People say this but none of the games of theirs I've played yet have "units" in the sense of things that have to be pathfind and be arranged tactically, which is what is difficult for the 5/6 AI (the old AI can't use 1UPT argument basically). Or does Stellaris? I haven't played that yet.
Can we do anything to make Firaxis devote more resources to improving AI?
- Thread starter Krajzen
- Start date
Usama84
Chieftain
Let's face it people, Civ is not just what it used to be. The AI was never perfect, but at least it knew how to play the basic game. Civ 6 AI is a joke... diplomacy makes absolutely no sense in this game. They make peace, and the next turn, they declare war. They don't know how to use air units??? Are you kidding me? I wouldn't know, because I lose interest in every game before I even make it to the gunpowder era. The civ fanatics like us can keep moaning and complaining, but as long as we are stupid enough to buy these unfinished products, they will just keep selling them and laughing all the way to the bank. I woudln't hold my breath for them to address the shortfalls of this game... I see very little difference besides a few bug fixes since this game came out over a year ago. Civ 6 is the last civ game I will waste my money on.
Ondolindë
Emperor
Well, I think that this very thread is a good thing to do. We may not be the majority, but we certainly are interested and are, as a whole, probably more knowledgeable and invested in the game than the majority. If what Ed Beach said is true about following this and other forums, and listening to them, and there has been some evidence to that, then let's keep the conversation going. Organizations can entertain self-defeating behavior without doing anything about it. Not paying attention to the complaints of the AI is an indication of that. Firaxis should pay attention to this group, a section of their client base because if they don't the brand will slowly lose a valuable perspective they cannot hire or hire for cheap, at least. I am seeing an increasing amount of people becoming disenchanted with Civ VI and they are voicing that here and in other places. For one, I won't buy their products blindly before understanding the good and the bad in them. I think I deserve to get a great product in each release, but we as customers promote such behavior by not voicing and being critical, even withholding our money to make a point. Now, I believe that the reason why they added two more DLC to the Deluxe portion is because they heard the complaints loud and clear.
I am actually quite interested and concerned in how Rise & Fall will be integrated with the AI. They could potentially be setting themselves up for a backlash if the AI doesn't manage what will be a more complex system of core mechanics. I, personally, don't buy the idea that as customers we should wait and expect for a game to be improved as time goes by, not in basic and core game mechanics involving the AI and other fundamental things. I do expect the game to be improved by adding mechanics that will enhance the basic gameplay and give you more to play with, but again, not in fundamentals.
So, in conclusion, voicing and discussing this topic in an intelligent manner, always supported by facts and a consensus of opinion as well as voicing praise for what they do get absolutely right and what each of us enjoys in Civ in general will be a good way to let them know our perspective. Who knows, they might open up directly to us and actually acknowledge mistakes. That would go a long way with me and I presume with others to reestablish confidence and renew loyalty in a game we would love to love for actually being what it can be now, and not in the distant future.
I am actually quite interested and concerned in how Rise & Fall will be integrated with the AI. They could potentially be setting themselves up for a backlash if the AI doesn't manage what will be a more complex system of core mechanics. I, personally, don't buy the idea that as customers we should wait and expect for a game to be improved as time goes by, not in basic and core game mechanics involving the AI and other fundamental things. I do expect the game to be improved by adding mechanics that will enhance the basic gameplay and give you more to play with, but again, not in fundamentals.
So, in conclusion, voicing and discussing this topic in an intelligent manner, always supported by facts and a consensus of opinion as well as voicing praise for what they do get absolutely right and what each of us enjoys in Civ in general will be a good way to let them know our perspective. Who knows, they might open up directly to us and actually acknowledge mistakes. That would go a long way with me and I presume with others to reestablish confidence and renew loyalty in a game we would love to love for actually being what it can be now, and not in the distant future.
hdbhdbvGFG541
Chieftain
- Joined
- Jul 16, 2015
- Messages
- 25
Well for starters,maybe hire somebody who has a portfolio of actually making ai for 4x games.
The current ai lead seems to have never worked on a 4x game before,or any other strategy game before.
The current ai lead seems to have never worked on a 4x game before,or any other strategy game before.
Mr Jon of Cheam
Emperor
- Joined
- Oct 26, 2017
- Messages
- 1,358
I wish I could dislike this post. Such cynicism and negativity is unhelpful and sad to see.
