This document will be the closing capture statement of the Concept Development Phase. Requires customer approval.
I need a vehicle for testing my AI creations. Also, I need something that can convey what I know about AI. I finally found a competition that is designed in a way that will satisfy my requirements. It's called the 2k BotPrize, created by 2k games. The idea is to create an AI system that can play UT2k4 online against real people. To win the prize you're bot needs to beable to fool three judges into believing that you're bot is human. I plan to have two stages to this project. The first stage will implement a rules engine. This should be close enough to creating a system that could compete for the prize but not necessarily win it. The second stage will be used to create the TrueAI system I'm working on.
Win the 2k BotPrize.
The judges of the 2k BotPrize need to think that my bot is human.
Major selling points.
The judges of the 2k BotPrize. I don't know anything about them, they are judges.
For the first stage, this will need a machine capable of running windowsXP. The faster the hardware the better.
The second stage will need a network of computers, see TrueAI for hardware specs.
The BotPrize competition is in 1 September 2010, which will give 1 year of development time. This may be enough time for the first stage of the bot.
There is strong competition in this field. The competitors are most likely dedicated AI game developers. The last winner used a combination of a rule engine and a learning algorithm to win. Although the winner did not win first prize(there were no first place winners) he took 2nd prize. The reason there weren't any first place winners was do to the inability of any one bot to prove to all the judges that it was human.
Skills needed to accomplish this task are a good understanding of C++, Threading, Networking, rule engines, and other game AI techniques. For the second stage the team outlined in TrueAI are the same.
Not having a clear understanding of the competition will most likely hinder the level of performance that this system will hit. Also, this is a win or lose proposition. The only gain will be my better understanding of game AI.
The goal is to create an AI system that will essentially beat the Turing Test. The risks are high but the payoff could be a career in the field. Also, if I can pull of stage two, that will most likely reach news level.