Why do Robots Need Pets?

In Stray‘s final moments, natural light floods a world otherwise lit by neon signs, streetlights, and sterile, carceral fluorescents.1 The companions, a robot species evolved over centuries into surrogates for humanity, adopting everything from procreation to shopping sprees, are overjoyed. The Zurk, a parasitic waste-disposal solution, are annihilated. As in most video games, the good guys win. But in this case, the good guys are only marginally human. The sky door is opened by a cat with a drone living on its back. The drone houses the consciousness of a human scientist, who bears the enviable privilege of having been uploaded to the cloud. As media theorist Jason Wallin posits, the notion of animal surrogacy pervades discourses of animal avatars in media–from video games, to movies, to the looping, pet-centric attractions of TikTok.2 Simply put, as animal species disappear due to the effects of man-made climate change, they reemerge as vehicles for human consciousness. Such an impulse is enabled by recent game engine development. The tools of game development are increasingly geared toward autonomous environments, whose realism is a function of machine learning. These tools, I will go on to suggest, echo the relationship between computation and animal physiology. The imbrication of human, animal, and machine in contemporary gaming is an aspiration to imagine entire surrogate environments. The world of the robot companions, figured as both a trace of humanity and a new environmental surrogacy, needs a pet to finally cultivate respect for nature–a respect that capitalist society, both in game and out, has failed to establish.

Stray’s penultimate cutscene.

The Environmental CPU

We traverse a vertical, urban landscape to locate traces of human life. Stray‘s cat, we find out, is a vessel for an artificial intelligence–which is in turn a vessel for the human scientist who developed the intelligence. Our cat navigates the game-world swaddled in a language processing human machine. While the game includes both typical gaming fare–running, combat, stealth missions–as well as peripheral feline pastimes–kneading, playing, sleeping–the most interesting function of Stray‘s playable character is the combinatory effect of Stray‘s animal, human, and machine language. Where the first chapter of the game is played through without language–only through the most catlike of mechanics–once the android B-12 is introduced, the gameplay turns toward decoding and tracing the disappearance of human life. B-12 makes up for the cat’s lack of language, allowing the player to read messages, unlock doors, and interact with non-player characters (NPCs). This meshing of animal and human machine is reminiscent of the broader, extra-diegetic relationship between animals and computational frameworks.

Unlike Stray‘s protagonist, real cats can have math without language.From an evolutionary perspective,” writes logician Helen De La Cruz, “stable properties of the environment that yield potentially useful information can exert selective pressures on nervous systems in animals…we can expect that modules that infer gravitational pull, color, or numericities occur in many species.”3 Here, the drone-cat imbrication within Stray‘s gameplay mirrors the ability of fauna to infer the properties of an environment through numerical reasoning. But the cat’s perception extends beyond the physiological capabilities of its real counterpart. The protagonist’s nervous system, imbricated with its drone companion, is tuned not only to the selective pressures of the environment, but also to those traces of human life stored in B-12’s hard-disk.

The numerical staging of digital objects in a game engine mirrors the evolutionary roots of numerical abilities. The math that lies behind processual images–images that function outside human perception–originates not in the disconnect between sight and computation, but instead in an adaptation of the natural environment. Through gameplay, three distinct perspectives merge: the machine, the feline, and the human–or more precisely, the trace of the human. This trace is present in your robot companion as an uploaded consciousness, as well as existent in the landscape through constant reminders of human disappearance. 

The words RIP Humans, scrawled on the wall of Stray’s first level.

Natural Data

The enmeshment between animal and machine remains latent in much of media studies, which instead centers on the way we perceive–or remain unable to perceive–the computation behind digital images. Here, digital images lead to an increased disconnection between the object and human perception–the surplus flows of binary code are beyond the image. Mathematics remains a black box closing us off from the image’s inner workings. Digital images, writes theorist Shane Denson, are “no longer tuned to (the temporality) of their human receivers.”4 The images Denson chooses as his examples are mechanistic, automatic, and lack legible continuity–images of the transforming robots in the endless Transformers series contorting into new forms through thousands of CGI nodes, functioning at a rate that cannot be intercepted by human cognition. For Denson, such images incite a novel shift in viewing subjects–what Denson calls “a transitional reconfiguration of affective embodiment itself.”5 He refers to these images as processual, insofar as their processes are outside the human spectator’s perception.

