Sunday, December 4, 2011

Wouldn't you prefer a nice game of chess?

All games involve solving problems. Sometimes it's just you and some cards, trying to figure out how to get all the suits lined up in order as you play Solitaire.

Sometimes it's you and another person, matching wits as you try to create obstacles for their pieces while simultaneously marching yours into your home base. That's backgammon.

Yet other times, you're talking about dozens of people, all of whom can be utilized in different ways to get an advantage in either advancing or stopping a ball from moving down the field. That's football.

But these days, mostly thanks to the Internet, there are games that involve thousands of people trying to solve a problem. Sometimes it is something as straightforward as an online role-playing or combat games. Other times it is a so-called "alternate reality game," where participants interact in the real world to reach an objective, often the solution to a mystery. Think of it as a Host A Murder dinner party with a million guests.

Those types of challenges can be fun. But they can also be productive. Much as projects like SETI@home use thousands or millions of computers to analyze data, having multitudes of people analyzing a problem can lead to a quick solution.

For instance, how long do you think it would take one person to put this paper-based Humpty Dumpty back together again?

Word spaghetti, served al dente.

DARPA--an increasingly often-cited agency on Read Ink--took the picture above to the masses, offering a US$50,000 prize for whoever could accurately piece back together five shredded documents in roughly one month.

They got a winner, too (solutions here). Some of the 9,000-ish entrants used brute force, simply having a whole bunch of people sifting through strips of paper. But the victorious team created an algorithm that essentially compared all the strips to one another and generated a "probability of fit" score. That meant the human operator could ignore pieces with a low score and examine only the ones that were likely to go next to each other.

And it still took a month to solve.

But besides being an interesting challenge, it is also an effective way to solve a bigger problem. Need a method? Consult the masses. DARPA essentially paid US$50,000 to solve a meta-problem that probably would have taken millions of dollars to examine with a smaller team of dedicated researchers.

And because this is a writing blog, I should of course note the impact this will have on the world's thriller authors. Need your spy to destroy some documents? Better have her burn them... otherwise, no secret is safe.

No comments: