Introduction

The Smartest Horse in the World t the end of the nineteenth century, Europe was captivated by a horse called Hans. “Clever Hans” was nothing less than a marvel: he could solve math problems, tell time, identify days on a calendar, differentiate musical tones, and spell out words and sentences. People flocked to watch the German stallion tap out answers to complex problems with his hoof and consistently arrive at the right answer. “What is two plus three?” Hans would diligently tap his hoof on the ground five times. “What day of the week is it?” The horse would then tap his hoof to indicate each letter on a purpose-built letter board and spell out the correct answer. Hans even mastered more complex questions, such as, “I have a number in mind. I subtract nine and have three as a remainder. What is the number?” By 1904, Clever Hans was an international celebrity, with the New York Times championing him as “Berlin’s Wonderful Horse; He Can Do Almost Everything but Talk.”1 Hans’s trainer, a retired math teacher named Wilhelm von Osten, had long been fascinated by animal intelligence. Von Osten had tried and failed to teach kittens and bear cubs cardinal numbers, but it wasn’t until he started working with his own horse that he had success. He first taught Hans to count by holding the animal’s leg, showing him a number, and then tapping on the hoof the correct number of times. Soon Hans responded by accurately tapping out simple sums. Next von Osten introduced a chalkboard with the alphabet spelled out, so Hans could tap a number for each letter on the board. After two years of training, von Osten was astounded by the animal’s strong grasp of advanced intellectual concepts. So he took Hans on the road as proof that animals could reason. Hans became the viral sensation of the belle époque. But many people were skeptical, and the German board of education launched an investigative commission to test Von Osten’s scientific claims. The Hans Commission was led by the psychologist and philosopher Carl Stumpf and his assistant Oskar Pfungst, and it included a circus manager, a retired schoolteacher, a zoologist, a veterinarian, and a cavalry officer. Yet after extensive questioning of Hans, both with his trainer present and without, the horse maintained his record of correct answers, and the commission could find no evidence of deception. As Pfungst later wrote, Hans performed in front of “thousands of spectators, horse-fanciers, trick-trainers of first rank, and not one of them during the course of many months’ observations are able to discover any kind of regular signal” between the questioner and the horse.2 The commission found that the methods Hans had been taught were more like “teaching children in elementary schools” than animal training and were “worthy of scientific examination.”3 But Strumpf and Pfungst still had doubts. One finding in particular troubled them: when the questioner did not know the answer or was standing far away, Hans rarely gave the correct answer. This led Pfungst and Strumpf to consider whether some sort of unintentional signal had been providing Hans with the answers.

As Pfungst would describe in his 1911 book, their intuition was right: the questioner’s posture, breathing, and facial expression would subtly change around the moment Hans reached the right answer, prompting Hans to stop there.4 Pfungst later tested this hypothesis on human subjects and confirmed his result. What fascinated him most about this discovery was that questioners were generally unaware that they were providing pointers to the horse. The solution to the Clever Hans riddle, Pfungst wrote, was the unconscious direction from the horse’s questioners.5 The horse was trained to produce the results his owner wanted to see, but audiences felt that this was not the extraordinary intelligence they had imagined. The story of Clever Hans is compelling from many angles: the relationship between desire, illusion, and action, the business of spectacles, how we anthropomorphize the nonhuman, how biases emerge, and the politics of intelligence. Hans inspired a term in psychology for a particular type of conceptual trap, the Clever Hans Effect or observerexpectancy effect, to describe the influence of experimenters’ unintentional cues on their subjects. The relationship between Hans and von Osten points to the complex mechanisms by which biases find their ways into systems and how people become entangled with the phenomena they study. The story of Hans is now used in machine learning as a cautionary reminder that you can’t always be sure of what a model has learned from the data it has been given.6 Even a system that appears to perform spectacularly in training can make terrible predictions when presented with novel data in the world. This opens a central question of this book: How is intelligence “made,” and what traps can that create? At first glance, the story of Clever Hans is a story of how one man constructed intelligence by training a horse to follow cues and emulate humanlike cognition. But at another level, we see that the practice of making intelligence was considerably broader. The endeavor required validation from multiple institutions, including academia, schools, science, the public, and the military. Then there was the market for von Osten and his remarkable horse—emotional and economic investments that drove the tours, the newspaper stories, and the lectures. Bureaucratic authorities were assembled to measure and test the horse’s abilities. A constellation of financial, cultural, and scientific interests had a part to play in the construction of Hans’s intelligence and a stake in whether it was truly remarkable. We can see two distinct mythologies at work. The first myth is that nonhuman systems (be it computers or horses) are analogues for human minds. This perspective assumes that with sufficient training, or enough resources, humanlike intelligence can be created from scratch, without addressing the fundamental ways in which humans are embodied, relational, and set within wider ecologies. The second myth is that intelligence is something that exists independently, as though it were natural and distinct from social, cultural, historical, and political forces. In fact, the concept of intelligence has done inordinate harm over centuries and has been used to justify relations of domination from slavery to eugenics.7 These mythologies are particularly strong in the field of artificial intelligence, where the belief that human intelligence can be formalized and reproduced by machines has been axiomatic since the mid-twentieth century. Just as Hans’s intelligence was considered to be like that of a human, fostered carefully like a child in elementary school, so AI systems have repeatedly been described as simple but humanlike forms of intelligence. In 1950, Alan Turing predicted that “at the end of the century the use of words and general educated opinion will have altered so much that one will be able to speak of machines thinking without expecting to be contradicted.”8 The mathematician John von Neumann claimed in 1958 that the human nervous system is “prima facie digital.”9 MIT professor Marvin Minsky once responded to the question of whether machines could think by saying, “Of course machines can think; we can think and we are ‘meat machines.’”10 But not everyone was convinced. Joseph Weizenbaum, early AI inventor and creator of the first chatbot program, known as ELIZA, believed that the idea of humans as mere information processing systems is far too simplistic a notion of intelligence and that it drove the “perverse grand fantasy” that AI scientists could create a machine that learns “as a child does.”11 This has been one of the core disputes in the history of artificial intelligence. In 1961, MIT hosted a landmark lecture series titled “Management and the Computer of the Future.” A stellar lineup of computer scientists participated, including Grace Hopper, J. C. R. Licklider, Marvin Minsky, Allen Newell, Herbert Simon, and Norbert Wiener, to discuss the rapid advances being made in digital computing. At its conclusion, John McCarthy boldly argued that the differences between human and machine tasks were illusory. There were simply some complicated human tasks that would take more time to be formalized and solved by machines.12 But philosophy professor Hubert Dreyfus argued back, concerned that the assembled engineers “do not even consider the possibility that the brain might process information in an entirely different way than a computer.”13 In his later work What Computers Can’t Do, Dreyfus pointed out that human intelligence and expertise rely heavily on many unconscious and subconscious processes, while computers require all processes and data to be explicit and formalized.14 As a result, less formal aspects of intelligence must be abstracted, eliminated, or approximated for computers, leaving them unable to process information about situations as humans do. Much in AI has changed since the 1960s, including a shift from symbolic systems to the more recent wave of hype about machine learning techniques. In many ways, the early fights over what AI can do have been forgotten and the skepticism has melted away. Since the mid-2000s, AI has rapidly expanded as a field in academia and as an industry. Now a small number of powerful technology corporations deploy AI systems at a planetary scale, and their systems are once again hailed as comparable or even superior to human intelligence. Yet the story of Clever Hans also reminds us how narrowly we consider or recognize intelligence. Hans was taught to mimic tasks within a very constrained range: add, subtract, and spell words. This reflects a limited perspective of what horses or humans can do. Hans was already performing remarkable feats of interspecies communication, public performance, and considerable patience, yet these were not recognized as intelligence. As author and engineer Ellen Ullman puts it, this belief that the mind is like a computer, and vice versa, has “infected decades of thinking in the computer and cognitive sciences,” creating a kind of original sin for the field.15 It is the ideology of Cartesian dualism in artificial intelligence: where AI is narrowly understood as disembodied intelligence, removed from any relation to the material world.

