The Algorithms of Collapse

AI is being diffused throughout society under chatbots, models, and agents which are explicitly reactionary and create communicative and physical walls to defend the status quo.

Capitalism has elected AI as the next tool to distribute and dismantle labor, create a new power structure in the world, and repress social and political movements. Unchecked, it will bring us right up to a collapse brought on by war and climate chaos.

Forget about the Terminator stories of Artificial General Intelligence and Artificial Superintelligence. These are closer to sci-fi than to reality. We don’t need to speculate about things that don’t exist in the AI realm. What we do need to look at are the things that already exist and are being deployed massively.

The main objective of AI is the automation of historical automation itself. AI holds an irresistible promise for capitalist elites: to be able to automatically direct most of the instructions that guide human activity, reducing the power of social classes other than the owners of the algorithms. Complete economic and social planning for the rich. In particular, they want to reduce the power that the working class has exercised in the past, the power to push toward the future and gain the political, social, and economic transformations that reduce or eliminate inequality and injustice.

Data centers today are nightmare factories.

A key and complementary objective of AI is to create an overwhelming monopoly over knowledge, codified via Large Language Models, Computer Vision, Convolutional Neural Networks, and other Machine “Learning” models. This monopoly is being designed to utterly transform social relations and install a reactionary hegemony that widely surpasses neoliberal capitalism and feeds a far-right dystopia.

The third essential objective has to do with the control of violence and political repression. For that effect, AI provides different tools to be used in declared and now mostly-undeclared states of war. During the Gaza genocide, human targets were chosen with AI, its models were used to determine the biggest impacts for sequences of targets in order to achieve maximum infrastructure and human suffering consequences. Obviously, AI is used to maximize efficiency in all war logistics, calculating payloads, schedules, and material distribution. In Ukraine, most of the war is being conducted with drones, many of them autonomous and with self-selecting target capabilities powered by AI. Automated killing machines that don’t question orders or targets are not only available, but already deployed in different war fronts. On the other hand, automated political repression and persecution in the streets and protests is growing, though it is currently at the data gathering and training phase. In the USA, Immigration and Customs Enforcement is deploying different apps developed by companies like Palantir to maximize social disruption and to capture the most vulnerable people in the country.

There is huge pressure to prevent any meaningful regulation of AI, in particular for AI used by police and the military. Surveillance with facial (FRT) and body recognition is used outdoors to map out movements and participants in protests and actions. Mapping of movement connections and alliances can be done via online pattern recognition, as well as out in the streets. Automatic protest repression combined with purposeful miscommunication and disinformation might make the usual protests simply nonviable.

And of course, AI can is being used for hacking by private companies and states. Considering the hackable systems now in place throughout society and the economy—banking systems, social security, electric and transport systems, aviation and navigation systems, pension management, surveillance apparatus, healthcare systems and, of course, all the internet and the data in public and private servers—massively disruptive events at large or small scale are inevitable. Many political and social movements will be targeted. This can mean accounts erased, financial assets blockaded, and growing personal political repression via the suppression of communication capacities. This can also happen at a much bigger scale, targeting cities, countries or entire regions.

For the most important investments and political efforts, AI is being introduced as a labor replacement tool, a cultural hegemony monopoly creator, a military and surveillance weapon. Most of this is being done with people actively engaging and inviting the models into their everyday life (even more than it already was). The resistance to large data center projects is important and inspiring, but the overwhelming threat of AI goes well beyond its emissions, water consumption, and land occupation (although they plan on multiplying by many factors the current numbers, especially in Europe). Data centers today are nightmare factories.

So far, AI hasn’t been able to deliver on a key aspect: successfully automated processes that allow for the mass firing of people, substituted by effective algorithms. This is clear: 95% of all investment made by companies in AI has led to no profit, which is making capitalists nervous. But it hasn’t in any meaningful way stopped its spread.

When we say AI, we mean Machine Learning, Robotics, and Expert Systems. Currently AI is mostly a process of recognition, classification, and very high probability calculation, based on massive amounts of data with a good human interface. The interface is the most important trick for the general public. The public debate surrounding this issue is deeply anti-historical and anti-materialist, almost entirely it is white noise.

AI is not replicating or reproducing human intelligence. It is trying to encode human activities into repeatable procedures that can create reproducible algorithms. As it is not imitating our biological intelligence, it is trying to imitate what it can more or less “comprehend” about the previously referred algorithms—it is copying labor and social relations, their mechanisms and their predictable outcomes. Like other abstractions that rule our lives, such as money, algorithms produce real outcomes. AI ushers an irresistible promise for capitalist elites: to be able to automatically direct most of the instructions that guide human activity, reducing the power of social relations, in particular the power of the working class to impose political, social, and economic transformations that reduce inequality and injustice.

AI’s neural networks don’t mimic the human brain at all, but instead automate the “labor of perception,” classifying and interpreting written, numeric, and visual data and establishing associations. This creates a synthesis of knowledge, of the collective form of knowledge that comes from social cooperation. As explained before, another of its objectives has been to establish a monopoly over knowledge, scrapped from every website, database, online encyclopedia, and bite it is fed. It is then no wonder that Elon Musk and the far-right are going after Wikipedia.

These are some of the reasons why attempting to hard-code ethical rules or constraints into these models will not work, as they will not change the underlying political and economic functions of the data it is trained under and the algorithms generated and fabricated. Of course we understand that language itself is an algorithm, all the data as well and, of course, the internet as well. But with AI, we’re talking about a new level of control. The fundamental abstract purposes of AI as it exists now are the extension of quantification, control, and exploitation. The Labor Theory of Automation posits that AI is the result of a set of technological advancements that have abstracted automation to the point where it can automate itself. As we now have the technical ability to make such machines and capitalism has the economic incentive to massively deploy them, they want to use it to reorganize the division of labor even further in their favor. It is the apex of automation: Automation of Automation.

Facing such seemingly insurmountable odds, social and ruptural movements cannot but ask what to do about AI. There are basically two options: Drop out of the grid or acquire tech capabilities that allow us to resist the onslaught of these algorithms of collapse.