Whatever they do I hope the AI is substantially upgraded

Discussion in 'Civ6 - General Discussions' started by Zinowolf, Dec 16, 2018.

  1. SammyKhalifa

    SammyKhalifa Deity

    Joined:
    Sep 18, 2003
    Messages:
    6,021
    Yeah, no doubt, didn't mean to make it sound that way. "Don't do dumb stuff" is fine, because let's be honest, as much as I would like a challenge I also don't want to lose all the time anyhow. ;)
     
    TahamiTsunami and SupremacyKing2 like this.
  2. steveg700

    steveg700 Deity

    Joined:
    Feb 9, 2012
    Messages:
    3,624
    What I find encouraging is when I hear that the AI behavior is tweaked for specific civ's so that they don't all follow the same script. So, in R&F, not all civ's pursued religion with equal fervor. And in the Hungary stream, we hear that now not all civ's will default to conquering every city-state they can reach before the classical era. We now need a pass for wonders, so the AI isn't trying to build civ's just because it can.

    Even if the AI can't be highly adaptive, it can at least have variety in which script it uses. That provides some challenge in the form of variety, at least.

    Of course, then there's the problem with combat. I can get why the AI can't used 1UPT effectively, but why can't it use aircraft?
     
  3. SupremacyKing2

    SupremacyKing2 Deity

    Joined:
    Oct 25, 2014
    Messages:
    4,495
    Location:
    Indiana
    Honestly, I think that is either a strategic resource issue or a priority issue. Most likely a combo of both. 1) The AI needs the strategic resources to build them and 2) it needs to make it a priority in its build order. It is very likely that the problem lies in the lack of priority. The AI has a list of stuff it needs to build and aircraft is placed at like #10 on the list so it never gets around to building them.
     
    steveg700, acluewithout and Elhoim like this.
  4. TahamiTsunami

    TahamiTsunami Prince

    Joined:
    Sep 27, 2017
    Messages:
    458
    Gender:
    Male
    I begin to feel quite self-conscious when people describe the AI as too weak yet here I am getting wiped out by turn 40 by an Aztec surprise war on king difficulty! I guess it shows how much of a newbie I am to the series!
     
  5. SammyKhalifa

    SammyKhalifa Deity

    Joined:
    Sep 18, 2003
    Messages:
    6,021
    But once you learn that one trick from them it's easy to side-step, and if you get through the first 50 turns not much else is going to get in your way.
     
  6. steveg700

    steveg700 Deity

    Joined:
    Feb 9, 2012
    Messages:
    3,624
    See, to me that's realistically addressable.

    The AI has a build order, and it just goes through it and checks for capacity and then proceeds. Stop prioritizing building every wonder so highly, and start prioritizing defending itself from having its wonder-filled cities captured. Aircraft would be an easy way to do this, since they are fire-and-forget defense.
     
    SupremacyKing2 likes this.
  7. Council 13

    Council 13 Chieftain

    Joined:
    Aug 8, 2013
    Messages:
    74
    Location:
    Chicago IL
    It may be beating a dead horse, but the AI (IMO) is worse at this stage than it even was in Civ 5. It doesn't repair tiles, it doesn't recapture workers and settlers from barbs, it barely creates any kind of real armies (even the war mongering civs), it doesn't use city states during war... on top of the combat issues I mentioned earlier. I mean come on... its terrible and possibly worse than ever.
     
    Babarigo likes this.
  8. Dave "old computer guy"

    Dave "old computer guy" Chieftain

    Joined:
    Sep 20, 2017
    Messages:
    36
    Gender:
    Male
    Location:
    Minnesota
    I do think that the issue of AI abilities in this (and previous) threads is missing one extremely important element. IBM spent a goodly number of years and a pile of money teaching a computer to play passable chess. Chess is a game of 32 "units" played on a field of 64 tiles with two combatants. Standard game settings for Civ presume 8 combatants, 80 plus units and a thousand tiles. A complexity factor in the tens of thousands. All of the possible moves must be evaluated within the confines of the device capabilities in your personal computer. Would you settle for 5 minute wait times while the AI makes up it's mind where to put it's units? As a product of the tech industry I am amazed at how the AI performs given the limitations of our personal computers and don't see major improvements until we move up a generation or two or three in personal computer capability.
    Question for those playing multi-player, are your human opponents that much smarter than the AI?
     
  9. SammyKhalifa

    SammyKhalifa Deity

    Joined:
    Sep 18, 2003
    Messages:
    6,021
    plus the AI has to pretend like it can't see the whole board.
     
