Artificial intelligence

The AI challenge

It is a common grief made to digital versions of board games that the Artificial Intelligence (AI) is poor.

The reasons are simple:

• developing AI is a costly thing: it takes a lot of time which means a lot of money
• even if the project budget allows enough time for the development of a good AI, the developers do not necessarily have the knowledge or capacity to program AI well. Only a tiny fraction of developers have ever written an AI and have any clue how to do it.

Our expertise

Happy Meeple AIs are developed by our expert Nicolas Guibert who has a long experience in AI.

In 1998, he started to work on a 10x10 draughts program. In 2003, his program Buggy beat world #8 Grand-Master N’Diaga Samb in a match organized in Cannes during the international games festival (see www.buggy-online.com).

Since then, Nicolas developed several AI engines including real-time bots that take instant decisions on fluctuating markets.

According to Nicolas Guibert,

"Artificial Intelligence is not only a question of techniques (alpha-beta, scoring, genetic algorithms, neural networks, expert systems, etc.), it is also a question of using the right one(s) and building the right model for the specific task. There is no unique recipe for designing an AI, one must build a different mathematical model for each specific case."

"Computer strength is computation. For all tactical games, that require a lot of computation, programming an AI is easy and the human player will always have a huge disadvantage. For games that require strategy, which means long-term planning, the computer struggles to compete with the human brain and this is where the AI challenge resides. When the program manages to beat a human player on the strategy battle field, then the time has come for the programmer to be proud of his work! Before that happens however, the human players can often have a very good laugh watching the program play very poorly!"