AI Primer

Because the issue at hand is about AI I thought I’d begin with a short explanation of what AI and Machine learning are. Traditional computing algorithms give explicit instructions to the computer. Eg. If input A then output B, where B = 2xA. However in AI and ML, the rules are not set a priori, they are learned by the algorithm. In AI we give the computer many inputs of A and the corresponding correct outputs B. The computer then tries some operations applied to A and generates a guess G. If the computer guessed correctly then G and the truth B would be the same. However in the begining it will be wrong so the so called loss function would be (G-B)≠0, thus we have a contradiction, a tension inside the computer. However, the computer keeps guessing at other things it can do to A and eventually it will get better until (G-B) ≈ 0. Thus AI in general needs training data, it cannot learn otherwise.

There are two main components to AI, there is the ‘architecture’ and there is the training data. The architecture is the structure of the machine learning code, neural networks were the first major development, followed by convolutional networks and now we are in a new era of transformer models which have caused a boom in Large Language Models, like chatGPT. Training data as the name suggests is the data or information used to train the parameters of the model architecture. The extent of the training data used for chatGPT is approaching all written and digitised text up to this point in history. It is huge!

While training data is unlikely to increase in the future. Advancement in AI will most likely come from new developments in model architectures. But there may also be computational advances that aid the development. So this leads me into the latest development in AI which originated with some researchers in China.

Let’s begin

Last month, a Chinese company launched a new AI language model called DeepSeek. It scored higher than every other Large Language Model (LLM)—including OpenAI’s ChatGPT—on multiple performance benchmarks. What shocked the tech world even more was the cost: DeepSeek was built using far less hardware and power, cutting development expenses to a fraction of what U.S. firms typically spend. The market reaction was swift: nearly a trillion dollars in tech stock value vanished, and Silicon Valley CEOs went into a panic at the prospect of China leading the AI race.

In response, many tech executives praised the U.S. government’s recent restrictions on advanced chip exports to China. Eric Schmidt, former CEO of Google, has long been explicit about the geopolitical stakes. As he once said, “In your lifetimes, the battle between the U.S. and China for knowledge supremacy is going to be THE big fight.” That view has now translated into policy: the CHIPS Act effectively banned the sale of advanced NVIDIA chips to China, with Schmidt hoping it would stall Chinese AI development by ten years. But DeepSeek’s success has already shown that those efforts are failing—and has proven Schmidt’s claim that the U.S. was “a few years ahead” tragically wrong.

Now we see Sam Altman, CEO of OpenAI, lobbying the Trump administration for a staggering $500 billion in subsidies to “stay ahead” of China. This proposed investment, dubbed the Stargate Project, involves OpenAI, SoftBank, Oracle, and the investment firm MGX pouring half a trillion dollars into AI infrastructure by 2029. Comparisons to the Manhattan Project are no accident—this is framed as a matter of national survival, a desperate attempt to maintain U.S. technological dominance.

But let’s be clear: what we are witnessing is not some grand new wave of innovation. It is a capitalist crisis in real time. The AI bubble has burst, and tech giants are scrambling to re-inflate it by leveraging the power of the state—rallying around nationalist sentiment while insisting on the “free market” in theory.

We could consider the following question – Will AI usher in a new era of freedom, or will it deepen the crises of inequality, imperialism, and repression? I will approach this topic through 3 lenses: Imperialism, Capitalist contradiction and AI within Socialism.

Imperialism and the AI Arms Race

The conflict over DeepSeek has been portrayed as a U.S.-vs.-China duel for technological supremacy. But this framing misses the deeper logic of imperialism, which emerges from the interplay of private capital and state power. Governments step in with bans, tariffs, and subsidies when big corporate profitability is threatened, and corporations pivot to nationalism and “security concerns” to justify their demands to government. As Chris Harman notes in The State and Capitalism Today, “the groups of capitals and the state with which they are associated form a system in which each affects the others… [They] are intertwined, with each feeding off the other.”

We’ve seen this pattern before. In the 1980s, Japanese automakers captured a significant share of the U.S. auto market, and General Motors, Ford, and Chrysler lobbied Washington to negotiate import limits. While preaching about “free enterprise,” they orchestrated a “Buy American” campaign that drew on nationalist sentiment—ultimately securing protection for their profits. Harman emphasizes that “No capitalist wants to face alone a world of bitter, unregulated competition… The state…sees its job as being to carry out lifeboat operations designed to forestall such collapse.” So even as corporations celebrate the free market in theory, they demand active state support in practice.

