GG is under appreciated
by Michael Cozzolino · in General Discussion · 03/16/2005 (8:24 pm) · 105 replies
I was going to post this in the Snapshot of the day for the Constructor tool but I didn't want to drag that down.
Here are my thoughts.
I always start to feel nervous when GG announces something new and cool. Reason is most people jump for joy but their is a small segment that really ruins it. I mean I don't get it. These guys could be working in the industry making alot more money but they choose to do this because it is their passion. They want us to succeed. They are sacrificing alot in my opinion to help us realize our dreams. It just sucks when people complain. I can see that it is disheartening to GG and it saddens me that they are so under appreciated. I know some of you will just call me a fanboy but I don't give a shit.
Here are my thoughts.
I always start to feel nervous when GG announces something new and cool. Reason is most people jump for joy but their is a small segment that really ruins it. I mean I don't get it. These guys could be working in the industry making alot more money but they choose to do this because it is their passion. They want us to succeed. They are sacrificing alot in my opinion to help us realize our dreams. It just sucks when people complain. I can see that it is disheartening to GG and it saddens me that they are so under appreciated. I know some of you will just call me a fanboy but I don't give a shit.
About the author
Indie Developer in the Albany NY area. iOS, PC, Mac OSX development. http://itunes.apple.com/us/artist/michael-cozzolino/id367780489
#102
if they did 20-25% in AI, that is assuming they truly set the time for it as low as you state, then they probably would either have weaker graphics, story, content, rendering code, etc etc (or more time to develop the game)... time has to come from somewhere
I'd have to respectfully disagree... with a ridiculous ammount of hours spent playing Unreal Tournament (yes the old original one) the AI does not take the same path... from playing on it at many difficulties it didnt... even on capture the flag where theres only 2-3 paths, it would alternate at times... I used to also put the difficulty on Godlike to practice...
also note that the AI in UT gave the bots a tendancy to like a certain weapon, a rashness, a tendancy to camp, yes those damn bots would camp in strategic spots!... and some more configurations that give it more than what you would seem to give it credit for.
(this is even for the original UT a pretty aged game ... theres been 2003 and 2004 since then!)
Again this is the original Unreal Tournament, quite a dated game
These "simple" and "basic" improvements seem simple and basic, yet they are far from it...
"The AI should realize the walls are in the way of its target and not storm after you as soon as you enter the map"
easy statement to say, but very hard to do... have you attempted pathfinding ? then I'd ask have you attempted A* pathfinding or other more complicated algorithms? Then I'd ask you if you have attempted these in a game that would truly demonstrate this techinique you simply state in real time...
Its insanely hard... you either have some very hefty processing to do this or predefined paths (like UT)... The hefty processing limits the ammount of other features you can display on screen at one time, those fancy graphics, the ammount of enemies... particle effects, etc... and is that really worth an AI that avoids walls a bit smoother ?
predefined paths give it a fake and predictable feel... you can avoid this somewhat by setting multiple predefined paths, but this adds time to the level design (massive ammounts of time... have you ever tried adding those damn path apples in UT =/ then correct lift objects... then script objects to tell them to take shortcuts to jump etc etc...)
Not trying to come down on you, just realistically stating the challenges that I think you are misinterpreting
04/17/2005 (3:07 am)
Quote:IF game devs put 25% of their total effort into the AI (hell, even 20%), it would be awesome.
if they did 20-25% in AI, that is assuming they truly set the time for it as low as you state, then they probably would either have weaker graphics, story, content, rendering code, etc etc (or more time to develop the game)... time has to come from somewhere
Quote:
Trying a different tactic is something I rarely see the AI do...say, if it's a shooter game and a bot's coming to blow your head off, it'll take the same approach, follow the same 'plan' every time
I'd have to respectfully disagree... with a ridiculous ammount of hours spent playing Unreal Tournament (yes the old original one) the AI does not take the same path... from playing on it at many difficulties it didnt... even on capture the flag where theres only 2-3 paths, it would alternate at times... I used to also put the difficulty on Godlike to practice...
also note that the AI in UT gave the bots a tendancy to like a certain weapon, a rashness, a tendancy to camp, yes those damn bots would camp in strategic spots!... and some more configurations that give it more than what you would seem to give it credit for.
