Nightrainsrt4
Limp Gawd
- Joined
- Dec 15, 2008
- Messages
- 435
What I want to know is why despite such huge advances in tech A.I is still pretty bad; yet games made many years ago have better A.I.
AI is one of the tougher things to program. Making code act like a human is difficult, and can be extremely costly cycle-wise.
Another issues is that companies don't tend to share with one another. You have one company that comes up with some sweet AI, but they aren't going to share it with other companies. Each company has to make these strides on their own. Yes, devs sometimes move from one company to another, but they can't just take the implementation they came up with at the first company to their new employer. Unless there is a really good open source / previously documented AI code, the devs on a project essentially have to rebuild it for each game, or try and tweak an existing code the company has the rights to in order to have it work for the new game. Reusable code is also something that is difficult to come by. The more reusable you make something, generally the more difficult/expensive it is to implement it the first time.
It is similar to game physics. These concepts involve difficult computations. You need people who really know what they are doing, and that is expensive. It is why game companies resort to things like Physx or Havok instead of spending months/years and a lot of money on an in-house implementation. Most of the in-house physics stuff you see is generally pretty basic, as the algorithms and computations representing natural physics can be difficult to represent in a game; you really need someone who understands both physics and software engineering.