Re: Storing data and code in a Db with LISP-like interface
Date: 30 Mar 2006 08:37:43 -0800
paul c wrote:
> I agree with Marshall, without agreeing that he's 'right'! More that
> his statement is apropos. Stephen Hawking mentioned how Bertrand
> Russell couldn't argue against the lady in the audience who claimed the
> earth was the shell of a giant tortoise or the Lisp man, John McCarthy I
> think, who on the topic of the natural capabilities of computers said
> that trying to make them think was like trying to make a submarine swim.
Nice quotes. About 20 years ago the concept of 'Situatedness' in artificial intelligence revolutionised (saved?) the field from the expert system/chess playing dirge it was ploughing through at the time. (see "Elephants don't play Chess" by Brookes for an influential text). Intelligence does not exist outside the environment the actors and agents it concerns survive in. Context is just too important, and far too complex.
As far as database theory is concerned, I see grave parallels with a lot of today's knowledge engineering movements. Semantic networks, graphs containing 'meaning', brittle XML global ontologies. They just do not and cannot work. Its like some horrific 70's AI flashback.
RDBMS on the other hand take the pragmatic view of being a tool for the user to extract 'meaning' from. Its up to us to decide what the relationship in a table 'means' - in fact this is the only way of doing it, as we are the only ones who can apply appropriate contextual information. That's why RDBMS, and systems with the same outlook, work in the field, while things like CYC and the Berners-Lee's Semantic Web do not.
To the OP - add in the fact that not all associations are binary (irreducible tuples anyone) and you can throw away your graph as a logical model (use it as a physical construct by all means).
I clung to the seeming purity of binary relations for a long while before conceding I was wrong.
Don't waste the time I did. Received on Thu Mar 30 2006 - 18:37:43 CEST