Let me reply to everyone here...
Pei: You said non-numeric heuristics (such as endorsement theory) may run into problems. Yes, but I believe those problems can be solved using further heuristics (eg see wikipedia article on Nixon diamond). If you resolve the Nixon diamond by referring to probabilities like "how many % of republicans are pacifist" then you're still relying on domain-specific knowledge.
I believe that all AGIs are basically heuristical. Thus there are many ways to design AGIs. One way is to assign numerical truth values to all statements, universally. Another way is to use mostly symbolic qualifiers. Yet another way is the hybrid of these 2. There isn't a single correct way.
Richard: I now think that having NTVs for everything is an acceptable heuristic.
Now, figuring out all the heuristical NTV / symbolic qualifier's update rules, such that an AGI will always be internally consistent, and provably increasing in accuracy, is a very non-trivial task. It's akin to an axiomatic formulation of AGI. What we got here so far is a bunch of heuristics (eg Dempster-Shafer theory) that people claim to be able to deal with all situations, without proof.
Ben: I think the problem of contextuality may be solved like this:
Examples:
John and Mary have many kids. (like, 10)
This Chinese restuarant has many customers. (like 100s)
Many people in Africa have AIDs. (like 10s of millions)
so I propose a rule like this:
IF
n is significantly > the average / usual number of a thing X
or n is significantly > the majority of elements in a set Y
THEN
"there are many Xs"
or "there are many in Y"
Is that similar to your solution? =)
YKY
To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/[EMAIL PROTECTED]
