It never ceases to amaze me how Hito Steyerl is able to take some of the most complex and pressing issues of our contemporary world and disassemble them in a playful yet contemplative, insightful, and satirical manner. This attitude is encapsulated in the title of her most recent exhibition, Humanity Had the Bullet Go In Through One Ear and Out Through the Other, organized at the MAK – Museum für angewandte Kunst. The multimedia installation uncovers the complex relationship between advances in the field of artificial intelligence, their use in the war economy, and the changing role and position of civilians. At their intersection, we come face to face with the results of data extraction and accumulation enhanced by the extensive, yet largely invisible network of image production.
The title of the exhibition evokes an ominously vivid mental image, originally penned by the Austrian essayist and journalist Karl Kraus. A vocal critic of the cultural and public affairs of the Austro-Hungarian Monarchy, Kraus dedicated himself to publishing satirical accounts exposing the lies and deceptions propagated by the official press at the beginning of the twentieth century. Many of his publications and books address the moral, societal, and political doom of humanity, brought about by World War I and the rising power of autocratic regimes, further enabled by rapid technological developments. By choosing this line from Kraus, the exhibition signals a conceptual continuity with his satirical and critical stance, while also disassembling the mechanics of the impending doom he envisioned.
Hito Steyerl, Hell Yeah We Fuck Die, 2016 Video installation (Edition 6_7) at Humanity Had the Bullet Go In Through One Ear and Out Through the Other © MAK
One does not exist without the other
Upon entering the white cube space, we find ourselves in an installation built up of large trapezoidal metal sheets and railings, evoking an obstacle course. Several outsized, illuminated words read Hell Yeah We Fuck Die – the signs of collective neurosis become painfully obvious by looking at the most frequently used words in English-language songs on the Billboard pop charts of the 2010s. Created in the style of a music video, a three-channel video work is omitting the peculiar sound of electronic music, while the footage shows a real-life version of the training course, used to train robots. Their treatment seems almost cruel: they are pushed and kicked around and get ruthlessly beaten by scientists. By playing out potential, violent scenarios, the aim is to enhance their performance so that they can be deployed to save people trapped in disaster zones. The video highlights the helpless, almost pathetic situation of robots: especially by showing the digital models trapped in an endless virtual scene, their sole purpose being to fail while practicing different movements. Their depiction evokes a continuation of slavery, extended over anthropomorphic nonhuman agents, which urges us to feel something close to empathy. Simultaneously, the style of the video renders it both ridiculous and frightening to envision a scenario, where people’s lives would be dependent on the intervention of robots.
Hito Steyerl, Hell Yeah We Fuck Die, 2016, © Hito Steyerl, Hell Yeah We Fuck Die, 2016, Loan LBBW Collection Video stills, © Hito Steyerl, VG Bild-Kunst, Bonn 2025
The video is juxtaposed by another, Robots Today, turning the relation seen previously on its head, and our neurosis into psychosis. As we are transported to the Kurdish region of Turkey, close to the Syrian border, the voiceover introduces the historical figure of Ismail al-Jazari, who lived and worked in the region in the 12th century. Called ‘the father of robotics’, he was a polymath, whose inventions, such as machines operating based on sequential programming, laid the foundation for modern day robotics and automation. Adopting the style of a first-person video game, the camera moves through of the military conflict-ridden towns, Diyarbakır and Cizre. Roaming the streets, we appear to be trapped in a simulation, with human figures moving like robots, while the narrator keeps questioning Siri about whose fault is the surrounding destruction. By centering al-Jazari’s legacy, the work draws up a temporal and spatial contradiction. First, it destabilizes Western hegemony in science and innovation by referencing the contribution of a historical figure from the East. Second, it contrasts the overall hopeful outlook of al-Jazari’s inventions to the present-day use of artificial intelligence, as it has become an essential part of military economy, wreaking havoc in the region it has originally sprung from.
Hito Steyerl, Hell Yeah We Fuck Die, 2016 Video installation (Edition 6_7) at Humanity Had the Bullet Go In Through One Ear and Out Through the Other © MAK
By pairing Hell Yeah We Fuck Die and Robots Today, we see two sides of the same coin: systems designed to save lives also contribute to the deaths and displacement of civilians – we just tend to turn a blind eye to the equivalence. In the age of AI warfare, ongoing conflicts play an essential role in turning entire regions into training course modules for multibillion-dollar tech companies to develop, deploy, and promote automated military technologies. Reality collapses into itself and gives way to an endless, nightmarish simulation.
The workers shoved inside Kempelen’s machine
The connection between warfare and the image is also highlighted in the final video installation. Long after the emergence of photography – but not so long after the exclusive domination of the optical/indexical paradigm – the technical image has shifted away from its representational and world-altering functions and toward its newfound ability to create visuals that could never be seen by the naked eye. Around the year 2000, Harun Farocki coined the term operational image to describe a new function of images: as instruments that carry out tasks – such as tracking, navigating, detecting, and identifying – as part of an operation. He argued that these images do not reflect how humans see, but how and what machines can see, and the kinds of images they can produce, all of which are completely invisible to the human eye.
Not long after, Daniel Rubinstein and Katerina Sluis introduced the term networked image, which, through the operation of vernacular imaging, essentially “expands” the idea behind the operational image to the way image-sharing platforms in the early days of Web 2.0 functioned. The uploading and tagging of several thousand images per minute models an early stage in users teaching the software how to ‘see’ images. [1] The logic of the platform renders the image to be in a constant state of flux: it is exposed to non-linear navigation, it appears and disappears, and it connects and disconnects with other images under various categories. From our viewpoint, it appears that we simply display our images, organized in carefully curated collections; however, in the depths of the system they become part of an expanding database and raw material for training algorithms. [2] Now, two decades after the first studies on this new understanding of imaging, it is more nonhuman than ever: from automated drone footage to prompt-based, artificially generated imagery, our position as the main producer and recipient of images has drastically shifted. We are completely lost as to what happens to the images we push into circulation on the network.
