Howdy, Stranger!
It looks like you're new here. If you want to get involved, click one of these buttons!
Quick Links
Utility-Based AI and EverQuest Next
TheRealFantomex
Member Posts: 36
I wrote a new article about utility-based AI and EverQuest Next. I am pretty certain this is the type of AI that is being implemented in EverQuest Next. It explains a lot about what they have said so far will be possible. I hope you all enjoy it. Let me know what questions you might have.
EverQuest Nexus - http://www.eqnexus.com
Head of Social Media and part time writer/interviewer.
Comments
Very good read, thanks for taking the time to write it. I am very excited for the game for the reasons you explained in the article. This next step is something i have been waiting years for and to finally be on the verge of having a game that is able to hopefully have to balls to release it...such an exciting time.
Dear OP..
KUDO's on a well written article explaining your position on AI.. However..... Regardless what you call the type of AI, from a players perspective the combat encounter will either be controlled or random (aka chaotic).. Each have their own pros and cons, depending on your preference of play.. Using the common example of class roles vs. smart AI.. If we are using roles, each person has restrictive duties to fill, whether it be tanking, crowd control, healing, DPS, etc etc.. With this form of combat the encounter is often controlled by the players, such as mezing, taunting.. The players have the ability to control the fight of who is killed, and in what order.. (normally).. There is nothing wrong with this type of play.. Professional sports have used roles for decades..
Then we have smart AI, where the MOB is so called in control of it's actions.. At this point, the players are unable to dictate the fight.. Instead the players must go into "response" mode.. This is more chaotic then what most want to accept or admit.. Given the fact that a smart AI will sometimes target soft targets (isn't that their purpose?), you either have to make the soft target durable enough to withstand a beating, or give them abilities to avoid damage.. What you normally end up with as this point are classes that are all able to mitigate damage, and often the best playstyle once that happens is zerg.. Except a few encounters that devs might require more strategy, but this will have to be isolated (instanced?).. Once the combat is out in the open, it's the players vs. the mobs as it is in GW2 and other games..
The 3 mobs too smart to defeat by 2 players, go get some buddies and make it a 5 on 3 fight then.. ZERG.. Power of numbers will always trump smart AI.. But that doesn't mean that smart AI is wrong either.. They are just 2 different styles , hamburger or hot dog.... I prefer burgers myself..
IMO the smart AI is to simulate PvP type of mechanics, which also encompasses MOBA games.. I don't enjoy that type of gaming, it's why I never played Call of Duty, and don't care for Planetside or LoL.. For me a trinity style is like playing football, whereas smart AI is like playing rugby..
NCAA football for the win... I hope Johnny Football wins the Heisman again, just to upset the critics.. LOL
A good read, but the dragon example makes me worry.
So if you have the same dragon and the same party, how much can a fight differ from the previous one? Unless it is total randomnes and dice throwing, which is a bad thing and not really AI, it will be pretty much identical.So in other words, what happens now:
You attack monster with a learnt pattern, you are ready for it's every predefined response and using the same tested party you farm that monster.
How it would look with the new AI:
You attack monster with a learnt pattern, you are ready for it's every predefined response and using the same tested party you farm that monster.
...
Since the idea is the same, monster still has a set of predictable behaviours that may differ a little bit from time to time, and by approaching the monster you learn those patterns and based on that you form a party that is most efficient, to reduce potential changes in behaviour and problems, and it is pretty much the same system only with a new name and nice technical explanation for marketing.
Explain where I'm wrong, please.
How about factors that can fluctuate?
How about the Dragon knows where other mobs are locally and can bring them into the fight?
How about variable skills they can bring to the fight?
How about the Dragons desire to fight based on vicinity to what they call 'home'?
How about variations in a fight based on how recently they ate?
How about variations based on their exposure to magic or combat from Dragon to Dragon (based on experience from a previous fight)?
There are literally hundreds of things that could be added to create variables to change a fight so the main thing to consider is just how many can be added to each mob without slowing the game down with too much info or calculations.
