Summary: The history of chess has involved changing various rules of the game to make it more interesting and popular. Recently Alphazero, a modern neural network chess engine, has been able to change how humans approach chess. This engine learns superhuman strategies from scratch and can compose new games and can create new opening theory that appeals to chess grandmasters. This research study, “Reimagining Chess with AlphaZero” by Nenad Tomasev, Ulrich Paquet, Demis Hassabis, and Vladimir Kramnik (https://dl.acm.org/doi/pdf/10.1145/3460349), used alphazero to provide a qualitative and quantitative analysis of atomic changes to the rules of chess. These rules include disabling castling and allowing the pawns to move in sideways or backwards, all of which are not allowed in a classical game of chess. The statistical and aesthetic properties of these new variants can provide insight into possible rules for the game, making it more appealing to players.
Reaction: This article was eye-opening since, previously, I had thought that alphazero was only used to analyze the classical game of chess, not look into other variants. At first, I was slightly confused as to why alphazero was being used to find new possible rules for the game. As an avid chess player myself, I believe that alphazero’s development of new ideas in classical chess makes it unnecessary to bring up ideas for new variants of the game. However, after some deeper thinking, I have realized the true greatness of alphazero, and that its ideas from the other variants can be used to change the approach to classical chess (for example, through the non-castling variant). My research questions involved how chess engines change the approach to chess and chess training. By comparing variants to classical chess, grandmasters can try to build bridges based off of alphazero’s ideas and try to apply them to classical chess.
Citation:
TOMAŠEV, NENAD, et al. “Reimagining Chess with AlphaZero: AI Is Driving the next Evolution of Chess, Giving Players a Glimpse into the Game’s Future.” Communications of the ACM, vol. 65, no. 2, Feb. 2022, pp. 60–66. EBSCOhost, https://doi-org.eznvcc.vccs.edu/10.1145/3460349.
Hey Barzin, I completely agree with your reaction as you were surprised. I think that chess is a very interesting game and should be played the old ways. However, I am glad that there is some changing and evolving in such a classical game and is being accepted.
This is a fascinating development for the chess world. Personally, I am not too good at chess but I definitely enjoy the sport.
Also, does the Alphazero engine create infinite variants of chess or is it limited? If it is limited, how does it create its variants? Does it use an automatic algorithm or does it need human input to develop a new variant?
Hallo Barzin. Those variations on classical chess do sound pretty fun and interesting. Do you think if you studied a computer like AlphaZero enough, you could be better than somebody who didn’t? I see how you were initially skeptical about how variations would actually effect actual chess, but some variations seem like they have promise. Nice approach!
The ideas that Alphazero can bring along are definitely interesting. I am interested to see exactly how many variants of the game are actually viable, since only so many would make sense (a game with only one pawn each, for example, may be a bit too far-fetched to have any real impact on the classical version). And with those variants, how are chess masters able to actually translate their findings with their gameplay? Would they have to wait for a situation in a match that perfectly emulates a simulation of their variant?
I learned chess from my grandfather and play occasionally, however i’ve never been involved enough in it to know there are different ways, so I found this very interesting. When do you believe these variants would actually be applied if they catch on, and do you believe some will take much longer, if ever, to catch on as applicable varients?
I find this blog post fascinating because, although I am familiar with chess, I am awful at it. From reading the blog, I think chess, just like many things in the modern world, is getting dominated by technology such as Alphazero. I am not sure exactly whether it is for the better or not for chess, but I am excited to see what Alphazero has in store for the future of chess.
Barzin, I find your topic very interesting, as I am also interested in the use of artificial intelligence to benefit modern life. As you stated, it seems that at first glance the program “AlphaZero” creates frivolous ways to play chess. However, what if one of those “pointless” ways to play chess becomes the new norm for the game? I am aware that there are well over 2,000 variants of chess that some people find entertaining, so AlphaZero seems to develop on that. Your next step of research could be researching simpler forms of technology that affect chess. For example, online chess, or even the timer that chess players use for their turns. I believe both of those sub-topics have had a great impact on chess, maybe even greater than AlphaZero! Overall, you have a great blog post with insightful observations.