Review: <i>The Synthetic Eye: Photography Transformed in the Age of AI</i> , by Fred Ritchin

When, in 1982, National Geographic doctored a horizontal photograph of two pyramids of Giza to make them appear closer to one another and render the scene fit for the magazine’s vertical cover, Fred Ritchin was one of the first commentators to sound the alarm bells over the crisis of veracity in photojournalism.1 Writing for The New York Times Magazine in 1984, where he served as picture editor from 1978 to 1982, Ritchin contemplated the need to label photographs as either “fiction” or “nonfiction” due to their increasing likeness while relying on the imagemaker instead of the image to discern their difference.2 Of course, image manipulation in some form or other has existed since the early days of photography. The problem, however, was the scale and pace at which manipulation became possible following the digital image revolution of the 1980s. Ritchin revisits this episode at the outset of his new book, The Synthetic Eye: Photography Transformed in the Age of AI, to connect our epistemic anxieties around “synthetic” or AI-generated photorealistic images to half a century of changing image-making technologies.Photo alteration became widely accessible with the introduction of Photoshop in the 1990s, which replaced the hefty Scitex workstations of the preceding decade. But algorithmic systems soon automated much of the editing process. Image applications today offer one-touch tools to whiten teeth, reduce waist size, and remove background clutter. And camera processors in smartphones make autonomous decisions based on what constitutes a good photo, compositing multiple frames to render skies bluer, reduce noise, or retrieve details from shadows—often beyond our command. Ritchin argues that such developments in computational photography advance “a form of proprietary looking in which what we want to see is realized as a form of consumerist entitlement” instead of establishing a “dialectical relationship” with the world (24). Generative AI has joined the latest phase of this continuum. Except, if fabricating photos allows one to momentarily escape reality, with image generators one no longer even needs to encounter the world one seeks to flee.The question of predictability emerges as an underlying thread of inquiry in Ritchin’s attempt to trace a techno-historical genealogy of AI image generators, or systems where algorithmic probability meets the monotony of photographic representation. Text-to-image generators like DALL-E, Midjourney, and Stable Diffusion tend to reproduce predictable tropes because programmers have trained them on vast datasets of generic photographs found online. A prompt asking for “a writer” will likely show a person surrounded by books, one asking for “CEOs” will portray men wearing suits, and so on. For Ritchin, such a “rudimentary lexicon” precisely regurgitates the preconceptions that have long predominated editorial decisions in photojournalism (44). To feed the 24/7 news cycle, which began in the 1980s with CNN, photojournalists are required to produce a few formulaic and illustrative images that can accompany a story without needing sufficient contextualization. In due process, photography “loses some of its lyricism” and ends up being “less intuitive and more predictable” (47).Toward the end of the twentieth century, the growth of stock photo agencies such as Corbis and Getty Images also consolidated enormous archives of creative and editorial photography for digital distribution. In 2023, Getty Images sued Stable Diffusion for the company’s unauthorized use of twelve million copyrighted images and their keywords to train its AI system. Yet now that standardized AI-generated “photos” are themselves appearing in stock libraries, predictability has fallen into an ouroboros loop. In this regard, Ritchin draws attention to the controversy surrounding Amnesty International’s use of AI-generated stock images in their social media teasers for reports on the 2021 protests in Colombia. The human rights organization reportedly used those images to protect the protesters’ identities (93).Ritchin draws on years of experience as a picture editor, curator, and photography professor to offer an eclectic range of references and case studies on the medium’s history in light of ongoing conversations about synthetic images. In the opening pages of The Synthetic Eye, for instance, a discussion of Italo Calvino’s 1958 short story “The Adventure of a Photographer” leads to a point in Charles Baudelaire’s 1859 essay “On Photography,” which then transitions to a recent observation on AI photography by tech journalist Stephen Shankland—and the citational mélange continues throughout. More rewarding are Ritchin’s insightful anecdotes that fill the chapters, referring to his classroom exercises with students (119), his Pulitzer-nominated project on the Bosnian genocide with the photographer Gilles Peress (146–49), his interview with Alvy Ray Smith—whose company Lucasfilm first attempted to generate a photorealistic image after “100 hours of calculation”—in 1984 (60), and his appearance on NBC’s The Today Show to discuss image manipulation “the day after Iraq had invaded Kuwait” (37), among others.Nonetheless, the never-ending flow of