A super intelligent computer just became the best player at "Go" which is a thousand year old game from China which is also considered the oldest board game still played today. Now, scientists have developed an artificial intelligence system known as AlphaGo that has already mastered this ancient, abstract strategy game. 

Like Us on Facebook

Scientists also consider the game of Go as the most difficult challenge for any artificial intelligence than chess. IBM's DeepBlue became famous as it defeated world chess champion Garry Kasparov in 1997, that heralded a new found respect for computer artificial intelligence.

To play the game, it requires complex strategies to place board positions quickly that also involves a potentially infinite number of moves. Researchers say that not unitl recently that supercomputers were playing Go at a skill level of human amateurs.

Two players will use black and white checkers type pieces onto a square board with squares equally dividing the entire board, as players take turns to place their pieces. The goal of the game is to fill more square areas on the board than your opponent.

Researchers from Google's DeepMind revealed that AlphaGo became successful at this game since past systems failed due to its usage of "value networks" in evaluating board positions where it also utilizes "policy networks" when deciding upon specific moves. 

In order to train for this, AlphaGo underwent through supervised learning from human expert games and reinforcement learning from the games it plays versus itself. Researchers believe that this simulation of complex human decision making and strategy application, improved the system's skills significantly.

AlphaGo's decision making skills are already comparable to human cognition where the system learned from trial and error, and as it plays against itself, it gets better every time.

During tests, AlphoGo yielded a 99.8 percent win rate against other computer programs where the AI ultimately defeated top human European Go champion during a tournament. AlphaGo won five games against this human Go expert which marks the first time that a computer program won against a human professional player.  

Researchers plan to test AlphaGo further as the artificial intelligence system is soon to play against Lee Sedol who is an expert Go player, known for his playing prowess all over the world for the past ten years or so. This AI versus human match will take place in Seoul, Korea this March.

This new study is published in the journal Nature.