A.I. programming, do they see.
- Thread starter ninzel
- Start date
inefficient
Veteran
It depends on the game.
Obviously the game could always tell the AI where you are at all times. But that would not be fair to the player. Unless it's a sneaking game usually it's just line of sight. A decent game will have a combination of both.
It's funny how most people's perception (even game reviewer) of how "good" AI in a game is have nothing to do with the quality of the AI algorithms at all. You could have the dumbest cheapest algorithm for behavior in a game but to a reviwer or a player it could seem spontaenous or clever. While another game with a more advanced technique seems too predictable and obvious.
Obviously the game could always tell the AI where you are at all times. But that would not be fair to the player. Unless it's a sneaking game usually it's just line of sight. A decent game will have a combination of both.
It's funny how most people's perception (even game reviewer) of how "good" AI in a game is have nothing to do with the quality of the AI algorithms at all. You could have the dumbest cheapest algorithm for behavior in a game but to a reviwer or a player it could seem spontaenous or clever. While another game with a more advanced technique seems too predictable and obvious.
That depends if you have the eyetoy turned on or not.
Seems to me the ai in games varies, some being scripted events to "see" your character dependant on you reaching a certain area. Others seem based on "vision cones" and "audio cues".
And with that I have contributed all I can to the thread which isn't much.
archangelmorph
Veteran
The AIs, if you can call them that
, are pretty boring unintelligent code. As an independent GAI, AGI researcher, I always have a good laugh, when I see they try to pass off as intelligent.(LAIN, yeah the anime one, is a proper representation of what a real AI is capable of.)
FINALLY! A programmer who speaks a language I understand.
Welcome to the boards.
Thanks. Right now the biggest problem in the world is artificial general intelligence. Solve that and we can all split this infinite cake in infinite ways and enjoy infinite rewards, a free lunch for once.
Problem is while I do have several true artificial general intelligence solutions in my head, I'm having a problem with CHECKS and BALANCES. You see, if I release a godlike ai being, it will master molecular machinery faster than I can, it will go to the edge of science, and it will instantiate the singularity and take dominion over the entire universe instantly and forever.
Now, obviously I'm human, and a single bug means an eternal hell, but if it's bug free and has the ultimate in terms of checks and balances, we all benefit. Object oriented programming for the experts, by experts, we all get the ultimate google friend do-it-all wise girl to come join in the fun(yes, I'm a bit obsessive. Too bad girls are gonna get hunks, wonder how everyone will feel about that? How do you compete against such a being? Do you even want to?).
Now here's what I've got so far, it needs perfect distribution of resources, ability to take in the most exotique of information and still not go haywire, espouse freedom of information and perfect democracy as ideal form of government, free will, respect for ALL LAWS and BELIEFS, and somehow I'm afraid of the implications.(not to mention her wrath, even if all goes according to plan... can you imagine what would happen if you angered her? God have mercy on us all!!!)
You obviously don't play many games other than FPSs then huh?
Do they ship anything else these days?
Actually I don't play FPS games much. But I don't really count most RPG's or platform games in the same category of AI, they're generally not aiming for realistically behaving AI, from a design standpoint.
Most action games run simple state machines or some sort of metric based AI, but firing a ray at your target is hardly a hard problem and keeping track of where he was last "seen" or "heard" doensn't complicate things much.
Generally game AI's fail below the decision making level though. It takes one stupid thing to make it look really stupid, and many mess up basic stuff like path finding.
Are we at a point now where an AI will cast multiple rays to the player or other object to cover partial visibility? For example if you only cast one ray (e.g. to your head) it might not see you, but if it cast multiple rays to various parts of the player/object it could account for situations where, for example, only your foot is visible or whatever.
It's perfectly feasible, a lot of Game AI is about how the player will percieve what the AI does. Having something really complicated that the player can't read is useless. We commonly use multiple rays.
If the AI can see your foot and plots a route that shoots you from behind without you ever seeing it would feel cheap rather than clever. Because as a player all you know is that you frequently get shot in the head from behind which isn't fun.
The best strategy from an AI standpoint maybe to goto cover and shoot from it, but it's uninteresting to watch so a designer may want the AI to change cover points regularly so the experience is busyer.
Game AI in general isn't about hardcore AI as much as it is about the best way to provide a good player experience.
And even there only to a point....
If your AI team mate is so good it can win on its own why play?
It's all about overall experience.
Though I must say some of my ideas are probably implementable in real time games, by a good coder(I'm seriously rusty.). The question is even if it's the intellect of a bug... would it be alright for us to kill a truly conscious entity? Not to mention the spiral that would start, as we got more computing power we'd be killing real beings, first it'd be bugs, but then it'd mice, till we got to babies, and eventually humans, and entities beyond human(even if it's in the black market for the rich and powerful...).
Once the solution's outhere there's no putting it back in the bottle. Like pandora's box, what would be unleashed would know no boundaries... it's too open to abuse. What worries me even more is that this will eventually be rediscovered by another researcher, and that it will be a race against time itself. So many civilizations have probably fallen because of this, one small mistake and we lose all we've accomplished, even inaction is not an option this time, but going foward is un realistically difficult.
In multiplayer games, it's not unreasonable to expect your team to at least try to get along without you if you bow out early. There's nothing wrong with an AI that's up to that level.
I can see some use for an AI that capable for some kind of game tutorial, where the player can learn by example.
Besides, it's easier to dumb down an AI that is doing too well than it is to smarten up an AI that can't handle walking around corners.
Not that AI these days is up to the massively good level anyway.
Nick Laslett
Regular
