Anatomy of the synthetic proletariat - 2

Last week, we defined the concept of the "Synthetic Proletariat," in which humans are dispossessed by their own data. The concept of the synthetic proletariat is a tragic mutation of the "prosumer" definition—conceptualized by Alvin Toffler in the 80s with an optimistic outlook—which has increasingly suffered under surveillance and platform capitalism due to artificial intelligence. This week, we will map out the digital landscape of this synthetic proletariat. 

Layer 1: Ghost Workers

Ghost workers, who scrub traumatic data to make artificial intelligence (AI) appear human, are the segment that performs the heaviest emotional labor in mostly invisible factories, receiving the lowest wages while ensuring the hygiene of the algorithmic system. According to data that emerged in 2023, companies like OpenAI (ChatGPT) and Invisible Tech paid workers in Kenya and Thailand between $1.32 and $2 per hour to tame their models and filter out toxic content.  Ghost workers were forced to label the most severe content, ranging from child abuse to descriptions of torture. 

Layer 2: Digital Signal Workers 

The mass of people who establish the logic of algorithms and label the world for them. For example, thousands of young people in the Philippines and Venezuela spent 15 hours a day marking road lines, traffic signs, and pedestrians, or labeling images via the Remotasks platform of the company Scale AI, so that autonomous vehicles could perceive the world. 

Layer 3: Click Workers and Micro-Taskers 

This category includes the vast masses who generate data through "I am not a robot" tests that verify the accuracy of algorithms, as well as gig-economy workers who are precarious, work on a piece-rate basis, and are constantly rated by an algorithmic manager. This also includes millions of global workers who fill out surveys lasting seconds or transcribe audio recordings under the guise of "crowdsourcing" on platforms like Amazon Mechanical Turk (MTurk), as well as work-from-home ad evaluators used by Appen, Telus, Google, and Meta to improve search results.

Layer 4: Raw Content Providers (Voluntary Collective Labor)

The majority who make the internet what it is, yet whose labor is nationalized by being turned into data. A vast mass whose labor is hidden under the label of 'user' and who have transferred the ownership of their collective cultural production to platform owners. The sale of the values we produce as a community is an example of the 'enclosure' of digital dispossession. Examples of this include the data of communities that have shared information for free for years being sold by the owners of platforms like Reddit and Stack Overflow to Google and OpenAI for millions of dollars, and Elon Musk using the entire history of the X platform to train his own artificial intelligence, Grok.

Layer 5: Theft of Creative High-Quality Labor

The theft of the styles of artists, writers, and academics. Professionals whose cognitive capital is devalued by being copied by machines. The use of photographers' images by Adobe Stock and Getty Images to train image generation models without permission, and the licensing of news archives produced by journalists for years by global media organizations (Axel Springer, News Corp, etc.) to OpenAI  are the most current examples.  From an economic-political perspective, the most critical and ironic point is that AI capital is dependent on organic human data.

According to a study published in the journal Nature by researchers from the universities of Oxford, Cambridge, and Toronto, when an AI model is trained on the output of another model, it forgets reality after a few generations and begins to produce nonsense.  This is the law of diminishing informational returns in the digital economy. Here, the hidden power of the synthetic proletariat is revealed: when human labor is reduced or completely eliminated, AI-generated platforms turn into digital junkyards. In other words, for AI to sustain its existence, it is ironically very much in need of the organic human labor it exploits. 

In short, behind the shining storefront of the AI era lies a synthetic proletariat growing in layers. It is a global labor chain that stretches from the ghost data workers who teach AI ethical boundaries and make it good and smart, to volunteer data producers, and from there to the dispossession of creative and collective labor. This picture, which includes psychological trauma, low wages, the usurpation of common knowledge property, the theft of style, creativity, originality, and talent, and much more, demonstrates the systematic re-commodification of human knowledge and labor in the age of data colonialism, digital enclosure, and platform capitalism.

How can a line of defense be established against the form that Negri calls "immaterial labor" merging with Standing's "precariat" for AI and AI systems, and the synthetization of labor in its social form of existence within Baudrillard's simulation universe? Is it a dream to stop cognitive dispossession and reclaim the sovereignty of data?