An algorithm enforced AI unit has managed to solve Rubik’s Cube in a mere 1.2 seconds.
Researchers at the University of California, who created the artificial intelligence system, has named the algorithm DeepCubeA.
In the study published in Nature Machine Intelligence, academics realised that the robot solved the 3D puzzle in a different way to how humans solve it.
Author of the report and Professor of computer science at the University of California, Pierre Baldi said, “It learned on its own,”. He added, “My best guess is that the AI’s form of reasoning is completely different from a human’s,”.
Solving the puzzle
The reason for using this specific game in the first place was because it cannot be played by guessing moves, there has to be a strategy in place.
The robots had no-prior game training to solve the brainteaser.
The AI device was given 10 billion different combinations to the problem and was told to decode the puzzle within 30 moves.
1,000 of these methods were then tested, all of which were correctly solved and done with an average of 28 moves. 60% of the time, the quickest path had also been taken.
In humans, even experts take around 50 moves to solve the cube.
The robots were also tested with other games such as Lights Out and Sokoban.
However, this is not the fastest machine that has been devised to solve the conundrum that was originally designed in 1974.
Last year, the Massachusetts Institute of Technology built a machine that solved the puzzle three times faster, named the min2phase algorithm. This cracked the cube in 0.38 seconds.
Why is DeepCubeA different?
The big breakthrough with DeepCubeA is that it mimics how the human brain works, which the min2phase did not.
Researchers were also looking at why these bots made the decisions they did with the goal being to build the next generation of AI systems.
“The solution to the Rubik’s Cube involves 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,” said Prof Baldi.
“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.”
Using this type of tech, researchers and engineers will be able to apply this kind of puzzle solving in AI to real-world problems. Such as enabling robots to be able to think, plan and respond to problems.