Do we have a use for “AI” after capitalism? This question is – of course – more a rhetorical device than a genuine question. It serves as a starting point for discussing two propositions: The first is that “AI” in its current form is little more but condensed capitalism and a technology created and shaped by it. The second is that in a society that will have moved beyond capitalism there is no use for what we currently see as “AI”. In order to understand both propositions I will explain the fundamental dynamics of capitalism, how they thrive in “AI” and why there is no place for it in a post-capitalist future. In case you prefer a spoken (German) version, you can check out the talk “KI nach dem Kapitalismus: Hat ChatGPT in der besseren neuen Welt einen Platz?“. This article draws mainly on that talk which Sandra Sieron and I gave at the 38th Chaos Computer Congress.
What defines capitalism
Without a good fundamental understanding of what constitutes capitalism we cannot begin to understand how it shaped contemporary “AI”. Let’s start with what capitalism is NOT. It is not a synonym for greed or lust for money. Both are individual character traits that may help to adapt to capitalism, yes. But greedy people existed before capitalism and capitalism is not just the sum of individual greed. Capitalism also is not a synonym for individualism, egoism or selfishness, either. While these are attributes forced upon people living under capitalism, capitalism itself is not a moral category. Finally, capitalism is not synonymous with democracy or freedom. The regimes of Pinochet (Chile), Trump (USA) or Putin (Russia) were and are full-blown authoritarianism or approaching it fast, but their economies are still market-based capitalism. On the other hand, Chile under Allende (the last democratically elected president before the US-backed Pinochet coup) was building a socialist economy while consciously respecting civil liberties and democratic institutions.
Capitalism can fundamentally be described by three aspects: private property, labour exploitation and profit-driven competition.
What, then, is capitalism? It is a social and economic system. Since its beginning it has always adapted to changing conditions and every attempt at defining it will inevitably fail to catch its complexity. Still, I do feel comfortable that capitalism can fundamentally be described by three aspects: private property, labour exploitation and profit-driven competition.
Private property means that individuals are owners of resources and production facilities. Under capitalism, land, corporations and natural elements are property. This property is legally safeguarded by the laws of capitalist societies and enforced by police and state power. Those that do not own resources or production facilities have no way to secure their everyday needs like food, shelter and clothes as the land and resources that would sustain them are under exclusive control of the capitalists. All these people – the vast majority of people in capitalist societies – have just one thing of value: their body and time. In order to survive, they sell both to the capitalist who pays them a wage for their labour. The product of this labour is then sold on markets which forces every capitalist to compete with other capitalists and forces all capitalists to optimise their production for profit. As a result, wages are never equivalent to the value of what workers produce but (much) less. The difference between the value of the goods produced and the wage is kept by the capitalist. The capitalist has little choice in this. She may be willing to pay higher wages but is forced into grinding every bit of profit out of her workers to avoid being outcompeted by other capitalists. In order to stay in business, every capitalist is incentivised to produce what is most profitable. The market does not care about what is actually needed. This is why under capitalism, production of consumer goods is prioritised over unprofitable care or social work. These get externalised via patriarchal norms or are extraordinarily underpaid thanks to racist or colonial exploitation.
How capitalism is embodied in “AI”
Having defined the fundamental aspects and dynamics of capitalism, we can look at how “AI” does embody it. You will have noticed that I use “AI” in parentheses. That is not only because there is nothing intelligent about “AI” (it is a very big probability machine) but because “AI” can mean very different things. In the current hype cycle, “AI” means large deep learning models and generative use cases. To be more specific: large generative models like ChatGPT, Google Gemini and Apple Intelligence. These are the products that are usually on people’s mind when you say “AI”. And this is what this article is referring to as “AI” as well.
Looking at this kind of generative large model “AI” the capitalist DNA becomes quite apparent. The corporations behind the respective models and products are privately owned for-profit commercial giants that are not under public control but are the property of few individuals and answer to their shareholders alone. When the heads of OpenAI, Softbank and Nvidia decide that the limited resources of our planet should be put into a $500 billion data centre infrastructure, they just can do that. These resources could better the daily lives of actual humans in uncountable ways, but are not under collective democratic control. Instead, very few corporations and individuals control them.
“AI” is shaped by capitalism and encapsulates its undemocratic and exploitative dynamics like few other techno-social phenomena of our time.
