Artificial Intelligence (AI) is a hot topic in photography and particularly in relation to photography software. This seems to have happened suddenly as a number of companies (Luminar, On1, Adobe) have adopted AI as a key part of their marketing, but the same trend is being seen in relation to some camera features, especially phone cameras.In response, there have been a number of “serious” photographers who have seen this development as either mainly fictitious, marketing hype that is promising what it cannot possibly deliver, or as a serious threat to the future of photography from the further intrusion of technology to dumb things down and destroy creativity. Some have seen it as both, which seems to be at least a little self-contradictory. It’s only a threat if it is real.
Like a lot of photography discussion online, there’s a mess of confusion and vague and ill-defined terms, not only in relation to AI itself, but what I think is a fundamental misunderstanding of what photography is.
I honestly don’t know yet whether any of the software actually does employ AI, and if so in what sense and whether what is being described as AI is going to be helpful to me or other photographers, or not. It’s also hard to be clear-sighted in a discussion in which there is obvious huge commercial and marketing stakes. These are clearly in play with the recent upsurge in using AI to label what some software does, to differentiate it from a market dominated by a few large corporations and to re-sell what is largely existing software with a new twist. There have been “AI” features in some software for a long time, but the developers have not always used or emphasised that terminology. For example, you can search in Apple Photos, not only for things like filenames or capture dates or camera type, but for elements of the visual content of the photos e.g. searching for “dog” will probably show you photos that contain a dog in the image. It’s far from infallible, but it would have been impossible on any level even a couple of years ago and is very definitely employing techniques adopted from AI research.
Even without the understandable confusion around AI, what it means, how it works and how far what is in software is actually AI, there is a widespread confusion about the nature of photography (and all visual art) itself. This is deeply philosophical, but important. I’ll over-simplify (and like all philosophy everything is disputed) to make my point.
People make images. Images are not the same as what they represent. For example, if I draw or paint or take a photograph of a famous building (e.g. the Taj Mahal), quite obviously the drawing, painting or photograph is not the Taj Mahal. At the same time, I would hope that people seeing my image would agree that it portrayed the Taj Mahal in some way. It might show its proportions and colours so it can be recognised and so you would agree that I had “captured” it.
Depending on how I had drawn, painted it or photographed it, some of the people who see the image I made might feel that it portrayed some perceptions, meaning or emotions about the Taj Mahal too - its beauty or timelessness perhaps, or a contrast between its size and the people near it or even the half-felt necessity of love and death. I’d expect some people who saw the image to have a range of reactions to it, just as I would expect them to have a range of reactions to visiting the Taj Mahal itself and I’d hope that at least some of them would react as I hoped they might, and that the overall consensus was positive, but I’d probably feel that my judgement of my own image was the most important one.
There are whole libraries devoted to discussing the issues arising from those last two paragraphs. Are reactions to images purely subjective? Is there any such reality as “beauty” or is that just a way of saying that we like something. Are some ways of making images more “valid” than others and what is the role of the image-maker’s conscious intention? It’s all very interesting, but I hope the basic point is clear. We **make** images and images are representations of something, not the thing itself. It follows, that different people will make different representations of the same thing. We do that by making an almost infinite number of choices about how to make an image from what medium to use, what to include and leave out, what viewpoint to adopt, how big to make the image and that’s before we make what are often called “artistic” choices, for example to adopt a colour palette, to blur some details, to draw lines that we feel, rather than just what we see, to consciously echo other people’s or our own previous images and so on.
Turning to photography. A photograph (on screen or printed) is also a representation of something, not the thing itself. Because it is possible to aim a camera at a scene, press a button and have an image (especially today) without having to consciously make anything, we can tend to see photography as intrinsically different to other way of making images. We talk about “capturing” photographs and making drawings or paintings. The idea is that a photograph can be a kind of “pure” representation. That the camera can’t lie. That there is something uninterpreted about photographs, at least unless someone manipulates them.