It would be nice to think that some people at Firaxis read these forums, perhaps we could make a list of constructive criticisms and suggestions. Unlikely to help but better than simply complaining.
I don't know if that portfolio exists. I think the "AI engineer" is just a normal game developer they roped into doing the AI (Ed Beach was actually bthe AI developer on Civ 5 vanilla apparently, Soren Johnson I believe was 3's AI developer before leading 4).
Anyone who is interested in AI specifically is probably going to Google or the like for 3 times the money. Plus most AI development is currently focused on machine learning techniques and the like which isn't really feasible with Civ 6.
hdbhdbvGFG541
Chieftain
- Joined
- Jul 16, 2015
- Messages
- 25
Yeah i understand that,but here is the weird thing.
His name isn't mentioned in any firaxis game credits(except civ6),or even other game credits.
So where the hell did this guy come from i wonder?
ggalindo001
Prince
- Joined
- Jan 4, 2002
- Messages
- 362
I agree, I think this is a good thing to do and I hope that Ed and others read and react appropriately.
Would I love an AI that is "human-like" on the higher levels -- absolutely, but I think that is unrealistic at present. And probably in my lifetime.
However, there is a lot missing in the current iteration of the AI that I believe could be addressed without significant overhaul. And would make 80%/90% of the issues we have with the AI go away. I'm only going to focus on diplomatic/war/tactical behavior, and not some of the other systems. All of these observations are in a current deity game I have going right now as the Khmer, playing a religious game against Poland, Russia, Macedon, Germany, Egypt, Nubia and Rome. Science victory turned off.
1. AI diplomatic and "declaration of war" behavior not aligned to agenda and my actions. I have Rome and Germany to the south of me, Egypt to the west on a shared continent, and Nubia with a toehold to my northeast. I'm slowly rolling over Egypt --- and outside of denunciations, no declaration of war from Germany or Rome. At deity, I should have expected one of two things:
a. Germany and/or Rome JOIN in on the carving up of Egypt (keeping me from totally snowballing)
b. Germany and/or Rome use this as an opportunity to attack me (both have superior weapons tech to me, and would create serious problems for me)
In addition, both Nubia and Macedon could surprise attack me and create havoc, but haven't (and won't) -- although I have been denounced by both.
Because of the AI, I am not afraid of a declaration of war from either and as such, have a "free hand" against Egypt. Even the -threat- of one of the other powers declaring war (much less doing it), would cause me to have to maintain troops back and likely not be able to steamroll Egypt and ultimately snowball to a victory.
I'm hoping that the "emergency" feature that gets implemented with R&F will help with this and at least cause the AI to take some actions to help the snowballing. And I hope it is more than just a number of pre-defined events (use of nuke, take over of holy city) but identification of a snowballing civ and ability to counteract)
2. Air and naval -- Good that the AI is building both, but they aren't using both. In my current war against Egypt, without any air units of my own, their jet fighters sat idle while I kept pounding walls with my artillery supported by observation balloons. Also, I don't think I have ever seen the AI use an observation balloon, and the destroyers that could have hit back at my artillery never bothered shooting at it (or moving for that matter). Had Egypt used her fighters, it is likely I would not have been able to take their cities.
I will say that naval tactics in the ocean are a bit better, but still could be improved.
3. Attacking out of a walled city or encampent to an adjacent unit -- AI is still not doing this -- and passes on an excellent opportunity to severely damage a surrounding unit. AI hasn't realized that attacking out of an encampment or walled city essentially is a free shot as it cannot be hit back (unless the city is taken). AI is still afraid to attack out.
-------
I'm hoping that the new loyalty, emergency and governor mechanics will ASSIST the AI vs. further crippling it. From the two playthroughs so far, we have seen minimal AI tactical and strategic depth, so it is too early to tell. I do think that if these mechanics further cripple the AI, it will be hard for the series to recover -- but if they strengthen the AI or at least balance the player, it will be a huge success.
However, a few other suggestions (that are more macro in nature).
1. Difficultly levels should not just be based on providing AI "bonuses" but also address types of behavior vis-à-vis agenda. For example, on Prince or King, leaders should play to their agenda(s) in a very consistent way. But on higher difficulty level, the faster you go against a leader's agenda, the more likely it is they will react much more strongly. And potentially unpredictably as well.
