Because for a finite problem space (IE, shoot player in head), it's much much better than you. Not to mention exponentially faster, and probably looks better in a dress.
Because for a finite problem space (IE, shoot player in head), it's much much better than you. Not to mention exponentially faster, and probably looks better in a dress.
There's a distinction between "intelligent AI" and "perfect AI." I can't think of anyone on this forum who wouldn't like to see more intelligent enemies in the games they play.
Zombiemambo on
0
Options
AxenMy avatar is Excalibur.Yes, the sword.Registered Userregular
Because for a finite problem space (IE, shoot player in head), it's much much better than you. Not to mention exponentially faster, and probably looks better in a dress.
That's not really making them more intelligent though. At least IMO I would say (in the context of a FPS) making them more intelligent means having them use cover effectively, flanking, group strategy, maybe laying traps in advance, or simply falling back to a better position.
Axen on
A Capellan's favorite sheath for any blade is your back.
I mean that is a blanket statement of course and I'll totally admit that not all fun AI are super dumb.
But!
More often than not, we have to continually dumb down the AI in the playtesting process.
I have to relearn this with every new project actually. It's not something you think about on the other side of the fence I don't think.
It would be super super easy to create an AI (and I mean in most, not all cases), that you would never ever win against. In fact that is a tendency, because all you have to do is think logically. Most of my time spent in AI programming is allowing the player to win on a regular basis, hah. Not all the time of course. We are masochists at heart aren't we?
But of course, by intelligent AI, you might be referring to an AI with human-like tendencies. This is perfectly awesome if you can pull it off. The same process is involved though. Start with super human perfect and add some flaws.
I think when people want more intelligent AI they want more human AI. They want to recognize emotion, or emotional strategy. This is something that has always been strived for but limitations for how much we could get across were there. This is something I believe we can make leaps and bounds in.
If all FPS games had the enemy AI from F.E.A.R., I would be so very very happy.
FEAR had some cool AI. It's still pretty rudimentary, and I can understand why making the AI too "intelligent" could be an issue, but I'd love to see enemies working together, performing more advanced maneuvers and running away/surrendering when the odds are really stacked against them.
Yeah, what people want is something that can basically pass the turing test for video gaming: play against it online and you don't realize it's not a human. Humans forget, humans can only see part of the playing field, humans interpret things in specific ways, to make a "good" AI you need to incorporate all of this into it. Unfortunately, the cost to move from "basic pathfinding with simple value judgement system" to "full fledged senses system on top of a dynamic route planning system coupled with the ability to work in concert with other AIs, all with the ability to misjudge and forget as well as having feelings such as getting angry and charging in or holding a grudge on someone or getting revenge for a teammate's death" is huge while the overall play difference is minor for 90% of the playerbase.
If all FPS games had the enemy AI from F.E.A.R., I would be so very very happy.
FEAR had some cool AI. It's still pretty rudimentary, and I can understand why making the AI too "intelligent" could be an issue, but I'd love to see enemies working together, performing more advanced maneuvers and running away/surrendering when the odds are really stacked against them.
The first Medal of Honor game had Nazis jumping on grenades you threw to save the lives of their buddies. :?
emnmnme on
0
Options
AxenMy avatar is Excalibur.Yes, the sword.Registered Userregular
edited September 2009
Okay so all this AI talk got me to searching the net.
I found this guy http://www.elbot.com/. Fuck that guy. I spent 2 or so minutes playing "word association" with fucking 1s and 0s.
FYI that is thewinner (or maybe it came in third I dunno) of the 18th Loebner Prize for artificial intelligence.
edit- The thing made 3 out of 12 judges think it was a human, despite the fact that Elbot has a "robot" personality. :?
Axen on
A Capellan's favorite sheath for any blade is your back.
Wait, someone actually thought that it was a person? WTF
Three people in fact. Apparently one more and it would have "passed" the Turing Test. Which goes to show that the Turing Test is a bit worthless. Or at least it's execution is.
