Man versus machine: AI beats human in mental and physical game for the first time



In a groundbreaking achievement, artificial intelligence (AI) has surpassed human performance in both physical and cognitive domains for the first time. Researchers at ETH Zurich University in Switzerland revealed their latest creation, the AI robot named CyberRunner, which demonstrated remarkable proficiency in a labyrinth game requiring a combination of dexterity and strategic thinking.

Unlike previous instances where AI excelled in mental tasks such as defeating chess grandmasters, CyberRunner engaged in a physical task involving the manipulation of two knobs to navigate a marble through a maze without falling into holes. Through a learning process marked by trial and error, the AI robot honed its skills, even attempting to devise unconventional strategies. Impressively, it completed the labyrinth challenge in a mere six hours, surpassing human performance.

The researchers employed a model-based reinforcement learning algorithm, enabling the robot to enhance its abilities through experience. Notably, CyberRunner demonstrated refined fine motor skills and spatial reasoning, showcasing its adaptability and mastery over the use of its robotic hands.

This achievement extends beyond prior accomplishments, with the AI robot surpassing the fastest recorded time set by a highly skilled human player by more than 6 percent, emphasizing its prowess in both physical and cognitive realms.

Professor Raffaello D’Andrea, one of the researchers, expressed optimism about the project’s broader impact by announcing plans to open-source the technology. By making it accessible on their website for less than $200, the researchers aim to democratize cutting-edge AI research, envisioning widespread engagement in large-scale experiments globally. This move towards citizen science reflects the potential for thousands of CyberRunners to contribute collectively to advancing AI capabilities on a global scale.