What Is AI? Neither Artificial nor Intelligent Let’s ask the deceptively simple question, What is artificial intelligence? If you ask someone in the street, they might mention Apple’s Siri, Amazon’s cloud service, Tesla’s cars, or Google’s search algorithm. If you ask experts in deep learning, they might give you a technical response about how neural nets are organized into dozens of layers that receive labeled data, are assigned weights and thresholds, and can classify data in ways that cannot yet be fully explained.16 In 1978, when discussing expert systems, Professor Donald Michie described AI as knowledge refining, where “a reliability and competence of codification can be produced which far surpasses the highest level that the unaided human expert has ever, perhaps even could ever, attain.”17 In one of the most popular textbooks on the subject, Stuart Russell and Peter Norvig state that AI is the attempt to understand and build intelligent entities. “Intelligence is concerned mainly with rational action,” they claim. “Ideally, an intelligent agent takes the best possible action in a situation.”18 Each way of defining artificial intelligence is doing work, setting a frame for how it will be understood, measured, valued, and governed. If AI is defined by consumer brands for corporate infrastructure, then marketing and advertising have predetermined the horizon. If AI systems are seen as more reliable or rational than any human expert, able to take the “best possible action,” then it suggests that they should be trusted to make high-stakes decisions in health, education, and criminal justice. When specific algorithmic techniques are the sole focus, it suggests that only continual technical progress matters, with no consideration of the computational cost of those approaches and their far-reaching impacts on a planet under strain. In contrast, in this book I argue that AI is neither artificial nor intelligent. Rather, artificial intelligence is both embodied and material, made from natural resources, fuel, human labor, infrastructures, logistics, histories, and classifications. AI systems are not autonomous, rational, or able to discern anything without extensive, computationally intensive training with large datasets or predefined rules and rewards. In fact, artificial intelligence as we know it depends entirely on a much wider set of political and social structures. And due to the capital required to build AI at scale and the ways of seeing that it optimizes AI systems are ultimately designed to serve existing dominant interests. In this sense, artificial intelligence is a registry of power. In this book we’ll explore how artificial intelligence is made, in the widest sense, and the economic, political, cultural, and historical forces that shape it. Once we connect AI within these broader structures and social systems, we can escape the notion that artificial intelligence is a purely technical domain. At a fundamental level, AI is technical and social practices, institutions and infrastructures, politics and culture. Computational reason and embodied work are deeply interlinked: AI systems both reflect and produce social relations and understandings of the world. It’s worth noting that the term “artificial intelligence” can create discomfort in the computer science community. The phrase has moved in and out of fashion over the decades and is used more in marketing than by researchers. “Machine learning” is more commonly used in the technical literature. Yet the nomenclature of AI is often embraced during funding application season, when venture capitalists come bearing checkbooks, or when researchers are seeking press attention for a new scientific result. As a result, the term is both used and rejected in ways that keep its meaning in flux. For my purposes, I use AI to talk about the massive industrial formation that includes politics, labor, culture, and capital. When I refer to machine learning, I’m speaking of a range of technical approaches (which are, in fact, social and infrastructural as well, although rarely spoken about as such). But there are significant reasons why the field has been focused so much on the technical—algorithmic breakthroughs, incremental product improvements, and greater convenience. The structures of power at the intersection of technology, capital, and governance are well served by this narrow, abstracted analysis. To understand how AI is fundamentally political, we need to go beyond neural nets and statistical pattern recognition to instead ask what is being optimized, and for whom, and who gets to decide. Then we can trace the implications of those choices.

Seeing AI Like an Atlas How can an atlas help us to understand how artificial intelligence is made? An atlas is an unusual type of book. It is a collection of disparate parts, with maps that vary in resolution from a satellite view of the planet to a zoomed-in detail of an archipelago. When you open an atlas, you may be seeking specific information about a particular place—or perhaps you are wandering, following your curiosity, and finding unexpected pathways and new perspectives. As historian of science Lorraine Daston observes, all scientific atlases seek to school the eye, to focus the observer’s attention on particular telling details and significant characteristics.19 An atlas presents you with a particular viewpoint of the world, with the imprimatur of science—scales and ratios, latitudes and longitudes—and a sense of form and consistency. Yet an atlas is as much an act of creativity—a subjective, political, and aesthetic intervention—as it is a scientific collection. The French philosopher Georges DidiHuberman thinks of the atlas as something that inhabits the aesthetic paradigm of the visual and the epistemic paradigm of knowledge. By implicating both, it undermines the idea that science and art are ever completely separate.20 Instead, an atlas offers us the possibility of rereading the world, linking disparate pieces differently and “reediting and piecing it together again without thinking we are summarizing or exhausting it.”21 Perhaps my favorite account of how a cartographic approach can be helpful comes from the physicist and technology critic Ursula Franklin: “Maps represent purposeful endeavors: they are meant to be useful, to assist the traveler and bridge the gap between the known and the as yet unknown; they are testaments of collective knowledge and insight.”22 Maps, at their best, offer us a compendium of open pathways—shared ways of knowing—that can be mixed and combined to make new interconnections. But there are also maps of domination, those national maps where territory is carved along the fault lines of power: from the direct interventions of drawing borders across contested spaces to revealing the colonial paths of empires. By invoking an atlas, I’m suggesting that we need new ways to understand the empires of artificial intelligence. We need a theory of AI that accounts for the states and corporations that drive and dominate it, the extractive mining that leaves an imprint on the planet, the mass capture of data, and the profoundly unequal and increasingly exploitative labor practices that sustain it. These are the shifting tectonics of power in AI. A topographical approach offers different perspectives and scales, beyond the abstract promises of artificial intelligence or the latest machine learning models. The aim is to understand AI in a wider context by walking through the many different landscapes of computation and seeing how they connect.23 There’s another way in which atlases are relevant here. The field of AI is explicitly attempting to capture the planet in a computationally legible form. This is not a metaphor so much as the industry’s direct ambition. The AI industry is making and normalizing its own proprietary maps, as a centralized God’s-eye view of human movement, communication, and labor. Some AI scientists have stated their desire to capture the world and to supersede other forms of knowing. AI professor Fei-Fei Li describes her ImageNet project as aiming to “map out the entire world of objects.”24 In their textbook, Russell and Norvig describe artificial intelligence as “relevant to any intellectual task; it is truly a universal field.”25 One of the founders of artificial intelligence and early experimenter in facial recognition, Woody Bledsoe, put it most bluntly: “in the long run, AI is the only science.”26 This is a desire not to create an atlas of the world but to be the atlas—the dominant way of seeing. This colonizing impulse centralizes power in the AI field: it determines how the world is measured and defined while simultaneously denying that this is an inherently political activity. Instead of claiming universality, this book is a partial account, and by bringing you along on my investigations, I hope to show you how my views were formed. We will encounter well-visited and lesser-known landscapes of computation: the pits of mines, the long corridors of energy-devouring data centers, skull archives, image databases, and the fluorescent-lit hangars of delivery warehouses. These sites are included not just to illustrate the material construction of AI and its ideologies but also to “illuminate the unavoidably subjective and political aspects of mapping, and to provide alternatives to hegemonic, authoritative—and often naturalized and reified—approaches,” as media scholar Shannon Mattern writes.27 Models for understanding and holding systems accountable have long rested on ideals of transparency. As I’ve written with the media scholar Mike Ananny, being able to see a system is sometimes equated with being able to know how it works and how to govern it.28 But this tendency has serious limitations. In the case of AI, there is no singular black box to open, no secret to expose, but a multitude of interlaced systems of power. Complete transparency, then, is an impossible goal. Rather, we gain a better understanding of AI’s role in the world by engaging with its material architectures, contextual environments, and prevailing politics and by tracing how they are connected. My thinking in this book has been informed by the disciplines of science and technology studies, law, and political philosophy and from my experience working in both academia and an industrial AI research lab for almost a decade. Over those years, many generous colleagues and communities have changed the way I see the world: mapping is always a collective exercise, and this is no exception.29 I’m grateful to the scholars who created new ways to understand sociotechnical systems, including Geoffrey Bowker, Benjamin Bratton, Wendy Chun, Lorraine Daston, Peter Galison, Ian Hacking, Stuart Hall, Donald MacKenzie, Achille Mbembé, Alondra Nelson, Susan Leigh Star, and Lucy Suchman, among many others. This book benefited from many inperson conversations and reading the recent work by authors studying the politics of technology, including Mark Andrejevic, Ruha Benjamin, Meredith Broussard, Simone Browne, Julie Cohen, Sasha Costanza-Chock, Virginia Eubanks, Tarleton Gillespie, Mar Hicks, Tung-Hui Hu, Yuk Hui, Safiya Umoja Noble, and Astra Taylor. As with any book, this one emerges from a specific lived experience that imposes limitations. As someone who has lived and worked in the United States for the past decade, my focus skews toward the AI industry in Western centers of power. But my aim is not to create a complete global atlas—the very idea invokes capture and colonial control. Instead, any author’s view can be only partial, based on local observations and interpretations, in what environmental geographer Samantha Saville calls a “humble geography” that acknowledges one’s specific perspectives rather than claiming objectivity or mastery.30 Just as there are many ways to make an atlas, so there are many possible futures for how AI will be used in the world. The expanding reach of AI systems may seem inevitable, but this is contestable and incomplete. The underlying visions of the AI field do not come into being autonomously but instead have been constructed from a particular set of beliefs and perspectives. The chief designers of the contemporary atlas of AI are a small and homogenous group of people, based in a handful of cities, working in an industry that is currently the wealthiest in the world. Like medieval European mappae mundi, which illustrated religious and classical concepts as much as coordinates, the maps made by the AI industry are political interventions, as opposed to neutral reflections of the world. This book is made against the spirit of colonial mapping logics, and it embraces different stories, locations, and knowledge bases to better understand the role of AI in the world.

