A Societal Debate on the Impact of AI is Essential
Photography and AI-generated Images
Artificial intelligence (AI) is currently developing at a rapid pace when it comes to the generation of images (text-to-image technology) due to the possibility of creating images with the help of keywords (prompts). The AI learns to create images with the assistance of links to billions of photographs found on the internet and their metadata. In this way, internet users can use prompts to create images that, although fictitious, can also show a kind of seemingly authentic reality.
This form of generative image creation has completely different implications to the AI we use in cameras and software to take better photos or to edit them. It poses blatant risks to the credibility of the media and thus to our democratic social order.
Photographs are created exclusively through camera-based exposure. AI-generated images are therefore not photographs.
Authenticity vs. Artificial Intelligence
Generated images do not show reality. Authentic photographs are vital for the credibility of images in journalistic media. Documentary photography provides proof of the existence of people and situations. This is also relevant in terms of the image of institutions and companies: if they and their representatives and products are not represented by authentic photographs, but instead by an idealisation and an invention by AI, their credibility is called into question.
Fraudsters and manipulators will take advantage of text-to-image AI and will deliberately distort facts and manipulate viewers with photo-realistic AI images. This has already happened in the past, but in the future even people with no prior knowledge will be able to create deepfakes (realistic-looking media content that has been falsified by AI) that are no different in quality from authentic photos. A flood of these falsifications of reality could have implications that endanger our democracy.
Yet again, the tech companies who provide this technology deny any responsibility.
We as photographers and as a society will have to come to terms with the fact that an image could be an authentic photograph or equally an AI-generated fiction.
Taking Responsibility – Ensuring Authenticity
Creators of all forms of work accept their responsibility. Photographers take responsibility for what they photograph, how they photograph it, how they edit it, and who they share their images with.
The challenge for photojournalists and the media in the future will be to give recipients the means to distinguish between authentic photographs and entirely photo-realistic AI-generated images.
FREELENS is a member of the Content Authenticity Initiative (CAI). We expressly welcome the provision of internationally uniform and consistently open technical standards for everyone involved (verify.contentauthenticity.org). IPTC standards state that photographers should label their images according to their digital source (https://cv.iptc.org/newscodes).
We call on all media to label image material with: authentic photo [A], manipulated photo [M] or generated image [G] – directly on the image and as clearly as the copyright notice.
We call for a ban on the deletion of metadata so that the origin and the creatorship remain traceable.
We remind political decision-makers that democracy is dependent on reliably true news. Those who want to protect the rule of law from deepfakes must promote the publication of authentic and verifiable information.
Ethics & Morality vs. Artificial Intelligence
There are other ethical issues related to photography-based and AI-generated images beyond the matter of authenticity.
AI is certainly capable of providing moral or even political guide rails. These are necessary and may prevent negative excess growth. However, since the AI firms are based in the USA and China, these guide rails will be designed according to their beliefs. Our consensus on what constitutes a free, democratic society differs greatly from the political agenda of the Chinese government. On points of morality, e.g. nudity, the same goes for the USA.
Experts also complain that AI-generated works threaten diversity, equality and inclusion because AI conveys and manifests human prejudices in line with the training images. AI applications do not take responsibility. Tech corporations only accept responsibility when it benefits the business model. Users can rarely keep track of the consequences of their actions.
We call for the establishment of a global supervisory authority which is capable of acting, and a globally valid agreement on ethical standards in AI applications.
Creativity vs. Artificial Intelligence
AI image generators can create any image requested by users with the right prompts. This will undoubtedly have an impact on the business models of photographers.
Nevertheless, we believe that creativity cannot simply be replaced by AI, because what AI cannot do is deviate from the input requirements. But this is what real creatives do: they push back against the rules, break with tradition, take risks, think unconventionally, overcome resistance. They are curious and sometimes visionary.
That is what we can do, what we enjoy doing and what our clients pay us for. Creativity is a chaotic process that is not based on a calculation.
AI may be able to combine old things in a new way (e.g. the style of van Gogh with an image by Helmut Newton) and the result may be surprising, but it can only ever create within the boundaries of what already exists – its originality is limited. It can evolve or combine things in a different way, but it cannot create something entirely new from scratch.
Preserve Cultural Diversity, Protect Copyrights
Photography and photojournalism not only depict our culture – they are a central part of it. They shape our visual memory.
We can all remember photographs that made a strong impression on us, whether we saw them at an exhibition or in a newspaper, whether they showed people or perhaps a landscape. A culture without authentic photography is scarcely conceivable. This is why it must be protected.
We are calling for the rights of creators whose images are used as training material by AI generators (data mining) to be respected. Under the pretext of scientific research, links and metadata for currently 5.85 billion images are being made available to the big tech corporations via non-profit associations. This training data will create a huge amount of value, which will remain with these few tech corporations in the USA and China. The AI business is not some small and friendly start-up. It will be IT giants like Microsoft, Google and Amazon – which dominate the infrastructure with their vast computing capacities – who will benefit from the development economically.
The original approach of keeping this technology »open source« and thus transparent has since been abandoned. Open AI is no longer »open« since the release of GPT-4. So the mask comes off: it’s not about research, it’s about vast sums of money. It’s about dominating the market and creating dependencies.
The creatives of the world who are essential in feeding this system are left out. No permission is sought from them, and they are not paid, as the Copyright Act (Urhebergesetz) states they should be.
A parasitic system has emerged in a very short space of time. Tech corporations first gobble up creativity without payment, to then replace the creatives with AI. The parasite kills the host, even though it will itself perish without fresh blood in the form of new creative works. A dystopia of creative stagnancy and eternal repetition looms.
The first step in counteracting this cycle would be to withdraw the non-profit status of the data mining associations. Works may not be used as training material without the consent and appropriate payment of their creators.
If copyright-protected works are identifiable in the AI-generated images (photomontage/editing) copyright and personality rights must be applied in full.
Legal regulations must be created that keep a check on AI applications, regulate the superior power of the big tech corporations and ensure transparency in data mining.
Photographers must clearly label images and websites to prevent them being used as training material for AI systems and destroying their own business model.
Sam Altman, CEO of OpenAI, told ABC News: »It is going to eliminate a lot of current jobs, that’s true. We can make much better ones.« We can rule out the second sentence for the creative industry.
We photographers are always open to new technologies. Large cameras have been replaced by compact cameras, analogue photography by digital photography.
But it is important to be aware of the consequences of the revolutionary innovations brought about by artificial intelligence. They will change our lives significantly. They will offer opportunities, but they also pose a danger to our society and our democracy.
A broad societal debate on the impact of AI is essential. We are ready for this.
The FREELENS position paper is available as a pdf here.