Abstract

Rather than a recording of the visible that shows aspects of the world around us that may not be in accord with our preconceptions, software will allow us to reshape the digital photograph to conform to what we want the world to look like. For example, an exhibition on display in Ashkelon, Israel, shows synthetic, quasi photo-realistic images that were produced to represent the memories of elderly Holocaust survivors recounting traumatic memories from their childhood some eight decades before. It was done, as the organizers put it, to remember this horrific era in human history. But does this method also encourage forgetting? How will future generations be able to discern the difference between the lens-based visual record compiled by photographers and filmmakers versus the photo-realistic imagery that is based upon explicitly subjective memories? The past becomes increasingly uncertain, as does the present. What will become of the photographic record? What, if anything, can be done about it? 

Recently I was sitting alone in a public garden in New York City basking in an early spring sunshine when I received a photograph of two Ukrainian soldiers huddled for protection in a raw wintry landscape of snow and rocks just a few feet away from the camera. The photograph, which I received on my cellphone, had been taken only a few hours before. The sunlight around me disappeared.

I could imagine the bombs and the bullets, the fatigue and the fear, the drones circling above ready to strike and kill. And then yesterday the same photographer, a young Ukrainian who is a friend of mine, sent me a five-minute GoPro video made by a colleague running right behind him. One could hear the huff and puff of his breath as he ran forward, and his shouted voice as he yelled at my friend to drop flat onto the ground; if he had not shouted, my friend might have been wounded or killed. “You saved my life,” my friend said after a few moments, turning around inside a trench. He then let me know that thirty seconds after the video ended one of the two Russian drones that had been hovering above launched a grenade at him; it exploded a meter away, fortunately not in the trench where he was trying to conceal himself.

Much goes into making a photograph, into being there where events are occurring. “If your pictures aren’t good enough, you aren’t close enough,” the twentieth-century war photographer Robert Capa famously said. My friend in Ukraine who has been covering the Russian invasion since it began was concerned that so much of the horror of war cannot be seen in the photographs; it happens in the in-between, outside the frame, in the air around you. But that initial photograph he sent me made me deeply shudder, as did the photographs that would follow.

Photographic representation is always fragmentary—a fractional second within a rectangular frame—but the viewer can be provoked to imagine the larger whole while reflecting upon what is being shown. Robert Doisneau, the legendary French street photographer, titled a book of his life’s work as “Three Seconds of Eternity,” given that the camera’s shutter is only open for a small moment. But those few seconds can also invoke a much larger sense of the world, even an eternity.

While the photograph is interpretive, made differently by photographers according to each one’s understanding of the scene before them, it is as a recording of the visible that each photograph, and the photographers who made them, are reliable as evidence, as credible witnesses. If it were not essentially a recording, the photograph largely becomes an expression of opinion, explicitly subjective, and no longer a photograph (at least not in the journalistic or documentary sense). As one of the founders of Doctors Without Borders put it, "Without photography, massacres would not exist. Nothing can be done without pressure on politicians.” The surviving eyewitness would be discredited as too subjective.

Now with the advent of artificial intelligence, it is increasingly the surviving eyewitness whose testimony is being used to produce photorealistic images. These are not recordings (no camera is involved) but are generated by text prompts, and the images that are outputted will as they become indistinguishable from photographs, confuse, and eventually displace much of the lens-based record. For example, recently in Ashkelon, Israel, an exhibition was presented showing synthetic images of childhood traumatic memories from the testimony of elderly Holocaust survivors. The exhibition was characterized as a way “to preserve the memory of Shoah victims.” But while potentially therapeutic for the individuals involved, the project may also promote forgetting. As more of these colorful, vivid, dramatic images made in response to memories decades later begin to populate the historical record, they will make the actual, more dispassionate black-and-white photographs seem somewhat banal, expected, and mechanistic. In an age when selfies brand and celebrate one’s own experience, displacing the self-portraits that introspectively tried to explore one’s inner being, would one rather believe an eyewitness or a camera?

In a similar vein, recently the Auschwitz-Birkenau Memorial and Museum had a Brazilian specialist colorize photographs of the prisoners who had been there in an attempt to engage with visitors for whom the historical record may not have been sufficient. To quote the website facesofauschwitz.com, “By bringing color to the original black and white registration photos and telling prisoners’ stories, “Faces of Auschwitz” commemorates the memory of those who were murdered in the name of bigotry and hate. It acts as both a memorial to their passing and a warning to the world at a time when the memory of the Holocaust becomes increasingly abstract and remote.”

An Irish artist similarly colorized portraits of Cambodians before they were tortured and killed by the Khmer Rouge; in some cases he made them appear to smile. Vice Media, which published the altered photograph, said the colorization was intended to “humanize the tragedy.” But according to the New York Times, “’The colors do not add humanity to these faces,’ said Theary Seng, a survivor of the Khmer Rouge who has written a book about her childhood experiences. ‘Their humanity is already captured and expressed in their haunting eyes, listless resignation, defiant looks.’” Instead, the inhumanity was in the artist’s “inexplicable adding of makeup and a smile, as if to mock their suffering.”

