AI-Generated Images are Here
Humboldt County, California
Artificial intelligence is changing the way we create images. AI-generated images can win photography competitions, and real photographs are mistaken for AI-generated images and rejected. I find my opinions continually changing the more I consider the new technology, work with it, and see how others use and feel about it. In this article, l share some current thoughts from my perspective as a photographer and teacher of Digital Photography at College of the Redwoods in Humboldt County, California.
A little background first.
When my folks gave me my first camera as a high school graduation present I had no idea of photography’s place in history. Photography was new to me, it was exciting, it was fun and wonderful, and it gave me a way to express my creativity that felt liberating and energizing.
Taking classes at Humboldt State University, now Cal Poly Humboldt, gave me some historical context. I was shocked and disappointed to learn that as photography gained popularity in the early 1800’s, my beloved creative outlet received a frigid reception from the established art world. The art establishment had felt threatened by it: photography used strange and new processes, it made images unlike any others in the art world, and it seemed that nearly anyone could do it. Many traditional artists felt that there was neither skill nor art to it, nor ever could be.
It was a fearful reaction, and in general an overreaction, but in some cases the threat was real. In one obvious example, business fell for painters of portraits as business boomed for the new wave of photographers who could make less expensive photographic portraits for their clients. But if portrait painters suffered, think of the priceless boon this was for people who had never before been able to afford to own painted portraits of themselves and their loved ones. Now with photography, it was financially in reach for people to record their likenesses for themselves and their posterity. It is hard to imagine today, but there was a time when few people could afford to have any images of their family, and most would not have known what their ancestors looked like. Photography changed that forever. Indeed, thanks to photography, I have photographs of my own ancestors dating back to the mid-1800s.
Photography opened up image-making to the masses, and from our hindsight 200 or so years later, we see the obvious benefits of photography. Beyond making art, photography has become an indispensable and unparalleled tool for documenting visual events, people and things in the physical world around us.
Photography didn’t destroy the art world. Rather, it added its unique flavor and made its own place in the art world, for with skill and creativity a photographer can make works of art after all. Photography is taught in college art departments, including at College of the Redwoods, where I teach Art 35, Digital Photography.
Now come images generated by Artificial Intelligence (AI), and I find that I’m one of those feeling iffy about the new form of image-making. That’s some delicious irony, isn’t it?
Long before the new AI boom, I taught techniques for compositing photographs — the combining of multiple photographs into a new image — using Photoshop’s traditional tools. So that students would feel the greatest ownership of their images, I encouraged them to create their compositions using photographs that they took themselves, rather than scouring the internet for images to use. I also allowed students to incorporate internet images that were marked as free to use, but I shared with them that when I make a photomontage that I have composited entirely from my own photographs, it feels more rewarding, more fulfilling, than when I create an image using other people’s photographs. When all the elements in the final composition are my own photographs, as in “What Lies Within,” then every aspect of the final image is from my own hand, my own eye, my own imagination. I made it, and it feels like mine. But will people now mistake my composited images as merely created by AI?
Artificial Intelligence can generate images for us from a description that we provide. I have experimented (played) with making whole images from text prompts with Firefly, Adobe’s AI image generation software within Photoshop, and it is fun. It’s really fun — but currently only the beta version of Photoshop has this ability. I think using AI to generate images will be a tremendous source of entertainment for many people, and will provide a creative outlet for those who would not otherwise have experienced creating images.
But having been a photographer for a long time, and having used Photoshop to manually edit and combine photographs for almost 30 years, I can say that the processes of creating a good photograph — and for making a composite image in Photoshop using my own photographs — feel more real, visceral, and more fulfilling to me than making requests of an artificial intelligence and seeing what it gives me. And there is the rub for me: if I’m waiting to see what the AI gives me, then that image cannot possible be completely mine.
As an artist, the way I feel about my images comes down to ownership for me. Did I make the image, or did the AI make it from the typed-in phrase I gave it? Making an image while using too much AI feels similar to the way I feel about making a photomontage while using photographs that other people took: it’s not 100% mine. When I lean too much on AI to generate parts of an image (or a whole image), I no longer feel like the sole creator, but a collaborator with something else, because some of it was chosen by an AI. What constitutes “too much” will be a matter of opinion for each of us, both as viewers and as artists.
As a viewer of art, I have a feeling of creative connection when I know that an artwork was made by a human artist. We humans share a lot of experiences that resonate with each of us. When I know an image was made wholly by AI via text prompts, I feel that much less human connection with the artist. The artificial thing that generated it had no experiences in common with me, and it doesn’t reach me as deeply on an emotional level. How can I connect on a human level with an image made by an artificial intelligence?
Yet AI can absolutely be a tremendous tool for photographers and other visual artists. AI can be used as little or as much as one likes in image generation; its uses span a spectrum from simple tool for streamline a process to an “artist” that can make entire images. As a tool, AI can help in such things as restoring damaged photographs, removing power lines, pimples and unwanted people from your photos. AI can be used creatively to enhance and alter original human-made images when creating fine art: for instance, it can make very quick work of selecting a subject you want to modify, or removing a background, or filling in a gap, among a myriad of other uses. And beyond all of that, we can use AI to generate entire images for us from a text prompt, or description.
When we have Artificial Intelligence generate an image for us, AI becomes the artist, and we’re the client making a request. When a person or business needs some artwork, they often go to an artist. The person or business is a client with an idea or request, and the artist then brings that idea to life. What about when an artist asks AI to make an image, or part of an image, and the AI brings that idea to life? To the AI, this is a request from a client. The artist becomes a client/artist requesting artwork, and will accept or reject what is offered, while the AI is the artist presenting the client/artist with image options.
I’ve seen wholly AI-generated images mistakenly called photographs. An image generated entirely by AI is not a photograph, though it may look real. A photograph by definition is the result of a process in which light [“photo”] falls onto a photosensitive surface such as film or a digital sensor, and is converted chemically or digitally to an image [“graph”]; thus a photo+graph is a “light image,” an image made from light itself. An AI-generated image is its own form of digital image, not a photograph, though it can have a photographic look.
AI has recently caused confusion at photography exhibitions. In one photography competition, an AI-generated image won top honors. It turned out that the photographer entered what can be called a synthograph (an image generated synthetically, as by AI) to make this point: “I applied as a cheeky monkey, to find out if the [competitions] are prepared for AI images to enter. They are not.” ( Scientific American “How This AI Image Won a Major Photography Competition .”) In another photography competition, an actual photograph was rejected because it was incorrectly assumed to have been made by AI ( The Times “Judges apologise after contest photo disqualified as AI fake.” ). How will judges know when an image is a photograph, or something generated by AI? How will you know?
My creative feelings sink to think that soon my own digital photomontages, and even some of my pure photographs, will be mistaken for and dismissed as AI-generated images that were made from text prompts. My photographs and photomontages and all of the thought and feeling I put into them — my real, human, creativity as an artist photographer — and all of the skills built from years of practice, will now sometimes (often?) be mistaken for and dismissed as artificial art with a casual, “Oh, that’s AI-generated.” Now that is disheartening.
I know that the Art world will reshape itself to accommodate AI, as it did with Photography. We are in the early stages of this technology; its place in art is evolving even as the technology itself is evolving. Its ability to produce realistic images will dramatically increase, artists will figure out how they want to use it, it will work itself into the fabric of our society, and we’ll all carry on.
In my Art 35 Digital Photography classes at College of the Redwoods, one of the many things we will explore will be how Artificial Intelligence fits into photography, both philosophically and as a practical tool.