Researchers at the University of California have developed an artificial intelligence (AI) capable of solving a Rubik’s Cube in just under one second.
Known as ‘DeepCubeA‘, the AI completed the puzzle as part of a study that has been published in Nature Machine Intelligence this week.
In the study, the algorithm was given 10 billion different combinations of the tricky puzzle with the aim of solving the problem in less than 30 moves. More than 1,000 subsequent tests saw the AI solve the Rubik’s Cube every time. During around 60% of the tests, DeepCubeA was able to find the shortest possible path to completion.
A key focus of the UCI research was to understand how and why the AI made its moves – and how long it took to perfect its method. Starting off with a computer simulation of a completed puzzle, the researchers then scrambled the cube. Once the code was established and running, DeepCubeA trained in isolation for around two days; during which it solved a series of combinations that escalated in difficulty.
“It learned on its own,” according to Professor Pierre Baldi, the author of the report and professor of computer science at UCI.
“Our AI takes about 20 moves, most of the time solving it in the minimum number of steps. Right there, you can see the strategy is different, so my best guess is that the AI’s form of reasoning is completely different from a human.”
- Artificial intelligence: ethics, regulation and public perception
- Deepnude AI photo app taken down by creators
- Chinese AI firm develops dog facial recognition app
The Rubik’s Cube has been challenging people, young and old, for nearly half a century since its launch in 1974. Generally, humans take around 50 moves to solve the puzzle, however, this latest AI system solved it at an impressive average of 28 moves.
Baldi explained that the key solution to solving the logic puzzle involves more “symbolic, mathematical and abstract thinking” – and thus a deep learning system capable of solving the puzzle edges nearer to becoming one that is capable of thinking, reasoning and making complex decisions.
“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,” he said.
Essentially, completing this challenge marks one of the first steps toward developing artificial intelligence that can do more than just tackle games and move into solving real-world problems.
“This work presents an AI system that can automatically learn how to solve the iconic Rubik’s Cube and other problems characterised by a vast number of possibilities and a very small number of solutions, and where random moves are unlikely to lead to a solution,” Baldi explained.