AI Sparks Controversy and Debate at World Photography Awards

The art world is buzzing after German artist Boris Eldagsen made history last month as the first person to win a prestigious photography competition with an image generated entirely by artificial intelligence (AI). In a shocking twist, Eldagsen proceeded to turn down the Sony World Photography Award, transforming his win into a powerful statement questioning AI‘s role in the future of photography.

Eldagsen‘s striking black-and-white portrait of two women, crafted using AI tools like Midjourney or Stable Diffusion, took first prize in the Creative category. But shortly after the announcement, Eldagsen published an open letter on his website declaring that he could not accept the award.

"I‘m grateful for selecting my image, but calling AI images photography is a mistake," Eldagsen wrote. "I don‘t need this prize, I need a discussion. And the photo world needs to have this discussion."

The World Photography Organisation stated that when Eldagsen submitted his image, he made clear it was AI-generated and argued it still fit the Creative category‘s mission of celebrating experimental approaches and "cutting-edge digital practices." After Eldagsen‘s public refusal, however, the organization announced it had suspended activities with him, citing "attempts at misleading" that prevented constructive dialogue[^1].

[^1]: May, Ashley. "An artist declined a $7,000 award after his AI-generated artwork won a contest." USA Today, 19 April 2023, https://www.usatoday.com/story/life/2023/04/19/artist-rejects-award-ai-generated-image/11685537002/. Accessed 8 May 2023.

Eldagsen‘s principled stand has sparked passionate debates about what exactly qualifies as photography in an era when artificial intelligence is capable of generating uncanny photorealistic images from mere text descriptions. As AI image synthesis models like DALL-E, Midjourney and Stable Diffusion grow exponentially more sophisticated and accessible, artists and photographers are grappling with profound questions about authorship, creativity, skill, and reality itself.

The Rise of AI Image Generation

Over the past two years, a new wave of "generative" AI systems trained on massive datasets of images and their text descriptions have made it possible for anyone to create striking visuals from natural language prompts. The most renowned models, from OpenAI‘s DALL-E to Midjourney and Stability.AI‘s Stable Diffusion, are built on diffusion, an approach that learns to iteratively denoise and construct images from pure noise[^2].

[^2]: Dhariwal, Prafulla and Alexander Quinn. "Diffusion Models Beat GANs on Image Synthesis." ArXiv, May 2021, https://arxiv.org/abs/2105.05233. Accessed 8 May 2023.

These text-to-image AI models have demonstrated remarkable abilities to generate high-resolution, aesthetically compelling, and sometimes uncannily realistic images from open-ended prompts. Given descriptions like "a majestic lion, by Rembrandt, 4K, trending on Artstation," the latest systems can produce vivid, detailed pictures that mimic distinctive artistic styles, eras, mediums, subjects, and moods[^3].

[^3]: Ramesh, Aditya et al. "Hierarchical Text-Conditional Image Generation with CLIP Latents." ArXiv, April 2022, https://arxiv.org/abs/2204.06125. Accessed 8 May 2023.

AI-generated images from Midjourney v4. Prompt: "a majestic lion, by Rembrandt, 4K, trending on Artstation."

The user bases and outputs of these AI tools have grown at a breakneck pace. Since launching in beta last July, more than 3 million people have signed up for Midjourney, producing over 20 million images per day[^4]. DALL-E 2, released to 1.5 million users last year, now generates an estimated 2 million images daily[^5]. And the open source Stable Diffusion, downloaded over 200,000 times since its weights were released in August 2022, has spawned countless derivative models and been integrated into popular apps and services[^6].

[^4]: Kantrowitz, Alex. "The inside story of Midjourney." Alex Kantrowitz, 4 April 2023, https://www.kantrowitz.com/p/the-inside-story-of-midjourney. Accessed 8 May 2023.

[^5]: Vincent, James. "OpenAI says ChatGPT has over 100 million users." The Verge, 2 February 2023, https://www.theverge.com/2023/2/2/23582974/chatgpt-100-million-users-openai-sam-altman. Accessed 8 May 2023.

[^6]: Romero, Andreas. "Stable Diffusion reaches 200,000 downloads." Medium, 28 September 2022, https://andreasromero.medium.com/stable-diffusion-reaches-200-000-downloads-5cec36a6dd4e. Accessed 8 May 2023.

This flood of AI-generated images has provoked awe, outrage, and existential anxiety across creative industries. Some see generative AI as a powerful tool to augment artists‘ capabilities and unlock new forms of expression. Others perceive it as an existential threat, capable of automating creative labor, laundering the styles of human artists, and deceiving viewers.

As the technology races ahead of governance frameworks, urgent questions are emerging around the ethical and legal implications of AI art. How will we attribute images created through complex collaborations of humans and machines? What should the standards be for disclosing the use of AI? How can we mitigate biases and harms embedded in AI systems trained on web-scraped imagery?

Photographers Grapple with AI‘s Possibilities and Pitfalls

For photographers specifically, generative AI presents both incredible artistic possibilities and risks of devaluing or diminishing their craft. As Eldagsen‘s contest refusal illustrates, the medium faces a quandary in whether pictures produced purely from code and data, without a lens or sensor capturing light, even qualify as "photography."

Many argue that while AI can imitate the appearances and aesthetics of photographs, it represents a fundamentally different approach to image-making, more akin to digital art or photo manipulation. Since photography‘s inception, theorists have emphasized its unique relationship to the real, its mechanical objectivity, and the notion that photographs are physical traces of light interacting with film or sensors[^7].

