To program a robot to play Jenga, traditional machine-learning schemes might require capturing everything that could possibly happen between a block, the robot, and the tower — an expensive computational task requiring data from thousands if not tens of thousands of block-extraction attempts.
You are viewing a single comment's thread from: