200 Words: Evolutionary Algorithms

It seemed like a simple enough study - find a more efficient wing surface, to increase the maneuverability and lift of regular single-man ultralight craft. And we set about it in the normal way, a controlled orbital evolutionary environment with some avian seed organisms and some artificial physical properties. Crank the ambient radiation up to increase the rate of mutation, put some unusual minerals in their food chain to allow exotic physical chemistry. Give them some contrived 'prey' species, force the avians to hunt them in very enclosed spaces, and leave the environment to run for a few thousand generations. We expected they might find some new ways to fly, perhaps reshape their wings or take advantage of ultralight chemistry in their bones.

We didn't expect their gripping limbs to develop, or that the minerals would migrate into their nervous system and dramatically increase their effective intelligence. We didn't expect one of them to try to break out of the environment. We didn't expect another to succeed.

We're going to try to shut the environment down tonight. I hope we succeed this time. Our life support controls run awfully close to the observation dome.