How the awesome computational power of video game cards has transformed the study of evolution.
Question: Could your crowd simulations of smaller organisms facilitate similar studies of humans?
Iain Couzin: Yes, we actually simulate across the spectrum from cells up to schooling fish and to human crowds. And one of the challenges is and has always been the computational power that you require when you're actually simulating these individuals. Individuals have to look at each other and see, you know, am I within range to interact with you, and so on. And as we start increasing the number of individuals, the computers tend to sort of chug to a stop and it's very difficult to work.
But there's a breakthrough. In the last couple of years, there's been programmable video game cards. So these cards that have been developed for, you know, gamers, so they can, you know, live in these virtual environments and so on, actually have hundreds of processing cores on them. And this has been an absolute revolution in terms of scientific computing for us. So we're investing heavily in our efforts to try and program all of our simulations on these video game cards. And to give you sort of a rough impression, we're getting around 300 or more times faster. You know, and if computing gets twice as fast or three times as fast, then that's wonderful. But if you can get 300 or 500 times as fast, if what used to take a month now takes you an afternoon, that changes the way we work.
And also, because we can harness this vast computational power, we can start asking questions about evolution, we can start simulating these groups of reasonable size with the reasonable resolution in how they interact in space over such long time scales that we can now start, you know, having a sort of virtual process of evolution to understand how and why collective behavior has evolved.
Recorded on December 15, 2009
Interviewed by Austin Allen