It’s a co-creation, combining strengths.
The winners of the 2023 Sony World Photography Awards (SWPA) – reputedly the world’s largest photography competition – were announced in March. The SWPA Open Competition awards an individual image in each of ten categories assessed in terms of composition, creativity, and visual language. This year, the Open Competition had a single judge, Eric Schlosser, the Art Director of Tbilisi Art Fair in Georgia, and the image he selected to win the Creative section was by the German artist Boris Eldagsen. Entitled ‘The Electrician’, it is a haunting image of two women: the older hiding behind the younger. A man’s hand appears mysteriously from out of frame. The vibe is 1940s retro and, while the image has its own enigmatic beauty, there is something uncanny about it.
Perhaps what underlies this sense of the uncanny are two facts: the young women and her shy companion never existed… and this is not a photograph. The image is from a series called ‘Pseudomnesia’ created by the artist using an Artificial Intelligence (AI) image-generation system called DALL E2.
In recent months, the image-generating capabilities of AI have grown exponentially, as has the uptake in using this form of image creation by amateur and professional alike. The winning of this award highlights some fundamental questions that have wide-reaching implications. When does something that looks like a photograph stop being a photograph? When it stops being a photograph what is it? How do we parse this new visual language? And what impact will this have on photography itself?
To explore these issues in greater depth, I reached out to Boris Eldagsen to find out more about how an image created using AI came to win a major photographic award and what he sees as the future of AI images that look like photographs.
How would you describe the overarching approach and themes of your artmaking prior to engaging with AI?
My creative journey has had a philosophical and psychological approach, travelling inwards. I was interested in the human condition, what we humans have in common beyond cultures and centuries. That’s why, in the thirty years that I photographed, I never documented what was in front of my lens. My goal was to transform what was happening into a timeless symbol drawn from the common human psyche.
What first drew you to working with AI and how did this influence your work in terms of the visual outcomes?
AI also works with the human condition, learning from the images in which the feelings, experiences, and disorders of many people have been expressed. Putting it philosophically: AI has internalised our primordial ideas in order to understand the archetypes of the collective unconscious.
For me, working with AI is a logical continuation of my artistic journey inwards. I now work purely from my imagination. I no longer need models, a location, or a camera; my technical, art historical, and creative knowledge becomes the creative material itself. This frees me from the constraints of budget and the limits of photography.
What is it that an AI image generator does?
Unlike classic computer programs that execute specific predefined processes, an AI is capable of machine learning. It builds on a complex network of pattern recognition that links things by type and interaction, modifying its own programming upwards as it goes. It is similar to the way that a child growing up sees many chairs and begins to form an understanding of what the abstract concept ‘chair’ means. In time the child can recognise something they have not seen before as being a chair or is able to draw a chair that does not look like any specific chair that child has seen before. When an AI responds to a prompt by producing an image, that image is not a collage of fragments drawn from its database, but something new constructed from scratch by the AI based upon what it has learned about images and the meanings associated with the terms used in the prompt. The image is created not pieced together.
How do you see your relationship with AI?
It’s a co-creation, combining strengths. And that is something I have always loved to explore. From the early nineties on, I have been collaborating with other artists. Since 2017, I have been working with the artist Tanvir Taolad from Bangladesh. We no longer take photographs, but work exclusively with and from our archives, which we deconstruct and reassemble, mainly in analogue forms. In this way of working, Tanvir will send me five or ten variations of a motif. I choose the one I think best and continue working with it. Funnily enough, this corresponds exactly to the way one works with an AI image generator, which in this case usually suggests four versions of a given prompt. So, for me, the workflow with AI is not new. I am simply building on the process I developed with Tanvir. And now we are exploring how to include the AI and develop a workflow between the three of us.
In broad terms, how does one go about generating an image with an AI?
You can create images with a text description or ‘prompt’ (text-to-image), vary existing images (image-to-image) or blend images together. Multimodal prompts combine all three approaches. There are then many ways to further refine the image. By erasing individual areas and generating new ones (inpainting), extending the result beyond the image border (outpainting), or defining areas of the image that should be formally identical in the original and final image (ControlNet). In addition, there are various pre-sets and plug-ins with which to control the image aesthetics.
You recently won the creative section of the 2023 Sony World Photography Awards [SWPA] open competition with an AI image called ‘The Electrician’. Why did you enter an AI image into a photographic competition and were you surprised to win a prize?
