CyberRunner AI is a robot that learns through a modular learning process. It gradually stores the 'memories' of previous mistakes and successes. The challenge is to move a ball from one part of the maze to the finish line while avoiding traps and dead ends on the game board. Researchers at the Swiss Federal Institute of Technology in Zurich released a competition video this week, showcasing their new modular learning creation. Engine and Muscles This AI robot uses two small motors to simulate the use of hands and a camera to represent its eyes. In the video released by CyberRunner, we see the robot practicing the game for only six hours, just like human competitors. Modular Learning According to CyberRunner's website, the robot does not stop to learn during the game; the algorithm runs in parallel with the robot's gameplay. As a result, the robot gets better after each run. These 'memories' enabled CyberRunner to complete the maze test and set a record faster than any human. Interestingly, the researchers state that the AI robot has also developed "shortcuts." The company responsible for this robot states, "The collected experiences are stored in memory. Using this memory, model-based reinforcement learning algorithms can learn the system's behavior."