What was the Mechanical Turk, the chess machine with a person inside?

For more than eighty years, a wooden figure dressed in a turban and robes sat behind a cabinet and played chess against all comers. It shook its head at illegal moves, nodded when it threatened its opponent's queen, and usually won. Audiences in Vienna, Paris, London and, later, American cities lined up to see it. Many smart observers were convinced it was a pure machine.

It was not. A human chess master sat hidden inside, guiding every move. The Turk is one of history's great hoaxes, and also one of the first public debates about whether a machine could do something that seemed to require a mind.

When and where it began

The inventor was Wolfgang von Kempelen (1734–1804), a nobleman from Pressburg in the Kingdom of Hungary. According to the usual account, he watched a French conjurer named François Pelletier perform at Schönbrunn Palace in Vienna and promised Empress Maria Theresa he would return with something more impressive. The result made its debut there in 1770. Accounts differ slightly on the date: Edgar Allan Poe, writing in 1836, said it was invented in 1769, so 1769 may be when it was built and 1770 when the court first saw it.

The first challenger was a courtier, Count Ludwig von Cobenzl, who lost quickly. Kempelen himself seemed embarrassed by the fame of what he reportedly called a "bagatelle", a trifle, and for years he avoided showing it. In 1781 Emperor Joseph II ordered it rebuilt for a state visit, and from 1783 it toured Europe. In Paris it lost to François-André Danican Philidor, the strongest player of the day, and played Benjamin Franklin.

After Kempelen's death the machine was bought by Johann Nepomuk Mälzel, a Bavarian musician with a passion for machines, who had patented a kind of metronome. Under him the Turk played Napoleon in 1809 and later toured the United States. It was destroyed by fire in Philadelphia on 5 July 1854.

How the trick worked

The cabinet was about 110 centimetres long, 60 wide and 75 high, with a chessboard on top. Before each game the presenter opened its doors one at a time, revealing gears and clockwork and even letting the audience see straight through. Meanwhile the hidden player slid along a moving seat to stay out of view.

The clever part was how the player followed the game. The board on top was thin, and each chess piece had a strong little magnet in its base. Under every square, inside the cabinet, hung a second magnet on a string. A piece standing on a square held its magnet up; lift the piece and that magnet fell; set it down elsewhere and the magnet under the new square was pulled up. The player, working by candlelight, read the game from these hanging magnets, using numbers 1 to 64 marked under the board, and replied using a small pegboard linked to the figure's arm.

Here is one move, step by step:

A pantograph is a set of linked rods that copies a movement from one place to another, the way old drawing tools let you trace a picture at a larger size. The operators over the years included strong players such as Johann Allgaier and William Lewis; the person inside during Kempelen's first tour is still unknown.

Why it mattered

The Turk forced people to think about what a machine could and could not do. The sharpest example is Edgar Allan Poe's essay "Maelzel's Chess-Player", published in April 1836 after the machine visited Richmond, Virginia. He compared the Turk with Charles Babbage's calculating engine. Arithmetic, he argued, moves in fixed steps from its starting data, so a machine can follow it. Chess is different: no move follows necessarily from the one before, and the opponent keeps changing the situation. Therefore, he concluded, a mind must be directing the Turk.

The essayist was right about the Turk and wrong about chess. He assumed that anything without a single fixed path was beyond machinery. In the 20th century, researchers showed that a computer can search through possible moves and weigh them, which needs no fixed path at all. On 11 May 1997 IBM's Deep Blue became the first computer to beat a reigning world champion, Garry Kasparov, and interest in the old Turk revived with it.

The Turk also gave its name to something very current. In 2005 Amazon launched Amazon Mechanical Turk, a web service that hands small tasks computers still struggle with, such as comparing colors, to human workers. The name is a knowing wink: behind a screen that looks automatic, people are doing the part the software cannot.

How did real computers finally learn to play chess?

The first real chess machines came in the 20th century. In 1912 the Spanish engineer Leonardo Torres y Quevedo built El Ajedrecista, which used electromagnets to play a simple endgame, king and rook against king, with no hidden human. After electronic computers arrived, researchers wrote programs that look ahead several moves, score the resulting positions and choose the most promising line.

That approach, made faster and deeper over decades, produced Deep Blue's 1997 victory. Its success came less from a brand-new idea than from careful engineering and enormous gains in computer speed: Deep Blue could examine around 200 million positions per second.

So Edgar Allan Poe's question has a twist ending. A machine can play chess far better than the Turk's hidden masters, but not by imitating their thinking. It wins by searching and calculating at a scale no person could manage, which raises a new question the Turk never did: is that the same as understanding the game?