Think it will work???
Author
This is my first post on this site, and since I can never think small I will ask for every ones oppion is it possible to create NPC on a game act like real people? Like Random stuff we humans do? Any suggestions on if it can work?
systemerror! systemerror! systemerror!
ALAS... There are far to many things that we as ''humans'' take for granted to program into ANY game. For instance, if you were to make a sim of the world and use AI to make all of the little people do what they do, how would you make a list of all the possible functions for them to perform on a daily basis? You can''t. There are just too many.
Tah for now...
WitePhyr
Jagged Industries
\^/¦†€¶hÿr
[Jagged::Industries]
Tah for now...
WitePhyr
Jagged Industries
\^/¦†€¶hÿr
[Jagged::Industries]
I always believed that coding an NPC to make random stuff is the easy part (you could just code about a hundred patterns of behaviour and just trigger them every now and then).
The difficult part comes when you want the actions of your character to make sence and to be context sensitive. If the player doesn''t say anything, the NPC might actually break the uncomfortable silence by saying "so, who might you be then?"
etc.
Of course you can still code something like that, but always trying to predict all possible situations before hand. A real NPC (like a real human) is still far from possible.
Blade Runner must wait.
The difficult part comes when you want the actions of your character to make sence and to be context sensitive. If the player doesn''t say anything, the NPC might actually break the uncomfortable silence by saying "so, who might you be then?"
etc.
Of course you can still code something like that, but always trying to predict all possible situations before hand. A real NPC (like a real human) is still far from possible.
Blade Runner must wait.
I consider an NPG a creature as us: he has one goal, one seen and one mind. He must have an istructor system that says what is good and what bad. He must know how to do some primitive action: go in that place, make this, ecc.
Feelings? Talking as an human? ohh thi is too for a small CPU taht hasn''t a life for learn it.
I never develop this thinks, but i suggest u to learn about neural nets... and many many variuos algoritm that can allow you character to well do the actions he want.
Feelings? Talking as an human? ohh thi is too for a small CPU taht hasn''t a life for learn it.
I never develop this thinks, but i suggest u to learn about neural nets... and many many variuos algoritm that can allow you character to well do the actions he want.
what you need to consider is that memory for memory we, as humans, have more than a computer will ever have for a long time yet.
and thats with a massively consecutive pre-optimised neural network in approx greater than it''s (n^(1000000000^100000000))th generation.
AND with all that we still take up to a year just to learn to walk!
even if, "in theory ", you managed to create an AI system that deals with the complexity that humans have to deal with then you''d probably need more computing power than there is in the entire world!
However, what you have to also remember is that in games, you really don''t need to bother with much outside the games parameters which are usually pretty limited!
As for random stuff, fuzzy neural networks are probably a place to look at.
"Bad Day... F**K it!" -Stephen Baldwin (Usual Suspects)
Author
Hmmm interesting...guess it might be harder than I thought...but now I have something to strive for, now time to learn some languages, thanks for the in put though.
Actually many of the most basic problems for creating a realistic AI are solved by game encapsulation.
1) By haveing a very simple means by which the AI can gain knowledge of there enviorment you solve one of the truly huge hurdles in AI design. For example, creating a routine like
Get Surroundings() the AI could go and get all the information about the "room" they are in. Belive me this is a truly huge issue.
2) Objectives, each NPC would really only have a very narrow focus for objectives, you really don''t even have to plan what the NPC would want. I would start with a Hierechy of needs(pavlovian) then you could assign values according to stimuli. For example
Hunger, each hour that goes by simply increase this value by one, when you reach X points, of hunger, it increases by y points on the hierarchy for action.
But what I think you were mainly looking for is quirks, character attributes which reflect personality, this would be somewhat simple, make a massive list of quirks. Have a random sampling of these quirks assigned to certien soft structure attachments to certien stimuli. For example:
Stress (variable) when stress reaches X, quirk y has z chance of begining.
It "could" be done, but I would stay away from both expert system structure, and neural networks, both are possible, but simply not neccesary.
If you do decide to undertake such a project, remember that the most valuable tool for creating multiple intellegences is the reuse of "personality algorthims".
1) By haveing a very simple means by which the AI can gain knowledge of there enviorment you solve one of the truly huge hurdles in AI design. For example, creating a routine like
Get Surroundings() the AI could go and get all the information about the "room" they are in. Belive me this is a truly huge issue.
2) Objectives, each NPC would really only have a very narrow focus for objectives, you really don''t even have to plan what the NPC would want. I would start with a Hierechy of needs(pavlovian) then you could assign values according to stimuli. For example
Hunger, each hour that goes by simply increase this value by one, when you reach X points, of hunger, it increases by y points on the hierarchy for action.
But what I think you were mainly looking for is quirks, character attributes which reflect personality, this would be somewhat simple, make a massive list of quirks. Have a random sampling of these quirks assigned to certien soft structure attachments to certien stimuli. For example:
Stress (variable) when stress reaches X, quirk y has z chance of begining.
It "could" be done, but I would stay away from both expert system structure, and neural networks, both are possible, but simply not neccesary.
If you do decide to undertake such a project, remember that the most valuable tool for creating multiple intellegences is the reuse of "personality algorthims".
yeah, appologies for seeming over the top there, I was trying to dissuade you from trying to design a fully working world on the first go.
As the Anonymous Poster said in a game you simplify.
There's no need to reinvent the world just to make a character smile when another character says or does something (oo-er )
anyway, you get the point!
To make his answers even easier basically,
cheat! There's no need for the end user to understand how things work, in fact most won't care, so you can cheat, have hooks, triggers, etc, to cause reactions, etc.
Although, I do like neural networks (you may have noticed )
I agree, start with simple hard coded stuff.
Then if you getting somewhere, take a look at working out scripting methods, which is always an easy way to chop and change things!
oh and never be afraid to look at other peoples stuff to see if you can use something!
(provided its open source of course )
hope this helps,
"Bad Day... F**K it!" -Stephen Baldwin (Usual Suspects)
Edited by - Nutter2000 on March 19, 2001 9:38:50 AM
As the Anonymous Poster said in a game you simplify.
There's no need to reinvent the world just to make a character smile when another character says or does something (oo-er
)anyway, you get the point!
To make his answers even easier basically,
cheat! There's no need for the end user to understand how things work, in fact most won't care, so you can cheat, have hooks, triggers, etc, to cause reactions, etc.
Although, I do like neural networks (you may have noticed
)I agree, start with simple hard coded stuff.
Then if you getting somewhere, take a look at working out scripting methods, which is always an easy way to chop and change things!
oh and never be afraid to look at other peoples stuff to see if you can use something!
(provided its open source of course
)hope this helps,
"Bad Day... F**K it!" -Stephen Baldwin (Usual Suspects)
Edited by - Nutter2000 on March 19, 2001 9:38:50 AM
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement
