Re: What do the ontologists want

> > I grew up in a world that was full of people with vague
> > ideas about how anarchic love-ins were going to change society, and
> > most of those people are now stone cold dead. I prefer to live in the
> > real world.
>
>I guess I hit a personal button with the word "anarchic."  I'm sorry.
>I mean it in the sense of "without government or formal authority"
>(sense 1a in my dictionary) not "the absense of order" (sense 2b).

Ah, sorry I misreacted to a misunderstanding. (There is another jibe 
in a message crossing with this which should be edited out.)

> I
>mean it in the same way that the Web is by design different from most
>previous hypertext systems: it is free of any central authority or
>bottleneck (beyond those few in the Internet).
>
>It's a lot harder to make system without that central bottleneck, and
>sometimes you have to give up features (like link consistency).  But I
>think the Semantic Web is supposed to share this kind of anarchic
>quality with the existing web.  (And I think Tim B-L and probably Jim
>Hendler would agree with this goal.  I'm not familiar with Matthew
>West.)

Well, I agree that it is free of central *authority*. In fact my own 
view  is that the whole idea of languages like DAML is easy to 
misunderstand, since it can be seen as trying to create a 'single 
standard' language, and that the very nature of the Web is such that 
there arent going to be any single standards. The moral being (yes, 
from AI/Krep work) that it would be better to focus on finding 
universally acceptable protocols for locating and negotiating 
semantic disconnects, rather than facilitating semantic transfers. We 
have plenty of formalisms for stating complex facts already, many of 
them in widespread use in various communities. In fact it is a 
constant ongoing problem of babelization, and the knee-jerk reaction 
to it is always to define a 'standard'. That is what KIF was supposed 
to be (Knowledge Interchange Format), but it never succeeded in that 
way. Nothing ever does, and nothing probably ever will. Right now 
some very smart and well-meaning people are trying to create a 
standard upper ontology, which is like trying to legislate all of 
human thought; not a chance in hell it will be any real use. So there 
will always be lots of languages, and yet we need interconnection. 
What we need are ways of translating between the different languages, 
and things like logic-scrapers that try to extract as much content as 
they can but know that they arent going to get it all, and so on. And 
these things already are starting to happen. But what we don't have 
is anything like a content transfer protocol, and thats what the 
semantic web REALLY needs, imho. People will write the translators 
and so on: thats not our job. But what are these things going to have 
to be able to say to one another? Like, you ask me for information 
about foo in language L, and I say, everything I know is in language 
M, sorry. And then you have to find someone who can translate L into 
M, so you go to your favorite translation service and say to it, can 
you translate L into M? and it says yes, so you say to it, OK, please 
tell Pat to send you what he knows about foo and then send me the 
translation, OK? And then it says to me, Sandor wants me to translate 
what you know about foo for him, is that OK? and I say, sure, and 
send it what I know about foo (in L) , and so on. Now, all the actual 
content-manipulation is being done by these various engines out there 
(Stanford already has a service like this, but less automatic, and 
there will be hundreds soon), and this is all just file-transfers as 
far as the CTP is concerned; but all the web-based negotiating is 
taken up with the meta-level discussions about who knows what 
languages and what language something is written in and who knows how 
to translate this into that and who can verify that this in L really 
does mean the same as that in M, and so on. This is where the 
semantic web is going to get really interesting, and what will really 
enable it to be anarchic-2b. It could support an entire economy of 
information trafficking and checking and brokering. And as far as I 
can tell, nobody is even thinking about this seriously at all. (Maybe 
the XML world has people who are. I confess to finding that world 
very hard to penetrate.)

> > > What lessons have such systems taken from AI?  What
> > >lessons should they have taken?
> >
> > There are no such systems, so how can I answer you?
>
>Well, that was the retorical part of my question.  You said RDF was
>doing old stuff and doing it badly.   I'm saying RDF is trying to
>borrow from old stuff to do something new, which isn't as "old hat" as
>you seemed to be claiming.

See my reply to Jonathan Borden. The webbish aspects of RDF are 
interesting - what exactly do URIs mean, that kind of thing, and the 
challenges that anarchy-2b introduce (though I think they have been 
overstated, but thats another discussion).  But I am not griping 
about that aspect of it.  The side of it that talks about assertions 
and expressions and consistency and so on IS old stuff, and it DOES 
do it badly. Which is a shame, because we know how to do it properly, 
and there are far more interesting things that we could be doing and 
that need to be done, instead of sitting next to the railway tracks 
making toy trains.

Pat Hayes

---------------------------------------------------------------------
IHMC					(850)434 8903   home
40 South Alcaniz St.			(850)202 4416   office
Pensacola,  FL 32501			(850)202 4440   fax
phayes@ai.uwf.edu 
http://www.coginst.uwf.edu/~phayes

Received on Thursday, 17 May 2001 23:09:21 UTC