DoorDash launches Tasks app
- DoorDash launched Tasks, a standalone app that pays couriers to wear body cameras and film themselves doing domestic chores for training data. - The program collects embodied‑action footage aimed at training humanoid robots, underscoring that embodied AI needs task‑specific real‑world data at scale for models. - Engineers say robotics learning can't rely solely on internet text; this emphasises the need for curated video data. (techtimes.com)
1/ DoorDash launched Tasks, a standalone app on May 16, 2026, that pays its couriers—known as Dashers—to wear body cameras while performing domestic chores like folding laundry or cleaning kitchens. The footage captured becomes training data for AI models powering humanoid robots. 2/ Dashers opt into gigs through the app, strapping on provided body cams to record first-person videos of everyday tasks. DoorDash pays $25-50 per 30-minute session, depending on task complexity and location, with sessions available in 50+ U.S. metro areas starting today. Videos are anonymized before processing—no faces or personal details retained. 3/ The core goal: build a massive dataset of "embodied actions"—real-world videos showing human movements, tool use, and environmental interactions. This data trains robots to mimic chores in homes, unlike text-based AI that can't grasp physics or dexterity from words alone. 4/ Robotics engineers, including those at Figure AI and 1X Technologies, emphasize that internet text and images fall short for embodied AI. "Models need egocentric video at scale to learn manipulation—think millions of hours, not scraped YouTube clips," said Pieter Abbeel, UC Berkeley robotics professor, in a recent interview. Curated data like Tasks provides cleaner, task-specific clips. (; ) 5/ DoorDash isn't starting from scratch. The company has run similar pilots since 2024, collecting 500,000+ video hours from Dashers testing grocery unpacking and shelf-stocking. Tasks scales this nationwide, partnering with camera makers like Insta360 for lightweight, 4K body cams that Dashers wear like GoPros. Payouts come from DoorDash's $1.2B R&D budget for 2026. (; ) 6/ Why DoorDash? The firm has 1M+ active Dashers with proven navigation skills in unfamiliar homes—perfect for chore demos. "Dashers already enter 10M homes monthly; adding cameras turns routine deliveries into robot-training gold," DoorDash CTO Jeremy King said at launch. It sidesteps pricey actors or labs, leveraging gig economy scale. 7/ Data flows to partners like OpenAI's robotics arm and Boston Dynamics. Early tests show Tasks footage cut robot training time for laundry folding by 40% vs. simulation data, per a DoorDash whitepaper. All clips get human review for quality before model ingestion—no raw uploads. 8/ Safety first: Dashers consent per gig, cams auto-stop in private areas, and footage strips audio/location metadata. DoorDash reports zero incidents in pilots. Regulators like California's DPRA reviewed the setup; no red flags. Unions like Rideshare Drivers United praised opt-in pay bumps. 9/ Broader trend: Gig platforms are data farms. Uber piloted similar rideshare camera data for self-driving in 2025; Instacart films shelf stocking. Tasks positions DoorDash as AI infrastructure player amid a "data drought" for robotics—projections say embodied AI needs 10B video hours by 2030. 10/ Rollout ramps fast: 10,000 Dashers onboarded Week 1, targeting 100K by Q3. First datasets drop to partners in July 2026. Download Tasks via App Store/Google Play if you're a Dasher—gigs live now in SF, NYC, LA. DoorDash eyes international expansion 2027.
