Moritz Neumüller in conversation with Boris Eldagsen - Images: Boris Eldagsen - March 31, 2023
Moritz Neumüller: A few days ago, on 14 March 2023, you won the Sony World Photography Award in the "Creative" category with an image generated by artificial intelligence. We're talking about The Electrician from the series PSEUDOMNESIA, according to the press release «a haunting black-and-white portrait of two women from different generations, reminiscent of the visual language of 1940s family portraits».
During the selection process, you already drew the organisers' attention to how the picture had been created and offered to debate publicly how to deal with this kind of image creation in photography competitions. It should be said that you have been working intensely with AI for some time now. You also give workshops and are part of technical advisory boards, so you are a person ideally positioned to lead this discussion and point out the acute need for action for us as a photographic community.
Boris Eldagsen: Yes, I started applying with my synthetic images last autumn to find out whether festivals and competitions were prepared for them. At that time, AI was nothing new, so it could have been taken into account. I was among the finalists in three competitions with the same image and explained the authorship of the image to the organisers.
The result was always the same: They felt cheated, but didn't want to talk about it. In the case of the Sony Awards, I was contacted in January and provided the details to my Instagram account and the webpage where the creation was described in detail. If anyone had looked at that, it would have become obvious that the images were AI creations.
Step 1: «Text-to-Image»: create the image with a text specification that defines how the portrait of the two women should look: Description of expression and emotional quality, timing and technique of the photo, lens, shutter speed, aperture, lighting, composition, resolution. The making-of of the winning image of this year's Sony World Photography Award in the "Creative" category by Boris Eldagsen.
«The result was always the same: They felt cheated, but didn't want to talk about it.» — Boris Eldagsen
How many people were contacted?
There are ten finalists in the «Creative» category, all of whom can be found on the award website.
That is a manageable number.
Exactly. Then I said: Beware! If you had looked at all the things I’ve sent you, everything would have been clear, but I now feel the need to re-emphasise that the image was created using artificial intelligence. I did compete with photographs at the Sony World Photography Awards for several years, but without much success. Now I have used my photographic knowledge from 30 years of experience to generate images instead.
I've been doing my experiments for more than a year, a kind of stress test for Artificial Intelligence, and if you come to the conclusion that you don't want that, we'll give a prize to someone else, it would be okay for me. But if you’d say we want to use this as an opportunity to think into the future, then I'd be happy, and would offer myself as a conversation partner to discuss how festivals and competitions should prepare themselves for the current challenge.
And their response was?
My offer to discuss the matter was not accepted. On the day of the press release I have received a lot of enquiries and the agency CREO, which organises the award, contacted me via email. I sent them a statement, which they did not use. Instead, they communicated a generic response that said something along the lines of «we support the dynamics of photography and artistic freedom». So they weren't actually interested at all in a discussion. I followed up again this week and was told, well, we have interviews on our blog from time to time, maybe we can do something there. I think this is a historic opportunity.
Photography has taken on a life of its own as a visual language. The visual language of AI, reminiscent of photography, will live on independently - separate from photography - but will in turn permeate photography pervasively, influencing it either directly or indirectly! A precise term, however, for the emerging AI-generated images is still lacking.
Festivals, prizes, institutions that deal with photography, even universities must take this into account. What do we do with it, how do we go into the future? No one knows, and I think the more we exchange our knowledge about it openly, the better we can set or at least influence the right course.
The problem permeates all areas of photography and visual? culture, not only artistic photography, which is your realm of work and in which you won the award. Of course, it specifically affects journalistic work and professional photography, for example classic product photography, stock photography and fashion.
Yes, it concerns all areas: Professional photography will have a very difficult time because many professions are no longer needed In the field of art, people will be pleased that there are no longer any boundaries; journalism, on the contrary, will be filled up with fake documents that are produced in a deliberately manipulative manner or simply for fun. This is a problem for our democracies that they are not prepared for. It pains me, that I don't know how to solve it either. And then, of course, there is private photography. We will still photograph our children, dogs and cats, but we will use AI to represent our babies in space.
As a member of the Technical Progress Working Group in the Deutscher Fotorat, you speak in front of many different audiences.
A fortnight ago I gave a talk to the German Association of Photojournalists to illustrate how easy it is to fake documentary photography.
This has become even easier since last week with the fifth version of Midjourney 5.
Exactly. These manipulated images are being created at a rate faster than ever before and faster than any of us can grasp. I think by the end of the year, even experts will no longer be able to tell the difference between real and fake. I am complicit in this development, for example with the images I generated this week with the fake Pope or Olaf Scholz sleeping in the German Bundestag.
With the ignorance on the one side and the speed of the technological development on the other, there are probably two fast trains heading towards each other on the same track. You have separated the problems along professions; but I also see an ethical problem in the area of imparting knowledge and truth, both in schools and universities, as well as in politics and reporting, especially photojournalism.
Your winning image was created by prompting, i.e. by text input, but also by techniques like inpainting and outpainting. That is, taking certain parts of the image and replacing them with something else, or telling the computer to enlarge the image in a fairly autonomous way to create things outside the frame. It is part of a series that you present on your website in the category «AI-generated». You emphasise in this context that you deliberately call your images «images», not «photographs». They are synthetically produced and use «the photographic» as a visual language.
You exhibit these works frequently and also offer them for sale in classic photo formats and framed. So you push boundaries yourself and build bridges between the worlds of the photographic and the synthetic, but you always try to be transparent about it.
