Saturday, October 14, 2006

Perceiving Patterns

I'm continually amazed by the human mind's way of finding meaning in randomness.  I've been trying to learn Japanese characters lately, using the Heisig method, where you make up little stories to remember what pieces go together to make each character.  It's a little weird how well this works.  For example the character for "hemp" (麻) is made up of parts that mean grove (林) and cave (广).  Makes sense, doesn't it, that you'd have to grow pot in a cave, to hide it from the police.  Others aren't so easy: the character for "feminine" (雌) is made of parts meaning footprint (止), spoon (匕), and turkey (隹).  So I've got this silly picture in my head now of a very feminine turkey who holds her toes in when she walks to make her footprints look like spoons, thinking that shape is more feminine than normal turkey-foot-shaped footprints.

So, I'm stuffing my head with hundreds of these stories, and it's working well for me so far, but I'm probably stuck with that strange association for the rest of my life.  The thing is, as the flashcard for "feminine" comes around 3, 4, 5 times, the association comes quicker and quicker, and starts to make a weird kind of sense. 

I've noticed this playing around with tarot cards as well.  Each card has a whole bunch of vague meanings, and the card's position in a layout has a vague significance of its own.  So it's no surprise that a layout of cards can be interpreted as meaning something.  What's is a surprise, is that when I lay the cards out and start trying to make a consistent message out of them, a message often tends to appear quite strongly.  I get the distinct sense that I've "found" what the cards are "trying" to tell me.  I don't believe there's anything supernatural going on -- it's a demonstration of how our minds search for meaning, and goose us with a satisfying "aha" when a best-fit interpretation is settled on.

One more example: a program called ELIZA was written in the 1960's that you could type at and it would respond in a crude imitation of a psychotherapist, by feeding you leading questions peppered with words taken from your own responses.  Remarkably, it sometimes could fool people for a short time into thinking there was a real person behind it.

It's my belief that when we attribute intelligence or consciousness to another being, this same sort of mechanism is at work filling in the gaps.  If we ever do succeed in building software systems that people consider to be artificially intelligent, it may be partly because we've come to understand better how this creative perception mechanism works, and how to spoof it for longer and longer periods.

No comments: