When the computer plays a game against a human player, a piece of software is constantly calculating what move is the most profitable, given the current state of the game and the game's history. How easy this task is depends of course on the competitor's ability, but also on the nature of the game itself: puzzles for example are games where the player is trying to overcome the environment.
Software playing agents however are not useful only in the strict sense of gaming! For example a lot of today's financial markets are run by them, as well as less competitive tasks, such as dispatching tasks over a network to the node that is most suitable to complete them. Or, given the hardware architecture of a machine and a set of tasks that have to execute, find the best ordering to maximally utilize the hardware resources available.
None, it is only a project idea.
The idea of this project is to develop an "artificial intelligence" agent that could address a real-world issue here at CERN.