Start with a pathfinder that works on 10,000 MOB's interacting with 1000+ players on a client/server architecture accounting for latency with client side prediction. Once you've got that one licked and you still think AI is simple, then we can talk about making the MOB's realistic.
I think that what most of you are wanting is something more like RTS AI rather than genetic AI. I would agree with that. Roving bands of MOB's setting up camps, collecting resources, eating, sleeping, etc. That would be cool. But again, the first challenge is making a regional A* pathfinder that will work on that scale in a massively multiplayer setting.
If you already have an AI system based on individual MOBs, can't that be used also to guide movement? You have perception, you have the AI to say do this or do that (or don't do this or that), can't there be a recognition in movement just the same way?
By words/design alone, Rome can be built in one day too.
What is 'easy' on paper doesn't mean it is 'easy' when it is made. The amount of processing to deal with the pathfinding alone for each mobs + is it a flying/walking/swimming mob? + What is the terrain? + should the mob 'see' the player? (Line of sight) etc etc.
I only dealt with AI's in a very high-level generic sense but what you think 'appear to have intelligence' doesn't mean 'it has intelligence.'
No, it's Artificial intelligence. Why can't a MOB "know" it wants to go to X, check the landscape for hazards and means, and go there if possible?
Hey, if you think you can make one that can scale with the amount of mobs that MMOs have, go for it.
What you see in RTS vs the amount of mobs you see in MMOs are vastly different. example: How many mobs are there in WoW? Just in the world not in instances. Millions? o_O
First off, no, I can't make such a game alone. It would take the resources of a large company to do it. This rediculous challenge to anyone with an idea outside of a WoW clone to "make it themselves" is childish and beneath a snake's belly.
Secondly, processing power these days is far ahead of what it was a few years ago. Yes it would take more. So what? Charge another $1 a month subscription and pay for the extra it would take. And save some money in production by not having to make these lame static quests. And save some money down the road by not having to add "new content". Because you already have an ever evolving world that provides new experiences all the time. And spend your future development time making the world go 'round instead, making it much richer and deeper of an experience for the players.
Well, again, I don't see where you are coming from here. You seem to be just trying to be negative.
What is "transparency" in your context? Predictability? Set actions that the players then have set reactions to overcome?
Your point about MOBs getting better at "beating" the players has a little bit of merit, but not much. What exactly do you want here, MOBs that never do anything but stand there and get beat down in the same way over and over again? To you, that must be what's fun. But to me, and I'm guessing many other players, it would be more fun and more interesting if the MOBs acted with more intelligence through better AI. Do they become harder to beat? Yes, and good! That's more fun than MOBs that just perform a predictable script.
You are also ignoring all the other posssibilities. Most games that have hiding/stealth also have reveal abilities. This isn't that cut and dried as you are trying to imply.
Lets take a look at your list there...
"AI of varying types are spread across the world. It's somewhat random what you'll face."
Isn't that what this idea does?
"After fighting it you rate the fun."
Ok, what's "fun"? Does "more interesting" add to the fun? Does a little unpredictability add to the fun?
"Very gradually, AIs with the lowest fun ratings are removed."
And leave us with what, with the predictable behavior and set patterns to "win"?
And frankly, I'm tired of being told what fun is. I have my own ideas, thank you. You sound like a present day game developer, insisting that what you make is what's fun, and that's it. Well, no thank you.
Transparency means exactly how I defined it in the last thread: it's clear to the player what's happening.
When a mob jukes left instead of right, you see that. It's transparent.
When some complicated behind-the-scenes genetic mechanic is the reason he juked left, you don't see that. It's opaque.
Nobody's suggesting mobs that just stand there. I'm suggesting that varied mob AI is great, but a convoluted, complicated-to-implement way of getting to varied mob AI is not great.
In my proposed system, your own ideas for what's fun would go into voting all AIs to "stay on the island". Although personally I still think the strongest system is what games already do: making a few core distinct AIs and giving mobs variance primarily through their abilities. Obviously as core gameplay differs between different games, so must game AI, but for typical MMORPG-style gameplay it's a reliable way to vary fights pretty substantially.
"What is truly revealing is his implication that believing something to be true is the same as it being true. [continue]" -John Oliver
Hey, if you think you can make one that can scale with the amount of mobs that MMOs have, go for it.
What you see in RTS vs the amount of mobs you see in MMOs are vastly different. example: How many mobs are there in WoW? Just in the world not in instances. Millions? o_O
First off, no, I can't make such a game alone. It would take the resources of a large company to do it. This rediculous challenge to anyone with an idea outside of a WoW clone to "make it themselves" is childish and beneath a snake's belly.
Secondly, processing power these days is far ahead of what it was a few years ago. Yes it would take more. So what? Charge another $1 a month subscription and pay for the extra it would take. And save some money in production by not having to make these lame static quests. And save some money down the road by not having to add "new content". Because you already have an ever evolving world that provides new experiences all the time. And spend you future development time making the world go 'round instead, making it much richer and deeper of an experience for the players.
I can't even begin to imagine the processing power + the data transfer needed to make this all work in an MMO.
Will a 2MB line suffice for one player? multiply that by all the players who are playing.....
You can talk about it all you want though.
Like I said before, on paper, Rome can be built in a day.
Gdemami - Informing people about your thoughts and impressions is not a review, it's a blog.
Hey, if you think you can make one that can scale with the amount of mobs that MMOs have, go for it.
What you see in RTS vs the amount of mobs you see in MMOs are vastly different. example: How many mobs are there in WoW? Just in the world not in instances. Millions? o_O
First off, no, I can't make such a game alone. It would take the resources of a large company to do it. This rediculous challenge to anyone with an idea outside of a WoW clone to "make it themselves" is childish and beneath a snake's belly.
Secondly, processing power these days is far ahead of what it was a few years ago. Yes it would take more. So what? Charge another $1 a month subscription and pay for the extra it would take. And save some money in production by not having to make these lame static quests. And save some money down the road by not having to add "new content". Because you already have an ever evolving world that provides new experiences all the time. And spend you future development time making the world go 'round instead, making it much richer and deeper of an experience for the players.
I can't even begin to imagine the processing power + the data transfer needed to make this all work in an MMO.
Will a 2MB line suffice for one player? multiply that by all the players who are playing.....
You can talk about it all you want though.
Like I said before, on paper, Rome can be built in a day.
What data transfer? It's no more than any other MMO. In both cases, the player is being sent data that MOBx is moving here and doing this.
I don't know where you're coming from with any of that.
First, who says it's supposed to wipe out the players? It's only to supply less predictable AI behavior and make things more interesting. And part of having a randomness to choices for MOB AI is to allow for "mistakes" by them, so they aren't always doing the best thing, giving them perfect tactics (or predictable tactics).
