The Labors of AI

Summary

The underbelly of current machine learning (ML) and machine vision applications consists of millions of underpaid workers around the world. In this chapter, the author attempts to dive deeper into the multiple layers of labor conditions related to tasks related to ML and other artificial intelligence (AI) applications to demonstrate current working conditions. There are larger, and newer examples, such as the LAION 5‐B dataset, but the goal is to rely on the well‐documented historical account of the creation of a dataset in the context of labor. Despite the high number, research conducted in 2018 disclosed that of the over 100,000 workers subscribing to work on Mechanical Turk, only about 2000 can be considered active accounts working on tasks such as labeling and annotations. The creators of ImageNet created a precedent, with after effects to this very day, by ignoring the fact that MTurk is a crowd‐working platform.