I can’t think of any technology right now that has more hype around it than artificial intelligence. It’s not entirely unearned hype, either – I tend to be pretty skeptical of “the next big thing,” but some of the feats in the AI world are nothing short of stunning. It’s already impacted almost every field in some profound ways, from medicine to art – including, very visibly, photography.
To that end, I’ve seen a few questions float around in various forums, including:
- Is artificial intelligence harming photography?
- Are AI-generated (or AI-supplemented) photos really “photos”?
- Can AI-generated images be considered art?
Those are some deep questions, and I think they call for a moment of introspection as photographers. What really makes a photo a photo? People have been manipulating photos since the dawn of photography, and it ramped up dramatically with Photoshop and similar tools. But right now is the first time that “photos” can be generated from thin air – type a short prompt, and you’ll soon immediately have a picture of whatever you want. Say, a photorealistic image of a ladybug on purple hydrangea flowers.
I can’t see the future, but I have been wrestling with these questions for a while. Photography is my hobby and my livelihood – it’s also a source of many of the most important moments of my life. Of course I don’t want to see it decline or be replaced by AI.
It’s for that reason that I would pose a different, but closely related question to help us answer the previous questions more clearly: Why is it that you do photography in the first place?
Figure out that question, and the rest follows.
Table of Contents
A Brief Background
Before anything else, I want to make it clear what I mean by “artificial intelligence.” It’s a term you’ll see everywhere, but not everyone agrees what it means, and a lot of marketing teams seriously stretch the definition.
For this article, the artificial intelligence that I’m referring to is image generator type – where the AI invents an image from scratch (well, from a database of existing photos) based upon keywords or prompts that you type. At a later date, either I or one of our other writers will publish an article discussing other AI/machine learning tools, like upsampling, noise reduction, and so on. But it’s the image generator AI that has the potential to rewrite photography as we know it.
That brings me back to my question from a moment ago: Why is it that you do photography in the first place? Depending on your answer, you’ll gain a lot of insight into what these AI images will mean for you.
Why Do You Do Photography?
1. Photography for Its Own Sake
To put it plainly, I like photography. Whether I get a great photo or not, I feel the same – it’s really enjoyable just to go out with a camera in hand.
As hobbies go, photography gets full marks from me. It’s creative and analytic at the same time. It rewards your effort and gives you tangible results. It also blends well with other hobbies and professions – I know many scientists, car enthusiasts, art collectors, and others, who turned to photography to supplement those areas of their life.
Taking pictures also gets you moving. I’ve always liked hiking and spending time outdoors, but as a landscape photographer, I find myself going out into nature more often than I ever would otherwise. Many of my favorite memories are moments that I only experienced because I’m a photographer – seeing bear cubs in the wild, getting caught in a sandstorm, hiking a nine-day trek in Iceland, and more.
In other words, photography doesn’t need to be a means to an end. It’s worthwhile in and of itself, even if you don’t bring back any photos that will go on your wall. Many of my favorite days of photography didn’t result in a single good picture. I love the process, and that’s enough.
How does this relate to AI? Quite simply, AI is a different process. Maybe you’re someone who gets enjoyment from typing keywords into an AI generator, seeing what pops out, and modifying your prompt until you have an interesting image. I’m not. The novelty was fun the first time I tried it, but the idea of sitting in front of a computer any more than I already do is almost repulsive to me now.
2. Photography to Record Memories
Do you know what subject I photograph the most? It’s not my favorite landscapes in Colorado, or even product photos for my reviews at Photography Life. It’s iPhone pictures of my cat.
No, that answer does not evoke the spirit of Ansel Adams or Henri Cartier-Bresson and the etherial magic of photography – but in my defense, he’s a cute cat. When he does something silly, I take a photo, end of story.
For most people, especially non-photographers, photography is about recording a memory. How many of us got into photography in the first place because we wanted a way to capture memories from a trip or pictures of our family?
I don’t need to write a thousand words to say that AI simply cannot touch this side of photography. At best, one day, you’ll feed an AI generator some photos of yourself, type a few sentences, and get something vaguely resembling a memory you have. Any time that you want to really document a trip, a concert, a wedding… forget it. AI will never replace photography in this area.
3. Photography to Earn Money
Let me start by saying, if you went into photography to earn a lot of money, I question your judgment (although I congratulate you if you succeeded). Every photographer I know started photography first because they loved it, and only later – if at all – turned it into a career.
But let’s say that you don’t care about photography, ethics, real images, or any of that nonsense. It’s all about the money, baby!