Axen on
A Capellan's favorite sheath for any blade is your back.
I'm always impressed when the AI just starts running away from me.
Best option really.
This.
Guess what the appropriate action is when I cut down your 3 bandit buddies in a single swing each or walk by you wearing equipment that is not only worth more than the GDP of your entire kingdom but is also glowing? Hint: It's not to attack me mindlessly. :P
It's one of the things I always loved about sitting around a table playing DnD with friends. When we were outnumbered 2-1 and we killed half of the attackers in the first round the rest would scatter while swearing/screaming.
Space Empires V has a surrender option and a demand surrender option.
After countless games of SEV, I'm still entirely unsure as to why these exist. Once I was in a massive game that by the end only had three empires remaining. I had about 2/3rds of the sectors under my control, and had spent outrageous amounts of materials and time blockading the fuck out of my borders. It would take forever to break through, and I had a drone defensive grid as well as some retaliatory "cruise missile" style drones(also used to punch through any defense they have).
Ok this isn't so bad. Sure, it'd kick their ass to attack, but if they had enough construction planets they could. And I turned the AI all the fuck the way up. Well, I get bored, and of course take out one drone group breaking through their borders and bombing one races shit without any actual risk to any fleets. Then I start sending in the dropships and a few.... "peacekeeping" fleets. So at this point I control almost 5/6ths of the galaxy. I figure considering the obvious way the war is going they're fucked, and send a message to one opposition race, being benevolent as I am, and demand their surrender. They refuse.
So I take them down to one planet. Another message. Nothing.
Well then the drones took them out. Bastards.
Next race! Same goddamn thing I controlled every sector, had them down to one planet, which was blockaded and surrounded and there was literally nothing they could possibly do to make any difference. There was literally 0% chance, unless I decided to lose, that they would be able to cause any damage to me.
Seriously why even have the surrender option.
If it's like Civ4, then they will never surrender to you if they hate you enough.
I was thinking the exact same thing when I was reading this, and if you hadn't of said it here I was going to by the time I caught up on the discussion
Wait, someone actually thought that it was a person? WTF
Three people in fact. Apparently one more and it would have "passed" the Turing Test. Which goes to show that the Turing Test is a bit worthless. Or at least it's execution is.
The Turing Test is worthless, but most people don't understand how it actually works. There's actually three participants. The tester, and two testees, one computer and one human. The test is whether or not the tester and pick out the human testee, but what's actually happening is that the computer is trying to lie to the tester and appear more human. So it has to imitate shit like typos, bad grammar, misunderstanding things, getting insulted, etc. If the human testee is a really smart guy who makes very few typos or grammatical errors, that's a very different test than if you have Ms Tiddles from accounting.
If all FPS games had the enemy AI from F.E.A.R., I would be so very very happy.
My two favorite AI stories com from F.E.A.R., actually. One, how good it can be, the other, how bad.
We'll start with the bad so the good seems more good.
Doing the usual when there's Replicas around, I was sneaking through a hallway and came across two soldiers having a bit of radio chatter to pass the time. I figured I could take them out with a grenade; they were fairly close, but still unaware. I toss it, it nicks the corner in front of me, bounces a bit, and lands on my foot. Naturally, I sprint in the direction that leads "away." Couple seconds later, massive explosion as the grenade blows up a fire extinguisher in addition to itself. AI response: "Did you hear something?"
Now, good:
Again, dealing with the Replicas, I come across a small team. Sneak attack one in the head, the other two dive for cover. One goes behind a concrete pillar, the other ducks behind a tipped-over coffee table. Seeing an opportunity, I start unloading into the table. Replica 2 starts taking some serious damage, and makes a run for it to a blind spot on my left. I don't kill him in time, but he's hurt, bad. Replica 1, the apparent commander, calls for reinforcements, then gives Replica 2 an order: "Head around and flank him!" Replica 2's reply?
