A New Arrangement of the Earth

There is a photograph. I glimpsed it on Twitter years ago, and I have never been able to find it since. Perhaps it is an image without substance, a product of my imagination. The photo depicted various stones—large and small—on a beach, wet and glistening: pebbles, gravel, and round cobbles. The photographer, presumably on a walk, must have crouched down to capture something. At first glance, the image appeared to be a harmonious composition of colorful, naturally shaped stones. Only upon closer inspection could one notice a fragment of a printed plastic panel—flooring material designed to resemble stones—camouflaged beneath them. The synthetic stone pattern was seamlessly integrated among the natural rocks, forming an uncanny unity. Judging by its slightly faded color and irregular edges, it was not something deliberately placed. Rather, it seemed to have drifted ashore after floating on the sea—an authentic found object. This image can be seen as an episode or allegory of contemporary visual culture that illustrates the hybrid harmony between nature and the artificial.

For modern viewers, witnessing artificial objects blend seamlessly into natural ones to form a harmonious new assemblage has become a familiar sight. To borrow the language of Bruno Latour, this may be just a simple example of the proliferating hybrids of modernity. Over tens of millions of years, massive rocks break down into beach pebbles, while remnants of organisms buried underground over similarly vast periods are transformed into petroleum, which in turn becomes plastic fragments. Today, these two materials lie side by side, providing an indistinguishably similar visual experience. The everyday presence of artificial objects like computers, mobile phones, and electric vehicles often obscures the fact that they are composed of natural materials. Nature and culture are already so entangled that even the binary distinction between them holds little inherent meaning.

Like any other natural substance, rocks neither despair nor harbor vague hope; they endure the Earth's timeline that spans billions of years. Over the course of deep time, a single boulder may have undergone multiple cycles—transforming from rock to pebble to sand to clay and back to rock again. Even if we worry about acid rain corroding it, the rock will likely retain its solidity long after humanity has gone extinct, perhaps until a new form of life—potentially non-biological—dominates the Earth. But perhaps such a judgment is too hasty, too modern. What if rocks and stones are not so enduring after all? What if all of Earth’s stones are mined—used up either for extracting the materials necessary for digital media devices or as sources of energy? Then the natural rocks might be replaced by synthetic ones, or merely images resembling them. In contrast to the time of the Earth and stones, humanity has, in just a couple of hundred years, accelerated and compressed the vast geological process of “deep time” through development, mining, and consumption. In pursuit of energy, modern humans have indiscriminately mined fossil fuels such as coal and oil, along with minerals like copper, aluminum, cobalt, silicon, nickel, and lithium for manufacturing and powering all kinds of industrial products and digital devices.
 
 
 
Between Data Extraction and Resource Mining

Recently, many have referred to data as a “new natural resource,” and this is more than just a metaphor. While data isn’t literally mined from the ground, it serves as the raw material that feeds global digital platforms and algorithmic networks—keeping them alive and functioning. Data is omnipresent, and once its necessity is recognized and measurement devices (sensors) are developed, it becomes an infinitely mineable substance. Although the very existence of data—whether it is an object, a concept, or an event—remains ambiguous, there is no doubt that anything in the world can be turned into data, making it an exceptionally plastic resource.

Just as the fusion of natural stones and artificial stone patterns has become “natural” to us, so too has the combination of silicon (as material) and algorithms (as data). The transition from industrial capitalism to cognitive capitalism has occurred through the inevitable linkage between rare minerals extracted from rocks and the data that can be extracted from every facet of the world. While digital or cognitive capitalism seems to reduce everything into the immaterial and intangible—such as data—detached from the physical earth and matter, in truth, it demands even deeper entanglement with rocks and the planet.

Matteo Pasquinelli summarizes this technological and social condition with the concepts of the “carbosilicon machine” and “cyberfossil capital.”¹ Our society is moving toward a cybernetic world where everything is automated through the integration of energy-producing fuels, materials that make up semiconductors, and information that gives form to things and energy alike.

