Global Network of Micro Workers Feed the AI Pattern Matching Industry

20.03.2024|Christian Kreutz

The AI industry has already created a global network of companies linked through a myrad of subscontractors as Antonio Casili and his team uncovers in a global research.

Casili conducted a field research in Venezuela and Madagascar, who millions of micro task workers not rarely with higher qualifications (e.g. academic background) to improve AI models. Content moderators are one part of the group and their difficult working conditions were widely reported. Another group are data checker, who are critical in improving AI models such as the like of ChatGPT. It is way too little acknowledged that a human check of such models and the improvement of data are critical for higher quality of AI models. Casili reports how a global wide network of companies have bind millions of micro task workers what once started with Amazons mechnical turk. They are fixing for a few cents per micro tasks the model output and increasingly make these models more "intelligent". Without these microtasks the model would lack precision and as they often lack data to work better.

According to Casili workers from around the globe know pretty well how important their contribution is and how much money is made with these models at the other end of the supply chain. Across the line they complain about the working conditions like certification processes that are critical to apply for jobs but do not offer any jub guarantee.

Source: German: