Re: Patterns and intelligence (was: DGI paper)

From: Doug Keenan (
Date: Mon Apr 15 2002 - 05:43:21 MDT

> Another example: Who has the stronger patterns, Kasparov or Deep Blue?
> Under the simplicity definition, Deep Blue's computational processes are
> vastly simpler than Kasparov's, and so Deep Blue has a much stronger chess
> pattern. Under the tractability definition, Kasparov and Deep Blue were
> roughly evenly matched in skill when Kasparov was searching 1.8 moves per
> second and Deep Blue was searching 2 billion moves per second, making
> Kasparov's chess patterns enormously stronger than Deep Blue's. Deep Blue
> plays the Game of Actual Chess, searching through the tree structure of
> actual chess positions. Kasparov plays the Game of Regularities in Chess,
> using an enormously more complex algorithm to execute a "search" that
> roughly matched the gameplaying power of Deep Blue even when considering
> only 1.8 positions per second - if such a measure can be said to apply at
> all to Kasparov. Kasparov's chess patterns are vastly more complex than
> Deep Blue's chess patterns, and they render the game far more tractable
> for Kasparov than for Deep Blue.

This is not a dodomorphic fizzbin at all. :-)

Are we concerned with the rough loading of these patterns?

Human intelligence loaded Kasparov's complex patterns at what,
call it 25-35 years.

Human intelligence loaded Deep Blue's simpler patterns at what,
call it 5 years (I don't know how long they took to build it).

Other factors might fudge this (like the genius of Kasparov) but it's
interesting to note the pattern storage rate is similar but the retrieval
and compare rate is vastly different. Only a seventh his complexity
beat him when accessed 10^6 times faster.

(May I ask a typo-fix for DGI please? Garry has two 'r's.)


This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:38 MDT