In the age of AI slop, what is the future of art and photography?

In a recent interview, artist and writer Victor Burgin was asked to reflect on the state of contemporary image production in relation to his seminal 1982 book Thinking Photography. The most fundamental change with the arrival of AI, Burgin said, is that photography is no longer just shaped by ideology, but has also become its vehicle. When we generate media using AI platforms, we are reproducing the worldviews embedded within its training data.  

The synthetic outputs of AI models are the result of aggregating billions of images scraped from all corners of the online sphere, converged as an average. Filmmaker and writer Hito Steyerl calls these mean images, understanding the AI-generated image as an approximation of society’s views of itself. We might understand mean to refer to the average and therefore the mediocrity of images generated using machine learning tools. But it can also refer to their intention. 

Get the latest news and insight into how the Big Issue magazine is made by signing up for the Inside Big Issue newsletter

As an artist working with photography, this notion feels ripe for interrogation. I have spent recent years exploring the complex layers of the AI-generated image – an investigation that began with acknowledging the infiltration of AI slop (as it is known) into popular media. 

But as AI scholar Kate Crawford highlighted recently, slop is not only waste. It is also fuel. And this inquiry developed to reveal the vast infrastructures on which these images rely – bringing the politics of surveillance, data extraction, and the exploitation of land, resources and labour to the surface of the image. 

From energy-hungry data centres consuming millions of gallons of water to radioactive waste dumped in the mining process, the material footprint of AI-generated images has huge planetary consequences. Through a four-chapter evolving installation work titled Variations, I have been attempting to unveil the machine that drives contemporary image production. 

The project began with V1: Content Aware – a shipping container plastered in collage imagery and fitted with a surveillance camera, sited in central Brighton. The intention was to gather surveillance data from an audience while drawing attention to the planetary scale of extraction. By using the surveillance images as training data for the chapter that followed, I could build my own generative system.

After a few days, I took the surveillance images (including lots of images of people engaging with the container with phones or cameras) and blended them in generative AI software to create an average or mean image.  

The outputs left me stunned. The four images generated depicted men holding hybrid machines – cameras as weapons. It felt as though these images were starting to reveal the violence of the contemporary camera from within its own system. Cultural theorist Susan Sontag explored the analogous connection between camera and gun (we shoot, we aim, we load).

Read more:

• Is AI pushing jobless young graduates to the far right?
• How do we stop AI becoming an ‘engine of inequality’, and make it a force for good?
• Teething problems with VAR and handball rules serve as a warning about AI

These images offered an amalgam of this metaphor, epitomising the violent force of photographic technology in the age of AI. From military contexts to algorithmic content, the image is increasingly weaponised. 

These images informed V2: Rigged, an installation that asked the audience to confront and interact with the aggressive processes that enable machine learning tools to operate. Part-camera, part-drill head, part-processing plant, the machine at the centre of the installation performed acts of taking (from the brutal processes used to extract rare earth minerals to aggressive surveillance systems and the scraping of personal data) that lie at the centre of computational image-making. The images gathered were used to explore the focus of the next iteration; the phenomenon of model collapse. 

Model collapse arises when a model is trained on AI-generated content. It results in a degenerative learning process caused by a feedback loop of data. V3: Model Collapse, which exhibited at The Photographers’ Gallery, London this summer, explored this hallucinatory process by feeding data from previous variations of the project into machine learning programmes. The resulting AI-generated images began to fall apart, full of impossibilities and inaccuracies, depicting ghostly figures and distorted perspectives. 

For this chapter, I re-enacted these images using my camera, constructing painted sets and printed props, inserting my own body into the image.  

AI generated content begins with images scraped from platforms we trust with our data. This is used to train systems that enact violence and disrupt political democracy whilst accelerating environmental collapse. We need to see ourselves mirrored in these images and understand the system to which we contribute. 

The final chapter of Variations is exhibited at Stills Centre for Photography in Edinburgh in November. The project was made possible by the Ampersand/Photoworks fellowship.

Do you have a story to tell or opinions to share about this? Get in touch and tell us more. 

Reader-funded since 1991 – Big Issue brings you trustworthy journalism that drives real change.

Every day, our journalists dig deeper, speaking up for those society overlooks.

Could you help us keep doing this vital work? Support our journalism from £5 a month.