A Network of Global Incinerators

Rob Arcand reviews Hito Steyerl’s new essay collection, “Medium Hot: Images in the Age of Heat.”

By Rob ArcandOctober 19, 2025

• Science & Technology

• Art & Architecture

The computer can be used to direct a network of global thermostats to pattern life in ways that will optimize human awareness. Already, it’s technologically feasible to employ the computer to program societies in beneficial ways.
—Marshall McLuhan, Playboy interview (1969)

THE DIGITAL IMAGE is dead, long live whatever fluid vessel has taken its place this week. It has been more than 15 years since the artist and writer Hito Steyerl first mounted her defense of the grainy, deep-fried, low-fidelity photos that populated early social networks and file sharing services in her 2009 essay “In Defense of the Poor Image.” That text fundamentally shifted the discursive frame of contemporary art and christened Steyerl an authority on the unruly digital vernacular of the late 2000s and early 2010s as it was just beginning to seep into the mainstream. Her approach, however, was never limited to discussion of media formats and the modes of circulation they’ve enabled. Across a sprawling catalog of texts, films, and multimedia artworks, she stretched the language of critique to new limits, playfully prodding surveillance technologies mere months after Edward Snowden revealed the existence of the National Security Agency’s vast domestic spying apparatus, or mapping the role of art in a global order defined by collapsing nation-states, widespread inequality, and other aspects of 21st-century neoliberal capitalism. It was, no doubt, because she seemed to grasp something essential about the logic of the present—its political, economic, technological, and visual turbulence—that Steyerl, in 2017, topped ArtReview’s Power 100 list, becoming the first woman to receive the honor of the art world’s most influential person.

It seems obvious, then, that Steyerl would be drawn to recent developments in artificial intelligence, which, in the span of a few short years, have made the politics of visual culture suddenly everyone’s problem. Yet Medium Hot: Images in the Age of Heat, her new essay collection, feels smaller and less substantial than much of her previous work, even as it ostensibly approaches some of the biggest issues of our time. At roughly 160 pages in length, the text isn’t discernibly thinner than Duty Free Art: Art in the Age of Planetary Civil War (2017), her previous essay collection. But across its 11 essays—spanning a sprawling assortment of diverse yet interconnected topics, including AI image generation, rising global temperatures, and the chaotic writing of newly influential far-right thinkers like Curtis Yarvin, Nick Land, and Peter Thiel—the author seems adrift, moving from topic to topic in search of new objects worthy of scrutiny. The texts suggest clear limits to Steyerl’s approach as a critic, as she struggles to meet the present crisis with the urgency it demands. Yet in their shortcomings, the essays also unwittingly ask what function criticism still serves in a moment of right-wing ascendance, making clear a pressing need for new modes of transformative, full-voiced critique.

From the beginning, the strength of Steyerl’s writing seemed to lie not just in its ideas but also in its ultracontemporary style, which seemed to mirror the forms it was addressing. In earlier essays such as “In Defense of the Poor Image” and “Duty-Free Art,” she used short sentences and punchy language, rapid digressions and fragmentary structures, drawing as much from the distracted language of advertising and social media as from critical theory. The collage-like quality of much of her most celebrated writing entails casual, slangy citation from the critical canon (Benjamin, Adorno, Deleuze) while prioritizing stylish turns of phrase over sustained engagement with the references. “In Defense of the Poor Image” is riddled with declarative statements about digital photos, but Steyerl’s sentences never really cohere into a larger argument beyond what’s present in the title. In place of a thesis, she organizes her essays around the production of new keywords. Stylish terms like “poor image” and “duty-free art” can be easily taken up by others, and her essays appear to be built from the ground up with this kind of virality in mind, as she elliptically circles themes as an exercise in personal branding.

Medium Hot exposes the fundamental cynicism of this approach. Initially published in the New Left Review, “Mean Images” hinges on a weak pun. Steyerl begins with the observation that AI images are often developed through computational processes of averaging and adds more pointed statements about the “hallucinated mediocrity” of these averaged images, the confused responses they stir in human viewers, and the “means of mean production” through which such images are produced. “Mean images are far from random hallucinations,” she writes. “They visualise real, existing social attitudes that align the common with lower-class status, mediocrity and nasty behaviour.” Yet the central claim here—that AI images reflect a populist spirit that’s viewed by some as “nasty” and “lower-class”—is deeply muddled, a poorly sourced generalization that she doesn’t seem especially interested in expanding upon. Insights like this one tend to operate throughout the collection as disconnected maxims, stylish language addressing broad themes that are rarely developed into larger arguments.

