AI solves a Rubik’s cube without human help
Rubik’s Cube is one of the world’s classic puzzles. First created in 1974, some people can solve it in seconds, others can only solve it by peeling off the stickers and replacing them. Now AI can solve it too in under a second and without human help. We’ve got the details.
How AI solved the puzzle
While computers have been solving Rubik’s cubes with regularity, they typically required some coaching and interaction from humans. This time, at the University of California, Irvine, researchers programmed a machine-learning algorithm, called DeepCubeA, to solve a Rubik’s Cube completely on its own. The algorithm did astonishingly well. At best, the puzzle was solved in under a second and with less than 20 moves. According to reports, results were consistent as the test was repeated, with AI solving 100 percent of the tests and identifying the fastest route to a solution 60 percent of the time.
How does this compare to human solutions?
When it was first made, the puzzle took its creator over a month to solve. As the cube was made with lighter materials, time to solve it decreased. The first Rubik’s Cube world championships were held in 1982 and the winner of that competition solved the cube in under half a minute. Since then, human solving times have only gotten faster. In 2019, champion Phillip Weyer solved it in less than seven seconds. Generally, it takes humans 50 moves or less to reconfigure the cube so all of the sides are the same color.
The lure of being able to solve the puzzle as quickly as possible, called speedcubing, created a large community of Rubik’s Cube fans. The internet, with its ability to share solutions and talk about the math behind the puzzle, caused the community to grow even further. Currently, speedcubing’s popularity is enormous. The Los Angeles-based World Cube Association hosts hundreds of worldwide competitions for more than 100,000 people. Competitions can get pretty creative including categories for people to solve it with their feet or with just one hand.
Implications of DeepCubeA’s Success
This latest success continues to demonstrate just what artificial intelligence is capable of. It comes on the heels of headline-making experiments when a computer has defeated the world’s best players at complex strategy games of chess and Go. Researchers on this project say that this is another step towards demonstrating the depth of higher-level thinking that machines can accomplish.
Want help? Solutions for humans are published online
Have you tried to solve the puzzle but didn’t have the success of DeepCubeA or any of the top human puzzle masters? Driven by the interest in speedcubing, it is pretty easy to go online an find solutions to the classic puzzle and its multiple variations. A google search turns up a number of options and even the official Rubik’s Cube website has answers. Think twice about going there, however. Like any other spoiler alert, finding out the answer without getting there yourself can wreck the fun.
In case you wondered: The history of the Rubik’s Cube
Hungarian professor Erno Rubik first created the cube in 1974 to help his students learn to solve 3-dimensional puzzles. The puzzle, which includes 54 stickers placed over 6 sides, became popular almost immediately in Hungary but was difficult to export because Hungary was a Communist country. Over time, the cube made an appearance at toy fairs and was eventually manufactured and marketed by the Ideal toy company. With the help of television advertising, it became a worldwide sensation by 1981 and was selected for the permanent collection at the MOMA by 1982. It is beloved to this day.