[om-list] Core model
Luke Call
lacall at onemodel.org
Mon Oct 9 09:17:16 EDT 2000
Mark Butler wrote:
> Luke,
>
> Do you have any comments on selecting AI-traditional model instead of a
> RDBMS-traditional model for our base data representation yet?
Mark,
Unfortunately my last week hasn't allowed for much time on this. I am
maybe 2/3 finished reading the first article you sent out, and pondering
as I go. I still need to carefully read the other email from you & Tom
since then.
Here are some quick (but disjointed) initial thoughts: I shouldn't have
said specifically RDBMS, though that was one example (and they are a
mature substrate)--an object database may be just as good, don't know
yet until more analysis done at a higher level. The URL/paper you
recommended on various approaches to knowledge representation is very
interesting--a good recommendation for which I thank you. One way to
represent "goals" or an idealized state of something is to have an
idealized object--like an instance of the person class that represents
the user in some desired future state (financially, habits, or with
whatever goals), together possibly with various calculated (suggested by
the system) or anticipated (planned by the user) paths for getting
there. In reading about rule-based things I wonder if objects would also
benefit from internal rules--sort of like database constraints or
triggers, but in a context more like real life and in a form matching
the system's architecture: So if the "ideal" man (let's say, Christ)
would never do something that another man might, the object representing
ideal behavior might have a rule that would cause certain actions to be
impossible.
I'm thinking that even with the currently limited time I have (1.3 hours
5 days a week under best conditions, hardly any last week), I may need
to take some time to learn more about the various AI areas (logic,
rules, nets, statistics etc.) mentioned in the paper you sent, either
before or simultaneously with developing the personal organizer software
mentioned earlier. If simultaneously, the benefit is using the software
sooner and having a framework within which to employ the techniques
learned as I go; a risk is to mis-design. I think the bigger risk is in
doing nothing--waiting for all the answers means getting nowhere--but am
currently unsure how the balance between learning vs. doing will best
play out.
Now that is all self-centered, and I don't want to slow down others, but
I don't feel qualified to say much beyond that at this point. As usual,
you may have useful insights, and this week I expect to get done reading
them and the other stuff you've sent. My sincere regret and apologies
for the slow pace.
On a more practical note, after reading your email better, I hope to
discuss more intelligently the limits vs. advantages (vs. possibility of
combining) our two preferred approaches. I don't see how to represent
the entire world and its behaviors with a logical system, but maybe to
use a logical system inside such a representation. But I will read &
reply more. Again, sorry for the lame response.
Luke
More information about the om-list
mailing list