The latest development in the world of artificial intelligence comes from an unusual source: Unreal Tournament 2004. Five years ago, 2K Games sponsored a yearly competition challenging programmers to develop multiplayer bots capable of passing a Turing-style test modeled on a good old deathmatch. This year’s competition was the first in which bots were judged more human than actual human players. In fact, two of the bots met that standard.
Romanian doctoral student Mihai Polceanu created one of the bots that achieved a "humanness rating" of 52%. The other came from the University of Texas at Austin’s UT^2 team, which consisted of Professor Risto Miikkulainen and doctoral students Jacob Schrum and Igor Karpov. The best human score was a measly 40%.
Physorg has more information on the UT^2 bot, which mimics even irrational behaviors like recklessly pursuing an opponent just because it has a grudge against them. The team also used neural networks derived from an evolutionary process:
Some behavior is modeled directly on previously observed human behavior, while the central battle behaviors are developed through a process called neuroevolution, which runs artificially intelligent neural networks through a survival-of-the-fittest gauntlet that is modeled on the biological process of evolution.
Networks that thrive in a given environment are kept, and the less fit are thrown away. The holes in the population are filled by copies of the fit ones and by their "offspring," which are created by randomly modifying (mutating) the survivors. The simulation is run for as many generations as are necessary for networks to emerge that have evolved the desired behavior.
The UT^2 bot is available for download if you have a copy of Unreal Tournament 2004 kicking around and feel some pwnage coming on. I wonder how it would fare at one of Scott’s LAN parties.