We all know AI can regularly beat humans at intellectual games like chess. But now, AI has proven itself in the physical realm. That annoyingly hard wooden Labyrinth game? AI can beat you at that, too.
According to The Next Web, researchers at ETH Zurich have created an AI robot that beat the Labyrinth world record held by Lars-Goran Danielsson. The Labyrinth-playing robot, called “CyberRunner,” safely maneuvered the marble to the opposite side of the board in 14.48 seconds compared to Danielsson’s 15.41 seconds.
CyberRunner accomplished this feat by mimicking the human learning process. It has motors, or “hands,” that operate two knobs controlling the slope of the board and a camera that acts as “eyes” to capture observations and learn from them. By playing the game, CyberRunner learns by “collected experience,” or as we humans call it, “practice.”
“The model-based reinforcement learning algorithm learns how the system behaves, and based on its understanding of the game, it recognizes which strategies and behaviors are more promising,” the research website said. After merely six hours of learning, CyberRunner successfully beat the human world record.
Research leads Thomas Bi and Raffaello D’Andrea are going to make the robotic system’s hardware and software open-source so that others can build their own AI experiments. In fact, the researchers encourage others to use this experiment as a jumping-off point for their own research. Thanks to relatively low-cost and widespread access of AI models, it’s never been easier for scientists and engineers to develop their own products.
You can watch CyberRunner show off its Labyrinth prowess in the video below.
Topics
Artificial Intelligence
Robotics