(this is even for the original UT a pretty aged game ... theres been 2003 and 2004 since then!)
Again this is the original Unreal Tournament, quite a dated game
Quote:
Some simple, basic improvements don't seem that hard to put in--only requiring some time and effort...for example, the "banging your head against a brick wall" mentality that I've seen over and over in games. The AI should realize that walls are in the way of its target and not storm after you as soon as you enter the map, which a lot of times end up with the AI running against a wall that inbetween it and you. This is just one example of it. But that's more on the pathing side of it, I suppose. But the AI never seems to have anything resembling patience.
These "simple" and "basic" improvements seem simple and basic, yet they are far from it...
"The AI should realize the walls are in the way of its target and not storm after you as soon as you enter the map"
easy statement to say, but very hard to do... have you attempted pathfinding ? then I'd ask have you attempted A* pathfinding or other more complicated algorithms? Then I'd ask you if you have attempted these in a game that would truly demonstrate this techinique you simply state in real time...
Its insanely hard... you either have some very hefty processing to do this or predefined paths (like UT)... The hefty processing limits the ammount of other features you can display on screen at one time, those fancy graphics, the ammount of enemies... particle effects, etc... and is that really worth an AI that avoids walls a bit smoother ?
predefined paths give it a fake and predictable feel... you can avoid this somewhat by setting multiple predefined paths, but this adds time to the level design (massive ammounts of time... have you ever tried adding those damn path apples in UT =/ then correct lift objects... then script objects to tell them to take shortcuts to jump etc etc...)
Not trying to come down on you, just realistically stating the challenges that I think you are misinterpreting
#103
As they explore their environment, they will learn how to use their 'senses' to determine where to go as well as how to go about getting there without getting noticed by players. One of the first aspects of this is their use of sight - some dumber zombies, if they've never seen a tree before, might be found gouging out the bark of an oak as you approach. But as you get closer, if you aren't concentrating on the fact that these aren't your regular zombies, you may gain his attention making too much noise...or maybe it's that musk that your character uses which catches the zombie's nose (what's left of one!). We have a pack of zombified panthers that will make use of scent quite well!
This is all still without the higher intelligent zombies, which will make use of their situations even more dasterdly! In fact, it's these high AI zombies which will stop to monitor your tactics before they mysteriously pop up in front of you in the middle of a field without you noticing them untill it's too late. In this example, a sort of false cheating is used to allow the zombies to move around when you aren't looking! I say it's false because you can do it yourself just sneaking up on someone just outside their line of sight, so why can't intelligent zombies?
They can see and understand their surroundings. They can hear a snapping twig. They can smell the distinct odor of fresh blood. They can plot on their victims. And sometimes, they can even use a little inherent ESP to find out where the action is. They can choose to ignore or react to just about anything we allow them to recognize. That's what the AI should be.
It's no walk in the park, that's for certain, but being an indie here at GG allows me the time to make it work. That is priceless! Thanks GG!
- Ronixus
04/17/2005 (6:58 am)
AI is actually one of the issues I've been trying to address as well. Right now I'm working on making zombies, both wandering and intelligent, realistic enough as well as scary. Granted AI can be a very time consuming task, but I believe what's come to be known as the average AI is really starting to bug me as well - which is why we're taking a new approach with these zombies.As they explore their environment, they will learn how to use their 'senses' to determine where to go as well as how to go about getting there without getting noticed by players. One of the first aspects of this is their use of sight - some dumber zombies, if they've never seen a tree before, might be found gouging out the bark of an oak as you approach. But as you get closer, if you aren't concentrating on the fact that these aren't your regular zombies, you may gain his attention making too much noise...or maybe it's that musk that your character uses which catches the zombie's nose (what's left of one!). We have a pack of zombified panthers that will make use of scent quite well!
