An AI job boom? Here's what the tedious, temporary work in data labeling is actually like

Artificial intelligence (AI) technology is creating and eliminating jobs, but the unseen reality is that marginalized workers in developing nations perform laborious tasks for AI systems. These workers are often underpaid, overworked, and lack job security. Data labeling is a critical component of building AI models, as humans must categorize, label, test, and moderate vast amounts of data before AI can learn from it.

This low-skill, low-wage work is performed by a growing global workforce in data centers in China and Australia, among other places. Data workers often face precarious labor markets and low social status, which has pushed digitally literate young people into the industry. While some workers with advanced qualifications and certifications can earn higher rates, most perform basic tasks such as drawing bounding boxes around objects in images or annotating audio files.

These tasks often result in chronic eye strain, back pain, and insufficient pay to cover daily expenses. Data work is similar to other gig economy jobs, with no formal contracts or employee status. Instead, workers are classified as "users" and can be dismissed at any time. As AI advances, data work becomes more difficult and time-consuming, with many workers receiving less work over time.

The future of these workers is uncertain, as AI may eventually replace their jobs, but many are content with their situation due to the lack of viable alternatives.