Monday, December 5, 2011

The Limits of Reason

In the late eighteenth and early nineteenth centuries, a device known as the Mechanical Turk toured Europe and the United States, astounding it’s audiences by playing chess with an impressive degree of skill.  Many leading thinkers and engineers of the day were utterly convinced that it was a true chess playing machine, although there were persistent suspicions that perhaps the device (which included a large cabinet, apparently filled with gears and cogs) hid a real human chess player, possibly a dwarf, in its interior.

One of the most convincing exposes on the subject was written by Edgar Allen Poe in 1836.  His article was printed in the Southern Literary Messenger, and elicited responses from numerous newspapers and magazines including the New Yorker and the Baltimore Gazette.  It was well written, and completely correct in it’s conclusions – the Mechanical Turk was eventually revealed as a fraud.  However, it’s interesting to note that some of the most critical arguments employed by Poe were utterly incorrect.

His primary point was that as a mechanical device, it must necessarily be 100% fixed and determinate.  From this foundation, he makes several observations about the nature of the machine: “The moves of the Turk are not made at regular intervals of time, but accommodate themselves to the moves of the antagonist – although this point (of regularity), so important in all kinds of mechanical contrivance, might have been readily brought about by limiting the time allowed for the moves of the antagonist.”

He goes on to make a very interesting argument about the nature of the machine's purported algorithm: “The Automation does not invariably win the game.  Were the machine a pure machine, this would not be the case – it would always win.  The principle being discovered by which a machine can be made to play a game of chess, an extension of the same principle would enable it to win a game; a further extension would enable it to win all games – that is, to beat any possible game of an antagonist.  A little consideration will convince any one that the difficulty of making a machine beat all games is not in the least degree greater, as regards the principle of the operations necessary, than that of making it beat a single game.”

I’m not aware that any of Poe’s contemporaries had any disagreements with these points.  However, from our modern perspective, it’s easy to see that he was utterly incorrect.  We have chess playing machines today.  And what we know about the principles of making them play chess did not inevitably lead to an understanding of how to make them play perfect chess – it took decades of improvement (mostly on the hardware side) to evolve from the first primitive chess program to Deep Blue, which finally defeated world chess champion Gary Kasparov.  Even now that we've left human players far behind, nobody would claim that any computer in the foreseeable future could play a “perfect” game of chess (a feat which would require calculating every conceivable move on the chess board – a trivial exercise with a game like tic-tac-toe, but infeasible on a 64 square chess board).  Nor has our success at developing computer chess programs led to similar success teaching computers to master the much more fluid game of Go.

What would it have taken to convince Poe that he was wrong, especially when his ultimate conclusion was correct?  We could try to introduce him to the concept of heuristics, or allowing variable response times through the use of triggers, or even introducing the concept of indeterminacy to a system by seeding a complex algorithm with an external (and completely random) value.  Maybe this would have worked.  But none of this would be nearly as effective as simply having him grow up in a world where computers play chess all the time, and nobody thinks very much about it.

We like to think of ourselves as living in the age of reason and critical thinking.  We like to point at our technological progress and say “Here is where logic and reason have taken us.  They will solve any problem.”  We forget how much of what we know is built not on reason, but on trial and error, iterative improvement, and pure luck.

No comments:

Post a Comment