Artificial intelligence programs can easily beat human competitors in games like chess and poker. This performance gap is starting to shorten, especially in games requiring physical dexterity.
Switzerland’s ETH Zurich researchers recently introduced CyberRunner, their new robotic system that learned to play Labyrinth faster than a human by using precise physical controls, visual learning, and AI reinforcement training.
Labyrinth is a box topped with a flat wooden plane that tilts across an x and y axis using external control knobs. CyberRunner mastered the dexterity required to complete the game in just 5 hours. Not only that, but its creators say it can now complete the maze in just under 14.5 seconds—over 6 percent faster than the existing human record.
CyberRunner learned how to navigate the marble successful along its route through hours’ worth of trial-and-error Labyrinth runs, stored in its memory. An algorithm runs concurrently with the robot playing the game, allowing it to keep getting better, run after run.
CyberRunner not only learned the fastest way to beat the game, but it also identified shortcuts that allowed it to shave time from its runs.
CyberRunner’s designers have made the project completely open-source, with an aim for other researchers around the world to utilize and improve upon the program’s capabilities.
“Prior to CyberRunner, only organizations with large budgets and custom-made experimental infrastructure could perform research in this area,” project collaborator and ETH Zurich professor Raffaello D’Andrea said in a statement this week. “Now, for less than 200 dollars, anyone can engage in cutting-edge AI research. Furthermore, once thousands of CyberRunners are out in the real-world, it will be possible to engage in large-scale experiments.
