Computer Science Majors Develop Artificial Intelligence for Video Games
Lauren Gillespie, Class of 2019, is a computer science and chemistry double major working alongside Assistant Professor of Math and Computer Science Jacob Schrum to create intelligent agents for video games.
“I thought it was really interesting that Professor Schrum’s ideas encompassed a lot of what I learned in biology,” says Gillespie.
Gillespie is one of few sophomores chosen to participate in SCOPE, and brings an interdisciplinary mindset to the research.
“That’s what made her stand out to me—that she has a broader interest in science in general, and that interdisciplinary thinking is really important in mixing biology and computer science,” says Schrum. “So I have a student who isn’t just good at programming, but also appreciates the bigger ideas behind this research.”
Schrum and Gillespie, along with a team of students, are testing whether a computer can process information using strictly visual information, which forces intelligent agents to see the game in the same way that humans do. Video games are a popular testbed for Artificial Intelligence (AI) techniques because they are simulated, controlled environments, but have a level of complexity that makes optimal decision making by in-game entities difficult.
“What we want to do is take visual patterns on a screen and have an agent playing a game learning to use them,” says Schrum. “For example, what we’re working on right now is getting the computer to evolve Tetris players, but using just the raw game screen.”
This research entails implementing cutting-edge Artificial Intelligence techniques like Evolutionary Computation. However, in addition to evolving intelligent agents, their research also involves creating art with programs such as Picbreeder.
“When you’re doing that procedure, the natural outcome is creativity and innovation,” says Schrum. “It encourages you to use your imagination, and harnesses the creative power of evolution.”
Schrum and Gillespie are the innovative forces behind the evolutionary system by acting as a filter through choosing designs that appeal to them—a subjective process.
“The idea is that evolution isn’t just moving in a specific direction for a specific pre-defined goal,” says Schrum. “It starts where it is, works with what it has, and it branches out in all possible directions, and what works survives—and what doesn’t work, dies off.”
The research will result in a peer-reviewed conference publication, with the potential for further expansion into a journal article.
“I have hopes that this research will reach an international audience,” says Schrum.