In her recently published book Medium Hot: Images in the Age of Heat, Steyerl attempts to make the invisible process visible. She identifies this new generation of nonhuman visuals as data necessary for the development of AI models – data that is extracted, privatized, and stored in vast databases owned by corporations that made themselves at home on the Internet with the neoliberal turn. Instead of light reflecting off surfaces, the primary source of the images circulating has become energy, generated by large infrastructures, systemically heating up the globe. We are witnessing an ontological, social, economic, and ecological shift in the nature of the image and its production. Since the processing of visual data consumes a large amount of energy, and algorithms can still be flawed, certain tasks are outsourced to humans – more precisely, to people in vulnerable positions due to conflict-induced displacement. Data populism and warfare reshape the conditions of labor and produce cheap, disenfranchised workers – in other words, data proletarians.
Hito Steyerl, Mechanical Kurds, 2025 Single screen installation, 13 min, at Humanity Had the Bullet Go In Through One Ear and Out Through the Other © MAK
Returning to the exhibition, this issue becomes more palpable in the final work, a documentary-style video titled Mechanical Kurds. It is based on interviews Steyerl conducted with Kurdish microworkers living in refugee camps in northern Iraq. Since displaced and vulnerable groups – especially women – have limited access to employment, they become prime targets for tech companies seeking machine-learning microworkers whose labor is used to train AI models. The interviewed refugees were commissioned to label various elements – such as cars, humans, and buildings – appearing in approximately seven thousand images of streets, which would later be fed into machine-learning systems.
The Mechanical Turk, source: Wikimedia Commons
As they complete small, repetitive tasks under the guise of fully automated ML infrastructures, the microworkers become invisible cogs in a corporate machine, appearing almost robotic. Steyerl uncovers their complex situation by paraphrasing the 18th century polymath, Wolfgang von Kempelen’s invention, “Mechanical Turk” or “The Turk”. It was a chess-playing automaton made of wood, in the form of an Ottoman pasha that could defeat anyone in the game. The widely sensationalized invention later turned out to be an elaborate hoax: the machine was hiding an experienced human chess-player, who was operating “The Turk” via magnets and levers. In Steyerl’s video, the attempt at disguising human labor as automation corresponds to the situation of Kurdish refugees. As the chess board turns into a refugee camp, and the logic of the game turns into the process of labeling targets, the “Turk” shatters and a Kurdish woman emerges from inside, signaling the invisible and severely underpaid work refugees – especially women – are forced to do under the conditions created by displacement and the digital fragmentation of labor.
Hito Steyerl, Mechanical Kurds, 2025, © Hito Steyerl Videostills, © Hito Steyerl, VG Bild-Kunst, Bonn 2025
The most blood-curdling realization emerges in a scene featuring a group of men riding in auto rickshaws. As they move through the partially AI-generated streets of a refugee camp, they suddenly come under a drone attack, and the tuk-tuks explode. Meanwhile, in the narration, one of the interviewees recalls his time working as a journalist in the region. He and several of his colleagues were involved in a car crash caused by a military drone, leaving him as the sole survivor. As a refugee, he was later recruited to complete visual labeling tasks as a microworker, believing that the purpose of the job was to assist in training self-driving cars. However, unbeknownst to him, he also contributed to the processing of data for the pattern-recognition systems of military drones – potentially similar to the one that had previously targeted him.
Hito Steyerl, Mechanical Kurds, 2025, Single screen installation, 13 min, at Humanity Had the Bullet Go In Through One Ear and Out Through the Other, © MAK
Steyerl’s exhibition vividly conveys the all-encompassing simulation in which we are trapped, where the value of human agency – and life itself – comes under relentless questioning with the advancement of artificial intelligence. The continuous, largely invisible production and circulation of data and conflict which we as humans become part of without fully realizing it, dissolves the boundaries between human and nonhuman agents, and exacerbates social inequality, resulting in vulnerable groups becoming not just the victims, but also the exploited, invisible engines of automation. Amid the blurring of the distinction between reality and simulation, only one thing is certain: once we identify the correlations, we cannot turn a blind eye to them. Steyerl’s exhibition forces us to stare down the dark barrel of the future, so that we might better anticipate when the next bullet will be fired.
Hito Steyerl: Humanity Had the Bullet Go In Through One Ear and Out Through the Other, organized by MAK – Museum für angewandte Kunst can be visited till the 12thof April 2026.
The text is inspired by Medium Hot: Images in the Age of Heat by Hito Steyerl that was published in 2025 by Verso Books.
[1] It might be worth noting that we could also find similar examples of this process predating the open-source era. For instance, the Clickworker experimental program launched by NASA in 2000 was based on crowdsourcing to process the vast number of images of the surface of Mars taken by the Viking probes and the Mars Global Surveyor. At the time, the tagging and cataloging of images for scientific analysis was time-consuming and, more importantly, not yet easily automated. In the first month of the experiment, 800 volunteers took part in processing the images, generating a large amount of useful data at minimal cost. More information about the experiment can be found here.
[2] Daniel R., Katrina S. (2008) A Life More Photographic; Mapping The Networked Image. Photographies vol. 1 (1), p. 9-28. https://www.researchgate.net/publication/299854265_A_Life_More_Photographic_Mapping_The_Networked_Image
Featured image: Hito Steyerl, Mechanical Kurds, 2025, © Hito Steyerl Videostills, © Hito Steyerl, VG Bild-Kunst, Bonn 2025