Let me think about this...
discussions about AI made by 2 guys who do it for a living...or you...a biased detractor without anything positive to say about the game....
Hint - You lose.
This reminds me of GW2, where people heard about dynamic events and started describing how live and amazing the world will be, but when the actual game released, it wasn't as big step as everyone said. And I see the same here, where people start to bring out the entire AI mechanics that professional companies research, while lets be honest, everyone knows that only a tiny procent of those can be added to a game. To many calculations as you said yourself.
So my worries still remain. If you look at it realistically, and base it on previous games expectations/reality ratio.
To respond to your reply, factors fluctuation depends on how many will be taken into consideration by monster, which we don't know yet. Monster calling allies is just a new name, for a simple minion summoning alogirtm that is in plenty of games now and is a predictable mechanic.
What my point is. I think that given the technological limitations we currently have as gamers, the only AI we can bring into a game world is simple enough, that it could be emulated with simple algorithms that were used in games up to this point. So I think that the emergent AI thing is just great marketing made by great company to fool people into thinking they will get something new.
Interesting article but i see 2 issues. First you didn't consider possible mob-emotions. Things like being enraged make a huge difference and in case of a dragon it could be arrogance or pride that change the the dragons decissions to be much different than your thoughts. Second issue is the dragon itself as chhoosing one of the most intelligent creatures isn't the best thing to do. A creature that is wild or more driven by instincts will react much different and can't analyse the combat as you described.
Of course all my issues could be integrated into the AI. Actualy storybricks are supposed to do something like this but people need to realize that as well and don't turn the new AI into every mob is a genius and knows all game mechanics thing.
Forgive us about being sceptical here but the proof is in the pudding and whilst I'm looking forward to EQN based on what I seen and what the dervs have said, what is worrying me is what the dervs have not said or seem to be shying away from that bothers me.
Asbo
I personally see no link whatever to how GW2 worked. GW2 AI was your standard fare, nothing remarkable about it.
But what is being discussed here a different way of doing it.
Let me give you an example.
A Mob gets hungry. Is it AI if, after x minutes the mob performs the 'eat food' animation?
How about if certain creatures like deer or squirrel, plants like apple tree or strawberry bush or farmyard crops are marked with the label food. And a hungry mob, when getting close to a location or item labelled food checks to see if they are hungry and if they are proceed to perform the eat food animation?
What if that mob now has selective likes and dislikes for food. So an Orc doesn't stop where there are Strawberries but will certainly hunt deer.
Take it a step further and what if you encounter an Orc in the wild, who is hungry and who also spots that you have some deer on you. Now the Orc has an additional thing to consider, to satisfy it's hunger or not. So you might fight 10 Orcs but find that the 10th Orc, simply because it is hungry, now acts differently because it has decided to risk attacking you to try and take you food.
When you transpose that kind of structure into combat you have a lot more options. And the complexity of each fight, in fact the complexity of each individual mob in each fight, can be created to react totally different simply by scaling the number of possible choices available with the respective IQ of the mob. And you could also have smarter mobs then the norm, with more options, eventually becoming boss mobs if they survive.
Of course till release we will not know if they have hit the mark but I certainly think they have made the correct choice in how to deal with AI at last.
Did they say anything at all about what you write here being in EQN, or are you just making it up yourself after reading about the utility-based AI. Because if it is just you, I will say once more, I don't believe any of this will be implemented in the actual game unless I hear it from the company itself, so you can say all you want about wonders of AI theory. Try to approach this more realistically, than dragon looking for food, because I think it is not gonna happen anytime in the next few years.
And I say it might be the same as in GW2, because of the nature of the situation, I'm not talking about AI in particular. People speculated about wonders of a mechanic, while in the end it had to be dumbed down and much simpler in order to work and it is probably the same in this case, so there is link.