Topographies of Computation How, at this moment in the twenty-first century, is AI conceptualized and constructed? What is at stake in the turn to artificial intelligence, and what kinds of politics are contained in the way these systems map and interpret the world? What are the social and material consequences of including AI and related algorithmic systems into the decision-making systems of social institutions like education and health care, finance, government operations, workplace interactions and hiring, communication systems, and the justice system? This book is not a story about code and algorithms or the latest thinking in computer vision or natural language processing or reinforcement learning. Many other books do that. Neither is it an ethnographic account of a single community and the effects of AI on their experience of work or housing or medicine—although we certainly need more of those. Instead, this is an expanded view of artificial intelligence as an extractive industry. The creation of contemporary AI systems depends on exploiting energy and mineral resources from the planet, cheap labor, and data at scale. To observe this in action, we will go on a series of journeys to places that reveal the makings of AI. In chapter 1, we begin in the lithium mines of Nevada, one of the many sites of mineral extraction needed to power contemporary computation. Mining is where we see the extractive politics of AI at their most literal. The tech sector’s demand for rare earth minerals, oil, and coal is vast, but the true costs of this extraction is never borne by the industry itself. On the software side, building models for natural language processing and computer vision is enormously energy hungry, and the competition to produce faster and more efficient models has driven computationally greedy methods that expand AI’s carbon footprint. From the last trees in Malaysia that were harvested to produce latex for the first transatlantic undersea cables to the giant artificial lake of toxic residues in Inner Mongolia, we trace the environmental and human birthplaces of planetary computation networks and see how they continue to terraform the planet. Chapter 2 shows how artificial intelligence is made of human labor. We look at the digital pieceworkers paid pennies on the dollar clicking on microtasks so that data systems can seem more intelligent than they are.31 Our journey will take us inside the Amazon warehouses where employees must keep in time with the algorithmic cadences of a vast logistical empire, and we will visit the Chicago meat laborers on the disassembly lines where animal carcasses are vivisected and prepared for consumption. And we’ll hear from the workers who are protesting against the way that AI systems are increasing surveillance and control for their bosses. Labor is also a story about time. Coordinating the actions of humans with the repetitive motions of robots and line machinery has always involved a controlling of bodies in space and time.32 From the invention of the stopwatch to Google’s TrueTime, the process of time coordination is at the heart of workplace management. AI technologies both require and create the conditions for ever more granular and precise mechanisms of temporal management. Coordinating time demands increasingly detailed information about what people are doing and how and when they do it. Chapter 3 focuses on the role of data. All publicly accessible digital material— including data that is personal or potentially damaging—is open to being harvested for training datasets that are used to produce AI models. There are gigantic datasets full of people’s selfies, of hand gestures, of people driving cars, of babies crying, of newsgroup conversations from the 1990s, all to improve algorithms that perform such functions as facial recognition, language prediction, and object detection. When these collections of data are no longer seen as people’s personal material but merely as infrastructure, the specific meaning or context of an image or a video is assumed to be irrelevant. Beyond the serious issues of privacy and ongoing surveillance capitalism, the current practices of working with data in AI raise profound ethical, methodological, and epistemological concerns.33 And how is all this data used? In chapter 4, we look at the practices of classification in artificial intelligence systems, what sociologist Karin Knorr Cetina calls the “epistemic machinery.”34 We see how contemporary systems use labels to predict human identity, commonly using binary gender, essentialized racial categories, and problematic assessments of character and credit worthiness. A sign will stand in for a system, a proxy will stand for the real, and a toy model will be asked to substitute for the infinite complexity of human subjectivity. By looking at how classifications are made, we see how technical schemas enforce hierarchies and magnify inequity. Machine learning presents us with a regime of normative reasoning that, when in the ascendant, takes shape as a powerful governing rationality. From here, we travel to the hill towns of Papua New Guinea to explore the history of affect recognition, the idea that facial expressions hold the key to revealing a person’s inner emotional state. Chapter 5 considers the claim of the psychologist Paul Ekman that there are a small set of universal emotional states which can be read directly from the face. Tech companies are now deploying this idea in affect recognition systems, as part of an industry predicted to be worth more than seventeen billion dollars.35 But there is considerable scientific controversy around emotion detection, which is at best incomplete and at worst misleading. Despite the unstable premise, these tools are being rapidly implemented into hiring, education, and policing systems. In chapter 6 we look at the ways in which AI systems are used as a tool of state power. The military past and present of artificial intelligence have shaped the practices of surveillance, data extraction, and risk assessment we see today. The deep interconnections between the tech sector and the military are now being reined in to fit a strong nationalist agenda. Meanwhile, extralegal tools used by the intelligence community have now dispersed, moving from the military world into the commercial technology sector, to be used in classrooms, police stations, workplaces, and unemployment offices. The military logics that have shaped AI systems are now part of the workings of municipal government, and they are further skewing the relation between states and subjects. The concluding chapter assesses how artificial intelligence functions as a structure of power that combines infrastructure, capital, and labor. From the Uber driver being nudged to the undocumented immigrant being tracked to the public housing tenants contending with facial recognition systems in their homes, AI systems are built with the logics of capital, policing, and militarization—and this combination further widens the existing asymmetries of power. These ways of seeing depend on the twin moves of abstraction and extraction: abstracting away the material conditions of their making while extracting more information and resources from those least able to resist. But these logics can be challenged, just as systems that perpetuate oppression can be rejected. As conditions on Earth change, calls for data protection, labor rights, climate justice, and racial equity should be heard together. When these interconnected movements for justice inform how we understand artificial intelligence, different conceptions of planetary politics become possible.

