In the 1760s a certain Baron Wolfgang von Kempelen was the sensation of his day, amazing the crowned heads of Europe with his invention, the Maezal Chess Automaton. The machine was nicknamed the Turk, due to the turbaned marionette attached to the cabinet, which actually made the moves. Sadly, the good Baron’s wonder proved to […]
In the 1760s a certain Baron Wolfgang von Kempelen was the sensation of his day, amazing the crowned heads of Europe with his invention, the Maezal Chess Automaton. The machine was nicknamed the Turk, due to the turbaned marionette attached to the cabinet, which actually made the moves. Sadly, the good Baron’s wonder proved to be no breakthrough in machine reasoning, but merely a subterfuge for hiding a vertically-challenged chess genius who was secretly moving the Turk’s levers. Two hundred and thirty years later, and the modern day equivalent of the Turk could conceivably hold one, perhaps two, average sized chess dervishes – but the twin six foot black processor towers of IBM’s Deep Blue, the world’s greatest chess playing machine, hold, so far as we trust, no hidden human assistants, only a massively parallel processor farm with some dedicated chess playing chips.
By Gary Flood
Yesterday Garry Kasparov took on Deep Blue for the fourth in the six game contest of the rematch between the world’s human chess champion and the best ever AI (artificial intelligence) rival. On the 50th floor of the Equitable building in the Upper West side of Manhattan BarbedWire joined a curious mix of gray-haired tweed-jacketed Grand Masters, lolling on some suitably comfortable couches, programmer weenies in polo shirts and glasses, and excited power-dressed media types, in viewing on closed circuit television the mock-up of a cultured gentleman’s library a few feet away. There sat a madly frowning blocky-faced man, Kasparov, next to a small flag of the Russian Federation, facing over a standard chess board with fairly ordinary wooden pieces a much calmer representative of his opponent, who spent his time glancing at the laptop which conveys the thoughts and reactions of the chess computer. Here at the very end of the 20th century is a contest that marks how far computer science has come in the fifty-odd years it has been a practical tool. If Deep Blue can beat the world’s best living chess player, some Rubicon of machine reasoning will be forever crossed, and one of the attributes of cognition that seemed hardest for silicon based entities to match will no longer be unique to humans alone. Kasparov himself seems to have ambiguous feelings about the contest, from which he can hope to win $700,000; on the one hand he feels he is champion of all that is human, in another he thinks he’s actually fighting Deep Blue’s programmers. In one respect, I am trying to save the dignity of mankind by playing in this match, he has said. In another respect, it is a team of research scientists who created this computer system, and they are really my opponents. Let’s see what they’ve come up with in terms of hardware and software to challenge the power of the human mind. The contenders fact sheet does a deadpan heavyweight versus heavyweight style comparison. Under ‘height,’ we see Deep Blue has the edge on Kasparov, at six feet five inches versus five feet ten; and it also has the advantage in ‘weight,’ at 1.4 tons versus 176 pounds. Kasparov has experience on his side – he’s thirty years older than the box of silicon, at 34. Kasparov can count on using his 50 billion neurons as his ‘processors,’ while Deep Blue must trust to its 32 P2SC microprocessors and 512 dedicated chess chips, granting Kasparov an estimated two moves a second versus Deep’s 200 million (or 50 billion for the maximum alloted three minutes per move). Deep Blue marks a high-point in a particular area of computer science, certainly. The idea of using chess as a platform for exploring and developing machine intelligence goes back as far as 1950, with the work of Claude Shannon, drawing on the so-called minimax algorithms for artificial game play devised by John von Neumann and Oskar Morgenstern, though there had been earlier suggestions and even some programs from figures like Charles Babbage, Alan Turing, and Konrad Zuse. Like many AI founding principles, the idea of developing a general purpose intelligent program to solve a problem in a way comparable to humans proved impossible. Instead, chess computers had to rely on brute force computational oomph to hope to compete with their carbon-unit antagonists, and a set of rules of thumb, rather than generate an arbitrarily clever chess brain. It became a kind of law for a while there that for every five fold increase in hardware a chess program could leap two hundred points in the international scaling system for chess players. Bell Labs’ Belle, developed by Ken Thompson, one of the fathers of Unix and a Technical Judge for the Deep Blue-Kasparov rematch, was the first ever program to reach the master’s level, with 2,200 points, in 1983. An earlier model IBM chess computer, Deep Thought, was the first non-human to be granted Grandmaster-level rating in 1988, and much of Deep Thought’s team graduated onto Deep Blue from this project. IBM sees future applications of the kind of technology Deep Blue represents – throwing computer resources at highly complex problems which are still bounded by simple rules – in areas like pharmaceutical drug development (molecular dynamics), financial risk assessment (speedy assessment of the values and risks associated with a large number of stocks and portfolios) and decision support (by which IBM means Data Mining). For example, the first application might see machine analysis of compound interactions at the molecular level; IBM speculates that this could bring the average time of development of a new drug down to eight or even six years, from today’s 12 (at a typical $12m price tag). But useful as such applications will be, they lack the drama of Man versus Machine in a sporting contest such as this. The competition is not yet decided, at one game each and one draw, and we went to press before Wednesday’s game was decided. IBM believes Kasparov played cat and mouse with Deep Blue in the first game, making the computer react in a way the excitable writer of the press release found a little bemused. The win by his AI opponent on Monday provoked a reaction that Deep Blue will never be able to copy, at least – the irritated Azerbaijani (and former contender for the mayorship of that Russian republic’s capital, Baku) left the playing area at great speed without comment. Plus, Deep Blue will certainly never be able to copy the frowning power of Kasparov, a demonstration of brute forehead muscle strength that left us fearful of the chap’s ability to support his eyebrows in future. Seriously, though, Kasparov is no slouch. He is hailed in chess circles for an amazing capacity intuitive play and an ability, which outfoxed Deep Blue in 1996, of being able to switch tactic in mid-game. One can cheer for the achievement Deep Blue represents, but one cannot help secretly rooting for the moody, jowly Azebaijani – in other words, us, the Earthlings. Who cares for the egos of Wall Street arbitrageurs if they get beaten by computers – chess is still a definitional aspect of human intelligence for most of us carbon units, and we still want to believe that will always remain so.