Second, what do you mean that players can't see it? If MOBs change their behavior from the norm, that's very noticeable. If MOBs move into a cave and make it "home", and build structures, and roam the nearby areas hunting, that's noticeable. Especially to anyone who knows the area very well. Who makes their own home nearby.
If things change in the game world, players will notice. And I think most players would be thrilled, as opposed to the standby design of predictable, boring, pre-designed sameness.
AI #1 casting spells and AI #2 meleeing is great, and yes players can tell the difference. Anything else you've mentioned (which seemed to be the core of your thread) isn't great, because players can't readily see how it works ("transparency"), are extremely limited in how they can influence it, and it introduces a huge spread of balance issues.
As for the first bit, it seems like the intent is for mobs to learn/adapt to become "better" (the OP orc example). But "better" in this case isn't "more fun to fight against" but rather "better at defeating the player" (and when those orcs successfully hide to evade the player, they've essentially defeated the player.)
AI in current games make mistakes. AI in your proposed game becomes better at avoiding mistakes. It becomes more adept at beating the player.
Really if you wanted to inch towards a fun game, you'd be like:
AI of varying types are spread across the world. It's somewhat random what you'll face.
After fighting it you rate the fun.
Very gradually, AIs with the lowest fun ratings are removed.
We need AI which evolves to be more fun. AI which evolves to defeat players better (even if "defeat" is 'flee successfully') is not more fun.
Well, again, I don't see where you are coming from here. You seem to be just trying to be negative.
What is "transparency" in your context? Predictability? Set actions that the players then have set reactions to overcome?
Your point about MOBs getting better at "beating" the players has a little bit of merit, but not much. What exactly do you want here, MOBs that never do anything but stand there and get beat down in the same way over and over again? To you, that must be what's fun. But to me, and I'm guessing many other players, it would be more fun and more interesting if the MOBs acted with more intelligence through better AI. Do they become harder to beat? Yes, and good! That's more fun than MOBs that just perform a predictable script.
You are also ignoring all the other posssibilities. Most games that have hiding/stealth also have reveal abilities. This isn't that cut and dried as you are trying to imply.
Lets take a look at your list there...
"AI of varying types are spread across the world. It's somewhat random what you'll face."
Isn't that what this idea does?
"After fighting it you rate the fun."
Ok, what's "fun"? Does "more interesting" add to the fun? Does a little unpredictability add to the fun?
"Very gradually, AIs with the lowest fun ratings are removed."
And leave us with what, with the predictable behavior and set patterns to "win"?
And frankly, I'm tired of being told what fun is. I have my own ideas, thank you. You sound like a present day game developer, insisting that what you make is what's fun, and that's it. Well, no thank you.
Transparency means exactly how I defined it in the last thread: it's clear to the player what's happening.
When a mob jukes left instead of right, you see that. It's transparent.
When some complicated behind-the-scenes genetic mechanic is the reason he juked left, you don't see that. It's opaque.
Nobody's suggesting mobs that just stand there. I'm suggesting that varied mob AI is great, but a convoluted, complicated-to-implement way of getting to varied mob AI is not great.
In my proposed system, your own ideas for what's fun would go into voting all AIs to "stay on the island". Although personally I still think the strongest system is what games already do: making a few core distinct AIs and giving mobs variance primarily through their abilities. Obviously as core gameplay differs between different games, so must game AI, but for typical MMORPG-style gameplay it's a reliable way to vary fights pretty substantially.
And again, what are you talking about? Transparancy of what causes a MOB do do something? Why would a player need that? Unless it's for cheating purposes. What's important is that in both cases the player sees the MOB "juke left". But in fact, in an AI system like this the MOB can in fact juke left or right. Or not at all and do something entirely different, unpredictably.
Darkfall has mobs that see you from a very long way off, co-operate and "level" when they manage to do enough damage to one or more players. It is pretty neat. I've come across some seriously leveled spiders that proceeded to hand me my ass, and of course grow even stronger. A PvE version of Darkfall is actually one of my dream MMOs.
Also, I wonder about the AI in Rift with its invasion forces that emerge from Rifts, wander off and proceed to attack towns etc. And whether there is a link there with RTS AI? Anyone know?
And again, what are you talking about? Transparancy of what causes a MOB do do something? Why would a player need that? Unless it's for cheating purposes. What's important is that in both cases the player sees the MOB "juke left". But in fact, in an AI system like this the MOB can in fact juke left or right. Or not at all and do something entirely different, unpredictably.
I just gave you the two examples of exactly what was and wasn't transparent about your idea. How are you possibly still confused?
"What is truly revealing is his implication that believing something to be true is the same as it being true. [continue]" -John Oliver
specifically on the arguement that it is hard to hit the sweetspot on mob AI for a great variety of people.
What the OP is suggesting can be accomplished however, but I would see it as the following scenario:
Mobs get certain skills and attributes, but at first they may focus on brute strength. Now when they're defeated a lot, they might reconsider and shift to another tactic. There by when you would later return to these mobs they would use different tactics and/or skills.
Dev's could add stuff to certain mobs over time because coding this from the get go for 100's of types of mobs would require a lot of time, and isn't really necessary at launch.
Mobs however who truly adapt and would be capable of avoiding the player are simply no fun. As this takes the fun out of encounters. Mobs ganging up on the players would add to the fun if there is a low chance of this happening.
Ps: in the main post it should "wandering" instead of wondering. Wondering made it a funny read
What you describe is essentially a type of AI called a neural network. It is used to teach the AI over a period of time how to do something so it constantly improves. This is most often used in racing games, you set the cars up, tell them to reach the finish lime and the AI does this over and over and over again until it gets a certain score or you think it has learn't enough.
It is possible to apply these to realtime games although the implementation can be quite tricky. Arguably the hardest thing to control is how well the AI performs. You can run into situations (if the network is not set up right) where the AI stays stupid or (on the flipside) it becomes too smart.
I would certainly like to see this kind of AI implemented in realtime in more games (rather than used to set up the AI before shipping) but i think it needs to be done in single player games before it makes the jump to the MMO field.
MMO wish list:
-Changeable worlds -Solid non level based game -Sharks with lasers attached to their heads
And again, what are you talking about? Transparancy of what causes a MOB do do something? Why would a player need that? Unless it's for cheating purposes. What's important is that in both cases the player sees the MOB "juke left". But in fact, in an AI system like this the MOB can in fact juke left or right. Or not at all and do something entirely different, unpredictably.
I just gave you the two examples of exactly what was and wasn't transparent about your idea. How are you possibly still confused?
I'm not confused. I simply don't understand why it's an important point to you. It's not something players would want unless they want to control the reactions of the MOBs. And that's entirely opposite of the whole principle of unpredictability.
What you describe is essentially a type of AI called a neural network. It is used to teach the AI over a period of time how to do something so it constantly improves. This is most often used in racing games, you set the cars up, tell them to reach the finish lime and the AI does this over and over and over again until it gets a certain score or you think it has learn't enough.
