I'm sorry if this is off topic in Elemental forum, but with Brad posting a great deal on Elemental AI, I feel that it's is related.
A great presentation on AI http://www.youtube.com/watch?v=IJcuQQ1eWWI from the Civ4 designer.
I wonder what would be Brad's opinion on the points touched in the presentation. What kind of AI is Elemental AI compared to Civ4 AI? What are common? What are differences? What are the game differences that lead to differences in AI?
True. However, all the games in question are trying to present the same facade - an intelligent opponent doing something interesting while following a given set of loosely defined rules. Though they aren't the same rules of economics as the human, they're rules nevertheless. After all, attacking a target remains the same regardless of how you're doing it and supporting that unit through a system of constraints remains the same.
So while they work with different points of data, they work on the same sort of objectives.
So far we've mostly been hacking back and forth about Civ4 AI vs Elemental AI.
BUT,
As far as games go, what would be the top 3 gaming AIs?
GalCiv/GalCiv2 has a mention.
I'll put in a mention for the recent Paradox games. I'm amazed an AI can handle them at all (although maybe that's because I can scarcely handle them).
Third - how about Thief/Thief 2.
What's the criteria for the AI? Good? Fun? Smart? Appropriate? Are we trying to compare FPS versus 4X versus Stealth?
Nice video, im clueless on these kinds of things but it explained alot
I think an AI should be challenging. I say open to all genres, though if that's too broad, then limited to Strategy (4X, TBS, RTS).
I can't see why some people insist this is so complicated. Elemental has tactical battles, but the model doesn't make it that complex. There are no formations, concept of flanking or defence by direction. The pattern for melee units to move forward can be decided mathematically. Granted the formula will be difficult to optimise, but it's less complicated than many of the strategic decisions the Civ AI makes.
Unit customisation - where is the decision making? You build the strongest unit of each class: melee, ranged and mounted you can afford. The AI will be preprogrammed with the ratios it prefers for combined arms based on the tactical algorithms it uses, but there is no reason not to use the strongest possible units at reasonable cost in each class (with some marginal utility applied so it doesn't spend 12 materials to turn a 30 defence unit into 31). There is no advantage to mixing your melee group with some 20 attack and some 26 attack units.
Creature economy - again nothing special. Hunting down creatures for gold is the same as hunting down barbarians to stop them pillaging your lands. It's even easier because the creatures don't enter your territory so you go after them when you want to, rather than having to be proactive. The decision is simple, the AI decides how much military it needs for expansion and how much for the home guard. Whatever resources is left over can be used to create hunting stacks. The composition of the stacks will be based on the strongest monsters that are currently roaming. How many stacks it builds depends on how much resources it has - it will always build a stack of a given minimum strength. These stacks then patrol a given region, or just randomly move around the unclaimed territory while staying close to friendly ground. They come back to heal and reinforce when damaged beyond a threshold.
Questing - this keeps being brought up as being complicated. It's not. You pop a quest location, it tells you to either go to another location, fight the monsters and bring back a goodie for reward, or to escort someone or something to another location for reward. Hunting stacks can be used to pop the quests. There's no time limit or anything special that needs doing, so if the target isn't too far away, it goes after it. The hunting stack that normally just patrols temporarily has a destination and it avoids monsters till it has done the quest.
Building - much less difficult than optimum city placement in Civ. Not even the avid players can agree on the best system for that, because terrain is so important. There is no special terrain in Elemental, only strategy resource which you should grab. You don't have to do city specialisation - it's not hard to realise you only need a market if you have a gold mine, or a school if you have a library. Otherwise, just expand the city in the direction with the most space. If you get a few conflicts, it's no big deal, not every city needs to grow beyond level 2 - in fact they shouldn't.
Strategic spells - much less complex than the strategic decisions for a Civ AI. Everything that AI has to do is strategic, because every decision has trade offs that are hard to quantify. Choosing civics based on current plans, setting the gold/science/culture/espionage ratios, deciding when to stop a city's growth to train a settler or worker... all these choices are hard enough for the player to make, they are even more difficult for the AI. Currently in Elemental, as a player your decisions are easy, because there is no penalty for making any decision. Essentially you can do everything, build everything, research everything - mostly because, due to the lack of penalties for expansion, your resource gathering grows exponentially. The AI can chose its order of research based on current priorities, just like a human. Currently, adding penalties for expansion would be a bad thing, because there is very little to do in the game as a builder and expanding is fast as you can is the only thing satisfying.
