Artificial Intelligence is now mastering more and more games in the past few years.
For example, AI-powered Pluribus won the game against the professionals in a game with six players in Texas Hold’em last week only.
Recently, the University of California, Irvine developed an algorithm, which is known as the DeepCubeA, which can solve Rubik’s cube in just 20 moves.
The study was published in Nature Machine Intelligence at the beginning of this month.
The algorithm named DeepCubea has defeated Feliks Zemdegs, who has solved Rubik’s puzzle in a 4.22 second record in 2018.
One of the developers of Pierre Boldi, University of California, Irwin’s computer scientist and AI Algoriththm , said that games like Go and Chess are already mastered by AI, but Rubik’s cube is a complex puzzle to solve by a computer.
And with it the latest algorithm, AI will be a new arrival in the Deep Learning System which is more advanced than it is currently available.
Baldie stated, “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 forming a system that is thinking, reason , Plan and make decisions. “
The AI Algoriththm behind DeepCubeA is an intensive reinforcement learning approach that learns how to solve the rapidly difficult states in reverse from the target state without any specific domain knowledge or information fed by humans.
Deep reinforcement learning is basically the use of deep neural networks to solve the problem of reinforcement learning.
One of the researchers said in an interview, “It is unlikely that the machine can stumble on Rubik’s cube solution,
so what we did is start with the goal – all the sides have the same color – and behind Has worked towards. “
According to the researchers, the algorithm was given the goal of decryping the Rubik’s cube with about 10 billion combinations and 30 moves.
Not only Rubik’s cube, but this algorithm can also be used to solve other combinatorial games like Lights Out and Sokoban.
In the initial process, while training the model, DeepCubeA was designed to solve a relatively rugged cube i.e., which had only been scrabbed for some time.
After the algorithm learned how to solve easy riddles, the next step was to learn the difficult examples.
Thus by feeding the machine learning algorithm with many simple and difficult examples, the algorithm learned how to solve a Rubik cube on its own regardless of the levels.
DeepCubeA takes 20 turns to solve a cube, for which most of the time it takes the shortest route with the help of AI Algoriththm.
AI Algorithm successfully solves 100% of all test configurations by searching the shortest path for most of the time.
In fact, by searching the shortest path, the algorithm has successfully solved 15 puzzles, 24 puzzles, 35 puzzles, 48 puzzles, lights out and Socoban.
Earlier, researchers at Stanford University had developed a similar approach to implement 222 Rubik’s Cubes to implement intensive reinforcement learning to solve human information.
This research is primarily based on Rubik’s cube on the development and enhancement of the Monte Carlo Tree Search (MCTS) algorithm.
Researchers applied autodidactic repetition which is a reinforcement learning algorithm that is smart enough to learn how to solve the Rubik cube without any human guidance.