It is possible to apply these to realtime games although the implementation can be quite tricky. Arguably the hardest thing to control is how well the AI performs. You can run into situations (if the network is not set up right) where the AI stays stupid or (on the flipside) it becomes too smart.
I would certainly like to see this kind of AI implemented in realtime in more games (rather than used to set up the AI before shipping) but i think it needs to be done in single player games before it makes the jump to the MMO field.
Not directly. This wouldn't be a network directly connecting the AI of MOBs in a "Hive Mind" sort of way. That would come as a sort of incidental thing, where a group of MOBs "learn" to use the same tactics, which are different than the normal tactics.
Let me try this with another example. It might be easier to understand, including the simplicity of it.
Lets take a large Orc Clan. The Wander (correction duly noted *grins*) into a new terrain type and make a home. Lets give these Orcs a normal starting skills of 40% in Swords and 25% in Spears. So when they attack something, they are 15% more likely to use a sword than a spear. But also 25% more likely to use a Spear than some other option.
Now lets say that there's a common enemy in this new terrain that they fight often, and lets say that the general armor typr of this other MOB is very good against sword attacks, but not good at all against Spear attacks (Leather Armor, perhaps).
So as time goes on and conflicts happen, more and more of the Orc MOBs who use spears win, as opposed to those who use swords. The Spear wielders survive better. But also, they gain skills in Spears, and that becomes their weapon of choice as that skill overtakes their Swords skill. In time, most of the Clan of Orcs will be those who use Spears.
And again, what are you talking about? Transparancy of what causes a MOB do do something? Why would a player need that? Unless it's for cheating purposes. What's important is that in both cases the player sees the MOB "juke left". But in fact, in an AI system like this the MOB can in fact juke left or right. Or not at all and do something entirely different, unpredictably.
I just gave you the two examples of exactly what was and wasn't transparent about your idea. How are you possibly still confused?
I'm not confused. I simply don't understand why it's an important point to you. It's not something players would want unless they want to control the reactions of the MOBs. And that's entirely opposite of the whole principle of unpredictability.
It's important because I work in the game industry so I see dumping hundreds of hours of dev time into something the players will never notice (or appreciate) as a bad move.
"What is truly revealing is his implication that believing something to be true is the same as it being true. [continue]" -John Oliver
And again, what are you talking about? Transparancy of what causes a MOB do do something? Why would a player need that? Unless it's for cheating purposes. What's important is that in both cases the player sees the MOB "juke left". But in fact, in an AI system like this the MOB can in fact juke left or right. Or not at all and do something entirely different, unpredictably.
I just gave you the two examples of exactly what was and wasn't transparent about your idea. How are you possibly still confused?
I'm not confused. I simply don't understand why it's an important point to you. It's not something players would want unless they want to control the reactions of the MOBs. And that's entirely opposite of the whole principle of unpredictability.
It's important because I work in the game industry so I see dumping hundreds of hours of dev time into something the players will never notice (or appreciate) as a bad move.
Ok, well, tell me Mr. Almighty Game Maker, what do I, as a game player of MMORPGs, want? What would I notice? And what would I appreciate? I really need to know and I cannot go on without your expert consent.
I shall post no more ideas or desires until I hear from you. It would probably be best to put it in writing and have it notarized, just to be official and all documented up and all.
For the love of God, some of these guys are so full of themselves.
What you describe is essentially a type of AI called a neural network. It is used to teach the AI over a period of time how to do something so it constantly improves. This is most often used in racing games, you set the cars up, tell them to reach the finish lime and the AI does this over and over and over again until it gets a certain score or you think it has learn't enough.
It is possible to apply these to realtime games although the implementation can be quite tricky. Arguably the hardest thing to control is how well the AI performs. You can run into situations (if the network is not set up right) where the AI stays stupid or (on the flipside) it becomes too smart.
I would certainly like to see this kind of AI implemented in realtime in more games (rather than used to set up the AI before shipping) but i think it needs to be done in single player games before it makes the jump to the MMO field.
Not directly. This wouldn't be a network directly connecting the AI of MOBs in a "Hive Mind" sort of way. That would come as a sort of incidental thing, where a group of MOBs "learn" to use the same tactics, which are different than the normal tactics.
Let me try this with another example. It might be easier to understand, including the simplicity of it.
Lets take a large Orc Clan. The Wander (correction duly noted *grins*) into a new terrain type and make a home. Lets give these Orcs a normal starting skills of 40% in Swords and 25% in Spears. So when they attack something, they are 15% more likely to use a sword than a spear. But also 25% more likely to use a Spear than some other option.
Now lets say that there's a common enemy in this new terrain that they fight often, and lets say that the general armor typr of this other MOB is very good against sword attacks, but not good at all against Spear attacks (Leather Armor, perhaps).
So as time goes on and conflicts happen, more and more of the Orc MOBs who use spears win, as opposed to those who use swords. The Spear wielders survive better. But also, they gain skills in Spears, and that becomes their weapon of choice as that skill overtakes their Swords skill. In time, most of the Clan of Orcs will be those who use Spears.
You missunderstand what a nerual network is. Think of it as a brain with lots of nodes that are all linked up (like a network). If we took your idea each orc would have it's own neural network to draw upon and learn.
As for the guy who thinks it is silly to sink time and effort into devopling things the players don't see, that is just silly. If the games industry took that approach games would be terrible. The player doesn't see what goes on behind the loot tables or the monster AI, they see the end result and that is what is important. The OP's idea is basically expanding upon the monster AI (more work required for AI) but the result for the players is a new and different experiance with the monster AI, not the same result as just using the stock standard AI.
MMO wish list:
-Changeable worlds -Solid non level based game -Sharks with lasers attached to their heads
The idea of MOB 'evolution' is quite interesting, if I think about navigate a map and see a bunch of low level orcs, ignoring them and, after some time, if I'll go again that map and see now a full fledged orcs willage there, with high level Orc Boss(es) too, makes it really interesting.
I think a better AI makes its role only when mobs are grouped, but it drives back 90% of players. Facing a group of mobs who are in dps-interupter-healer config, it is troublesone to most players if the mobs use even only the bare minimum of AI, as healer healing (and not wanding the tank), dps trying to kill the support player and interrupter stopping the long-evil-cast-mage
About same MOB being more strong/skilled over time, it may be difficult for starters areas, if for some reason game not enough populated, the newcomers will become dead meat even in lv.1 fields
Overall, nice idea. Hope see something like that implemented.
The idea of MOB 'evolution' is quite interesting, if I think about navigate a map and see a bunch of low level orcs, ignoring them and, after some time, if I'll go again that map and see now a full fledged orcs willage there, with high level Orc Boss(es) too, makes it really interesting.
