A History of Computers in Chess.
In the late 1940’s, 2 of the creators of the computer, Alan Turing and David Champernowne, wanted to explore different avenues of machine learning, and one of these avenues happened to be chess. The pair would then develop what would be the first iteration, albeit basic, of a chess engine. They attempted production on Turochamp, but the program was too complex for the technology of 1950.
Chess engines really had their beginning moments in 1957, when IBM engineer Alex Bernstein created the first fully produced chess engine, called the Bernstein Chess Program, capable of just about 1 move per 8 minutes, which is still superbly subpar compared to someone with even the most basic chess knowledge. Engines started evolving throughout the 1960’s, but were still a ways away from beating skilled players. In 1968, International Master David Levy challenged anyone who thought their chess engine could beat him, and when the group of researchers behind Kaissa, an advanced engine for the time, challenged him in 1977, Levy obliged. He played 2 matches against Kaissa where he wouldn’t lose either. He had also played against engines Chess 4.5 and MacHack, where he dominated the board as well.
But in 1986, a new engine, ChipTest, showed potential in chess after having an even score in the North American Computer Chess championship. The next championship, in 1987, ChipTest won with a 4-0 sweep. ChipTest was built by the Carnegie Mellon University team of Feng-hsiung Hsu, Thomas Anantharaman, and Murray Campbell. Later, this same team would improve upon ChipTest into its new form, Deep Thought. Deep Thought was entered into the NACCC in 1988 and 1989, where it won both times, and it was the first engine to have beat a grandmaster, when it played against Bent Larsen in 1988. This put the Carnegie team on the radar of IBM, and 2 of them joined IBM’s team where Deep Thought was able to be modified more thoroughly. That win also put Deep Thought on the radar of world champion Garry Kasparov, where the champion and the engine played a game in 1989, which led to a disappointing loss for Deep Thought.
(Left to right, Murray Campbell, Feng-hsuing Hsu, Thomas Anantharaman, Mike Browne, and Andreas Nowatzyk.)
IBM had seen the loss as a learning experience, and they got back to improving Deep Thought, as well as putting up a contest to rename Deep Thought, where it would then take the name of Deep Blue. IBM started Deep Blue out strong, getting a practice game against grandmaster Joel Benjamin, and that's when the team behind the engine decided that Benjamin has enough skill to “train” Deep Blue. It played in a few NACCC games, but in 1996, it played against world champion Garry Kasparov once again, where in the first game of six, Deep Blue won against Kasparov, marking it as the first chess engine to beat a world champion under the standard game rules. However, Garry then won 3 and drew 2 of the subsequent games, making the score 4-2 for Kasparov. Deep Blue’s engineers then upgraded its hardware, making it essentially double the speed, and the 2 faced off once again in 1997, where Deep Blue would barely take the game, with a score of 3 ½ - 2 ½ , making it the first engine to beat a tournament against a world champion. Sadly, shortly after this achievement, Deep Blue went through a major technical error, and had to be dismantled later in the year.
From 2008 to 2022, one engine has controlled the game: Stockfish. Stockfish has led modern chess for years, mostly due to its advanced neural network, where it learns from previous games, and its availability, where you can play against the engine with a mere search on the internet. It has helped players of all skill levels, from absolute beginners to the highest grandmasters. Even current world champion Magnus Carlsen uses Stockfish for practice in high level play. However, in the middle of 2022, a new chess engine has allegedly beaten Stockfish in as little as 4 hours of analysis. That engine is AlphaZero, a new project by the team at DeepMind. To put Alpha Zero’s skill level into comparison, Magnus Carlsen has an elo rating of 2859, while Alpha Zero is said to sit around 4650. AlphaZero trains through self-play and neural network improvement, where it runs probability simulations to coordinate the best possible move, rather than other engines, which plot their moves according to the position of the enemy pieces and try to achieve the quickest checkmate. It isn’t released to the public, but some were invited to play against AlphaZero, like Carlsen, but until it’s brought into the world, we cannot know how it could affect chess.
Sources:
“Deep Blue,” IBM, seen Jan. 26, 2023, https://www.ibm.com/ibm/history/ibm100/us/en/icons/deepblue/
“Deep Thought,” Britannica, seen Jan. 23, 2023, https://www.britannica.com/topic/Deep-Thought
“About: Turochamp.” DBpedia, seen Jan. 23, 2023, https://dbpedia.org/page/Turochamp
Larry Greenemeier, “20 Years After Deep Blue: How AI has Advanced since Conquering Chess.” Scientific American, June 2, 2017, seen Jan. 26, 2023, https://www.scientificamerican.com/article/20-years-after-deep-blue-how-ai-has-advanced-since-conquering-chess/
Shiva Maharaj, Nick Polson, Alex Turk, “Chess AI: Competing Paradigms for Machine Intelligence.” April 14, 2022, seen Jan. 22, 2023, https://www.mdpi.com/1099-4300/24/4/550
“What makes Stockfish Chess the best AI engine?” Square Off, Sept. 7, 2022, seen Jan 26, 2023, https://squareoffnow.com/blog/stockfish-chess-engine/
Johnnie Rios, “What did Magnus Carlsen learn from the most advanced Chess engine?” Medium, June 7, 2022, seen Jan. 26, 2023, https://medium.com/@jr4216232/what-did-magnus-carlsen-learn-from-the-most-advanced-chess-engine-d8907c47e091#:~:text=AlphaZero%20developed%20enough%20to%20defeat,neural%20network%20to%20evaluate%20positions.
David Silver, Thomas Hubert, Julian Schrittwieser, Demis Hassabis, “AlphaZero: Shedding new light on chess, shogi, and Go.” DeepMind, Dec. 6, 2018, seen Jan. 27, 2023, https://www.deepmind.com/blog/alphazero-shedding-new-light-on-chess-shogi-and-go
Maxim Khovanskiy, “AlphaZero Chess: How It Works, What Sets It Apart, and What It Can Tell Us.” Towards Data Science, May 5, 2022, seen Jan. 27, 2023, https://towardsdatascience.com/alphazero-chess-how-it-works-what-sets-it-apart-and-what-it-can-tell-us-4ab3d2d08867
Mark Anderson, “Twenty years on from Deep Blue vs Kasparov: how a chess match started the big data revolution,” The Conversation, May 17, 2017, seen Jan 26, 2023, https://theconversation.com/twenty-years-on-from-deep-blue-vs-kasparov-how-a-chess-match-started-the-big-data-revolution-76882
“Deep Thought team with Fredkin Intermediate Prize,” Computer History Museum, picture taken 1988, seen Jan 26, 2023, https://www.computerhistory.org/chess/stl-430b9bbd52f71/
David Ewalt, "Kasparov vs. Deep Blue." Forbes, May 3, 2011, seen Jan 27, 2023, https://www.forbes.com/sites/davidewalt/2011/05/03/kasparov-vs-deep-blue/?sh=48a8f49530f8
