Lists Home |
Date Index |
Bullard, Claude L (Len) wrote,
> Can you tell me why decidability was given that level of importance?
> This is what I meant by "over-engineered" when I could have said
No idea. Decidable systems are guaranteed to give answers in finite
time. That sounds great until you remember that "finite" extends well
beyond the expected lifespan of the universe.
I think it's just a cool formal property rather than having any real
practical import. It's also a hot research topic at the formal end of
academic AI, so there are papers to be written and research projects to
be done. I guess that's attractive to some.
But it's unnecessarily constraining: for all that inferences in an
undecidable system aren't guaranteed to terminate, in many practical
cases they will. After all, when was the last time you worried about
the halting problem wrt a business system?
> There is a paper from TimBL on the issue of FOL and the SemWeb at
> which I have trouble following not being well-trained or just not able
> to determine his conclusions. Dan Connoly comments on it elsewhere.
As far as I can make it out, TimBL and I are more or less in agreement
here. I've no idea if that represents his current position.
> At first, I thought it the case that the notion was to not require FOL
> so that other logic systems could be used as needed, or as in the
> spirit of HTML, to make initial fielding easy and get some mind share.
> Your arguments make me unsure because requiring decidability would
> seemingly make it harder and cut the SemWeb away from uptake of where
> the majority of the world's business semantics reside: relational
> databases with client or server based business rules and logic in the
> form of code libraries. This would appear to limit the role of the
> semweb to being merely metadata about web pages, and that isn't that
> useful. So that can't be right.
Can't it? ;-)