Everyone loves a few rounds of a classic video game, but why should humans have all the fun? The Ms Pac-Man vs Ghost Team Competition serves to redress the balance by putting AI controllers in charge of video game characters in an effort to see which plays the game best.
Competitors could submit AI controllers for either the titular Ms Pac-Man or the team of four ghosts and each entrant faced off against the rest to determine a winner. The Ms Pac-Man AI had to maximise its score, while the ghost AI had to prevent Ms Pac-Man from scoring. The competition was organised by Philipp Rohlfshagen and Simon Lucas, two computer scientists at the University of Essex, with the results announced today at the Congress on Evolutionary Computation in New Orleans.
How did the AI controllers do? Compared to a human, not great – the highest scoring Ms Pac-Man controller was 69,240, while the world record stands at more than 900,000 points. “I would assume that ‘professional’ human Ms Pac-Man players will be better than any AI controller at this stage,” says Rohlfshagen, though he added that the ghost teams were also much harder to play against than those found in the original game: “The original ghost team was developed to engage and entertain the human player whereas the ghost teams submitted in the competition were designed to eat Ms Pac-Man as efficiently as possible.”
Developing an AI to play video games for us isn’t really the aim though, and there is some serious research behind the competition. “Games are usually seen as a valuable test-bed for new technologies in computational intelligence as they are well defined yet very challenging,” explains Rohlfshagen. He says the multi-agent algorithms behind the ghost controllers could be used for transport or military applications, or even modelling biological predator-prey dynamics.
New Scientist reports, explores and interprets the results of human endeavour set in the context of society and culture, providing comprehensive coverage of science and technology news.