how will the Ai work

[okkoko]

hello i like to know is ther any fact laying around on how Ai will work out and such

[teknoman2]

with lots of "if" commands in the code i guess

[SophosTheWise]

Or maybe just "not"

[obyknven]

how will the Ai don't work

 

fix your topic.  Obsidian traditionally cant make AI, NWN2 (and Tony_K with his Companion and Monster AI) is best proof of this. For modern games AI is weakest point, dev's usually fix lack of AI by scripted events and cheats ( invulnerable enemies etc ). It's really bad because with good AI gameplay totally different (macro instead of boring micro, it's my own experiense when i play NWN with Tony_K mod and without him).

[teknoman2]

you only need to think of BG2 where npc mages had like 3 or 4 contingency spells active, so if you dispelled and hit them, they would recast their defences over and over, forcing you to have a full page of dispells every time you went against a mage. meanwhile the rules said "only one contingency at a time"

Edited October 29, 2013 by teknoman2

[nikolokolus]

 

how will the Ai don't work

 

fix your topic.  Obsidian traditionally cant make AI, NWN2 (and Tony_K with his Companion and Monster AI) is best proof of this. For modern games AI is weakest point, dev's usually fix lack of AI by scripted events and cheats ( invulnerable enemies etc ). It's really bad because with good AI gameplay totally different (macro instead of boring micro, it's my own experiense when i play NWN with Tony_K mod and without him).

 

 

Just to be a completely pedantic butthole, you really didn't "fix" anything?

[Lephys]

Just to be a completely pedantic butthole, you really didn't "fix" anything?

I'm pretty sure that quote can no longer have puppies, so... clearly it's fixed now. u_u

[okkoko]

at it base line AI is a mirro of action a enemy can take  in BG2 there where tons of Spells an enemy could react to but in PE it different so it should be ezyer to make a Ai that can react to the player choice i bet ?

[teknoman2]

usually the AI is made of scripts that predict every possible action of the player, and indicate to the npc to perform an appropriate reaction. if there is little time or money to spend on making the AI, usually they go for a level design that restricts the player options or/and cheats for the npc, like having unlimited spells, being able to always hit the player, no matter his defenses, being imune too friendly fire(balls) and so on

[Shepard]

Why don't Obisidian borrow some AI?

[Pipyui]

I don't know anything about video game AI, but a popular trick (or at least the one I know) is Min-Maxing.  Min-Maxing creates a tree of which each level represents a turn (though the game need not necessarily be turn-based, just turn-abstracted), and each node a possible action of that turn.  Thus the tree maps all possible immediate futures of the game, and the computer chooses it's own move based on some heuristic that either minimizes its own potential loses after X moves, or maximizes it's potential gain after X moves.

 

Lecture over.  Anyway, what I'm saying is, it's not unreasonable to think that the AI is calculating your party's potential responses to its own actions, and using this to minimize or maximize its own harm or damage dealing, respectively.

 

I'm almost certainly completely wrong though.

[okkoko]

so im simple themr is

 

 

ai will be like the dev playing or making a chess like AI you know like the electronic chess boards. they give us restriction in what we can do by that they can acturly make good opponent

[AGX-17]

There will certainly be less problems in the compiling than you personally encounter when coding.

 

I don't know anything about video game AI, but a popular trick (or at least the one I know) is Min-Maxing.  Min-Maxing creates a tree of which each level represents a turn (though the game need not necessarily be turn-based, just turn-abstracted), and each node a possible action of that turn.  Thus the tree maps all possible immediate futures of the game, and the computer chooses it's own move based on some heuristic that either minimizes its own potential loses after X moves, or maximizes it's potential gain after X moves.

 

Lecture over.  Anyway, what I'm saying is, it's not unreasonable to think that the AI is calculating your party's potential responses to its own actions, and using this to minimize or maximize its own harm or damage dealing, respectively.

 

I'm almost certainly completely wrong though.

