RE: SUO: ELP's summary of MRW's standards experience

["John F. Sowa" <sowa@bestweb.net>]

Eric,

Some comments:

>[ELP] I might restate that as saying that an ontological model can
>reconcile OODB's and RDB's.  OODB logic is clearly different logical
>language from RDB logic - with much overlap.  But I suspect we are not
>disagreeing here.

I agree that the languages are very different.  But what I
was trying to say is that an RDB and an OODB are capable of
expressing logically equivalent information.  The query
languages, however, are very different.  Many people who
use OODBs access them with SQL, but that is partly because
the programmers are more familiar with it.

JS> ....  The main
> problem with existing ontologies is that the main candidates
> don't have solid foundations. In fact, Doug Lenat is the first
> to admit that the Cyc upper levels are not very important.  

>[ELP] I've found that John S. the consensus builder bends over backwards
>to makes conciliatorily general technical statements for building
>consensus.  But I've also got the sense that the specific technical
>opinions of John S. might be far from what could be wrongly interpreted
>from his general consensus building statements.
>
>Perhaps Doug is falling victim to this here.

I don't know what you mean by that.  I frequently criticize
Cyc for many different reasons, and I say good things about it
for many other reasonhs.

In any case, Doug has repeatedly said that the upper levels
are not very important, and he repeated that statement last
week in Austin.  So I asked him why he didn't throw away the
upper levels and just use the middle and lower levels.  He
said that there are many important inferences that are implied
by the general principles included in the upper levels.

JS>... I
> agree with him -- primarily because I think that the Cyc upper
> level is a disaster zone and that a better foundation would
> simplify much of the lower levels.

>[ELP] I was told that OpenCyc was or is an ANSI draft standard for upper
>ontologies.  So if you really want to agree with him, it appears to me
>that you need to view OpenCyc as at least a reasonable draft standard
>for upper ontologies.

I don't want to agree with Doug for the sake of agreeing.
And I certainly stand by the claim that the Cyc upper levels
are seriously wrong.  Doug's position is that he is not
enamoured with the upper levels, but he doesn't see any
pressing need to change.  But I believe that we can do
something much better for the SUO.

>Also, I have no problem with Cyc-bashing if it is done in concrete
>verifiable specifics.  I think they would appreciate that kind of
>feedback -- but one should warn them first or they might be quite
>startled ;^)

There are quite a few Cyclers who subscribe to this list and
post comments regularly.  I'm sure they're aware that many
of the Cyc features have been criticized.

>  But if ontology is science as you say, and if you want to
>be fair minded, which I fully believe you do, shouldn't your criticisms
>reference specific axioms and make claims for which you have already
>supplied concrete proofs.

There is an enormous amount of work to be done in analyzing Cyc.
I and many other people have done some of that work over the
years, including the joint review of the Lenat and Guha book
on Cyc (AI Journal, vol. 61, 1993).  Lenat told me that my
review raised some painful issues that they were aware of
and hated to see discussed in public, but he considered it
the fairest one of the bunch.

>This sort of exercise would be indistinguishable from the sort of
>interchange that would be needed in the merging that you have advocated
>and in the comment/revision period for a chosen start ontology that I
>have advocated.

I certainly agree that there will have to be a great deal of
analysis and discussion on a point-by-point basis of essentially
every category and axiom at the top levels and probably at
many of the middle levels as well.

John