RPG - you didn't mention it but it keeps coming up. Heroes level up and you distribute some stats. Again, easy. The AI simply assigns a class to each hero. Spell Caster, Generic Melee, Generic Ranged, City Attacker, Hunter, Imbue Guy (this would be the Sovereign). The number of each the AI uses would depend on its falvour and play style. Each time one levels, the stats chosen are weighted towards improving the hero's effectiveness in its class, with a bit of randomness for variation. This is the same as how the Civ AI decides which promotions to give units (and in my opinion, deciding between extra jungle defence vs hills defence is more complicated than +2 strength vs +2 dexterity).
I'm not saying the AI is easy to make for any old programmer. It's still a very difficult job and needs a specialist. But based on how difficult it is as a player to "figure out" the ideal strategy for Elemental (1 game - maybe 2) vs Civ (years), I can safely assume that those strategic decisions that are hard for the player in Civ are even harder for the AI.
I know, it's laughable falconne...
The elemental fanboys are out in this thread and for some reason they have a bug up their ass that Elemental has to 'beat' civ for whatever stupid reason.
Elemental probably will be more complex than civ at some point, of course that's not necessarily a good thing either. So to have a 'good' AI will be more difficult, but honestly, I doubt elemental will have a 'good' AI. It's not as though the AI in GC didn't cheat as well, and playing on whatever the 'normal' difficulty was was a trivial exercise in boredom. Yeah, you can get a bad start and be screwed, but that's no different than in any game really. Get a decent or good start and crushing the galciv AIs was as easy as beating up on any other game.
I admire Brad for his desire to make strong AIs, but he limits his ability to do this by making games with too much complexity for his AIs to deal with. From what I've seen Elemental will be no different than GC, unless the AI is open to modders.
Do not use the "F" word lightly. Even if I concour with falconne, I'd say that is the fastest way to start a flame war.
As for me, I'm simply shocked by the amount of people that posted WITHOUT reading the thread. Who has ever said that one AI was better than the other? The intention was never to compare the two AI, but [again] the COMPLEXITY of both games.
And btw, no one will EVER convince me that Elemental, in it's CURRENT state, is more complex than Civ. Because it is not.
Will it be? I do not know... It's not even important, complexity is not gameplay. I like Elemental and I'm sticking with it.
Yes it's annoying many people are just skimming the thread and assuming it's a question of what game has the best AI. If we were talking about good AIs, then neither Elemental nor Civ would even be a contender in such a discussion.
Much of the annoyance for me has been the disappointment with the gameplay in Elemental. This being Stardock, I was hoping for a game with the complexity of EU III, or Hearts of Iron 3, or Victoria II, but with the presentation and approachability of Civ. With all the hype going on about how it would be the best TBS game ever, I think that's justified. I could have dealt with bugs and gameplay issues and all that, I'm used to them in every game.
Except we got something which is, right now, an extremely simple and casual game. A sort of lightweight puzzle adventure. Then I realised it's supposed to be a platform to be built upon, so it will be a year or more before enough depth can be added. I'm fine with that now... it would be interesting watching it grow from these roots. Being able to make and comment on suggestions that make it into the game will be awesome.
However, it's strange to me how a strategy gamer can find any depth in the game right now, let alone say it's more complex than Civ.
Actually this is to do with AI skill, not game complexity. There's a difference between the time you know you've won and the time you actually win. Civ is too complex for the AI after the industrial revolution, which happens 100-200 turns in. When you get there your production increases exponentially and if no AI has got there long before you, you know you'll most likely win. It's several hundred turns later that you actually get into the endgame. However, in that first couple of hundred turns, you can actually lose.
Currently in Elemental, the scale goes down to 10%. If in the first 10 turns you don't lose your sovereign to several extremely unlikely dice rolls against a wolf, you know you're going to win. In effect, you know you'll win before you even start. Two hundred turns is what you need to reach the end game mop up.
I always stop playing strategy games when I know I'm going to win. Despite the AI not being clever enough in Civ, the complexity of the game means I can play at least till reaching the modern era before knowing for sure I'll win, at which point it's no longer fun. At this point, Elemental doesn't offer enough depth that requires you to experiment or go by trial an error at the start to figure out how to win. You know exactly how to win and it's one linear path every single game, which is what annoys me. You don't get the thrill that's there at the start of a Civ game, as you explore your surrounding territory and you start to see how the terrain is going to affect your early to mid game. You formulate an entirely unique expansion strategy based on the terrain and neighbours you discover. In Elemental, the terrain doesn't matter, creating new cities isn't a mammoth effort like in Civ and the first neighbour you discover - will be dead in about 20 turns.