The tools you use go a step further than those that occupied us not so long ago in the context of post-photography, that is, the discussion about the nature of the photographic image with its purity, originality and character of truth. We’ve had this discussion for years at almost every «World Press Photo Award», whether it is allowed to photoshop or not.
But now we have another dimension. It's about a technique that creates images ex nihil, through a specific poetic language, which can then be refined through further automated techniques. Your intention is to differentiate this photo-poetry from pure photography and to say: Look, this is an image that I as an author have created using DALL-E or Stable Diffusion or Midjourney, but it is not photography. Am I right about that?
Exactly, it's not photography, and I think that distinction is important. I use my photographic language as a starting point to create the image. I use my artistic background. But the result is not created by going out into the world, it comes directly from my brain, using the tools that AI offers. That is the difference.
On the one hand, you can do it the way I've been doing it for the past year, experimenting a lot with text prompts and then trying to «hack» the AI as well and see what you can get out of it. Prompts are complex recipes with 10 or 11 elements. But the latest AIs produce results that are quite photo-realistic with even the simplest text input, such as «Donald Trump gets arrested». And that's something that irritates me on a political level, because I worry about the democratic system. That's one of the reasons why I'm very active in this field.
On the artistic level, on the other hand, it means absolute freedom for me. I am free from matter, I can simply create what I have always had in my head and could not photograph in the real world.
On the subject of selling my images in the gallery, I would like to say that the gallery Photo Edition Berlin approached me when they heard about my AI experiments. They specialise in generative photography and computer art going back to the 60s and 70s and represent artists like Gottfried Jäger and others. So the idea of generating images beyond reality is something that has interested this gallery for years. Hence they find it exciting to explore what is possible as a next step.
«The rules on Facebook regarding the display of nudity were very strict from the beginning, but they were not nearly as strict with regards to fake news.» — Moritz Neumüller
The question of the truth of an image also comes up in another context. The rules on Facebook regarding the display of nudity were very strict from the beginning, but they were not nearly as strict with regards to fake news. And that, as we now know, interfered in political decision-making processes, such as Brexit and the election of Donald Trump. Today, it is no longer about retouching or about false captions, but about being able to create a deliberately false image out of nothing, but which could well be true. And thus the question of ethics has become more urgent and complicated.
Yes. I would like everyone, whether you are the director of a photo museum, a festival or an award, to realise that everything you have done in the past is now changing dramatically. And this at a speed that is enormous. It's happening faster than you can keep up if you want to explore all these new possibilities. It has become a full-time job for me, and yet I would need a 48-hour day to play with all the options and the new tools that are coming out every week. So it's really time for the photo community to begin thinking about the future: we need to think ahead, not lag behind.
Just as politics slept through the internet in the late nineties, the same could happen now with AI: There are no rules, not nationally and certainly not internationally. Legislation is lagging behind. The risk is to accelerate problems that already existed?
Artistic means such as retouching and collages are as old as photography itself, however, what is important is how you deal with them and where you draw the boundaries. At the photo book festival in Aarhus, of which I am the artistic director, we are incorporating AI this year. We are working with students from the University of Dresden and, in addition to a thought experiment and a small show, we also want to organise some kind of competition. Personally, I'm not afraid of mixing different media, including AI, especially in the field of the photobook, artistic photography, visual arts and image culture.
My concern is mainly that the use of AI is not reflected and is not transparent. If festivals and competitions come to the same conclusion as you, that's fine. If they come to a different conclusion, that's fine too. I just want them to be aware of it, reflect on it and then come to a conclusion that they can share.
Not talking about it, not acting on it, is the worst. That's why I continue to consciously apply for festivals where the regulations regarding AI are fuzzy. If the conditions say any device is allowed, then AI generators are allowed and I try it out.
UPDATE 1, April 16, 2023.
Boris Eldagsen has declined the Sony World Photography Award previously given to him in the Creative category at the awards ceremony in London on April 13, 2023.
Excerpt from his public justification:
«AI images and photography should not compete with each other in an award like this. They are different entities. AI is not photography. Therefore, I will not accept the award. I applied as a cheeky monkey, to find out, if the competitions are prepared for AI images to enter. They are not. We, the photo world, need an open discussion.»
Full explanation here
Der ReVue Newsletter erscheint einmal im Monat. Immer dann, wenn ein neuer Artikel online geht. Hier en passant abonnieren.
Sie möchten unsere Arbeit
mit einer Spende unterstützen?
Hier en passant spenden!
Fotografie ist allgegenwärtig, wird aber in den journalistischen Medien noch wenig hinterfragt oder erklärt. Wer an Journalismus denkt, denkt an Texte. Das digitale Magazin ReVue verfolgt einen anderen Ansatz: Es nähert sich den Themen vom Bild her. In unseren Beiträgen untersuchen wir die Rolle und Funktion von Bildern im Verhältnis zum Text, zur Wahrheit, zum politischen oder historischen Kontext. Wie nehmen wir Bilder wahr? Welche Geschichte steckt dahinter?
Unsere Beiträge erscheinen auf Deutsch, wir übersetzen aber auch fremdsprachige Texte und erleichtern so den Wissenstransfer zu einer deutschsprachigen Leserschaft.
ReVue ist unabhängig. Die Redaktion arbeitet ehrenamtlich. ReVue ist ein Projekt der gemeinnützigen DEJAVU Gesellschaft für Fotografie und Wahrnehmung e.V. in Berlin.
ReVue wird unterstützt von