Mind.Forth Programming Journal: 11 August 1999 
Author Message
 Mind.Forth Programming Journal: 11 August 1999

HOW THE TIME-CLOUD WILL WORK

Culling out pertinent passages in the Mind.rexx ARPJ of SUN.27.-
NOV.1994 has reconstituted for me the thinking that went into the
time-cloud.

Suppose that the AI KB has recently learned, "Horses eat hay."
Now suppose that any two of those concepts have been freshly
entered, and the third must be found ("inveniend").

The time-cloud technique depends upon the "pre" and "seq" flags
of the http://www.*-*-*.com/ ~mentifex/aisource.html Mind.Forth
INSTANTIATE subroutine.  Temp{*filter*}proximity alone will not do the
trick, because a jumble of simultaneous concepts does not declare
which one is the subject and which is the object.

If all recent nodes of a concept like "horses" are activated during
HOLODYNE, then SPREADACT (and not TRANSFERACT) must follow "pre"
and "seq" flags to increase by some measure the activation of all
related concepts.  So the concept of "eat" will be activated, and
send activation to all its recent direct objects, including "hay."
However, here "hay" does not yet have an exceptional or special
activation.

This line of thought is getting really weird.  I just realized
that you could use reentry to drive the search for the "inveniend"
third concept.



Mon, 28 Jan 2002 03:00:00 GMT  
 
 [ 1 post ] 

 Relevant Pages 

1. Mind.Forth Programming Journal: 11 August 1999

2. Mind.Forth Programming Journal: 17 August 1999

3. Mind.Forth Programming Journal: 17 August 1999

4. Mind.Forth Programming Journal: 17 August 1999

5. Mind.Forth Programming Journal: 2 September 1999

6. Mind.forth Programming Journal: 19 JAN 1999

7. Mind.forth Prog Journal 20 Jan 1999 (Attn: computer architects)

8. Mind.Forth progress report Mon.30.Aug.1999

9. Mind.Forth PD AI: 10.Aug.1999 Progress Report

10. Mind.Forth AI status report thurs.1.jul.1999

11. Mind.forth State of the Art: 6 JAN 1999

12. Mind.Forth PD AI: 10.Aug.1999 Progress Report

 

 
Powered by phpBB® Forum Software