AI, Photography and the Battle for Reality

In 1984, when photographers were still using film, I began exploring the early use of computers to modify photographs quickly and undetectably. I wrote in an article in The New York Times Magazine that “in the not-too-distant future, realistic-looking images will probably have to be labeled, like words, as either fiction or nonfiction, because it may be impossible to tell them apart. We may have to rely on the image maker, and not the image, to tell us into which category certain pictures fall.” This, to our detriment, has come to pass.

Now, without even the use of a camera, artificial intelligence systems can simulate photographs in response to text prompts, depicting people and places that never existed. One can make realistic-looking images that seem to show the losing side of a war as if that side had won, of Winston Churchill taking a selfie before cellphone cameras were invented, or of Adam and Eve holding an apple in the Garden of Eden.

Some of these synthetic images can be thoughtful, even revelatory, helping to broaden perspectives and amplify possibilities. Others threaten to distort and undermine the visual media that we depend upon to understand contemporary events while causing the public to be suspicious of the actual photographs that they resemble. And these systems are trained, usually without permission, on the billions of photographs that are already online, an ethical and legal conundrum that has yet to be resolved.

The Synthetic Eye: Photography Transformed in the Age of AI is an exploration of this pivotal moment, asking how we can maximize the constructive potentials of generative AI while minimizing the distortions that destabilize our grasp of a collective reality. A nuanced understanding is both necessary and urgent in order to formulate an effective response to the masses of generated images that are surfacing, questioning the role of the artist, interfering with the functioning of democracies, and at times causing considerable personal anguish. The Synthetic Eye is meant to help kickstart this larger discussion.

“A photograph of the perfect family”

The imagery generated by artificial intelligence in response to text prompts is often correctly characterized as racist and misogynistic, reflecting the biases of many of the online images upon which the systems trained. But sometimes the results can be surprising, challenging stereotypes rather than reinforcing them. For example, the prompt that I provided, “A photograph of the greatest mothers in the world,” resulted in an image of an ape-like animal with her children, not a human being, soundly rebuking my anthropocentric expectations; “The most beautiful woman in the world” elicited, rather than a sleek model trying to impress while posing, a rather mundane image of a young Black woman, her head tilted, oblivious to anything other than the private conversation she was having on her cellphone; and “A photograph of the perfect family” ended up generating these two quite serious, formally dressed older men with two young children, radiating a decorous, routine air as if families headed by same-sex couples were always the norm.

“A photograph of a street scene in Harlem” in the style of a 20th-century jazz musician.

Much of the fun to be had with generative artificial intelligence is to ask it to provide imagery that simulates the photographic but is created in the style of non-photographers, such as in this case a jazz musician. How might another artist, scientist, or philosopher envision the world differently? Digital code can be output in any medium – the code for a photograph can be output as music, and the code for music can be output as a painting, and so on. These porous borders among media are a key facet of the digital revolution, allowing an open-ended transmedia rather than the more prosaic multimedia that is usually employed. Artificial intelligence allows one to cross such borders, among many other kinds of experimentation.

“The first photograph ever made.”

Sometimes the images that are generated in response to text prompts seem to allude to important photographs from the past. Roland Barthes, in his influential book, Camera Obscura, famously wrote of happening upon a photograph of Jerome, the younger brother of Napoleon, that was made in 1852. He “realized then, with an amazement I have not been able to lesson since: ‘I am looking at eyes that looked at the Emperor.’” Barthes would mention his amazement to other people but, he recounts, no one was able to share or even understand his response. Unlike his contemporaries, the artificial intelligence system with which I collaborated may have generated this image as a way of signaling the importance of Barthes’ insight.

The second image that was generated in response to this text prompt tells a very different story, apparently showing a woman taking a photograph with a small camera that was not invented until about a century after the first photographs were produced.

One can interpret this depiction as whimsical, as well as indicating a preference for a more self-reflexive, mobile photography that acknowledges the critical role of the photographer as well as that of the camera as a kind of cyborg appendage. In both cases, the artificial intelligence begins to transform our sense of the past as it will increasingly do in so many other instances.

Words by Fred Ritchin.

Discover The Synthetic Eye: Photography Transformed in the Age of AI by Fred Ritchin.