ShankarArtificial Intelligence, The hardest thing to create?
January 16th 2011 05:35
I am a full-time gamer. I'm also a student of Game Development. And as a member of both these occupations, I must say, nothing has ever made me more curious and intrigued as Artificial Intelligence. These simple two words, according to me, entirely define the whole experience of a game. A bad AI, whether accompanied with 'mind-blasting' graphics or not, will most certainly lead to a disastrous game-play.
As a Developer and a Gamer, I'm often torn between the choices of how to make the AI in my game.
For Instance, As a gamer, I have NEVER been able to tell a good Artificial Intelligence from a bad one (Unless of course it REALLY SUCKS ASS!!!!). I just don't know how. I consider myself a very good Racing Game player, and hence I assume that any AI that can beat me is either good or, if he beats me by a long way, is unnecessarily hard and hence bad. But then again, there is always someone else who can beat that AI. So that get's me thinking, is the AI bad or do I just stink at this game.
As a developer, I'm constantly looking for something that'll challenge the average gamer enough to keep him excited and on the edge of the seat, but not challenge him so much that he keeps losing and throws away the joystick in disgust and frustration! However, all this pertains to the 'average game'. One who is 'OK/Decent' at playing games. But, at the back of my mind there is always this fearful thought running: "What if only Hardcore, extremely skilled gamers with lightning fast reflexes play my game. They'll think my Artificial Intelligence is lame and stop playing my game!!"
I always try to make the AI in my (mini-)games just they way I would want to see it as a gamer. But the problem is, as a gamer, I really have no idea what an Ideal AI is! And this puts me in a HUGE dilemma when I program my AI!
There is a solution to this though, It's called neural networks. It's a technique in which an AI starts of at Lvl 0 and then as he interacts more and more with the player, learns about how the player plays the game and what mistakes it made previously and tries to omit them. It's an amazing way to program an AI, but I always wonder, How many levels can the AI really go? And to what consequence? And I don't see many games out there with a neural network AI system out there, and it makes me wonder as to why they don't have it?? Is it because it doesn't work? And that inkling of doubt is enough to put me off the idea of neural networking (That and the fact that I'm a little lazy, and it takes a lot of work!)
As a Developer and a Gamer, I'm often torn between the choices of how to make the AI in my game.
For Instance, As a gamer, I have NEVER been able to tell a good Artificial Intelligence from a bad one (Unless of course it REALLY SUCKS ASS!!!!). I just don't know how. I consider myself a very good Racing Game player, and hence I assume that any AI that can beat me is either good or, if he beats me by a long way, is unnecessarily hard and hence bad. But then again, there is always someone else who can beat that AI. So that get's me thinking, is the AI bad or do I just stink at this game.
As a developer, I'm constantly looking for something that'll challenge the average gamer enough to keep him excited and on the edge of the seat, but not challenge him so much that he keeps losing and throws away the joystick in disgust and frustration! However, all this pertains to the 'average game'. One who is 'OK/Decent' at playing games. But, at the back of my mind there is always this fearful thought running: "What if only Hardcore, extremely skilled gamers with lightning fast reflexes play my game. They'll think my Artificial Intelligence is lame and stop playing my game!!"
I always try to make the AI in my (mini-)games just they way I would want to see it as a gamer. But the problem is, as a gamer, I really have no idea what an Ideal AI is! And this puts me in a HUGE dilemma when I program my AI!
There is a solution to this though, It's called neural networks. It's a technique in which an AI starts of at Lvl 0 and then as he interacts more and more with the player, learns about how the player plays the game and what mistakes it made previously and tries to omit them. It's an amazing way to program an AI, but I always wonder, How many levels can the AI really go? And to what consequence? And I don't see many games out there with a neural network AI system out there, and it makes me wonder as to why they don't have it?? Is it because it doesn't work? And that inkling of doubt is enough to put me off the idea of neural networking (That and the fact that I'm a little lazy, and it takes a lot of work!)
| 27 |
| Vote |
Add To: 
subscribe to this blog