AI as the Logic of Capitalist Domination

Two recent books on Artificial Intelligence (AI) offer rich research and key conceptual frameworks that can help progressive activists comprehend and analyze this phenomenon. They also offer ways of combating AI’s harmful and dangerous effects.

Emily Bender and Alex Hanna’s The AI Con: How to Fight Big Tech’s Hype and Create the Future We Want, begins by discussing the origins of AI in the 1956 Dartmouth College workshop on “thinking machines” led by Marvin Minsky and John McCarthy, who coined the term “Artificial Intelligence.” They situate this development in the context of the U.S. military competition with the USSR and the view that the human mind, too, is a machine. The authors cite the work of MIT professor Joseph Weizenbaum, who invented the first chatbot, named ELIZA in the late 1960s, and became a critic of AI; he developed his critique in a 1976 book entitled Computer Power and Human Reason: From Judgment to Calculation.

After briefly discussing the 1940s’ origins of Large Language Models (LLMs) and word-sequencing with machine-learning algorithms, they demonstrate that ChatGPT (2022) is a “synthetic text extruding machine” generating language, “but without any intent or thinking mind behind it” (p. 31). Bender and Hanna do not discuss John Searle’s “Chinese Room Argument,” which had offered a serious challenge to Alan Turing’s argument that computers could imitate the human mind. However, their description of AI as a synthetic text extruding machine sheds light on Searle’s argument by showing that large “language models only manipulate the form of language, with neither understanding nor communicative intent” (p. 32). Chatbots only “link together word patterns” through the use of “probabilistic algorithms” trained by other people’s stolen works. ChatGPT as a “text synthesis machine” and its “learning by rote” are contrasted by the authors to the act of human writing, which involves critical thinking and requires communication, conversation, and in-person and written dialogues with other people, as well as with writers and experts in the case of scientific and academic writing (p. 50).

Furthermore, the authors emphasize that although it appears that AI is doing all the work, AI requires a great deal of hidden human labor usually by low-paid workers in the Global South who train AI and monitor it. Two examples are remote human workers monitoring self-driving cars and content moderators monitoring LLMs.

In contrast to Sam Altman, the CEO of OpenAI, who thinks that humans are only machines that string text, Joseph Weizenbaum argued in his 1976 book that computers restrict rather than enlarge our humanity. Bender and Hanna agree with Weizenbaum’s critical views that extended to a critique of capitalism and its militarism. They also believe that reducing the human mind to codes can only strengthen militarization and authoritarian rule.

Furthermore, they view AI’s massive use of energy and water and its intensification of the climate crisis as its existential danger. In their opinion, this problem and the surveillance state created by AI are the fundamental dangers.

Thus, they call for a “strategic refusal” (p. 190), meaning that anyone concerned about the dangers of AI should not use it and not feed it. They especially call on librarians and information science experts to train their patrons to do regular internet searches by looking up verifiable links, identifying sources such as books, articles, and websites, and evaluating the information in the context of the sources from which they arise.

In calling for “building socially situated technology” created to “strengthen and empower communities” and protect the environment (pp. 188–9), they argue that instead of all-purpose tools, we need specific tools for specific tasks. These can include tools for editing papers. People should be able to decide whether they want to “opt in” to AI or not. Technology should be created with the input and engagement of the people that it affects. People’s data rights would have to be protected. Technologies for facial recognition and data centers that are wasting our water and energy and contributing to climate catastrophe should be stopped.

Kate Crawford’s Atlas of AI, which was published five years ago, offers a comprehensive account rooted in philosophical, historical, and global research, including travels to mining towns in the United States and abroad, close scrutiny of the labor process at Amazon Fulfillment Centers, and a trip to Jeff Bezos’s Blue Origin suborbital launch facility in West Texas.

She begins with the idea that intelligence cannot be disembodied because it develops in a body and in relation to other human beings and nature. AI is not about “thinking” or “intelligence.” It is about “mimicking” based on LLMs. It serves existing relations of domination under capitalism. Hence, the title “Atlas of AI” refers to the “map of relations of domination” involving labor, capital, resource extraction, and promoting policing, militarization, racism, and sexism.

The first chapter is devoted to the extractivism of AI and its rape, plunder, and pollution of the environment in places ranging from a mining town in Nevada to countries such as Bolivia, Congo, China, and Australia. We learn that by 2040, at least 14 percent of the world’s greenhouse gases will be produced by Big Tech and that a single data center now uses an estimated 1.7 million gallons of water per day.

