Why Film Photography is the Antithesis of AI Art
The camera is a reasonably recent image-creation tool; compared to millennia of paintings, drawings, carvings, and illustrations, we have only a few hundred years of photographs and photographic development. What photography offers compared to those ancient arts is (relative) immediacy and accuracy.
The camera is the tool that can offer immediacy and accuracy in the depiction of whatever is in front of the lens. While digital paintings, hyper-realistic renderings via CGI, and even freehand pencil work can result in an image that represents reality as accurately as a camera, they can also offer something the camera cannot – imagery that exists beyond the scope of the lens, or even physical, current reality.
These forms of art can allow the user to depict imagery that would otherwise exist only in their mind’s eye. When Dali wants melting clocks, he prepares his canvas; when a photographer wants melting clocks, they must prepare some kind of industrial oven.
The Allure of AI Image Generators
For creators who wish to conjure imagery from their mind’s eye, AI generation offers an incredible value proposition. The only ability needed from the operator is translation, in order to describe and refine the right adjectives and parameters leading to the desired results.
AI imagery is as good as the prompt, or input, allows it to be. Someone with a poor ability to articulate their vision will struggle to bring their imagination to the screen. You may be an incredible visual artist, but a poor grasp of linguistics will prevent you from being able to describe in such a way that the machine system can produce the visual.
In creative fields, the imagination is not a boundary to compromise with, and exploring the potential for expressing that frontier via art is truly impressive.
However, there are some areas where imagination is not the ideal tool to put into practice. Describing a fantastical situation to illustrate a fictional scenario is obviously a deeply different context to an eyewitness account of a car crash while in the witness box in a courtroom. During the latter you would hope for as little involvement of the imagination as possible, especially considering how commonplace inaccuracies are with eyewitness accounts, and just how fallible the human brain is as an information storage/recovery unit. When the eyewitness may describe a car, they are using imperfect language to represent an underlying reality – the image will show it.
In that courtroom setting even a very eloquently articulated AI formulation from eyewitness testimony would be less valuable than a CCTV recording, or mobile phone snapshot. Imagine for a moment any historical photograph you have seen, hold it in your mind and think about how well you can describe what you are thinking about.
Think about something that happened to you yesterday, and whether or not you would be able to prompt an AI to recreate it in the same way that a photograph of that moment would offer. The photograph can offer near-perfect fidelity, representations of reality not just in the aesthetic but also to a degree of description that the eye and brain can rarely achieve in optimal conditions, let alone when under stress.
Having said that, there is a very analogous situation where testimony is used to produce something usable in an investigation – police sketch artistry. These roles work hard to determine what aspects are imagined and what is derived from memory, in order to rotate between features and fine-tune a likeness to then put out as part of an investigation. This could be a very appropriate application of AI as they already operate in a similar way, using adjectives and “prompts” to render visually what is being described linguistically. Even so, a photograph would be more valuable than an extensive description from memory.
Trust in Photography
Some areas of photography require grounding in reality. Some kinds of photographs contain such significant information as to demand a level of trust that other photographs from other genres simply never would. A passport (or any ID) photograph must be a depiction of the passport holder’s likeness; if it does not it may be an issue to use that passport as a passport. Some photographs may alter public opinion on food safety, animal welfare, and even active combat. If AI were ever hinted towards as playing a role in these images, it may result in life-or-death consequences.
I don’t photograph in order to show off my imagination. I photograph what I could never imagine, even when that turns out to be something mundane. Photojournalism and documentary photography require that level of trust that what you are being shown was really what happened in front of me. Not just my vision, but a slice of real life, unprompted and unadulterated.
If I make a claim in one of my photo essays that I documented something in another country, then I need the plane tickets to back that up. I need the unbroken chain of events that saw me travel, arrive, work, and make the images I have claimed to make. My workflow with film offers me a way to keep this chain of custody entirely physical, with me present at every stage from exposure to development to print, without ever needing to involve a digital electronic step.
The film negative is not ambiguous or ephemeral. If someone questions one of my images, is it real? Is it manipulated? The negative will prove my result while my contact sheet will prove the context. Both negative and contact strip can be compared against the alleged story of their provenance. Any fraud or fakery will be exposed.
A digital file is removed from context as soon as it is shown in isolation, separated from sequence and source. A film negative will always exist alongside the images from the same roll, inescapable from what came before and what occurred after. The physical nature of film means it can be examined in person, and an “independent” print can be produced from it. A digital image has no such method for verification.