A few problems arise from this philosophy. For one thing, as theorist and media archaeologist Jacob Gaboury writes, while much of the history of computing is concerned with the problem of simulating images–a fact that new game development technologies certainly corroborate–”…the computer is not a visual medium.”6 The medium of the computer is computation, not graphics. Yet Gaboury rightfully identifies much of art history and film studies’ preoccupation with the digital image in a perspective-based paradigm where computation is only a proxy for the images it constructs. The application of computer graphics to visual representation masks a number of significant differences between the way visuality is “constructed by computer graphics and captured by a camera.”7 Furthermore, the imagistic preoccupations of these disciplines limit “investigations into the technical materiality of the algorithms that produce” Denson’s processual images. The traces of the animal present in computed visual constructs is one means of reuniting the study of digital media with its material origins–cognitive, retinal, biological, and physiological.

The second problem with the study of digital images as outside human perception stems from the moral dilemma raised by autonomous machines. As scholar Louise Amoore writes, “the ethical problem is thought to dwell in the opacity of the algorithm and in its inscrutability.”8 The solution to this inscrutability is traditionally “addressed through remedies of transparency and accountability,”9 but is undermined by the increasing autonomy of machines. Where “the scientist at the terminal was supposed to ground the output and assign answers,”10 as theorists Leif Weatherby and Brian Justie write, “we have plugged our machines into each other and let them take on systemic autonomy.”11 Amoore goes on to write that “the principal ethicopolitical problem does not arise from machines breaching the imagined limits of human control but emerges instead from a machine learning that generates new limits and thresholds of what it means to be human.”12 Indeed, the invisible calculations of rendering technologies do disrupt the human perception of what makes images digital, but the fragmentary and often non-linear evolution of such technology is far from centered around the perception of a human subject in the first place. There is nothing new about automation, or automatic images. “As living beings, we are based on the automatic behaviors of our cells. All those cells are not machines but devices producing processes of automatic repetition,” writes Bernard Stiegler in a 2015 interview.13 In light of this provocation, the basis for Denson’s processual images is then usefully figured as a biological one.

In Stray’s diegetic world, the statistical disintegration of animal diversity has come to pass, and humanity is one of the casualties. In his writing on animals and extinction in video games, Jason Wallin draws on recent biological studies of wildlife populations. He reports that the rate of species disappearance is at least 100 times greater than it was decades ago.14 For Wallin, such a disappearance leads to a profusion of animal simulation in video games and elsewhere. “The annexation of animal life by digital models is symptomatic of the Anthropocene imaginary,” he writes, “or in league with the ‘standardization’ of matter intimate to Anthropocene thought is the recreation of nature as an object of management and control.”15 Virtual animals constitute a replacement for the real-life, disappearing animal, suggesting a teleological formula: the fewer real animals reside on Earth, the more we wish to generate and take control of animal proxies. Stray‘s diegesis takes this surrogate logic to an extreme. The companions, former robot servants to a human population devastated by a pandemic, do not have facial expressions, so much as their screens display images of facial expressions. Words scrawled, scratched, and sprayed across Stray‘s environment reveal the story of organic life’s dissolution. The plants genetically cultivated by the companions remain the only vestige of the outside world. The only natural environments are surrogate.

