Your Uber driver has a new side hustle: Training AI for cash

Follow ZDNET: Add us as a preferred source on Google.

ZDNET’s key takeaways

• Uber launches a new, optional “digital tasks” program in the US.
• Drivers can upload content to help companies train AI models.
• Compensation depends on a task’s time commitment and complexity.

There’s more to being employed by Uber than simply driving around passengers — at least, that’s what Uber says. Following the expansion of its courier service in 2020, Uber is now offering drivers the opportunity to help companies train their AI models for extra cash.

How it works

According to Uber, beginning later this year, drivers and couriers who opt into the program can complete “digital tasks” within Uber’s Driver app. These tasks can include submitting a video of themselves speaking in their native language, uploading pictures of specific everyday items, or presenting documents written in a different language. 

Also: This free Google AI course could transform how you research and write – but act fast

After tasks are completed, the earnings will be in the users’ balance within 24 hours. Compensation depends on the time commitment to complete tasks and their complexity.

Uber says that over time, more digital tasks across a broader range of requests will be available, giving users the opportunity to earn more money without having to drive their car.

Also: You’re reading more AI-generated content than you think

In a blog post, Uber says this program connects users to “companies that need real people to help improve their technology” — as in AI companies that need real-life resources to train their models. Opportunities to complete digital tasks depend on the demand from the companies that need them, so it may not be a reliable source of income.

Evading copyright issues 

Calling on everyday people to submit photos, videos, written documents, and voice clips could be a workaround for AI companies hoping to circumvent the copyright legal battle between themselves and publishers. Typically, AI companies scrape training data in the form of text, audio, video, images, and more off the open internet; as a result, record labels, social media companies, newspapers, independent artists, and movie studios have sued AI companies like OpenAI, Suno, and Anthropic, usually claiming copyright infringement. 

Also: Is art dead? What Sora 2 means for your rights, creativity, and legal risk

Instead of training an AI image generator on professional, credited photos of bananas, perhaps the average person can help in exchange for a few extra dollars. It’s a gig-economy spin on the existing method of training AI models, in which companies pay workers in the global south very low wages to tag and sort data for models to ingest. 

Privacy concerns 

It’s unclear how much of a cut Uber will take from participants in this program, and the privacy policy around the program is also unclear. Uber says that it will not disclose the names or business goals of AI companies to participants. You can assume that AI companies may sell, transfer, or retain any content received in the program.