Computer Science: Faster, Smarter, More Robust

My wife and I like to play video games together cooperatively. Generally, she plays the exploratory parts, and I play the parts the require quick reflexes.

Recently, we've been enjoying a game called iNinja for the GameCube. (Thanks Ben Bangert!) As we neared the end of the game, we reached a level called "Egg Shell Skull" that seemed impossibly hard. We each spent a couple hours trying to beat it, but to no avail. It required 100% accuracy, very fast reflexes, and a little bit of multitasking. I could "program myself" to have fast reflexes with high accuracy, but every time I mixed in the multitasking, my reflexes went out the window.

Finally, I had the idea of playing the level together. I used the left side of the controller which involved ducking with 100% accuracy and very fast reflexes, while she focused on the right side of the controller which involved planning, combos, and monitoring a timer. After several tries, we finally beat the level :)

Since this is a purely technical blog, you might have guessed that I'm not trying to focus on how great a gamer I am ;) I'm trying to point out something. In a real-time system, you don't always have enough clock cycles for the best algorithm possible. That's doubly true for "coding in wetware". I was able to program my brain to have high accuracy and low latency, but only if the task was very, very simple. There's just not much decision making you can do in low latency situations. This is backed up by what little I know about the amygdala and the neo-cortex.

There's another case of this pattern that I find enlightening. I saw a talk from an air traffic controller, and he was explaining that there are three systems that are used simultaneously to help planes avoid hitting each other.

There is a very stupid and robust system that kicks in when two planes are about to collide. This system is in charge of implementing the correct "duck!" algorithm. The next system is in charge of avoiding a "loss of separation" within the next 5-7 minutes or so. It's more complicated, but doesn't have to worry about wind, etc. The highest level system is in charge of avoiding a loss of separation in the next 12 minutes. It is extremely complicated, and has to account for things like wind, etc. These separate systems have separate code bases, naturally.

In the future, I wonder how often we'll see cases like this. Sometimes, neither the simple, fast, and stupid solution nor the elegant but complex solution will do the trick. Sometimes you need both--just like the human brain needs both.


Anonymous said…
Very nice entry. Interesting and an inspiration for a current little programming project of mine.
Cheers, Jan
writeson said…
Really like this entry, though it completely reinforces my belief that driving and talking on a cell phone are in a mutually exclusive domain. I've also read articles about the programs on the space shuttle being kind of like this; complicated, but short and only responsible for this range of five minutes before another program takes over.
Anonymous said…
It brings about a zen like realization...somethings like these are the ones that you think about, know about, yet when you read it in words, it does bring out something.
jjinux said…
Glad you liked it guys :)