Data Labelers Fuel AI Boom But Face Uncertain Prospects

Data annotation involves labeling data to train AI systems. Demand has grown as more companies build AI models. Workers can make $20/hr or more, but the industry is poorly regulated.

To get started, workers must pass an assessment that can take 1-3 hours. If accepted, they start receiving paid tasks. But some report never getting work after passing assessments.

Many report positive experiences and good pay, but some have accounts deactivated with large unpaid earnings. Sites often use algorithmic management which can result in poor treatment.

Some data annotation sites are actually subsidiaries of larger companies like Scale AI and Surge AI. This secrecy makes it harder to link companies to potentially exploitative labor practices.

The data annotation industry is fundamentally undervalued, paying workers by the task rather than sharing royalties, even though it is workers' intelligence training lucrative AI systems.