In March the Sony World Photography Awards announced the winning entry in their creative photo category: a black-and-white image of an older woman embracing a younger one, entitled PSEUDOMNESIA: The Electrician. The press release announcing the win describes the photograph as “haunting” and “reminiscent of the visual language of 1940s family portraits.”
But the artist, Berlin-based Boris Eldagsen, turned down the award. His photograph was not a photograph at all, he announced: he had crafted it through creative prompting of DALL-E 2, an artificial intelligence image generator.
“I applied as a cheeky monkey, to find out if the [competitions] are prepared for AI images to enter. They are not,” Eldagsen explained on his website. His stunt has sparked controversy and conversation about when AI-generated or assisted images should be considered art.
On supporting science journalism
If you're enjoying this article, consider supporting our award-winning journalism by subscribing. By purchasing a subscription you are helping to ensure the future of impactful stories about the discoveries and ideas shaping our world today.
Scientific American spoke with Eldagsen about the image’s creation and the future of AI-aided “promptography.”
[An edited transcript of the interview follows.]
How did you get started with AI art?
I started with photography because drawing was a lonely job. I was always experimenting. So when AI generators started, I was hooked from the very beginning. For me, as an artist, AI generators are absolute freedom. It’s like the tool I have always wanted. I was always working from my imagination as a photographer, and now the material I work with is knowledge. And if you are older, it’s a plus, because you can put all your knowledge into prompting and creating images. If I were 15, I would have probably just generated Batman.
Where did the inspiration for The Electrician come from?
I did it for myself as an exercise, and I just love the result. It sparked off from a project that started years back. My father was born in 1924. So he went to war when he was 17 but, like most of that generation in Germany, never talked about it. After his death, I found some images from the forties my mom and I hadn’t seen before. I learned a lot about their time just looking at these images, and I started to collect images from the forties at flea markets, and also on eBay, but didn’t know what to do with them.
So my first experiment was: Can I re-create images of that time using AI? And then “The Electrician” just grew. The best images are those you didn’t have in your mind before. They came out of the process. You start, and it leads you somewhere—with AI it’s the same. You start somewhere and then you make many different decisions. You delete elements, you add frames. Sometimes the AI has very good suggestions. Sometimes it’s just crap. That takes time and patience, so it’s not finished in 20 seconds. It can take days.
So how did you actually make this image?
I used DALL-E 2, and it was all done by text prompts and inpainting and outpainting. For inpainting, you could say, “I don’t like his tie,” and you erase it and write, “I want him to have a white tie.” Then you get suggestions. And if you don’t like any of those suggestions, you start again. Outpainting [is what] you do when the frame is not large enough. You put in an additional frame so you can see his whole tie, his pants, the chair, the floor. It’s endless.
Why did you decide to submit it to a photography competition?
I’ve been very involved in AI and photography—I’ve become one of the experts in Germany—so it’s not just me poking fun. I wanted to test if a competition has taken into account that AI-generated images can be sent in. I applied to three different competitions, and the image always was a finalist. There’s something about the image. When I apply, I don’t say it’s AI-generated. I keep the information very short: just the image and the title. Then when it was selected, I said the art is AI-generated.
What I was hoping for has happened: the conversation has started, and it was basically with the help of the community. I did not expect it to be that big; I thought it would have been a conversation for a week in the European photo scene.
Do you think someone could have figured out that it was AI-generated just by looking at the image?
Of course. There’s a difference in color that comes through outpainting. On the left side it’s too yellow, and on the right side it changes into black and white. And then maybe the fingers, but also part of the arm on the right side, you can tell. If you work with it on a daily basis, like I do, you can tell.
Have you ever been fooled by an AI image?
There’s a German magazine called GEO; it’s something like National Geographic in Germany. They had an online test with images, asking “Is it generated or authentic?” I failed once.
I think with The Electrician, it’s very easy to tell because it’s an old image from early last September. But I think by the end of this year, we won’t be able to tell.
Does that alarm you?
As an artist, I just love it. As a citizen, I’m deeply concerned. Most kinds of photography can be augmented by AI but not the photojournalism part. The press needs to come up with a system to make it clear what is authentic, manipulated or generated. The Pope Balenciaga AI-generated photo should have always been notated. If you don’t do that, democracy will be manipulated and misinformed by anyone who can write five words.
But to fact-check is a lot of work. That takes time. So for the press to do that, to pay all the staff and to also have AI technology to help—who is going to pay for this? Now as the citizens, I say, we cannot let the press work alone. It’s very important for a democratic society [to be able to distinguish real photos from fakes]. So we have to think about the structure to co-finance that [fact-checking] as citizens, as a democratic state. But how can we co-finance it and still maintain the freedom of press? This is something we need to think about.
So the future of democracy and journalism aside—how will AI fit in the art world?
One thing I propose is to clean up the terminology and not call realistic AI art “AI photography” anymore, because it’s not photography. And one suggestion that came out of the community was “promptography,” and I just love it. It is large enough to encompass that the result can look like a drawing, like a painting, like a photo.
The next step would be to talk about the relationship between promptography and photography. Do they belong into one basket—one museum, festival, gallery, competition? It’s very complex. And I don’t have any answer for that. The only thing I can say is that the easy answers on both sides—those who want to go back to analog times and those who say promptography is photography—are nonsense. We need to think deeper than that.