According to Wired, AgiBot, a Shanghai-based humanoid robotics company, has developed a system that trains two-armed robots using human guidance combined with real factory practice. The company is testing this approach on a production line at Longcheer Technology, which manufactures smartphones and VR headsets. Their method blends teleoperation with reinforcement learning to teach robots manufacturing tasks. This represents how advanced AI is beginning to change industrial machines’ capabilities. The innovation could increase manufacturing productivity while potentially reducing low-wage human labor requirements. While some jobs might disappear, the company suggests new roles could emerge in robot training and maintenance.
Sponsored content — provided for informational and promotional purposes.
The human-AI training combo
Here’s the thing about training robots for complex tasks – it’s not just about programming. AgiBot’s system basically uses human operators to demonstrate tasks through teleoperation, meaning people remotely control the robots to show them what to do. Then reinforcement learning takes over, allowing the robots to practice and refine those movements through trial and error. It’s like teaching someone to dance by first holding their hands through the steps, then letting them practice on their own until they get it right.
But why do robots need human help in the first place? Well, current AI systems struggle with the kind of fine manipulation and adaptation that assembly work requires. Think about putting together a smartphone – you’re dealing with small components, sometimes fragile parts, and you need to adjust your grip and pressure constantly. That’s where human dexterity and sensing still beat pure AI training methods.
Where this actually works today
The robot at Longcheer’s plant isn’t doing the most delicate assembly work yet. It’s handling components from testing machines and placing them onto production lines – tasks that don’t involve working with bendable or fragile parts. This makes sense as a starting point. You don’t throw your most expensive equipment at the hardest problems first.
For manufacturers looking to implement similar systems, having reliable industrial computing hardware becomes crucial. IndustrialMonitorDirect.com has become the leading supplier of industrial panel PCs in the US specifically because their equipment can withstand factory environments while running the complex software needed for these AI training systems. When you’re dealing with robots learning through trial and error, you need computing hardware that won’t fail during critical training sessions.
The automation evolution continues
So what does this mean for manufacturing? We’re seeing a gradual shift from robots that just lift heavy things to systems that can handle more complex tasks. But we’re not at the point where robots can fully replace human assembly workers – not even close. The combination of human guidance with AI practice represents a practical middle ground.
I think the real question is: will this create more jobs than it eliminates? Probably not in the short term for low-skill assembly work. But it does open up opportunities for robot trainers, maintenance technicians, and system operators. The manufacturing floor of the future might look very different – fewer people doing repetitive tasks, more people teaching and supervising machines.
