By Daphne Barroeta
April 22, 2024

Part 1: Digital labor and material

What is AI — artificial intelligence? According to the University of Illinois, Chicago, artificial intelligence describes a type of software design that “enables machines to learn from collective human experience, adapt to new inputs, and execute tasks resembling human capabilities.

By leveraging AI technologies, computers can undergo training to perform particular tasks through the analysis of extensive data sets and the identification of patterns within the data.” (tinyurl.com/334b9r2k)

AI has been a subject of research since a summer research project at Dartmouth initially published in 1955. (tinyurl.com/muwd59xd)

Generally, the form of machine intelligence described in news articles and encountered in daily life operates by taking a large data set and iterating over it, exporting suggested solutions to a problem at each iteration and receiving human feedback with each batch. Each iteration − or repetition of a step of the process − brings the process closer to a final acceptable solution. This is not the only way to design AI, but it is often used.

To give a concrete example of one of the major processes mentioned above, let’s say there is a computer program designed to recognize the handwritten letter “A.” The researcher would assemble a collection of scans of samples of handwritten letters, including but not limited to the letter “A.” Every time the program successfully identifies “A” it gets a passing grade, and every time it misidentifies something else as an “A” it gets a failing grade.

As the program goes through batch after batch of this process, it slowly learns how to identify the letter “A” with a significant degree of accuracy. The process of grading and guiding the program starts off as a human-directed process and eventually can be transitioned into an automatic process.

Built-in bias

An important thing to take note of is that a dataset of some type is acquired and used as the foundation for building the “understanding” of what constitutes “truth” within the context of some objective. When it comes to AI systems that handle data relating to human demographics, this is an area where the underlying biases of humans often come into play.

A classic example comes in the form of facial recognition software. “Facial recognition software has a long history of failing to recognize Black faces. Researchers and users have identified anti-Black biases in AI applications ranging from hiring to robots to loans” (Vox, June 2023).

Capitalism not only facilitates the perpetuation of biases in algorithmic technology in the pursuit of the profit motive, it also exploits the workers who develop and train these systems.

Increasing exploitation of workers

As a former Tesla worker in the autopilot department, I experienced first hand the factory process laborers are subjected to and the way in which the exploitation of our labor physically and psychologically harmed the workforce.

Data annotators — the people who process data for machine learning — were tracked using keystroke detection software, and their productivity was ranked in an automated system. Workers who were not putting in at least 6.5 hours of keystrokes in an 8 hour day were subjected to disciplinary action, held back from pay raises and even terminated. Obtaining a medical exemption in cases when the individual had migraines, ADHD or a medical condition that required bathroom or insulin breaks was a near impossibility.

The bourgeois image of the computer programmer who owns a part-time business and works at Google where they have lemon water and cucumber sandwiches on demand is largely a myth — a myth that benefits the tech firm. Make no mistake, while there are petit bourgeois individuals who work in the tech industry, the general trend in computer programming labor is a trend towards industrializing computerized labor so that proletarians can be exploited and paid lower wages.

While this trend has been brewing since the 1990s when the personal computer exploded on the market and transformed the office landscape, it has only been within the last decade or so that this trend has really kicked off — with the COVID-19 pandemic further accelerating this process. This trend is also not limited to the Global North. Companies like OpenAI (developer of ChatGPT) and Amazon Web Services rely on the cheapest paid labor they can exploit, anywhere in the world.

Reliance on gig labor

A platform is a packaging of software and data meant to control the inputs and outputs to create a desired experience or manufacture useful information or outcomes. One of the major styles of platforms organizing computer labor for the benefit of the tech-capitalists is called “the Mobile T- -k.” The term refers to an old and racist device called the “Mechanical T- -k” in which a hidden human operator would pilot a “robotic” puppet which was dressed as a stereotypical facsimile of a Turkish man.

The reference to the human operated “robot” (puppet) stands today as a bourgeois fantasy image for computer labor that abstracts the role of the human laborer to the human users of a platform (like Facebook or Coursera) so that the capitalists can super-exploit the laborers and create a false sense of both ethical practice and intellectual savvy on the part of the business — obviously a lie that needs to be dispelled. A strong example can be found in the book “Ghost Work” which explains “invisible” computer labor to readers in the Global North.

The following excerpt is discussing a family in India and their experience with one of Microsoft’s gig tools:

“In Bangalore, India, Kala works from her makeshift home office, tucked away in the corner of her bedroom … Kala picks up work from an outsourcing company that supplies staff to the Universal Human Relevance System (UHRS), an MTurk-like platform used internally by its builder, Microsoft. Kala, a 43-year-old housewife and mother of two with a bachelor’s degree in electrical engineering, calls her two teenage sons into the room, points to a word displayed inside a large text box on her LED monitor, and asks them, ‘Do you know what this word means? Is it something you shouldn’t say?’

“They giggle as she reads the text out loud to them. They make fun of her pronunciation of ‘chick flick.’ Together they decide that, no, this sentence does not contain adult content. Kala clicks ‘no’ on the screen, and the window refreshes with a new text phrase to read to her sons … 

“Though she’s typically unable to find enough tasks to fill more than 15 hours of work in a given week, Kala returns to UHRS almost every day to see if there are any new tasks that she feels qualified to do … ” [italics in the original]

“Companies like Google, Microsoft, Facebook, and Twitter use software to automatically remove as much ‘not safe for work’ content as they can, wherever possible. But these software filtering systems, powered by machine learning and artificial intelligence, aren’t perfect. They can’t always tell the difference between a thumb and a penis, let alone hate speech and sarcasm.” (“Ghost Work,” Gray and Suri)

The exploitation of labor globally is one of the three major avenues through which the ruling class uses artificial intelligence, and computer technology more broadly, to oppress and exploit the global proletariat. The other two ways are surveillance technology and mechanized warfare.

The working class must recognize our consistent relationship to these tools and enhance our study of them, so we can use them to our advantage in the fight against the ruling class. We must also be aware that the “digital” world and “digital” technology have a material basis, the physical components of and labor behind computing, which must be studied through the dialectical lens of Marxism-Leninism.

Machine labor and living labor

A useful description Marx gave for this kind of technology comes in his description of the difference between machine labor and living labor. To paraphrase sections of Marx’s “Grundrisse” (1861): “Living labor seeks to exchange its productive capacity for wages for its own survival, and machine labor serves to abstract some aspect of living labor and reduce it in other areas.”

Doing an analysis of digital technology also means recognizing that every minute of battery life, every minute of email, every minute of YouTube videos or Facebook feeds requires mass extraction of resources and human labor, produces carbon emissions contributing to ecological damage, steals water away from local municipalities to cool servers and produces digital “material” (information) for the ruling class to study and/or generate profits from.

These technologies can be useful to our class interests and do hold the potential for helping us to organize our lives in transformative ways, but the prerequisite to this is proletarian revolution. The working class and oppressed of the world must own and reform the means of production — including the digital means. Anything less kicks the cycle of exploitation around the globe.