Artificial intelligence defeated the European Go champion

Go is said to be one of the most complex games that people can play but recently a computer program was able to defeat this chess champion in Europe, which 2 years people think it takes 10 years or more to become a reality.

Go is really a complicated game, requires a lot of thinking for people and for artificial intelligence, it is of course a huge challenge. Until recently, it was believed that with the current level of development it would take a decade or more for the computer algorithm to be strong enough to solve a Go game. If solving the problem of Go is synonymous with other issues, AI has the potential to overcome people.

Although not popular, especially in Western countries, there are still around 40 million people playing Go. The game will take place on a 19 x 19 square chess board, and 2 people will hold 2 black or white troops, place them on the chessboard and try to invade as much land as possible by encircling the chess pieces of enemy. Although there are still some Go versions available on computers, they are mostly easy to beat even for beginners.

Picture 1 of Artificial intelligence defeated the European Go champion
Go is really a complicated game.

And a completely new program called AlphaGo developed by Google's DeepMind team has used two artificial intelligence platforms to find a way to win people in the Go game. Specifically, the first platform allows AI to analyze each move based on probability and combine with a set of rules, allowing the program to rely on predicting the right move. A neural network designed to simulate human brain activity, allowing the AI ​​to go beyond predictions, can deduce the image of the chessboard, which determines which moves will open and surround and choose the right path, finally determine what is the method to win.

Demis Hassabis, a member of the research team at Google, said: "For us the most important aspect of AlphaGo is not just a dedicated system with built-in rules, but instead. It also uses learning techniques to determine how to win the Go game . " On the other hand, concurrent optimization of the two AI techniques is a significant effort by scientists.

In a Go game, there are more than 10 ^ 359 valid moves, the number of possible transformations is 10 ^ 761, in a number of seconds, the number of seconds in 3 years is less than 10 ^ 8, the number of variables chemistry in chess is about 10 ^ 43 to 10 ^ 50 and even the number of variables in the Go flag is greater than the number of atoms in the universe (physicists in the past have calculated this visible world There are more than 10 ^ 90 protons.) And AlphaGo program can defeat the European Go champion with the name Fan Hui . And the AlphaGo development team will not stop there either. They continue to accept the challenge from world champion Lee Sedol in the upcoming match in March. Let's see.