  10. Elhoim

    Elhoim Iron Tower Studio Dev

    Joined:
    Nov 3, 2004
    Messages:
    2,639
    Gender:
    Male
    Location:
    San Isidro, Argentina.
    Sure, nobody disputes that. But you should download Vox Populi for Civ V (which is basically the same combat system), and what they managed. It's MUCH better than the stock Civ AI. Nobody is asking for a superhuman level of AI, but a competent one for which we already have a working example.
     
    Civrinn, oedali, footslogger and 2 others like this.
  11. Gorbles

    Gorbles Load Balanced

    Joined:
    Nov 24, 2014
    Messages:
    6,122
    Location:
    UK
    Sure, but it's easy (easier, specifically) to mod in an AI on top of an existing framework (and an existing functional AI) than create it all from scratch, from the ground-up, during a typical development cycle. I've been a modder. You simply don't have deadlines in the same way, regardless of the arguments of "but it's your free time". I don't have the luxury at work of working on things forever. I have that luxury with things I do in my spare time - however many days of that spare time it takes.
     
    bbbt, cvb and SammyKhalifa like this.
  12. Wingednosering

    Wingednosering Prince

    Joined:
    Dec 12, 2011
    Messages:
    536
    Gender:
    Male
    Location:
    Toronto
    I said something similar back on page 1. Obviously even an AI with ten minute turn times can't out think a player with a game this complex with this many layers. As a game dev that has coded AI though, I can tell you there's a lot of low hanging fruit here that could quite reasonably achieve better results with minimal overhead.

    Why did we wait so long to get a religion preference list? Why is city state destruction only a variable factor now? Why aren't certain civs more/less inclined to get certain wonders? Why does the AI rarely use spies? Why do they rarely repair pillaged land? Why don't warmongers have respectable armies?

    As for the game being harder on Prince, what I meant was that the AI should be 'better' on Prince. I don't remember what word I used before, but I believe asking for Deity alone to have a 'smarter' AI isn't reasonable. They need to make the AI smarter across all difficulties and then balance their starting bonuses at each level accordingly so the change is minimal until you get to Emperor+.

    Coding different levels of intelligent play from the AI based on the level would require way more resources than it's worth, unfortunately.

    What previous posters have said is correct. I don't believe Civ VI is meant to give a challenge, it's aim is roleplaying. Additionally, a big issue is that no matter who you play as, certain techs, civics and production just waste your time and thay shouldn't be the case. The AI wasting resources on Wonders and religion is a big part of why they fail to challenge human players.

    I once sat back and waited on Emperor for the AI to win and it timed out. A score victory should never, ever happen on Emperor (this was vanilla, not R&F just to clarify).
     
    Last edited: Dec 17, 2018
  13. Gorbles

    Gorbles Load Balanced

    Joined:
    Nov 24, 2014
    Messages:
    6,122
    Location:
    UK
    The answer to all of your questions, frustratingly I'm sure, is "because there was higher-priority stuff to do". The average backlog for most products of a size and complexity of an AA / AAA video game is long. There are so many bugs you don't see, just so you can see the few you do. Or the many, I'm sure, over time at least.

    AI is not only a pretty-specialised field, but there isn't a huge amount of space for it in video games; that's why it's mostly apparent in other fields (sometimes educational, sometimes military, sometimes random stuff like whatever automated robot Google is assembling this month). And AI programmers (much like engine programmers, or network programmers) tend to not only be experienced, but good programmers. Which makes them invaluable on things that aren't just AI. I mean, I have no idea if this is the case with Firaxis. I'm a software dev myself; not games. But that happens in my field too. Folks get stretched, as much as they let themselves be (in the best situations).

    I wish we could have a discussion on the potential of AI in games (and why difficulty-based resource scaling is so incredibly common without having to reduce it to "because AI is hard") without having to go down the "why aren't the professionals considering these very obvious and seemingly easy improvements". Which we kinda seem to do, most times ;)
     
  14. PhilBowles

    PhilBowles Deity

    Joined:
    Nov 20, 2011
    Messages:
    5,333
    The Civ V AI could handle air combat and by the latter stages of the game's life it was programmed with a reasonable unit mix. It was a bit hit or miss with naval invasions - Civ VI is actually more consistently able to at least send units across the ocean en masse - but mostly adequate towards the end.

    Programming the AI to use units more effectively than it does now is unlikely to be worth the resource investment, but there's no clear reason I can see why it can't be programmed to at least build artillery, default to a more effective unit composition than anti-cavalry spam (does the late game AI build anything but AT teams?) and exhibit some capability to deploy aircraft (it can sometimes build them, just not use them except fighters on intercept - which happens automatically). And if it can't use air units it could at least be programmed to build AA - that was Civ V's workaround since, although the AI could use the system, it was not good at it and rarely used fighters to defend.

