Originally posted by Urban Ranger
I much rather they develop competent AIs that don't cheat - okay, cheat as little as possible - instead of tagging on MP.
I much rather they develop competent AIs that don't cheat - okay, cheat as little as possible - instead of tagging on MP.
Unfortunatly, there is this little tiny thing called profit. Gaming companies would spend less money by taging on generic MP code and call it replay value, then to take the time to make a decent AI. Few companies have not followed this path, one being Rebellion.
After looking at the AI code in the released code of AvP GE, granted, the AI does cheat, but only when not in the same "cell" as the player (Far AI), at which point they behave a different (Local AI). For example, if I play as an AL|EN in a level filled with marines, and I pass by a few marines and go up a pipe or somplace, the marines will eventualy be in the same cell as I am. They followed me. Now they switch to Local AI and act like they can't see me because I busted all the lights.
What I worked on a little (all I did really was search for a few bugs they wanted me to look for and fix up the text and add comments when needed), was a generic AI code that documents mistakes it makes. It's not some miracle learning AI, it would... say, learn not to charge the player head on, seeing that it's odds of "living" are low. It won't the AI from trying though.
I have one of my friends applying this to the AI code in AvPGE (since it's really the only game we have that has released source code and we all have). Perhaps we can ship this over to Rebellion and let them have some fun. Fun times ahead.
Comment