Extraction, Power, and Politics Artificial intelligence, then, is an idea, an infrastructure, an industry, a form of exercising power, and a way of seeing; it’s also a manifestation of highly organized capital backed by vast systems of extraction and logistics, with supply chains that wrap around the entire planet. All these things are part of what artificial intelligence is—a two-word phrase onto which is mapped a complex set of expectations, ideologies, desires, and fears. AI can seem like a spectral force—as disembodied computation—but these systems are anything but abstract. They are physical infrastructures that are reshaping the Earth, while simultaneously shifting how the world is seen and understood. It’s important for us to contend with these many aspects of artificial intelligence—its malleability, its messiness, and its spatial and temporal reach. The promiscuity of AI as a term, its openness to being reconfigured, also means that it can be put to use in a range of ways: it can refer to everything from consumer devices like the Amazon Echo to nameless back-end processing systems, from narrow technical papers to the biggest industrial companies in the world. But this has its usefulness, too. The breadth of the term “artificial intelligence” gives us license to consider all these elements and how they are deeply imbricated: from the politics of intelligence to the mass harvesting of data; from the industrial concentration of the tech sector to geopolitical military power; from the deracinated environment to ongoing forms of discrimination. The task is to remain sensitive to the terrain and to watch the shifting and plastic meanings of the term “artificial intelligence”—like a container into which various things are placed and then removed—because that, too, is part of the story. Simply put, artificial intelligence is now a player in the shaping of knowledge, communication, and power. These reconfigurations are occurring at the level of epistemology, principles of justice, social organization, political expression, culture, understandings of human bodies, subjectivities, and identities: what we are and what we can be. But we can go further. Artificial intelligence, in the process of remapping and intervening in the world, is politics by other means—although rarely acknowledged as such. These politics are driven by the Great Houses of AI, which consist of the halfdozen or so companies that dominate large-scale planetary computation. Many social institutions are now influenced by these tools and methods, which shape what they value and how decisions are made while creating a complex series of downstream effects. The intensification of technocratic power has been under way for a long time, but the process has now accelerated. In part this is due to the concentration of industrial capital at a time of economic austerity and outsourcing, including the defunding of social welfare systems and institutions that once acted as a check on market power. This is why we must contend with AI as a political, economic, cultural, and scientific force. As Alondra Nelson, Thuy Linh Tu, and Alicia Headlam Hines observe, “Contests around technology are always linked to larger struggles for economic mobility, political maneuvering, and community building.”36 We are at a critical juncture, one that requires us to ask hard questions about the way AI is produced and adopted. We need to ask: What is AI? What forms of politics does it propagate? Whose interests does it serve, and who bears the greatest risk of harm? And where should the use of AI be constrained? These questions will not have easy answers. But neither is this an irresolvable situation or a point of no return—dystopian forms of thinking can paralyze us from taking action and prevent urgently needed interventions.37 As Ursula Franklin writes, “The viability of technology, like democracy, depends in the end on the practice of justice and on the enforcement of limits to power.”38 This book argues that addressing the foundational problems of AI and planetary computation requires connecting issues of power and justice: from epistemology to labor rights, resource extraction to data protections, racial inequity to climate change. To do that, we need to expand our understanding of what is under way in the empires of AI, to see what is at stake, and to make better collective decisions about what should come next.

Conclusion – Power
Artificial intelligence is not an objective, universal, or neutral computational technique that makes determinations without human direction. Its systems are embedded in social, political, cultural, and economic worlds, shaped by humans, institutions, and imperatives that determine what they do and how they do it. They are designed to discriminate, to amplify hierarchies, and to encode narrow classifications. When applied in social contexts such as policing, the court system, health care, and education, they can reproduce, optimize, and amplify existing structural inequalities. This is no accident: AI systems are built to see and intervene in the world in ways that primarily benefit the states, institutions, and corporations that they serve. In this sense, AI systems are expressions of power that emerge from wider economic and political forces, created to increase profits and centralize control for those who wield them. But this is not how the story of artificial intelligence is typically told. The standard accounts of AI often center on a kind of algorithmic exceptionalism— the idea that because AI systems can perform uncanny feats of computation, they must be smarter and more objective than their flawed human creators. Consider this diagram of AlphaGo Zero, an AI program designed by Google’s DeepMind to play strategy games.1 The image shows how it “learned” to play the Chinese strategy game Go by evaluating more than a thousand options per move. In the paper announcing this development, the authors write: “Starting tabula rasa, our new program AlphaGo Zero achieved superhuman performance.”2 DeepMind cofounder Demis Hassabis has described these game engines as akin to an alien intelligence. “It doesn’t play like a human, but it also doesn’t play like computer engines. It plays in a third, almost alien, way. . . . It’s like chess from another dimension.”3 When the next iteration mastered Go within three days, Hassabis described it as “rediscovering three thousand years of human knowledge in 72 hours!”4 The Go diagram shows no machines, no human workers, no capital investment, no carbon footprint, just an abstract rules-based system endowed with otherworldly skills. Narratives of magic and mystification recur throughout AI’s history, drawing bright circles around spectacular displays of speed, efficiency, and computational reasoning.5 It’s no coincidence that one of the iconic examples of contemporary AI is a game.