incidental references in the book becomes tedious after a point since no sustained line of theoretical intervention seems to bind them. This occurs primarily due to Ritchin’s overemphasis on a few aspects of technological mediation and sheer impatience with social analysis. For example, when Ritchin claims that photographs have transformed from being “referents” to “desirents,” with the increasing efficiency of altering and synthesizing images (205), he conveniently sidesteps analyzing the deeper social structures or economic and political conditions underlying the thesis. His perfunctory compensation: this transformation charts people’s desire to have a “sense of control” in chaotic times and to “re-make the world” (205), or be the “masters of their own destiny,” amid the general erosion of trust in traditional institutions and experts (36). That’s all. No scientific corroborations follow.Ritchin’s argumentative framework also hinges too heavily on the platitudinous binary between analog and digital photography, the former being “reassuringly continuous and Newtonian,” the latter more “quantum-like” and characterized by “non-linearity” (36). In historicizing this binary, Ritchin relies on Baudelaire’s concerns about photography’s “absolute factual exactitude” in the 1850s, contrasting them with the uncertainty and ambiguity that inform the medium’s understanding today (23). He neither factors in contemporary scholarship that has complicated the linear analog-to-digital history of photography nor, for that matter, reconsiders Baudelaire in a new historical light. Ritchin had earlier relied on this binary in his 1990 primer In Our Own Image: The Coming Revolution in Photography, which summarized how the digital age might challenge people’s trust in photography. There, too, as cultural critic David Levi Strauss wrote in a review of the book, Ritchin misses that the “crisis of belief” in photography is not a result of “new electronic imaging technologies in themselves” but reflects larger patterns of surveillance, control, and visuality assaulting “our public right to know.”3Related to believability, the principal apprehension for Ritchin is the potential of synthetic images to “destabilize a sense of a shared history” (79). He experiments hands-on with DALL-E and DreamStudio to illustrate how AI can generate convincing photorealistic images of improbable situations, such as those depicting “happy people in the Warsaw Ghetto during World War II” or “happy prisoners on a Texas chain gang” (81–82). The scenario turns more problematic when his prompt for “segregationists in the South after the Civil War” returns a photorealistic image of exclusively Black men, as if Black men resisted an integrated society, or when DreamStudio generates an image showing African Americans as plantation owners in the United States (84–85). Here, one might contest Ritchin’s anxiety: are these images sufficient to derail people’s knowledge of the past? What role do ideological factors play in determining the index of truth and believability around images? What about AI-generated images that viewers consume and operationalize with a willing suspension of disbelief, be it for humiliating others or deriving pleasure? Well, Ritchin does not tread into these territories. Nor does he investigate viewers’ engagement with synthetic imagery.Despite voicing his worries about the status of photojournalism today, the tone of Ritchin’s assessment is not overwhelmingly despondent. He also shares several cases where artists and institutions have used image generators toward reparative measures, including the law firm Maurice Blackburn’s use of synthetic images to represent the abuse of refugees in Australia, based on “300 hours of interviews with 32 detainees” (96). But, without active and collaborative participation, such approaches “can portend a digital colonialism” in which those with means can photorealistically render the lives of those without any to the point of mass expropriation (93). After all, current models of image generators mostly “learn” from photographic archives that have “consistently demeaned, erased, and marginalized groups of people” (102).From cover to cover, synthetic images out of Ritchin’s experiments with image generators recur frequently alongside the prompts that generated them. He remains hopeful about the occasionally “serendipitous” outputs of these AI tools that can “circumvent truisms” and explore “parallel universes” (211). Given that they “riff on ideas, thoughts and dreams” as well as a miscellany of nonvisual materials, Ritchin suggests collaborating with image generators to challenge our worldview rather than putting them to instrumental means (211). Be that as it may, the drab and unremarkable outputs of “challenging” prompts like “a photograph of the Salem witches on trial in 1692” or “Romantic Martians in love,” inspired by the work of Virginia Woolf, barely seem to hold up his claim (226–27). Ritchin eventually falls back on the liberal humanist bottom line that synthetic imagery can be relevant when used with “transparency and respect” (220). In all, The Synthetic Eye takes stock of photography in the age of AI with an arsenal of quick facts and historical snapshots, but struggles to establish a theoretical position for itself.