Editor's take: Human workers quietly underpin AI's apparent intelligence, performing low-paid labeling tasks that correct errors and refine outputs. As AI grows across industries, these micro-jobs are spreading rapidly, forming an invisible workforce that powers generative models and highlights the growing intersection of human labor and the expanding AI gig economy.
Uber recently introduced a new way for US drivers to earn extra income, launching what is essentially an AI-oriented version of Amazon Mechanical Turk. The company's new Work Hub allows drivers to complete brief, low-effort tasks while waiting for their next passenger.
These minute-long "jobs" include uploading documents, recording short audio clips, or labeling data for AI training. Uber says the new service gives drivers another way to earn money, functions offline while the car is parked, and requires no particular experience beyond using a smartphone.
Uber Chief Product Officer Sachin Kansal told Bloomberg that Uber will add more tasks over time, with pay depending on the effort and time required to complete them. He emphasized that the new AI jobs do not replace work lost to autonomous vehicles, adding that Uber's recent partnerships with AV firms are purely coincidental.
Before introducing it to select US drivers later this year, Uber is testing its new task-based program in India. Although the system could eventually expand to non-drivers, Kansal said the company's immediate priority is to engage drivers already using the Uber platform.
Human input to correct AI errors is becoming an increasingly significant part of the gig economy, and Uber has steadily expanded its investments in this area. In 2024, the company began offering data-labeling opportunities to independent contractors, aiming to address the tendency of AI models to hallucinate, generate low-quality content, or fabricate information with algorithmic confidence.
Meanwhile, critics lament the lack of regulation or safety measures for this growing workforce – known as "Turkers" in Amazon's parlance – tasked with making AI appear less error-prone than it actually is. Recent valuations of AI data-labeling startups such as Scale AI and Surge AI have reached $30 billion, suggesting that Uber and other gig economy companies will continue vying for a share of that market.