Games without Frontiers Games have been a preferred testing ground for AI programs since the 1950s.6 Unlike everyday life, games offer a closed world with defined parameters and clear victory conditions. The historical roots of AI in World War II stemmed from military-funded research in signal processing and optimization that sought to simplify the world, rendering it more like a strategy game. A strong emphasis on rationalization and prediction emerged, along with a faith that mathematical formalisms would help us understand humans and society.7 The belief that accurate prediction is fundamentally about reducing the complexity of the world gave rise to an implicit theory of the social: find the signal in the noise and make order from disorder. This epistemological flattening of complexity into clean signal for the purposes of prediction is now a central logic of machine learning. The historian of technology Alex Campolo and I call this enchanted determinism: AI systems are seen as enchanted, beyond the known world, yet deterministic in that they discover patterns that can be applied with predictive certainty to everyday life.8 In discussions of deep learning systems, where machine learning techniques are extended by layering abstract representations of data on top of each other, enchanted determinism acquires an almost theological quality. That deep learning approaches are often uninterpretable, even to the engineers who created them, gives these systems an aura of being too complex to regulate and too powerful to refuse. As the social anthropologist F. G. Bailey observed, the technique of “obscuring by mystification” is often employed in public settings to argue for a phenomenon’s inevitability.9 We are told to focus on the innovative nature of the method rather than on what is primary: the purpose of the thing itself. Above all, enchanted determinism obscures power and closes off informed public discussion, critical scrutiny, or outright rejection. Enchanted determinism has two dominant strands, each a mirror image of the other. One is a form of tech utopianism that offers computational interventions as universal solutions applicable to any problem. The other is a tech dystopian perspective that blames algorithms for their negative outcomes as though they are independent agents, without contending with the contexts that shape them and in which they operate. At an extreme, the tech dystopian narrative ends in the singularity, or superintelligence—the theory that a machine intelligence could emerge that will ultimately dominate or destroy humans.10 This view rarely contends with the reality that so many people around the world are already dominated by systems of extractive planetary computation. These dystopian and utopian discourses are metaphysical twins: one places its faith in AI as a solution to every problem, while the other fears AI as the greatest peril. Each offers a profoundly ahistorical view that locates power solely within technology itself. Whether AI is abstracted as an all-purpose tool or an all-powerful overlord, the result is technological determinism. AI takes the central position in society’s redemption or ruin, permitting us to ignore the systemic forces of unfettered neoliberalism, austerity politics, racial inequality, and widespread labor exploitation. Both the tech utopians and dystopians frame the problem with technology always at the center, inevitably expanding into every part of life, decoupled from the forms of power that it magnifies and serves. When AlphaGo defeats a human grandmaster, it’s tempting to imagine that some kind of otherworldly intelligence has arrived. But there’s a far simpler and more accurate explanation. AI game engines are designed to play millions of games, run statistical analyses to optimize for winning outcomes, and then play millions more. These programs produce surprising moves uncommon in human games for a straightforward reason: they can play and analyze far more games at a far greater speed than any human can. This is not magic; it is statistical analysis at scale. Yet the tales of preternatural machine intelligence persist.11 Over and over, we see the ideology of Cartesian dualism in AI: the fantasy that AI systems are disembodied brains that absorb and produce knowledge independently from their creators, infrastructures, and the world at large. These illusions distract from the far more relevant questions: Whom do these systems serve? What are the political economies of their construction? And what are the wider planetary consequences?

The Pipelines of AI Consider a different illustration of AI: the blueprint for Google’s first owned and operated data center, in The Dalles, Oregon. It depicts three 68,680-square-foot buildings, an enormous facility that was estimated in 2008 to use enough energy to power eighty-two thousand homes, or a city the size of Tacoma, Washington.12 The data center now spreads along the shores of the Columbia River, where it draws heavily on some of the cheapest electricity in North America. Google’s lobbyists negotiated for six months with local officials to get a deal that included tax exemptions, guarantees of cheap energy, and use of the city-built fiber-optic ring. Unlike the abstract vision of a Go game, the engineering plan reveals how much of Google’s technical vision depends on public utilities, including gas mains, sewer pipes, and the high-voltage lines through which the discount electricity would flow. In the words of the writer Ginger Strand, “Through city infrastructure, state givebacks, and federally subsidized power, YouTube is bankrolled by us. The blueprint reminds us of how much the artificial intelligence industry’s expansion has been publicly subsidized: from defense funding and federal research agencies to public utilities and tax breaks to the data and unpaid labor taken from all who use search engines or post images online. AI began as a major public project of the twentieth century and was relentlessly privatized to produce enormous financial gains for the tiny minority at the top of the extraction pyramid. These diagrams present two different ways of understanding how AI works. I’ve argued that there is much at stake in how we define AI, what its boundaries are, and who determines them: it shapes what can be seen and contested. The Go diagram speaks to the industry narratives of an abstract computational cloud, far removed from the earthly resources needed to produce it, a paradigm where technical innovation is lionized, regulation is rejected, and true costs are never revealed. The blueprint points us to the physical infrastructure, but it leaves out the full environmental implications and the political deals that made it possible. These partial accounts of AI represent what philosophers Michael Hardt and Antonio Negri call the “dual operation of abstraction and extraction” in information capitalism: abstracting away the material conditions of production while extracting more information and resources.14 The description of AI as fundamentally abstract distances it from the energy, labor, and capital needed to produce it and the many different kinds of mining that enable it. This book has explored the planetary infrastructure of AI as an extractive industry: from its material genesis to the political economy of its operations to the discourses that support its aura of immateriality and inevitability. We have seen the politics inherent in how AI systems are trained to recognize the world. And we’ve observed the systemic forms of inequity that make AI what it is today. The core issue is the deep entanglement of technology, capital, and power, of which AI is the latest manifestation. Rather than being inscrutable and alien, these systems are products of larger social and economic structures with profound material consequences.

