In a press release from NVIDIA today, researchers have developed a deep-learning system that can teach robots how to complete tasks by observing the actions of a human. The article states that once the robot sees the task, it generates a human-readable description of the necessary tasks to re-perform the task.



First GPP and now Skynet, way to go NVIDIA. Aside from the very real future of robots killing us ALL, this is pretty cool tech.

“The perception network as described applies to any rigid real-world object that can be reasonably approximated by its 3D bounding cuboid,” the researchers said. “Despite never observing a real image during training, the perception network reliably detects the bounding cuboids of objects in real images, even under severe occlusions.”

Discussion

Source: [H]ardOCP – Nvidia Creates AI to Enable Robots to Learn From Human Demonstration