AI & CHESS by Arya Shah

Author:Arya Shah [Shah, Arya]
Language: eng
Format: azw3
Publisher: PaperTrue Ltd.
Published: 2020-07-14T16:00:00+00:00

Chapter 4

AlphaGo Zero

When we talk about the future, we must talk about AlphaGo Zero owing to its impressive technology. For those of you who do not know what is AlphaGo Zero, imagine this: you teach a small child to play chess, then you ask the little child to go in a small room and ask her to keep playing chess by herself. You then go and see what progress the little child has made. Suddenly, she is better than a world champion. That is exactly what AlphaZero did. It learned chess in four hours just by playing it on its own and beat the best chess engine in the world. Just unbelievable. Time is just a tagline for AlphaGo but the important part is that it just learned by playing it by itself. In this chapter, we will discuss the uniqueness of AlphaGo Zero and what makes it special.

We already know how AlphaZero has made such tremendous contributions to chess with such amazing games and its technology. Now we need to know how AlphaZero works. AlphaZero learns from itself by using a method called reinforcement learning. The software already has a neural network installed in it, which knows nothing about the game expect the rules of the game. The software has a very substantial algorithm. The neural network is designed to work in such a way that it can prognosticate the moves. In each game played by the software, its strength increases, leading to stronger neural networks. Hence, the software would remember its mistakes and get its own type of intuition. It also uses its own neural network to evaluate the current position on the board. I now realize what made Vidit Gujrathi say that it was like “aliens” playing chess. Even the CEO of ChessBase India and IM Sagar Shah said that people thought aliens have come to our world. Mr. Shah also said that he had not seen anything play chess like AlphaZero. In fact, Mr. Shah claimed that even ChessBase was trying to imply similar techniques while developing Komodo 12 which is a chess engine by ChessBase. Mr. Shah also predicted that brute force will soon take a step back and AI will take the stage. Mr. Ramesh also said that chess engines would get more powerful once AI is incorporated, just like we have seen in AlphaGo Zero. Of course many chess players would be concerned if AI could solve chess, which in fact was also one of the main concerns of Mr. Ramesh when he spoke about AI being helpful to chess. I believe that it is highly unlikely that AI will be able to solve chess as humans have been trying it for the last six centuries, and the machines have been trying for only the last century.

As stated earlier, AlphaGo Zero can perform its spectacular feats using the subset of machine learning called reinforcement learning. So let’s try to decode how reinforcement learning works and how it has been used in AlphaZero.

Reinforcement learning

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.