I think a better AI makes its role only when mobs are grouped, but it drives back 90% of players. Facing a group of mobs who are in dps-interupter-healer config, it is troublesone to most players if the mobs use even only the bare minimum of AI, as healer healing (and not wanding the tank), dps trying to kill the support player and interrupter stopping the long-evil-cast-mage
About same MOB being more strong/skilled over time, it may be difficult for starters areas, if for some reason game not enough populated, the newcomers will become dead meat even in lv.1 fields
Overall, nice idea. Hope see something like that implemented.
Yes, now imagine you go back again later on, after seeing these Orcs living there for a long time. They've survived assaults by players, and made a living there. There can be good reasons why these Orcs aren't simply wiped out like in other games. I believe strongly in worldly games where there's not such a big gap in abilities, more in a percentage system. So no level 80's easily wiping out level 10 Orcs. Also, if these Orcs used AI to build defensive structures, these could be quite formidable. And they could require expensive efforts in construction and time for sieges to overcome them, or a massive assault for an alternative. And maybe it's just not worth it at this point because of the location, and lack of threat.
At any rate, imagine you go back, and you've always had no problem avoiding or defeating their scouts before. But this time, you meet no scouts. You get closer and meet no resistance of any kind. And when close enough, you discover the ruined remains of their hold. You might explore the ruins, looking for loot or clues to what happened, and might find one or the other or both. You might discover:
That the Orcs were wiped out by a Dragon
That the Orcs left for some unknown reason, and you could try tracking them to their new home
That the Orcs fell victim to any number of things, other Orcs, disease (may mean some evil abouts), lack of food (why?), etc.
That the Orcs mysteriously dissappeared, no signs of leaving, but there's a big burnt spot on the ground in the center of a religious shrine of evil nature
The more AI, and the more abilities built into a game, the more it can be very interesting. On an ongoing basis, no need for development added.
I'm a strong believer in that you build things in a worldly nature. You don't build a canon in a quest that sits there waiting to be used for a specific purpose, and when fired, a particular wall falls down. You build canons that can be moved, placed, built, traded, and shot with.
Oh, edit to add, for starter areas. In a game like this, there are no static spawns. Players can drive the MOBs out of their community areas, making it safer for newbs. You can have reasons why players might do this, in the form of community social needs. New players help build things, increase tax coffers, etc.
And again, what are you talking about? Transparancy of what causes a MOB do do something? Why would a player need that? Unless it's for cheating purposes. What's important is that in both cases the player sees the MOB "juke left". But in fact, in an AI system like this the MOB can in fact juke left or right. Or not at all and do something entirely different, unpredictably.
I just gave you the two examples of exactly what was and wasn't transparent about your idea. How are you possibly still confused?
I'm not confused. I simply don't understand why it's an important point to you. It's not something players would want unless they want to control the reactions of the MOBs. And that's entirely opposite of the whole principle of unpredictability.
It's important because I work in the game industry so I see dumping hundreds of hours of dev time into something the players will never notice (or appreciate) as a bad move.
Ok, well, tell me Mr. Almighty Game Maker, what do I, as a game player of MMORPGs, want? What would I notice? And what would I appreciate? I really need to know and I cannot go on without your expert consent.
I shall post no more ideas or desires until I hear from you. It would probably be best to put it in writing and have it notarized, just to be official and all documented up and all.
For the love of God, some of these guys are so full of themselves.
This has nothing to do with pride. It's straightforward, logical criticism.
"Full of themselves" would be if I automatically rejected logical criticism of my ideas. One person's doing that here, and it isn't me.
"What is truly revealing is his implication that believing something to be true is the same as it being true. [continue]" -John Oliver
Having behavior evolve sounds very interesting. But all AI does is mimic real players.
Maybe even more interesting would be if the orcs behavior was controlled by players, in a kind of rts type fashion. By incorporating a completely different type of play style, you would appeal to more types of gamers within the same game and also achieve the diversity that is currently lacking in mob behavior.
And again, what are you talking about? Transparancy of what causes a MOB do do something? Why would a player need that? Unless it's for cheating purposes. What's important is that in both cases the player sees the MOB "juke left". But in fact, in an AI system like this the MOB can in fact juke left or right. Or not at all and do something entirely different, unpredictably.
I just gave you the two examples of exactly what was and wasn't transparent about your idea. How are you possibly still confused?
I'm not confused. I simply don't understand why it's an important point to you. It's not something players would want unless they want to control the reactions of the MOBs. And that's entirely opposite of the whole principle of unpredictability.
It's important because I work in the game industry so I see dumping hundreds of hours of dev time into something the players will never notice (or appreciate) as a bad move.
Ok, well, tell me Mr. Almighty Game Maker, what do I, as a game player of MMORPGs, want? What would I notice? And what would I appreciate? I really need to know and I cannot go on without your expert consent.
I shall post no more ideas or desires until I hear from you. It would probably be best to put it in writing and have it notarized, just to be official and all documented up and all.
For the love of God, some of these guys are so full of themselves.
Irony.
-------- "Chemistry: 'We do stuff in lab that would be a felony in your garage.'"
The most awesomest after school special T-shirt: Front: UNO Chemistry Club Back: /\OH --> Bad Decisions
I think your "Genetic AI" concept is great (something I call Artificial Life). But do you have any programming experience?
You insist that it's easy to do, but how much about AI programming do you actually know? Have you ever tried programming AI?
It may be possible, but is definitely not easy.
Axehilt had some great observations, and was thinking like a programmer, not a dreamer. To truly appreciate how difficult your proposal actually is, you should try writing your own AI.
PS:
"Wonder" means to contemplate, "Wander" means to travel around aimlessly. I think you meant "wandering MOBs."
Having behavior evolve sounds very interesting. But all AI does is mimic real players.
Maybe even more interesting would be if the orcs behavior was controlled by players, in a kind of rts type fashion. By incorporating a completely different type of play style, you would appeal to more types of gamers within the same game and also achieve the diversity that is currently lacking in mob behavior.
There's actually some random korean game that does just this(based on starship troopers). One person is in charge of the AI, everyone else tries to beat them.
Probably would be berated by "this" community as not an MMO if it was on this side of the ocean.
_______________
Personally I think the OP is going to the "acedemic" level of genetic AI where the purpose is to actually innovatively solve a problem. Then expecting the AI to be a little not AI like too life like.
If I were in charge of genetic AI(in an MMO) I would give it around 30 loadout option(individual skills/gear/stat buffs All with preset actions), then let the AI choose 5 from the 30. The choosen loadouts based on the genetic part of the AI. This means you get an AI that actually does change, and with some clever design(building the loadouts) adapt to most situations players toss at it over a few days to a month.
Practice doesn't make perfect, practice makes permanent.
"At one point technology meant making tech that could get to the moon, now it means making tech that could get you a taxi."