"FUCK YOU!"
He then disobeyed orders and stayed put. I really liked this, because self preservation is something that adds a lot of "intelligence" to AI.
Okay so all this AI talk got me to searching the net.
I found this guy http://www.elbot.com/. Fuck that guy. I spent 2 or so minutes playing "word association" with fucking 1s and 0s.
FYI that is thewinner (or maybe it came in third I dunno) of the 18th Loebner Prize for artificial intelligence.
edit- The thing made 3 out of 12 judges think it was a human, despite the fact that Elbot has a "robot" personality. :?
Eh. He could remember my name, but not much else. He would also frequently meander off into meaningless conversation. He also couldn't seem to follow lines of conversation well.
If all FPS games had the enemy AI from F.E.A.R., I would be so very very happy.
My two favorite AI stories com from F.E.A.R., actually. One, how good it can be, the other, how bad.
We'll start with the bad so the good seems more good.
Doing the usual when there's Replicas around, I was sneaking through a hallway and came across two soldiers having a bit of radio chatter to pass the time. I figured I could take them out with a grenade; they were fairly close, but still unaware. I toss it, it nicks the corner in front of me, bounces a bit, and lands on my foot. Naturally, I sprint in the direction that leads "away." Couple seconds later, massive explosion as the grenade blows up a fire extinguisher in addition to itself. AI response: "Did you hear something?"
Now, good:
Again, dealing with the Replicas, I come across a small team. Sneak attack one in the head, the other two dive for cover. One goes behind a concrete pillar, the other ducks behind a tipped-over coffee table. Seeing an opportunity, I start unloading into the table. Replica 2 starts taking some serious damage, and makes a run for it to a blind spot on my left. I don't kill him in time, but he's hurt, bad. Replica 1, the apparent commander, calls for reinforcements, then gives Replica 2 an order: "Head around and flank him!" Replica 2's reply?
"FUCK YOU!"
He then disobeyed orders and stayed put. I really liked this, because self preservation is something that adds a lot of "intelligence" to AI.
With your typical AI, I notice it as being bad when it does something completely dumb, like not running away from a grenade that's about to explode to save itself.
To me intelligent AI means the AI mirrors the expected behaviors of whatever entity it's representing. If it's a drafted grunt who just wants to go home in one piece, he's going to stay behind cover all the time, and run/surrender if it's more likely that will keep him alive than something else. If it's a bunch of soldiers guarding an area and they hear shots or explosions nearby, they'll stop lounging around and go into full alert or whatever instead of looking around once and going back to their coffee.
Okay so all this AI talk got me to searching the net.
I found this guy http://www.elbot.com/. Fuck that guy. I spent 2 or so minutes playing "word association" with fucking 1s and 0s.
FYI that is thewinner (or maybe it came in third I dunno) of the 18th Loebner Prize for artificial intelligence.
edit- The thing made 3 out of 12 judges think it was a human, despite the fact that Elbot has a "robot" personality. :?
I fooled around with that. Question, where those 3 judges robots? I do not see how you could think it was human. I asked it if it wanted to hear a joke, it said yes, so I was like knock knock... and it responded telling me it doesn't like having dents on it.
It would be super super easy to create an AI (and I mean in most, not all cases), that you would never ever win against.
This depends on the problem. It is, for instance, much harder to create a competitive, non-cheating RTS AI than it is to create one for a puzzle game. Search problems are, at least at the moment, much more suited to AI than are planning ones.
I'm always impressed when the AI just starts running away from me.
Best option really.
This.
Guess what the appropriate action is when I cut down your 3 bandit buddies in a single swing each or walk by you wearing equipment that is not only worth more than the GDP of your entire kingdom but is also glowing? Hint: It's not to attack me mindlessly. :P
It's one of the things I always loved about sitting around a table playing DnD with friends. When we were outnumbered 2-1 and we killed half of the attackers in the first round the rest would scatter while swearing/screaming.
