English 
6 min of reading 
0 comments 
DoorDash Tasks pays couriers to film household tasks, like washing dishes and folding clothes, and uses the videos to train AI, humanoid robots, and automation systems.
According to TechCrunch, DoorDash announced on March 19, 2026, the launch of DoorDash Tasks, a standalone app that allows the platform’s 8 million American couriers to earn money by filming everyday tasks, such as folding clothes, hand-washing dishes, and making the bed. Payment is displayed before each task and varies according to complexity.
More difficult tasks, such as pruning and replanting plants, pay more than simple activities. The Washington Post reported that workers can earn up to $25 per hour with the available tasks, creating a new form of on-demand work linked to data collection for artificial intelligence.
The collected data will be used to train DoorDash’s own AI and robotics models and those of partners in the retail, insurance, hospitality, and technology sectors. Potential interested parties include humanoid robot manufacturers like Tesla, Figure AI, and Agility Robotics.
-
The government is studying a new phase of Desenrola Brasil to benefit those who pay their bills on time.
-
Government launches panel to monitor pesticides in rivers and expand environmental control over water resources in Brazil
-
A man builds a homemade raft with twelve 50-liter containers, spends R$ 260, and creates a fishing platform that supports four people on a calm lagoon without needing to buy a traditional boat for weekend leisure, with a simple structure.
-
Chile blocks billion-dollar mining megaproject near Humboldt penguin colony, halts iron and copper mine and turns endangered birds into an environmental symbol
DoorDash Tasks turns household tasks into data for artificial intelligence and robotics
One of the tasks documented by Bloomberg asks the worker to use a camera pointed at their hands while scrubbing at least five dishes. Each cleaned dish must be held in front of the lens for a few seconds before the worker continues the recording.
The app was launched across the United States, except for California, New York, Seattle, and Colorado, regions with stricter rules on data privacy and platform worker rights. This exclusion is one of the most revealing details of the launch.
DoorDash is turning couriers into data collectors of the physical world, using real household tasks to feed AI systems that need to learn how humans manipulate objects, clean surfaces, fold fabrics, and organize environments.
Robots need to see humans washing dishes because there is no internet for physical data
The central problem that DoorDash Tasks tries to solve is not hardware, but data. Chatbots like ChatGPT have learned from trillions of words available on the internet, but robots that need to manipulate physical objects don’t have such a broad and accessible equivalent.
“There is no internet for robotics data,” said Ken Goldberg, a roboticist at UC Berkeley, to the Washington Post. A robot folding a shirt needs to learn about force, angle, texture, fabric reaction, and error correction in real situations.
Computer simulations still do not accurately reproduce the variability of kitchens, clothes, dishes, plants, and household objects. Therefore, videos recorded by workers in real homes may be more valuable than clean, expensive, and controlled laboratory data.
Worker videos are cheaper than specialized robotic teleoperation
Researcher Simar Kareer from Georgia Tech explained that robotic teleoperation data usually has higher quality because it includes real robot movement commands. The problem is that this type of collection requires expensive equipment, trained operators, and a lot of time.
Filming humans performing household tasks is cheaper, faster, and more scalable. Paying workers to record movements with cell phones, head cameras, or body mounts creates a volume of data that robotics labs could hardly produce on their own.
According to estimates cited by the Washington Post, one hour of human video performing tasks can cost less than $5, while robotic teleoperation data can exceed $100 per hour. This cost difference explains why AI companies are turning to gig workers.
DoorDash uses 8 million Dashers as a sensor network for the physical world
The logic of DoorDash Tasks goes beyond selling robotics data. The company has spent more than a decade building an infrastructure capable of deploying workers to specific locations, confirming tasks, and processing payments on a large scale.
Ethan Beatty, general manager of DoorDash Tasks, stated that there are more than 8 million Dashers capable of reaching almost anywhere in the United States and interested in earning flexibly beyond deliveries. For him, this is a powerful capability to digitize the physical world.
This human network is difficult to replicate. No traditional data company can quickly assemble a base of millions of people distributed across the country, available to film tasks, verify local events, or act as physical support for automated systems.
Privacy of domestic data is still the biggest question about DoorDash Tasks
The launch did not publicly address central questions about consent, retention, and use of the data. It is unclear how long the recordings will be stored, what rights workers will have over images of their own homes, and who exactly will receive the videos.
Reports from The Next Web and IBTimes highlighted these omissions as relevant. In a program that encourages workers to bring cameras into kitchens, bedrooms, laundries, and private areas, the absence of details is not a minor point.
The issue is not just labor-related, but also domestic and familial. Videos made inside homes can capture personal objects, voices, faces, routines, and sensitive information that go beyond the contracted task.
How much is a human washing dishes worth to train robots
The market value of this data helps explain the rush. Human video data is cheap, abundant, and captures real variations of environment, lighting, objects, posture, rhythm, and improvisation.
For robotics, this is decisive. A domestic robot doesn’t just need to know that a plate is dirty; it needs to understand how a hand holds the plate, what force is applied, how the movement changes when there is foam, grease, or running water.
The bet of DoorDash Tasks is mathematical: paying up to $25 per hour for humans to film tasks can generate enough data to replace part of the much more expensive robotic data, especially when combined with smaller sets of specialized teleoperation.
DoorDash Tasks exposes the paradox of workers training the automation that may replace them
The structural irony of DoorDash Tasks is that workers vulnerable to automation are being recruited to build part of this automation. A delivery person can earn extra money filming household tasks while simultaneously feeding models that may eventually reduce the demand for human labor.
The process involves two fronts: training domestic robots to perform activities like washing dishes, folding clothes, and making beds, and supporting automated systems in delivery, retail, hospitality, and mobility.
Professor Antonio Casilli, from the Institut Polytechnique de Paris, summarized the risk by saying that the danger is not just robots taking human jobs, but humans having to work for robots. For now, the task pays up to $25 per hour. The question is how long this balance will last.
The new digital work has left the screen and entered the workers’ kitchens
The DoorDash Tasks show a significant change in the data market for artificial intelligence. The first phase of AI consumed texts, images, and videos available on the internet; the next requires human movements in real physical environments.
Washing dishes, folding clothes, closing robotaxi doors, and making beds seem like mundane activities, but they have become raw material for AI models and humanoid robots. The human body, filmed in common tasks, has become a valuable source of training for machines.
The question now is not just how much these tasks pay, but who controls the data, who profits from the trained models, and which workers will be benefited or replaced when automation learns exactly what they were paid to teach.
Be the first to react!