In that case, you might have something to fear from AI, although only if you shoot for some very specific industries. There will always be room for wedding photographers, real estate photographers, high school portrait photographers, and most of the other types of photography that earn decent money today. No matter how convincing an AI “photo” may look, it won’t be very useful if your client needs a photo of their house, and you’ve just generated a different house from scratch.
I’m sure, however, that real photography will lose ground in advertisements and product images. That’s already been the case for a while. (Many of the generic product photos you see on sites like Amazon today are not actually “photos,” and they haven’t been for years – although for now, they’re mostly 3D models rather than AI-generated images.) I’m also concerned for stock photography, which already has declined massively as a profession and will likely see an even greater loss in revenue in the coming years.
What if you’re clever and want to turn the situation on its head… to earn more money by creating and selling AI work yourself? It might work if you’re just trying to gain a following on Instagram. But if you’re selling to clients, not a chance. No client will pay you $1000 to generate an AI image to match their vision once they realize they can generate it themselves for free. There’s maybe a 1-2 year window right now where you can make money this way, because a lot of people don’t know how to generate these images themselves. After the technology has progressed just slightly further, you’ll need to find a new job.
In short, if earning money is your main purpose with photography, there is a chance of AI upending your career – it really depends on what type of work you do. Still, for most professional photographers, apart from advertisement photographers, I wouldn’t worry any time soon. Few people would want a computer algorithm (which wasn’t even there) to imagine their wedding from scratch, or to invent the winning moment in a football game based on a description of that moment.
4. Photography to Create Art
One of the main reasons why I do photography is to make art. You can call it pretentious if you want, but when I put creativity into my compositions and my prints, I see that as art. It is a way to create something tangible out of ideas and emotions, putting an image from my head into reality.
I’ve always had a broad definition of art. I think that any process which involves our creativity can be art – whether that’s cooking a good meal, singing a song, taking a photo, or painting a classical painting. Some people have much more limited definitions, and I’m not here to change your mind on that. But for someone like me who is willing to call a lot of things “art,” how does AI imagery fit in?
Despite my broad definition, my first reaction is to dismiss AI imagery as not art because it lacks the human side of things. That’s basically the view that the US legal system is taking right now in not allowing people to copyright AI images. Then again, I can’t deny that there is still some human element involved in making an AI image. You need a bit of creativity in order to choose the right keywords to type, then cull through the results and modify your keywords to get an image that you like, or that matches what’s in your head.
I would make the following comparison. Let’s say you decide that you want to display a seashell on a shelf in your home. You go to the beach and pick up a nice seashell, then discard it when you find a better seashell, and so on. When you get back home and put up the seashell, is that art?
AI is basically the same. You didn’t create the images, but you did find some that matched the idea in your head. Maybe that’s art to you, and I’m fine if you think so. Personally, I think it lacks the creativity that goes into other art forms, so I don’t consider AI-generated images to be art. And yes, people said that about photography in the early days and were proven wrong. But even if I’m proven wrong about this, it’s still how I feel about AI imagery today.
Conclusion
Let me return to the questions I posed at the start of this article. Hopefully the answers are clearer now.
- Is artificial intelligence harming photography? In some areas, yes. If you have a career as an advertising photographer, you may need to find another job in the next few years. And if you’re trying to gain a quick following on Instagram, you could be crowded out by AI-generated work. But in most regards, no. AI will never replace the fundamental reason that most people take photos: to capture a memory. Nor can it replace the photographic process – just going out and taking pictures – that attracted many of us to photography in the first place. And if you do photography as a form of art, AI is likely to be an unsatisfying substitute.
- Are AI-generated (or AI-supplemented) photos really “photos”? No, they aren’t. It’s a separate medium – you could call it photorealistic digital imagery. Photography is specifically defined as “drawing with light.” AI-generated elements don’t involve any more light than something like drawing a smiley face in Microsoft Paint. If you put even a small AI illustration into a photo, it turns the photo into digital art. It’s akin to replacing the sun in a photo with a clip art drawing. Whether that matters to you is one thing, but no, AI-generated images aren’t photos.
- Can AI-generated images be considered art? Sure, they can – everyone’s definition of art is different. I have a lenient definition of art and am still not willing to count it, though. Art is about creativity, and there’s not much room for creativity in the AI-generation process (basically choosing keywords and culling the results). Making a sandwich is a higher art form with more room for creativity. Although, AI images could always be a form of art if you add some creative flare of your own to the mix, like printing it out and shredding it.
