Artificial Intelligence Gamers Win Prize For ‘Human-Like’ Gameplay
Lee Rannals for redOrbit.com — Your Universe Online
[ Watch the Video ]
“The idea is to evaluate how we can make game bots, which are nonplayer characters (NPCs) controlled by AI algorithms, appear as human as possible,” Risto Miikkulainen, professor of computer science in the University of Texas at Austin College of Natural Sciences, said in a statement.
Miikkulainen created one of the two winning bots, UT^2 game bot, along with doctoral students Jacob Schrum and Igor Karpov.
During the competition, the bots face off in a tournament against one another and about an equal number of human players. Each player in the tournament tries to score points by eliminating its opponents.
The players also have a “judging gun” in addition to its usual complement of weapons, which is used to tag opponents that are either human or bots.
The bot that is scored to be the most human-like by the human judges is named the winner. UT^2 and MirrorBot, created by Romanian computer scientist Mihai Polceanu, were the victors.
The winning bots achieved a humanness rating of 52%, while human players received an average humanness rating of only 40%. The two winning teams will be splitting the $7,000 prize.
“When this ‘Turing test for game bots’ competition was started, the goal was 50% humanness,” Miikkulainen said in the statement. “It took us five years to get there, but that level was finally reached last week, and it’s not a fluke.”
The complex gameplay and 3D environments of the game require that bots mimic humans in ways including: moving around in 3D space, engaging in chaotic combat against multiple opponents, and reasoning about the best strategy at any given point in the game.
“People tend to tenaciously pursue specific opponents without regard for optimality,” Schrum said in the statement. “When humans have a grudge, they’ll chase after an enemy even when it’s not in their interests. We can mimic that behavior.”
In order to mimic as much of the range of human behavior as possible, the team takes a two-pronged approach. Some of the behavior is modeled directly on previously observed human behavior, while the central battle behaviors are developed through processes called neuroevolution.
This process 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 with copies of the fit ones and by their “offspring,” which are created by randomly modifying the survivors.
“In the case of the BotPrize,” Schrum said in the statement, “a great deal of the challenge is in defining what ‘human-like’ is, and then setting constraints upon the neural networks so that they evolve toward that behavior.”
He said if they just set the goal to eliminating an enemy, a bot will evolve toward having perfect aim, which is not human-like. Instead, they impose constraints on the bot’s aim, such as rapid movements and long distances to decrease accuracy.
“By evolving for good performance under such behavioral constraints, the bot’s skill is optimized within human limitations, resulting in behavior that is good but still human-like,” Schrum said.
Miikkulainen said methods developed for the prize should eventually be useful in developing both games that are more entertaining, but also in creating virtual training environments that are more realistic.