The artificial intelligence industry continues to grow and cause discussion. Giants such as Amazon, Facebook, Google and Microsoft have presented tools to understand, analyze and create images and text in their most used services. But behind these advanced AI algorithms there is an army of workersOften underpaid, which train and label the data that powers AI models. Ad hoc companies recruit collaborators in Southeast Asian and Latin American countries, paying them a few cents.
Underpaid workers to train AI, labeling data
The investigation by the American editorial team of Wired underlines the enormity of this phenomenon starting from a particular story, that of Oskarina Fuentes and the Australian data services company Appenone of the major players in the industry of data labeling for training AI algorithms. It has over a million collaborators, many of whom perform “microtasks”. They tag images or texts to make money a few cents for each completed task. In an industry that was already worth 2.22 billion dollars in 2022 and will be worth 17.1 billion in 2030.
According to Wired, Appen works for Amazon, Meta (Facebook), Google e Microsoft: All the major Silicon Valley companies that are investing in AI. All companies are among the top ten in the world in terms of financial value: if we exclude Meta (tenth at 765.28 billion dollars), the others are among the top five with over a trillion dollars in value. But the story of Oskarina Fuentes and the other underpaid workers to train AI does not amount to trillions of dollars, rather it is counted in cents.
Eighteen hours a day to avoid making the minimum wage
Oskarina Fuentes began working for Appen with enthusiasm, also due to the country’s runaway inflation and the problems with the electricity grid that led her to emigrate to Colombia, where she continues to work for Appen. But the economic reality of her job soon dampened her enthusiasm. Her salary varies from 2.2 cents to 50 cents per taskand typically earns approx $280 a monthwhich is just below Colombia’s minimum wage.
His job, however, depends on how much homework he can get in a week. So he spends 18 hours on his laptop every day, sometimes starting at two in the morning in order to accommodate the first requests from Appen’s international customers.
In addition to platforms like Appen, there are companies like Clickworker e Scale AI who find collaborators in East Africa, in Venezuela, in India, in the Philippines. Or there are data centers like the Sama one in Nairobi, with three thousand people, already the subject of a Time investigation for the very low salaries used to train OpenAI models (we told you about it here).
Low-paid workers and AI, all over the world
Most workers live in the “Global South” and it is difficult to think of any other reason other than low labor costs. In Venezuela, hyperinflation has caused the economy to plummet, with many educated workers having to accept even very low wages to cope with the economic crisis. The lower cost of living and unemployment rate make it possible to find low-paid workers for AI in India, Philippines, East Africa, and other countries.
Wired has heard Saiph Savage, director of the Civic AI lab at Northeastern University, who defines this approach as a form of “data colonialism”. Which explains that the rapid evolution of AI makes it difficult to develop skills that allow workers to advance their careers. If until a few months ago they were working on labeling images for autonomous driving, in recent months they have been working on ChatGPT-style language models.
Without timetables and continuity
The workers interviewed by Wired explain that the main problem is not so much the low rate, but… rather the payment system of the platforms. Many complain about the uncertainty surrounding when homework will be assigned, since forces you to stay near the computer all day. Although there is no additional compensation for the time spent waiting. One worker says he has to work long hours to earn a few dollars, describing this situation as “digital slavery”.
Appen told Wired that he works for reduce the amount of time spent searching for tasks, balancing customer needs for fast-paced business with the need to provide a steady source of income to associates.
But as Fuentes points out, after working for Appen for seven years he still has to sit in front of a screen all day to earn a few dollars. If he had been working directly for one of the Silicon Valley companies all this time, What would your salary be today? And he closes by hoping for unionisation of the sector and says: “I would like them to consider us not only as work tools to be thrown away when we are no longer needed, but as human beings who help them in their technological advancement”.
If AI is the future, these workers are making it possible. For a few cents.