Thu 3 Dec 2009
Well, the day of reckoning is almost upon us – Dec 5th, DARPA Network Challenge Day! I've previously blogged about this here, and make sure to take a look at the comments because there's some good stuff there.
Anyway, I was looking again at the rules recently, and noticed some things I didn't notice before. The prize is to the 1st team that submits the most correct balloon locations. The balloons will only be aloft during the daylight hours of Dec 5th, unless there are weather difficulties, in which case the balloons will go up on the 6th or later. It's unclear whether they'll delay all the balloons, or whether it means that depending on the weather there could be balloons up on different days in different parts of the country (more bad news for teams that are thinking of driving around looking!!!). But teams have until Dec. 14th to submit winning entries. And they've revised the accuracy of the location to be within 1 mile (huge!). To me, this suggests that DARPA thinks the challenge will be won algorithmically. I think they might be right.
On Dec. 5th, I'm guessing we'll hear about the locations of 3-6 of the balloons. There may be a team here or there that has some private information, but I suspect most of the locations will be known by all teams who are paying attention. (Of course, only one of them will be first to submit…) But then the real fun begins. Starting at nightfall, the balloons are gone, but we can still find them. How? Well, I don't know. How much media is uploaded to the cloud each day? How many pictures, videos, that might show a balloon in the background, even if the photographer never noticed it? Of course, that media would have to be geotagged. It's possible to infer the location of balloons after the fact with some accuracy, especially now that we only need to be right within 1 mile. After all, teams will have more than a week to look, and then even to go out to these locations and pace off distances if they choose.
So, what do you think? Is this how the challenge will be solved?