Clean House in Exchange for Data: Shift Collects Videos to Train Humanoid Robots
Free house cleaning in exchange for useful data for artificial intelligence. This is the proposal from Shift, a German startup specializing in data collection for training AI systems, which has launched a service in New York aimed at transforming regular household chores into valuable material for the development of domestic robotics.
The offer is simple: an operator visits the client's home and for about two hours performs tasks such as cleaning bathrooms, washing floors, organizing the kitchen, folding laundry, or tidying the refrigerator. Throughout the work, the operator wears a head-mounted camera that records in first-person movements, actions, and interactions with the home environment.
Today, we're launching shift. We're starting by cleaning your apartment in New York City, for free. Here's how it works. Book a shift cleaning. A vetted shift operator comes to your home wearing one of our devices. They clean. They leave. You pay nothing. In exchange, we record…
According to the company, the economic value of the data collected is sufficient to fully cover the cost of the service. On its website, Shift emphasizes that there are no hidden costs: the cleaning is provided for free and the equivalent is represented solely by the data generated during the activity.
The goal is to build increasingly realistic datasets to train AI models and future robots capable of performing household tasks autonomously. The recording of activities allows for acquiring detailed information on human movements, operational sequences, and the ways in which real environments—often quite different from the controlled scenarios of laboratories—are approached.
The company claims to already operate in over 15 countries and collaborates with thousands of people engaged in collecting videos for AI training. In addition to the free cleaning service, Shift actually offers a platform that allows users to register while they work or engage in daily activities. Those who join the program can receive a recording device and use a dedicated app to upload footage, obtaining compensation that the company quantifies at about 20 dollars an hour.
The startup also states that it adopts specific measures to protect privacy. Before being used in training processes, videos are processed to blur sensitive elements and anonymize the content. The company states that the collected material is used exclusively for training AI models, without being published or used for advertising purposes.
A curious aspect concerns the quality of domestic environments. In the informational documentation published online, Shift explains that more complex situations can be particularly useful for training intelligent systems. Disordered or particularly challenging environments indeed provide richer and varied scenarios, helping to improve future robots' ability to adapt to real contexts.
The initiative is part of the growing race to develop humanoid robots and autonomous domestic assistants, which is increasing the demand for data from the real world. Companies like Tesla, Figure AI, and Agility Robotics are investing in the development of increasingly sophisticated robotic platforms, while the market dedicated to collecting the necessary data for training them is also growing.
The case highlights how the strategic value of data is assuming an increasingly central role in the economy of artificial intelligence. In some contexts, the information collected during seemingly ordinary activities can become such a valuable resource that it completely replaces traditional monetary payment for a service.