London, Jan 23 : Researchers at Eotvos University in Budapest, Hungary, have worked out a way of getting a computer to play the 1980s video game Pac-Man, using tactics that it developed by itself.
In the game, the player controls a blob that eats dots and fruit, while trying to stay away from several pursuing ghosts.
If the blob munches a special 'power' dot, it can eat the ghosts for more points.
Andras Lorincz and Istvan Szita at Eotvos began by giving the program a selection of possible scenarios, such as 'if ghost nearby', and possible actions, such as 'move away'.
Ms Pac-Man program randomly combined scenarios with actions to produce rules, and then played games using combinations of those rules to figure out which ones work best.
The program also decides its own priorities, important for situations in which two rules conflict.
It decided that the most important rule was to avoid being eaten by ghosts, followed by pursuing any edible ghost.
The next rule said that if all moves seem equally fine, never turn back as you have already eaten the dots in that direction.
The researchers said that the resulting program narrowly outperformed average human players and the work is part of broader strategy for analysing the weaknesses of AI compared to human intelligence when using video games.
"Games are interesting and challenging for human intelligence and therefore an ideal means to explore what artificial intelligence is still missing," New Scientist quoted the researchers, as saying.
For instance, the program performed better than the average human player, but it failed to develop certain tactics that humans find useful, like waiting for ghosts to come near before eating a power dot to increase the potential effect of the dot.
"No such behaviour evolved in any of our experiments," the researchers said.