Like teknoman touched on, there will be a lot of if/then/else statements. I've played plenty of games where even that fails and NPCs target a character they should know has high resistance to a particular type of damage despite the presence of an equally viable target that is weak to the given damage type. Video game AI is a misnomer, it's not intelligence in any true sense of the word. And last I heard, Min-Maxing was a metagaming/powergaming strategy to minimize the known dump stats and maximize the known useful ones in order to maximize combat capabilities at the cost of the RPing experience. i.e. 9 INT (+ INT bobblehead,) 1 Charisma in Fallout 3, in which Charisma's only effect was to boost the speech and barter skills, which Bethesda, either deliberately or out of incompetence, designed as a dump stat by providing ways of maxing the Speech and Barter skills out without Charisma's involvement. Not that there was much RPing to be done there beyond the karma system.

 

ai will be like the dev playing or making a chess like AI you know like the electronic chess boards. they give us restriction in what we can do by that they can acturly make good opponent

Computers are not very good at Chess. Games like P:E or the IE games are not Chess, or comparable to it in terms of game design, rules, mechanics, etc. It took a warehouse-filling supercomputer devoted solely to playing Chess to defeat a human grandmaster at Chess. And the rules of Chess, which are quite ancient, are what give you restrictions, not the AI.

Edited October 30, 2013 by AGX-17

[okkoko]

 

ai will be like the dev playing or making a chess like AI you know like the electronic chess boards. they give us restriction in what we can do by that they can acturly make good opponent

Computers are not very good at Chess. Games like P:E or the IE games are not Chess, or comparable to it in terms of game design, rules, mechanics, etc. It took a warehouse-filling supercomputer devoted solely to playing Chess to defeat a human grandmaster at Chess. And the rules of Chess, which are quite ancient, are what give you restrictions, not the AI.

 

yes the rules im talking about. lest say i have an attack that gets stronger the more  i add of others skill effect  also they eater encounter the enemy's attack or i am engaging in special attack on my enemies. that a rules just like in chess .

and you try play chess against the pc or tablet it can be quit challenged now battle in EP will be like in BG2 hit space bar an all and you can see that as you taking a tinkerer on your next move in real time..

 

Ai store in its buildup what kind of move it can accomplish to word defending or attacking or add special attribute to its skill all to what it oppserv the player do to an extent depending on the deficulty

[Lephys]

I think essentially, the best thing to do is to have the AI basically calculate, objectively, the best choices it can take. But then, throw in a slight dynamic element to its choosing so that sometimes it pulls one of those "no one would try this because it would incur such a disadvantage, so I'm not even going to prepare to defend against that" moves that catches you off guard.

 

It needs to be challenging, so the AI needs to not run around doing stupid stuff all the time, like letting you easily kite it, or always charging even super-speedy characters, etc. So, it needs to be efficient and tough in its decisions. But then, you don't want it to be SO predictable that you can basically juke it every time. So, while you can't make it actually THINK and consider things, like "Wait, even though this is the most sound tactical option, perhaps this is a trap...", you CAN at least simulate that by tossing in a bit of randomness to its decision-making process. That way, you don't actually know what it's going to do, exactly, until it does it, so you're required to watch and react to it as opposed to setting up the same trap every single time, etc.

 

Now, as to how to technically produce that, I dunno. But, I've seen PLENTY of the Skynet "we always pick the best choice, no matter what" AIs in games, so, I would think it wouldn't be too big of a leap to simply have a controlled chance of it NOT always picking the best choice, and/or sometimes picking one of a handful of good choices that result in differing circumstances for the player to deal with.

[Pipyui]

And last I heard, Min-Maxing was a metagaming/powergaming strategy to minimize the known dump stats and maximize the known useful ones in order to maximize combat capabilities at the cost of the RPing experience.

 

While this is also true, minmaxing (or minimaxing) is an algorithm to create aritficial intelligence in turn-based systems.  It is infeasible to calculate more than a very few moves into the future, as the complexity is exponential.  It almost certainly won't be used in PE, but it is relevent to the AI topic and how AI generally works.  (it was the only one related to games we learned in AI)  Wikipedia

 

It is also true that we haven't yet (or likely ever will) solved chess.  The warehouse computer example is decades old, though.  Recent chess programs can beat grandmasters not uncommonly, and fit in one or two server blades (not much space at all).  Wikipedia