Skip to content

DARPA Thursdays: Looking for bugs and through your wall

December 1, 2011

DARPA is looking for a way to detect code defects via crowdsourcing. As it is right now, trolling through code to find problems is not an easy task and generally requires someone with decent coding skills and even maybe very specialized skills to find bugs in software. They’re hoping to find some mechanism of turning a search for these issues into a game that an average person could play. I think the idea is generally interesting but it’s difficult for me to imagine how some tool would be made more effective by having a non-technical person on the other side playing some kind of game or looking for particular solutions. It doesn’t seem like it’d be any more effective than just building a better and more automated verification tool to begin with. But then I’m not a software engineer so maybe I’m missing the angle.

In other police state news, DARPA is looking for better machines to watch people through walls. There are current tools available, but DARPA would like to see better range and better ability to differentiate between individuals if they are in a crowd. Biometrics are sure to play a role in this. (Bug photo via Creative Commons from Anne Bennett)

Advertisements
3 Comments leave one →
  1. Dave Vandenbout permalink
    December 1, 2011 8:57 pm

    There has already been similar work using gamers to solve protein folding: http://arstechnica.com/science/news/2010/08/gamers-beat-algorithms-for-finding-protein-structures.ars

    Most gamers aren’t molecular biologists, but given a few simple rules and using their innate pattern recognition skills they were able to make improvements to protein structures that the AI program could not. If they can transform computer code into a similar visual form (perhaps graphs and tree structures?), then DARPA might succeed with this.

  2. frautech permalink*
    December 1, 2011 9:20 pm

    Dave- Thanks for reminding me about that program, I had forgotten. Still, I guess I have a hard time visualizing how an AI program looking for problems would be very effective. The protein folding makes a little more sense to me, I can see how with a little bit of user feedback a program can do most of the heavy lifting and go in the right direction. But it seems like if your program is already clever enough to know what to look for, the user input almost becomes superfluous. But again, I’m not a programmer, so maybe I’m just not seeing it. Should be pretty interesting to see what kind of solution they arrive it and what form it takes.

  3. December 27, 2011 7:34 am

    I’d tend to agree. Bugs in programs are a different type of problem than protein folding. I can see it for protein folding or other things, but programming is strictly logic and I’d think computers or experience programmers would be a better way to deal with it than crowd-sourcing.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: