Re: SUO: *Date 06 Apr 2002 -- Formalism Engineering
On 4/5/02 23:00, "Jon Awbrey" <jawbrey@oakland.edu> wrote:
> That's SCIENCE, not ENGINEERING. Engineers just build. Look at the Segway.
>
> Segue?
No, the Segway - the little gyrostabilized electric two-wheeler that you can
drive around on sidewalks.
> You can't possibly mean the things you just said here.
>
> JA: But how do we find a place to stand in regard to logic?
> Where is the examen of the balance to weigh our reason?
> Where is the fulcrum of the lever to shift for thought?
>
> BA: For the purposes of building an ontology system (or systems),
> there is only one necessary condition: that the logic be able
> to do any computation we can do on our current von Neumann-style
> digital computers. Anything less, and we lose the ability to
> choose the point in the expressivness-efficiency space in
> which we'd like a given ontology to operate.
Well, perhaps there should be more necessary conditions. But I have named
one at least and I encourage others to name more - perhaps this is where we
all should have started out on this list.
> The question was: How to design a better theorem-trap,
> one that is somewhat higher up in the efficiency space?
Jon, I fear you may have some important point here - please speak plainly.
Are you saying that we should favor ATP approaches that are by some standard
more efficient (and less complete) than unrestricted resolution?
> I am talking about the practical conditions for practical success in applying
> logic.
So am I.
> That just means: It's hard to improve something that you think is perfect.
> But any person who thinks that FOL is GOL, in any of their secular avatars,
> is a finitely informed critter with a serious awareness-of-fallibility gap.
Don't jump to conclusions. If you were actively working with people in the
field who are IMPLEMENTING ontology software rather than just talking about
it, you would not find the love affair with pure FOL that you imagine
exists. However, I think John Sowa (and others) have the right idea - that
FOL has pretty much all the power you need to say what you want. Using that
as a basis, you can then choose how you want to make computational
compromises. In my view, the problem with languages like DAML/OIL is that
the computational compromises are built in -- fine for an implementation
language to be used for a special application, but completely undesirable as
an ontology authoring language.
It would make my life simpler (I don't have time to read a screed) if you
would just say what features you think a logic for ontology ought to have.
If you take this on, please try and use standard mathematical vocabulary. I
for one would like to see what you have in mind.
-bill