Pal, you and I are gonna get along nicely !
I too feel that map and terrain have too little value in Elemental. In Civ it's ultimately the map that dictates your strategy! While I'm all for a "unique" gameplay in Elemental, a little more importance on terrain would do a lot to restore that "Exploration X" that Elemental, being a 4X game is lacking ATM.
honestly, i think elemental features one of the worst AI i've ever played in a 4X game. it's really laughable. it doesn't expand, barely builds formidable armies, his sov is wandering aimlessly around the battlefield without leveling up... it barely exists. the fact that i can steamroll the 7 "ridiculous" AIs in about 200 turns speaks for itself.
other than that, i think elemental is also a very shallow 4X game. it's supposedly a tile-based game, yet stardock didn't nail down the very fundamental concepts of tile-based games. besides some scarce strategic resources (gold mines, material mines, farms...) almost all the other tiles on the map are just blank. no features, no production capabilities, no yield... and the whole city management model is just too simplistic and shalllow. this is not how you make a strategic tile based game. i think it's lightyears behind civ. not civ 4 mind you, but even civ 1 or 2.
honestly, i think this game should be redesigned from scratch. i've been waiting for it since the initiial announcement, and it's undoubtly my most disappointing game this year.
Great talk. I especially found interesting the division of players into 3 groups, Challenge, Sandbox, and Narrative, and how he treated those 3 groups equally, trying to cater to their individual needs and treating their specific concerns as valid ones.
I do not believe I did. We know who they are, and once they showed up the original point of this thread was going to be swamped by pointless and occasionally nonsensical attacks on Civ. Even if the merit of some criticisms leveled at the Civ AI is believable, the idiotic comparisons to Elemental show a clear bias which undermines the initial intent. Now you will note that I have not and will not be defending the civ AI.
If you find me speaking too bluntly then so be it. I do not need to suffer fanboys lightly.
See here is the problem.. When one break out the Fanboy or hater one looses an element of credibility.. Any points one makes can become a second class citizens to the fact we a labeled people who disagree with what can be viewed as a derogatory term... (i like both Civ and Elemental and am firmly in the camp that we can learn for other games successes. by the way) .. I am not trying to start crap in fact a good few of things posted I support..
My only point is when we stop calling each other fanbois and haters or say someones opinion is stupid just because we cant understand it or disagree with it we will have better communication..
It proven by how many threads devolve into arguments between a few people who have insulted each other some how and the valid points of the thread often suffer..
Back on track for me at least.. Holding Element up the standards of other well established 4x game is a tad lopsided.. However pointing out what 4x "y" did better, we liked more, or had that Elemental lacks, only provides solid ideas for debate.. and IMHO helps Elemental and us the players in the end..
I just wanted to point out that you'll get a bigger crowd by being kind. I actually am with you in many of the point you made about the game, but "fanboy" is an offense, or as such it is perceived. I couldn't care less if you're blunt or not but in doing so, you're shooting your own feet.
If you don't care about that, by all means, feel free to ignore my words! And please, do not take this as a personal insult!
Peace!
None of the CIV's had "good" AI. They were all easily beatable, as is Elemental's. It's hard to think of a game in recent memory that had "good" AI.
SSG (many years ago, 1980s) used to build their games around the AI. Meaning, if the AI could not handle the game mechanic it was not put in the game.
The sad truth is AI does not sell games; graphics, interface, gameplay (and of course adverstising) sell games.
Most players will never discover the game has poor AI, as they will stop playing the game long before that happens. Most players will never visit a game's website, or forums either. In fact, it has been my experience that most game buying consumers will not ever patch their games, unless it is done automatically for them via the game software.
The strength of Elemental is it will continuously get better for years to come. At present it is a sandbox game. You can play it, there is AI, but it functions at a very basic level.
I find it interesting how the entire game development paradigm changed once the Internet became widely available. There were VERY complex war, strategy and RPGs released throughout the 80s. They worked out of the box, there were no patches.
I apologize if I have strayed off topic or hijacked this thread in anyway.
Cheers!
Wow....just wow. So, if someone disagrees with your opinion, that person is a fanboy? If someone says that the Civ4 gameplay mechanics are simple compared to the core [unpolished] gameplay mechanics of EWoM, that person is a fanboy or that person is just posting his/her subjective opinion? Take a look at my #2. note in my first reply. I will translate it for you -> complexity of the Dominions 3. gameplay mechanics > EWoM's > Civ4's. -> Which is why I said, that writing a decent AI for Doms 3. would be the hardest imo. This is my opinion, and it won't make me a Dominions 3. or EWoM fanboy at all. Am I wrong? You should chill out.