I want to talk about this new AI more realistically, how can it differ from current solutions, given that it is a game for home PC. They can't make a true AI in a game, so they will for sure try to emulate it, and I can't see how is that different than current solutions. It might be more complicated and rich, but still the same way with a new name, and when people find out, it will be a disappointment again.
For me more realistic apporach, not hyped looks like this. You will have a very simple tag system, where there will be few factors for every fight, like player classes. Monster will have few different actions that trigger on certain tag combinations. Simple system, not AI, but has feeling of it. I would like for this game interest to be slightly increasing with time and reveal, and not drop drastically after hype, but unfortunately this is not possible, as people love to speculate and dream.
perhaps they will run or call for help when they see the zerg coming?
In some games they simply boost spawn. It's really no different than that, except that "calls" for nearby MOBs offer a chance for the players to have a final victory.
I think the whole point is to make a more "realistic" simulation, and force players to use tactics rather than simply overrun the "rooms". Players will be trying different strategies, like "training" or hit and runs. Would love some grenades for this. Especially if the MOBs can use them too.
Once upon a time....
from the sounds of it mobs don't spawn. they are released. You may find that all orcs originate from crushbone, all gnolls from splitpaw and all goblins from runnyeye. This would make these places extremely dangerous even for an army of players.
personally I hope they use the monster generator strategy where there are a couple of places that enemies come from and spread out over the land. Places you just cant walk into and survive.
I would also like to see a recall behavior when crushbone is overrun by players all orcs in the world head back there to retake it.
Yes. Blowing booming horns to recall all the orcs, spreading the word, so to speak, with horn blast to horn blast. A summons to return to the homeland for defense. And players could try to cut off returning hoards on their way back. Or force such a recall in order to achieve some other goal in a recently departed area, against whatever remains.
The larger scale tactics could be a lot of fun in a war setting style of strategies.
Once upon a time....
Very interesting!
I hope DEVS are able to take good advantages from Utility-Based AI and release to us a fresh, fun and smart new game.
I don't want to see another "more of the same" mmorpg.
All we can do is keep our fingers crossed...
Thanks for all of the discussion everyone. I do want to remind everyone that you are right I do not have an insider with information, this is an opinion, one that I personally think is correct in terms of the system they are using, but still an opinion. Utility-based AI is something Dave Mark knows a LOT about and also the Storybricks team, so why bring them into the game if you wanted to use a different AI? You wouldn't. That is a big reason why I think this is article is correct in the type of AI being used. The other part is that it does account for a lot of what we have been told will happen, where other AI systems would be too bulky and risky to use.
A big thing to mention is that while this article was largely based on combat, this AI applies to everything else in the game where AI is involved, not just combat. Things mentioned are mob spawns, economy, ecology, social interactions, etc. So think broadly about this system.
There are some valid concerns for combat though. The first thing I would say is read the article and watch the two GDC conferences from Dave Mark. I think this will help calm some of those concerns.
The combat example I used was one of a very simplified encounter with an intelligent monster so please do not think this to mean every encounter works in this fashion. They won't. We know that they have said AI can be scaled to change the difficulty of the encounter. And with this system it can. For harder encounters you give the monster more tags to read and more abilities to chose from. For simple encounters you remove a bulk of that intelligence, other than "hey, something is hitting me, I should hit back." It is a very broad spectrum that could be used here.
For the concern of controlled vs random. That is an all or none statement. There will be plenty of room in the middle. You just have to understand how the AI can work and be adjusted. The use of the statement about controlled being class roles that need to be filled. Well they will still need to be filled, but with differences in strategy of how they will be used. This is partially covered by the introduction of a dual class system. You can mix and match skills that help you control the encounters a little better. However, there will need to be an element of responsiveness as well as was mentioned because the added catch here is that the monster could be changing its strategy and shouldn't it? This is not at all entirely random either. Is there room for some "Did not expect that to happen but it makes sense?" Yes, there is, but it is still within the control of the devs based on what tags they want to put into the encounter. Again, read that article and watch those presentations. That concerns is actually directly commented on by Dave Mark.