This dynamic is not just about economics; it is also about securing military and political dominance. Both the U.S. and China recognize that AI is critical for the next stage of automated warfare, state surveillance, and labor control. The U.S. has already integrated AI into military operations, with the Pentagon investing billions into autonomous weapons, cyber warfare, and AI-driven battlefield decision-making systems. Google and OpenAI, despite their public rhetoric about “ethical AI,” have been deeply involved in military contracts—while Amazon and Uber harness AI to surveil and discipline their workers. Meanwhile, Israel’s use of AI in the oppression of Palestinians exemplifies how AI-driven technologies are already reshaping modern conflict and policing.

China, too, is expanding AI-based authoritarianism. Facial recognition is used to track and control the population, particularly in regions like Xinjiang, where AI facilitates the surveillance and repression of Uyghurs. Yet, just like in the U.S., China’s AI development arises from the fusion of government priorities and capitalist enterprise: private Chinese AI firms such as Baidu and Tencent benefit enormously from state subsidies, just as U.S. firms benefit from Pentagon contracts.

Ultimately, imperialism is not just about war between states—it is about capitalists waging war against workers at home and abroad. While AI companies receive billions in subsidies, public services remain underfunded. Governments that claim they “cannot afford” universal healthcare or wage increases somehow find limitless resources for AI development, military applications, and corporate bailouts.

But capitalism cannot resolve its own contradictions. The more it tries to maintain control—via subsidies, tariffs, arms races—the clearer its unsustainability becomes. As AI intensifies competition and exposes the fragility of high-tech markets, the system plunges further into crisis. Whether it is U.S. or Chinese capital, the outcome is the same: heightened militarism, deepening exploitation, and escalating risks for the global working class. The stage is set for AI to fuel both external confrontations and the internal decay of the capitalist system.

AI and the Contradictions of Capitalism

The intensification of AI competition is not just sharpening the struggle between imperialist states—it is also accelerating capitalism’s most fundamental contradictions. AI has the potential to drastically reduce human labor, to increase efficiency, and to automate away many of the most grueling tasks of modern society. And yet, under capitalism, these very developments threaten workers with mass unemployment, economic insecurity, and deepening inequality.

A recent study found that around 3 million jobs in South Korea alone are at risk of being automated by AI, disproportionately affecting highly educated, white-collar workers—doctors, accountants, lawyers, and financial analysts. This overturns the traditional assumption that automation primarily impacts low-wage, manual labor jobs. In reality, AI threatens workers across all sectors, from factory employees to software engineers. And yet, the impact is not evenly distributed: while tech executives rake in billions, workers face greater precarity, worsening working conditions, and the constant threat of redundancy.

This is the core contradiction of automation under capitalism. AI has the capability to reduce labor time and improve social well-being, but because production is not organized for human need, but for profit, it results in job losses, deskilling, and economic instability. The technology that could be used to liberate workers from drudgery instead deepens their exploitation.

This paradox is most visible in the rise of gig work and algorithmic management. AI is not simply replacing workers—it is being used to intensify and degrade work conditions. Amazon warehouse employees, for example, are subjected to AI-driven surveillance that tracks their every move, penalizing them for inefficiencies and forcing them to work at breakneck speed. Uber drivers are managed by opaque AI algorithms that dictate their pay, control their working hours, and manipulate their behavior, all while denying them the legal protections of traditional employment. The result is a new digital proletariat, where workers are governed by AI systems they do not understand and cannot challenge.

The recent $1 trillion loss in U.S. tech stocks following the release of DeepSeek is an early warning sign that the AI bubble, like every capitalist boom before it, is built on shaky foundations. Tech firms, eager to secure state subsidies and venture capital funding, have vastly overpromised AI’s short-term economic potential. Yet “as capitalism ages, it finds it more and more difficult to overcome the pressures leading to stagnation and deep crises,” notes Chris Harman (Explaining the Crisis), emphasizing the system’s mounting contradictions. In AI’s case, “the scale of investment tends to rise much more quickly than the source of profit (labour power), so producing a decline in the rate of profit.” This imbalance threatens to create the very form of overaccumulation—where excess capital chases increasingly limited profitable outlets—that has repeatedly led to speculative bubbles. As with the dot-com crash and recent crypto collapses, “crises are not mere accidents; they express deep-seated conflicts that arise whenever capital seeks to expand beyond the limits set by profitability.” Despite the hype, AI’s inflated valuations and reliance on cheap credit suggest that today’s tech euphoria could quickly unravel once confronted with the hard limits of profitability.

