IP: Economist on Deep Blue

[David Farber <farber () cis upenn edu>]

> From May 3 Economist


Fool's mate


IBM's chess-playing computer may see beating Garry Kasparov as a leap
forward for artificial intelligence. Sucker




SOMETHING about chess champions scares ordinary mortals. Even the title
"grandmaster" seems chosen to emphasise their superiority to most people.
In the minds of many, playing chess well ranks alongside understanding
higher mathematics or the more obscure by-ways of Wittgenstein's philosophy
as a testing ground of human cleverness. The idea that a machine could be
cleverer is therefore disturbing. If machines can beat even champions at
chess, what area of human endeavour is safe from their encroachment? Over
the coming weeks most humans are therefore likely to be cheering for Garry
Kasparov, the current world champion (and, in the opinion of many, the best
chess player in recorded history), during his latest duel with Deep Blue,
an IBM computer. In their last meeting, in February 1996, Mr Kasparov won
the contest 4-2. But Deep Blue has been upgraded since then, and there is a
good chance that it may win this time round, even though Mr Kasparov has
meanwhile been busy upgrading himself. If Mr Kasparov does win, despite the
odds, it may give his audience a warm feeling. But, in truth, if Deep Blue
were to beat him it would not really matter one jot. It would not show that
Deep Blue was cleverer than any human alive. It would merely confirm
something that has been known for a long time: that chess is trivial.


One mark, Watson, of a scheming mind? That does not mean that playing chess
is easy. But as John von Neumann, one of the founders of modern computing,
observed over 40 years ago: "Chess is not a game. Chess is a well-defined
form of computation. You may not be able to work out the answers, but in
theory there must be a solution." What IBM's programmers will have done is
to press the quest for that solution (which is merely a matter of logic and
number crunching) beyond the point where human thought processes can keep
up with it.


What they have not done is to reproduce human thought, which is the usual
declared goal of artificial intelligence. No computer programmer has done
that. Programs intended to play real games -- those where bluff and
deception as well as calculation are involved -- have not made nearly so
much headway as those that play chess. There is certainly room, as Sherlock
Holmes observed to his biographer, for scheming in chess. Indeed, it was by
observing his opponent and outscheming it that Mr Kasparov beat Deep Blue
last year, even though he lost the first match of the series. The
significant point was that Deep Blue could not scheme back; it just had to
follow its imperfect algorithm. Human intelligence is strong on scheming.
There is a respectable line of evolutionary argument that scheming, and the
manipulation of the behaviour of one's fellow humans that goes with it, is
the primary biological purpose of intelligence. Outmanoeuvring that by
means of number crunching -- the only way that computer programmers
currently know -- is not the same as emulating it. In the case of chess,
success has not been an indication of anything remotely approaching a
scheming (and so truly intelligent) mind. Programmers do not know how to
tackle that problem. There is another reason not to feel too crestfallen
even if Deep Blue does beat Mr Kasparov. For it, too, is a product of human
intelligence, in all its scheming cleverness. Only when people have built
machines that can themselves design and build machines such as Deep Blue
(and, more to the point, want to) will people's pre-eminence be seriously
threatened. Of course, Deep Blue might object, if it had been so
programmed, that this definition of intelligence is deeply unfair. Despite
von Neumann's stricture, outplaying people at chess has after all been a
sub-goal of artificial intelligence for years, and now that the goal is in
sight, people are moving its posts once again. This may look to a computer
like a particularly Machiavellian example of changing the rules in
mid-game. But that's people for you. Clever, aren't they?


The Economist Home Page   -   Search   -   Contents -   Next article


Copyright 1997 The Economist Newspaper Limited. All Rights Reserved