AI image generators enable people without photographic training to produce photo-like images that they could never have made otherwise. Inevitably, competitions are going to be flooded with AI-generated images. In my view, these competitions should have already changed their rules before October. But they didn’t. That’s why I submitted my picture to three different competitions, to hack the system and see how far I could get. All three times I was among the finalists, and now I won…
Ironically, I have applied several times in the past with ‘real’ photographs and never once got shortlisted. I am surprised that it was so easy when I submitted an AI generated image. Even more so that I suspect it was partly because the central figure is a beautiful woman.
Were the organisers aware that this was an image created using AI?
When I applied, I submitted three images from the ‘Pseudomnesia’ series, the other two being ‘The Connoisseur’ and ‘The Deal’. Those other two images were obviously AI, with all the flaws and mistakes that AI images had back in August 2022.
I checked if the regulations excluded AI generated images and, if not, applied without giving further information. In January, the award organisers asked for links to my website, my Instagram account, and for the exhibition history of the picture. If they had looked at the information I sent them, it would have been obvious that the image was AI generated. What information the jury was aware of, I don’t know. Why they chose it has not so far been communicated.
Sony informed me of the award in advance of making it public. At that time, I made the situation clear to them. I offered to disqualify myself, freeing them to give the win to a ‘real’ photographer. Or, as an alternative, to use this as an opportunity for an open discussion on the complex relationship between photography and AI, which could take the form of an online discussion or an on-stage panel.
How did they respond?
Four days later, I was told I could keep the award, but my suggestion to use this to launch an open discussion was not addressed.
When the press release went out on 14 March, there were many enquiries about the nature of my picture. The PR department of Creo, which works for SWPA, appeared unaware of my previous emails and asked me for more information. Within forty minutes I had emailed them a statement and repeated my suggestion that this situation be used as an opportunity for a debate. Later, a journalist friend who had contacted Creo for information told me that they did not send him my statement, but rather this generic reply:
“Many thanks for your inquiry.
Please find below a quote from Scott Gray, Founder and CEO of the World Photography Organisation:
‘As a medium photography has always been at the forefront: constantly adapting and evolving, it has a singular ability to transform itself and push boundaries. We are interested in photography as an art form, and within the Sony World Photography Awards we have our Creative categories in the Professional and Open Competitions which welcome photographers to experiment and explore the dynamism of the medium. With technological advancements, a wider audience of creators are engaging with lens-based work and we look forward to seeing how this can expand the reach and impact of photography.’
Later that evening, Creo responded to my statement and suggestion with a short email saying thanks and a smiley. Since then, they have neither used my statement nor taken up my suggestion. In my view it would have been better if the organisers had taken a stand and made explicit whether they included or excluded AI images. But they appeared to take an approach of ‘don’t mention the war’. After I made my case for a third time, they offered me an interview for their blog. It’s a start… but so far nothing has yet happened…
What do you think should happen?
It is important to realise that the ‘photo-like’ visual language of AI-generated images has separated itself from the medium of ‘photography’ and is now free-floating as something in its own right. It is crucial to clearly separate the ‘photo-like’ and ‘photography’ as terms that describe distinctly different visual forms, so that we have a vocabulary with which to define more precisely the criteria of future photo competitions, festivals, museums, exhibitions, and so on…
How easy is it to generate an AI image?
With Midjourney 5 it is very simple. You don’t need any technical skill in order to get a good result. Just type ‘Trump getting arrested’. The AI will do the rest, taking the lead and delivering a convincingly photorealistic image. It is a more complex undertaking to create an aesthetic formular that is unique. I consider that there are essentially eleven elements from which a prompt can be built. Six of those relate to the technical and art-historical knowledge of photography. Beyond those variables, nobody knows how an AI will respond when you start to play around with words. This is what I aim for, to be experimental. For example, while it is easy to get a result with the prompt ‘a pizza painted by Van Gogh’, what will be the result if I ask for ‘a pizza photographed by Van Gogh’? or ‘Van Gogh photographed by a pizza’?
As an example of how you work with AI, could you talk me through the making of your winning image, ‘The Electrician’.
There were twenty stages in all in the creation of that image. Here are some of the key ones…
The development of AI image-making capabilities has been startlingly rapid over the past twelve months, as AI generated images become ever more persuasively realistic. Where do you see this headed?
