A black contraption comes into my sight, I quickly calculate the size of it, then recalculate a split second later to find that the image has increased in size. Then I add the facts that some round black and moving things are placed in all corners, round glass things in the middle, noise arising - volume increasing.
First conclusion: I have headlights, wheels, motor running in front of me, ha, a car!
Second conclusion: Volume increasing, must be approaching, and that fast. Oops.
Final conclusion: Get the hell out of here!
Well, probably not. With the feeble calculating MIPS of my brain-cells I would've been roadkill.
My brain is far better at recognising patterns, immensely so. A few clues, probably merely the increasing sound volume of an engine would tip me off and make me jump even "before thinking". And good is that.
"Umbrella, children, film" - enough clues to give most an image of the film "Mary Poppins". Hardly what a programmer would have used as logic for a film recognition application. Absolutely not the analytical computational method.
Recognising an emerging pattern representing an object from a few clues...
A movement, a smile, a few words from an otherwise unknown person and you choose him to be your partner in the canyon transversing exercise on that company training camp. And usually it works out fine.
Recognition of a known pattern featured by such sought after objects...
At work, in the usual structured setting of structured organisational information structures we use the tree structures of manuals, catalogues and hierarchies when we are to find, choose and use a resource. Something that requires real training and knowledge. And a manual.
Otherwise you'll get lost at the second branch choice and end up with a goat as canyon-transversing-partner.
A tree structure is information structured at source, requiring a structured, logical, one-step-at-a-time method to find, choose and use resources. A path-to-the-object, not a pattern representing an object.
Perfect for computational and paper-based information repositories.
No good for humans.
We.are.the.best.at.pattern.recognition. Let's use that.
What we need is:
- A system where the "characteristics" or "features" of an "object" or "subject" are independent of standards and structure.
- Where all these "features" are presented and can be chosen.
- That any combination choice of features is enabled for "pattern building".
- With any set of features chosen (pattern) all objects corresponding to such "features patterns" shows for a quick and simple pattern recognition by a human user.
[Note: An early practical example of the method can be found here. Multiple tags ("features") choices as "stencils" to find patterns in an unstructured information repository.]
If I were to choose the canyon-transversing-partner from all company employees (say IBM... 234.000 employees) I think I would quickly tick off "nimble", "strong", "no vertigo", "English speaking" - and end up with 13.467 choices. OK, add "available" and I have 578. Being male and we're having fun so why not tick off "female" and... then ending up with 7 that I could zoom in on as in shaking hands with. You get the drift, I would probably end up with a pretty cool partner for the outing :)
Why not use the same method when choosing members to a project group? Instead of the typical set of "departments" and "titles" that we all know does not tell much.
Or for parts, suppliers, advisors... any resource?
["right", "rear", "estate", "light", "red", "audi A6", "2003", "April"] instead of ["XP87-pfGT-0009"]. No training required, quicker, no manual, probably more accurate as "pf" could easily become "pF" while "red" is not easily mistaken as "blue"! And of course more flexible in the case one starts making "white" and "orange" rear lights as well.
Easier, simpler and above all - a better chance of getting the best resource at the right time.
And that, ladies and gentlemen, is what business is all about, best use of resources.
Equal to more value added, margins increased, net profits exploding and general financial bliss.
Increase profits, make life easier, beat your competition, save resources - dump any notion of the need to structure the information at source, rely on patterns when choosing from the unstructured mess instead.
I'm still alive and bouncing after many years of traversing roads, purely relying on precisely that method. QED.
[P.s. Been a bit absent, much going on, a bit too much, do see end of tunnel, etc....]
I was sorry to hear you were robbed, heartened to see you take a posisitve out of a negative.
Back however to the matter at hand. I believe what your attempting to do combined with a method of creating that pattern data, would allow humans to quickly and easily locate that which they seek.
I've been thinking about streamlining this process, creating those patterns so that search engines and humans alike have a friendlier time of it. All using freetags. I believe by doing what I call Typetagging, allows me to create those patterns and paint my own scene for later perusal by an observer, either human or machine.
You can read more of what I call Typetagging here:
http://neuraxon77.wordpress.com/2005/10/26/this-is-an-example-of-typetagging/
It may be nothing new, but the more I think about it the more I like it. And I can't really see a better way at this point in time.
Posted by: craig | October 26, 2005 at 08:57
Craig, appreciate your kind words - yep, no other way to handle setbacks than twist'em into something better :)
Interesting idea you're writing about, kindred to what we're dabbling with - and I will keep my eyes peeled on any progress in the Flock world and otherwhere!
As you can see, I'm a bit on the simplistic side - no search nor tree structures at all - only one tiny little practical issue left - make it useful in practical life :)
Will use the method though in our run-your-business system, found it to open up many interesting possibilities far beyond "finding" stuff... but we'll see, soon I hope...
Posted by: sig | October 27, 2005 at 08:58