2. AI Tech tree beelining -- I'm concerned about how this will play out on the higher difficulty levels with the golden/dark ages. This has to get addressed. I could add in the speed by which the tech tree is played out, and the need for intermediate units inside of the tree, but this is generally addressed through mods.
3. In Civ 4, I always believed that I was fighting everyone else....and early in Civ 5 the same thing. Somehow, the AI needs to get back to this -- fight each other, but also present a more "unified" front against the human player. In the long run, this will help with replayability. Again, maybe this is a setting based on difficulty level.
4. Hopefully the warmonger penalty will be further tweaked down. At the moment, wars are so infrequent once you hit renaissance, and that is not realistic.
Just my $.02 I'm very optimistic about the systems in R&F.....I just hope that we get a good rollout with it.
Mr Jon of Cheam
Emperor
- Joined
- Oct 26, 2017
- Messages
- 1,358
Great post.
Archon_Wing
Vote for me or die
- Joined
- Apr 3, 2005
- Messages
- 5,257
Nothing really, besides not supporting a product you don't approve of.
Bad AI has never stopped the series from selling. This is just true of most games. Reading posts online may make you think that everyone's trounching the game with ease, but in reality most players play very casually and don't necessarily seek higher challenges or even care about learning mechanics and stuff. A lot will play on like chieftain because they don't want or need to spend the time learning mechanics, and for them, the AI doesn't need to be challenging.
This is why I ask for perspective when certain things are discussed, namely that beating the AI is a given. Of course when you've played the entire franchise for thousands upon thousands of hours it might seem a bit easy....
Bad AI has never stopped the series from selling. This is just true of most games. Reading posts online may make you think that everyone's trounching the game with ease, but in reality most players play very casually and don't necessarily seek higher challenges or even care about learning mechanics and stuff. A lot will play on like chieftain because they don't want or need to spend the time learning mechanics, and for them, the AI doesn't need to be challenging.
This is why I ask for perspective when certain things are discussed, namely that beating the AI is a given. Of course when you've played the entire franchise for thousands upon thousands of hours it might seem a bit easy....
Ever Adrift
Warlord
- Joined
- Sep 28, 2010
- Messages
- 131
Quite frankly, Firaxis is horrible at customer service and at looking for and reacting to feedback. The best evidence of this is their refusal to hotfix anything, including game-breaking bugs that require then require the community to fix via mods. The most recent patch's map generator bug demonstrates this clearly - it took weeks to hotfix, and the hotfix didn't even work. Since then? Nothing. So now you basically have to use YnAMP if you want to play Civ 6. Meanwhile, games like WoW receive hotfixes at least weekly - and for major bugs it may only take hours. Of course, WoW is an mmo with a larger, dedicated team so you'd expect hotfixes faster. But no bug fixes outside of major patches? Really?
empresskiova
Warlord
- Joined
- Feb 19, 2017
- Messages
- 170
Lol if the live streams have been anything to note, you can’t make good AI if you don’t know how to play the game... 
- Joined
- Mar 3, 2011
- Messages
- 1,128
Yeah well I'm not convinced it isn't feasible. A full-time dedicated brain in a R&D set-up job , and not the easiest job but I'm convinced you can achieve quite some stuff with ML. The best would be to convince Fanatics to send data from their games, set-up the proper API in-game and start building models.
Temppu
Warlord
- Joined
- Oct 16, 2016
- Messages
- 160
I think this forum should be somewhat more tolerant to the AI of Civ than it is, and I mean both this thread and Civfanatics forums in general. Civilization franchise, and strategic games in general, is the most difficult possible game type for a competent AI, because of the super complex and dynamic nature of these games. Let me try to incentivice some tolerance.
Intelligence is very different for a machine than it is for a human being. For a machine, intelligence means quantitative decision rules, while for a human being, intelligence is an outcome of our brain, which is much more capable in many aspects than any supercomputer built to date. For example, babies who have just learned the word "dog" have no difficulty in distinguishing dogs in photos, but machines require very elaborate ML algorithms to be able to make that distinction. When you think that the AI does badly in some Civ waring decision and when you have a clear idea in mind what the AI should have done, try to think of a quantitative decision rule for making that decision in general. While thinking of that, try to think how that rule would affect hundreds or thousands, or tens of thousands, other decision rules the AI may have. This little exercise will turn out to be quite a lot more difficult than pointing out particular situations in which your super computer brain outperformed the decision rules made by the AI developers.
