Artificial Intelligence and photography
- ‘Experts Call for a Pause on ‘Dangerous AI Experiments’,
- ‘Photo Editing App Bazaart Adds New AI Background Feature’,
- ‘Romance Scammers are Using AI Images to Trick Victims’ and
- ‘AI Can Recreate Images From Human Brain Waves’.
These are all recent headlines from the past few weeks, highlighting opportunities and fears around artificial intelligence (AI).
As you might imagine AI has been a topic of conversation within the RPS prompting questions about what it is, what it means for distinctions and competitions, and more fundamentally is it photography? We’re still working out exactly what it means for the RPS and its members, and for photography more generally, and determining what’s acceptable in particular contexts and what is not. AI is evolving rapidly and there are no definitive answers but with the heightened interest around Chat GPT, Google’s Bard and a host of other apps it seems appropriate to take stock and look at the impact on photography and photographers.
What is AI?
Artificial Intelligence (AI) refers to the development of computer software that can perform tasks which would typically require human intelligence, such as recognising speech, interpreting images, making decisions, and learning from data. AI involves the creation of algorithms and computer programs that can analyse and interpret data, learn from it, and use that knowledge to make decisions or take actions. This can be done using a variety of techniques, including machine learning, natural language processing, and computer vision.
Some of these techniques have been around for some time, but what has come to the fore recently are significant improvements in software that will learn from large datasets and generate new data based on that learning, and the making of that software freely available to the public.
What does AI mean for photography?
Unless you are still using film and traditional darkroom printing the chances are you have probably been using AI without realising it. It lies behind in-camera image processing and increasingly in software such as Photoshop. New applications have taken this further, based on ‘learning’ from large datasets (collections of images) to generate new images that have never existed in the real world. It is this that is exercising the photography world.
What are the issues?
At a basic level using AI to automate aspects of image editing, such as colour correction, exposure adjustments, and object removal can save photographers time and improve an image. This is simply automating what many photographers would do themselves in Photoshop.
Generating new images based on text descriptions is where the bigger concerns are. Firstly, the software needs to ‘learn’ and most trawl large sets of images online with or without permission. There are unresolved issues around copyright and the law has yet to catch up with this ‘data mining’. Several court cases are pending to try and determine how existing laws might apply. Companies with large image sets have a commercial asset that could be valuable. There are wider concerns around privacy and personal data although these affect images less directly.
Secondly, the generating of images based on text to produce a picture of a person, place or situation that has never existing raises basic questions for organisers of photography competitions and for organisations such as the RPS which assess photography, and for others such as the media with questions of authenticity and what is real.
Impact
AI around imaging is already impacting photography. Winners of competitions have had awards rescinded after images were found to be created by AI, a well-known blogger revealed his portraits were created using AI and there are almost certainly other cases that we simply don’t know about.
For photographers, the advertising world is already using AI created ‘people’ instead of using real models, with Levi jeans the latest company to make use of them. Images can be dropped on to new backgrounds and into real or fictional ‘places’.
What does AI mean for RPS?
For the RPS, then Distinctions submissions and competition entries are the areas most likely to be impacted. Both have definitions that will need updating to reflect the availability of AI tools. What is acceptable? Should it be flagged? Can such images even be identified?
Raising awareness amongst members and the wider public of what AI can do and how it might be used is a first step.
We’ll be providing further updates over coming months.
Dr Michael Pritchard
RPS Director of Programmes and ChatGPT
April 2023