This contradictory dynamic helps drive a wider crisis: as AI-driven automation expands, it undercuts labor—the very source of surplus value—thereby sapping workers’ buying power. “The basic contradiction in the system,” Harman writes, “is the way in which the scale of investment tends to rise much more quickly than the source of profit (labour power).” When capitalists replace workers with AI, they slash costs but simultaneously weaken the market for goods, intensifying the risk of overaccumulation. However we should look deeper here, according to Harman, there are three main factors that could offset this tendency “the destruction of certain capitals to the benefit of others through periodic crises; the flow of investment away from old areas of capitalist development into new ones through imperialism; the employment of a growing proportion of the investable surplus value in ways which aid particular sections of capital in their competition with other sectors but do not contribute to productive accumulation – such as marketing costs and arms.” Yet, while such measures can delay an outright slump, they do not resolve the underlying tensions.

Thus, capitalism enters a new phase of speculative investment, financial bubbles, and increasing reliance on state subsidies. Governments are stepping in to pump billions into AI research and development, not because they believe it will benefit society, but because AI has become a new frontier for capital accumulation, a temporary fix to capitalism’s worsening contradictions. But as history has shown, such interventions are only short-term solutions to deeper structural crises.

All of this points to a system in profound crisis. AI does not free people from capitalism’s instability; it intensifies the contradictions of automation, falling profitability, and social inequality. The capitalist class cannot fully embrace automation, because it depends on the exploitation of labor to extract profit. At the same time, it cannot afford to halt automation, because the competitive drive for technological advancement forces every firm to keep investing, regardless of the social consequences.

AI Under Socialism

The crisis of AI under capitalism is undeniable. Instead of being a tool for human liberation, it has the potential to deep exploitation, drive mass unemployment, and accelerate economic instability. But this trajectory is not inevitable. AI is not an autonomous force—it is shaped by the social and economic system in which it develops. The question is simple: who controls AI, and for whose benefit? Under capitalism, AI is a tool for maximizing profit, deepening imperialist rivalries, and expanding surveillance. But under socialism, it could be repurposed to reduce working hours, eliminate scarcity, and reorganize production to meet human needs.

This is why the response to AI must go beyond regulation, ethical oversight, or social welfare policies. The capitalist class will not voluntarily create a system where AI benefits workers. Even the most progressive proposals—such as Universal Basic Income—are designed to pacify workers rather than empower them. A world where AI is used to genuinely free humanity from unnecessary labor will not emerge from government policy alone—it must be won through class struggle.

We should remember that crises are also moments of struggle, and the response of the working class will determine the future AI creates. In South Korea, workers are already fighting back against AI-driven exploitation. Baemin delivery drivers, whose routes and wages are dictated by AI algorithms, have organized against increasing workloads and declining pay under gig economy platforms. Rider Union official said. “Their dangerous working environment is pushing them to ignore traffic lights and drive at unsafe speeds to save delivery time. Their safety, meanwhile, is not guaranteed at all.”

But this resistance must go further. Rather than merely resisting AI’s deployment, workers must demand democratic control over AI itself—not as a tool for corporate profit, but as a force to reduce working hours, improve conditions, and serve social needs. AI should be a means to free humanity from drudgery, not to intensify capitalist exploitation. The task ahead is not just to oppose job losses or surveillance, but to fight for a world where workers—not billionaires or tech monopolies—determine how AI is used.

If AI remains in the hands of the ruling class, the future is clear: a world where AI automates repression, deepens inequality, and accelerates capitalist decay. But in the hands of the working class, AI could be a tool for liberation—a means to eliminate drudgery, expand human freedom, and finally break from a system that reduces every technological advance to another instrument of exploitation.

We started with the question “Will AI usher in a new era of freedom, or will it deepen the crises of inequality, imperialism, and repression?”. I would say that the choice is stark. AI within Capitalism means a future of surveillance, unemployment, and imperialist wars fought with autonomous drones. An alternative, Socialist, AI means a future of human-centered technological progress, where workers decide how technology is used in their workplace and beyond. This is not a question for the distant future—it is a struggle unfolding now.