Discover more from dots per inch
The reality problem
AI is not the end of photography, it's a solution
For a while, I worked in an advertising agency as a photo producer. One of the things I learned from my time at the agency was that images that I had assumed to be 3D renderings were in fact photographs. And photos that I thought were taken with a camera were actually rendered or digital composites. The areas where this was most intense were product, food and beverage, and automotive photography. I was surprised at the actual weeks of work that went into preparing a bottle of champagne to be shot with water droplets sliding down the surface in just the right way or at the high cost of having a food stylist prep ice cream so that it would melt at just the right amount under the bright lights of the shoot. As a kid growing up with Photoshop, it was amazing to me: this was real, someone shot that burger over the counter at McDonald's, or that "flying water macro" in your soda ad. I never wondered where these images came from.
But there was a flip side to this experience. I was involved in multi-day product and campaign shoots that required weeks of calls and preparation to find the best photographer for the brief. Large teams of advertising experts, producers, photographers, art directors, set stylists, food stylists and lightning technicians were flown in to exclusive locations, all for the one image, the "hero shot". But then, after the shoot, these very expensive images were disassembled and reassembled like a puzzle in post-production: elements were moved to fit different ad formats (and those formats are wild!) or to make room for typography. Label positions were shifted, textures were flattened, objects from one image were recombined with those from another. The end result was often strangely off-perspective, over-polished, hyper-real.
As a photo editor with an editorial background, this excessive use of post-production was puzzling to me. If you were going to spend so much money, why not choose the best photo? There seemed to be no reason other than convention. Sometimes it was necessary because the campaign concepts didn't consider whether things could be photographed before client approval, and then had to be photographed as approved. More often, the practice seemed to be related to an obsession with detail on the part of creative directors and clients who wouldn't accept anything less than perfection in the execution of a brief. It was very much related to the high production costs of these shoots: when you spend that much money, the result better be flawless. In a food shoot for a major supermarket, we had to take a fork out of the frame next to a plate of Mozzarella we were shooting because the client thought it was "too spiky, too aggressive".
I would like to draw your attention to this very gray area in photography that has been around for years: images that we call photos, but that don't resemble what was originally shot. Most cars in automotive photography today are 3D renderings inserted into a photo landscape that was shot with a camera, but then digitally blurred to look moving. Watches in watch ads are rendered into multi-exposure sharpness. There is a whole field of conceptual photography that provides an exaggerated version of reality, often used in movie posters. I recently came across an ad that showed Cara Delevingne standing in 3D water in a weird sci-fi hydrogen world - or maybe it was a digital Cara-model-scan, who can tell? In these cases, photography only provides the raw material, and when you look at the final images, it is often quite difficult to tell which parts are photographic and which are computer generated, because everything is blurred into a painterly appearance. We could call it "AI aesthetics" to be cheeky. My point is, we think of AI images as realistic or photographic today, because we are already so used to thinking of images as photos that are objectively no longer photos. If we're honest, most realistic AI images look like heavily retouched photos, or why could Jos Avery convince so many people his portraits were shot on a Nikon D810?
We, as an industry, are (at least partly) responsible for this mess. Moving on, how can we distinguish photography from other images that just look photographic? Photography is seen as the technical process of taking a photo with a camera, but this doesn’t take into account that the examples above consist of images shot on camera. There is the „move no pixel“ rule that works for documentary photography, but would forbid any post production or retouching everywhere else.
So I have a proposal: I recently came across a text by Max Bense, pioneer in the philosophy of technology, basically one of the fathers of Gen AI, whose students in Stuttgart were among the first computer artists (Bense coined the term “Generative Aesthetics” and in the 1960s, his student Frieder Nake programmed one of the first commercial computers, the Zuse Z64, to autonomously draw like Paul Klee). In 1958, Max Bense wrote about the difference between photography and painting. In his text, he points out that one thing that distinguishes photography from painting is how quickly the content of a photograph is determined by the circumstances in front of the camera:
“Photography is distinguished by the fact that, in principle, every dot of the image plane corresponds to a dot outside the image plane. Thus, it is possible to infer real conditions. ... much faster, almost with the choice of location, the aesthetic process of photography reaches the point where the gapless determination of the image begins, a point that in painting is sometimes won with the last or penultimate brushstroke".
Whereas a painting has time to evolve as it's being painted, in photography you choose a frame, you adjust details, but you have to work with what's in front of the camera at any given moment. From this simple observation, Bense continues:
"So while painting does not necessarily have a "reality problem" (Außenweltproblem) and can be a representation or imitation of not only the real, but also possible worlds, photography always has to deal with physical reality. There is always a minimum, at least a minimum of this reality in it; absolute world annihilation is impossible for it, but so is absolute world invention."
So his suggestion for a distinction or definition of photography is from the perspective of production. Bense proposes that photography, unlike other forms of image production, always has a “reality problem”. I really liked that and I think it makes a lot of sense because as a photographer, you're always working in physical space and you can't "annihilate" its conditions: you have weather, you have moody subjects, you have time constraints. Painting, he says, does not have the same problem. Computer images, he implies, don't have these problems, just as AI images don't have them.
Now you can argue whether the "reality problem" is really a problem. I once saw a talk by Vivian Sassen, where she said she used to be stressed out by photo shoots and their realities. You can plan everything, but when you get to the set, everything can be different - no, she even said, everything will be different than you expect. Today, that’s what she she loves about photography: the role that chance plays in it, as the best results are often not anticipated.
When it comes to the commercial sector, this blessing of photography is often perceived as a threat. So much so that you would spend 30,000, 50,000, 100,000 Euros on a shoot and still feel that the results are not satisfactory. So I would like to make the following argument: all the areas of photography where reality is a problem are the areas that already heavily rely on post-production and CGI. Looking to the future, these are also the fields that will heavily utilize photographic AI.
Let me briefly explain what I mean by using the example of automotive photography, an industry that began to heavily incorporate CGI around 2007 and for which the technology is now standard practice. In car photography, every detail of the physical existence of the object in space is a problem to be solved: shipping large objects to beautiful remote locations is expensive, keeping the new models top secret while shooting them on location: a problem, placing a car in the middle of an impressive landscape or natural reserve: not easy to get permission for. Also, cars are shiny, they not only reflect light, but also the world around them. When you photograph a car, you see a lot of things on the car that are not part of the car. Photographing cars in a way that makes them look beautiful and worth the money they cost, especially on location, is a highly technical process and very difficult. CGI and post-production have helped solve a lot of these problems in the past and probably in the future, with Generative AI, it might be cheaper and easier to render the locations instead of shooting them. This could, for example, save you the cost of architectural rights for the buildings you see in the background.
Looking at current developments from this perspective, it gives one a certain trust that Gen AI is not the beginning of the end of photography. Rather, AI imaging will bring to an end what digital photography started 30 years ago: the freedom of absolute world annihilation, the freedom of absolute world invention. And in that sense, if we put aside for a moment the legal issues of current applications, it offers a solution to a problem within photography for which, one can argue, photography was never the best solution. After all, why should we continue to produce something in a very expensive, difficult and time-consuming way when what photography can deliver is never good enough? When what is needed is a better version of photographic reality. When we can't or won't tolerate photography's "reality problem".
Thanks for reading dots per inch! Subscribe for free to receive new posts and support my work.