IRVINE, Calif. — The Rubik’s Cube has been frustrating would-be solvers for decades, and whenever someone is able to conquer it, it’s an achievement worthy of recognition. Now, researchers from the University of California have created an extraordinary artificial intelligence system capable of solving the Rubik’s Cube — in a fraction of a second.
The AI system is a deep reinforcement learning algorithm named DeepCubeA, and — perhaps even more astonishing — it doesn’t need to be provided with any specific information on the Rubik’s cube or in-game coaching to beat the puzzle. It learns all on its own.
For reference on just how impressive beating a Rubik’s Cube in under a second really is, consider that each cube features billions of possible moves and completion paths.
According to researchers, the AI solved the Rubik’s Cube in 100% of all test runs, and found the absolute shortest path to victory 60% of the time. DeepCubeA doesn’t just play with Rubik’s Cubes either, its creators say it can also play games such as Lights Out and Sokoban.
“Artificial intelligence can defeat the world’s best human chess and Go players, but some of the more difficult puzzles, such as the Rubik’s Cube, had not been solved by computers, so we thought they were open for AI approaches,” explains senior author Pierre Baldi in a release. “The solution to the Rubik’s Cube involves more symbolic, mathematical and abstract thinking, so a deep learning machine that can crack such a puzzle is getting closer to becoming a system that can think, reason, plan and make decisions.”
Baldi and his team say that some extremely bright individuals, specifically teenagers, are able to solve a Rubik’s Cube in around 50 moves. The AI, on the other hand, usually only takes a mind-bending 20 moves. This had lead Baldi to hypothesize that the AI goes about solving the puzzle in a completely different way than a human ever would.
The AI’s creators say that the end goal of projects like this is to build the next generation of AI systems like Siri and Alexa. However, while current AI system are certainly impressive in some ways, they are still very much a work in progress.
“But these systems are not really intelligent; they’re brittle, and you can easily break or fool them,” Baldi comments. “How do we create advanced AI that is smarter, more robust and capable of reasoning, understanding and planning? This work is a step toward this hefty goal.”
The study is published in the scientific journal Nature Machine Intelligence.