Oookaay... now that we've all had a chance to vent some steam shall we get the thread vaguely on topic, like some ideas on how the AI can be improved so we'd have some things to try out when the API is released? Or at least, such ideas would indicate how detailed the initial Python API should be.
Obviously I believe the game's mechanics are simple, but even if you disagree, I'm pretty sure most people here can steamroll the game at its most difficult inside of two or three hundred turns and you don't exactly have to be that creative to do it. This implies we've all figured out some simple rules to maximise output.
In reply 57 I mentioned a few rules around mechanics that are claimed to be overly complex. Perhaps others have some further (or better) ideas for solving those mechanics?
I think there's confusion over the word 'complex.' If a game has really only one system you need to manage, but it's a difficult system that can be dealt with in different ways and is possible to completely screw it up if you're not cautious, is that a complex game? If a game has many relatively simple systems that one must balance, sometimes prioritizing one over the other, sometimes vice versa, is that a complex game? I think from an AI perspective, the first is much easier - computers are nothing if not amazing at crunching numbers and figuring out how accomplish one goal optimally (i.e. how do I generate the most gold from my cities?). And from a human perspective, the second feels much easier - we do better at balancing different priorities that cannot be directly compared in a numerical way (i.e. should I spend limited resources developing magic or economy?). Anyway my point is that it's impossible to compare the "complexity" of two such games, since the difficulty of dealing with that complexity varies between human and AI, and even from one human to another or one AI to another.
I think the current problem with Elemental - and the reason so many seem to dismiss its "complexity" out of hand - is that while it has the second sort of complexity, i.e. many individually simple systems that need to be juggled at once, balance issues currently render this complexity meaningless. You can choose between many different priorities - gearing up champions, developing magic, building regular units, developing your city economy, and so on - and these priorities share limited resources (and I don't just mean gold, materials, etc. - research is also a limited resource one must choose how to "spend"), forcing you to make difficult choices. The problem is the balance and AI is so horrible that the difficult choice is trivialized. You only really have one viable way to develop your empire if you want to win, which is far better than the alternative due to balance problems.. and even if you don't make the obvious best choice, you can still win easily against the current AI. So we have two camps - the realists that judge Elemental as it currently is, and justly deem it simple, and the idealists that see the underlying complexity of the game mechanics, even if balance issues currently ruin that complexity.
I'd rather not join either camp, the only difference is point of view - how you define complexity, and how much you credit potential complexity that's tarnished by superficial flaws like numerical balance, something we're not likely to reach any consensus on. Let's just say that Elemental in its current state isn't fit to be compared to a polished, perfected game with years of expansions and patches, and we all have high hopes for where Elemental will be following an equal amount of patching/expansions.
Exactly. Note: The problem isn't only balance, but many of those "simple systems" are far from being polished enough.
Perhaps I haven't played the same game as some former posters, but I never found strategies in Civ to be endlessly evolving.
My starting position and my civilization choice determined my strategy, be it Slingshot, Great Profit(Prophet), etc... and from there flowed the game. Every several games exploration might have changed my strategy (oh balls, I'm sandwiched between Montezuma and Tokugowa - no Great Pyramid for me), but most games I knew what the first 50-100 turns would bring from turn 1.
On Emperor difficulty and above, one early mistake means destruction. Hell, even doing everything perfect still meant destruction if the situation was unfavorable. Moreover cheating on higher difficulties (two archers) made rushing impossible. AI cheating is all well and good, but winning against a cheating AI doesn't provide the same satisfaction as winning against an adaptive AI.
Digression:
The point I took from the OP's questions (and subsequent posters) is that this is a discussion on what we can do to create a superior AI. How do the two games compare - which has been admirably investigated. Inevitably opinions will arise over which is better, or which has the greater potential. Dismissing opinions of those who favor Elemental as "fanboy" rubbish is wretchedly ignorant; opinions from every side of a conversation are crucial to answering the questions at hand.
I think Elemental's AI has greater potential over Civ's because the community can be involved in making it - it's not just Soren Johnson or Brad Wardell writing the bloody thing - it's all of us!
I like Civ4 a lot - but its AI is limited. Elemental's AI can surmount such limitations so long as discussions like these continue.
Unfortunately, at this point, nothing has actually been discussed. Sure, a lot of words have been said, but they've all amounted to "This game and that game have poor AI."
There are many great features available to you once you register, including:
Sign in or Create Account