I like the idea but it is too soon for it IMO. I will give some reasons:
First is that casual mmo (WoW for example) players dont wanna it to be too hard, they like story, lore, quests, mobs to be easy/bit challenging (not all, but I am guessing plenty of casual players (have job/life, family etc)). They are playing and have only few hours to play, they want enjoy and relax in that time, they want linear and sure path to go by. From that comes that all WoW type (using just an example and dont ahve anything against/for it, but it has biggest playerbase, and plenty of games like it) games wouldnt want to use sophistacated AI as they DONT need it. That leaves us with sandbox games. Those are usually done by indies with low budgets and they have plenty of other things to work.
So after seeing that only sandboxes would use such system, think on what sandboxes need to improve before doing genetic AI. I instantly think about larger world, I would imagine 3-4x times bigger then darkfall. Reason for that is that many players still wouldnt grind up all mobs and players would be scattered. Also games should work on wildlife AI like Ryzom did, it is far from above good, but still it is start, but we havent seen it implemented it anywhere alse yet. And that would be just basics. Then you need to improve even basic AI senses to go with world. Somewhat normal visual range (not like now that you stand 36m from monster it doesnt see ya, but if it is 35m it attacks you), hearing, smell, visual blocks (so you can get next to it by going behind rocks, or crounching in tall grass, or just sitting behind tree and it comes by without noticing you). Another thing that would be done before AI is world places placement and creation. No more non-stop mob grinds in same place (I mean that they respawn in few meters area all the time), locations should be a lot larger, large enough not even to cover your visual range. Mobs should wander looking for challenges/fights/(food/drink)(generic ideas why mobs can be wandering)/even maybe adventures LOL. They should interact (fighting/grouping etc) with other mobs as well...... So somewhat to sum it up, world should be TRULY DYNAMIC. I use word "dynamic" for word "changing", and not like GW2 or Rifts trying to sell it. I mean if change happened it happened and there is normal chance of things to never be same as they were. For example: you exterminated huge settlement of humans, so for days/weeks when you come there you will find scavangers trying to find something to loot (may be humans, may be other races if there would be any), there would be chance that if someone truly put INSANELY HUGE ammount of gold and people provide enough raw mats for that gold it might be rebuild, if not after few more weeks/months it grow more and more into ruins, while nothing except few strongest/enduring things can tell that there was something there, and latter even that is gone...
Just throwing idea from the top of my head. "Genetic AI" is nice idea, but we wouldnt feel it much unless plenty of other things are introduced/improved.
Also very smart AI isnt much needed in how MMORPGs goes these days. Basicly MMORPG is synonime (to most of players) to "killing monsters". And when you look it is easy to see why, main point of most mmorpg are to kill monster so you get stronger and so you can kill EVEN MORE monsters. Killing to just kill... "Genetic AI" wouldnt be noticeble (maybe minimally) in such games. For players to appreciate they need to change themselves (small chance of that) and their objectives. Each encouter with enemy should be adventure in itself not that going through game and doesnt even feeling like you fought. Damn there is plenty of games which you can go through without dieing once and putting much though it in combat. Fighting is like breathing nowdays in MMORPGs... That comes very much from "Levels" in games. They give too much power too fast. You level to lvl5 and lvl1 mobs are no challenge to you, they can barely freaking scratch ya doing few damage/sec when you have hunders/thousands... That was just example from mind and not reference to any game btw... So genetic AI would also really work only in skill based games or with REALY REALLLLY reworked level system...
And again, what are you talking about? Transparancy of what causes a MOB do do something? Why would a player need that? Unless it's for cheating purposes. What's important is that in both cases the player sees the MOB "juke left". But in fact, in an AI system like this the MOB can in fact juke left or right. Or not at all and do something entirely different, unpredictably.
I just gave you the two examples of exactly what was and wasn't transparent about your idea. How are you possibly still confused?
I'm not confused. I simply don't understand why it's an important point to you. It's not something players would want unless they want to control the reactions of the MOBs. And that's entirely opposite of the whole principle of unpredictability.
It's important because I work in the game industry so I see dumping hundreds of hours of dev time into something the players will never notice (or appreciate) as a bad move.
Ok, well, tell me Mr. Almighty Game Maker, what do I, as a game player of MMORPGs, want? What would I notice? And what would I appreciate? I really need to know and I cannot go on without your expert consent.
I shall post no more ideas or desires until I hear from you. It would probably be best to put it in writing and have it notarized, just to be official and all documented up and all.
For the love of God, some of these guys are so full of themselves.
This has nothing to do with pride. It's straightforward, logical criticism.
"Full of themselves" would be if I automatically rejected logical criticism of my ideas. One person's doing that here, and it isn't me.
Take it easy, guys. The problem here is communication. Axehilt's point is very valid, and is related to efficiency once you get that idea beyond text and into a game scenario. What I think Axehilt is missing is that this genetic AI affects mobs in the long run throughout multiple, possibly infinite situations. Sure, you can have various mob scripts for different types of terrain and so on, but there's a limited (still huge, but limited) number of unique situations that can arise from combat with those mobs in the terrain they belong to. If the AI is learning all the time and trying new things, then the number of unique situations grows exponentially.
Think of it like this: it's giving the AI a more human-like way of doing things. When you see a person doing something, you unconsciously know that there's a reason for the way in which he or she is doing it, but you don't know it. In current AI and games, if you see a mob doing something, you know it's just a script. If you see an orc wandering around a forest, you know it's not gonna react until you approach it at a certain, very precise distance. In general, they're utterly predictable. With this new type of AI, the possibilities rise, and even if you only see the orc moving left or wandering in the forest, you can't be exactly sure why. Under these circumstances you can't even be sure if it will actually attack you if you approach it a certain distance. Maybe it will run away. Maybe it will talk. Maybe it'll call its friends hidden in the trees and it turns out it's a trap.
Now, in a current game, the traps might surprise you once or twice. But once you're going through a replay or even just facing the same mobs again, that trap will stop surprising you, it will stop working because you know you have to bring a team, or a certain item, or whatever, because it will always be the same. This is what happens with RTS AI: once you've faced it enough times, you know it doesn't build X or Y, or send X or Y troops to attack, or whatever, and those mistakes are key to defeating it. In a multiplayer match of RTS, you generally don't know what to expect, unless, of course, the system has been figured out in detail enough to let the players use the same tactics every game. Regardless of that. in multiplayer matches you have to actively think about what's going on and what it is you're going to do, all the time. In MMOs this rarely happens, especially when there are no scripts in play as in instances and so on. You set up a 'winning combo' against X or Y mob group, and it's only a matter of rinse and repeat.
So I believe that's why it would be a good idea to implement, because it's not a mechanic that affects only one type of encounter - it's a mechanic that generates several types of encounters. The orc might move left, and that's all you'd see, but you wouldn't know for sure why, or if it will do it again. Games would be a lot more interactive, a lot more thoughtful, and not as reliant on visual cues as you're demonstrating they currently are.
