This past 5th of december Google's DeepMind team published a paper called "Mastering Chess and Shogi by Self-Play with a General Reinforcement Learning Algorithm"
Based on a previous approach, AlphaGo Zero which just a few months ago turned the world of Go completly upside down, now they implemented a more general neural network which is able not only to play and master Go, but Chess and Shogi also.
Alpha Zero is how they called it, and like its name suggest, it learned to play completely from scratch, without any knowledge about chess other than the game rules. Only by playing against another instance of itself during millions of games and thanks to the implementation of a reinforcment learning algorithm, the program "teaches" itself and becomes stronger and stronger each time.
In a 100 game match against the 2016 TCEC world-champion program Stockfish 8, Alpha Zero managed to win 28 games, with 72 draws and...0 losses. A completely wipe out for the (previously consider) almighty and almost unbeatable Stockfish.
Even more impressive is the fact that Alpha Zero achieved this level of play in only 4 hours of intensive training using 5000 TPUs.
Out of the 100 games of the match, 10 has been released to the public and obviously all are wins by Alpha Zero. But they are not just wins. Every single game is a masterpiece of the highest caliber.
Watch Alpha Zero play is like being in the precence of an alien or an entity on a completly different level of chess, that leaves even Stockfish runing on a cpu with 64 threads, out of comprehension or response.
Another interesting result about this massive achivement is that Stockfish, was able to calculate 70 million positions every second while Alpha Zero only looks at 80000 positions per second running on 4 TPUs. This is maybe a hint that Alpha Zero's strenght is based on a more "human like aproach" than traditional engines nowadays. Something that also gets reflected in the long-term positional type of play displayed on some of the games, where Stockfish just got outplayed from start to end.
You can see the full paper published by DeepMind here. It also contains the notation of the 10 selected games at the bottom of the article.
This certainly is going to be a revolution in the world of chess and maybe in a not so distant future, in many other areas of the human knowledge.
Congratulations @ancalagon! You have completed some achievement on Steemit and have been rewarded with new badge(s) :
Click on any badge to view your own Board of Honor on SteemitBoard.
For more information about SteemitBoard, click here
If you no longer want to receive notifications, reply to this comment with the word
STOPCongratulations @ancalagon! You received a personal award!
Click here to view your Board of Honor
Do not miss the last post from @steemitboard:
Congratulations @ancalagon! You received a personal award!
You can view your badges on your Steem Board and compare to others on the Steem Ranking
Vote for @Steemitboard as a witness to get one more award and increased upvotes!