The concern about, well if it is the same party and the same mob it will be the same encounter, is understandable. But you have to take into account that first the mob AI level of intelligence will not be the same across the board. Some will be smarter, some dumber. Second, there are all kinds of factors here, things you would not always think about. What if your healer was standing 5 yards closer or 7 yards further to the side? What if the damage dealers slightly changed their timing or pattern on accident? What if you are wearing a new sword that you won from the first time you encountered this monster instead of your previous one? There are a lot of factors here that can change encounters tremendously. Now, what level of change will the devs put in place? I have no idea, I am just talking about what is possible. They might not want to make encounters that difficult. But they have said they do not want people to be able to go to a wiki or watch a video to know exactly what to do in every situation.
The comment about mob emotion. I love it. You are thinking 100% as I am however like I said the article I did not want to write a short book about everything possible. You are correct here. This can and likely will play into what happens. This dragon may be more determined to fight if you are knocking at its door than if you met it 300 miles away on a random mountain side.
Again you are correct, will the devs confirm this anytime soon? Heck, no, not likely at all. But actions often speak louder than words and why go outside of your company to contract with people who specialize in this area if you were not going to use this AI?
The comparison to GW2 is I think unfair. To the best of my knowledge they did not use an AI anything near like a utility-based AI and instead used the same as previous trying to fit it into dynamic content. A lot like fitting a circle into a square hole. Actually, I do not believe any MMO on the market has ever used this type of AI system in their game.
EverQuest Nexus - http://www.eqnexus.com
Head of Social Media and part time writer/interviewer.
There was a brief comment (unless you watched every presentation you might have missed it) where they mention the fact that everything in the game can be tagged. So things you tag can be used as hooks to trigger possible decision making processes. So yes, it has been mentioned. My example of how it would work was simply my own personal interpretation.
Well, as I mentioned above they have mentioned precisely that so I am being realistic.
I disagree...personal opinion so leave it at that.
The difference is that there are numerous triggers that can prompt various reactions to situations. Without getting into the nitty gritty of how the calculations are done it is simply a change from having a list of scripted, if when statements to a list of multiple optional when possibilities. The more possibilities available the greater variety in each scenario and the greater perceived AI.
Again, your personal opinion. I disagree based on the research shown and my own personal research.
Define what AI is...
Actually define what the 'I' is.
We Humans have huge lists of things to consider for every action we take. If you think the human mind is more complex then an insane number of potential actions based of an insane number of choice based decisions then of course you wouldn't see it.
Human intelligence is, when you get down to it, just a very complex form of AI. Just a series of on/off electrical impulses...beyond our ability to replicate of course but not impossible. To provide a computer game with an AI that will make a big difference is not that hard. The hard part is making the AI dumb enough for players to have fun and getting the dumb players to actually accept it. As we all know, Humans are not very good at accepting new stuff or change...just look at these forums for the level of acceptance given to any form of innovation....the first month of posts about anything new is negative, fearful, reactionary and ultimately wrong.
Dave Mark responded to your article with "Well done, Sir." I'd say that's a pretty good clue you're on the right track
I had watched pretty well all the video presentations you linked and used as the basis for your article a few weeks ago, triggered by a discussion here about the trinity.
You did a great job of explaining it in your article.
Here's another video that you may not be aware of from 2009 with some of the basic principles and motivation behind the decisions to focus on changing the AI: http://www.youtube.com/watch?v=w7pktyeFFic
“Microtransactions? In a single player role-playing game? Are you nuts?”
― CD PROJEKT RED
*shrugh* from the point of view of an AI programmer (which I was, once upon a time, I even wrote my diploma about it.) the article is full of wonderful gibberish. The language of the article... Well, the same language I've used to explain to my elderly grandmother why the computer in our house will not become alive on Y2K error and will not eat her in her sleep.