The Map Is Not the Territory How do we see the full life cycle of artificial intelligence and the dynamics of power that drive it? We have to go beyond the conventional maps of AI to locate it in a wider landscape. Atlases can provoke a shift in scale, to see how spaces are joined in relation to one another. This book proposes that the real stakes of AI are the global interconnected systems of extraction and power, not the technocratic imaginaries of artificiality, abstraction, and automation. To understand AI for what it is, we need to see the structures of power it serves. AI is born from salt lakes in Bolivia and mines in Congo, constructed from crowdworker-labeled datasets that seek to classify human actions, emotions, and identities. It is used to navigate drones over Yemen, direct immigration police in the United States, and modulate credit scores of human value and risk across the world. A wide-angle, multiscalar perspective on AI is needed to contend with these overlapping regimes. This book began below the ground, where the extractive politics of artificial intelligence can be seen at their most literal. Rare earth minerals, water, coal, and oil: the tech sector carves out the earth to fuel its highly energy-intensive infrastructures. AI’s carbon footprint is never fully admitted or accounted for by the tech sector, which is simultaneously expanding the networks of data centers while helping the oil and gas industry locate and strip remaining reserves of fossil fuels. The opacity of the larger supply chain for computation in general, and AI in particular, is part of a longestablished business model of extracting value from the commons and avoiding restitution for the lasting damage. Labor represents another form of extraction. In chapter 2, we ventured beyond the highly paid machine learning engineers to consider the other forms of work needed to make artificial intelligence systems function. From the miners extracting tin in Indonesia to crowdworkers in India completing tasks on Amazon Mechanical Turk to iPhone factory workers at Foxconn in China, the labor force of AI is far greater than we normally imagine. Even within the tech companies there is a large shadow workforce of contract laborers, who significantly outnumber full-time employees but have fewer benefits and no job security.15 In the logistical nodes of the tech sector, we find humans completing the tasks that machines cannot. Thousands of people are needed to support the illusion of automation: tagging, correcting, evaluating, and editing AI systems to make them appear seamless. Others lift packages, drive for ride-hailing apps, and deliver food. AI systems surveil them all while squeezing the most output from the bare functionality of human bodies: the complex joints of fingers, eyes, and knee sockets are cheaper and easier to acquire than robots. In those spaces, the future of work looks more like the Taylorist factories of the past, but with wristbands that vibrate when workers make errors and penalties given for taking too many bathroom breaks. The uses of workplace AI further skew power imbalances by placing more control in employers’ hands. Apps are used to track workers, nudge them to work longer hours, and rank them in real time. Amazon provides a canonical example of how a microphysics of power—disciplining bodies and their movement through space—is connected to a macrophysics of power, a logistics of planetary time and information. AI systems exploit differences in time and wages across markets to speed the circuits of capital. Suddenly, everyone in urban centers can have—and expects—same day delivery. And the system speeds up again, with the material consequences hidden behind the cardboard boxes, delivery trucks, and “buy now” buttons. At the data layer, we can see a different geography of extraction. “We are building a mirror of the real world,” a Google Street View engineer said in 2012. “Anything that you see in the real world needs to be in our databases.”16 Since then, the harvesting of the real world has only intensified to reach into spaces that were previously hard to capture. As we saw in chapter 3, there has been a widespread pillaging of public spaces; the faces of people in the street have been captured to train facial recognition systems; social media feeds have been ingested to build predictive models of language; sites where people keep personal photos or have online debates have been scraped in order to train machine vision and natural language algorithms. This practice has become so common that few in the AI field even question it. In part, that is because so many careers and market valuations depend on it. The collect-it-all mentality, once the remit of intelligence agencies, is not only normalized but moralized—it is seen as wasteful not to collect data wherever possible.17 Once data is extracted and ordered into training sets, it becomes the epistemic foundation by which AI systems classify the world. From the benchmark training sets such as ImageNet, MS-Celeb, or NIST’s collections, images are used to represent ideas that are far more relational and contested than the labels may suggest. In chapter 4, we saw how labeling taxonomies allocate people into forced gender binaries, simplistic and offensive racial groupings, and highly normative and stereotypical analyses of character, merit, and emotional state. These classifications, unavoidably value-laden, force a way of seeing onto the world while claiming scientific neutrality. Datasets in AI are never raw materials to feed algorithms: they are inherently political interventions. The entire practice of harvesting data, categorizing and labeling it, and then using it to train systems is a form of politics. It has brought a shift to what are called operational images—representations of the world made solely for machines.18 Bias is a symptom of a deeper affliction: a far-ranging and centralizing normative logic that is used to determine how the world should be seen and evaluated. A central example of this is affect detection, described in chapter 5, which draws on controversial ideas about the relation of faces to emotions and applies them with the reductive logic of a lie detector test. The science remains deeply contested.19 Institutions have always classified people into identity categories, narrowing personhood and cutting it down into precisely measured boxes. Machine learning allows that to happen at scale. From the hill towns of Papua New Guinea to military labs in Maryland, techniques have been developed to reduce the messiness of feelings, interior states, preferences, and identifications into something quantitative, detectable, and trackable. What epistemological violence is necessary to make the world readable to a machine learning system? AI seeks to systematize the unsystematizable, formalize the social, and convert an infinitely complex and changing universe into a Linnaean order of machinereadable tables. Many of AI’s achievements have depended on boiling things down to a terse set of formalisms based on proxies: identifying and naming some features while ignoring or obscuring countless others. To adapt a phrase from philosopher Babette Babich, machine learning exploits what it does know to predict what it does not know: a game of repeated approximations. Datasets are also proxies—stand-ins for what they claim to measure. Put simply, this is transmuting difference into computable sameness. This kind of knowledge schema recalls what Friedrich Nietzsche described as “the falsifying of the multifarious and incalculable into the identical, similar, and calculable.”20 AI systems become deterministic when these proxies are taken as ground truth, when fixed labels are applied to a fluid complexity. We saw this in the cases where AI is used to predict gender, race, or sexuality from a photograph of a face.21 These approaches resemble phrenology and physiognomy in their desire to essentialize and impose identities based on external appearances. The problem of ground truth for AI systems is heightened in the context of state power, as we saw in chapter 6. The intelligence agencies led the way on the mass collection of data, where metadata signatures are sufficient for lethal drone strikes and a cell phone location becomes a proxy for an unknown target. Even here, the bloodless language of metadata and surgical strikes is directly contradicted by the unintended killings from drone missiles.22 As Lucy Suchman has asked, how are “objects” identified as imminent threats? We know that “ISIS pickup truck” is a category based on hand-labeled data, but who chose the categories and identified the vehicles?23 We saw the epistemological confusions and errors of object recognition training sets like ImageNet; military AI systems and drone attacks are built on the same unstable terrain. The deep interconnections between the tech sector and the military are now framed within a strong nationalist agenda. The rhetoric about the AI war between the United States and China drives the interests of the largest tech companies to operate with greater government support and few restrictions. Meanwhile, the surveillance armory used by agencies like the NSA and the CIA is now deployed domestically at a municipal level in the in-between space of commercial-military contracting by companies like Palantir. Undocumented immigrants are hunted down with logistical systems of total information control and capture that were once reserved for extralegal espionage. Welfare decision-making systems are used to track anomalous data patterns in order to cut people off from unemployment benefits and accuse them of fraud. License plate reader technology is being used by home surveillance systems—a widespread integration of previously separate surveillance networks.24 The result is a profound and rapid expansion of surveillance and a blurring between private contractors, law enforcement, and the tech sector, fueled by kickbacks and secret deals. It is a radical redrawing of civic life, where the centers of power are strengthened by tools that see with the logics of capital, policing, and militarization.

