AI's next dataset is your apartment
Summary
German startup MicroAGI's Shift app has launched a free apartment cleaning service in New York City, where customers receive complimentary cleaning in exchange for data. Cleaners wear head-mounted cameras, recording point-of-view footage of the roughly two-hour jobs. This human-generated data is then sold to AI labs and utilized for MicroAGI's internal AI research, effectively turning customers and labor into training material for automation. Shift claims to have already paid over \$5 million in Q1 to individuals globally for filming everyday chores at \$20 per hour. Following its successful New York launch, drawing "thousands and thousands of bookings," Shift plans to expand to London, Munich, and Zurich. This initiative exemplifies a growing trend of sourcing AI datasets from ordinary human work rather than traditional internet scraping. The brief also covers Nebius Token Factory for open-source LLM deployment, AI use cases from The Rundown staff, a guide for building a video workstation with Higgsfield and Claude, and Inherent Labs, an ex-DeepMind group, securing \$50M to develop a self-improving AI science platform.
Key takeaway
For AI Product Managers exploring new data acquisition strategies, you should consider models that directly monetize human activity. MicroAGI's Shift app demonstrates how offering free services, like apartment cleaning, can generate valuable first-person training data for AI models. This approach allows you to subsidize service costs while building proprietary datasets. Evaluate the ethical implications and user consent mechanisms carefully when integrating such data-for-service exchanges into your product roadmap.
Key insights
The next frontier for AI training data involves monetizing human labor and daily activities through direct capture.
Principles
- Human activity can be a direct data source.
- Data capture can subsidize services.
- Recursive self-improvement applies to research.
Method
MicroAGI's Shift app records cleaners via head-mounted cameras during free apartment cleanings. This first-person footage is then used for AI training and sold to other AI labs, creating a new data acquisition model.
In practice
- Offer services in exchange for data.
- Use POV cameras for task data collection.
- Explore AI agents for scientific discovery.
Topics
- AI Training Data
- Physical AI
- Data Monetization
- Robotics
- Self-improving AI
- AI Agents
Best for: Investor, Computer Vision Engineer, AI Scientist, Entrepreneur, AI Product Manager, Tech Journalist
Related on AIssential
Editorial summary, takeaway, and curation by AIssential. Original article published by The Rundown AI.