This is all still without the higher intelligent zombies, which will make use of their situations even more dasterdly! In fact, it's these high AI zombies which will stop to monitor your tactics before they mysteriously pop up in front of you in the middle of a field without you noticing them untill it's too late. In this example, a sort of false cheating is used to allow the zombies to move around when you aren't looking! I say it's false because you can do it yourself just sneaking up on someone just outside their line of sight, so why can't intelligent zombies?
They can see and understand their surroundings. They can hear a snapping twig. They can smell the distinct odor of fresh blood. They can plot on their victims. And sometimes, they can even use a little inherent ESP to find out where the action is. They can choose to ignore or react to just about anything we allow them to recognize. That's what the AI should be.
It's no walk in the park, that's for certain, but being an indie here at GG allows me the time to make it work. That is priceless! Thanks GG!
- Ronixus
#104
Well, everyone I have known who plays games, including dozens of people from modding communities, would be glad have less flashy graphics for a challenging, capable AI. I doubt people would mind...in the last decade or so, I've seen better and better graphics and yet the same AI I played against when I was a kid, some twenty years ago.
Yeah, the time has to come from somewhere, but how many guys on a given team are working on the AI, compared to the models or 2D art? My point is: The most under-developed part of games is the AI, and I'm not the only one who notices (probably 80% of all the complaints I've ever heard about games is directed towards the AI).
Take a look around GG, for example. Look at the number of posts--
Graphics...1382;
Animation...1663;
3D Modeling...7851
Mapping...3636
Programming...3666
Artificial Intelligence...641
Audio...349
At least it beat out audio ;)
I've played the original Unreal, too. I wouldn't say the AI was great in that game, but yeah, there was 'some' variation in its attack. I seem to recall several bone-headed moves, however, which had nothing to do with pathing (yes, I have worked on pathing in real time, back on Star Trek Armada...mind you that was in space, not in tight corners of rooms and hallways, though--and, right, predefined paths was what I was getting at, numerous, so it doesn't look that contrived); I recall bots spawning over and over again in the same spot, while I splattered their heads a few dozen times until I finally got bored and went to a new location...mind you, every thirty or forty times, they would not spawn and then run directly into my awaiting rifle sight, rather they'd run in a different direction and avoid getting shot (how hard is it to add: if you get oblitered fifteen or more times at one spawn location, spawn somewhere else?!); and I recall many other marvels of AI strategy, such as avoiding me by running face-first into a corner and standing perfectly still (the "I'm invisible cuz he can't see me" strategy? =) ), so I could approach and shoot em in the back a few times. Yeah, I remember they'd camp at good spots...most would just stand there like idiots, though, after they missed me, but you're right, they weren't complete twits.
However, whenever I was in an area which had two or more entrances, yes, they would attack the same route--then die, respawn, and come back the same way...every time. Not once would they fall back and try to flank around me. Didn't matter how much I tweaked those bots myself, either. Again, the AI was better than other games at that time--but it wasn't very good overall.
Anyway, shooters are not widely known for challenging AI--guys who play them usually don't want an intellectual challenge...they want to blow up some bots, so why make them smart? Yet for strategy games...there's no excuse for a dumb AI.
I remember one game a while back called Liberty or Death and that game always managed to surprize me; most of the time, the AI was predictable and stupid, but every once in a while it would execute just a brilliant move and totally mess me up. It was almost like I was being lulled into a mind-numbing stupor by its usual moves and wham! from out of nowhere it would do something really unexpected. It only happened a few games, but it would pull a MacArthur-type maneuver (like his invasion during the Korean War, an amphibious assault far from the frontlines), exactly where I was the weakest. Civ 1 and 2 had moments like this, too.
