Developing the ability of machines to manipulate small objects is a crucial direction in the study of robotics. Scientists from the Massachusetts Institute of Technology have managed to achieve a significant milestone in this area by teaching a robot to play Jenga – a popular board game that develops fine motor skills and reasoning. It’s worth noting that the robot wasn’t programmed with any particular ‘winning’ algorithms – instead, the scientists focused on developing its ability to learn from practice.
This approach is significantly different from those commonly used in the scientific community. At the outset, the machine can make some mistakes, but as the game progresses it learns which pieces determine the structure’s stability, and which strategy it should adopt in order to win. This learning process mirrors the one used by humans. According to the scientists working on this study, the machine needs to pull out around three hundred blocks to fully master the game.