Now in order to teach the robot something new, you often need to spend a lot of time and effort of people in the process of learning. But NVIDIA offers significantly change this approach. Their new system for deep machine learning will allow robots to learn on their own, just watching people.
The new system is based on interconnected neural networks. Some are responsible for the perception of others in planning, others for the management and processing of data and so on. Due to the interaction of different systems and combining them with control, recognition of movements and objects, robots can learn by just watching people. And to be retrained to perform new tasks. As stated by the chief scientific officer of NVIDIA Stan Burchfield,
“In a production environment, robots are very repeat the same actions, but they are not able to adapt to changes occurring around, are not trained, performing only the specified program. And in order to teach the robot to perform a new task, you have to invite the expert who preprogrammed robot that is not cheap”.
The development of NVIDIA aims to ensure that even non-experts could teach a robot new tricks.
“We are now at the stage where with the help of new technologies we can create essentially limitless amounts of labeled data, and then develop and test algorithms. All this will help to allow us to create systems which will adapt to any conditions”.
The Corporation even released a video demonstration of the technology. In it a man moves objects, but, “looking at it”, the robot learns to do the same. If you believe the employees of NVIDIA, the technology is only “in the beginning”. Therefore, on the one hand, it is interesting, and a little scary, what else robots can learn from us in the future. Full version demonstration video is available below.