Re: Deep Blue

[ddyer@xxxxxxxxxxxx]

Shortly after Deep Blue vs Kasparov-II I attended a seminar given by one 
of the deep blue team members, and had the opportunity to ask a few pointed
questions.  Here's what I learned:

(1) they are employing many of the standard tricks for making Alpha-Beta 
    pruning more effective.
(2) they have special tactical units for various well-understood chess situations.
(3) they have used grandmaster advice to hand tune their static evaluations.
(4) they have lots of raw speed.

They're not using any AI, machine learning, or sophisticated state-of-game modeling
techniques.

As it applies to Go, this means that the deep blue team doesn't know
anything not known to most of the members of this list.  What they do 
know, they've applied well; and it helps to have large budget.

---
I'm preparing to switch ISPs.  To make this the last time
you have to care (HA! Fat Chance!) please use ddyer@xxxxxxxxxxxxxxxxx