There were some comments in this thread on a machine learning approach of improving the AI. Pure learning approaches would never be able to work in Civ, because of the excessive randomness of the game. The number of different game situations is absurdly large due to, for example, the different map possibilities, different outcomes in each map after different decision sequences of all the civs in the map, etc. etc.. This excessive complexity makes development of learning algortihms simply infeasible.
Intelligence is very different for a machine than it is for a human being. For a machine, intelligence means quantitative decision rules, while for a human being, intelligence is an outcome of our brain, which is much more capable in many aspects than any supercomputer built to date. For example, babies who have just learned the word "dog" have no difficulty in distinguishing dogs in photos, but machines require very elaborate ML algorithms to be able to make that distinction. When you think that the AI does badly in some Civ waring decision and when you have a clear idea in mind what the AI should have done, try to think of a quantitative decision rule for making that decision in general. While thinking of that, try to think how that rule would affect hundreds or thousands, or tens of thousands, other decision rules the AI may have. This little exercise will turn out to be quite a lot more difficult than pointing out particular situations in which your super computer brain outperformed the decision rules made by the AI developers.
There were some comments in this thread on a machine learning approach of improving the AI. Pure learning approaches would never be able to work in Civ, because of the excessive randomness of the game. The number of different game situations is absurdly large due to, for example, the different map possibilities, different outcomes in each map after different decision sequences of all the civs in the map, etc. etc.. This excessive complexity makes development of learning algortihms simply infeasible.
I think it would be more honest to say that pure learning approaches would never be able to work in Civ with current hardware. There's just too many neurons and connections between them, that we don't really understand how they do what they do. Some parts, like the primary visual cortex (V1 or Brodmann's area 17) and primary auditory cortex (A1 or Brodmann's 41 & 42) are better understood, since the neurons there are arranged topographically, i.e. the center of our visual field is represented in the center of V1 and low frequency sounds are represented in one end of A1, while high Hz sounds are in the opposite end, and the rest in between in an orderly fashion.
However, there are clear rules, be it how a single neuron works, or how the brain as a whole decides whether some object is a dog or not. E.g. a neuron doesn't really have a choice about whether it will "fire" or not. Slightly simplified, if a certain voltage threshold between a neuron's intracellular and extracellular matter is reached, certain voltage-gated channels will open and the neuron will send a signal forward.
The way I see it, it will probably take an AI to create an AI which will be able to simulate the functions of the brain. I'm not expecting Firaxis to create such an AI. Maybe some Google, IBM and Microsoft collaboration?
ggalindo001
Prince
- Joined
- Jan 4, 2002
- Messages
- 362
Hopefully, I'm not coming across intolerant of the AI, as for the most part, I do think that the devs have done good job, in particular with the complexity of the game. However, as I pointed out earlier, there are some clear misses that get many on this board revved up about the quality of the AI. Air power, declarations of war in renaissance era and beyond, and use of units within cities that are being sieged are three areas that I believe are quick fixes that can go a long way in providing for a more challenging opponent.
I also think there are some game systems that handicap the AI even further. Two items in particular -- war weariness and separate health bars for walls. On war weariness, I think it is a mechanic that the AI becomes so afraid of, it will never launch a war. This significantly handicaps the AI, and allows for the human player(s) to snowball without much penalty. I'm wishful that R&F will bring better balance to the mechanic.
Second, I would prefer a solitary health bar for cities (and encampments) vs. dual health bars. In my perfect world, walls and encampments would do 3 things:
1. Allow for a ranged shot every turn (no walls, no ranged shot)
2. Increase overall singular city defense value to make the city harder to capture
3. Slow down damage to city based on attacking units
I would also increase city defense value if a garrisoned unit is stationed as well.
Making this change would do two things:
1. It would allow for cities to be captured without having to bring loads of siege units. You could effectively knock down a city through concentrated melee/horse. This would significantly help the AI that cannot currently figure out siege tactics
2. It would also allow for potentially better attack-counterattack play outside of the city as a more fluid battle could take place with melee units are trying to take the city outright.