The photography scene needs to reposition itself and face the fact that it is no longer photography itself but AI that is defining the future of the medium. The innovations come thick and fast. Since January, I have been working full-time on knowledge transfer and consultancy with respect to AI image generators. It is becoming increasingly difficult to maintain an overview. As a single individual, it has become impossible to try out every new innovation. No one can say where we will be at the end of this year. But I expect ‘text-to-anything’ to become reality – with the ‘anything’ encompassing video, 3D, music… whatever.
In terms of image-making, AI is raising some complex questions around authorship and copyright. How do you see these issues ethically and practically?
An AI learns by sampling a vast number of images online. That scraping of data is still largely covered by existing legislation. Whether claims for damages can be upheld will be decided by the ongoing lawsuits in the USA and UK [see, for example, here]. But soon this discussion will be a thing of the past. Adobe Firefly and Stable Diffusion 3.0 have already set themselves up so that they will be much more difficult to challenge legally. OPEN AI does not reveal where they source the data from which the AI learns and thus remains effectively unassailable. What no one discusses is that the responsibility in matters of copyright has already passed from the providers of the platforms to the users. They have innumerable ways in which to work with other people’s image material, without it ever being possible to control or prove it. Ethically, this is all highly problematic; practically, no one can prevent it. With the open-source model of Stable Diffusion, Pandora’s Box was opened and can no longer be closed.
With machine learning tending to an exponential growth, what do you see as the potential benefits, challenges, and dangers over the next few years?
Like every other tool invented by human beings, it will be used for good and bad. And, as with every technological revolution, old professions will disappear and new ones will emerge. I know the hardships of my commercial photographer colleagues and understand their existential fears. But I am far more concerned about the demise of photojournalism in an avalanche of fake documentary images than I am about the collapse of advertising photography.
Images created as satirical comment from Eliot Higgins’ Twitter post, which nonetheless demonstrate the potential plausibility of fake AI-generated imagery.
We need photojournalism as a pillar of democracy, but we are in danger of being overwhelmed in a tsunami of ‘alternative facts’. Democratic societies will find it harder and harder to agree on the factual foundations on which to take up positions that can then be deliberated and compromises found. As a citizen, I care about this deeply – and unfortunately, I cannot offer a solution. A seal of approval, modelled on the World Press Photo Award, could strengthen photojournalism, if all the news media were to commit to it and establish a common set of rules. However, it cannot stop the flood of fake pictures. And this flood will far exceed the number of ‘real’ pictures. It is the wet dream of autocrats and conspiracy theorists come true.
What impact do you see AI imagery having on photography as an art practice?
One hundred and seventy years ago, Baudelaire pronounced the nascent medium of photography to be the “mortal enemy of painting, (…) refuge of all failed painters, the untalented and the lazy”. Today, AI artists are berated in much the same way by photographers. Back then, photographers took the likeness, the realistic representation, away from painters. It proved a liberating blow, the prerequisite for the development of modern painting.
I assume that, like me, many artists will welcome the development as a liberation from the material and integrate it as a new tool. Those who know how to deal with AI will receive a co-pilot capable of taking them to a higher level. To borrow a term from Professor Peter Kabel, I see AI as a form of ‘augmented intelligence’. [Peter Kabel is a professor at Hamburg University of Applied Sciences. He set up the Cogniwerk platform, which provides an overview of AI generative tools for creatives.]
What have you learned about yourself over the past year making images? What has AI taught you?
AI was the right tool at the right time, and the only one I could use last year. My partner was unexpectedly diagnosed with cancer in late summer and image generation via smartphone was the only way to be creative and recharge my batteries in the autumn, when we spent a lot of time in waiting rooms.
I learned that very difficult times and very good times can happen in parallel and that some things just have to be endured.
UPDATE: 13 April 2023
The Sony World Photography Awards night was held at the London Hilton, Park Lane. While Boris Eldagsen was not invited to make a speech, he took to the stage anyway in order to publicly declined to accept the prize for the creative open section. “Thank you for selecting my image and making this a historic moment,” he began, “as it is the first AI-generated image to win in a prestigious international photography competition.” Going on to say, “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.”
He used the opportunity to call for “an open discussion. A discussion about what we want to consider photography and what not.” And concluded by suggesting that the prize money might be donated to the photography festival in Odessa, Ukraine.
