AlphaGo Computer is an outstanding achievement recently made. How was it made possible to defeat human GO champion.
AlphaGo is a computer program developed by Google DeepMind in London to play the board game Go. In October 2015, it became the first Computer Go program to beat a professional human Go player without handicaps on a full-sized 19×19 board. Wikipedia
Deep Mind with Masterminds:
Deep Mind is an artificial intelligence company founded in 2010 and acquired by Google in 2014. The company captured the media this March after its AI (Artificial Intelligence) program deep contested with the game of Go champion Lee Sedol.
AlphaGo: Horizon of Artificial Intelligence
Go is two player ancient games originated round about 2500 years ago in China. Go uses a 19 X 19-line grid board with white and black stones in the game scene. Both players are required to cover as much space on board as they can by surrounding empty space or capturing the opponent space. In the end and one having larger space wins.
Why it’s a Challenge:
Go was a challenge for AI experts, as though it looks simpler, yet it has enormous search space. Some trillions possible moves are possible at each step. After the first move in chess, there are 300 possible moves, but after the first move in Go, there can be some 130,000 moves. It has been reported by Deep Mind official site that there are a possible moves in Go than there are atoms in the Universe, How complex it is Man.
Thinking and then implementing a so much diverse number of moves was really a challenge. Google has recently succeeded in defeating the world champion of GO, Lee Sedol through its AlphaGo Computer. It is the first computer program to beat the professional human player of Go. Many thoughts it would become possible in the next decade or at least 5 years from now. Even the co-founder of Google Deep mind stated he was “”stunned and speechless” at the victory. High praise to deep mind experts, they made it possible.
How do they do it?
It has used machine learning with monte Carlo tree search and deep neural networks to solve the problem at hand. The AlphaGo program has run near about 99% of its games with other similar Go computer program.
Two neural networks have been used by AlphaGo to make it possible. In simple terms, we can say AlphaGo has to predict cleverly for the next possible moves of the opponent. AlphaGo uses Monte Carlo search tree, both value and policy networks and an advanced search algorithm to play. A value network is used to evaluate board positions and policy network helps in selecting the appropriate move among many possibilities. According to Deep Mind official blog,
We first trained the policy network on 30 million moves from games played by human experts, until it could predict the human move 57% of the time (the previous record before AlphaGo was 44%).
Later on, Deep mind was trained through self-play for improvement.
The Game of Leo & Deep Mind:
The recent game of Deep Mind and Leo sedol was played in South Korea on 9, 10, 12, 13, and 15th of March 2016. Deep Mind wins the first three games but loses to sedol in the fourth game. Though sedol apologizes to Go lovers for losing, yet he says that he will not exchange his priceless victory for anything. The game gained significant importance as it was the first time since the birth of AI that a human beats an AI based machine.
AI community has deeply praised for Google victory and it has set a base for further research in the field. Both Mark Zuckerberg and Elon Musk has praised deeply to the AlphaGo team for the milestone.