One other point made earlier, which I think is a strong one -- it would be great for there to be a facility for users to send game "scripts" to Firaxis for analysis and processing. Right now, there are significant limits on what QA and the devs can do to gameplay test -- in the wild, there would be a tremendous amount of data that can be used to potentially best calibrate AI/coding to drive a stronger experience.
Hauptman
Chieftain
- Joined
- Jan 25, 2015
- Messages
- 68
Some of the concepts of this civ are awe inspiring. Whoever it was who conceived the overall game has my respect.
But for implementation, this game is absolute garbage. The list of failures is long and painful. From not being able to exit the game without task manager, to the lack of an AI, to complete and utter imbalance of tech speed vs build speed, and OH MY GAWD the absolutely horrible world builder that had to be modded in to even be usable. It's really hard to truly enjoy the game for all the cool features it brought. I waited a year to buy it, at half price, and still feel like I got robbed, but I'm still playing... so there is that.
I just quit a game before I came here because I lost a wonder... but in that game, here's what I saw...
Egypt and India declare joint war as soon as it's available (early game) then proceed to do nothing but send workers at me. They send one-charge workers to watch me build a wonder... which my horseman gladly takes. But not one single military unit. I sent 6 eagle warriors to go harass Egypt, maybe get some pillaging in, steal more workers, etc... and run smack dab into crossbows. Seriously Egypt? You have crossbows... and arent even raping me with them? In general, the AI will never attack you unless you share a border... never. I've played a number of games in this past month and not seen it happen once, besides the occasional privateer. Even when you do share a border... all the AI will do is run it's horsemen and pikes straight into your walls... Seriously believe you could play, and win, without ever building a single military unit... Civ 4 this game is not.
The real problem is of course... they hire dumb 20 somethings to make this stuff. Some 20 year old kid isnt going to know Civ, or care... The old timers, who grew up with Civ, and love it... went out and got real jobs by now. The guy who did Civ 4's AI personalities (Blake was it?) would be a godsend here... but somehow I doubt he'd come do it. So we are stuck with this crap... it wont get better, until we, the players, make it better... but that's how every game is these days. Once a good modder makes a good mod, then the game will be great. Hell I still play Total War; Medieval 2 Stainless Steel, and happily... just like I play Civ 4 still... and will never get tired of those old games because the mods made them what they are. So get to it troops... where's the Radius, or Darth Mod guy for this game... it needs it.
But for implementation, this game is absolute garbage. The list of failures is long and painful. From not being able to exit the game without task manager, to the lack of an AI, to complete and utter imbalance of tech speed vs build speed, and OH MY GAWD the absolutely horrible world builder that had to be modded in to even be usable. It's really hard to truly enjoy the game for all the cool features it brought. I waited a year to buy it, at half price, and still feel like I got robbed, but I'm still playing... so there is that.
I just quit a game before I came here because I lost a wonder... but in that game, here's what I saw...
Egypt and India declare joint war as soon as it's available (early game) then proceed to do nothing but send workers at me. They send one-charge workers to watch me build a wonder... which my horseman gladly takes. But not one single military unit. I sent 6 eagle warriors to go harass Egypt, maybe get some pillaging in, steal more workers, etc... and run smack dab into crossbows. Seriously Egypt? You have crossbows... and arent even raping me with them? In general, the AI will never attack you unless you share a border... never. I've played a number of games in this past month and not seen it happen once, besides the occasional privateer. Even when you do share a border... all the AI will do is run it's horsemen and pikes straight into your walls... Seriously believe you could play, and win, without ever building a single military unit... Civ 4 this game is not.
The real problem is of course... they hire dumb 20 somethings to make this stuff. Some 20 year old kid isnt going to know Civ, or care... The old timers, who grew up with Civ, and love it... went out and got real jobs by now. The guy who did Civ 4's AI personalities (Blake was it?) would be a godsend here... but somehow I doubt he'd come do it. So we are stuck with this crap... it wont get better, until we, the players, make it better... but that's how every game is these days. Once a good modder makes a good mod, then the game will be great. Hell I still play Total War; Medieval 2 Stainless Steel, and happily... just like I play Civ 4 still... and will never get tired of those old games because the mods made them what they are. So get to it troops... where's the Radius, or Darth Mod guy for this game... it needs it.