04/17/2005 (7:17 am)
Matthew, I'm not sure what you mean by "assuming they set the time for it as low as you state." What time did I state?Well, everyone I have known who plays games, including dozens of people from modding communities, would be glad have less flashy graphics for a challenging, capable AI. I doubt people would mind...in the last decade or so, I've seen better and better graphics and yet the same AI I played against when I was a kid, some twenty years ago.
Yeah, the time has to come from somewhere, but how many guys on a given team are working on the AI, compared to the models or 2D art? My point is: The most under-developed part of games is the AI, and I'm not the only one who notices (probably 80% of all the complaints I've ever heard about games is directed towards the AI).
Take a look around GG, for example. Look at the number of posts--
Graphics...1382;
Animation...1663;
3D Modeling...7851
Mapping...3636
Programming...3666
Artificial Intelligence...641
Audio...349
At least it beat out audio ;)
I've played the original Unreal, too. I wouldn't say the AI was great in that game, but yeah, there was 'some' variation in its attack. I seem to recall several bone-headed moves, however, which had nothing to do with pathing (yes, I have worked on pathing in real time, back on Star Trek Armada...mind you that was in space, not in tight corners of rooms and hallways, though--and, right, predefined paths was what I was getting at, numerous, so it doesn't look that contrived); I recall bots spawning over and over again in the same spot, while I splattered their heads a few dozen times until I finally got bored and went to a new location...mind you, every thirty or forty times, they would not spawn and then run directly into my awaiting rifle sight, rather they'd run in a different direction and avoid getting shot (how hard is it to add: if you get oblitered fifteen or more times at one spawn location, spawn somewhere else?!); and I recall many other marvels of AI strategy, such as avoiding me by running face-first into a corner and standing perfectly still (the "I'm invisible cuz he can't see me" strategy? =) ), so I could approach and shoot em in the back a few times. Yeah, I remember they'd camp at good spots...most would just stand there like idiots, though, after they missed me, but you're right, they weren't complete twits.
However, whenever I was in an area which had two or more entrances, yes, they would attack the same route--then die, respawn, and come back the same way...every time. Not once would they fall back and try to flank around me. Didn't matter how much I tweaked those bots myself, either. Again, the AI was better than other games at that time--but it wasn't very good overall.
Anyway, shooters are not widely known for challenging AI--guys who play them usually don't want an intellectual challenge...they want to blow up some bots, so why make them smart? Yet for strategy games...there's no excuse for a dumb AI.
I remember one game a while back called Liberty or Death and that game always managed to surprize me; most of the time, the AI was predictable and stupid, but every once in a while it would execute just a brilliant move and totally mess me up. It was almost like I was being lulled into a mind-numbing stupor by its usual moves and wham! from out of nowhere it would do something really unexpected. It only happened a few games, but it would pull a MacArthur-type maneuver (like his invasion during the Korean War, an amphibious assault far from the frontlines), exactly where I was the weakest. Civ 1 and 2 had moments like this, too.
#105
"and is that really worth an AI that avoids walls a bit smoother ?"
Well, that depends on what type of game you want to make. The typical FPS, I guess players don't mind too much if AI bots are bouncing off walls in one spot--makes em easier to shot ;)
I think it's worth some extra effort to have the AI as something other than a moron. So, you're saying that making a "smart" AI is too much work, so don't bother? Or are you saying that it's not possible to make a smart AI? Well, on either point, I guess we disagree.
I admit I don't fully know all the finer details of programming--I'm not a programmer. I have worked on the AI in several games, mostly fixing other people's mistakes or shortsightedness, adding tons of stuff that should have been there in the first place. Activision's version of Civ (CtP1) was the worst--spent three and a half months fixing that AI. Yes, it was hard work at times (and sometimes it was pretty easy), and mostly trial and error, at times very frustrating and time-consuming, but it was worth it, I think.
"Not trying to come down on you, just realistically stating the challenges that I think you are misinterpreting"
No problem. I do appreciate your comments--got me thinking.