The full text of his non-acceptance speech, not all of which it was possible to deliver on the night, is published here.
UPDATE: 20 April 2023
A lot can happen in a week. Following Boris Eldagsen’s public non-acceptance of the award, his image was replaced in the exhibition and removed from the competition website, but no statement was issued by the organisers. Coverage on social media escalated until, on 18 April, The Guardian and the BBC covered the story followed rapidly by other media outlets across Europe and North American and as far away as Zimbabwe.
In a statement to the mainstream media an unnamed spokesperson for the World Photography Organisation made the following key points:
That SWPA “welcomes various experimental approaches to image making from cyanotypes and rayographs to cutting-edge digital practices.”
And that the artist had expressed his interest in “the creative possibilities of AI generators” and had emphasised that “the image heavily relies on his wealth of photographic knowledge”.
The spokesperson did not go so far as to say whether the organisers believed that an interest in AI and a knowledge of photography were enough to classify an image as a photograph. However, they did suggest that his actions – rejecting the award and making a statement explaining why he had set out to test the system – constituted “deliberate attempts at misleading us” and that “we no longer feel we are able to engage in a meaningful and constructive dialogue with him.”
The spokesperson went on: “We recognise the importance of this subject and its impact on image-making today. We look forward to further exploring this topic via our various channels and programmes and welcome the conversation around it.” But not, it was made clear, with Boris Eldagsen. [The full SWPA statement is included in the Channel 9 News report.]
One wonders what the corporate Goliath fears the artist has in his intellectual slingshot…
Meanwhile, Boris Eldagsen issued a forensically detailed rebuttal of the points made in the organiser’s statement.
Whatever the outcome, it is clear that (to slip figuratively from the biblical to the pantomime) the genie is well and truly out of the bottle on this one.
Online response to the story has divided along familiar lines. For those with a more traditional understanding of photography, it seems clear that AI-generated imagery never can be included within the frame of the medium. For the laissez-faire and technologically adventurous, AI is seen simply as an extension of what are already pretty porous notions of ‘the photographic’. For those with a broad systems view, the arguments about what is and what is not photography pale in contrast to the larger question of just how much, already, our experiences across the board are shaped by AI technologies, often without us being aware of it.
In terms of the organisers, Sony entered the camera market with the advent of digital, which was at the time itself a contested ‘photographic’ space. They currently sponsor a photographic award because that brings them to the attention of a large potential market for their digital cameras. It may well be that the tech giant will find new and even more lucrative opportunities as AI becomes more widespread. So, one could perhaps recognise why they might be reluctant to make any definitive statements about a field that is changing rapidly by the day. On the other hand, in a crowded tech market, it is often the mavericks unafraid to take on controversial subjects that achieve the competitive edge.
* Images captioned with an asterisk are courtesy of Photo Edition Berlin
Boris Eldagsen was born in Pirmasens, West Germany, in 1970. He studied media and drama, philosophy, and German studies at the University of Cologne (1991–92). From 1992 to 1998 he concurrently studied philosophy at the Johannes Gutenberg University Mainz, and visual art at Mainz Academy of Fine Arts. During that time, he also studied visual art at the Sarojini Naidu School of Arts and Communication in the University of Hyderabad, India (1994); and conceptual art at the Academy of Fine Arts in Prague, Czech Republic (1995). He has presented nine solo exhibitions in Australia, France, Germany, and the United Kingdom including those at the Rencontres d’Arles, Edinburgh Art Festival, and the Australian Centre for Photography, Sydney. His work has featured in twenty-one group exhibitions and festivals in Asia, Europe, and South America including Chobi Mela in Bangladesh; Encontros da Imagem in Braga, Portugal; FestFoto in Porto Alegre, Brazil; FORMAT Festival in Derby, United Kingdom; and PhotoVisa in Krasnodar, Russia. He has won a number of awards including the Voies-Off Award, Arles (2014); best show at FORMAT festival, Derby (2015); and the Open Competition Creative prize at the Sony World Photography Awards (2023), which he declined to accept. Boris Eldagsen is a member of Deutsche Fotografische Akademie, responsible for their online activities, and, since 2019, a digital consultant to Roger Ballen. He lives and works in Berlin.
This interview is a Talking Pictures original.