DoorDash Pays Gig Workers to Film Themselves Doing Chores to Train AI Models
The new Tasks app turns everyday activities into training data and offers a glimpse of where the gig economy is headed
The concept is simple but unsettling: humans performing mundane tasks on camera so that AI systems can learn to replicate those tasks. WIRED's hands-on review describes the experience as a window into the bleak future of AI gig work, where the value of human labour is reduced to its utility as machine learning input.
The Tasks app represents a growing trend of companies finding creative ways to source the vast quantities of real-world data needed to train multimodal AI models. Rather than relying on synthetic data or professional studios, DoorDash is leveraging its existing gig workforce to capture authentic human behaviour at scale.
The pay structure and working conditions raise familiar questions about gig economy exploitation, now applied to an entirely new category of work that exists solely to make human workers eventually unnecessary.
Analysis
Why This Matters
This is perhaps the most literal example yet of workers being paid to train their own replacements. It raises profound questions about the trajectory of AI-era labour.
Background
AI companies have been paying for human-generated training data through various means, from Mechanical Turk to RLHF contractors. DoorDash's approach is novel in using its delivery network for data collection.
Key Perspectives
Critics see this as exploitative — low-paid workers creating the training data that will eliminate their jobs. Supporters argue it creates new income opportunities in a transitional period.
What to Watch
Whether other gig platforms follow suit and how regulators respond to this new category of AI training labour.