Other essays, on the “operational image” or “burnt-out image,” also adopt this style. In an essay titled “The Digital Rift: From Poor Images to Power Images,” Steyerl returns to her earlier “poor image” coinage, affirming the prophetic accuracy of her 2009 essay, and drawing a line to the second, novel image type of the title. “Low-res styles have been fully integrated into all aspects of digital capitalism, from 8-bit Balenciaga to right-wing memes,” she writes. The poor image may have once been a useful stand-in for the social and material conditions of “communicative capitalism,” but Steyerl suggests that it has now become so ubiquitous as to have been practically forgotten. The changing political-economic landscape now requires new terms, which Steyerl thankfully provides in spades. She isn’t wrong to suggest that the same photos and videos she once praised became important groundwork for the modes of “accumulation and extraction” required to produce AI images. Yet her insistence on the prescience of her previous term (which is notably applied quite flexibly) tends to overshadow any genuine curiosity she may have once maintained about AI images, their politics, and the deranged and distressed society that produced these images to begin with.

To be fair, there is an abundance of interesting research here. Her essays on entropy and thermodynamics situate the devastating ecological impact of AI image generation within existing scholarship on technology’s effect on the environment, drawing on Bernard Stiegler’s work on the Anthropocene, Nicole Starosielski on heat and information theory, and Donald MacKenzie on the mathematical models used throughout the global banking sector to price financial derivatives. Other essays, on the use of artificial intelligence in contemporary military technologies and the cultural narratives used to excuse the short-term consequences of AI as a necessary expense for long-term visions of progress, show that Steyerl recognizes that there is an implicit connection between today’s AI images and the larger role that artificial intelligence has come to play as part of efforts to maintain systems of global violence.

Steyerl’s view of the AI image as an index and spectacle—both in the sublime visions of the future it promises and in the devastating military violence it already helps to enact in the present—is perhaps the strongest claim of this collection, and her writing is at its best when playing to this beat. Her essay on the “war economies” of statistical image production provides the most focused instance of this argument. In that essay, “‘The Hardest Part’: Statistical Image-Making and War Economies,” she highlights the ideological function that AI images currently serve, masking the larger project of artificial intelligence—which is premised upon the vast extraction of mineral resources, which actively contributes to rising global temperatures in measurable (and widely measured) ways, and which is used to identify and select targets for an ongoing genocide—as something relatively benign. Yet Steyerl’s account of this violence remains tepid throughout the collection; beyond referencing defense contractors such as Palantir in passing, her discussion does little to grapple with the scale and intensity of what she’s describing, listing off a few brief examples of this horrifying reality before moving on.

Other instances of this central claim—that the AI image is a Trojan horse for more menacing uses of artificial intelligence generally—are burdened by the theories of others, which Steyerl rarely builds upon in any substantial way. In the introduction, she highlights the work of the media theorist Jussi Parikka, whose 2023 book on “operational images” goes much further in its efforts to historicize machine vision and to provide methodological guidance as to how one should approach efforts to “read” machine images that aren’t primarily intended for human viewers. “Operational images organize the world,” he writes, “but they also organize our sense and skills in terms of how we are trained to approach such images, from the photogrammetric mapping of landscapes to pattern recognition, astronomy datasets to Mars Rover imaging practices.” His extended passages on the astronomer Henrietta Swan Leavitt, whose elaborate datasets measuring how light travels through outer space made detailed photographs of the night sky possible for the first time, and on the sprawling prehistory of photographic metadata are genuinely novel and highly relevant to the applied study of machine vision today. But all of this goes unmentioned in Steyerl’s text beyond her use of the term “operational images,” which, alongside mentions of the work of Bernard Stiegler, Nicole Starosielski, Donald MacKenzie, and others, feels like a missed opportunity to make a more ambitious intervention.

It is also striking, amid so much discussion of images and visuality, how few artists and artworks are featured in these essays. Parikka highlights Steyerl’s artistic practice—and specifically her 2013 film How Not to Be Seen: A Fucking Didactic Educational .MOV File—as exemplary of art’s potential to address the politics of visibility and image production. But Steyerl rarely draws on the creative works of others, a substantial omission from such a celebrated artist. When she does, it is to scour the works for traces of the AI technologies used in their production and the political valences of such use. In her “war economies” essay, for example, she makes extended use of “The Hardest Part,” a music video directed by the filmmaker Paul Trillo for the American indie band Washed Out. The four-minute film, which was billed as the first AI-generated music video made with OpenAI’s groundbreaking text-to-video application Sora, follows a teenage couple from behind, racing toward an elusive vanishing point on the horizon as the scene changes rapidly around them. Steyerl reads the video’s forward motion as representative of the race to produce new advances in artificial intelligence, pushing what is, at best, a tenuous connection between the film and its surrounding ideology firmly into the space of tortured metaphor:

