DoorDash Tasks App Turns Gig Workers Into AI Training Data Collectors

The platform represents a notable shift in gig work toward direct AI training support, where workers strap cameras to their bodies to film household chores, cooking tasks, and navigation activities. According to DoorDash, this video data helps AI and robotic systems understand physical world interactions.

How DoorDash Tasks Works

The Tasks app offers five main categories of recording assignments: household chores, handiwork projects, cooking demonstrations, location navigation, and foreign language conversations. Workers receive upfront payment based on task complexity and effort, typically around $15 per hour with maximum time limits of 20 minutes per task.

Tasks range from simple activities like loading laundry or scrambling eggs to more complex projects like pouring cement or exploring public spaces. The app requires workers to keep their hands visible throughout recordings and follow strict guidelines about filming locations and consent.

Wired's testing revealed that basic tasks like loading a washing machine with 10 items of clothing paid approximately $0.37 for 90 seconds of work. Navigation tasks in parks proved challenging due to requirements to avoid filming people without consent, making compliance difficult in populated areas.

Geographic Limitations and Market Access

The DoorDash Tasks app launched with notable geographic restrictions, explicitly blocking residents of California, New York City, Seattle, and Colorado from participation. The selective rollout suggests potential regulatory or legal considerations around data collection and gig worker classification in these jurisdictions.

European AI teams should note that the platform's current US-only availability and state-specific restrictions may indicate future regulatory compliance challenges for similar services operating under GDPR and EU AI Act requirements.

Training Data Economics and AI Development

The platform addresses a critical need in AI development: high-quality human demonstration data for training computer vision and robotics models. Thousands of videos showing hand movements during specific tasks can improve robot manipulation capabilities through imitation learning techniques.

For AI builders, this approach offers an alternative to synthetic training data or expensive professional data collection services. However, the economics reveal significant cost pressures on workers, with complex tasks potentially earning less than minimum wage when factoring in setup time and equipment requirements.

Implications for AI Training Workflows

DoorDash Tasks represents a scalable model for collecting real-world training data that could influence how AI companies approach data acquisition. The platform's emphasis on hand visibility and specific task completion suggests targeting for robotic manipulation training rather than general computer vision applications.

Technical teams should consider the trade-offs between crowdsourced data collection and controlled laboratory environments for training data quality and consistency. The platform's success may indicate growing demand for human-in-the-loop training approaches as AI systems tackle more complex physical world tasks.

The DoorDash Tasks app demonstrates how traditional gig work platforms are evolving to support AI development infrastructure, though questions remain about worker compensation and long-term sustainability of human-powered training data collection. This analysis is based on Wired's hands-on testing and reporting of the platform.