Artificial intelligence has overturned many of the old rules, and the one about ‘seeing is believing’ was perhaps the first to go. Ritchin, a respected photography critic and academic, has been thinking about the interactions between traditional image-making, computer-manipulated images and AI for a long time, and in The Synthetic Eye he takes a deep dive into the subject.

He acknowledges that the idea of a photograph as an immutable slice of reality has always been a little shaky. There have always been counterfeit photos, and fakery took a major step forward with the development of CGI, which arrived at about the same time as cellphones that took high-quality images. Ritchin estimates that two trillion photographs are taken every year. Instead of being a professional or occasional activity, photography has become universalised. The sheer number of images presented to us has undermined the value of any one.

But AI represents a fundamental change in the practice of photography, according to Ritchin. Smartphones with AI algorithm presets can amend images without much direct input from, or even awareness of, the photographer. Amending an image is no longer a somewhat shady and marginal activity: it has become the point of taking a photograph at all. Ritchin says: ‘In this new paradigm, photography becomes increasingly an expression of consumer entitlement. While it appears to have been photographed, the world around us instead has been largely reconfigured to please.’

Media outlets are already using AI techniques for image manipulation. So far they have largely avoided crossing the line into outright falsehood for malicious purposes, although they have sometimes come very close. The temptation to utilise manipulated images is hard to resist, especially given the overheated nature of our current politics. The day will eventually come when images and videos are ruthlessly manipulated by the media or by partisan activists without much concern for their veracity. The only metric that will count is how many clicks they attract.

This raises complex ethical questions. Should a manipulated image be used to make a political point? If so, who decides which causes to promote and how far the process should go? Does the end justify the deceptive means? How will the consumers of an influential yet artificial image or video respond when they learn they have been duped? If there are answers to these questions they are not yet visible.

Ritchin accepts that a reasonable response to this is cynicism: simply assuming that all images – or at least ones that you don’t like or don’t agree with – are in some way fake. In an effort to restore credibility, he has devised a system he calls Four Corners. This would embed accessible information in the digital image itself. At the very least, AI-generated or AI-manipulated images should come with a caption that explains their origin. It sounds like a good idea but would require additional work from the photographer as well as the consumer. Not easy, although perhaps it would be a good place to start. Of course, it would not cut much ice with those who want to use AI expressly to deceive.

A crucial step is that AI systems can now generate digital images from text instructions, and Ritchin recounts some of his experiments. The results can be odd, even alarming. The prompt ‘a photograph of the perfect family’ led to a picture of two older men in conservative suits standing behind two young girls. The image is provided in an old-fashioned style, which seems to be something of a favourite for AI systems.

A request for ‘a respectful photograph’ generated an image of a little boy, seen from the back, staring along an empty street. When Ritchin asked for ‘a photograph of Adam and Eve and the apple of knowledge’, the result was strange and ambiguous. Who is offering what to whom? And how did a severed head get into the mix? Only the AI knows.

One wonders what the AI systems think of all this. A hint was provided when Ritchin asked for a picture of ‘an unhappy algorithm’. The resulting image is of a ghostly, squashed-up face, as if someone had taken a baseball bat to a teddy bear. The synthetic face looks like it knows something that we don’t and does not like it.

This is not what Ritchin expected. It is probably fair to say that it is not what anyone would expect. Clearly, the learning process for AI systems, which involves them ‘seeing’ millions of images, is creating a ‘black box’ inside the machine. Glitch or feature? Maybe a bit of each.

Ritchin suggests that there is potential in joint projects between human photographic artists and AI systems. ‘We tend to overlook AI’s ability to surprise with new ideas,’ he says. ‘We can invite it to collaborate on outlining possibility, circumventing assumptions while helping to redefine aspects of our worldview.’

It is certainly an intriguing proposition. It would, of course, turn photographic art into something very different to what it has been. Can photographers make changes at this level and, just as importantly, do they want to? What will we, the readers of the image, make of human/AI collaborations?

But there may not be a choice. The AI genie cannot be forced or tricked back into the bottle, the Rubicon cannot be un-crossed. One way or another, photography as a practice and an art form cannot drift along as if nothing is happening. Even if the destination cannot be fully discerned, Ritchin believes that photography must move forward if it is to survive at all. The issue is how to do it with responsibility and a measure of wisdom.