Ben:
> The procedures contained inside nodes are expressed as tree structures
> which may be textually expressed in a language called Combo. These
> Combo procedures may be expanded into semantic nodes and links for the
> purpose of reasoning on them. Also, inferentially derived knowledge
> expressed as semantic nodes and links may be "compiled" into
> executable Combo procedures via a process called "predicate
> schematization."
> which may be textually expressed in a language called Combo. These
> Combo procedures may be expanded into semantic nodes and links for the
> purpose of reasoning on them. Also, inferentially derived knowledge
> expressed as semantic nodes and links may be "compiled" into
> executable Combo procedures via a process called "predicate
> schematization."
First, can you define "procedural knowledge"? I take you don't mean "motor" knowledge such as how to walk.
IMO procedural facts can be stored as declarative knowledge using standard KR. For example, "eggs should be boiled for eating" or "one cannot put on socks after putting on shoes" or "2 + 2 is 4".
When we need to find the sum of "2 + 2", it could be draw from the declarative knowledgebase.
It seems that your approach is to store the function "add(x,y)" directly *inside* a node. This destroys the nice uniformity of the KR. Secondly, the AGI should be able to process addition just like ANY other concept. "add(x,y)" is inside a node and therefore requires special processing (not AGI!).
Planning and action generation should be done by searching for the goal state within a state-space, by drawing facts from the declarative knowledgebase. For example the goal is "eat something" and there are some eggs. Declarative knowledge says that eggs can be boiled or fried for eating. So boil / fry some eggs.
> I agree this is more complex than having a flat knowledge
> representation, but we have found that for several reasons it is
> valuable to have separate representations for procedural and
> declarative knowledge, and then methods for converting between the two
> representations.
Do you store procedural and declarative in 2 different places?
It sounds like cheating because you may store the entire procedure for solving some fixed problems instead of searching for solutions in the declarative knowledgebase.
YKY
To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/[EMAIL PROTECTED]
