The past few weeks have seen a flurry of media coverage of artificial intelligence (AI) image generation becoming increasingly compelling. Should the industry worry?
Hamish Crooks, president of the Society of Photographers, shares his views with Rachel Erskine, co-chair of Bond’s Film People Working Group.
For those not working in the photography industry, the growth of AI image generation may seem inevitable and irreversible. how do you see things?
To be brutally honest, it’s exciting! Seeing what is possible in other fields such as health, medicine, and science is, of course, very exciting from an inventive and creative point of view. But what I’ve learned is that most AI systems do only one thing, or at least do one thing well. No one has yet created a multifunctional AI that is interconnected across different disciplines. No keywords yet. Much of what is out there is very early. But AI will improve and find its niche. We’re never going to put the genie back in the bottle. I think there is reason to be alarmed when you see the way social media companies have reached out to governments around the world and gone largely unregulated for 20 years. Regulation will be needed. When companies themselves ask for regulation, it certainly worries me. There are certain photographic styles that NGOs like and the AI learns from them. It doesn’t exist yet, but it will.
Join Bond’s “People in the Pictures” working group.
This community is an advisory group on ethical approaches to image collection and use. This group shares best practices and knowledge to advance the NGO image debate. It also provides a space for peer-to-peer discussion and support on best practice approaches.
learn more
What is the potential impact of AI on how we evaluate photos and the people who make them?
Computer-generated images are nothing new. Before we had AI, we had CGI, but with CGI you build an image from scratch. It’s very expensive because you’re starting from scratch. AI image generation uses existing ones. In other words, scrutinize datasets that are in the public domain. This makes the price significantly lower. And here some arguments arise. Any publicly available platform can be scraped, with obvious copyright implications. Photographers had no opportunity to opt out and were never paid for the images taken.
I don’t think all photos suddenly disappear. News photos must be real, and there are very strict rules about that, so they can never be replaced. But he’s already seen his two major photo agencies, Getty and Shutterstock, work with AI image generators to cut costs. Those expenses may be the people who have created the photos so far. The struggle to get photographers paid fairly is ongoing and will become increasingly difficult. I’m not saying you shouldn’t use AI, I’m just saying you should think about it.
AI images do not show “real” people, which is believed to offer an opportunity to sidestep some of the ethical issues that charitable fundraising campaigns can pose. From this perspective, AI protects us from problems of fair or accurate representation. What do you think?
These images are not real, but you have to remember that they are based on real images. As a result, the final product will have the same bias as the dataset used to create it. If the AI is scraping poverty porn, the result will be poverty porn, so the problem is not solved. The first rule of photography still applies here. In other words, photographs can only be read according to one’s cultural inclinations.
What steps can NGO communicators take to protect existing photos and the people in them?
I think it’s more important than ever to have a sound metadata policy. It may seem tedious, but if an image contains no metadata, it has no ownership. No one knows it’s yours. Metadata may be stripped by AI, but if you want to know which dataset the image will end up in, the metadata must be present in the first place. You can also update your consent policy to clarify that metadata will be removed by AI. Scraping is one of the possible secondary uses of images.
One thing AI isn’t very good at yet is context. Previously, when I used image recognition to search for photos of buses, it showed a bus that had been blown up by a bomb. At the moment, AI only reads images from vast publicly available datasets that are mostly collected without permission, so these datasets have the potential for AI to make better value judgments based on human culture. It doesn’t have rich metadata such as keywords to make it possible. Image reading – but we are catching up. Getty and Shutterstock have these resources and will be fine-tuning the crude output of these AIs today with very good products for their image licensing skills. Also in this area, if the NGO has a certain aesthetic, the AI will start looking for it and provide it specifically upon request.
Do you think AI is an irreversible point in NGO communications?
Someone who attended one of our events wrote in chat that ‘real people with real stories are always important’ and I agree. You can’t use AI to write the human stories that INGO produces. AI will continue to grow and improve, but I think INGO will continue to tell stories and use real photographers. AI may just be a tool in your toolbox.
In the past I’ve seen NGO campaigns like the one Shelter produced in the 1990s. Since it would be unethical to show real people in such situations, real stories were used, accompanied by images of models posing. Will the NGO sector use AI in a similar way? I think that will depend on how successful those campaigns are.
Your industry has a very strong ethical foundation, and we’d like to know how you end up using that technology. How do you get past those ethical issues? Are you getting what you wanted out of it? These are the questions I would like to know the answers to!#
Above is an edited version of the conversation that took place between Hamish Crooks and People in the Pictures Co-Chairman Rachel Erskine on Thursday, March 16th. A full recording is available to members of the People in the Pictures working group. You can join here as a member of Bond. For more information on Bond membership, please contact Rachel Phillips. [email protected]
