[<< | Prev | Index | Next | >>]

Tuesday, September 28, 1999

Measuring tiles, and Concept Representation

I spent the whole day working around the house. Measuring slate tile mostly, to finish the master bathroom I won't be here to use. Add to that laundry, grocery shopping, and trimming back the thorny vines that have been consuming my potential guests before they could get to the front door, and there it went-- the whole day. Poof!

I guess the solution is obvious: Eat less and wear fewer clothes. Ok, ok, so perhaps I should start by not remodelling an entire house all by myself. I actually did call around for bids on the tile. The prices were quite reasonable until I said "slate" at which point I found myself bidding for their souls. "Please don't make me an offer I can't refuse" they'd say. But if all I wanted was ceramic tile, I may as well do that myself too, because relatively speaking that's easy--they're all the same thickness so you just slap down some thinset with a notched trowel and essentially glue the puppies to the floor. How could I hire someone to do that?

Besides, I'm poor.

That's why I'm moving to Maui. The nice thing about not having any money is it doesn't matter where you don't spend it. My friend Garrett* is going with me. We mailed his truck last week. He's got a Phd in physics, so he's really poor.


I've been thinking alot about concept representation lately. Serious philosophical, mathematical, logical ponderring. For weeks. Nay, months. Nay, years. And for what? My phone book. Yes -- I need something to keep my phone numbers straight. Especially lately, with this plague of dynamic area codes. A good phone book program ought to handle that, no? And my addresses. And my finances. And my todo list. And all those postit notes that defy categorization. And I have to figure it out, all of it, in the next couple of months, because my current schedule program has a Y2K bug! I know because I wrote it--thirteen years ago, in college. Who would have thought I'd still be using it today? You'd think somebody would have written something better by now, but nothing beats mine for ease of use. For example, say I just made plans to have dinner with Sally next thursday at 7pm. I just type (at my Linux shell prompt):

	se Dinner with Sally next thurs at 7pm
and that's it. Presto, the program picks out the date and time, assumes the rest is the description, and when the time approaches it sends me email to remind me. Or I just type "sc" (schedule check) at any time and it tells me what's coming up.

But when I wrote it, I didn't know all the rules for leap years and things, so it's probably going to get confused at the turn of the century, and all of my appointments will be off by a day.

And that just won't do.

So I want a way to represent knowledge, to stuff it into my computer in some useful and general manner. Sure, the entire field of A.I. has been concerned with this for decades, to no apparent avail, but heck, if I let other people's failures disuade me, I'd, just for instance, never have put on a parachute and tied myself to that truck. (Aside from the stitches in my forehead, it was a blast!)

So, consider the problem: What is Sally's phone number? How do you, a human presumably, answer that question? What if her area code just changed recently? What if you happen to know she's on vacation right now, or working at a different office this week? A simple predicate just doesn't cut it:

(PhoneNumberOf Sally 555-1234)
Because in truth, there are potentially many phone numbers for Sally, each correct in a different context.

So how do you encode that fact? If-Then's? (Prolog's Horn clauses are essentially If-Then's.) Explicit context variables in every expression? (E.g., "(PhoneNumberOf Sally 555-1234 Context_42)".)

The deceptive thing about introspection is that we freely intermingle symbolic reasoning with subconscious inference and association, and the latter by definition can never be fully observed so we are left always inadvertently taking some things for granted. The things which we observe as the "obvious" trail of our thoughts may be no more than the final conclusions subconsciously drawn at the end of each stage of processing -- the tip of the iceberg, with a whole world of connections and processing that we're just not aware of.

We're not aware of it because it is us. But that's another topic for another day.

[<< | Prev | Index | Next | >>]

Simon Funk / simonfunk@gmail.com