There currently isn't a comprehensive dataset available to train robots to reliably perform various tasks, such as grabbing complex objects or understanding and executing simple commands. To create this type of data, we need to allow robots to experience these tasks similarly to how a human baby learns, but in a more controlled environment.

To develop the general-purpose robot that humanity has long envisioned, we need a standardized type of robot distributed among researchers, industries for simple tasks, and households. By continuously updating a single AI system with insights from all these robot replicas, we can incrementally add more skills, building on existing ones, until we have robots capable of constructing buildings, cooking, washing clothes, and more.

The goal of this project is to create an affordable robotic platform optimized for AI and provide an open-source AI platform that integrates learning from multiple sources into a unified AI system continuously.