- From: Seth Russell <seth@robustai.net>
- Date: Fri, 28 Jun 2002 09:48:34 -0700
- To: "Danny Ayers" <danny666@virgilio.it>, <www-rdf-logic@w3.org>
From: "Danny Ayers" <danny666@virgilio.it>
> >>>In practice, an implementation could contain two knowledge bases - one
> >>>'live' and one 'sandbox'.
>
> A slightly embarrasing thought just occured to me - this is rather like
the
> short-term/long-term memory setup used in some of the old (untrendy)
> production systems. Let me know if flares come back into fashion
(again)...
Hmmm ... I've been toying with the same thought. The long term memory of
the quads is in a database accessable to the agent. The short term is in
fast access memory structures in RAM. There is a whole cadre of logic that
attempts on the one hand to make the short term and long term dicotomy
transparent to the user; on the other hand it attempts to minimize the
respose time to the user.
Seth Russell
{{
seth (wants to show you) (sailor agent).
(sailor agent)
isa (software agent), (social agent);
purpose "to sail the information ocean";
status (preview alpha);
seeUrl <http://robustai.net/sailor/>
comment "Sailor agents don't have long term memory yet".
}}
Received on Friday, 28 June 2002 12:54:58 UTC