The best answer for why the characters in “The Hardest Part” are running, therefore, is perhaps not even that they are so keen to win the race. Instead, perhaps they are running away from accepting that, though the race and its consequences are real, it remains unclear whether it bears social benefits for anyone other than major corporations, authoritarian states, emerging war economies, fossil and nuclear energy providers, extreme right-wing organisations and the like.

Elsewhere Steyerl attempts to read the output of image generators like Stable Diffusion as akin to works of art, reflecting the social, political, and aesthetic values of the models and their creators. In “Knuckleporn or Phocomelia: GenAI’s Dialectics of Enlightenment,” she examines the grotesque human forms spit out by Stable Diffusion 3, a model that struggles to generate appendages and joints with the realism now expected of AI image generators. This aesthetic limitation is partly the result of regulatory changes affecting the model’s training data; after child pornography, nudity, and other NSFW material was discovered and removed from its dataset, the model was unable to replicate the physical balance and weight distribution of human forms. Later, this NSFW training data was reintroduced, but these models, which Steyerl notes feature a “healthy dose of gay porn influence,” still produce phantasmatic chimeras—“a bestiary of basilisks, monopods and struthopodes”—leading her to suggest that the images are symptomatic of a fundamental flaw in Enlightenment thinking. Where Kant and Descartes believed that the most effective forms of governance and social progress would emerge fairly naturally from human thought of the most rational order, Steyerl holds up the horrifying AI images produced by Stable Diffusion 3 as an example of a perversity produced by reason. Echoing Dialectic of Enlightenment (1947), Max Horkheimer and Theodor Adorno’s foundational work of social theory that traces forms of rational thought to appalling end points in capitalism and Nazism, Steyerl notes that “rationality recoils into its absolute opposite.” Yet, somehow, Horkheimer and Adorno’s extended critiques of Nazi rationality go unmentioned in the essay.

Steyerl’s interest in the justifications for continued work on AI is further developed in “Roko’s Basilisk: Artificial Stupidities and Existential Risk,” an essay reflecting on the philosophies of the billionaire founders and venture capitalists who have made these technologies available. Like Horkheimer and Adorno, she is guided by an interest in rationality, wondering how ostensibly brilliant people come to participate in horrifying systems of exploitation, abuse, mass slaughter, climate denialism, and innumerable other perversions of rationality. Yet substantive critiques of capitalism are oddly nowhere to be found. Instead, Steyerl focuses on eccentric thought experiments like Roko’s basilisk; the philosophies of Curtis Yarvin, Nick Land, and Michael Anissimov; and the influence of their thought on the worldviews of tech founders including Peter Thiel and Balaji Srinivasan. Steyerl treats these ideas as harbingers of potentially terrifying futures to come, even as the many horrors of the capitalist present, like so much else about her worldview, go wholly uninterrogated within the text.

It’s a shame that Steyerl doesn’t more thoroughly engage with many of the ideas she cites since Horkheimer and Adorno’s account of fascism is deeply intertwined with theories of media and mediation. For these writers, radio and cinema provided the technological means through which the Nazi Party expanded and deepened its influence, and Dialectic of Enlightenment is as much concerned with Enlightenment rationalities and their perversions as it is with how these media technologies were used to mobilize support. A similar claim could be made of AI images, which, as Steyerl at times gestures toward, also perform an ideological function, turning user input into dazzling spectacles that conveniently elide the large-scale extraction and exploitation required for their production. Yet this kind of analysis, which has a lengthy history in Marxist scholarship on reification and commodity fetishism, would require that Steyerl take issue with capitalism, something she seems strangely reluctant to do. Her limited engagement with history and lack of systemic perspective ends up portraying contemporary horrors as new, unprecedented challenges, a critique that fails to achieve any distance from its subjects. When she finally hands the writing process over to OpenAI’s GPT-3 model in an effort to imagine new visions for the “disconnected archipelago of post-pandemic art worlds” in the book’s concluding essay, it feels like an indictment of the essays that came before it. Whether intended as a kind of détournement, the gesture comes off as an uncritical adoption of the very technology she goes to great lengths to critique. The incoherence would be galling if it weren’t present in her work all along.