This is an idea that I thought would be really cool in an RTS Battlezonish kinda game: What if you had bots in a fictional game that you used to fight with (ala Battlzone, but human figure AI bots). What if the skills of those bots could upgraded by hiring other real players from a commutiy website to fight with your bot in a game. In this situation your bots would be persistent bots that you could take into another game forum. Now, if the guy you hiring is a great dogfighter then the game would record the vital playing statisitcs of that player, and would adjust the skill of the bot to match that of the player when the bot next plays another game alone. In this model, you could have a bot that contains the skills, tactics and experience of multiple real players that have previously played with that bot. For instance if the bots were in game that was Tribesish, than in one game you could have hired a sniper, and in another an enginneer in heavy armor. In the next game, you decide that the sniper player made your bot a kickass sniper, so decide that you dont need a human to control the bot, because it contains the accuracy and tactics of the real-player that played with it before. And if you ordered to bot to be a heavy suited engineer, it would contain the skills of the real-engineer that used it before. I have ideas on how this might work from a larger paradigmatic perspective in regards to networking, security etc, but I do not have much experience in programming AI. Tell me what you think? Edited by - TheEnderBean on June 28, 2001 2:38:35 PM

The two methods of learning AI that I''ve played around with a little bit are genetic algorithms and neural nets. Both work well in my opinion, but learning how to do them is not an easy task. I had to do a lot of reading before I could code anything with them.

If you are actually writing this game, you will want to get yourself an AI programmer to do it, or plan on doing a lot of studying.

(Don''t let me scare you though. I enjoyed reading and learning about AI. It''s just not a simple, learn in a single evening type thing.)

Shaft: thanks for the reply. One question: What do you think of the idea though?.

I think it''s a good idea. On my current game learning AI is on our "Hopeful" list. Not only would we have to write AI that learns, but we have to setup the networking such that the learned state is saved somewhere. We don''t have money for a server farm, we are going to do things like doom, where one machine in the current game acts as the server.

So after a game is played and the AI has been updated, how does everyone else in the world playing the game get that new updated state, and how does their game AI state get sent out.

We have a solution, but it''s just going to sit on our hopeful list for a while.