Monday, February 14, 2011

The spiral game: the miracle of human efficiency

What immediately fascinated me about the “shell game” was how well it worked: one shell from a box of nearly identical other shells could be quickly identified and distinguished from the others. One of the impossibilities of this feat stems from the fact that the distinctions were made using questions, not characteristics.

One might argue that a question like “why is the shell so blue” isn’t a question at all, but a statement implying: “the shell is blue.” Still, reliable identifications were made from questions much more confusing, like “does someone miss you?” This yes/no question could have implied that no one missed the shell, or that someone did. The efficacy behind this question is that the answer actually doesn’t matter—either way, there was a sea creature involved, and that creature may or may not have missed its shell, so the observational implications were that the shell could have been home to a sea creature.

As I later discussed with the Doc, what’s even cooler is the lightning-quick judgement that the human brain exercises while narrowing down the selections after hearing questions. I immediately imagined a computer trying to do this selection: with such confusing questions, and a computer’s inability to find distinctions in the shells, it would be impossible. It reveals a fundamental aspect of what separates us from computers.

A question like “why does the shell look so spiral-y,” to a computer, wouldn’t be very helpful. How does a computer know what “spiral-y” means? The spiral on a shell’s surface isn’t actually a spiral (despite how geometrically precise nature can be), and a computer wouldn’t have enough prior reasoning experience to determine that a shell would generally resemble a spiral. The technical limitations would begin at imaging and end in the computer’s logic—how, even, would a programmer begin to teach a computer how to continually learn, amass useful knowledge, and be able to apply conceptual understanding of a spiral to an identification game? If it’s hard enough for humans to do it, it would have to be near impossible for a computer to (considering we program them).

No comments:

Post a Comment