The concept of animal surrogacy is a useful framework to consider recent advances in gaming technology, which facilitate more comprehensive, realistic environmental play. Where previous environmental rendering tools resided at the limit of both labor and computation, these technologies provide the game environments with autonomy, and therefore, more naturalistic surrogacy. Unreal Engine 5, and its even more recent iteration, 5.1, comes with a variety of tools game developers use to make virtual environments–and the non-human life that inhabits them–photorealistic and immersive. Nanite technology, used for environmental rendering, takes the trouble of constructing large numbers of assets, such as trees, buildings, and terrain, out of the hands of developers. Where billboards—simplified, 2-dimensional models that stand in for the numerous vertices of a digital object—once populated worlds viewed at a distance, saving computation time, now the game decides automatically which polygons are necessary to render. Nanite makes the game environment another observer, which, like the feline visual cortex, identifies which properties objects retain no matter the distance. While neural networks are not an explicit ancestor to game engine technology, a comparison between Nanite and neural networks’ physiological inspiration offers a means of figuring the imbrication of human, animal, and machine.

Comparison between Level of Detail (billboard) rendering and Nanite.

Synapses and neurons in the feline visual cortex make up the cellular framework of neural networks. In 1980, inspired by a landmark study in feline physiology, Kunihiko Fukushima developed a mechanism for image recognition he called the ‘Neocognitron’—an ancestor to today’s neural networks.16 The goal of the Neocognitron, Fukushima wrote, was to learn without a teacher.17 This learning process consisted of a hierarchy of simple “S-cells”—scanning the image data—and complex “C-cells”—consolidating individual traits from the S-cells into an output with its own unique, functioning behavior, like reflectivity, motion, and shape. Nanite’s predictive architecture renders the game’s assets more physically real–and more believable surrogates for natural environments. The physiological inspiration for the Neocognitron’s S- and C-cells enable cats to distinguish objects regardless of location or lighting conditions. This enmeshment between animal and machine permeates modalities of animal avatars, and by extension, surrogate environments.

Human Disappearance, Surrogate Environments, and Machinic Remains

In the game’s conclusion, the player’s AI-human intelligence hybrid sacrifices itself to open a door to the outside world. The twist is melodramatic, evoking the sense that it is too late to redeem humanity, and instead a more radical shift toward posthumanism must occur. B-12, by way of the scientist’s uploaded consciousness, is the last human on Earth. But the last human does not save humanity, so much as he saves its remains–reuniting a nonhuman, machine world with a natural one. Here, the animal world grieves the human one, rather than the other way around. In her study of game environmentalism, theorist Lauren Woolbright echoes object-oriented ontology, reminding us that “ecosystems, whether perfectly functioning or not, churn on without any concern whatsoever for us…we are neither alone nor special in this world.”18 While providing the happy ending that we’ve been working toward, level by level, memory by memory, the ending of Stray also displays a world that seems to churn by without us. The game’s final action involves walking your cat avatar out into the wilderness. The traces of the human have been reunited with sunlight. However, this reunion doesn’t signal an animistic harmony, instead, the idea of nature and its actual environment is cleaved apart. To gain the environmental surrogate is to lose the cat. At the end of the game, it’s the cat that is reunited with the rest of nature, rather than the human. Again, the game world seems to allegorize game development’s flow toward providing surrogates for animals, and finally, its flow toward providing surrogates for entire environments. 

The cat, rather than its human-machine companion, is reunited with nature. Stray

While the primary gameplay function of Stray is one of data and resource extraction, moving through the world to collect items and add to a memory bank, its optimism regarding the post-Anthropocene bears a correlation to an environmental turn in video game culture more generally. If the notion of surrogacy that Wallin demonstrates bears some truth–if indeed our affinity with animal images, tropes and avatars becomes more pronounced as the natural world is threatened–then it seems that our environmental surrogates also multiply. Alenda Chang writes that, “games remain largely untapped in terms of their potential to create meaningful interaction within artificially intelligent environments, to model ecological dynamics based on interdependence and limitation, and to allow players to explore manifold ecological futures—not all of them dystopian.”19 Likewise, Paolo Ruffino writes, “a future Earth without humans is, in other words, possible, and a video game that plays by itself helps us to imagine this unimaginable scenario.”20 Stray, even in its conventionality in terms of game mechanics–the traversal of a world ripe for extraction of material resources–exemplifies this kind of assistance, asking us to refigure the relationship between machine, human, and animal. 