Space Empires V has a surrender option and a demand surrender option.
After countless games of SEV, I'm still entirely unsure as to why these exist. Once I was in a massive game that by the end only had three empires remaining. I had about 2/3rds of the sectors under my control, and had spent outrageous amounts of materials and time blockading the fuck out of my borders. It would take forever to break through, and I had a drone defensive grid as well as some retaliatory "cruise missile" style drones(also used to punch through any defense they have).
Ok this isn't so bad. Sure, it'd kick their ass to attack, but if they had enough construction planets they could. And I turned the AI all the fuck the way up. Well, I get bored, and of course take out one drone group breaking through their borders and bombing one races shit without any actual risk to any fleets. Then I start sending in the dropships and a few.... "peacekeeping" fleets. So at this point I control almost 5/6ths of the galaxy. I figure considering the obvious way the war is going they're fucked, and send a message to one opposition race, being benevolent as I am, and demand their surrender. They refuse.
So I take them down to one planet. Another message. Nothing.
Well then the drones took them out. Bastards.
Next race! Same goddamn thing I controlled every sector, had them down to one planet, which was blockaded and surrounded and there was literally nothing they could possibly do to make any difference. There was literally 0% chance, unless I decided to lose, that they would be able to cause any damage to me.
Seriously why even have the surrender option.
If it's like Civ4, then they will never surrender to you if they hate you enough.
I was thinking the exact same thing when I was reading this, and if you hadn't of said it here I was going to by the time I caught up on the discussion
I think Alpha Centauri handled it a little better. There was a specific routine for the AI to surrender--from there, the AI could elect to avoid you at all cost (say, moving to another continent), or, and this made things more interesting, become your best friend. Of course, this usually meant it had to intent to betray you later on.
Then again, unless you were playing at a higher difficulty mode, that wasn't entirely relevant, but it was a really nice touch. I still think the game had better AI than the recent Civ games.
In Gears of War 2, the friendly AI was actually decent. Whenever I was down, my AI buddy would pick me back up with the exception of one rare occasion in which it moved right past me.
Also, the friendly AI in Left 4 Dead was nicely done and helpful. That game probably has the best friendly AI system.
In Gears of War 2, the friendly AI was actually decent. Whenever I was down, my AI buddy would pick me back up with the exception of one rare occasion in which it moved right past me.
Also, the friendly AI in Left 4 Dead was nicely done and helpful. That game probably has the best friendly AI system.
I don't like the bots in L4D. I'm pretty sure that they never use grenades, which is a little annoying.
In Gears of War 2, the friendly AI was actually decent. Whenever I was down, my AI buddy would pick me back up with the exception of one rare occasion in which it moved right past me.
Also, the friendly AI in Left 4 Dead was nicely done and helpful. That game probably has the best friendly AI system.
Except when you've been pounced by a hunter a foot away from them, and they proceed to kill every zombie within two blocks before helping you.
That said, they were really great at picking off infected, especially mid-air.
If all FPS games had the enemy AI from F.E.A.R., I would be so very very happy.
I dug up this article that I remembered about FEAR specifically. It's got some interesting points about how the AI was somewhat "faked" by, for example, the AI seeking cover, and that cover being there by very specific level design.
That being said, I'm not bashing FEAR. Even the illusion of good AI makes for a fun game experience, and that's what actually matters.
If all FPS games had the enemy AI from F.E.A.R., I would be so very very happy.
I dug up this article that I remembered about FEAR specifically. It's got some interesting points about how the AI was somewhat "faked" by, for example, the AI seeking cover, and that cover being there by very specific level design.
That being said, I'm not bashing FEAR. Even the illusion of good AI makes for a fun game experience, and that's what actually matters.
I'm fairly certain the same method was used in Half-Life. I'm not complaining either way. I remember the first time I saw a replica soldier see me, turn and run, and only stop after he'd knocked over a bookshelf for cover.
