Rupert Goodwins' Diary

Everyone has their weaknesses. Some gamble, some snort cocaine, some collect pickled onions. Among my many foibles -- none of the above, except the onions -- I spend far too long online reading about evolutionary biology and arguing with creationists. It's never terribly satisfying -- the creationists can just say: "Of course, the Grand Canyon was carved by the waters of the Flood" and then you have to go and dig up geological columns, dating methods, hydrodynamics and so on. Which gets roundly ignored in favour of: "But the amount of helium in the atmosphere proves the Earth's only six thousand years old," and so on, and so forth. Still, you learn a lot of biology, geology, physics and theology, and one must stand firm in the face of the counter-Enlightenment.

You can also get your code debugged for free. Take Avida, an artificial life program chuntering away in the non-throbbing non-metropolis of East Lansing, Michigan. This takes little lumps of self-reproducing code, randomly mutates them and then rewards those that can do a particular task better than the others -- in other words, it's modelling biological evolution. The good thing is, while the real thing takes place over millennia and is only partially self-documenting, Avida can churn through and log thousands of generations in less time than it takes to read a page of The Origin Of Species.

Pleasingly, the software is producing some truly bizarre results -- and the biologists are getting interested. The simplistic idea of evolution is that good mutations thrive while bad ones fail: however, in Avida bad ones can turn into even better ones down the line. Also, some of the successful artificial organisms are so complex that there's no way of working out how they got that way - until you look at the logs.

Now, creationists have a special place in their hearts for evolution -- although most branches of science contradict their ideas, this takes pride of place. And one of their trump cards is complexity: some things about living organisms are so complex, they say, that they must have been designed as complete systems by an intelligence. "What good is half an eye?" they ask, No matter that they can't begin to explain how to show this or actually point to anything that clearly lies outside the bounds of evolutionary possibility, least of all eyes.

They really, really don't like Avida. So much so, according to an article by the unrelentingly excellent Carl Zimmer, that thousands of them emailed the project after the first results came out to say: "You must be wrong. It can't be doing that." "Fair enough," said the researchers. "Here's the source code. Tell us where it's broken."

Two years later, Avida's going better than ever -- thanks in part to the incensed creationist programmers (engineers and computer bods being over-represented in their ranks) finding a handful of small bugs. There's no better way to fix code than to have it picked over by people highly motivated to find fault, as the battling egos of open source know only too well. Now, if only the religious wars of the operating systems could be turned to such mutual benefit…