DARPA Bums: Restoring Truthiness Tackles Top Secret Government Code!

DARPA Bums - Restoring Truthiness Tackles Top Secret Government CodeWe were robbed and we lost.  This is going to sound like sour grapes; it’s not but I’m pissed off and somebody’s got to pay for screwing up.

Restoring Truthiness has its finger in many pies, and you don’t want to know what the interns get up to with their pies.  Seriously, you guys are why our insurance premiums are as high as they are.

We can’t often tell you what pie we’re in because much of this is done in the interest of national security, but it’s safe to say that the night has a thousand pies and you can rest easy.  Unless you’re afraid of pies.  (Look under your desk or bed right now.  If you saw a pie sitting there, tell me it wouldn’t be scary.  Pies man, they freak me out.)

The Defense Advanced Research Projects Agency (DARPA), the same vast, cool and unsympathetic intellect that brought you the internet and free porn, recently tasked the RT R&D department with cracking a super-hard code.  Five documents of increasing complexity were shredded and the government agency asked us to find the fasted, most efficient way of reconstructing the originals.  To make things fair, they opened the contest to the general public and offered the winner a $50,000 prize.  Certain we would easily crack the problem, we opened a line of credit with Bank of America which we spent on hookers and blow a week-long team building retreat.  To that august institution we have no comment other than to turn out our pockets and say ‘bite us.’   Because like I said, we lost.

Ostensibly, the agency wanted to document how people went about solving the puzzles; the contest ending date and 50 grand payout occurring during the government’s end-of-year spending flurry being simply incidental to scientific interests.  More accurately, it doubtless provided the astute observer with a more revelatory schematic for the structure of the internet since it was just a couple of dozen computers and message board for dumb blond jokes.  Graduated entry points, lines of communications, clusters, orphans, white hats and black hats.  Really, somewhere right now, a government uberwonk is rubbing one out over the results.

So how did Restoring Truthiness represent you, our readers and our torch bearers?  We joined up with the previous winners, natch.  In 2009, the University of California San Diego (UCSD) team solved the DARPA challenge in just nine hours.  Dr. Manuel Cebrian, a member of that team and the 2011 team leader, decided to use crowdsourcing, or rather a distributed work model focusing the efforts of multiple participants over a wide geographic area, which is how USCD won in 2009.   In other words, the plan was to get 10s of thousands of people to piece together as much of the documents as they could, have the results analyzed and refined by a computer, then returned to the players to continue their efforts.

And it worked fine until three things happened: 1) the 4th puzzle was sabotaged, perhaps by ANONYMOUS; 2) 4,000 people joined RT in trying to solve the puzzle, instead of the 100,000 Dr. Cebrian thought would apply; and 3) the puzzle reached a complexity were even the efforts of 4,000 members of the UCSD team were slowed down, often hampered by the simultaneous work of other team members on particular areas.  It is RT’s determination that three factors contributed to this general fuckery and the attendant evaporation of interest: 1) No coed nudity.  C’mon, it’s UCSD, a black hole of hotness.  Would it have killed the leaders to occasionally show a clip of a buxom coed flashing her boobs to show her support?  We think not; 2) Access to the puzzle was restricted based on past levels of participation and; 3) it was easier to get a DMV agent on the line than to communicate with UCSD members.

Participants who’s score was greater than ours (0, because we came in late) were able to claim dibs on multiple chads and preclude anyone else from using them, and they hogged the pieces.  A ‘ghosting’ feature was added, where users could move a copy of a dibbed chad around to see where it might fit, but if you weren’t the first to touch the chad you couldn’t do anything with it.  Even if it was a perfect fit, the ghosted chad still flew back to its true location as soon as you released it.  The chat feature was unused.  And once, when RT had stumbled across a breakthrough and was busy piecing correct orphaned chads around a chad cluster that someone else had dibbed all over, the dibber scattered the core rather than accept the help he or she so desperately needed.  At no time did the RT R&D department witness any feedback from the UCSD team leaders.  And that was the fatal flaw in the whole tragic process- crowdsourcing depends on establishing and maintaining a community.  Playing on the UCSD team was as rewarding as playing bingo with a lot of really old, cranky people; the kind who have lucky seat cushions and a permit to carry.  No community, no progress.

So we lost.  Lessons learned.  We wish the UCSD team well in their future endeavors, and congratulate the winners, ‘All Your Shreds Are Belong to U.S.’  Three people and an algorithm.  Lessons learned, indeed.