That, and I swear that a lot of them wouldn't see you, but your light. Which basically forced you to rethink using it sometimes.
I think my favorite quote about AI is from CliffyB
"Good AI does everything it can to kill the player until it dies"
And then he elaborated that both parts of that were important. The AI needs to be using a full set of tools, using logic and thinking a lot, doing all it can but it also has to eventually get killed by the player.
I would disagree. At least in TBS and RTS I don't want the AI to do everything it can, since that usually involves using information that SHOULD be hidden the the comp.
The 'strongest' AI's tend to ignore stuff like espionage requirements, and Fog of war.
Takes all the fun out of it, when it is impossible to bluff or construct a 'hidden' army.
GooglyBoogly on
Into Alien Crossfire? It has been almost 10 years. Time to update your datalinks.
Try out my Comprehensive Datalinks Update. Now v1.3!
I don't play a lot of games any more, but one thing I've never seen is this:
When you're pew-pewing away at the mans from a distance, and they go down, why have I never seen AI that drops to the floor and plays dead, and then shoots you when you walk past?
Has such a technique ever been used, instead of the more common revenant/undead enemies that you can knock down and they get up after a time?
Okay so all this AI talk got me to searching the net.
I found this guy http://www.elbot.com/. Fuck that guy. I spent 2 or so minutes playing "word association" with fucking 1s and 0s.
FYI that is thewinner (or maybe it came in third I dunno) of the 18th Loebner Prize for artificial intelligence.
edit- The thing made 3 out of 12 judges think it was a human, despite the fact that Elbot has a "robot" personality. :?
I told it that I though it wasn't intelligent, and he asked why and when I tried to respond he stopped working.
You've got to be fucking kidding me.
That thing isn't even a particularly skilled AIMbot.
I hate these "AI contests". They're all so geared towards the idea that you can "cheat" the Turing Test. It's not like getting a passing grade grants your robot a soul, it's just an elegant rule of thumb.
I don't play a lot of games any more, but one thing I've never seen is this:
When you're pew-pewing away at the mans from a distance, and they go down, why have I never seen AI that drops to the floor and plays dead, and then shoots you when you walk past?
Has such a technique ever been used, instead of the more common revenant/undead enemies that you can knock down and they get up after a time?
Yes. I've seen this in Swat 4. Its a ridiculously simplistic system.
I had this thought about oblivion and Morrowind and came up with this thought. Now, I am a shitty programmer and someone more skilled than I will recognize the faults in this right away and the various logic holes so forgive me.
Now, we need a variable called "combat impression." We already had one in ESIII and ESIV that was something along the lines of "style" or "fashion," every time you went about in pauper clothes or rich clothes, you got a response from npcs. If you traveled in pauper clothes in ESIV, you wouldn't get robbed by bandits.
I say, we need to take the "fashion" variable further and take it to "combat Impression. A single traveler in nice clothes would be attacked by any kind of bandit on the road
"combat impression = pussy in nice britches"
Now, when you are travelling down the road in a plain leather armor and a single bandit sees you, he will assume you to be somewhat skilled and may run away (evil is a cowardly lot, etc.) A group of bandits however, would try to surround you and take you out as a group.
"combat impression=can tell his ass from a hole in the ground"
Now with each of these rankings, you have a value and if its less than the bandit or marauder, they attack blindly, if its about the same, they are cautious, and if its kind of high, they run to find friends or attack as a group.
Now, if the difference is super high, they all run at the first sight of you, some using chameleon or invisible plus fortify speed to get the hell out.
Of course, in my dream world, having a "combat impression" higher than a certain level would give you a speech penalty because a lot of people are scared of you. Hence bringing back the ability to wear a robe over your armor from ESIII, so that all that daedra armor doesn't scare the shopkeeper out of selling you anything (which it should have done anyway but whatever)
Unfortunately,
all I am asking for in that rant would require a lot more combat options for the enemy in the next ES game. I know I won't see this for a while but the idea of wearing robes to conceal my badassatude and so that I don't find 3 robed figures on the road to Elsewyre wearing ebony mail underneath their robes while archers and illusionists flank me from bushes smile.
