The team assigned specific tasks such as finding a specific pellet or avoiding ghosts to more than 150 agents, which worked together toward the ultimate goal of playing Ms. Pac-Man perfectly.
A "top agent" then collected input from the other agents to determine when and how to move the Ms. Pac-Man character through the on-screen maze.
"The top agent took into account how many agents advocated for going in a certain direction, but it also looked at the intensity with which they wanted to make that move," Microsoft said. "For example, if 100 agents wanted to go right because that was the best path to their pellet, but three wanted to go left because there was a deadly ghost to the right, it would give more weight to the ones who had noticed the ghost and go left."
While simple in concept, games like Ms. Pac-Man are difficult for A.I. to master due to the number of different situations that can arise over the course of the game.
"A lot of companies working on A.I. use games to build intelligent algorithms because there's a lot of human-like intelligence capabilities that you need to beat the games," Maluuba project manager, Rahul Mehrotra said.
Doina Precup, an associate professor of computer science at Montreal's McGill University, said the team had made a significant breakthrough in A.I. by splitting the agents up to accomplish the difficult feat.
"This idea of having them work on different pieces to achieve a common goal is very interesting," she said. "That would be really, really exciting because it's another step toward more general intelligence."