And if we take it a little bit further, it may also be the solution to THE GRIND, haha. Grinding is just a mechanic operation, the rinse and repeat, no thought, no interactivity, just taking in the visual cues and pressing buttons in certain orders. If mobs were deeper than that, if we would actually have to question their motives and imagine the possibilities (is it a trap? is that orc friendly or aggressive?) then I believe the feeling of nothingness and chore that the grind produces would be finally overcome. I'm not a game programmer or anything, though, so I don't know, haha.
And again, what are you talking about? Transparancy of what causes a MOB do do something? Why would a player need that? Unless it's for cheating purposes. What's important is that in both cases the player sees the MOB "juke left". But in fact, in an AI system like this the MOB can in fact juke left or right. Or not at all and do something entirely different, unpredictably.
I just gave you the two examples of exactly what was and wasn't transparent about your idea. How are you possibly still confused?
I'm not confused. I simply don't understand why it's an important point to you. It's not something players would want unless they want to control the reactions of the MOBs. And that's entirely opposite of the whole principle of unpredictability.
It's important because I work in the game industry so I see dumping hundreds of hours of dev time into something the players will never notice (or appreciate) as a bad move.
Ok, well, tell me Mr. Almighty Game Maker, what do I, as a game player of MMORPGs, want? What would I notice? And what would I appreciate? I really need to know and I cannot go on without your expert consent.
I shall post no more ideas or desires until I hear from you. It would probably be best to put it in writing and have it notarized, just to be official and all documented up and all.
For the love of God, some of these guys are so full of themselves.
Irony.
The real irony of it is that I'm a gamer saying I like this idea. And there are other gamers saying they like it too. But he's saying gamers don't want it because they can't see the underlying mechanic behind it, and he's a game programmer, and therefor gamers don't want it. Even though they say they do.
And yet, even though it appears that programmers are saying this might just be possible, the fact that I'm not a game programmer means that it shouldn't be done. Not even explored. It should just be dropped and forgotten. The main reason being that I'm not a game programmer.
Comments
First off, no, I can't make such a game alone. It would take the resources of a large company to do it. This rediculous challenge to anyone with an idea outside of a WoW clone to "make it themselves" is childish and beneath a snake's belly.
Secondly, processing power these days is far ahead of what it was a few years ago. Yes it would take more. So what? Charge another $1 a month subscription and pay for the extra it would take. And save some money in production by not having to make these lame static quests. And save some money down the road by not having to add "new content". Because you already have an ever evolving world that provides new experiences all the time. And spend your future development time making the world go 'round instead, making it much richer and deeper of an experience for the players.
Once upon a time....
Transparency means exactly how I defined it in the last thread: it's clear to the player what's happening.
When a mob jukes left instead of right, you see that. It's transparent.
When some complicated behind-the-scenes genetic mechanic is the reason he juked left, you don't see that. It's opaque.
Nobody's suggesting mobs that just stand there. I'm suggesting that varied mob AI is great, but a convoluted, complicated-to-implement way of getting to varied mob AI is not great.
In my proposed system, your own ideas for what's fun would go into voting all AIs to "stay on the island". Although personally I still think the strongest system is what games already do: making a few core distinct AIs and giving mobs variance primarily through their abilities. Obviously as core gameplay differs between different games, so must game AI, but for typical MMORPG-style gameplay it's a reliable way to vary fights pretty substantially.
"What is truly revealing is his implication that believing something to be true is the same as it being true. [continue]" -John Oliver
I can't even begin to imagine the processing power + the data transfer needed to make this all work in an MMO.
Will a 2MB line suffice for one player? multiply that by all the players who are playing.....
You can talk about it all you want though.
Like I said before, on paper, Rome can be built in a day.
Gdemami -
Informing people about your thoughts and impressions is not a review, it's a blog.
What data transfer? It's no more than any other MMO. In both cases, the player is being sent data that MOBx is moving here and doing this.
Once upon a time....
And again, what are you talking about? Transparancy of what causes a MOB do do something? Why would a player need that? Unless it's for cheating purposes. What's important is that in both cases the player sees the MOB "juke left". But in fact, in an AI system like this the MOB can in fact juke left or right. Or not at all and do something entirely different, unpredictably.
Once upon a time....
Darkfall has mobs that see you from a very long way off, co-operate and "level" when they manage to do enough damage to one or more players. It is pretty neat. I've come across some seriously leveled spiders that proceeded to hand me my ass, and of course grow even stronger. A PvE version of Darkfall is actually one of my dream MMOs.
Also, I wonder about the AI in Rift with its invasion forces that emerge from Rifts, wander off and proceed to attack towns etc. And whether there is a link there with RTS AI? Anyone know?
I just gave you the two examples of exactly what was and wasn't transparent about your idea. How are you possibly still confused?
"What is truly revealing is his implication that believing something to be true is the same as it being true. [continue]" -John Oliver
I mostly agree with Axehilt on this.
specifically on the arguement that it is hard to hit the sweetspot on mob AI for a great variety of people.
What the OP is suggesting can be accomplished however, but I would see it as the following scenario:
Mobs get certain skills and attributes, but at first they may focus on brute strength. Now when they're defeated a lot, they might reconsider and shift to another tactic. There by when you would later return to these mobs they would use different tactics and/or skills.
Dev's could add stuff to certain mobs over time because coding this from the get go for 100's of types of mobs would require a lot of time, and isn't really necessary at launch.
Mobs however who truly adapt and would be capable of avoiding the player are simply no fun. As this takes the fun out of encounters. Mobs ganging up on the players would add to the fun if there is a low chance of this happening.
Ps: in the main post it should "wandering" instead of wondering. Wondering made it a funny read
What you describe is essentially a type of AI called a neural network. It is used to teach the AI over a period of time how to do something so it constantly improves. This is most often used in racing games, you set the cars up, tell them to reach the finish lime and the AI does this over and over and over again until it gets a certain score or you think it has learn't enough.
It is possible to apply these to realtime games although the implementation can be quite tricky. Arguably the hardest thing to control is how well the AI performs. You can run into situations (if the network is not set up right) where the AI stays stupid or (on the flipside) it becomes too smart.
I would certainly like to see this kind of AI implemented in realtime in more games (rather than used to set up the AI before shipping) but i think it needs to be done in single player games before it makes the jump to the MMO field.
MMO wish list:
-Changeable worlds
-Solid non level based game
-Sharks with lasers attached to their heads
I'm not confused. I simply don't understand why it's an important point to you. It's not something players would want unless they want to control the reactions of the MOBs. And that's entirely opposite of the whole principle of unpredictability.
Once upon a time....
Not directly. This wouldn't be a network directly connecting the AI of MOBs in a "Hive Mind" sort of way. That would come as a sort of incidental thing, where a group of MOBs "learn" to use the same tactics, which are different than the normal tactics.
