Post by José Siles

Data Engineer @Nestlé | Ex-Amazon | +130k AI/Data Community | Trusted by 50+ Global Brands

This is how robots are being trained now. Not in labs. By watching humans. Stanford built UMI: A $400 gripper that lets you teach robots by demonstration. You hold it. Do the task. The robot learns. No teleoperation. No expensive robot-specific data. And it collects data faster: Teleoperation: ~35 demos/hour UMI: ~111 demos/hour Now Sunday Robotics is taking this further: A $200 glove. 500+ homes. Millions of real-world interactions. Robots learning dishes, laundry, espresso… From real human behavior. The robotics unlock might not be better models. It might be better data.

Post content

Video Content