Uber Launches Program to Let Drivers Train A.I. Models In Their Downtime
The life of an Uber driver often involves stretches of downtime—waiting on ride requests or charging an electric vehicle’s battery. To make the most of those idle moments, Uber is launching a pilot program that allows drivers and couriers to make extra money by completing digital tasks that train A.I. models for Uber’s enterprise clients.
“Drivers have asked for more ways to earn, even when they’re not on the road,” Uber CEO Dara Khosrowshahi said in a statement. To address this request, drivers will soon be able to opt in for quick, in-app tasks ranging from uploading documents—such as restaurant menus or receipts—to providing everyday images and recording audio samples.
The pilot will launch later this fall as part of Uber’s AI Solutions Group, a division created last November to offer data-labeling services to other businesses. Its client list includes Aurora, a self-driving software developer; Niantic, the company behind Pokémon Go; and Luma AI, a text-to-video generator. Until now, Uber AI Solutions has relied on independent gig workers to complete data-labeling tasks. The new program shifts those assignments to Uber’s own network of drivers and couriers, giving them access to additional income streams directly through the Driver app.
In addition to the upcoming U.S. launch, Uber has already been testing the initiative in more than 12 cities in India. “Until now, these tasks were completed by independent contractors outside the app,” said Megha Yethadka, the global head of Uber AI Solutions, in a September LinkedIn post describing the Indian pilot as “very promising.”
Before accepting a task, drivers will be able to see the expected pay rate and estimated completion time. They can only take on digital tasks while not actively signed in to drive or deliver for Uber.
While data-labeling is a relatively new area for Uber, it’s long been a critical part of A.I. development. One of the largest players in the space is Scale AI, which was valued at $29 billion earlier this year following a $14 billion investment from Meta. Other players include Surge AI, which counts Anthropic and Microsoft amongst its clients, and in-house data-labelling initiatives run by model developers like xAI.
Khosrowshahi first discussed Uber’s plans to introduce digital tasks at the Bloomberg Tech Summit in June, where he laid out a strategy to expand income opportunities of drivers and couriers over the next five to ten years. He described the data-labeling effort as a form of “knowledge work” emerging from the A.I. era and a way to provide new job options even as automation and autonomous vehicles threaten traditional driving roles.
Uber announced the digital tasks initiative yesterday (Oct. 16) during its annual Only on Uber event, which highlights new features inspired by driver and courier feedback. Other updates unveiled at the event included a new heat map tool showing demand hotspots, a rider rating filter that allows drivers to screen trip requests, and a delayed-ride guarantee offering extra pay when trips take longer than estimated.
Uber also announced an expansion of its women rider preference feature, which lets female drivers accept rides only from women passengers—a setting that has been used for more than 150 million trips and is activated weekly by one in four female drivers.
