Photography and Artificial Intelligence
Dr Michael Pritchard, Conference Chair
Bristol, 9 - 10 October 2023
Hosted by the Royal Photographic Society
October 9-10, 2023. RPS House, Bristol, United Kingdom and Livestreamed
Conference website: https://events.rps.org/en/4LrdQ66/conference-photography-and-artificial-....
The Photography and Artificial Intelligence conference, hosted by the Royal Photographic Society was a hybrid event with both remote and online participants. It set out to explore how the availability of generative AI has prompted concerns about AI’s impact on photography, as well as an interest in new ways of creating images in equal measure, notwithstanding ethical and legal issues. The conference heard from a series of 18 international speakers from the Association of Photographers, the National Union of Journalists, and the Royal Photographic Society, along with leading academics studying AI and its impact, its meaning, and its use; and international lawyers dealing with AI and intellectual property. There were also industry figures, and creatives and photographers using AI and the opportunities afforded by it.
The conference was arranged over two days in four panels: Introductory; History, Theory Application; Legal, Rights Practice; and Creative Practice. Each panel began with a keynote presentation that discussed the impact and opportunities of AI for photography, as well technical, legal, and philosophical issues. The panel discussions were then followed by a lively Q+A with the speakers taking questions from the audience and the Zoom chat.
An important keynote by Katrina Sluis, Associate Professor and Head of Photography & Media Arts at the Australian National University (ANU), provided a lucid overview of the recent advances in machine learning made possible by the labour of millions of photographers and viewing subjects. Sluis eloquently outlined the growing apprehensions around new AI text-to-image models which threaten to eliminate the photographer from photographic production raising vital ethical questions on authenticity, datasets biases and authorship. She asked what challenges this situation posed to public organisations committed to photography and called for a critical ethics of AI and an understanding of the photographic pipeline to scrutinise some of these hegemonic systems. Subhankar Poddar presented an argument for the ontological difference between a photograph and an AI-developed image, while panel speakers Sluis, Palmer and Farnun agreed that AI extended the logic of digital photography, since automated functions such as white balance and smile detection have been embedded into digital SLR cameras for decades.
The second keynote, by Matt Hervey, Head of Artificial Intelligence Law at Gowling WLG, provided clarity on the copyright of AI generated images. He pointed out that human authorship is the bedrock of copyright and, so, establishing who or what makes the decisions involved in the creative process is key to attributing legal ownership. AI raised the fascinating question that in some scenarios the software creator and the algorithm trainer may arguably have had more input than the individual who used the software to create an artistic work. Hervey however, clarified that a human may be the "author" for copyright purposes provided the work reflects their own intellectual creation and expression of ideas.
The problem pointed out by practitioners, Nicolas Malevé and Tim Flach, was that the prompt economy threatens to capture whatever value is in an artist’s signature style and to reprocess it into a generic look without any renumeration or benefit to the artist. In a further challenge to intellectual property Malevé reminded us that for AI “to produce ‘photographs’, algorithms must use large collections of images scraped from the internet without permission from their authors”.
In the follow up it was evident from the discussion between photographic practitioners that there are mixed positions about the intersection of photography and AI and how AI is impacting the creative process. There was also much debate about the notion of what constitutes a photograph, what it means to be a photographer, and what the future of photography looks like.
Some practitioners celebrated the affordances of AI-driven image processing for enhancing photo quality or for generating artwork that stage fantasies, or as in Andrew Robinson’s work, finding alternative explanations for the repositioning of the cannonballs in Roger Fenton’s famous photograph, Valley of the Shadow of Death. For some the ability for AI and photography to synergise opens up exciting possibilities for the future of visual storytelling and advertising. Other photographers discussed how AI can be integrated into traditional photography techniques, and the use of AI-driven photo editing tools, such as Adobe’s Generative Fill tool, demonstrate a less sceptical view of how this technology can be used to augment human creativity rather than replace it. On the other hand, practitioners involved in “truth seeking” photographic practices such as journalism, crime investigation, and medical imaging expressed deep concerns about audiences being influenced by something that appears to be actual, but in reality, is not. At stake here is the bias in image data sets can create a significant distorted truth, for example in the understanding of a presentation of a disease. The effect of the extensive use of AI photography could be that audiences may distrust photographic evidence, believing that all images are not records of reality but are artificially created.
Finally, Katrina Sluis left a sobering message in the Zoom chat to those who refer to AI generators as “just another tool”. “Gen AI” she wrote “is not just a tool, it is a political economy and part of a capitalist reconfiguration of the relations between seeing and knowing, built on the enclosure and extraction of the commons, with huge environmental costs, and major labour implications”.
This was an important conference that opened up important fissures as AI photography becomes naturalised and assimilated into the digital landscape. The hybrid format was also crucial since the in-person and virtual attendance options made it accessible to global participants and allowed some speakers suffering another round of Covid infection to participate. The virtual platform was well-organised, with smooth live streaming and interactive Q&A sessions, which enabled remote participants to thicken the discussion with some helpful references and links to websites live in the chats.