Startups in Shenzhen, China's manufacturing heartland, are pioneering a new wave of blue-collar work where humans remotely pilot humanoid robots using VR and motion-tracking tech.
At IO-AI Tech, just outside Shenzhen, workers don VR headsets and motion-tracking gear to control robots in simulated environments like factory floors and convenience stores. The goal is to have robots perform tasks such as stocking shelves and picking items, while also collecting data to enable autonomous operation down the line.
During a demonstration, the author experienced firsthand how intuitive this technology can be, controlling 10 different robotic hands with a custom motion-tracking glove. Finger movements were instantly mirrored by the robot's digits, and the system even provided haptic feedback, allowing the user to feel a ball placed in the robotic hand.
A trial with a Chinese convenience store chain involved using VR and grippers to restock medication shelves. Although initially disorienting, adjusting to the slight lag between personal movement and the robot's action quickly led to efficient shelf-stacking.
The tech is reminiscent of "Ready Player One," with workers in VR headsets and body sensors controlling robots. In one setup, a Unitree humanoid robot mimicked a person's movements in a mock apartment, demonstrating tasks like removing a shirt from a hanger and folding it, all viewed through the robot's cameras.
IO-AI Tech's innovation lies in its ability to translate human movements across various robot models, a crucial advantage given the diverse range of robots available in China. Their algorithms also blend human control with AI, ensuring robots maintain balance even when their physical form differs from the human operator.
Being based in Shenzhen, with its vast network of manufacturers, provides IO-AI Tech with an ideal environment for rapid prototyping and collaboration. The startup is partnering with companies like Jack Sewing Machines to train robots for tasks such as ironing shirts, aiming to integrate automation directly into existing production lines.
While some experts believe massive teleoperation datasets could lead to highly capable AI models, IO-AI Tech co-founder Si Chin advocates for a more measured, incremental approach. She likens it to the development of self-driving cars, where AI is gradually deployed in more complex scenarios. This data-driven approach is even finding its way into Chinese vocational schools.
As China continues to produce affordable, advanced robots, companies like IO-AI Tech suggest that AI might soon master the physical world through human-guided training.