There are many places where AI-generated photos concern me. Because it’s so easy to fake a scene (like the Pope image from a moment ago), it will be even easier to fool people than ever before. Propagandists will publish photorealistic AI images and contrived illustrations meant to inflame their readers. It’s already happening to a worrying degree. Not to mention that people will trust real documentary photos less and less, especially when they show something “inconvenient” to their beliefs.
As worrying as that is, it’s different from the question of whether AI will replace photography and render it obsolete. I simply don’t see that as possible. Did photography replace painting? It changed painting, but did not replace it. Likewise, artificial intelligence will surely change photography, but it can never capture genuine wedding photos or pictures of our families growing up. Nor can AI replace the reason that many of us take photos in the first place, which is simply that we enjoy it. To me, bringing a camera out into nature will always be more fun than typing words into a box. There are real reasons to worry about artificial intelligence, but outside of a few specific sub-genres, rendering photography obsolete is not one of them.
Based on the previous AI-related articles we’ve written on Photography Life, I’m willing to bet that most of you reading this are, to put it mildly, not fans of how the technology is being associated with photography. But I’d like to hear more. This is a nuanced issue considering that AI technologies have some obvious benefits in photography, like better autofocus tracking, despite all the potential concerns. Where do you draw the line? Are you willing to use software like AI upsamplers and noise reduction? What about AI spot-healing tools, or even more extensive edits than that? I’m curious to hear your thoughts in the comments below.
Thank you for this enlightening. I learnt photography and as I heard about AI I was dispirited. But with your article, I better pick it up again. I know I’d forgotten somethings in the photos, but I’ll sure get back on my feet.
Thanks for the thoughtful article, Spencer. It’s got to be the goin’ not the getting there that’s good. . . .
Yeah, this is a nice topic of conversation.
It’s quite hard to draw the line of where using AI in a photo makes it stop being a photo.
Myself, I use Noise Reduction, which I think that doesn’t alter the status of your photo, is more akin to color grading or something along the lines, as it doesn’t alter the image you took. Spot healing is a slippery slope. If taking a pimple from a face, a bit of garbage on the floor or something that doesn’t affect the end game of the picture, except by making it a bit “better”, I can see the end product as a photo. Taking big things like light lines and poles, cars, and so on, I think that makes it already a stretch.
Also sky replacement, and placing elements is a photo that weren’t there, that’s not a photo, it’s a montage. Nothing wrong with it, not diminuahing it, but I just don’t think photo is the real name of that.
Thank you! And absolutely – I don’t think anyone would judge a photographer for using AI noise reduction. But like you say, that’s just the very beginning of the slope.
Personally, I draw the line at the same place as a particular photography competition that I highly respect: “the integrity of the subject must be maintained.” Small cloning adjustments are as far as I’m willing to go, and even those I prefer to avoid.
That said, whether an AI or a more traditional cloning algorithm is what fills in those spots doesn’t particularly matter to me.
While I agree with most of the points you brought forth, I would disagree in the single instance where AI is used to improve the cloning tool. In the case of cloning something out from a photo, like a bird in the sky (which wouldn’t be there if you waited a second) or a freckle on a face (which is likely gone in a week), I would still consider it a photo as such small interventions have little impact on the reproduction of reality in an image. Anything beyond that I would agree should be considered digital art. Cloning in a snow leopard is obviously a no go 😂.
Nevertheless I hope the advent of AI will force some innovation in photo formats, perhaps a “secure” sRAW format which contains timestamp+photographer’s ID and ideally GPS coordinates could be introduced to help prove authenticity. Such a new file could be adopted by social media and a tick mark confirming an image is actually a photo would help us distinguish reality from fiction.
Definitely – I didn’t mean to imply that AI spot healing is always going to turn a photo into digital art. It would be nice if there were a simple answer, but I think it’s always going to be a question of degrees.
For example, getting rid of a speck of dust on the camera sensor using a good AI tool? I think hardly anyone would see that as digital art. Removing a distant power line the same way? A bit harder to judge. But as the size and importance of the removed object increases, the photo will inch closer and closer to digital art.
I think you’re right that some sort of verified checkmark will be desirable before long, to indicate that a photo on social media was not AI-generated. Figuring out how to verify that would be tough, of course.
You make a lot of good points and I generally agree. Only the analogy of AI beeing seashells you have found does not hold up in my view. You could easily argue that every photo you take is just as much like a seashell you have found. That would disqualify landscape photography as art es well.
In both cases, you see stuff, if in the real world or generated by AI, and then you arrange, i.e. compose, that stuff into a picture. And that picture can be good art, bad art or no art at all. The distinction is just as hard to make with AI images as with photos.
