On Photorealistic AI and Social Media
In 1993, three months after Israel and the PLO signed the Oslo Accords, the Canadian magazine Maclean’s published what was soon to become an iconic photograph. It shows the backs of two boys in brotherly embrace, overlooking a blurred-out Jerusalem cityscape. The boy on the right is sporting a yarmulke, the one on the left a keffiyeh. The picture came to symbolize hope for the future because, as everyone knows, the future belongs to the children. As it turned out, however, the symbol was hollow, and the hope was manufactured—both boys were Israeli Jews. When confronted about this deception, the creator of the image, American photojournalist Ricki Rosen, said that “it was never supposed to be a documentary photo,” but a “symbolic portrayal of the idea of a long road to peace.” Apparently, Rosen had been following the instructions of her photo editor, who even provided a sketch of the desired composition. The Toronto-based editor cannot have been familiar with Middle Eastern dress codes, since the boy’s keffiyeh was secured with an agal, normally reserved for older men. But in 2014, the image was compelling enough for Rihanna to tweet it out during yet another conflict between Israel and Gaza. The photo spread further, not only through some 46,000 retweets of her post, but also through other online publications. And that is when the two boys in the photograph (now thirty-something-year-old men) were interviewed about their modeling gig of two decades earlier.
The 2014 resurgence of the faked 1993 photo is an excellent example of how propaganda works. The Encyclopedia Britannica defines propaganda as “the more or less systematic effort to manipulate other people’s beliefs, attitudes, or actions by means of symbols (words, gestures, banners, monuments, music, clothing, insignia, hairstyles, designs on coins and postage stamps, and so forth).” Propagandists “deliberately select facts, arguments, and displays of symbols and present them in ways they think will have the most effect.” Hence the use of children (innocent of prior history and symbolizing the future) in the headwear of the two respective groups (representing the collective) in the Maclean’s photo. Propagandists often envision themselves as educators, believing that “they are uttering the purest truth,” while the recipients of propaganda may see the message as both truthful and educational. Not realizing it was a fake, Rihanna must have thought that the photo was a constructive symbol of the hope for peace, and a good way to compensate for her earlier #FreePalestine tweet, which she deleted due to backlash after only eight minutes. According to the Britannica, “‘true believers’—dogmatic reactors to dogmatic religious, social or political propaganda”—are conditioned to trust whatever is being preached, because they are already sitting in the choir stalls.
Our information environment has changed dramatically since 2014. The social jolts of #MeToo in 2017 and #BLM in 2020 were strengthened by the rapid rise of an attention economy that dictates social media use. Since our attention is a limited resource, social media platforms from X through Instagram and Facebook to Reddit endlessly hone their persuasive techniques to motivate users to revisit their sites, create friendship networks, and affirm their social virtue—by, for example, changing their avatars to black squares en masse after the murder of George Floyd in the summer of 2020. Following the massacre of 7 October, there has been a renewed push to assign everyone on these platforms to committed camps. And conveniently, there is a new contender for the most effective weapon in the propaganda wars: photorealistic, generative AI art.
About a week after Hamas slaughtered more than 1,400 Israelis, a user of the AI platform Midjourney created a reddit post entitled “Visions of Peace.” It features ten images, all of which show smiling and hugging people—both children and adults—who can be identified by the their dress as Arabs and Jews. In one image, a girl and a boy are sitting by the seaside. To ensure that there is no doubt about their respective group affiliations, the girl is wearing an oversized Star of David necklace, while the boy’s head is covered with a keffiyeh. This display of symbols presumably reflects the verbal prompt that generated the image—probably something along the lines of “Jewish girl and Palestinian boy smiling and embracing by the sea.”
While it is photorealistic, the image is replete with improbable details that result from the AI’s translation of language prompts into unrefined visual renderings. The girl’s necklace is preposterously large, and her Star of David is the size of the crosses worn by Orthodox priests. The boy is dressed like an old man—white button-up shirt, dress coat, and a keffiyeh secured with an agal—the same sartorial inaccuracy as in the staged 1993 photograph. The children’s oddly intertwined fingers are a telltale AI flaw, as is the mismatched landscape in the background. AI is still learning the tricks that were mastered by Western painters six centuries ago.
