Auteur : theoclamart

Previous versions of AlphaGo initially trained on thousands of human amateur and professional games to learn how to play Go.

AlphaGo Zero skips this step and learns to play simply by playing games against itself, starting from completely random play. In doing so, it quickly surpassed human level of play and defeated champion-defeating version of AlphaGo by 100 games to 0.

It is able to do this by using a novel form of  in which AlphaGo Zero becomes its own teacher.

The system starts off with a neural network that knows nothing about the game of Go. It then plays games against itself, by combining this neural network with a powerful search algorithm. As it plays, the neural network is tuned and updated to predict moves, as well as the eventual winner of the games.

If you want to learn more about reinforcement learning.

It also differs from previous versions in other notable ways.

AlphaGo Zero only uses the black and white stones from the Go board as its input, whereas previous versions of AlphaGo included a small number of hand-engineered features.

It uses one neural network rather than two. Earlier versions of AlphaGo used a “policy network” to select the next move to play and a ”value network” to predict the winner of the game from each position. These are combined in AlphaGo Zero, allowing it to be trained and evaluated more efficiently.

AlphaGo Zero does not use “rollouts” – fast, random games used by other Go programs to predict which player will win from the current board position. Instead, it relies on its high quality neural networks to evaluate positions.

All of these differences help improve the performance of the system and make it more general. But it is the algorithmic change that makes the system much more powerful and efficient.

 

Have you heard of AlphaGO?

It has been thousand of years since the creation of the game of Go. Since all this time, not a single pro-player of go has been defeated by a computer program, in opposition to other strategical games as chess for example. 

Three years ago, in October 2015 the program AlphaGO is the first program to win against a professional player, Fan hui. It is at this time a bomb in the world of GO. In fact, during a game, the number of possibilities are superior to the number of atoms in the known universe (Hubble space), thus the success of AlphaGo is a huge step forward in the AI domain but also a huge hit for the professional players.

However, the Asian professional players are not worried about AlphaGO. They are trashing Fan Hui as a French professional player he does not have the level of the Asian’s (And from my point of view, they were right). 

Due to the provocations, DeepMind -Google- challenge the number 1 player of the century : the Korean player, Lee Sadol. The challenge will be 5 consecutive games  between him and AlphaGo. The first « player » to win 3 games will win the challenge.

I am not telling you who won the 3 games, but I strongly recommend to watch the documentary « Alpha GO » on Netflix
The fight between AlphaGo and Lee Sadol had been a revolution in Korea where the game of Go was and still is very popular. 

Nowadays, DeepMind is creating AlphaGo Zero which will probably be the master of AlphaGO. All the Go community is watching this closely. 

Fan Hui is a Chines-born french Go player. He became a professional Go player in 1996 at the age of 15. When he was 18 years old, he had never left China. He then decided to stop playing go, and discover the other part of our world. He first went to London, to Italy and then stopped in France. 

However, no matter how hard he tried, the game of Go could not get our of his life. The go was everywhere. And for all go players across the world, we know how he felt, as go is in every one of our moves, every forces or energies in the universe. The Go came back little by little in Fan Hui’s Life. 

Finally, Fan Hui started to play Go in the European ligue. He quickly became the number one player in France, but also in Europe. Between 2005 and 2015 he was the most important player in the Occidental world, and he participated to the development of the game in France but in a more global way, in the whole world. 

Three years ago, after his defeat against the first version of AlphaGo, the professional world was very hard on him. Treating him as a « traitor » to his homeland , as an opportunist as the level of Go in Europe is not comparable to the Asiatic one. 

I wanted to write this article to bring my support to this amazing go player. I have met him once in Strasbourg, and I had the chance to talk about the reaction of the other professional players to his defeat. 

Even if this event was three years ago, I could really felt how Fan hui was still affected by this story, and I wanted people to know a bit more about his history and about what he did for the game itself. 

We are with you Fan !