What AI technology does a computer use to get better at playing chess

The checkmate heard round the world happened twenty years ago last month, when reigning world chess champion Garry Kasparov lost a game of chess to a computer, IBM's Deep Blue. Though Kasparov would go on to win the six-game match in Philadelphia four games to two, the point had been made. A computer had defeated the best chess-playing human in the world. Fifteen months later, "Deeper Blue"—Deep Blue's smarter, slicker, and better offspring—caused Kasparov to quit mid-match out of frustration and hopelessness. ''I'm a human being," Kasparov said at the time, "When I see something that is well beyond my understanding, I'm afraid."

"I lost my fighting spirit."

As world looks back on the Deep Blue matches with two decades of hindsight and considers their future implications, it's clear that game-playing computers have gotten even more fearsome. Just this past week, Google's AlphaGo dominated world champion Lee Se-dol in three straight games of Go, a game previously thought to complex for a machine to master, before the human struck back with a fourth-game win. But because chess holds such a grip on the human imagination, it is the game computer programmers have attacked since the beginning, and what they learned programming machines to play chess paved the way to AI that plays Go and other games. So let's take a step back and investigate how we got here in the first place. How, and when, did computers become so darn good at chess?

Getty Images

The dawn of the game dates back approximately 1,400 years to ancient India and Persia. Evolving from a game called "chatranj" or "shatranj" (which sometimes translates to "a four division army" in Sanskrit), it was played at high court and used to "test the wisdom of men." War and trade took the game global, to Europe, the Far East, and the Byzantine Empire. With the game's arrival in the West (despite the Church's initial misgivings due to its gambling potential), the queen was added. By the 15th century chess had crystallized into its modern form. 

Chess was always considered an intellectual game, so much so that people have long imagined the innate power of machines would one day topple the might of the human mind. Back in 1770, thanks to the wizardry of Hungarian inventor Baron Wolfgang von Kempelen, some people thought it had already happened.

The Mechanical Turk was supposed to be a simple parlor trick. Standing in front of the court of the Holy Roman Empress of Austria-Hungary Maria Theresa, Kempelen unveiled a life-sized carved wooden statue of a sorcerer wearing a fur-trimmed robe and a turban, seated behind a 4 x 2.5 x 3-foot box. On the box rested a chess set. Noting that his wooden friend "Turk" was an automaton, Kempelen opened the cabinet to reveal its inner workings were nothing more than complex clockwork machinery. Then, Kempelen brought up a volunteer, Count Cobenzl, to challenge his robotic friend to a game of chess. Kempelen turned a few a keys and, after a brief silence, the automaton whirled to life. The court cried out in amazement as the Turk moved its first piece. The game ended in a hurry as the Mechanical Turk soundly defeated the Count in what was the first instance of machine beating man at chess. 

Except, it wasn't.

The Mechanical Turk became a worldwide sensation, going on tour and taking on several famous challengers. The Turk defeated Catherine the Great in Russia, beat Ben Franklin in Paris, and triumphed over an amazed Charles Babbage. The Turk's most famous match came in 1809 against Napoleon Bonaparte, who tried to outsmart the machine by making deliberately illegal moves. Frustrated, the Turk took its wooden arm and knocked all the chess pieces on to the floor, apparently delighting Napoleon. Rarely defeated (though the Frenchman Philidor, renown as the world's best player, did win his match against the Turk), the machine confounded even the most learned about how it actually worked. Was it magnetic? Controlled by wires? Perhaps it was actually magic? In 1854, eight decades after the Turk's introduction to the world, it burned in a fire in Philadelphia. Shortly after that, the Mechanical Turk was revealed to be a hoax. The truth was that the Turk had a concealed operator, one that was hidden by trick doors in the undercarriage of the box. Fitting only a person of small stature, it required the person to not only diminutive, but very good at chess. Needless to say, the Mechanical Turk was not magic. Quite simply, it was nothing more than a human hiding in a machine.

The Mechanical Turk inspired copycat fakers, but also immense interest in chess-playing machines, including from a man who actually had played the Turk: Charles Babbage. In 1833, after developing his "Difference Engine," Babbage conceived of what is widely considered the first computer, the "Analytical Engine." While he never was able to construct a fully function prototype, Babbage theorized that his machine would be able to play chess, though he recognized that would require "the machine being able to represent the myriad of combinations related to (the game)." Given that there are seemingly infinite possible moves in a chess match, with complexity growing exponentially, Babbage instead he described in detail how his Analytical Engine could play a much simpler game, tic-tac-toe, and left chess for another century.

In 1950, a researcher at New Jersey's Bell Telephone Laboratories named Claude Shannon wrote the first paper on how to program a computer to play chess. Published in Philosophical Magazine, it lays out theory, strategy, and procedure for building a program that could calculate and consider all 10120 game variations. While Shannon himself never built the program, Alan Turing did. 

"Once or twice it played so well that it rattled me."

Turing created an algorithmthat he manually calculated himself with pencil and paper in a game against a friend (taking 30 minutes per move). Later, Turing attempted to program his algorithm into the Ferranti Mark I, the world's first commercial computer, but never completed the task. Turing's colleague Dietrich Prinz accomplished this for him in 1951, but Ferranti Mark I didn't have enough processing power to play a full game of chess - it could only calculate the best move if the chess game was within moves of checkmate. In other words, Turing's brain had much more processing power than a 1951 computer. Six years later, the IBM 704 became the first computer to play a full chess game, taking eight minutes between move. Although the computer specialist Alex Bernstein who programmed it was able to beat the IBM, he said, "On(ce) or twice it played so well that it rattled me."

The IBM 704 is the turning point in this story. Once it had been proven that computers could play chess against humans, it because a question of when, not if, they could beat us. The race to make the fastest and most competent chess-playing computer was on. 

In 1965, in the midst of the Cold War, a chess playing program built by Alan Kotok and John McCarthy went to the USSR to take on a Soviet-designed program, a precursor to KAISSA that would becomethe World Computer Chess Championship's first winner. In a game played over nine months, the Soviets defeated the Americans three matches to one. MIT's Greenblatt Chess Program became the first computer to compete "respectably" against humans in tournament play in 1967. By the mid-1970s, the first commercially available chess program was out for the home computer, MicroChess. It was followed by Chess Challenger, the first electronic chess computer. Soon, chess computers and programs for home computers filled the market, including the best-selling chess franchise in history, The Chessmaster 2000, which was introduced in 1986.

Popular Mechanics

Which brings us back to Kasparov and Deep Blue. As world champion and renowned chess personality, Kasparov himself had gotten into the chess computer game, putting his name on a number of Saitek chess computers, pricey devices that claimed they could teach you to play the game. (That includes 1993's Kasparov's RISC 2500, an ad for which is found in the pages of Popular Mechanics). But nothing had prepared Kasparov for playing a computer as advanced, as sophisticated as Deep Blue and Deeper Blue. "I lost my fighting spirit," Kasparov told the press that day about his battle against artificial intelligence. 

Twenty years after Deep Blue, computers have far eclipsed human ability. While computers have not mathematically solved the game (it's an equation that may never be solved), humans don't stand a chance against the top machines. Game-playing computers have moved on to other games, like Go, Jeopardy!, checkers and Texas Hold 'em. Soon the machines will probably smoke the top poker champions, too. That's what we built them to do.

This content is created and maintained by a third party, and imported onto this page to help users provide their email addresses. You may be able to find more information about this and similar content at piano.io