Artificial Intelligence continues to take steps towards its goal of replicating and surpassing human abilities. His last major challenge has been the resolution of the Rubik’s cube.
Rubik’s Cube is a game with billions of roads possible to reach a single end. The great difficulty this poses for machine learning is that the puzzle cannot be solved by sequences of random moves.
To meet this challenge, a team of scientists and mathematicians from the University of California has developed a deep learning algorithm with the ability to solve the problem of multiple combinations. The results of his work were published in the journal Nature Machine Intelligence.
In 60% of the tests found the fastest solution
DeepCubeA has solved 100% of the tests to which he has been subjected, and even managed to find the fastest route to solve puzzles 60% of the time. The solving capacity of the new algorithm is not limited to the Rubik’s cube, but has been successfully performed when it comes to solving in the fastest way other electronic puzzles classics like Lights Out or Sobokan.
The future of Artificial Intelligence
DeepCubeA has found the best way to solve the Rubik’s cube in barely a fraction of a second and without the need for specific knowledge or prior training by humans. This means that the algorithm has learned to solve the puzzle on your own and faster than any human being.
Faster than anyone
The algorithm developed by American scientists has completed the Rubik’s cube in a average of 20 movements against the 50 required by anyone. “Our Artificial Intelligence finds the solution most of the time in the minimum number of steps. We see that each strategy is different, and from this we have deduced that the reasoning method of AI is completely different according to Pierre Baldi, lead author of the study.
This project represents a new step in the ambitious goal of building the next generation AI that will mark the future. Today, these systems continue to show weaknesses in the face of creativity and improvisation capacity human.
The arrival of new machine learning algorithms will make a difference if it manages to advance in the field of intelligence, robustness, reasoning and comprehension capacity.