Still, I want to be worried about the bandits on the roads more than the monsters.
I don't think computer AI opponents will ever replace the fun and satisfaction of competing against human players for a few reasons.
* It's hard to find that sweet spot of difficulty that isn't too easy or too hard. Everyone has different skill levels so scaling AI to match the player is extremely difficult.
* AI is faked by algorithms that are goal oriented. You could spend enough time developing an AI that could beat anyone. But would anyone find it fun being beaten mercilessly by an opponent who programmed to beat you like a rented mule?
* Games are supposed to be fun. Is it really *fun* knowing you beat the computer opponent at AI level 7? Or is it more fun to say you beat your 13 year old cousin who talks too much and think he's the shit?
* The social aspect of competition is something ingrained in us. Similar to the reason above, it's more fun to compete against people you know, rather than a complete stranger or someone who may not even be a person?
How many single player games are selling a shitton of copies these days? Not many. Almost every best selling game in the past few years is based on multiplayer.
I think that until we reach the point of self-directed-learning AI, we won't ever be happy with AI as a whole. I'm not trying to kill the discussion, I'm just pessimistic about game AI being both fun & realistic until some serious technological leaps can be made in AI technology. And when it happens (notice I said "when") people will be more concerned about its non-gaming applications to notice whether or not it can beat you in a game of X.
The inherent flaw with AI is that AI has to be just flawed enough to seem human. We're not going to see that kind of thing in games for a really long time, if ever.
Posts
Because for a finite problem space (IE, shoot player in head), it's much much better than you. Not to mention exponentially faster, and probably looks better in a dress.
There's a distinction between "intelligent AI" and "perfect AI." I can't think of anyone on this forum who wouldn't like to see more intelligent enemies in the games they play.
That's not really making them more intelligent though. At least IMO I would say (in the context of a FPS) making them more intelligent means having them use cover effectively, flanking, group strategy, maybe laying traps in advance, or simply falling back to a better position.
Sure.
I mean that is a blanket statement of course and I'll totally admit that not all fun AI are super dumb.
But!
More often than not, we have to continually dumb down the AI in the playtesting process.
I have to relearn this with every new project actually. It's not something you think about on the other side of the fence I don't think.
It would be super super easy to create an AI (and I mean in most, not all cases), that you would never ever win against. In fact that is a tendency, because all you have to do is think logically. Most of my time spent in AI programming is allowing the player to win on a regular basis, hah. Not all the time of course. We are masochists at heart aren't we?
But of course, by intelligent AI, you might be referring to an AI with human-like tendencies. This is perfectly awesome if you can pull it off. The same process is involved though. Start with super human perfect and add some flaws.
I think when people want more intelligent AI they want more human AI. They want to recognize emotion, or emotional strategy. This is something that has always been strived for but limitations for how much we could get across were there. This is something I believe we can make leaps and bounds in.
Steam ID XBL: JohnnyChopsocky PSN:Stud_Beefpile WiiU:JohnnyChopsocky
FEAR had some cool AI. It's still pretty rudimentary, and I can understand why making the AI too "intelligent" could be an issue, but I'd love to see enemies working together, performing more advanced maneuvers and running away/surrendering when the odds are really stacked against them.
The first Medal of Honor game had Nazis jumping on grenades you threw to save the lives of their buddies. :?
I found this guy http://www.elbot.com/. Fuck that guy. I spent 2 or so minutes playing "word association" with fucking 1s and 0s.
FYI that is thewinner (or maybe it came in third I dunno) of the 18th Loebner Prize for artificial intelligence.
edit- The thing made 3 out of 12 judges think it was a human, despite the fact that Elbot has a "robot" personality. :?
I told it that I though it wasn't intelligent, and he asked why and when I tried to respond he stopped working.
