RICE AI is a platform for crowdsourced robotics data collection. It enables teleoperators to remotely control robots and gather data from robot vision, joint movements, and force measurements. Teleoperators can use webcams with skeleton detection, joysticks, VR controllers, or teleoperation rigs, earning token rewards based on the device's dexterity.
Individuals or companies can contribute robots -- such as humanoid robots, mobile bimanual robots, quadrupeds, or AI toys -- to designated data collection centers for token rewards. Higher rewards are offered for robot types in greater demand. These centers operate in real business settings or lab-built environments.
The collected data is sold to research labs and companies. RICE AI also trains robotics foundation models using this data, offering access to these models via a monthly subscription. Clients paying with the platform's native token receive a discount. Fees from data sales are used to burn tokens, reducing the total supply. The token also serves as a governance mechanism for the platform.
An AI evaluation model assesses data quality from teleoperators, supported by human data labellers who rank sample data based on task completion efficiency. An AI model predicts data episode similarity, and a clustering-based embedding database filters out redundant data to maintain quality.