You’ve probably heard of AI winning against humans in games like chess and GO that require intellect. That’s impressive, but not shocking. AI is good at crunching numbers and finding patterns. But what about physical skills? That’s something humans are supposed to be better at, right? Well, not anymore.
An AI labyrinth maze champion emerges
Researchers at ETH Zurich have created an AI robot with the task of learning how to play the popular wooden labyrinth maze game. The goal of the game is simple, using two knobs, you have to steer a marble ball from a start to an endpoint without it falling into the holes across the board. But if you’ve ever played it, you know it’s actually easier said than done. The scientific explanation behind its difficulty is that it requires acute motor skills, spatial reasoning abilities — and a lot of practice.
Credit: ETH Zurich
Credit: ETH Zurich
MORE: HOW THIS ROBOT HELPS YOU PROTECT AND CONNECT YOUR HOME
Meet the new robot champion
The robot, named CyberRunner, is equipped with two motors (its hands), a camera (its eyes), and a computer (its brain), allowing it to play the game just like a person would. Much like a human, CyberRunner learns through experience by leveraging recent advances in model-based reinforcement learning, which enables the AI to make decisions and choose potential successful behaviors by predicting the outcomes of different courses of action.
Credit: ETH Zurich
While playing the game, CyberRunner makes observations of the labyrinth and receives rewards based on its performance. It keeps a memory of the collected experience, or as we humans call it, “practice,” which the algorithm uses to learn how the system behaves.
Based on this knowledge, it’s able to recognize the most promising behaviors. As a result, the robot’s use of the two motors continuously improves, and CyberRunner keeps getting better while the algorithm runs every time it plays.
Credit: ETH Zurich
MORE: HUMANOID ROBOTS ARE NOW DOING THE WORK OF HUMANS IN A SPANX WAREHOUSE
The triumph of the robot in breaking the human record
The robot received 6.06 hours of practice. Impressively, it beat the previous world record set by Lars Göran Danielsson, who set a time of 15.41 seconds in 2022. CyberRunner completed the game in 14.48 seconds — faster by over 6% compared to the human record holder.
During the learning process, the researchers say the robot discovered shortcuts and found ways to cheat — a behavior that they are studying as an innate human trait. Therefore, the researchers had to step in and instruct CyberRunner not to skip parts of the maze.
Credit: ETH Zurich
MORE: THE NEXT GENERATION OF TESLA’S HUMANOID ROBOT MAKES ITS DEBUT
What can be learned from CyberRunner?
Researchers Thomas Bi and Raffaello D’Andrea are making available the robotic system’s hardware and software open-source so that others can benefit and learn from their breakthrough in AI skill acquisition. They want to spur other scientists and engineers to use their experiment as an opportunity to conduct their own research. For those interested, their research paper is available here.
Kurt’s key takeaways
CyberRunner’s victory marks a significant milestone in the field of artificial intelligence, demonstrating that AI can now outperform humans not only in tasks that involve information processing but also in physical skill games. The project’s open-source approach ensures that the benefits of this achievement are accessible to a wider audience, paving the way for further innovations in real-world machine learning and AI. As we witness AI breaking new ground, the line between human and machine capabilities continues to blur.
Are you amazed or frightened by the increasingly blurred line between human and machine capabilities? Let us know in the comments below.
FOR MORE OF MY SECURITY ALERTS, SUBSCRIBE TO MY FREE CYBERGUY REPORT NEWSLETTER HERE