It’s the same thing with music, where the term composition comes up all the time: in classical music we have 12 notes, in about 8 octaves. In the manner we purposefully arrange, i.e. compose, those notes lies the art making.
My point beeing: I can totally see art beeing created with AI imagery, even though I don’t expect to enjoy that process myself.
Great points. I think the big distinction between the seashell example and something like photography or composing music is arrangement.
In photography, you change your own position relative to the scene, your focal length, and the direction you point the lens in order to arrange the relative sizes and positions of the elements in the scene to match your vision. Something very similar is done in music in your arrangement of the notes.
The seashell example would be more easily described as “art” if you gathered a lot of them and arranged them in a particular way. And I think AI images could be a type of art if you modify or arrange them. Otherwise, the seashell and AI examples are just culling or identifying something that looks nice, not creating it.
At its very basic, AI is plagiarism. It is stealing other people’s work without their permission, and making something else out of it. For that reason alone, it will never be art. Real art does not only require creativity, it requires work. Making a beautiful photo is not easy, but as you say, making an AI picture just requires pushing buttons and a little judgment. I don’t really understand why we are letting AI in the door at all. Most people fear it and few people want it, especially in photography. Sometimes it is best to say no to new technology and just not go there.
That’s an excellent point. I can’t find myself thinking of plagiarism as art. It’s always been a tricky subject with artists like Andy Warhol approaching the line of plagiarism and others like Richard Prince unashamedly leaping over it. To me, if you’re not going to create something yourself, why even bother with it?
I also agree with you that art takes work. I’d even say that most people see it as “more work = better art.” And that’s partly why photography took so long to be accepted as an art form that can stand beside classical paintings. Personally, I don’t think that saying “more work = better art” is always correct, but there’s a correlation. I’d say the average symphony involves more creative, artistic effort than the average song.
I agree that more work = better art is not true. It is not how much work, but a question of degree, if I could put it like that. I once did a painting in a single afternoon (the only time that ever happened to me), but in that timespace it took thought, application, the vision of what it would be, the eye to rule out mistakes, emotion and the all important understanding of when to stop. It just so happened that things went right. The effort was exhausting but gratifying. The result was pretty good, and definitely art. It could be said that to create art, it takes what it takes. More work will not make something better and can ruin it if you don’t know when to stop. That phrase should never be used to define what qualifies as art. The symphony and song comparison is really one of apples and oranges. The two are just not comparable. Besides, if you are a McCartney, it may not take much effort to create a wonderful song. But if you are a Beethoven, a symphony might just pour out of you, and it is just a question of how long it takes to get it onto paper. But what if Beethoven tried to write a song? It might have taken him weeks. At the end of the day, I think that most of us just know what art is. And AI will never be art. Maybe it might look like art, but with no real creative process, made of stolen bits and pieces of other people’s efforts, it will not even be next door to art.
That’s a really good way to put it. And you’re right, sometimes overworking a photo, painting, or really anything, can cause it to lose the initial spark that drew you to it in the first place.
Like with all new Technology, AI frightens people for it is new and often not well understood.
However, in this time were fakenews is common it should be clear if something is (partly) generated with AI or not. In a few years time AI will be more refined and it will be harder too see if something is real or AI generated.
The same problem already exists with press photography where photoshopping is not done and a raw file has to proof that the image is genuine.
Ai can be used in many areas in a positive way, however we must always realize that Ai is not real intelligence and the outcome cannot be trusted to be right.
I saw an exampel hat AI could recognize a tractor in a photo, but by changing one pixel it was completely lost, while people would still see the same tractor.
We’re getting extremely close to that point, where AI images are indistinguishable from real images. It depends on the subject. AI still struggles with a few things, like people’s hands and ears. But in a lower resolution image, I think I’d be fooled more often than not, if a well-done AI image was shown to me.
The tractor example you mention is an example of an adversarial attack to fool an AI processor (www.dremio.com/wiki/…cks-in-ai/). I’ve seen something similar where changing the noise pattern in a very particular, calculated way has drastic effects on what an AI thinks is in the photo. A picture of a koala is labeled as a fire hydrant, or something like that.
The big concern is that someone with malicious intent could exploit this in disastrous ways if we as a society start to rely too strongly on AI. For example, placing a few small stickers with a particular pattern onto a stop sign can make a self-driving car think it’s a 45 MPH speed limit sign. This is not a hypothetical example, either. www.autoblog.com/2017/…-stickers/