Last edited:
It's the lack of data (being a primarily offline game) that makes it difficult. And the fact that it's a heavily modded game. Too many decision points, not enough data. Something like freeciv (online, no mods) should be much more feasible with that sort of thing.
If I had a Civ magic wand, I'd say pull out just the tactical combat rules of Civ and make it it's own game. Two player online duels only, each person chooses their units at the start. Say different scenarios: i.e. one person has a city and wants to hold it the other is the barbarian with twice as many troops. Then that could be used to generate pathfinding and tactical decision trees that could be filtered into the full game. Assuming it turns out that the full game could handle them from a not-take-10-minutes-a-turn perspective.
Spudsie74
Warlord
WOW! Lots of posts on a perennial favorite topic. 
Needless to say, I can't resist the chance to put my 2 cents worth in....again. 
a
The AI is really, really stupid.
The smiley kid of says it all...except: The game works, bad AI & all. 
IIMO, there are multiple ways to win, an incredible amount of depth (or, if you prefer...fluff/chrome). The game has a LOT of replay value, and every game is different (even though the AI is'nt).![[pissed]](/images/smilies/pissed.gif "Pissed [pissed]")
Increasing the difficulty level is a big deal...I've played 2,000 hours, and am amazed at the difference between King,Emperor, and Immotal. Lots more work to win as you move up the ladder.
I left out Diety because:
I suck at that level.
Mostly 'caus I'm too lazy to do the micromanagement necessary...rather play while enjoying (take your pick; (beer, hard liquor, cannabis, or [choose your euphoric or orher vice]). Hard on the liver, brain & other organs, but anyone who has spent ~5,000 hours playing Civ 5&6 probably has some major issues. I know I do. 
But I digress....often
While I think Civ. is a great game, and have little hope for an ideal AI, I do think its' most annoying faults could be easily corrected. Probably take a mod, as Firaxis (or 2K, Sid, or whoever) is too busy counting money, and prepping the next DLC/Expansion to worry about quality. (I mean FANATICS quality....we do tend to be a bit picky. 
Needless to say, I can't resist the chance to put my 2 cents worth in....again. a
The AI is really, really stupid.
The smiley kid of says it all...except: The game works, bad AI & all. IIMO, there are multiple ways to win, an incredible amount of depth (or, if you prefer...fluff/chrome). The game has a LOT of replay value, and every game is different (even though the AI is'nt).
Increasing the difficulty level is a big deal...I've played 2,000 hours, and am amazed at the difference between King,Emperor, and Immotal. Lots more work to win as you move up the ladder.
I left out Diety because:
I suck at that level.
Mostly 'caus I'm too lazy to do the micromanagement necessary...rather play while enjoying (take your pick; (beer, hard liquor, cannabis, or [choose your euphoric or orher vice]). Hard on the liver, brain & other organs, but anyone who has spent ~5,000 hours playing Civ 5&6 probably has some major issues. I know I do. But I digress....often
While I think Civ. is a great game, and have little hope for an ideal AI, I do think its' most annoying faults could be easily corrected. Probably take a mod, as Firaxis (or 2K, Sid, or whoever) is too busy counting money, and prepping the next DLC/Expansion to worry about quality. (I mean FANATICS quality....we do tend to be a bit picky. 
ChocolateShake
Prince
- Joined
- Dec 28, 2016
- Messages
- 525
From my perspective, I just want the AI to be a military threat again. They can sometimes feel like one, but they often make odd decisions like retreating instead of taking a city. If they could just make the military aspect of the AI smarter, that would go a long way to increasing the challenge of the game.
From my understanding of it, the AI in 6 has too much going on, it looks like there are many competing subsystems. I think like @Spudsie74 mentioned, modders will have to work on it. There's just not enough incentive to make the AI significantly more competitive, I think most 4X games have this issue, it's probably too much of an investment compared to things like new art and game systems.
From my understanding of it, the AI in 6 has too much going on, it looks like there are many competing subsystems. I think like @Spudsie74 mentioned, modders will have to work on it. There's just not enough incentive to make the AI significantly more competitive, I think most 4X games have this issue, it's probably too much of an investment compared to things like new art and game systems.
Last edited:
Similar threads