The comments on the initial post were overwhelmingly positive. Many Redditors recognized and welcomed the propagandistic nature of the image of the two children. One commentator even spells it out: “This is the kind of propaganda we need!” He had his wish, as the image was shared and reshared numerous times across social media, often accompanied by emoji-filled, inspirational comments, such as “Pray for peace,” “Hate is learned, love is our nature,” and “We used to be friends.” There is clearly a market for images like this—easily legible (cue the jumbo necklace) virtual modifications of reality. But, thanks to the photorealistic nature of generative AI art, not everyone who liked and shared the image realized that it was simulated. The image pretends to be photography, and by association it pretends to be true to life. The story of Ricki Rosen’s staged 1993 photograph caused a stir at the time because back then people expected such an image to be a snapshot of two real-life friends hugging. That would have been a sign of real-life hope and inspiration, epitomized by two boys growing up in the aftermath of the Oslo Accords to coexist peacefully. When the conflict was reignited in 2014, the same image, improbably enough, was again used to urge peace and reconciliation.
Now, in 2023, a second photo from the 1993 shoot is making the rounds. In it, two boys are hugging as they are walk towards the camera. Judging by the captions and comments on this second photograph, social media users don’t doubt its veracity. They assume it is an actual interaction because it is a photograph. Nor are they wrong to do so, since veracity has traditionally been key to photography, which has always proclaimed that truthfulness is its most fundamental feature. Painting—even the most accurately mimetic kind—is manufactured by the human hand directed by the human eye. Analogue photography, by contrast, relies on the machine that produces the image via light exposure and chemical reactions. When photography was pioneered in the 1830s, it purported to be the visual agent of truth. The famous pictorialist photographer Henry Peach Robinson was widely criticized for his infamous 1858 print Fading Away, composed of five separate negatives, showing a young woman in the last stages of consumption, surrounded by her grieving family. Since a photograph was meant to be recorded proof of an incident that took place in real life, Robinson’s contemporaries deemed that this “untruthful,” artistic, image was an insult to the duped viewer.
Photorealistic AI-generated images, if recognized as such, should function as symbolic representations that invite the suspension of our disbelief. They should not be seen as photographs, and as therefore “truthful,” and they must remain separate from any context in which they could be confused with straight analogue or digital photographs.
Among the images shared in the context of the current Gaza conflict was a photograph that I first spotted in one of my media feeds. It is a candid shot taken in Jerusalem seven years ago, by Russian-Israeli photographer Mikhail Levit (b. 1944). The photograph made me pause my doom-scroll and examine it carefully, because it contained something that Roland Barthes, in his gem of a book Camera Lucida, calls “punctum.” Barthes proposed that a truly great photograph must possess two components: studium and punctum. The former, he explains, is the result of a shared cultural interest, a reading of visual signs. The latter is a “sting, speck, cut,” an accidental detail that “pricks” and “bruises” the viewer. When we recognize the studium, we grasp the photographer’s intentions and share the photographer’s culture. The experience of punctum is of a different order. It is emotional. It provokes pity, love, disgust, empathy, anger.
Levit’s photograph shows two elderly gentlemen: a Jew and an Arab. They are engrossed in conversation—the Arab man is grasping his cane with his left hand and subtly gesticulating with his right, his face turned away from the camera; the Jewish man is holding a cigarette and leaning forward, listening intently. Neither of them is looking at the photographer. It is a candid snapshot of a moment in time. I found the photograph incredibly touching. I was struck by the similarity of their dusty shoes and worn-out hems, by the way their feet point towards each other. The two old men are of different cultures, but they could be brothers; so close are they to each other both physically and emotionally. Touching as it is, the photograph is hardly a poster advertising hope for resolving the Gaza conflict because the well-meaning sharer of the image was wrong in assuming that it depicted an Israeli and a Palestinian—the men are clearly an Israeli Jew and an Israeli Arab. But even if Levit’s photograph does not apply to the current conflict between Israelis and Palestinians, it is still a powerful symbol of coexistence, because it captures an instance of peaceful interaction in real life. It is, in short, the truth.
The real price of the faked hope represented by the overshared image from the “Visions of Peace” is that the audience is fooled. Instead of responding to the cultural complexity of the studium and the emotional jab of the punctum, the viewer fatuously follows a breadcrumb trail of propagandistic triggers. While a great photograph—be it analogue or digital—is likely to provoke thinking, a formulaic propaganda image tends to shut it down, sequestering its viewers in their respective echo chambers. Nuance and ambiguity don’t survive long when a picture is reduced to a visual slogan. All a visual slogan can do is keep those thumbs scrolling.