He data labeling It is a necessary step so that learning models can understand them and thus learn. It’s the ScaleAI business, Alexandr Wang’s company recently valued at $29 billion. However, not everyone involved in data labeling enjoys this status. Much of this work is carried out by workers in poor countries, poorly paid and involves very unpleasant tasks.
what’s happening. The advancement of artificial intelligence requires an enormous amount of data labeling. They count in AFP that this work is usually done by workers who reside in impoverished countries such as Kenya, Colombia or India. In addition to being very poorly paid, the job often requires them to review very unpleasant images. For example, for an AI to write an autopsy report, taggers must view hundreds of images of real crimes.
The work. It consists of reviewing and labeling files, most often images. It does not require a degree, just knowing how to use a computer and demonstrating that you can think analytically. The ease of access means that many people in vulnerable situations turn to this type of work. The problem is that, to get a decent salary, they have to work long hours, up to 16 hours a day in some casesand also many times the content they must label is violent and extreme.
AI moderators. It is a situation similar to the one that moderators of different platforms have been denouncing for years. We recently talked about the lawsuit that a former Chaturbate moderator had imposed on the company. There are other cases like that of Facebook moderators in Barcelona who denounced the company due to the psychological trauma that filtering all that content caused them.
Invisible. The data labeling market generated 3.8 billion dollars in 2024 and is expected to grow to 17 billion in the next five years. However, those who make it possible work in very poor conditions. A Colombian worker tells AFP that data taggers are “like ghosts. No one knows we exist even though we are contributing to the technological progress of society.”
Better conditions. There was no legislation in Kenya, but data taggers have been organized to achieve regulation and have better working conditions. They denounce the lack of psychological support they receive and demand formal employment contracts, a fair salary that reflects their work and the fundamental right to rest. This mobilization seeks to guarantee a more dignified work environment and protect the rights of these essential workers in the artificial intelligence industry.
The platforms. The most mentioned is Remotasks, a subsidiary of Scale AI that has been the subject of protests in countries such as Kenya, Venezuela and the Philippines for defaults and problematic practices. The company defends himself and ensures that they offer “fair and competitive remuneration.” Last year closed its doors in Kenya after workers complained publicly. There are more like the Australian Appen or Sama, a subcontractor of companies like Meta and OpenAI that was sued in Kenya due to poor working conditions and also ceased its activity.
The human cost. There is growing concern about the environmental impact of artificial intelligence, that requires large amounts of energy to run, especially due to the training and operation of complex models. However, there is not only an energy and natural resource cost, but also a significant human cost that seems to be going more unnoticed.
Image | Christina Morillo, Pexels
GIPHY App Key not set. Please check settings