In the second chapter, Crawford takes readers to an Amazon Fulfillment Center where she has carefully observed the ways in which workers have to keep up with “the picking rate,” which has no regard for the limits of their bodies. She speaks out against the capitalist “control over time” in the process of production and discusses the history of workplace automation and worker resistance to it. The work of Charles Babbage, Henry Ford, and Frederick Winslow Taylor are singled out to show how automation further and further intensified the division of labor to the point of quantifying every movement of workers’ bodies with the aim of maximizing output and minimizing cost for capitalism. In the twentieth-century Ford assembly line, humans were treated like robots with the main emphasis on speed of production. What she sees as new in the twenty-first-century Amazon Fulfillment Centers and similar production sites is the intensity of the “surveillance” of “every movement” of the body to increase speed.

Crawford writes that the process of automation has “exemplified Marx’s description of the domination of clock time, ‘Time is everything, man is nothing; he is, at most, time’s carcass’”(p. 74). Furthermore, a human becomes “an appendage of the machine” (The Communist Manifesto). Crawford’s emphasis throughout the book is on the control over workers’ production time as a distinctive feature of capitalist production. This critical view of capitalism not only as a system of unequal distribution of wealth but as an alienated and alienating system of production is one of the most important contributions of her book.

She demonstrates powerfully that “AI is neither artificial nor intelligent” (p. 8). Instead, it is the result of the onerous and low-paid labor of miners and other workers. Capitalist reform, in her opinion, at best allows for regulations at the margins, but leaves relations of production untouched (p. 72).

In subsequent chapters on data and classification, she offers an immense amount of research on how Big Tech companies have used data stolen from creators and without context, how they mix truth and falsehoods in their LLMs, and promote classification systems that are biased, racist, misogynist, and reify people as objects. AI, she argues, “flattens social, political and economic relations into quantifiable entities” (p. 144) and turns all data including human affect, feelings, and facial expressions into a quantifiable mass. This critique can remind us of Marx’s discussion in the first chapter of Capital on how capitalism’s extreme division of labor reduces concrete, specific, and differentiated labor into an abstract undifferentiated and congealed mass of human labor.

Chapter 6 on the state argues that AI has been shaped by a close relationship between Silicon Valley, the U.S. military, and NASA. AI was first a military technology. Now it is Big Tech companies such as Palantir, which is run by the billionaire and MAGA supporter Peter Thiel, that are designing the mechanics of mass deportation by ICE. All of these technologies, she emphasizes, are aimed at shaping a centralized state with a powerful surveillance system to dominate and exploit humans.

Crawford urges readers to see that “AI is not an objective, universal, or neutral technique” (p. 212). It is shaped by economic, social, and political forces designed to promote domination and state power. It offers “computational interventions as universal solutions applicable to any problem” (p. 214). Its aim is to speed up production at the expense of human physical and mental well-being; mechanize and automate to the point of destroying the human ability and potential for critical thinking; and “convert an infinitely complex and changing universe into a Linnaean order of machine-readable tables” (p. 221).

Crawford does not think that there can be “an AI for the people.” The very foundation of AI is the promotion of “centralization of control” (p. 223). In the words of Black feminist poet and librarian Audre Lorde, “the Master’s Tools Will Never Dismantle the Master’s House.” Her “renewed politics of refusal” asks us to not accept AI as “a technological inevitability,” but to “question the idea that everything should be subject to the logic of statistical prediction and profit accumulation” (p. 226). She cites philosopher Achille Mbembe’s call for “a new critique of technology” (p. 226) that examines the “interrelatedness of capitalism, computation, and control” and “bring[s] together issues of climate justice, labor rights, racial justice, data protection and the overreach of police and military power” (p. 227).

Anything short of this, she argues, will allow for the dystopia of Jeff Bezos and Elon Musk with the exhaustion of earth’s resources, creation of Nazi-style concentration camps, and a Brave New World scenario involving space colonies.

At a time when a movement seems to be emerging in the United States in opposition to AI’s mass replacement of workers, AI data centers, AI surveillance, and AI’s harmful effects on education, the above books by Bender, Hanna, and Crawford can provide fruitful ground for the effort to develop an affirmative vision and practical immediate activities.