Millions of workers train AI models for pennies
In 2016, Oskarina Fuentes got a tip from a friend that seemed too good to be true. Her life in Venezuela had become a struggle: Inflation had reached 800 percent under President Nicolás Maduro, and 26-year-old Fuentes didn’t have a steady job and had to juggle multiple side hustles to survive.
Her friend told her about Appen, an Australian data services company that was looking to crowdsource workers to tag training data for artificial intelligence algorithms. Most internet users have done some form of data labeling: identifying images of traffic lights and buses for online captchas. But the algorithms that power new bots that can pass legal exams, create amazing images in seconds or remove harmful content on social media are trained on datasets – images, videos and text – collected by gig economy workers on some of the cheapest labor markets in the world.
Appen’s customers include Amazon, Facebook, Google and Microsoft, and the company’s 1 million employees are just one part of a huge, hidden industry. According to consulting firm Grand View Research, the global data capture and labeling market was valued at $2.22 billion in 2022 and is expected to grow to $17.1 billion by 2030. As Venezuela descended into economic disaster, many college-educated Venezuelans like Fuentes and her friends joined crowdsourcing platforms like Appen.
For a while, it was a lifeline: Appen meant Fuentes could work from home at any time of the day. But then there were power outages – the power went out for days. Left in the dark, Fuentes was unable to take on tasks. “I couldn’t take it anymore,” she says in Spanish. “In Venezuela you don’t live, you survive.” Fuentes and her family emigrated to Colombia. Today she lives in an apartment in the Antioquia region with her mother, grandmother, uncles and dog.
Appen is still their only source of income. Pay ranges from 2.2 cents to 50 cents per task, Fuentes says. Typically, an hour and a half of work earns $1. If there are enough tasks to work a full week, she earns about $280 a month, almost equal to the Colombian minimum wage of $285. But filling a week with tasks is rare, she says. Lost days, which are becoming increasingly common, yield no more than $1 to $2. Fuentes works from her bed on a laptop and is glued to her computer for over 18 hours a day to make the first selection of tasks that could arrive at any time. Given Appen’s international clients, the days start when the assignments come out, which can mean starting at 2 a.m.
It is a pattern repeated in developing countries. Marking hot spots in East Africa, Venezuela, India, the Philippines, and even refugee camps in Kenya and the Shatila camps in Lebanon provide cheap labor. Workers complete microtasks for pennies each on platforms like Appen, Clickworker and Scale AI, or sign short-term contracts in physical data centers like Sama’s 3,000-person office in Nairobi, Kenya, which is the subject of a Time Investigating the exploitation of content moderators. The AI boom in these places is no coincidence, says Florian Schmidt, author of Digital labor markets in the platform economy. “The industry can flexibly move to where wages are lowest,” he says, and can do so much more quickly than, for example, textile manufacturers.