Let me try this with another example. It might be easier to understand, including the simplicity of it.
Lets take a large Orc Clan. The Wander (correction duly noted *grins*) into a new terrain type and make a home. Lets give these Orcs a normal starting skills of 40% in Swords and 25% in Spears. So when they attack something, they are 15% more likely to use a sword than a spear. But also 25% more likely to use a Spear than some other option.
Now lets say that there's a common enemy in this new terrain that they fight often, and lets say that the general armor typr of this other MOB is very good against sword attacks, but not good at all against Spear attacks (Leather Armor, perhaps).
So as time goes on and conflicts happen, more and more of the Orc MOBs who use spears win, as opposed to those who use swords. The Spear wielders survive better. But also, they gain skills in Spears, and that becomes their weapon of choice as that skill overtakes their Swords skill. In time, most of the Clan of Orcs will be those who use Spears.
Once upon a time....
It's important because I work in the game industry so I see dumping hundreds of hours of dev time into something the players will never notice (or appreciate) as a bad move.
"What is truly revealing is his implication that believing something to be true is the same as it being true. [continue]" -John Oliver
Ok, well, tell me Mr. Almighty Game Maker, what do I, as a game player of MMORPGs, want? What would I notice? And what would I appreciate? I really need to know and I cannot go on without your expert consent.
I shall post no more ideas or desires until I hear from you. It would probably be best to put it in writing and have it notarized, just to be official and all documented up and all.
For the love of God, some of these guys are so full of themselves.
Once upon a time....
You missunderstand what a nerual network is. Think of it as a brain with lots of nodes that are all linked up (like a network). If we took your idea each orc would have it's own neural network to draw upon and learn.
As for the guy who thinks it is silly to sink time and effort into devopling things the players don't see, that is just silly. If the games industry took that approach games would be terrible. The player doesn't see what goes on behind the loot tables or the monster AI, they see the end result and that is what is important. The OP's idea is basically expanding upon the monster AI (more work required for AI) but the result for the players is a new and different experiance with the monster AI, not the same result as just using the stock standard AI.
MMO wish list:
-Changeable worlds
-Solid non level based game
-Sharks with lasers attached to their heads
The idea of MOB 'evolution' is quite interesting, if I think about navigate a map and see a bunch of low level orcs, ignoring them and, after some time, if I'll go again that map and see now a full fledged orcs willage there, with high level Orc Boss(es) too, makes it really interesting.
I think a better AI makes its role only when mobs are grouped, but it drives back 90% of players. Facing a group of mobs who are in dps-interupter-healer config, it is troublesone to most players if the mobs use even only the bare minimum of AI, as healer healing (and not wanding the tank), dps trying to kill the support player and interrupter stopping the long-evil-cast-mage
About same MOB being more strong/skilled over time, it may be difficult for starters areas, if for some reason game not enough populated, the newcomers will become dead meat even in lv.1 fields
Overall, nice idea. Hope see something like that implemented.
Cheers^^
Yes, now imagine you go back again later on, after seeing these Orcs living there for a long time. They've survived assaults by players, and made a living there. There can be good reasons why these Orcs aren't simply wiped out like in other games. I believe strongly in worldly games where there's not such a big gap in abilities, more in a percentage system. So no level 80's easily wiping out level 10 Orcs. Also, if these Orcs used AI to build defensive structures, these could be quite formidable. And they could require expensive efforts in construction and time for sieges to overcome them, or a massive assault for an alternative. And maybe it's just not worth it at this point because of the location, and lack of threat.
At any rate, imagine you go back, and you've always had no problem avoiding or defeating their scouts before. But this time, you meet no scouts. You get closer and meet no resistance of any kind. And when close enough, you discover the ruined remains of their hold. You might explore the ruins, looking for loot or clues to what happened, and might find one or the other or both. You might discover:
That the Orcs were wiped out by a Dragon
That the Orcs left for some unknown reason, and you could try tracking them to their new home
That the Orcs fell victim to any number of things, other Orcs, disease (may mean some evil abouts), lack of food (why?), etc.
That the Orcs mysteriously dissappeared, no signs of leaving, but there's a big burnt spot on the ground in the center of a religious shrine of evil nature
The more AI, and the more abilities built into a game, the more it can be very interesting. On an ongoing basis, no need for development added.
I'm a strong believer in that you build things in a worldly nature. You don't build a canon in a quest that sits there waiting to be used for a specific purpose, and when fired, a particular wall falls down. You build canons that can be moved, placed, built, traded, and shot with.
Oh, edit to add, for starter areas. In a game like this, there are no static spawns. Players can drive the MOBs out of their community areas, making it safer for newbs. You can have reasons why players might do this, in the form of community social needs. New players help build things, increase tax coffers, etc.
Once upon a time....
This has nothing to do with pride. It's straightforward, logical criticism.
"Full of themselves" would be if I automatically rejected logical criticism of my ideas. One person's doing that here, and it isn't me.
"What is truly revealing is his implication that believing something to be true is the same as it being true. [continue]" -John Oliver
Having behavior evolve sounds very interesting. But all AI does is mimic real players.
Maybe even more interesting would be if the orcs behavior was controlled by players, in a kind of rts type fashion. By incorporating a completely different type of play style, you would appeal to more types of gamers within the same game and also achieve the diversity that is currently lacking in mob behavior.
I like the sound of the idea I must say.
Would do wonders for Sandbox games.
Irony.
--------
"Chemistry: 'We do stuff in lab that would be a felony in your garage.'"
The most awesomest after school special T-shirt:
Front: UNO Chemistry Club
Back: /\OH --> Bad Decisions
@Amaranthar
I think your "Genetic AI" concept is great (something I call Artificial Life). But do you have any programming experience?
You insist that it's easy to do, but how much about AI programming do you actually know? Have you ever tried programming AI?
It may be possible, but is definitely not easy.
Axehilt had some great observations, and was thinking like a programmer, not a dreamer. To truly appreciate how difficult your proposal actually is, you should try writing your own AI.
PS:
"Wonder" means to contemplate, "Wander" means to travel around aimlessly. I think you meant "wandering MOBs."
There's actually some random korean game that does just this(based on starship troopers). One person is in charge of the AI, everyone else tries to beat them.
Probably would be berated by "this" community as not an MMO if it was on this side of the ocean.
_______________
Personally I think the OP is going to the "acedemic" level of genetic AI where the purpose is to actually innovatively solve a problem. Then expecting the AI to be a little not AI like too life like.
If I were in charge of genetic AI(in an MMO) I would give it around 30 loadout option(individual skills/gear/stat buffs All with preset actions), then let the AI choose 5 from the 30. The choosen loadouts based on the genetic part of the AI. This means you get an AI that actually does change, and with some clever design(building the loadouts) adapt to most situations players toss at it over a few days to a month.
Practice doesn't make perfect, practice makes permanent.