Christopher:
Some good ideas there. I've seen a few games which utilize hearing for the AI, but I haven't seen any that use smell. I think Manhunt was the last one I played which had the AI use some sort of sensory information to detect you--sound of course and I was impressed with its use of shadows, varying degrees of shade for stealth--instead of black or white, either they see you or they don't. Maybe some other games have done this but that's the last one I recall.
Boy, this thread has gotten off-topic...maybe I should post in the AI forum ;)
Thomas.
04/17/2005 (7:21 am)
(continued...)"and is that really worth an AI that avoids walls a bit smoother ?"
Well, that depends on what type of game you want to make. The typical FPS, I guess players don't mind too much if AI bots are bouncing off walls in one spot--makes em easier to shot ;)
I think it's worth some extra effort to have the AI as something other than a moron. So, you're saying that making a "smart" AI is too much work, so don't bother? Or are you saying that it's not possible to make a smart AI? Well, on either point, I guess we disagree.
I admit I don't fully know all the finer details of programming--I'm not a programmer. I have worked on the AI in several games, mostly fixing other people's mistakes or shortsightedness, adding tons of stuff that should have been there in the first place. Activision's version of Civ (CtP1) was the worst--spent three and a half months fixing that AI. Yes, it was hard work at times (and sometimes it was pretty easy), and mostly trial and error, at times very frustrating and time-consuming, but it was worth it, I think.
"Not trying to come down on you, just realistically stating the challenges that I think you are misinterpreting"
No problem. I do appreciate your comments--got me thinking.
Christopher:
Some good ideas there. I've seen a few games which utilize hearing for the AI, but I haven't seen any that use smell. I think Manhunt was the last one I played which had the AI use some sort of sensory information to detect you--sound of course and I was impressed with its use of shadows, varying degrees of shade for stealth--instead of black or white, either they see you or they don't. Maybe some other games have done this but that's the last one I recall.
Boy, this thread has gotten off-topic...maybe I should post in the AI forum ;)
Thomas.
Torque Owner T G Hempstock
I have seen some small sparks of actual AI intelligence in some games...one small area which it in fact responded or initiated something clever. The Sims did do good things with its AI. So some games have done one or two things right with the AI--if only we could have a game which incorporates all those things and adds a few others...
Well, I've modded Civ games, the AI especially, and...well, I wouldn't say it's great. It's okay...what makes the AI more challenging in Civ games is merely "cheating"--it's coded to be faster, more aggressive, and have more stuff on certain levels; it has a massive advantage over you, so it appears smarter. I've studied the code for the AI on all difficulty levels and that's all it is--it cheats. It's programmed to.
There are personality files which distinguish one AI nation from the other, though these just set rules for diplomacy, as well as some behavior, not so much for strategy.
I think less attention on the appearance of the game would help (most games are like driving what appears to be a Ferrari when there's an engine from a Gremlin under the hood). Some simple, basic improvements don't seem that hard to put in--only requiring some time and effort...for example, the "banging your head against a brick wall" mentality that I've seen over and over in games. The AI should realize that walls are in the way of its target and not storm after you as soon as you enter the map, which a lot of times end up with the AI running against a wall that inbetween it and you. This is just one example of it. But that's more on the pathing side of it, I suppose. But the AI never seems to have anything resembling patience.
Trying a different tactic is something I rarely see the AI do...say, if it's a shooter game and a bot's coming to blow your head off, it'll take the same approach, follow the same 'plan' every time, no matter how many times it fails, no matter how utterly impossible it is to get you with this strategy. I don't see how difficult it would be to code in something instructing it to try something a few times, and if it seems futile, try *something else* and if that doesn't work a few times, then try something else. Maybe something radical, maybe something only slightly different. Perhaps even add some unpredictability--some goals are not accomplished by logic alone...have it get mad ;)
Having the AI adapt to your strategies, well, that would be more complicated...depending upon how many strategies it's programmed to recognize and how many counter-strategies it can try and in what situations to use them, relative to some probability of success, I guess.
Oh well. My two cents...