Agent researchers have joined this AI tradition to the extent of discussing the possibility of building agent-based chess programs. However, we know only a single attempt to create one: by A. Drogoul, with MARCH (Multi-Agent Reactive Chess). His paper, "When Ants Play Chess" was one of the most referenced papers in multi-agent systems (MAS) research in its time, which is significant indeed, because it combines the current favourite ant concept with reactive behaviour as well as with emergence and ideas on local and global strategies.
We have decided to attempt a rational reconstruction, and replicate and improve his algorithm in order to draw more detailed conclusions on the behaviour and capabilities of such a MAS. In addition have developed a sophisticated ant-based chess program and played several thousands of games on the Free Internet Chess site (FICS). We have concluded that these simulated ants with some cooperative behaviour can play at a level of an average chessclub player in spite of the total lack of a global strategy and the 1 move "visiblity horizont".
During the seminar I briefly introduce the audience how this agent/ant-based chess program was built and what is the real significance of treating a traditionally graph-search game problem as a reactive, agent-based one. Attendees will have the chance to test the chess program. If time allows I will talk about how a new model of the chess "environment" and the introduction of entropy (or "antropy") can help the ants to play more efficiently than without them.