Critical AI researcher Kate Crawford begins her book Atlas of AI² from the foundational layer—“Earth”⁴—drawing on Benjamin Bratton’s planetary-scale platform model in The Stack.³ The computational reality in which everything has become calculable is built across multiple layers and scales: energy and minerals, cloud infrastructure, smart cities, the Internet of Things, user interfaces, and AI algorithms. The abstract and immaterial systems of software and algorithms are layered directly above and below the strata where energy and minerals are mined. In Crawford’s collaborative project with Vladan Joler, Anatomy of an AI System,⁵ their holistic analysis of AI also begins and ends with the Earth. “Each product included in the extended network of an AI system—from network routers to batteries to microphones—is made of elements formed over billions of years. From the perspective of deep time, we are squeezing the history of the Earth to manufacture technological artifacts used for mere years. […] Geological processes mark both the beginning and end of the entire lifecycle, from ore extraction to disposal in electronic waste dumps.”

From this perspective, Sandro Mezzadra and Brett Neilson attempt to redefine the dominant paradigm of contemporary capitalism by expanding the meaning of “extractivism.”⁶ Historically, the term referred to the imperialist process by which natural resources and living beings (including humans) were forcibly displaced and exploited—mainly in the Global South—for value production and accumulation. But now, the scope of “extraction” is being broadened to encompass practices like data mining and the commodification of human life itself in the era of biocapitalism.

In other words, from the extraction of natural resources for manufacturing and sustaining AI devices, to the mining of data for training and refining AI algorithms, and further to the exploitation of human data labor that enables AI’s completion, the mechanisms of contemporary capitalism are all fundamentally extractivist in nature. Furthermore, given that today’s mining, extraction, and exploitation still overwhelmingly take place in the Global South and peripheral regions, it must be emphasized that the global inequality of capitalism remains historically unchanged.

The expanded notion of “extractivism” proposed by Mezzadra and Neilson is vividly visualized through the maps meticulously rendered by Joler and Crawford. In Anatomy of an AI System, they trace the macro and micro pathways of a small AI speaker device—Amazon’s Echo (equipped with the Alexa system)—from the mineral elements it is made of, to how its algorithms are trained and operated through human labor, and finally to how it is dismantled and buried in the earth after disposal.⁷

Through this comprehensive diagram, we come to realize a crucial, yet often overlooked fact: even something as seemingly immaterial and abstract as data—or AI algorithms—requires a material (mineral-based) foundation to exist, function, and circulate in digital form. This includes everything from generation, storage, computation, measurement, to mobility. The diagram reveals that mining and disposal are unavoidable components of the data-AI cycle.

Within a single AI system, the extraction of resources, labor, and data occurs simultaneously, generating value through a multi-layered process of exploitation and extraction. This structure mirrors the Marxist dialectical triangle of labor force, means of production, and product—a fractal pattern akin to a Sierpiński triangle, repeating complexity at every level. Whether at a rare earth mineral mining site in the Congo, an Amazon voice AI manufacturing plant in China, or a household using digital devices, it is impossible to escape this triangular dialectic.



Extraction of Knowledge and the Mind

This omnipresent extractivism now extends into the realms of our knowledge, consciousness, and the automation of perception. AI stands at the forefront of this mental extractivism. In their collaborative text, The Nooscope Manifested, Vladan Joler and Matteo Pasquinelli describe in detail—via another map (or diagram)—the full process by which biases in AI are trained on datasets through machine learning, structured into algorithmic models, and ultimately applied in real-world contexts.⁸

To strip AI of its ideological status as an “intelligent machine” and reposition it simply as a knowledge instrument, the authors suggest it is more rational to consider machine learning as a tool for amplifying human knowledge—one that helps identify features, patterns, and correlations in vast data spaces beyond the reach of human cognition. Rather than mythologizing AI as something autonomous and superior to human intelligence, we should see it as a kind of magnifying lens—an apparatus that refracts, reinforces, and filters knowledge extracted from various forms of data.

They outline, in three steps, how human biases (errors) and statistical (machine) biases are amplified during the training phase of machine learning, embedded within the algorithmic model, and eventually transferred into real-life application through AI’s “lens.”

In the final stage shown on the Nooscope diagram—where the algorithm is applied to reality—two directions emerge: classification, which performs “pattern recognition,” and prediction, which handles “pattern generation.” Today, however, the term “generation” would more appropriately be labeled as what we now call generative AI.⁹ This is because prediction, in essence, is less about generating patterns than it is about making judgments based on extrapolating future scenarios from past patterns.

In short, AI applications in reality today can be divided into “judgment” (classification, which includes prediction) and “generation” (creation or production). These two processes not only represent the primary roles of AI in contemporary society but also correspond to two key human mental faculties traditionally thought to be irreplaceable: judgment and imagination.