If we are to take the possibilities Chang and Ruffino lay out for environmentally conscious games and gaming as real possibilities, tools available in Unreal Engine 5.1, such as Nanite, are harbingers of a move toward more natural–more ecological–surrogate environments. They fit with the properties of natural life Chang identifies, writing that “it matters whether the plant is generated as a growing form that can evolve, change through seasons, grow old, and even die, or is merely the result of blatant cookie-cutting.”21 These new technologies, then, can be just as useful in simulating the cycles of the natural world as they are in automating object detection for tools of war and surveillance. This possibility only complicates their relationship to the human and natural world. As both Stray and the technology it allegorizes suggests, such technologies do not simply exist for us, instead they find precedence around us. They will exist after we’re gone. Computers will become rusted out extensions to the environment, like the network of drains and pipes our cat, machine, and human protagonist navigates. It becomes imperative that we consider the non-human paradigm of such a mediascape: an animal technology that, in a play of light and dark, imagines human extinction through animal eyes.


  1. Stray. Annapurna Interactive and BlueTwelve Studios, 2022.
  2. Jason Wallen, “Game Preserves: Digital Animals at the Brink of the Post-Anthropocene,” Green Letters 26, no. 1 (January 1, 2022): 103, PDF.
  3. Helen De La Cruz, “Why Are Some Numerical Concepts More Successful than Others? An Evolutionary Perspective on the History of Number Concepts,” Evolution and Human Behavior, no. 27 (2006): 310, PDF.
  4. Shane Denson, “Crazy Cameras,” in Discorrelated Images (Durham, NC: Duke University Press, 2020), 22.
  5. Ibid, 23.
  6. Jacob Gaboury, Image Objects: An Archaeology of Computer Graphics (Cambridge, MA: MIT Press, 2021), 36, PDF.
  7. Ibid.
  8. Louise Amoore, Cloud Ethics: Algorithms and the Attributes of Ourselves and Others (Durham, NC: Duke University Press, 2020), 5, PDF.
  9. Ibid.
  10. Ibid.
  11. Leif Weatherby and Brian Justie, “Indexical AI,” Critical Inquiry, Winter 2022, 382, PDF.
  12. Amoore, Cloud Ethics, 65.
  13. Stiegler, Bernard, with Anaïs Nony. 2015. “Bernard Stiegler on Automatic Society.” Third Rail 5: 16–17.
  14. Wallin, “Game Preserves,” 103.
  15. Ibid.
  16. Weatherby and Justie, “Indexical AI”, 400; D.H. Hubel and T.N. Wiesel, “Receptive Fields, Binocular Interaction and Functional Architecture in the Cat’s Visual Cortex,” Journal of Physiology 160 (1962): 106, PDF.
  17. Kunihiko Fukushima, “Neocognitron: A Self-organizing Neural Network Model for a Mechanism of Pattern Recognition Unaffected by Shift in Position,” Biological Cybernetics 36 (1980): 196, PDF.
  18. Woolbright Lauren, “Game Design as Climate Change Activism,” Ecozones 8, no. 2 (2017): 89-90, PDF.
  19. Ibid, 16.
  20. Paolo Ruffino, “Nonhuman Games: Playing in the Post-Anthropocene,” in Death, Culture, and Leisure: Playing Dead, ed. M. Coward-Gibbs, Emerald Studies in Death and Culture (Bingley: Emerald Publishing, 2020), 16, PDF.
  21. Chang, Playing Nature, 165.

About The Author

Max Oginz is an MA Candidate at San Francisco State University. His written and filmed work interrogates the ecological and cinematic implications of emergent technology. His fiction has been published in Sleepingfish and Fanzine, and his documentary short, The Seismic Properties of Cloud Computing, has shown in numerous film festivals across the US and abroad.

Related Posts