Capitalist AI

A Marxist case against its further development

José Clemente Orozco’s mural Prometheus (1930)

I’ve seen a lot of people criticize AI over the past year, and often in ways I agree with, but I haven’t read a Marxist case against it yet. I’m going to make one because I think AI’s critics sometimes conflate machine learning technology itself with the capitalists who control it, and that that criticism sometimes devolves into an anti-technology stance I disagree with. I support the Luddite position towards capitalist AI development, and I think we should do everything we can to stop it, but that’s because capitalists are the genocidal, ecocidal enemies of human life. I don’t want them gaining more power with AI for the same reason I wouldn’t have wanted the Nazis splitting the atom, but as with aspects of nuclear science, I think machine learning could be useful in a communist society.

When I talk about AI, I mean pattern-recognition software, large language model text generators, and image generators. Right now, the capitalist class is using AI pattern recognition to surveil working people, bomb schools and refugee camps, and Israel is using it to scan thousands of hours of Iranian street footage with facial recognition software and find Iranian leaders to assassinate. Bosses are using it to spy on their employees, men are using it to create fake porn of women they want to degrade, and AI companies are using it to steal real people’s voices. AI is environmentally destructive, it makes people dumber, and the capitalist class is using it as a pretext to fire as many people as they can. Capitalist AI has overwhelmingly made the West a worse place to live. Its current uses are indefensible, and no future society can ethically or ecologically countenance them.

However, I also think that’s because capitalists are the ones directing western AI use. They’ve used automation to inhumane ends since the Industrial Revolution. In chapter 15 of Capital Volume One, “Machinery and Large-Scale Industry,” Marx writes that the widespread adoption of the sewing machine in the 19th-century led to mass job losses for people paid to weave by hand, especially children and older women. “The fearful increase in death from starvation during the last ten years in London runs parallel with the extension of machine sewing,”¹ he writes. “World history offers no spectacle more frightful than the gradual extinction of the English hand-loom weavers... [and in India] the bones of the cotton-weavers are bleaching the plains.”²

Capitalists left these people to die for two reasons: first, because capitalism has no purpose beyond the maximization of profit, and second, because they need to keep the proletariat poor or they wouldn’t otherwise work for them. If the state stepped in to meet people’s material needs, most of them wouldn’t spend eight-plus hours a day working for a wage to make someone else rich. The threat of homelessness and starvation “rivets the worker to capital more firmly than the wedges of Hephaestus held Prometheus to the rock.”³ It’s been this way since the system’s violent inception and capitalists have historically used automation to make good on that threat.

In chapter 27 of Capital Volume One, “The Expropriation of the Agricultural Population from the Land,” Marx describes the enclosure of the commons, or when capitalists superseded feudalists, conquered the land that peasants lived on, and forced them into wage work. Capitalist merchants and aristocrats coordinated with the British state to send in soldiers, strip working people of their houses and livestock, and turn them into “vagabonds” without land who were then forced to work for wages to survive. If those people refused to work for capitalists, they could be “whipped, branded, and tortured by grotesquely terroristic laws into accepting the discipline necessary for the system of wage-labor.”⁴ King Edward IV granted capitalists the right to make “idlers” into slaves,⁵ and every person who wanted to live outside the system was forced into it as a capitalist or a worker. Capitalists later used automation to throw lots of those workers on the street when they no longer had use for them.

AI, or contemporary automation, presents the same economic problem to workers as its 19th-century counterpart. In a capitalist system, it pits workers against machines. Capitalists, acting as “capital personified with a will and consciousness,”⁶ have to maximize their profits and fight alternative systems of resource allocation that could compete with capital to meet people’s needs. This means that they’ll fight to ensure that anyone whose job is “replaced by AI” meets the same fate as those “replaced by the sewing machine.” You can see this in the tepidness of AI manufacturers’ answers about what will happen to workers their technology puts out of a job.

In a recent discussion about AI, weapons manufacturer Peter Thiel hesitated when asked if would “prefer the human species to endure.” Elon Musk, the pro-genocide leader of the U.S. government’s jobs purge, has said he wants AI to “replace” human workers and that “universal high income” would sustain a world without necessary work. He offered no details about how this basic income would function. Sam Altman, the U.S. military contractor head of OpenAI, similarly argued for a “universal basic income,” or for the government to pay people directly, but he hasn’t meaningfully pushed for it at scale. None of these people really believe in direct resource allocation. Their power results from their ability to disburse capital, and if that power were undercut by people not needing their money, they’d lose everything they’ve spent their whole lives trying to gain.

Despite all this, I still think machine learning itself could be used to remarkable ends. Once, on deadline at work, I used a free trial of an AI transcription service to transcribe a 30-minute interview I’d done. It did a near-flawless job in two minutes and stunned me. I felt the same sense of awe as when I discovered I could use a web browser on my iPod Touch at age 12 and realized I was carrying a touch-screen computer in my pocket. I think people could use both pattern-recognition software and text, speech, and image generation technology to do incredible things, among them medical sequencing, complex information retrieval, and that they could find plenty of other uses I can’t imagine. I also think scientists will develop ways to lower AI’s energy use, and that communists could ration energy to put machine learning towards useful ends. Unfortunately, that’s not systemically possible in a genocidal, ecocidal capitalist system, and I think we should fight to stop capitalist development of the technology rather than let people committing mass slaughter make further advances with it.

In the Capital Volume One section, “The Struggle Between Worker and Machine,” Marx writes that “Workers are in revolt against this particular form of the means of production (machines) because it is the material foundation of the capitalist mode of production.”⁷ He writes that “It took both time and experience before workers learnt to distinguish between machinery and its employment by capital, and therefore to transfer their attacks from the material instruments of production to the form of society which utilizes those instruments.”⁸ I think we should take a similar approach to machine learning technology. In a communist world where the state guarantees food, shelter, education and medicine to every human being, people wouldn’t be pit against machines for their survival. I think they could do remarkable things with automation if they were free. But until then, we should fight to stop its development by people whose only interest is death.

Recommended Reading

• I liked this essay by B.D. McClay about The Passenger Seat by Vijay Khurana, a novel about how the manosphere made two teenage boys into killers. I liked how she described its invisible presence and how a group is capable of things that people alone aren’t, namely in the lines “The push–pull of who can and can’t drive, who has and has not made out with a girl, who does and doesn’t have a gun license, of a thousand constant tiny competitions, are essential to the creation of Adam-and-Teddy.⁴ Who’s in the passenger seat of this friendship? Both of them, that is: Adam and Teddy are going along. Who is driving? Neither of them, that is: Adam-and-Teddy is driving.”

• I liked this essay in Dirt by Laura Adamczyk about third places and the egalitarianism of the pool hall, which is a topic I’m also invested in as my third places keep me from going insane.

1. Capital Volume One, page 601 (Penguin Classics Edition, 1976)

2. Capital Volume One, pages 557-558 (Penguin Classics Edition, 1976)

3. Capital Volume One, page 799 (Penguin Classics Edition, 1976)

4. Capital Volume One, page 899 (Penguin Classics Edition, 1976)

5. Capital Volume One, page 897 (Penguin Classics Edition, 1976)

6. Capital Volume One, page 225 (Penguin Classics Edition, 1976)

7. Capital Volume One, page 554 (Penguin Classics Edition, 1976)

8. Capital Volume One, page 554 (Penguin Classics Edition, 1976)