When grid is effective: narrow pipes, huge skulls

  • Article

It's like Jim says: "The ideal mobile task is stateless (needs no database or database access), has a tiny network input and output, and has huge computational demand". That's why SETI@home was such a good idea. However, I believe that the new NASA stunt is even more interesting. Check it out here, but in short the story goes as follow:

  • the Stardust project produced a HUGE amount of data
  • the information they are searching for is not easy to spot by automatic means

So they came out with a crossing between SETI@home and Amazon's Mechanical Turk (another very smard idea, if you ask me): in the coming months NASA will make available an app, with which volounteers will be able to easily spot the interesting samples just by watching small 40-frames movies. While the data may not be actually tiny, they will be certainly comparable to all the animated banners & rich content we are now used to deal with; and I have an hard time imagining something more computationally demanding than something that can't be crunched without a between-the-ears CPU...

NASA rightfully declares that the competence of each user is non-trivial to assess, so they will resort on some sort of voting system: every sample will be sent out to 4 users, and it will be considered only if 2 of them will agree. It's interesting to note the underlying assumption of good faith of the users, which basically makes something like those systems possible at all. While it's true that nobody would have an interest in clicking everything at random, I've seen too many ruined monuments & broken public phones to think that we are just PLB (perfectly logical beings) trying to figure out the color of the gem on our forehead... I still think that the vast, vaaast majority of users will make the right thing (why bother to register at all, otherwise?). However I also think that it's too easy to forget how this trust dramaticly simplify an otherwise very hard problem: when collaborative efforts are used for revenue generation activities, this kind of trust can be a luxury. And the effort for securing effectively the whole thing can sometimes quickly erode the other advantages that lured you in the adventure...