RE: NPR, Godel, Semantic Web

["Bullard, Claude L (Len)" <clbullar@ingr.com>]

Logical systems as closed systems don't 
usually trip too badly.   As I've said 
before, don't fly the first one.  As medium 
closed, semi-permeable, I'd 
say Danny is probably right; there are 
some feedback chaos-like issues, but 
even then, I don't see insurmountable 
obstacles except horrible expenses where 
the knowledge base is large and there are 
lots and lots of rules.  Remember that 
expert systems were the result of scaling 
down big ambitious knowledge projects to 
some reasonable domains.  I can definitely 
see a place for that on the web and I think 
we already are for example, in the financial 
systems for mortgage applications, etc.  

For a system to be semantically interoperable, 
someone has to have a very good domain concept 
(common information model or CIM) and a lot 
of bucks for the customized version that runs 
according to the local ricebowl legacies. We 
will get some of this by default from schema 
efforts, and lots of locally interoperable apps, 
but there is a time and the river problem where 
by the time you get the boat built, the river 
has moved on, so aim for the other shore, not 
an eddy in the current.   It may be something 
that emerges bottom up.  Still, pattern-seeking 
analyses have to cope with superstition and gaming 
so when hooking up semantic engines, eg, pattern 
seeking bots, make sure they go to only the 
*best* places for information then expect to 
be surprised by the infoGhettos (used to be 
called the bazaar).

In the Sci Am article, as I recall, the two main 
points were a common language (universal 
system... bleaaachh) and some simple 
and widely available tools for adding the 
information to the web site.  Now the 
problem becomes inter-site and inter-domain 
reasoning (logical layers, assertions of 
authority such as all the state rules for 
crime classification vs the federal rules and 
so forth).  The overlaps, the ecotones, are fascinating 
and dangerous places but we already build 
systems like this.  Expensive beasties.  The 
presence of a data standard (say NIBRS) makes 
it economic to do; the requirement for customization 
makes it a business. 

Len 
http://www.mp3.com/LenBullard

Ekam sat.h, Vipraah bahudhaa vadanti.
Daamyata. Datta. Dayadhvam.h


-----Original Message-----
From: Simon St.Laurent [mailto:simonstl@simonstl.com]

At 11:14 AM 5/7/01 -0500, Bullard, Claude L (Len) wrote:
>The semantic web doesn't trip on godel or
>incompleteness.   It trips on authority.

It may well trip on authority, but this claim suggests that it also trips 
on inherent limitations of logical processing.

>The crockness of it isn't the doability;
>it is the need to do it now.  Good
>topic for research, good topic for
>discussion; perhaps not the initiative
>by which all other tasks before the
>W3C et al should be measured or circumscribed.

I've got to agree with that, though the Semantic Web does strike me as a 
better yardstick (less likely to lead to horrible designs) than some of the 
other possibilities.