An Uber logo.
Credit: Reuters Photo
Uber Technologies Inc.’s gig-economy workforce now includes programmers. The company is expanding beyond its rideshare roots to enter a hot new market: helping other businesses outsource some of their artificial intelligence development to independent contractors.
Its new AI training and data labeling division, called Scaled Solutions, builds on an internal team that tackles large-scale annotation tasks for Uber’s rideshare, food delivery and freight units. According to its website, Scaled Solutions has begun serving other companies that also need high-quality datasets. Clients include Aurora Innovation Inc., an Uber-backed firm that makes self-driving software for commercial trucks, and Niantic Inc., the game developer behind Pokémon Go.
Uber’s efforts to sell data labeling services have not previously been reported. The move could allow it to gain a piece of a growing market, as global companies rely on humans to vet data to train AI models. Scale AI Inc, which offers similar services, is valued at $14 billion, making it one of the hottest artificial intelligence startups.
The rideshare giant has plenty of experience recruiting contractors, as it has done for years with drivers and couriers. Now the company is betting that it can help other businesses by getting enough skilled workers who can label images, text and videos with context for machine learning models to recognize patterns and make accurate predictions and recommendations.
To perform this work for more companies, Uber this month started signing up contractors with various skills in India, the US, Canada, Poland and Nicaragua. Earnings will be determined by each task they complete and paid out monthly, according to the FAQ section of its onboarding website, which is separate from the platform for recruiting drivers and delivery couriers.
It also has corporate job postings for San Francisco, New York and Chicago-based account executives who will manage Uber’s relationship with Scaled Solutions’ enterprise customers.
“Having performed these tasks at scale over the past decade as part of our own growth, we deeply understand the needs of companies requiring these services,” an Uber spokesperson said in an emailed statement. Hiring independent contractors aligns “with our expertise as one of the world’s largest providers of flexible work opportunities,” the spokesperson added.
The firm declined to share revenue or customer numbers, or elaborate on how much it charges per project.
Uber’s internal Scaled Solutions team is based across the US and India, and has employed both humans and automation to help validate maps and ensure accurate translation to localize its rideshare product in various markets. The group has also digitized photos of restaurant menus for the Uber Eats app and tested new features across thousands of mobile devices.
Outside the organization, the team is helping Aurora ensure accuracy when its software classifies objects on the road, including pedestrians, debris and other vehicles, an Aurora spokesperson said. Uber’s contractors have also evaluated location data for Niantic, which is trying to build a 3D map of the world for its augmented reality games and other game developers, said Erin Schaefer, Niantic’s senior vice president overseeing game publishing and operations.
In addition, Uber plans to enlist workers to use their cultural knowledge to adapt products to local markets, or use their programming knowledge to provide feedback on an AI chatbot’s responses to software engineering questions.
Prior to opening up the platform for public registration, Uber this year began recruiting freelancers with coding or language skills to help evaluate outputs of large language models, the technology underpinning generative AI tools like OpenAI’s ChatGPT and Microsoft Corp.’s Copilot.
An India-based software engineer, who requested anonymity citing fear of reprisal, said they were given two AI-generated responses to a complex coding problem and asked to compare and rate the correctness of the AI’s answers based on several criteria. Those include functionality, efficiency and formatting of the code, according to documents seen by Bloomberg News. The engineer reported completing three questions for a fee of 200 rupees ($2.37) per set.
It’s unclear how wages will differ in the various locations where Uber is onboarding contractors for Scaled Solutions. The tasks are similar to some offered by existing annotation platforms like Scale AI’s Remotasks or Data Annotation Tech that have appealed to remote workers hoping to earn extra income through online gig work.
The industry has also come under fire for underpaying outsourced workers in developing countries. Remotasks, for example, claims on its website to offer as much as $18 per hour to English-speaking, US-based workers, but The Verge reported last year that a Remotasks worker in Nairobi was paid about $10 for annotating 8-hour-long video footage for self-driving cars.