Toward Connected Movements for Justice If AI currently serves the existing structures of power, an obvious question might be: Should we not seek to democratize it? Could there not be an AI for the people that is reoriented toward justice and equality rather than industrial extraction and discrimination? This may seem appealing, but as we have seen throughout this book, the infrastructures and forms of power that enable and are enabled by AI skew strongly toward the centralization of control. To suggest that we democratize AI to reduce asymmetries of power is a little like arguing for democratizing weapons manufacturing in the service of peace. As Audre Lorde reminds us, the master’s tools will never dismantle the master’s house.25 A reckoning is due for the technology sector. To date, one common industry response has been to sign AI ethics principles. As European Union parliamentarian Marietje Schaake observed, in 2019 there were 128 frameworks for AI ethics in Europe alone.26 These documents are often presented as products of a “wider consensus” on AI ethics. But they are overwhelmingly produced by economically developed countries, with little representation from Africa, South and Central America, or Central Asia. The voices of the people most harmed by AI systems are largely missing from the processes that produce them.27 Further, ethical principles and statements don’t discuss how they should be implemented, and they are rarely enforceable or accountable to a broader public. As Shannon Mattern has noted, the focus is more commonly on the ethical ends for AI, without assessing the ethical means of its application.28 Unlike medicine or law, AI has no formal professional governance structure or norms—no agreed-upon definitions and goals for the field or standard protocols for enforcing ethical practice.29 Self-regulating ethical frameworks allow companies to choose how to deploy technologies and, by extension, to decide what ethical AI means for the rest of the world.30 Tech companies rarely suffer serious financial penalties when their AI systems violate the law and even fewer consequences when their ethical principles are violated. Further, public companies are pressured by shareholders to maximize return on investment over ethical concerns, commonly making ethics secondary to profits. As a result, ethics is necessary but not sufficient to address the fundamental concerns raised in this book. To understand what is at stake, we must focus less on ethics and more on power. AI is invariably designed to amplify and reproduce the forms of power it has been deployed to optimize. Countering that requires centering the interests of the communities most affected.31 Instead of glorifying company founders, venture capitalists, and technical visionaries, we should begin with the lived experiences of those who are disempowered, discriminated against, and harmed by AI systems. When someone says, “AI ethics,” we should assess the labor conditions for miners, contractors, and crowdworkers. When we hear “optimization,” we should ask if these are tools for the inhumane treatment of immigrants. When there is applause for “large-scale automation,” we should remember the resulting carbon footprint at a time when the planet is already under extreme stress. What would it mean to work toward justice across all these systems? In 1986, the political theorist Langdon Winner described a society “committed to making artificial realities” with no concern for the harms it could bring to the conditions of life: “Vast transformations in the structure of our common world have been undertaken with little attention to what those alterations mean. . . . In the technical realm we repeatedly enter into a series of social contracts, the terms of which are only revealed after signing.”32 In the four decades since, those transformations are now at a scale that has shifted the chemical composition of the atmosphere, the temperature of Earth’s surface, and the contents of the planet’s crust. The gap between how technology is judged on its release and its lasting consequences has only widened. The social contract, to the extent that there ever was one, has brought a climate crisis, soaring wealth inequality, racial discrimination, and widespread surveillance and labor exploitation. But the idea that these transformations occurred in ignorance of their possible results is part of the problem. The philosopher Achille Mbembé sharply critiques the idea that we could not have foreseen what would become of the knowledge systems of the twenty-first century, as they were always “operations of abstraction that claim to rationalize the world on the basis of corporate logic.”33 He writes: “It is about extraction, capture, the cult of data, the commodification of human capacity for thought and the dismissal of critical reason in favour of programming. . . . Now more than ever before, what we need is a new critique of technology, of the experience of technical life.”34 The next era of critique will also need to find spaces beyond technical life by overturning the dogma of inevitability. When AI’s rapid expansion is seen as unstoppable, it is possible only to patch together legal and technical restraints on systems after the fact: to clean up datasets, strengthen privacy laws, or create ethics boards. But these will always be partial and incomplete responses in which technology is assumed and everything else must adapt. But what happens if we reverse this polarity and begin with the commitment to a more just and sustainable world? How can we intervene to address interdependent issues of social, economic, and climate injustice? Where does technology serve that vision? And are there places where AI should not be used, where it undermines justice? This is the basis for a renewed politics of refusal—opposing the narratives of technological inevitability that says, “If it can be done, it will be.” Rather than asking where AI will be applied, merely because it can, the emphasis should be on why it ought to be applied. By asking, “Why use artificial intelligence?” we can question the idea that everything should be subject to the logics of statistical prediction and profit accumulation, what Donna Haraway terms the “informatics of domination.”35 We see glimpses of this refusal when populations choose to dismantle predictive policing, ban facial recognition, or protest algorithmic grading. So far these minor victories have been piecemeal and localized, often centered in cities with more resources to organize, such as London, San Francisco, Hong Kong, and Portland, Oregon. But they point to the need for broader national and international movements that refuse technology-first approaches and focus on addressing underlying inequities and injustices. Refusal requires rejecting the idea that the same tools that serve capital, militaries, and police are also fit to transform schools, hospitals, cities, and ecologies, as though they were value neutral calculators that can be applied everywhere. The calls for labor, climate, and data justice are at their most powerful when they are united. Above all, I see the greatest hope in the growing justice movements that address the interrelatedness of capitalism, computation, and control: bringing together issues of climate justice, labor rights, racial justice, data protection, and the overreach of police and military power. By rejecting systems that further inequity and violence, we challenge the structures of power that AI currently reinforces and create the foundations for a different society.36 As Ruha Benjamin notes, “Derrick Bell said it like this: ‘To see things as they really are, you must imagine them for what they might be.’ We are pattern makers and we must change the content of our existing patterns.”37 To do so will require shaking off the enchantments of tech solutionism and embracing alternative solidarities —what Mbembé calls “a different politics of inhabiting the Earth, of repairing and sharing the planet.”38 There are sustainable collective politics beyond value extraction; there are commons worth keeping, worlds beyond the market, and ways to live beyond discrimination and brutal modes of optimization. Our task is to chart a course there.

Coda – Space

A ountdown begins. File footage starts rolling. Engines at the base of a towering Saturn V ignite, and the rocket begins liftoff. We hear the voice of Jeff Bezos: “Ever since I was five years old—that’s when Neil Armstrong stepped onto the surface of the moon—I’ve been passionate about space, rockets, rocket engines, space travel.” A parade of inspirational images appears: mountain climbers at summits, explorers descending into canyons, an ocean diver swimming through a shoal of fish. Cut to Bezos in a control room during a launch, adjusting his headset. His voiceover continues: “This is the most important work I’m doing. It’s a simple argument, this is the best planet. And so we face a choice. As we move forward, we’re gonna have to decide whether we want a civilization of stasis—we will have to cap population, we will have to cap energy usage per capita—or we can fix that problem, by moving out into space.”1 The soundtrack soars, and images of deep space are counterposed with shots of the busy freeways of Los Angeles and clogged cloverleaf junctions. “Von Braun said, after the lunar landing, ‘I have learned to use the word impossible with great caution.’ And I hope you guys take that attitude about your lives.”2 This scene comes from a promotional video for Bezos’s private aerospace company, Blue Origin. The company motto is Gradatim Ferociter, Latin for “Step by Step, Ferociously.” In the near term, Blue Origin is building reusable rockets and lunar landers, testing them primarily at its facility and suborbital base in West Texas. By 2024, the company wants to be shuttling astronauts and cargo to the Moon.3 But in the longer term, the company’s mission is far more ambitious: to help bring about a future in which millions are living and working in space. Specifically, Bezos has outlined his hopes to build giant space colonies, where people would live in floating manufactured environments.4 Heavy industry would move off-planet altogether, the new frontier for extraction. Meanwhile, Earth would be zoned for residential building and light industry, left as a “beautiful place to live, a beautiful place to visit”—presumably for those who can afford to be there, rather than working in the off-world colonies.5 Bezos possesses extraordinary and growing industrial power. Amazon continues to capture more of U.S. online commerce, Amazon Web Services represents nearly half of the cloud-computing industry, and, by some estimates, Amazon’s site has more product searches than Google.6 Despite all this, Bezos is worried. His fear is that the planet’s growing energy demands will soon outstrip its limited supply. For him, concern “is not necessarily extinction” but stasis: “We will have to stop growing, which I think is a very bad future.”7 Bezos is not alone. He is just one of several tech billionaires focused on space. Planetary Resources, led by the founder of the X Prize, Peter Diamandis, and backed with investment from Google’s Larry Page and Eric Schmidt, aimed to create the first commercial mine in space by drilling asteroids.8 Elon Musk, chief executive of Tesla and SpaceX, has announced his intention to colonize Mars within a hundred years— while admitting that, to do so, the first astronauts must “be prepared to die.”9 Musk has also advocated terraforming the surface of Mars for human settlement by exploding nuclear weapons at the poles.10 SpaceX made a T-shirt that reads “NUKE MARS.” Musk also conducted what is arguably the most expensive public relations exercise in history when he launched a Tesla car into heliocentric orbit on a SpaceX Falcon Heavy rocket. Researchers estimate that the car will remain in space for millions of years, until it finally crashes back to Earth.11 The ideology of these space spectacles is deeply interconnected with that of the AI industry. Extreme wealth and power generated from technology companies now enables a small group of men to pursue their own private space race. They depend on exploiting the knowledge and infrastructures of the public space programs of the twentieth century and often rely on government funding and tax incentives as well.12 Their aim is not to limit extraction and growth but to extend it across the solar system. In actuality, these efforts are as much about an imaginary of space, endless growth, and immortality than they are about the uncertain and unpleasant possibilities of actual space colonization. Bezos’s inspiration for conquering space comes, in part, from the physicist and science fiction novelist Gerard K. O’Neill. O’Neill wrote The High Frontier: Human Colonies in Space, a 1976 fantasy of space colonization, which includes lush illustrations of moon mining with Rockwellian abundance.13 Bezos’s plan for Blue Origin is inspired by this bucolic vision of permanent human settlement, for which no current technology exists.14 O’Neill was driven by the “dismay and shock” he felt when he read the 1972 landmark report by the Club of Rome, called The Limits to Growth.15 The report published extensive data and predictive models about the end of nonrenewable resources and the impact on population growth, sustainability, and humanity’s future on Earth.16 As the architecture and planning scholar Fred Scharmen summarizes:

