A unique “kindergarten” for robots has been opened in China, where researchers are testing a new approach to machine learning. Instead of teaching robots every action in advance, scientists want them to learn independently through experiments, mistakes, and repeated attempts. The training center was created by the Chinese company Tashan Technology.
Researchers believe that this method could allow robots to continue adapting even after they are deployed in the real world, including in situations they were never specifically trained for. For example, if a robot crashes into a wall during training, it can analyze the failure, change its route, and try again. In this way, mistakes become useful learning data rather than simply being treated as failures.
Wang Peng, a researcher at the Beijing Academy of Social Sciences, said that independent exploration of the environment could help robots move beyond narrow training for specific tasks. Instead, they may gradually develop a more fundamental understanding of the physical world and how objects around them behave. However, this type of learning requires much more time.
Robots may need to make hundreds or even thousands of attempts before they learn how to deal with a new situation. Still, researchers believe that the ability to make mistakes and learn from them could become an important part of the future of robotics. In the long term, scientists also hope that robots will be able to share their experiences with one another.
If one robot learns how to overcome a particular obstacle or complete a new task, other robots could potentially use that knowledge instead of learning everything from scratch. Interestingly, this approach is similar to the way young children learn. They do not simply follow instructions — they explore, make mistakes, and gradually understand how the world works.
Perhaps the idea of a “kindergarten” is therefore more than just a funny name: it reflects a completely different way of teaching machines. Here we are to serve you with news right now. It does not cost much, but worth your attention.
Extract — continue reading at the source.
