According to researchers, the gleaming edifice of the artificial intelligence industry is built on a foundation that resembles a “pyramid scheme of human labor.” At the top are a few highly-paid engineers and executives. Below them is a vast and invisible workforce of contractors and subcontractors who perform the essential labor but see little of the rewards. This structure is the industry’s hidden secret to success.
At the base of the pyramid are data labelers, often in developing countries, who perform the most basic tasks for the lowest pay. In the middle are the “raters” and “trainers,” the focus of recent scrutiny. These workers, often in the US and Europe, are more skilled and better paid than the labelers, but are still treated as a disposable workforce by the tech giants at the pinnacle.
This layered system of contracting allows the companies at the top to distance themselves from the realities of the labor. They are not the direct employers of the raters, so they can avoid responsibility for working conditions, mental health support, and job security. The contracting firms, in turn, are squeezed by the tech giants to provide labor at the lowest possible cost, a pressure that is passed down to the workers.
This structure is highly efficient at extracting value and concentrating it at the top. The intelligence and hard work of thousands of people are funneled upwards to create a product that generates immense wealth for a select few. It’s a model that is both deeply effective and profoundly inequitable.