional and amateur photo sites, you can find quite modern stylizations in this genre. I’ve selected three images (all three are original and really old) to demonstrate why even professionally seasoned experts do not necessarily have to doubt this image — let’s simply call it neurophotography for ease.
I chose the first one to be compositionally similar to Eldagsen’s neurophotography. The next image demonstrates that completely authentic old shots may not look like such:
Finally, the third image, also reminiscent of Eldagsen’s neurophotography, demonstrates the significant imagination and even more significant technical mastery of past photographers who convincingly manipulated analog images:
Now it is easier for us to understand how the jury could have been deceived — it is quite possible to imagine a modern photographer having in mind some photos similar to these, and amateur shots from the 20s-30s, among which there are quite amazing ones, like this photograph from a photo salon archive in The Hague:
This story has caused a lot of noise — which its hero, of course, sought — and most often in discussions about neurophotography, as well as about what (in terms of video) is called deep fakes, — so most often in these discussions you hear various people expressing fears about our ability to distinguish images of “genuine reality” or even “historical past” from images that do not imply anything “real”, i.e., presumably existing independently of our perceptions of it. As someone noted the other day, “the advent of AI will cause us to lose our own past as we know it”.
It seems to me that, at least in relation to photography, these fears, on the one hand, are explainable, quite understandable, but on the other — they ignore, firstly, the modern state of photography’s relationship with reality, and secondly — how difficult these relationships have been formed throughout the entire existence of photographic technology.
Let’s try to recall the main lines of tension in these relationships. To start with, photography still enjoys an unjustifiably large credit of trust in our culture. What does this mean? We tend to perceive photography as a document — that is, as something that certifies what took place in the past — be it a few minutes or a hundred years ago. We do not have direct access to the past: this is most clearly expressed, perhaps, by the opening phrase of British writer Leslie Hartley’s novel “The Go-Between”: “The past is a foreign country”. We gain access to the past through various kinds of documents — but for about 150 years now, photography (and video since the advent of affordable video cameras) has been unrivaled here: we perceive photography as a priori more truthful source than memories, diaries, and even transcripts — not to mention drawings, which can, in fact, be very accurate — take the American genre of courtroom sketches, whose long life is explained by a legislative ban on photography in US courts.
The rapid technological progress of photography since its advent in 1839 allowed it to be used relatively quickly as an aid to memory and as a substitute for direct observation. Photographed documents also quickly began to be considered almost equivalent to the originals. From photographs, we get ideas of what we ourselves were like in the past, how other people lived, about distant events or places we visited but did not retain in memory. Photographs shape our understanding of culture, history, and the people depicted in them. Photography was perceived — and by default is still perceived as a tool of accurate and objective documentation. But why? Because it is “directly” linked to the real world: photosensitive film records what is in front of the camera lens: “The camera doesn’t lie”, the saying goes. The seeming absence of a human medium gives rise to the illusion that photography represents a pure “fixation” of reality. Neither the lens of the camera nor the chemical processes during development and printing add anything to the image.
At the same time, it has long been known that the latter statement is far from the truth. First, a photographic image can be manipulated, and this possibility has existed almost since the technology itself appeared. Here is one photograph and one neuro-photograph, both depicting ghosts, and the one that was created in 1897 seemed to contemporaries (not all, of course) very convincing. The next example is quite textbook:
As we can see, there were no technical obstacles for convincing image manipulation already in the 1930s. But there is no need to resort to direct manipulation. Every Soviet schoolchild knew that Lenin, having arrived at the Finland Station in April 1917, spoke there from an armored car. Right? Wrong. He didn’t. And the well-known photograph depicting this event is a frame from Sergei Eisenstein’s 1937 film “Lenin in October”:
In the next, actual photo, Lenin is speaking in Moscow, at the Theatre Square on May 5, 1920.
To what extent can we talk about a photographic image as a genuine representation of reality? Lens — lens, chemical processes — chemical processes, but the shooting point, time of day and year (light), and finally, the contents of the frame are chosen by the photographer. This means that the photograph, even the most documentary, inevitably carries the imprint of his, the photographer’s, personality — with all its prejudices, ideology, good intentions and so on. Susan Sontag cites as an example in this regard the well-known photographs commissioned by the US Farm Security Administration during the Great Depression. Here are two of them:
Theoretically, these photographs were supposed to present the government (in the form of the Farm Security Administration) with a “real picture” of rural America during the economic crisis. In fact, as Sontag says, they bore the imprint of the ideological program of their authors, and of course, the client, who paid the photographers. The Administration allocated (as part of Roosevelt’s “New Deal”) funds to support farm families: the “reality” that we see in these photographs is quite heavily skewed: life, of course, was very difficult then, but photographers consciously and unconsciously selected the most difficult plots. As Sontag writes, reality becomes for photography both an object and, at the same time, a subject of manipulation.
