Artificial Intelligence (AI) has once again showcased its prowess, not just in intellectual games like chess, but also in the physical domain. A team of researchers at ETH Zurich has unveiled CyberRunner, an AI robot that achieved a remarkable feat by breaking the world record in the challenging Labyrinth game.

Unlike traditional AI models that often defeat humans in strategic games, CyberRunner mastered a physical game that demands dexterity and precision. The wooden Labyrinth game, notorious for its difficulty, involves maneuvering a marble through a maze without it falling into holes. The AI-driven CyberRunner accomplished this task in an impressive 14.48 seconds, surpassing the human record of 15.41 seconds held by Lars-Goran Danielsson.

What sets CyberRunner apart is its ability to mimic the human learning process. Equipped with motors as “hands” and a camera acting as “eyes,” the robot learns through collected experience, akin to human practice. The model-based reinforcement learning algorithm allows CyberRunner to understand the game’s dynamics, identify effective strategies, and continually improve its performance.

The research leads, Thomas Bi and Raffaello D’Andrea, plan to make CyberRunner’s hardware and software open-source, encouraging others to use this breakthrough as a foundation for their experiments. They emphasize the affordability and widespread accessibility of AI models, enabling scientists and engineers to explore and develop their own innovative solutions.

CyberRunner’s achievement not only highlights the advancements in AI but also demonstrates its potential for real-world applications. The researchers envision large-scale experiments where AI-driven robots like CyberRunner can learn and adapt on a global scale. This opens the door to cutting-edge AI research, making it accessible to a broader audience and fostering collaborative efforts in pushing the boundaries of technology.

Check the video below,