This is not unique to AI-produced digital imagery. Photoshop and other forms of image doctoring have been with us for decades, and longer still have been the techniques used in the darkroom to alter the print from a raw projected negative. Dodging, burning, split toning, masking, double exposure collage, and many other techniques have their origins in the darkroom. However, the difference was, and remains, that that physical piece of film is an artifact that was physically present to be exposed onto, and can be compared against any print or scan to see what manipulation has occurred.
Photoshop fakery in the digital era is common, although frauds do not tend to last long. Reverse image search software makes it easy to see whether anyone else is claiming an image as their own, including when “pieces” of a photograph are used. Perhaps the most blatant example of this kind of fraud is the well-documented exploits of Souvid Datta who passed off collages of his own work spliced together with work from other photographers, as well as straight up passing off work by other photographers as his own.
Datta was caught out because someone very familiar with the works of one of the plagiarized photographers spotted one of their subjects, inverted and placed in the background of Datta’s image. All it takes is one suspicion for the lie to unravel, and from that point, everything the photographer has worked on has been in the shadow of this unethical decision. An unimpeachable reputation is a fundamental trait to a photographer who claims their work represents reality, and it only takes an allegation to dissolve such a reputation.
Why allow even the possibility to be questioned in such a way? Whether it’s Photoshop compositing/collaging or a prompted AI invention, if you’re pretending that your image is something other than an impressionistic rendition, you should recognize that you’re actually removing, not adding, value in that pretense.
Stories and Illustrations
A painting or drawing doesn’t need to pretend to be a photograph to have value as a painting or drawing. Before photographs were the main currency of accompanying imagery, we had illustrations – artistic renditions to supplement a piece of writing. If you produce an AI illustration to accompany a story or to depict something that you were not able to photograph, then it has that value as it is. If you need to pass it off as something it is not, then what value have you actually brought to your project?
Imagine, for example, a biographical article about the life of a recently deceased musician. The header image can be a portrait made during their life, and the copy can be populated with photographs from performances, backstage, recording sessions, and so on. This relies on the presence of a camera at these moments, which are all part of their performing persona. Someone may share an anecdote for which there is no photograph to represent that moment, which leaves four options: leave it un-illustrated, use an unrelated photograph, use an artist’s illustration, or use an AI illustration. The second two of these are effectively re-creations, not unlike the way the news may produce a “reenactment” of a story they do not have footage of to show.
However, those reenacted, recreated impressions cannot claim to be the exact moment the way a photograph can. In a way, in the context of that biography, the photograph IS the story, and the writing is the illustration. The photographs show real moments from the life, while the words elaborate and contextualize.
Does it matter to you if someone calls your image into question? If they accuse you of lying about the provenance of a photograph, or even a series of photographs, will it badly affect the impact you want that work to have? If yes, then what steps can you take to avoid this? Just how much transparency can you offer if you have an exclusively digital workflow? How would you “prove” your process? A doctored darkroom print is still beholden to the negative it was stenciled from, which may prove or disprove by comparison. What do you have in your digital workflow that provides this same foundation of felicity?
One of my current long-term projects involves documenting obscure religious rituals and artifacts. If I were to describe one of my photographs, a collection of sticks bound with string to act as a binding spell in a community apple orchard I am sure it would offer me an illustration, which may even be close to the photograph I made. However, I would only know to ask for this depiction because it is something I physically found, if I hadn’t had the real-life discovery of the object then I wouldn’t know such a thing existed or researched them further.
Without that physical process, I may have asked for an AI to offer me some still life of pagan artifacts, and I am sure it would have delivered, but it would only be based on things that other people had discovered, photographed, and keyworded in such terms. It would not be a grounded-to-reality depiction of a real thing in a real place.
In this way, what film represents is not just a different process to AI illustration of an idea, but a process at all. Physical travel, real conversation, interactions, and primary research resulting in filled notebooks and exposed rolls of film. Images that no one else can claim to have made; backed up and verifiable physically via the existence of the physical negative onto which light was exposed, and silver chemically processed to reveal. Moments that are beyond what I could imagine, and situations that I would not be able to describe even in my notebook afterwards without the camera to remind me exactly what happened, to whom, and in what order.
About the author: Simon King is a London-based photographer and photojournalist, currently working on a number of long-term documentary and street photography projects. The opinions expressed in this article are solely those of the author. You can follow his work through his documentary collective, The New Exit Photography Group, and on Instagram.