At IO-AI Tech, a startup near Shenzhen, workers use VR headsets and motion-tracking gear to remotely operate humanoid robots in industrial and retail settings. The company aims to train robots for tasks such as stocking shelves and picking items from bins while gathering data for future autonomous operation. During a demonstration, the author controlled 10 robotic hands using a custom motion-tracking glove, with movements instantly transferred to all 50 robotic digits. The experience included both playful gestures, like flipping the bird, and practical tasks, such as stacking shelves with robotic precision. The system also allows users to interact with robots in real-time, adjusting for minor discrepancies between human and machine movements.
The company's technology enables the transfer of human movements to various robot forms, a critical feature given the diversity of humanoid robots available in China. IO-AI's algorithms must balance human control with robot autonomy, as the two entities differ in shape, size, and weight. Without some level of independent movement, robots risk losing balance. Shenzhen, known for its manufacturing base, provides an ideal environment for the startup to develop and refine prototypes. IO-AI collaborates with local manufacturers like Jack Sewing Machines to automate tasks such as ironing shirts, which could integrate into existing production lines.
IO-AI Tech’s cofounder, Si Chin, explained that teleoperation data could eventually lead to advanced AI models but emphasized an incremental approach to automation. She compared this to the development of self-driving cars, noting the need for focused training data. Robot teleoperation is also gaining traction in Chinese vocational schools, indicating a growing interest in the technology. China’s existing manufacturing capabilities, which produce affordable, high-quality robots like Unitree’s, may further support AI’s integration into the physical world.
Source: wired