And the decision to use these vast resources is driven by the quest for profit. Right now “AI” benefits two groups: providers of processing technologies and corporations that gain a narrative for reducing wages. Microsoft and Nvidia are rejoicing over demand for processing capabilities. And companies that are in any kind of content business use it to lower wages or demand higher productivity. Generative “AI” is sold mainly on the premise of enabling workers to be more productive and churn out content more quickly and with less effort. Right now, the “product” that “AI” companies are selling is wage pressure. This product is created by appropriating public resources. Generative large models are possible only because the collective digital knowledge, data and information of the world wide web is privatised and used to “train” the models that are then used to generate profit. “Open Source AI” does not change this. It is just an attempt to “openwash” proprietary systems and a wedge to legitimise probably illegal behaviour.
As a final aspect, using the public resource of the open internet to train “AI” models requires enormous amounts of human labour. An “army of overseas workers in ‘digital sweatshops’” categorises, labels and filters the raw data to make it machine readable and digestible for pattern recognition (the heart of deep learning). There is no “AI”, there is only other people’s work. And these people usually aren’t white. This, in return, brings us right back to the colonial and racist exploitation that has been a foundation of capitalism since its beginning. “AI” is shaped by capitalism and encapsulates its undemocratic and exploitative dynamics like few other techno-social phenomena of our time.
What part could “AI” play in a post-capitalist future
Can a technology that embodies capitalism to such an extent have a place in a future that leaves capitalism behind? When we look at “AI” in its current form, the answer is a rather clear “no”. This is not merely because every technology inevitably carries its ideological legacy with it but because of what “AI” inherently is. Generative large models are giant pattern recognition and pattern reproduction systems. Prompted with unfamiliar inputs they will try to produce outputs that most likely fit the input based on data about the past. “AI” has no idea what a society looks like that is based on solidarity, equality and cooperation because it was trained on data about a society that is built on competition, exclusion and exploitation. “AI” is also a social homogenisation tool. Its ability to produce somewhat usable outputs requires it to produce outputs that are the most likely and most expected output. This is why “AI”-generated content looks so eerily familiar and boring. Finally, “AI” obfuscates responsibility. There is no way to know why exactly a model produces a certain output. The social credibility of an “AI”-based decision hinges on the promise that the “AI” was fed only “good data” and on the tendency to credit computer decisions with some kind of super-human rationality. That is why one of the most common use cases of “AI” today is to justify violence and cruelty. From denying access to health or social services to rationalising automated genocide, “AI” awards inhumane actions an aura of rationality.
Vivid pictures of what a world after capitalism could look like are the fuel that keep the spark alive.
Looking at all these inherent aspects of “AI” there is little room to imagine it being part of a world of solidarity, equality and cooperation. To the contrary: The societies built on the inbuilt ideologies of “AI” seem to be much more aligned with authoritarianism and fascism than anything else. Any political project that attempts at moving beyond capitalism needs to be deeply aware of these inherent traits of “AI”. If there are any lessons to be learned from the historic examples of oppression in the former USSR or enforced conformity in the former GDR , then it is that homogeneity and authoritarianism would spell doom for any post-capitalist project. This does not mean that pattern recognition and machine learning are categorically unable to be of service to social transformation. However, as Dan McQuillan said in his must-read book “Resisting AI“, it is “highly unlikely that this new apparatus will simply be a repurposing of existing AI for progressive social ends.”
Still, in recent times, there has been an encouraging revival of discussions about post-capitalist economies that embrace technological progress. Decentralised, democratic and cooperative models are being theorized to an encouraging extend. Leo Schlichter wrote an very useful overview about contemporary ideas while Eden Medina’s book “Cybernetic Revolutionaries” about Cybersyn, Chile’s attempt at cybernetic economic coordination, includes valuable lessons from historic experience. These examples are encouraging, but they should not seduce us into believing that there is a technological fix to social and political problems. The status-quo will not be transformed with an app or a protocol. Tech won’t save us. Organising, resistance and continuous democratic struggle will. But there has never been a better and more inspiring time for tech enthusiasts to join. Vivid pictures of what a world after capitalism could look like are the fuel that keep the spark alive. And these pictures cannot be created without those whose hearts and imagination is currently taken up by the hype around “AI”. We need their enthusiasm to construct and invent the future.