Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: HAL 9000

Author: Roberto Nerici

Date: 05:15:53 03/29/04

Go up one level in this thread


On March 29, 2004 at 08:05:28, Sune Fischer wrote:

>On March 29, 2004 at 07:25:54, Roberto Nerici wrote:

>>But back to the chess bit! If I remember correctly, Murray Campbell's point is
>>based on the observation that Hal plays a non-optimum "trappy" move. The
>>intelligence is that it knows a very effective move _to make againt Dave_ (or
>>whoever it's playing) because it knows Dave and it knows the kind of mistakes he
>>makes, rather than it just making the theoretical "best" move.
>
>Yes, patzer chess.
>Not something I'd recommend, your might underestimate your opponent, perhaps he
>has learned from last time.

Ah, but HAL could watch everyone wherever they were so it would know if its
opponent had been practicing :-)

>Still Crafty and many other engines do a bit of swindle mode and assymmetric
>evaluation which has a bit of the same smell to it.

I think the term for all this is "opponent modelling". I know nothing about it,
but Guy Howarth (in this forum) has mentioned it a few times with respect to
EGTBs.

Roberto/.



This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.