What is the impact of the high demand for artificial intelligence (AI) training data for autonomous vehicles on the working conditions of crowdworkers? This chapter will put an emphasis on the planetary dimensions of this particular outsourcing stack, its structural aspects, its layered and siloed qualities, its fractal and redundant features, as well as its fragilities and uncertainties for the different stakeholders.

The analysis focuses on the intersection of four interdependent developments that culminated in 2018: First of all, there was the race of several dozen very well financed automotive and technology companies to be the first to bring self-driving cars onto the streets. Second came the ensuing unprecedented demand for vast amounts of highly accurate training data necessary to teach the cars how to navigate traffic based on vision. Third, there was the restructuring of large crowdsourcing platforms in order to cater to the specific needs of the automotive industry and orchestrate the required workforce for them. And finally came the crash of the Venezuelan economy, because of which Venezuelans inadvertently ended up providing the brunt of the work for this mammoth task of manual image labeling in the service of AI development.