Borrowing loosely from Kantian terminology, we can consider these as the faculties of judgment and imagination: the ability to evaluate whether a given object conforms to a rule, and the intuitive capacity to represent an object even when it is not currently present. By this logic, contemporary generative AI can be seen as a model that attempts to simulate the human faculty of imagination.

Whether or not AI can actually “imagine” is ultimately a futile debate, but what matters is that generative AI (especially image-generating models) produces representations—images—in a way that resembles human imagination, however imperfectly.

 
 
Generation or Extraction?

At the beginning of her book Atlas of AI, right after the anecdote about the clever horse Hans, Kate Crawford identifies two dominant fantasies that operate powerfully in the field of AI. The first is “that nonhuman systems (whether computer or speech) are like human minds,” and the second is “that intelligence is something that exists independently, as if it were natural and distinct from social, cultural, historical, and political forces.”¹⁰ Crawford argues that these are illusions widely held by the public—that AI resembles human intelligence, and that intelligence itself is an independent faculty, detached from other human capacities.

Yet as the story of Clever Hans illustrates, even if AI lacks actual human intelligence and does not emerge from a harmonious coordination of other human faculties, the fact that AI performs something considered equivalent to human intellectual capacities (like judgment or imagination) inevitably leads people to interpret it as possessing those very capacities. Perhaps the critical perspectives that seek to reduce AI to something inferior compared to human superiority may in fact be based on a narrow conception of “intelligence” itself.

Therefore, our focus should move beyond whether AI “has” intelligence or “thinks.” AI algorithms today are trained on vast datasets accumulated from past and present human activities and are now capable of generating unprecedented texts, images, and even sounds using the algorithmic patterns constructed during that training. Artists and creators in the cultural industries increasingly utilize generative AI as a tool or collaborator in producing creative yet efficient outcomes. Although the issue of hallucinations will likely never be fully resolved, artistic creation is increasingly conducted through dialogue, interaction, and question-and-answer exchanges with generative AI—offering both artists and AI systems opportunities for mutual innovation and evolution.

This will likely be demonstrated clearly in the near future. However, what must not be overlooked is the fact that the generative capability of AI is made possible only through the extraction of human capacities. Generative AI, in essence, is an extractive machine. It does not generate creativity or imagination like a human—or beyond human capabilities—but rather extracts human imagination and curiosity by producing something that simulates those very qualities.

So how does AI extract even human creativity and imagination? Vladan Joler’s New Extractivism project presents a vast cartographic view that encapsulates every dimension of digital capitalism’s extractivist tendencies—from human physical and mental activity to the Earth’s subterranean resources, and from the cellular level to outer space.¹¹

In this map, a middle zone referred to as the “factory” is where data is provided to AI algorithms, processed, and applied to reality. Above this lies the world of SNS platforms and digital services, which, like irresistible gravity, pull individuals in and trap them in a space reminiscent of a hybrid between Plato’s cave and Foucault’s panopticon—called the “Platopticon.” While each person consumes individualized content in their own space, their responses and interactions are converted into data. Below this layer, the map shows how different forms of data, nature, and labor are mined and extracted from the dimensions of human life, society, logistics, and resource extraction—then sent upward to the factory.

From this perspective, it may seem as if human creativity and imagination no longer exist. They are entirely subsumed by the AI extraction machine.



Art in the Age of New Extractivism

The series of works involving Vladan Joler depart from conventional academic discourse by using the visual form of a “map”—an integrated arrangement of allegories and conceptual structures—to represent the immense apparatus of extraction in contemporary AI capitalism. How do we come to understand that AI plays a central role in this planetary machinery of extractivism? And how does art reflect on this, and ultimately present it to us?

For Joler, a map is a form of media, and the act of mapping itself constitutes discursive and artistic practice. He “thinks primarily through the method of mapping,” viewing maps as a non-linear storytelling format—a cognitive space that can be explored in multiple ways. “Each map has its own symbolic, relational, and semantic language,” pulling us into the map so that we may “create our own paths and interpretations.”¹² Thus, a map itself becomes a work of art.

