I Tried DoorDash’s Tasks App and Saw the Bleak Future of AI Gig Work

Imagine that instead of delivering a hot pizza to a customer's door, you spend the afternoon in your kitchen, recording the process of frying scrambled eggs with your phone. However, you aren't doing this for reach on TikTok or Instagram. You are doing it for DoorDash, and your only viewer is an algorithm learning to recognize the texture of setting egg whites. This is the new reality of the gig economy, where the line between a physical service and training artificial intelligence is finally blurring.

The food delivery giant is quietly testing the Tasks app, which sheds new light on how large corporations intend to acquire data to train their vision models. This is no longer just the work of a courier; it is the role of a "machine teacher," paid at rates that, when compared to the technical requirements, raise serious ethical and economic doubts. Testing this tool allows us to understand the direction in which the global labor market driven by AI is headed.

The comfort of home as a testing ground for AI

The DoorDash Tasks app differs from the standard interface for couriers. Instead of a map with pickup points, the user sees a list of tasks that resemble bizarre social media challenges. Examples? Recording a video showing laundry being sorted, walking through a park with the phone pointed at shoes, or the previously mentioned meal preparation. Each of these tasks has one goal: to provide high-quality, diverse video data that will be used to train Computer Vision.

Unlike static images, video provides AI models with information about movement dynamics, lighting changes, and interaction with objects in real-time. DoorDash, possessing an army of millions of "dashers," has a unique advantage over research laboratories — access to authentic, unscripted scenes from everyday life in thousands of different locations. This is data that cannot be generated synthetically in a way that is sufficiently reliable for advanced autonomous systems.

• Video tasks: Recording everyday activities (cooking, cleaning, walking).
• Precise guidelines: Requirements regarding camera angle, lighting, and the absence of visible faces of third parties.
• Low barriers to entry: No need for specialized equipment other than a smartphone.

Micro-earnings for macro-data

The most controversial aspect of Tasks is the compensation structure. While traditional deliveries in the gig model rely on trip rates and tips, AI training is valued extremely low. Users report rates of a few dollars for tasks requiring plan preparation, recording, and uploading heavy video files. After accounting for the time needed to review instructions, the real hourly rate often falls below the minimum wage in many regions.

This phenomenon is a classic data sweatshop (digital exploitation), but in a new, corporate version. DoorDash utilizes its existing workforce infrastructure to bypass expensive data annotation companies. For the worker, it is an illusion of easy earnings "without leaving home," which in reality is a transfer of unique knowledge about human behavior to systems that may automate those same professions in the future. Machine Learning feeds here on the efforts of people who are the least likely to benefit from technological progress.

"Gig economy workers are becoming the involuntary architects of their own obsolescence, providing the data that will allow algorithms to take over their current duties."

Technical challenges and privacy in the lens

From a technological standpoint, Tasks is a fascinating experiment in the field of data crowdsourcing. Collecting millions of hours of footage from different climate zones, apartment interiors, and weather conditions allows for the creation of AI models with unprecedented robustness. However, the price workers pay goes beyond low earnings. It is about privacy.

Recording the interior of one's own home, eating habits, or way of moving is handing over the most intimate biometric and behavioral data to a corporation. Although DoorDash assures data anonymization, the process of training neural networks often requires access to raw materials. There is a risk that data collected for one purpose (e.g., teaching delivery robots to recognize obstacles) will be used for consumer profiling or other, less transparent commercial purposes.

A new hierarchy in the tech world

The emergence of apps like Tasks suggests the rise of a new working class in the technology sector. These are no longer just engineers in Silicon Valley, but millions of people around the world performing repetitive, tedious actions in front of a smartphone camera. This is an extension of the Amazon Mechanical Turk model into the physical and video spheres.

In this new structure, the value of physical labor (moving a package) is equated with the informational value of that action. For the DoorDash platform, information about how a human avoids a pothole in the sidewalk may ultimately be more valuable than the fact of the delivery itself, as it allows for the scaling of autonomous technologies. We are witnessing a moment where the human becomes merely a "sensor" in a vast network gathering data for global AI players.

I predict that in the coming years, we will see a rapid increase in the number of platforms combining gig work with model training. Traditional service occupations will increasingly be supplemented with a "data collection" component, which will force regulators to redefine the concept of digital labor. If standards regarding minimum rates for data and the protection of biometric privacy are not introduced, we face a future where every action we take will have its (very low) market price in the databases of tech giants.