An even more complex issue with photography is that it disrupts the linguistic principle — this is specifically addressed in Roland Barthes’ book Camera Lucida. In language, the sign is the link between the signified (orange) and the signifier (the word “orange”). Photography does not break down reality into smaller units (as happens in speech) and does not reproduce the linguistic principle of exchange between the signifier and the signified. However, it has a symbolic nature: we all know that a photograph has a literal meaning (denotative) — and a metaphorical one (connotative). Dorothea Lange’s photograph, which we just looked at, literally represents a portrait of a woman with two children. Metaphorically, it’s a photograph about poverty, alienation, shame, and despair: however, our ability to read this meaning is determined not by the photograph, but by the framing, cultural, and visual context. But what reality does this photograph then reflect? The reality of the “outside world” or the reality of the visual culture of its time? Barthes believed that photography is a sign devoid of a signifier.
Photography has a complex relationship with space: no photographs of a place convey the sensation we get from being there. The fact is that photography fragments space, reducing its small segments to two-dimensional representations, and we cannot reconstruct a somewhat complete picture of space from these two-dimensional imprints. The space represented in a photograph is always very conditional — a striking contrast between a photo image and a three-dimensional environment is familiar to anyone who has ever worn a virtual reality helmet.
Why is this conversation about photography and reality? My assumption is that neurophotography — as we have agreed to call the result of creating images by large language models from text queries — is not a radical denial of photography as we know it. On the contrary: in neurophotography, photography goes to the end, finally consistently exposing the nature of its relationship with materiality. Neurophotography represents a significant step towards revealing the poverty of the philosophy of the photodocument, but at the same time, it offers us a new program of photography, in which the aspiration to a truthful reflection of material reality is replaced by the aspiration to a truthful reflection of social reality. The latter should in no way be confused with social construct: it’s not about there being no reality other than social, which, of course, is not the case — but about what reality “documents” photography, what kind of tool it is.
Intuitively, due to the seeming absence of a person (a writer, an artist with a brush or pencil), we have been thinking of photography as a technology of reality capture for a century and a half. However, photography is an art precisely because — and to the extent that — it is not a technology of fixation but a technique of artistic reflection. Images brought by the algorithm, which descends into the dark basement of latent space for them — much like a photographer in a darkroom — being already devoid of the signifier, according to Barthes, finally separate in neurophotography from the signified — if we understand it as physical reality. The signified of neurophotography is not a material object or activity as such. It manifests images, (potentially) existing in the world of representations, on which the model is trained. Where photography can still pretend that it captures the reality of materiality, neurophotography almost plainly tells us that the subject of its “capture” (and interest) are cultural representations, ultimately generated by us, people, humanity.
Here are a few examples for illustration:
What is depicted in this neurophotography? A family from the late 1980s — 1990s? No, you can’t say that. But it certainly captures the existing cultural (including visual) representation of a “strange studio family photo from the late 80s — early 90s.”
Similarly, the two neurophotographs above capture the existing cultural representations (in which exactly? I don’t know for sure) of everyday life in China in the mid-nineties, less than ten years after the start of Deng Xiaoping’s “open door policy.”
In the next two neuroshots, we see the same images of fashion photography from the late forties:
…and “popular among users of photo sites street photography of the 2010s”:
Fortunately, it is not limited to such simple examples: the mechanism of text requests, now also connected with the ability to “show” the algorithm several completely different images, opens up infinite possibilities for combining various elements, the results of which can be seen in the example of two series of neurophotographs, depicting the Burning Man festival in the 1920s…
…and in the early 1960s:
Notice that what we have before us is not simply a combination of “reality” from one time and imitation of image features from another, but a full-fledged (at least) attempt to transfer a complex social phenomenon (namely, Burning Man) to two different epochs, while maintaining recognizability and social practices as such, and practices of self-representation (see the poses of “neuro models”).
![Ksenia Klimanova. Untitled [On Identity]. Prague Media School Graduation project (2023).](https://assets-global.website-files.com/5a2d0fe7937442000184c574/65b53b7121078c5d273fcb18_1*B9AshyMuRuEB2qRPZluHzA.jpeg)
These new signs, offered to us by “text-to-image” technologies, are devoid of the signifier, like ordinary photographs according to Barthes. But they are not devoid of the signified. The signified is such that it deprives us of the possibility to hide behind the “real” for fear of seeing our reflection, our own image, apparently more truthful than the other one, appearing before us from the imagined global archive of those photographs yet. Well, or at least from the archive of all the selfies taken by you over the last year and a half — better yet, two years.