What then is to stop synthetic images from being produced by Nazis and their sympathizers, or by members of the Khmer Rouge, that are linked to their own memories, or the ways in which they might prefer to view the past? Might there be a revisionist history of benevolent guards playing cards and football with prisoners eating ice cream? Given that each image requires only a text prompt to generate, such takeovers of various histories may become all too commonplace, the image war reaching deep into the past.

After viewing the images from Ashkelon, as an experiment to provoke discussion I created smiling plantation owners from Mississippi in 1855 via text prompts. I showed them at a public lecture where they were viewed as convincingly realistic. Of course, photography in the 19^th century did not allow people to smile due to long exposure times. But by having them smile I had allowed them—these people who did not exist—to appear less rigid and more human. The far greater sin, of course, was that this computer-generated façade of geniality can camouflage and distort the sordid history of slavery in the United States. If these images were to be placed on the Internet, newer AI systems could then train on such falsifications, distancing future generations from the horrific realities of slavery in the United States. The profound sufferings of people throughout the world can be, to coin an unfortunate term, AI-washed.

It was then with a sense of both enthusiasm and trepidation that I clicked on a link to an online project by a law firm, Maurice Blackburn Lawyers, depicting, in both text and image, the abuse of refugees held in Australian offshore immigration processing centers, published after a discontinued pro bono class action lawsuit. The images are all synthetic, based upon more than 300 hours of interviews conducted with detainees. With very little visual evidence available due to restrictions on journalists and camera equipment, the viewer is informed that “the images in this project have been generated using AI technology. They are not real but the experiences they depict are.”

The project was shown at Melbourne’s Immigration Museum, made into a book, and also presented online as “Exhibit A-i, The Refugee Account,” prefaced by a content advisory warning that “The following pages depict details of physical assault, verbal abuse, coercion, racism, homophobia, religious discrimination, unlawful incarceration, rape, murder, paedophilia, self-harm, self-immolation and suicide. Readers may find it distressing.”

In a short accompanying film, the imagery is vivid, photorealistic, visceral, and grim: a man’s battered face, what is said to be a man burning himself to death, an assault by a guard, the aftermath of a rape, a bloody sink as a mother recounts her very young baby witnessing a suicide, police in riot gear milling about the encampment, and other situations.

The self-representation encouraged here undoubtedly empowers those who had been victimized, and the synthetic imagery protects their identities. Rather than relying upon a photographer visiting for a short while unable to photograph what had previously occurred, each detainee can now be consulted in how they want their experiences to be represented. Also, the personalization of the imagery allows a more visceral rendering of how the abuse was perceived by the individuals involved. Or, returning to the organizers’ somewhat paradoxical differentiation that the images “are not real but the experiences they depict are,” the primacy of the latter is asserted. But one wonders whether, even if cameras are present in similar circumstances in the future, synthetic imagery generated from victims’ testimonies might not be favored as being more real.

The statements introducing the film evoke the dominant role of the photographic or film image in the last century: “The most powerful evidence in history is visual;” “The right image not only exposes injustice but can help put a stop to it;” “Only by making injustice visible can we provoke change.” Applying these assertions to camera-less images based on the testimony of individuals is, at this point, wishful thinking. An article describing the project also asserts that the AI-generated images have been placed on an editorial website “to sit alongside photojournalism,” an initial step in photojournalism’s destabilization.

Synthetic imagery has emerged in an era of post-truth and “fake” news, when the factual is increasingly irrelevant and when few hierarchies remain which are still considered to be credible in the larger society. Photographs will be found to be too limiting for those intent on seeing the world “in our own image,” the title of a book I wrote that was published in 1990. In the short term, images that borrow from the previous authority of the photograph will be made as a kind of quasi-proof, following several decades of photo manipulation with software. In a consumer-capitalist era where the customer is always right, the visual record becomes an obstacle. In a society suffused with image, one can have, or at least pretend to have, any persona, family, friends, and history desired.

The need for guardrails to help sustain a believable sense of contemporary and historical realities is evident and urgent. Authorship, codes of ethics, statements of principles, media literacies, and labels to differentiate various kinds of imagery that look similar, all need to be asserted. And there must be repercussions, legal and otherwise, for those who intentionally mislead and victimize others using the rich array of tools now available, including those generated by artificial intelligence.

Otherwise, we will have gone from a useful twentieth-century bromide, “the camera never lies,” to a sense that nearly all is lies. In that case, one will have to ask whether my Ukrainian photographer friend, dodging bullets and bombs, should have stayed home and whether the war in his country, like the tree falling in the forest, would have made any noise.

Fred Ritchin is Dean Emeritus of the School at the International Center of Photography, former picture editor of the New York Times Magazine, and former professor of Photography & Imaging at New York University’s Tisch School of the Arts. At NYU he also taught at ITP (the Interactive Telecommunications Program) and co-founded the Photography and Human Rights Program. He created the first multimedia version of the New York Times in 1994-95, and in 1996 conceived and edited the online non-linear documentary, "Bosnia: Uncertain Paths to Peace," photographed by Gilles Peress, which the Times nominated for a Pulitzer Prize in public service. Ritchin has written three books on the future of imaging, including In Our Own Image: The Coming Revolution in Photography (1990), After Photography (2008), and Bending the Frame: Photojournalism, Documentary, and the Citizen (2013). He is currently writing The Synthetic Eye, to be published by Thames & Hudson in 2024.