The Club of Rome models calculate outcomes from different sets of initial assumptions. The baseline scenarios, extrapolated from then-current trends, show resource and population collapse before the year 2100. When the models assume double the known resource reserves, they collapse again, to a slightly higher level but still before 2100. When they assume that technology will make available “unlimited” resources, population collapses even more sharply than before due to spikes in pollution. With pollution controls added to the model, population collapses after running out of food. In models that increase agricultural capacity, pollution overruns previous controls and both food and population collapse.

Limits to Growth suggested that moving to sustainable management and reuse of resources was the answer to long-term stability of global society and that narrowing the gap between rich and poor nations was the key to survival. Where Limits to Growth fell short was that it did not foresee the larger set of interconnected systems that now make up the global economy and how previously uneconomic forms of mining would be incentivized, driving greater environmental harms, land and water degradation, and accelerated resource depletion. In writing The High Frontier, O’Neill wanted to imagine a different way out of the no-growth model rather than limiting production and consumption.18 By positing that space was a solution, O’Neill redirected global anxiety in the 1970s over gasoline shortages and oil crises with visions of serene stable space structures that would simultaneously preserve the status quo and offer new opportunities. “If Earth doesn’t have enough surface area,” O’Neill urged, “then humans should simply build more.”19 The science of how it would work and the economics of how we could afford it were details left for another day; the dream was all that mattered.20 That space colonization and frontier mining have become the common corporate fantasies of tech billionaires underscores a fundamentally troubling relationship to Earth. Their vision of the future does not include minimizing oil and gas exploration or containing resource consumption or even reducing the exploitative labor practices that have enriched them. Instead, the language of the tech elite often echoes settler colonialism, seeking to displace Earth’s population and capture territory for mineral extraction. Silicon Valley’s billionaire space race similarly assumes that the last commons—outer space—can be taken by whichever empire gets there first. This is despite the main convention governing space mining, the 1967 Outer Space Treaty, which recognizes that space is the “common interest of all mankind” and that any exploration or use “should be carried on for the benefit of all peoples.”21 In 2015, Bezos’s Blue Origin and Musk’s SpaceX lobbied Congress and the Obama administration to enact the Commercial Space Launch Competitiveness Act.22 It extends an exemption for commercial space companies from federal regulation until 2023, allowing them to own any mining resources extracted from asteroids and keep the profits.23 This legislation directly undercuts the idea of space as a commons, and creates a commercial incentive to “go forth and conquer.”24 Space has become the ultimate imperial ambition, symbolizing an escape from the limits of Earth, bodies, and regulation. It is perhaps no surprise that many of the Silicon Valley tech elite are invested in the vision of abandoning the planet. Space colonization fits well alongside the other fantasies of life-extension dieting, blood transfusions from teenagers, brain-uploading to the cloud, and vitamins for immortality.25 Blue Origin’s high-gloss advertising is part of this dark utopianism. It is a whispered summons to become the Übermensch, to exceed all boundaries: biological, social, ethical, and ecological. But underneath, these visions of brave new worlds seem driven most of all by fear: fear of death—individually and collectively—and fear that time is truly running out.

I’m back in the van for the last leg of my journey. I drive south out of Albuquerque, New Mexico, headed toward the Texas border. On my way, I take a detour past the rocky face of San Augustin Peak and follow the steep drive down to the White Sands Missile Range, where in 1946 the United States launched the first rocket containing a camera into space. That mission was led by Wernher von Braun, who had been the technical director of Germany’s missile rocket development program. He defected to the United States after the war, and there he began experimenting with confiscated V-2 rockets— the very missiles he had helped design, which had been fired against the Allies across Europe. But this time he sent them directly upward, into space. The rocket ascended to an altitude of 65 miles, capturing images every 1.5 seconds, before crashing into the New Mexican desert. The film survived inside of a steel cassette, revealing a grainy but distinctly Earthlike curve.

That Bezos chose to quote von Braun in his Blue Origin commercial is notable. Von Braun was chief rocket engineer of the Third Reich and admitted using concentration camp slave labor to build his V-2 rockets; some consider him a war criminal.27 More people died in the camps building the rockets than were killed by them in war.28 But it is von Braun’s work as head of NASA’s Marshall Space Flight Center, where he was instrumental in the design of the Saturn V rocket, that is best known.29 Bathed in the glow of Apollo 11, washed clean of history, Bezos has found his hero—a man who refused to believe in impossibility. After driving through El Paso, Texas, I take Route 62 toward the Salt Basin Dunes. It’s late in the afternoon, and colors are starting to bloom in the cumulus clouds. There’s a T-junction, and after turning right, the road begins to trace along the Sierra Diablos. This is Bezos country. The first indication is a large ranch house set back from the road, with a sign in red letters that reads “Figure 2” on a white gate. It’s the ranch that Bezos purchased in 2004, just part of the three hundred thousand acres he owns in the area.30 The land has a violent colonial history: one of the final battles between the Texas Rangers and the Apaches occurred just west of this site in 1881, and nine years later the ranch was created by the one-time Confederate rider and cattleman James Monroe Daugherty.3

nearby is the turn-off to the blue origin suborbital launch facility. the private road is blocked by a bright blue gate with security notices warning of video surveillance and a guard station bristling with cameras. i stay on the highway and pull the van over to the side of the road a few minutes away. from here, the views stretch across the valley to the blue origin landing site, where the rockets are being tested for what is expected to be the company’s first human mission into space. cars pass through the boom gates as the workers clock out for the day. looking back at the clusters of sheds that mark out the rocket base, it feels very provisional and makeshift in this dry expanse of the permian basin. the vast span of the valley is broken with a hollow circle, the landing pad where blue origin’s reusable rockets are meant to touch down on a feather logo painted in the center. that’s all there is to see. it’s a private infrastructure-in-progress, guarded and gated, a technoscientific imaginary of power, extraction, and escape, driven by the wealthiest man on the planet. it is a hedge against earth. the light is fading now, and steel-gray clouds are moving against the sky. the desert looks silvery, dotted with white sage bushes and clusters of volcanic tuff punctuating what was once the floor of a great inland sea. after taking a photograph, i head back to the van to begin the final drive of the day to the town of marfa. it’s not until i start driving away that i realize i’m being followed. two matching black chevrolet pickups begin aggressively tailgating at close range. i pull over in the hope they will pass. they also pull over. no one moves. after waiting a few minutes, i slowly begin to drive again. they maintain their sinister escort all the way to the edge of the darkening valley.