[^7]: Batchen, Geoffrey. "Photogenics." In Each Wild Idea: Writing, Photography, History, MIT Press, 2002, pp. 145-162.

When photography is reduced to a statistical process of finding patterns in pixels and captions, disconnected from any referent in the world, some worry it loses the truth value and emotional resonance that makes it meaningful. As photographer and podcaster Ben Smith puts it: "The ‘magic‘ of photography is that no matter how abstractly you present it, it still has one toe dipped in the pool of reality. But AI images, from start to finish, have no connection whatsoever to actual reality."[^8] [^8]: Smith, Ben. "What does AI mean for photography and photographers? (Spoiler: It‘s not good.)." A Small Voice Podcast, 23 March 2023, https://asmallvoicepodcast.com/conversation/ai-and-photography. Accessed 8 May 2023.

Others argue that this perceived dichotomy between camera-captured and computer-synthesized photography is misleading. Photography has always relied heavily on constructed realities and manipulated images, from staged daguerreotypes to spirit photography to the ubiquity of Photoshop. As artist and educator Kate Bingaman-Burt notes, "Every photograph is an edit, every photograph has some element of fiction."[^9] [^9]: Bingaman-Burt, Kate. "The Camera and The Cursor." Debris Journal, 24 March 2023, https://www.debrismagazine.com/features/camera-cursor. Accessed 8 May 2023.

The malleability of photos has only grown more apparent in an age of Instagram filters, deepfakes, and AI upscaling tools. Many contend that rather than a threat to photography‘s essence, AI image synthesis simply continues the medium‘s trajectory from mechanical reproduction to digital simulation[^10].

[^10]: Zylinska, Joanna. Nonhuman Photography. MIT Press, 2017.

Some photographers are embracing AI tools to push the boundaries of visual expression. By combining their technical skills and creative vision with AI‘s ability to generate and manipulate pixels programmatically, they‘re forging new hybrid forms that merge camerawork and code.

Artist Helena Sarin has gained renown for her experiments in "neural photography," using AI models trained on her own photos to generate uncanny variations and textures[^11]. Documentary photographer Mitch Epstein edits and reworks images from Midjourney to evoke climate anxiety[^12]. And an emerging movement of "promptographers" are investigating the poetics and politics of steering AI image synthesis through language[^13].

[^11]: Lufkin, Felicia. "AI/Ƀ/rr/t-ing – Uneasy Relationships in the Age of Machine Learning." The Photographers‘ Gallery, March 2023, https://thephotographersgallery.org.uk/whats-on/aiɓ/rr/t-ing-uneasy-relationships-age-machine-learning. Accessed 8 May 2023.

[^12]: Epstein, Mitch. "The Unmaking of Nature." Mitch Epstein Studio, 2023, https://www.mitchepstein.net/work/the-unmaking-of-nature. Accessed 8 May 2023.

[^13]: Karasick, Adeena. "Poetic Photography: Promptism and Human-AI Co-Creation." ArXiv, March 2023, https://arxiv.org/abs/2303.09958. Accessed 8 May 2023.

While AI remains an imperfect tool, prone to glitches, distortions, and alarming biases, it‘s clear that the technology will profoundly impact photography in the years ahead. Some predict a near future in which most commercial images, from ads to stock photos, will be generated by AI. Many portrait, product, and landscape photographers worry about AI rendering their skills obsolete.

At the same time, AI could also democratize image-making for those without access to expensive equipment or formal training. It could make it possible to generate photos of impossible scenes and camera angles. And it could free photographers to be more experimental and conceptual, focusing on prompts and art direction over technical execution.

Charting an Ethical, Thoughtful Path Forward

As photography grapples with this strange and exciting new chapter, Eldagsen‘s stand offers an important template for engaging with AI critically and transparently. Like any powerful technology, generative AI is not inherently good or bad for art – what matters is that we deploy it thoughtfully and ethically, with careful consideration for its implications.

The photography world needs to establish new norms, standards, and vocabularies for assessing the merits and meanings of AI-generated images. Competitions and professional organizations must develop clear rules and categories for submissions that disclose the use of AI tools. And as Eldagsen urges, photographers should proactively steer the conversation around these systems to foreground their social impacts and limitations.

While today‘s text-to-image AI models are capable of producing captivating visuals, it‘s crucial to remember that they are ultimately probabilistic systems constrained by their training data and prompts. Truly groundbreaking, provocative art will always demand human imagination, cultural fluency, and a deep understanding of why pictures matter beyond pixels[^14].

[^14]: Lyons, Kim. "The first AI-generated image to win a photography competition disqualifies itself." The Verge, 18 April 2023, https://www.theverge.com/2023/4/18/23686772/ai-generated-image-photography-competition-midjourney-disqualified. Accessed 8 May 2023.

Photography is not ending with the rise of AI, but it is undoubtedly evolving. As curator and critic Charlotte Cotton argues, "We‘re in a moment now where photography has become incredibly elastic. It can take many forms, many materials, many modes of production."[^15] Embracing that elasticity, while still preserving the medium‘s core values and ethics, will be key to ushering in an exciting new future for the photographic arts.

[^15]: Cotton, Charlotte. "AI Aesthetics." Fotomuseum Winterthur, 11 February 2021, https://www.fotomuseum.ch/en/explore/literacy/13/39216_ai_aesthetics. Accessed 8 May 2023.

Eldagsen‘s courageous provocation has ignited a vital dialogue on the interplay of artificial and human intelligence in photography. His story is sure to be remembered as a pivotal turning point, when the art world could no longer ignore AI‘s disruptive potential. Now, it‘s up to photographers, curators, and the public to decide what we want that potential to become.