Welcome To The Machine

Datta maps the chequered history of computer chess, from P T Barnums Turk to IBMs Deep Blue

The most paradoxical thing about computer chess is that it was first played some six decades before the first computer. By the 1880s, P T Barnums The Greatest Show on Earth was facing stiff competition, as well as opprobrium from disapproving Victorian intellectuals who deplored the marketing of entertainment to the lowest common denominator.

So, Barnum dreamt up computer chess. In between the Fat Lady and the Sword-Swallower, he installed the Turk a moustachioed, turbaned one-armed bandit which played and beat most comers at the Royal Game. It was only three decades later when the Smithsonian took over the Turk that all was revealed.

 

It was done with midgets and mirrors. A small human sat scrunched in the Turks abdomen, seeing through cunningly placed mirrors and moving the Turks arm by a system of levers, a succession of top class players manning it. The first operator was Isidore Gunsberg who stood four feet-eight inches tall and narrowly lost a world title match to Wilhelm Steinitz. Barnum was a good enough paymaster to even tempt another great, Harry Nelson Pillsbury, who was six foot plus to do a stint.

The concept of playing chess with a machine has always struck strange chords in the human psyche. It has also always fascinated computer experts. The cross fertilisation is not very surprising, since a lot of chessplayers are also computer experts and vice versa.

Chess is one of the most difficult activities to program. Given the mindboggling number of 10 to the 120 possible positions, brute force isnt a perfect solution even with massively parallel systems like Deep Blue. Intelligent pruning is vital and setting up an algorithm for that involves insights into the mental processes of great chessplayers. Developing such software has led to many serendipitous insights into the act of pattern perception and efficient data-basing to store and access those patterns.

The history of computers and chess have thus been linked in an intimate fashion. Alan Turing, computer genius and chess tyro developed a paper chess computer. This consisted of a set of rules an algorithm applicable to any chess position. Virtual software, as it were; it could be run by a non-chessplayer without a machine. Von Neumann and Sakharov also did some seminal work before being sidetracked into fusion bomb development.

The advent of the PC really set off the chess rave, processing time now being cheap and easy. Most PC-enthusiasts liked their wood-pushing they still do if the average of two million hits a day during the Kasparov- Deep Blue matches is any indication.

The first strong computer chess programme was Sargon. Named after the great conqueror of Akkad, it was a minor cult figure in the early 70s. It was good enough to beat the average coffee-house chess freak though it didnt really measure up to competitive levels.

Then in 1975, David Levy, minor International master and computer expert took a 10,000 bet that he could beat any computer program designed in the next ten years under standard tournament conditions. He won his bet, winning a couple of matches and sending a lot of software engineers back to the drawing board. One flaw he exploited effectively was the horizon effect computers only looked a specified number of moves ahead before making value judgements. So, if a human set up something that operated over the horizon ....

Despite Levys successful wager, Mikhail Botvinnik, multiple world champion and mathematician, predicted that chess programs would be better than the best humans by 2000. Most people said Botty had gone batty and he died in 1995, two years before his favourite pupil Garri Kasparov played supporting actor in his screenplay. Around 1975, Big Blue also offered a million-dollar award to any chess programme that beat a reigning world champion. That money has, of course, been channelled back to the Deep Blue program.

Chronologically, the next great leap forward in computer chess was not playing programmes, but databases. Chessplayers dont like reinventing the wheel, so they check out previously played games for ideas when preparing. With a corpus of half a million competitive games recorded since the 1850s and something like 30,000 new games annually, electronic indexing and pattern matching is essential. By the late 80s, average playing standards had improved dramatically because of the use of silicon chessbases.

But programs still played lousy chess. They were greedy, they lacked technical skills, and most important, they repeated the same mistakes over and over again. Hence, new programs would be released with much fanfare, they would have a couple of good results and, then they would turn into tournament fodder.

That situation only started being rectified with the Fritz series. Fritz was created by Anglo-Indian Friedrich Freidel, now Kasparovs advisor. Freidel added features such as flexible time controls combined with assessment outputs that made it an excellent test-lab for new ideas. He also found a way of marrying it to databases, giving it an opening flexibility unrivalled by anything until Deep Blue.

Every year he produces an update with new features and a more powerful playing algorithm. Most importantly, it was affordable at $150-odd, running well off a basic 386 and getting stronger with a Pentium platform. Fritz 3 was good enough to win occasional games against Kasparov, Karpov and Anand at blitz time controls rating out close to GM standards. Fritz 4 is even stronger, improving on Kasparovs play against Deep Blue when used in analysis mode. Fritz now has several rivals in Hiarcs and Chess Genius, which play at roughly the same strengths and possess most of the same databasing advantages. These also make Chess DTP (Desk Top Printing) much simpler since you can port games in and out.

By 1993, chess playing programmes had got strong enough to lead to several ingenious scams. While it is considered legit to use computers to database or analyse adjourned positions, Correspondence Chess has been vitiated by accusations of totally computer-aided play. It is still evolving a set of enforceable rules to limit computer usage.

Scams have happened in over-the-board play also. One of the most blatant was when a player who called himself John Von Neumann picked up the Best Amateur award at the 1993 Philadelphia World Open. Von Neumann wore a personal stereo to aid concentration, and he played at the same pace regardless of his time status. He looked blankly at the ceiling and muttered subliminally in between moves. He lost several games on time including one bizarre game which lasted for just eight moves. After the eighth move, he simply stared at the ceiling for 90 minutes. Upon being challenged to play with his earplugs off, or just furnish ID, he walked out without the $5,000 prize. The consensus was that an accomplice somewhere was relaying the moves to a computer running at a specific time control and some electronic glitch occurred in the 8-mover.

With the advent of pocket computers, this sort of scam will be increasingly difficult to pick up. Perhaps chessplayers will graduate to playing au naturelle in order to ensure naked unaided mental struggle. Which brings us to the redefinition of chess as a sexual metaphor. According to the Freudians, chess represents the Oedipal struggle with the desire to kill the father being symbolised by the idiom of checkmate. Not coincidentally, the queen represents the mother-figure and is the most powerful of pieces.

But Sexy Chess, which is a new playing program carries things further. Here the pieces are anthropomorphic and get all hot and bothered when there is game afoot. The queen goes topless into battle and every capture is depicted as an act of coition. Checkmate is a gang-bang. It isnt much of a programme so far as chess strength goes. Of course, for obvious reasons Im not going to mention the address of the website concerned.