Three people in fact. Apparently one more and it would have "passed" the Turing Test. Which goes to show that the Turing Test is a bit worthless. Or at least it's execution is.
Buwahaha, I got it to go into [ELIZAMODE] where it is infact exactly like a male Eliza.
You'd think someone performing a turing test would check for something like that.
http://elbot.blogspot.com/2004/03/when-im-good-im-good-when-im-bad-im.html
I was thinking the exact same thing when I was reading this, and if you hadn't of said it here I was going to by the time I caught up on the discussion
I tried to talk to it and all I got was a bunch of stupid non sequiturs.
The Turing Test is worthless, but most people don't understand how it actually works. There's actually three participants. The tester, and two testees, one computer and one human. The test is whether or not the tester and pick out the human testee, but what's actually happening is that the computer is trying to lie to the tester and appear more human. So it has to imitate shit like typos, bad grammar, misunderstanding things, getting insulted, etc. If the human testee is a really smart guy who makes very few typos or grammatical errors, that's a very different test than if you have Ms Tiddles from accounting.
My two favorite AI stories com from F.E.A.R., actually. One, how good it can be, the other, how bad.
We'll start with the bad so the good seems more good.
Doing the usual when there's Replicas around, I was sneaking through a hallway and came across two soldiers having a bit of radio chatter to pass the time. I figured I could take them out with a grenade; they were fairly close, but still unaware. I toss it, it nicks the corner in front of me, bounces a bit, and lands on my foot. Naturally, I sprint in the direction that leads "away." Couple seconds later, massive explosion as the grenade blows up a fire extinguisher in addition to itself. AI response: "Did you hear something?"
Now, good:
Again, dealing with the Replicas, I come across a small team. Sneak attack one in the head, the other two dive for cover. One goes behind a concrete pillar, the other ducks behind a tipped-over coffee table. Seeing an opportunity, I start unloading into the table. Replica 2 starts taking some serious damage, and makes a run for it to a blind spot on my left. I don't kill him in time, but he's hurt, bad. Replica 1, the apparent commander, calls for reinforcements, then gives Replica 2 an order: "Head around and flank him!" Replica 2's reply?
He then disobeyed orders and stayed put. I really liked this, because self preservation is something that adds a lot of "intelligence" to AI.
Eh. He could remember my name, but not much else. He would also frequently meander off into meaningless conversation. He also couldn't seem to follow lines of conversation well.
Sigh.
With your typical AI, I notice it as being bad when it does something completely dumb, like not running away from a grenade that's about to explode to save itself.
To me intelligent AI means the AI mirrors the expected behaviors of whatever entity it's representing. If it's a drafted grunt who just wants to go home in one piece, he's going to stay behind cover all the time, and run/surrender if it's more likely that will keep him alive than something else. If it's a bunch of soldiers guarding an area and they hear shots or explosions nearby, they'll stop lounging around and go into full alert or whatever instead of looking around once and going back to their coffee.
I fooled around with that. Question, where those 3 judges robots? I do not see how you could think it was human. I asked it if it wanted to hear a joke, it said yes, so I was like knock knock... and it responded telling me it doesn't like having dents on it.
I think Alpha Centauri handled it a little better. There was a specific routine for the AI to surrender--from there, the AI could elect to avoid you at all cost (say, moving to another continent), or, and this made things more interesting, become your best friend. Of course, this usually meant it had to intent to betray you later on.
Then again, unless you were playing at a higher difficulty mode, that wasn't entirely relevant, but it was a really nice touch. I still think the game had better AI than the recent Civ games.
Also, the friendly AI in Left 4 Dead was nicely done and helpful. That game probably has the best friendly AI system.
I don't like the bots in L4D. I'm pretty sure that they never use grenades, which is a little annoying.
Except when you've been pounced by a hunter a foot away from them, and they proceed to kill every zombie within two blocks before helping you.
That said, they were really great at picking off infected, especially mid-air.