"At one point technology meant making tech that could get to the moon, now it means making tech that could get you a taxi."
I like the idea but it is too soon for it IMO. I will give some reasons:
First is that casual mmo (WoW for example) players dont wanna it to be too hard, they like story, lore, quests, mobs to be easy/bit challenging (not all, but I am guessing plenty of casual players (have job/life, family etc)). They are playing and have only few hours to play, they want enjoy and relax in that time, they want linear and sure path to go by. From that comes that all WoW type (using just an example and dont ahve anything against/for it, but it has biggest playerbase, and plenty of games like it) games wouldnt want to use sophistacated AI as they DONT need it. That leaves us with sandbox games. Those are usually done by indies with low budgets and they have plenty of other things to work.
So after seeing that only sandboxes would use such system, think on what sandboxes need to improve before doing genetic AI. I instantly think about larger world, I would imagine 3-4x times bigger then darkfall. Reason for that is that many players still wouldnt grind up all mobs and players would be scattered. Also games should work on wildlife AI like Ryzom did, it is far from above good, but still it is start, but we havent seen it implemented it anywhere alse yet. And that would be just basics. Then you need to improve even basic AI senses to go with world. Somewhat normal visual range (not like now that you stand 36m from monster it doesnt see ya, but if it is 35m it attacks you), hearing, smell, visual blocks (so you can get next to it by going behind rocks, or crounching in tall grass, or just sitting behind tree and it comes by without noticing you). Another thing that would be done before AI is world places placement and creation. No more non-stop mob grinds in same place (I mean that they respawn in few meters area all the time), locations should be a lot larger, large enough not even to cover your visual range. Mobs should wander looking for challenges/fights/(food/drink)(generic ideas why mobs can be wandering)/even maybe adventures LOL. They should interact (fighting/grouping etc) with other mobs as well...... So somewhat to sum it up, world should be TRULY DYNAMIC. I use word "dynamic" for word "changing", and not like GW2 or Rifts trying to sell it. I mean if change happened it happened and there is normal chance of things to never be same as they were. For example: you exterminated huge settlement of humans, so for days/weeks when you come there you will find scavangers trying to find something to loot (may be humans, may be other races if there would be any), there would be chance that if someone truly put INSANELY HUGE ammount of gold and people provide enough raw mats for that gold it might be rebuild, if not after few more weeks/months it grow more and more into ruins, while nothing except few strongest/enduring things can tell that there was something there, and latter even that is gone...
Just throwing idea from the top of my head. "Genetic AI" is nice idea, but we wouldnt feel it much unless plenty of other things are introduced/improved.
Also very smart AI isnt much needed in how MMORPGs goes these days. Basicly MMORPG is synonime (to most of players) to "killing monsters". And when you look it is easy to see why, main point of most mmorpg are to kill monster so you get stronger and so you can kill EVEN MORE monsters. Killing to just kill... "Genetic AI" wouldnt be noticeble (maybe minimally) in such games. For players to appreciate they need to change themselves (small chance of that) and their objectives. Each encouter with enemy should be adventure in itself not that going through game and doesnt even feeling like you fought. Damn there is plenty of games which you can go through without dieing once and putting much though it in combat. Fighting is like breathing nowdays in MMORPGs... That comes very much from "Levels" in games. They give too much power too fast. You level to lvl5 and lvl1 mobs are no challenge to you, they can barely freaking scratch ya doing few damage/sec when you have hunders/thousands... That was just example from mind and not reference to any game btw... So genetic AI would also really work only in skill based games or with REALY REALLLLY reworked level system...
Take it easy, guys.
The problem here is communication. Axehilt's point is very valid, and is related to efficiency once you get that idea beyond text and into a game scenario. What I think Axehilt is missing is that this genetic AI affects mobs in the long run throughout multiple, possibly infinite situations. Sure, you can have various mob scripts for different types of terrain and so on, but there's a limited (still huge, but limited) number of unique situations that can arise from combat with those mobs in the terrain they belong to. If the AI is learning all the time and trying new things, then the number of unique situations grows exponentially.
Think of it like this: it's giving the AI a more human-like way of doing things. When you see a person doing something, you unconsciously know that there's a reason for the way in which he or she is doing it, but you don't know it. In current AI and games, if you see a mob doing something, you know it's just a script. If you see an orc wandering around a forest, you know it's not gonna react until you approach it at a certain, very precise distance. In general, they're utterly predictable. With this new type of AI, the possibilities rise, and even if you only see the orc moving left or wandering in the forest, you can't be exactly sure why. Under these circumstances you can't even be sure if it will actually attack you if you approach it a certain distance. Maybe it will run away. Maybe it will talk. Maybe it'll call its friends hidden in the trees and it turns out it's a trap.
Now, in a current game, the traps might surprise you once or twice. But once you're going through a replay or even just facing the same mobs again, that trap will stop surprising you, it will stop working because you know you have to bring a team, or a certain item, or whatever, because it will always be the same. This is what happens with RTS AI: once you've faced it enough times, you know it doesn't build X or Y, or send X or Y troops to attack, or whatever, and those mistakes are key to defeating it. In a multiplayer match of RTS, you generally don't know what to expect, unless, of course, the system has been figured out in detail enough to let the players use the same tactics every game. Regardless of that. in multiplayer matches you have to actively think about what's going on and what it is you're going to do, all the time. In MMOs this rarely happens, especially when there are no scripts in play as in instances and so on. You set up a 'winning combo' against X or Y mob group, and it's only a matter of rinse and repeat.
So I believe that's why it would be a good idea to implement, because it's not a mechanic that affects only one type of encounter - it's a mechanic that generates several types of encounters. The orc might move left, and that's all you'd see, but you wouldn't know for sure why, or if it will do it again. Games would be a lot more interactive, a lot more thoughtful, and not as reliant on visual cues as you're demonstrating they currently are.
And if we take it a little bit further, it may also be the solution to THE GRIND, haha. Grinding is just a mechanic operation, the rinse and repeat, no thought, no interactivity, just taking in the visual cues and pressing buttons in certain orders. If mobs were deeper than that, if we would actually have to question their motives and imagine the possibilities (is it a trap? is that orc friendly or aggressive?) then I believe the feeling of nothingness and chore that the grind produces would be finally overcome. I'm not a game programmer or anything, though, so I don't know, haha.
Great thread.
The real irony of it is that I'm a gamer saying I like this idea. And there are other gamers saying they like it too. But he's saying gamers don't want it because they can't see the underlying mechanic behind it, and he's a game programmer, and therefor gamers don't want it. Even though they say they do.
And yet, even though it appears that programmers are saying this might just be possible, the fact that I'm not a game programmer means that it shouldn't be done. Not even explored. It should just be dropped and forgotten. The main reason being that I'm not a game programmer.
Oh, and I misspelled a word, that too.
Once upon a time....