Actually, explanation is not typically thought of as a 'validation' (or
confirmation) although it is a necessary part of confirmation, which is
probably what I meant.  First of all, I am using 'explanation' here to
represent 'understanding'.  The problem with this is that 'understanding'
is the elusive goal that we are seeking.  We need some way for our programs
to attain understanding.  So if we use understanding as a primary means of
theorizing (or making a theory-like objects of thought) then we would need
another method to act as confirmation.  The act of confirmation is usually
seen as arriving to a conclusion using an alternative method of insight.
This becomes subtle because when we are talking about an essential ability
of an AGI program we are not using these terms to mean something like being
able to integrate new knowledge in a knowing way or something like that.
We are not talking about the same kind of thing like confirming an arrival
time or confirming a story that a newspaper wants to print.  However, the
goal of getting a computer program to acquire knowledge that can be used
'intelligently' probably requires a number of different methods to validate
knowledge.  But again, from my point of view, they all require 'explanatory
power' or something like that. The use of prediction as a form of
validation uses the perception of an event (found in input) as
confirmation, but the real problem there is that most useful observations
are based on complex assessments of input, so here explanatory power is a
necessary part of an observation.  So, at the very least I can say, with
some confidence, that multiple paths of reasoning are necessary for
confirmation, but the inclusion of an observation is an important method.

Jim Bromer



-------------------------------------------
AGI
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-f452e424
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657
Powered by Listbox: http://www.listbox.com

Reply via email to