Uber for AI labeling – The Verge

Uber is using gig workers to get into the AI labeling business, Bloomberg reports. It’s a sign that the ridehail and delivery company is interested in expanding its independent contractor-powered business model to meet the demands of the fast-growing world of machine learning and large-language models.

The company’s new “Scaled Solutions” division claims it can connect businesses to “nuanced analysts, testers, and independent data operators” using its platform. It’s an extension of an internal team with members based in the US and India that do new feature testing and other tasks like converting restaurant menus to Uber Eats selections.

Uber was already using artificial intelligence and machine learning for its own business and is now making that available to others for a fee. The company is hiring gig workers for data labeling, testing, and localization for other companies, including Aurora, Luma AI, and Niantic.

A big behind-the-scenes reality of AI model training is that it requires many human workers to do tedious tasks such as selecting the most human-sounding chatbot results or labeling obstacles like pedestrians in self-driving car footage frame-by-frame.

Companies building AI models often hire workers in developing countries to do these tasks and pay small amounts per exercise. One engineer in India told Bloomberg they were tasked with comparing and rating the correctness of AI-generated responses to complex coding problems, and were paid 200 rupees per set, or about $2.37.

Uber is currently signing up people from Canada, India, Poland, Nicaragua, and the US, paying different amounts per completed task with earnings issued to workers monthly. It’s also looking for people with different cultural backgrounds to help make AI more adaptable in various markets.