I dug up this article that I remembered about FEAR specifically. It's got some interesting points about how the AI was somewhat "faked" by, for example, the AI seeking cover, and that cover being there by very specific level design.
That being said, I'm not bashing FEAR. Even the illusion of good AI makes for a fun game experience, and that's what actually matters.
I'm fairly certain the same method was used in Half-Life. I'm not complaining either way. I remember the first time I saw a replica soldier see me, turn and run, and only stop after he'd knocked over a bookshelf for cover.
That, and I swear that a lot of them wouldn't see you, but your light. Which basically forced you to rethink using it sometimes.
"Good AI does everything it can to kill the player until it dies"
And then he elaborated that both parts of that were important. The AI needs to be using a full set of tools, using logic and thinking a lot, doing all it can but it also has to eventually get killed by the player.
The 'strongest' AI's tend to ignore stuff like espionage requirements, and Fog of war.
Takes all the fun out of it, when it is impossible to bluff or construct a 'hidden' army.
Try out my Comprehensive Datalinks Update. Now v1.3!
When you're pew-pewing away at the mans from a distance, and they go down, why have I never seen AI that drops to the floor and plays dead, and then shoots you when you walk past?
Has such a technique ever been used, instead of the more common revenant/undead enemies that you can knock down and they get up after a time?
You've got to be fucking kidding me.
That thing isn't even a particularly skilled AIMbot.
I hate these "AI contests". They're all so geared towards the idea that you can "cheat" the Turing Test. It's not like getting a passing grade grants your robot a soul, it's just an elegant rule of thumb.
I say, we need to take the "fashion" variable further and take it to "combat Impression. A single traveler in nice clothes would be attacked by any kind of bandit on the road
"combat impression = pussy in nice britches"
Now, when you are travelling down the road in a plain leather armor and a single bandit sees you, he will assume you to be somewhat skilled and may run away (evil is a cowardly lot, etc.) A group of bandits however, would try to surround you and take you out as a group.
"combat impression=can tell his ass from a hole in the ground"
Now with each of these rankings, you have a value and if its less than the bandit or marauder, they attack blindly, if its about the same, they are cautious, and if its kind of high, they run to find friends or attack as a group.
Now, if the difference is super high, they all run at the first sight of you, some using chameleon or invisible plus fortify speed to get the hell out.
Of course, in my dream world, having a "combat impression" higher than a certain level would give you a speech penalty because a lot of people are scared of you. Hence bringing back the ability to wear a robe over your armor from ESIII, so that all that daedra armor doesn't scare the shopkeeper out of selling you anything (which it should have done anyway but whatever)
Unfortunately,
Still, I want to be worried about the bandits on the roads more than the monsters.
http://www.fallout3nexus.com/downloads/file.php?id=16534
* It's hard to find that sweet spot of difficulty that isn't too easy or too hard. Everyone has different skill levels so scaling AI to match the player is extremely difficult.
* AI is faked by algorithms that are goal oriented. You could spend enough time developing an AI that could beat anyone. But would anyone find it fun being beaten mercilessly by an opponent who programmed to beat you like a rented mule?
* Games are supposed to be fun. Is it really *fun* knowing you beat the computer opponent at AI level 7? Or is it more fun to say you beat your 13 year old cousin who talks too much and think he's the shit?
* The social aspect of competition is something ingrained in us. Similar to the reason above, it's more fun to compete against people you know, rather than a complete stranger or someone who may not even be a person?
How many single player games are selling a shitton of copies these days? Not many. Almost every best selling game in the past few years is based on multiplayer.
I think that until we reach the point of self-directed-learning AI, we won't ever be happy with AI as a whole. I'm not trying to kill the discussion, I'm just pessimistic about game AI being both fun & realistic until some serious technological leaps can be made in AI technology. And when it happens (notice I said "when") people will be more concerned about its non-gaming applications to notice whether or not it can beat you in a game of X.