The dragon decides... Assigns percentage numbers... arrrgh... Well, it's a way to explain the AI to a slightly retarded child, I guess, but really, do people need that? The concept of "utility-based AI" was known, like, forever, under different names, when I was writing my diploma (what was it? Nineties? I don't remember already...) it was called, I think, "goal-oriented fuzzy logic" - but the point remains the same - actor evaluates the changes in the system, accordingly to the set of flexible weights, which are calculated by the rules programmed into the AI, and changes its behavioral state accordingly. It is indeed an emergent AI, fully capable of "evaluating" all those factors described, as long as programmers have programmed the rules accordingly, it's very hard to predict the behavior of such AI (the number of rules, influencing the weights dynamically may be extremely huge and constantly changing depending on the system), but there is absolutely nothing new or exciting in that.
You still need to programm and test the AI, and it's extremely human-time-consuming, for emergent AI may (and quite often does) emerge either a complete idiot or a boring genius (for example, if, by the rules of the game, the most profitable move in every situation would be a basic attack, all the computer will ever do is that basic attack, and that's it.), so you need constant human tweaking for it.
Nothing new here.
As for StoryBricks, as I've said already, they aren't even AI, they are INTERFACE, the way to programm AI with your mouse, not within the code; it doesn't add anything to actual AI, or do anything, the only thing it does - allowing to move in-world AI programming from expensive programmers to cheap mouse-clickers, probably outsorced third world, thus allowing larger quantity in AI rules, behaviors and scenarios. Especially scenarios. Thus StoryBricks.
*shrug* so far, nothing exciting at all.
Gee, I wish you were making MMOs so we could continue down the path of predictable boredom.
Once upon a time....
The concepts may not be "new" but can you point me to a mmorpg that has them already or to the scale at which EQN may?
As with all things tech, I'm sure they are talked about and imagined long before anyone gets their hands on the finished product. Why you wouldn't be excited to see in action something you knew about 20 years ago is beyond me.
Thank you for this article, very nice, I have been studying and talking about utility AI for MMORPGs for years now, even put together a design of an MMORPG using it. If you are correct about this, it will change the way MMORPGs are created. I have no doubt about that because the AI and the interaction within the game wil change across the board.
Zerg is only beneficial if players are rewarded for doing so.
Simple solution, every mob has a scale or number of players that are expected to be able to defeat it. Say a giant is expected to need 20 players to take it down safely. Less can do it, but 20 is the average.
If 50 people roll by and take it down in half the time, they are given no reward. The game would know that 50 players either attacked or healed or did whatever contribution in the fight and that it wasn't "fair."
This would most likely discourage those that plan on Zerging for progress. If people just want to be able to say they killed a giant, well let them bring 200 friends along and do it quick if they want, just don't reward them.
This still works with a system that frowns on kill stealing or other mob tagging issues from previous games.
I'm sure devs could think up even more sophisticated systems to keep everything in check. Unlike GW2 (at least when I played) where you only had to do a small amount of damage to get credit among a huge zerg of players. Not to mention spawns are so large and close together that even if you can't tag half the mobs, you still are able to tag the other half and do well in progression.
With the non-static spawns, camping and zerging should also run into some issues in EQN.
None of this is accounting for AI and all the systems that can be in place. Just simple, Zerg = no reward. But AI could alter and become harder or whatever to account for more players to make it a waste of time to attempt to zerg.
Personally, I find PVP and games that rely on creative active teamwork to be more like Football then the Trinity. Both require roles and everyone to do their part, but one allows for a lot more possibilities.
In Football, a QB throws the ball to someone to score. If the moment allows, they can also run for a TD themselves. In the trinity, you can't have a Healer all of a sudden be the Tank or the DPS turn into a Healer. Everyone is stuck doing a very specific role (unless you are some hybrid class). Football has a decent amount of flexibility, even though everyone is a particular role to do.
EQN will still have distinct roles, but there should be a bit more freedom due to class skill mixing and allowing everyone to be the QB, TE, WR, or LB depending on the situation. Instead of 5 friends picking the QB class on Day 1 and unable to do anything together because no one picked the Kicker and this first fight needs a Kicker.