Photo Courtesy: Pexels - Pavel Danilyuk.
Mukundan Gurumurthy
“When we play chess, we hatch grand plans, take risks, fall into traps, succumb to pressure, psych out opponents, and make bold sacrifices — all without knowing whether any of it will pay off.” Yoni Wilkenfeld
The advent of Artificial Intelligence (AI) has killed many strategic games, and the long-standing king is next on the guillotine. Research into AI has resulted in us “solving” many games, from simple ones -- like tic-tac-toe -- to ones as vastly complex as checkers, a game with 500 billion possible positions. Chess, though even more complicated, has already been conquered by our ingenious rival.
We face a dilemma, where we, as humans, are not certain about whether these “Chess AIs” are bringing about a positive or an adverse effect on us. Perhaps, it is best to see it as a sacrifice we are making, a gambit of sorts. But before we delve into these questions, we must answer: what are these “Chess AIs”?
Chess AIs, more formally known as chess engines, are computer programs that examine chess positions and generate possibly the best, or the most ideal move. Until the introduction of IBM's chess computer “Deep Blue”, which had an abundance of data fed into it, engines were lacklustre, often overrun with bugs and lacking positional understanding. Deep Blue, a machine running at a paltry fraction of computers we have access to today, was able to topple the throne of Gary Kasparov, the world chess champion for about 22 years, by brute forcing its way through with the help of the countless moves available to it. In today’s world of ever-growing computational power and highly optimized algorithms, computers have far overtaken humans in the game.
Ben Morse, writing in the CNN, stated that in recent years, chess engines have outperformed humans in terms of strength; with some now at higher than 3,000 Elo, a rating which indicates a player’s proficiency. To put things in perspective, Ben wrote, the record for the highest Elo rating ever attained by a human player is held by former world champion Magnus Carlsen who reached 2,882 in 2014.
The continuation of work on the development of these engines poses the risk of “solving” chess, which would effectively kill the game. Solving a game means that, given any position, a computer will be able to determine and generate an outcome for it. But what does solving a game mean in the real sense of the word? And what do we gain from seemingly “killing” the game, or, in fact, are we truly killing it in the first place?
The Middle Game
As far back as 1925, GM José Raúl Capablanca, a Cuban chess champion, feared the “draw death” of chess. He reasoned that as players continually improve, games at the highest level will eventually always result in a draw. This led him to create a new variant of chess --Capablanca Chess, effectively introducing more pieces and complexity.
The use of AI has contributed to players drastically improving their proficiency at the game in a short span of time, accelerating Capablanca’s predictions. Renowned players even use these engines to plan out the openings of their games. This results in a more orderly and “boring” game as both sides try to play the ideal moves as far into the game as they can.
In the paper “Can Chess Survive Artificial Intelligence?”, Yoni Wilkenfeld discusses how making mistakes, to blunder, is to be human. Chess engines, which are significantly stronger than human contenders, take this element of error out of their games. This makes the game more mundane and uneventful, since both sides avoid going for an interesting move that might carry some risk factor. This error-minimized way of playing chess will in most cases end in a draw; while as the “human” way of playing chess is far more dynamic: the game is not only on the board, but also as a mental battle between the opponents, and this clash yields far more interesting moves.
While it is established that we should not, or cannot, kill the game, it remains to be admitted that there are certain positive outcomes we can draw from using chess engines.
Chess has vastly increased in popularity during the pandemic due to the availability of websites such as chess.com and lichess.org. But with the advent of online chess, it has become harder to maintain the competitive integrity of the game, as players are not monitored onsite. AI engines can be applied in this scenario to identify potential cheaters, due to the unique, non-human play style of an engine. Various companies such as Deep Mind are working on engines that detect whether a player is cheating.
These engines are also heavily used to analyse games to provide better insights and point out mistakes made. They are useful to people of a varying skill levels because of how deeply and far-sightedly they can calculate. In fact, as Delia Monica Duca Iliescu discusses in the paper The Impact of Artificial Intelligence on the Chess World, chess, and other strategic games in general, also help us benchmark our progress in the development of Artificial Intelligence due to the sheer size of the decision tree that it can generate. Effectively spanning that tree to be presented with an optimal move is a monumental task, and a big achievement.
Checkmate
What we find of interest in chess lies in the risk-laden, error-prone style of playing that humans have. And even though players use engines to help prepare their moves for a game, this “preparation” is often limited to a few moves into the game. Since the player must memorize each optimal move, they cannot venture too far into any given sequence of moves without sacrificing knowledge of other potential “best” moves. There must be a balance between the breadth and depth of a player’s preparation, which simply cannot last throughout the game due to the substantial number of ways the game can go. This adds to the game, as it becomes a mental clash between two players, analysing and predicting what moves the other player could and will play, way before the game itself begins.
Continuing the trajectory of chess engine development might eventually lead to the “solving” of the game, akin to checkers, but just because checkers is “solved” does not mean we cannot enjoy playing it. Looking at the potential downsides of solving chess, it might take away the excitement of not knowing who is going to win and the mistakes players usually make.
On the other hand, what is interesting about using these smart engines is that they can teach us a lot and help us understand the game better. Embracing the idea of solving chess becomes a strategic gambit itself, one that prompts us to reassess our perception of the game, recognizing both the challenges and opportunities that lie ahead.
Mukundan Gurumurthy is a second year B. Tech student, focusing on computer science and artificial intelligence, at Plaksha University, Mohali.
