Exploring AlphaGo Teach
Written by Sue Gee   
Wednesday, 13 December 2017

DeepMind has released AlphaGo Teach, an interactive tool that can be used to explore the opening moves of a game of Go. It lets you try out alternative moves and shows the probabilities of a win for black for each and the moves that AlphaGo would take in response.

alphagoteachbanner

 

For those who don't know history, AlphaGo is the neural network-based program developed by Google's DeepMind team whose career as a Go player lasted  less than 18 months. The first reported win was when it beat the European Go champion by 5 games to 0 in January 2016. AlphaGo then beat Lee Seedol, a Korean player who ranked second in the world 4 games to 1 in face to face play in an event in Seoul in March 2016. The next news came that he was the mystery player that was beating the world's top ranking players last December and then, at the Future of Go Summit in May 2017 organized in a collaboration between Google, the China Go Association, and the Chinese government, it beat the world's No 1 ranking Go player, Ke Jie a 19-year old Chinese player, 5 games to 0.

One of the surprising effects of AlphaGo's swift rise to supremacy over human players is that it rekindled interest in the game among its existing players and attracted newcomers to the game. Lian Xioa, one of the top Chinese professionals who took part in a Pair Go exhibition match at the Future of Go Summit, in which both sides had an AlphaGo teammate summed up the positive sentiment towards introducing AI techniques into his country's traditional game of skill and strategy:

 AlphaGo could actually broaden the horizon of Go playing. It could bring more imagination into Go.

After its triumph at the Future of Go Summit, AlphaGo officially retired from playing Go, and now as is befitting one who retired after a successful career it is sharing its expertise. As soon as the Summit ended a series of self-play games, i.e. AlphaGo playing both the black stones and the white ones, were made available so that human players could study the surprising moves and novel strategies used by the AI. Now in a further contribution by DeepMind, AlphaGo has been  given the role of tutor to give human players the benefit of its wisdom for opening games. 

agoteachend

AlphaGo Teach isn't going to help newcomers to the game. Its intended role is to help those who already know its rules and have a level of expertise find new and creative ways of playing Go. It is an interactive tool that lets you explore how Alpha Go's moves compare to those of human players, using 6,000 of the most popular opening sequences using data from 231 human games and 75 games AlphaGo played against human players. .  

 

agoteach4

There are two ways to interact with the tool. You can click on any of the circles outline in purple (moves a human would play) or green (moves AlphaGo would play. As the game progresses you find that there are fewer, or no degrees of freedom in the choice to be made. If you find this mode unresponsive use Firefox Quantum which worked when Chrome didn't. The alternative is to click the arrow keys to make AlphaGo's choice of move by a single move, 10 moves or to the end. The backward arrows let you rewind to the point where you can make an alternative choice.

One of the interesting things you discover by rewinding a game is that AlphaGo doesn't necessarily choose the move with the highest probability. This behavior is something that human players can now study and can learn from.

 

alphagosq

 

 

More Information

AlphaGo Teach

AlphaGo vs AlphaGo: self play games

Related Articles

AlphaGo Triumphs In China

AlphGo Defeats World's Top Ranking Go Player

AlphaGo Revealed As Mystery Player On Winning Streak

Why AlphaGo Changes Everything

 

To be informed about new articles on I Programmer, sign up for our weekly newsletter, subscribe to the RSS feed and follow us on Twitter, Facebook or Linkedin.

Banner


Flox Releases Flox Hub
13/03/2024

Flox has announced that its Command Line Interface (CLI) and FloxHub are now generally available. The CLI is open source and FloxHub is free for anyone to use.



AWS Adds Support For Llama2 And Mistral To SageMaker Canvas
12/03/2024

As part of its effort to enable its customers to use generative AI for tasks such as content generation and summarization, Amazon has added these state of the art LLMs to SageMaker Canvas.


More News

raspberry pi books

 

Comments




or email your comment to: comments@i-programmer.info

Last Updated ( Friday, 02 September 2022 )