In an era where AI is believed to possess the unique creative faculties of humans, what can art do? On one side are artists who adopt generative AI technologies as tools and collaborators, devoting their efforts to adding novelty and speed to the creative process. On the other side are those who question or challenge the very premise of such technologies. While the former group may always constitute the majority, it is unlikely they will break through the boundaries of art as radically as the latter.

In this sense, Unmake Lab is a representative example of the latter. Through a series of works, they explore the subtle and perhaps unsettling relationship between the natural object of stone and artificial intelligence. Their works include performance videos like Utopian Extraction (2020), real-time videos such as Fresh Stones (2020), and more recent pieces employing GPT-3, motion tracking, and virtual engines, such as The Sisyphean Variables (2021), as well as Camouflage Ketchup (2022), which uses object-recognition AI.

Unmake Lab critically examines the inherently extractivist nature of contemporary digital technologies—often invisible yet pervasive—through the motif of the stone. They investigate how rocks, earth, terrain, and nature are subsumed into the domains of data and AI, how this extraction process creates problems, and how the process of AI “generation” ultimately amounts to another form of extraction—all through a series of representational gestures.

To return to the topic of stones and conclude this essay: in the age of digital technology, data, and AI, it seems increasingly clear that everything we experience remains fundamentally grounded in earth, soil, and rock. Donna Haraway speaks of the chthonic—the subterranean and earthy—as central to our time. She proposes that we are not “Homo” but rather “humus,” not human but compost, and suggests that we should call this era not the Anthropocene but the “Chthulucene”—the age of the Earth.¹³

Reflecting on this, we come to realize that the boundaries between the artificial and the natural, the human and the nonhuman, are ultimately fictional. And yet, in a world where everything is becoming a target of extraction and as we face the looming catastrophe of the Anthropocene, art still holds an open and unresolved space—where it can become something else, and do something more. This, at least, offers us a small measure of consolation.



 
1.  Matteo Pasquinelli, “The Automaton of the Anthropocene: On Carbosilicon Machines and Cyberfossil Capital,” South Atlantic Quarterly, Volume 116, Issue 2 (April 2017), pp. 311–326.
2.  Kate Crawford, Atlas of AI, trans. by Seungyoung Noh (Seoul: Sosoh Book, 2022).
3.  Benjamin H. Bratton, The Stack: On Software and Sovereignty (Cambridge, Massachusetts: The MIT Press, 2016).
4.  In the Korean translation, the term was rendered as “Earth,” but it can also mean “land” or “soil.”
5.  Kate Crawford and Vladan Joler, “Anatomy of an AI System: The Amazon Echo as an anatomical map of human labor, data and planetary resources,” AI Now Institute and Share Lab, September 7, 2018. Korean translation available via the 13th Gwangju Biennale (2021): https://13thgwangjubiennale.org/ko/crawford-joler/
6.  Sandro Mezzadra and Brett Neilson, “On the Multiple Frontiers of Extraction: Excavating Contemporary Capitalism,” Cultural Studies, Volume 31, Issues 2–3 (March 2017), pp. 185–204.
7.  See note 5.
8.  Vladan Joler and Matteo Pasquinelli, “The Nooscope Manifested: Artificial Intelligence as Instrument of Knowledge and Extractivism,” KIM HfG Karlsruhe and Share Lab, May 1, 2020. Full Korean translation available via the 13th Gwangju Biennale (2021): https://13thgwangjubiennale.org/ko/pasquinelli-joler/
9.  If Joler and Pasquinelli had written The Nooscope Manifested after the rise of contemporary generative AI, they likely would have used terms like “production” or “generation” instead of “prediction.”
10.  Kate Crawford, Atlas of AI, p. 13.
11.  Vladan Joler, “New Extractivism: An assemblage of concepts and allegories,” commissioned by Digital Earth for the Vertical Atlas publication (Ljubljana: Aksioma, Institute for Contemporary Art, 2020): https://extractivism.online/
12.  Vladan Joler and Sangmin Kim, “Conversation: Reading the Map of ‘New Extractivism’,” MMCA Research Journal, Vol. 14, 2022, p. 81.
13.  Donna Haraway, Staying with the Trouble, trans. by Yumi Choi (Seoul: Manonji, 2021), p. 99.

References

• [칼럼] 채굴되는 지구, 추출되는 데이터: AI 시대의 지도 그리기와 예술
  [Column] Mining the Earth, Extracting Data: Mapping and Art in the Age of AI