It should be a combination fo the two, even as Cyc is finding out now with
their use of Google to search out new terms and facts.
A really simple example of that is related objects... a book scraping can
generate a listed of objects it thinks are related, then it can ask the user
how they are related, or to correct it if it is unsure about a rule it learned.
Imagine though a very easy to use Cyc, that was in simple english (no Cyc-like
language) that any user could interact with in a nice interface.
Given a few million interactions, you could quickly gather much useful
information about any given topic. (re Wiki pages)
James Ratcliff
Mark Waser <[EMAIL PROTECTED]> wrote: >> Much of commonsense knowledge is
not explicitly stated in books or other reading materials for adults
Almost all commonsense knowledge *is* explicitly written down somewhere
-- just not restated constantly in "reading material" (where it would be
redundant). Dictionaries and encyclopedias (and even self-help books) are
wondrous things . . . . :-)
>> many logical rules would have to be inductively learned while the machine
>> is doing the reading.
Actually -- No, I don't believe so. Grammar, definitions, and
inheritance can all be used to *MASSIVELY* reduce the number of rules (as in,
by several orders of magnitude) that need to be learned (and inductive
learning is *not* a particularly effective form of learning at this scale/in a
text environment).
>> It seems much easier to simply ask humans to encode the facts / rules.
Why does it seem easier? No one has gotten it to work in the past. What
do you think you know that they don't?
To me, it seems much easier to have an automated system encode the
facts/rules.
Mark
----- Original Message -----
From: YKY (Yan King Yin)
To: [email protected]
Sent: Tuesday, April 17, 2007 3:30 PM
Subject: Re: [agi] AGI interests
On 4/18/07, James Ratcliff <[EMAIL PROTECTED]> wrote:
> Mark,
> This is the closest Ive seen so far to my work and what I believe in,
> Have you got some more specific information / code / algorithm / papers on
> gathering and processing world information and discovery of?
> I have been working with text processing and getting a bot to "read" and
> process books/ newspapers as a main method of learning.
Hi James, Mark,
I'd be very interested in collaborating on this kind of project
(knowledge harvesting). My thinking is that it is best done via collecting
simple facts / rules from humans (using natural language eg Basic English),
rather than from books or the web.
Much of commonsense knowledge is not explicitly stated in books or other
reading materials for adults. Also, many logical rules would have to be
inductively learned while the machine is doing the reading. It seems much
easier to simply ask humans to encode the facts / rules.
What are your thoughts about this?
YKY
---------------------------------
This list is sponsored by AGIRI: http://www.agiri.org/email
To unsubscribe or change your options, please go to:
http://v2.listbox.com/member/?&;
---------------------------------
This list is sponsored by AGIRI: http://www.agiri.org/email
To unsubscribe or change your options, please go to:
http://v2.listbox.com/member/?&;
_______________________________________
James Ratcliff - http://falazar.com
Looking for something...
---------------------------------
Ahhh...imagining that irresistible "new car" smell?
Check outnew cars at Yahoo! Autos.
-----
This list is sponsored by AGIRI: http://www.agiri.org/email
To unsubscribe or change your options, please go to:
http://v2.listbox.com/member/?member_id=231415&user_secret=fabd7936
