Scientists program AI to be more human than humans

Scientists at the University of Texas at Austin have created an artificial intelligence (AI) virtual gamer whose actions convinced a panel of judges that it was more human-like than half the humans it competed against, Phys.Org reports.

The scientist-programmed bots played in a 2K Games sponsored tournament against humans in the first-person shooter, Unreal Tournament 2004. Each of the human players were given a judging gun where they could tag whether their opponent was a human or a bot.

The player that scored the most tags was be crowned the winner. In this case, it was a bot named UT^2, which was programmed by Risto Miikkulainen, a professor of computer science in the College of Natural Sciences, and doctoral students Jacob Schrum and Igor Karpov.

UT^2 scored a humanness rating of 52 percent while the human players received an average rating of only 40 percent.

Schrum told Phys.Org that the complex gameplay and 3D environment in Unreal Tournament 2004 requires bots to mimic human players in a number of ways, from the way they move, the way they engage in combat, and also the way they perform the seemingly irrational actions of humans.

UT^2 scored a humanness rating of 52 percent while the human players received an average rating of only 40 percent.

"People tend to tenaciously pursue specific opponents without regard for optimality," Schrum said. "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 the case of the BotPrize, 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 behvior," he said.

"If we just set the goal as eliminating one's enemies, a bot will evolve toward having perfect aim, which is not very human-like. So we impose constraints on the bot's aim, such that rapid movements and long distances 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."

Risto Miikkulainen believes that not only can these bots help make certain games more entertaining, they can also help make virtual training environments more realistic and help robots interact with humans in more effective ways.