We see this idea reflected all over the place. The “no filter” tag on social media. The intense hatred of “photo-manipulation” from some “art” photographers. The proud boast that the image is straight “out of camera” or “no photoshop”. It’s also part of a strong reaction against any idea of AI or even single slider or other automatic or simple processing of images.
I think this idea is profoundly wrong.
If I point a camera at a scene I have already made quite a number of important decisions in order to make an image. I have decided what to photograph, when to photograph it, what camera and lens to use, where to aim it, what to include and what to exclude. I have probably decided what light I want to fall on the scene and I have a pretty good idea what the photographic image will look like. I still don’t have an image.
If I was working a century ago, I would have to take the latent image (invisible chemical changes) on the film or plate and treat it through a sequence of chemical and physical processes in order to have anything that someone might be able to look at. I might have intermediate steps like a negative. There are dozens to hundreds of decisions to make, every one of which might have some impact on the image. Even if I stuck absolutely to the “standard” development at every point, people (like photographic manufacturers) had made decisions: how black the blacks would be, what colours there might be, what the surface of a print would look like and so on.
Today, I can choose to allow my camera, phone or computer to take the captured data and to apply all kinds of software and hardware processes to make an image I can look at on screen or print onto paper. There are dozens to hundreds of choices made which impact on the image. A lot of those might have been made by the camera or phone or printer manufacturer (and which I chose to buy) but I will have the ability to control some of them and how they are applied and I can probably exercise a huge amount of control over the final image if I wish to.
I need to say that there’s no problem with adopting particular approaches to any art form. If your photographic practice involves changing nothing more than the aim of your camera and the framing of your scene, and therefore choosing your time and lighting very carefully from what nature provides, and then choosing not to change any settings in software, to always use the same printer and paper or screen, that’s fine. There’s a long and very honourable tradition of artistic practice which involves closing down the choices you make. Similarly, if your practice allows you to change camera settings but not computer ones, that’s fine, but be aware that there is nothing at all that makes that intrinsically better than another photographer’s decision to take control over every possible aspect, at every possible stage, from the initial idea to the final image.
Deceit is wrong, of course, but every photograph is the result of “artistic” choices and how well you have executed them, just as every drawing or painting is. It’s all art. Some is great, a lot is not, but fundamentally every time a human makes an image, there is an interaction of our intent and technology to portray something.
So what about AI?
I am not convinced there is a single field or approach that can fairly be called “AI”, partly because there is no single definition of “intelligence” anyway. Whether any of the current software is “really” artificially intelligent is far from clear. There is definitely some application of image recognition: so that objects like faces, or features like horizons are being selected in images by software without manual intervention. There is also software that is able to recognise image planes (foreground or subject and background), sky and such things as foliage from clues in the image. Some of this software has been “trained” through “machine learning” where algorithms are developed by building matrices of decisions from large collections of examples.
Like all previous developments in photographic software, as in all developments in technology, this can make new artistic decisions possible. New chemical technologies made photography possible in the first place. Digital technology created a vast range of new possibilities for image creation and manipulation. Software being able to recognise things within images will create new possibilities too, not least in being able to automatically adjust specific image elements or to apply conscious changes to specific elements without the need for the image-maker to select them. As we have already seen, this makes things like replacing backgrounds or skies, or smoothing and blurring skin, or adjusting eye or tooth colour, in software much more practical than it was before.
That will influence photography. When there were only black and white films, photography had to be in black and white, or hand-tinted. That shaped how photographers saw the world and what images they created. Black and White or tinted are still options for photographers, even though the technology no longer requires it. The whole panoply of existing photographic approaches and techniques will still exist and be available photographic choices however far software develops. Will we see more equalised, saturated, contrast-adjusted, neutrally balanced images with smoothed skin and idealised facial and landscape features? Of course. We already are. Will that be a bad thing? Probably not. Great photographers will adapt and develop with the new tools and make images that were never possible before. Many photographers will go with the flow.
It has always been like that. It always will be. But it will still be people making images in which we represent the world and as soon as we do that, we choose to represent it as we wish.