    Combat isn't a system that's become more complex than it was in Civ V, so I don't know why the AI is so much worse than it was there. It certainly has become better over time - at release the AI literally wouldn't attack cities, for a start, and garrisoned units would never fire at attackers - but it's still noticeably much worse than Civ V was even by Gods & Kings, when the 100 HP system was first introduced.
     
    Trav'ling Canuck likes this.
  15. acluewithout

    acluewithout Deity

    Joined:
    Dec 1, 2017
    Messages:
    3,447
    Inevitably for such a mainstream game, FXS must try to cover multiple bases. I agree one goal has been this “play it you way” goal. But there has also been a “play the map” goal, which is the complete opposite goal - if you have to play the map, then you can’t really play it your way.

    I don’t think the the poor AI is deliberate, and instead FXS do want a challenging game. But they’re not able to deliver.

    I don’t think the issue is lack of resources or attention either. People keep talking about one dedicated person working on AI. To me, that sounds like a lot of people focusing on one task. And I also don’t think having a dedicated AI person would mean they don’t get support from other people at FXS.

    My guess is that the AI is so poor because (1) it’s hard to get AI right, and (2) Civ makes it particularly hard for the AI because of all the various systems. There are some really good posts above - too many to quote - which make this point better. To get AI right FXS need the right people, and then need to get lots of other things right including organisationally. In fairness to them, that’s not easy, and there are lots of other things they’re being asked to deliver on too. I expect this AI stuff is frustrating for them just as much as it is for us.

    Yes, I agree. And I’m sorry I’m jumping into this topic again. I’m just cross about things.

    For what it’s worth, I am making two slightly different point. First. That is: rather than just “fixing” the AI, I really think FXS should make the game hard at Prince.

    It’s just my opinion. But I think the game would be much better - and would pull in way more players - if the game was hard out of the gate. The game should mess you up at Prince. The Darksouls (...yay, I’ve mention an almost modern game!) or MegaMan (doh!) or arcade Gauntlet (...okay, that’s deliberately anachronistic). Some games are popular precisely because they are hard.

    To be clear, the game would still need to be accessible - so, intuitive controls, easy to understand systems, lots of guidance. And you would need an assist mode - eg you can play Rome and they are really, really punchy, you can start with extra settlers and Builders, Gilgabro starts as your Ally etc.

    But Civ should be getting reviews on Steam like “Man, this game just wrecks you! Ghandi is the worst”.

    I know, I know. I have nothing to back this up with. No research. Maybe I’m wrong. But I really think that sort of difficulty would drive sales. And players that don’t want so much challenge would still be okay because they could play lower difficulties or play with an assist mode.

    Second, and lastly, I also think FXS need to think of challenge more generally. It’s not just about AI (meaning how well the computer plays the game).

    As I said in another thread: I think the games problems are deeper than the AI.

    Core problems include:

    1. Your opponents all start at the same time as you. This is the source of the whole snowball problem. Can you imagine any other game where you’re opponents basically never got more powerful. This is, more or less, how the game currently works absent some isolated Civ running away with the game. The game would be much more interesting if you started with some Civs already ahead of you and established (ie you’re the new guy on the block), and then later you had new Civs, some behind you (like you were) and some that rapidly expand (like America).

    2. You’re never required to actually manage your empire. Your only threats are external.

    3. You can’t lose, absent losing all your cities or someone else which is fairly extreme. So, you’re never under pressure, you can always grind out a victory, and you never get any feedback on how well or not you’re playing.

    RnF basically fixed none of these issues. Even loyalty and Eras don’t create any empire management.

    The game is designed like Monopoly. Once you’re established, there’s no one that can stop you.

    OMG this is the third [er, fourth] time I’ve mentioned Monopoly on this forum. Unbelievable.
     
    Last edited: Dec 17, 2018
    steveg700 likes this.
  16. steveg700

    steveg700 Deity

    Joined:
    Feb 9, 2012
    Messages:
    3,624
    No need to apologize.

    Someone doesn't like the topic, they can kick rocks. Nobody's forcing anyone to participate.
     
    Last edited: Dec 17, 2018
    Aristos and Trav'ling Canuck like this.
  17. Browd

    Browd Dilettante Administrator

    Joined:
    Aug 5, 2012
    Messages:
    11,936
    Gender:
    Male
    Location:
    Rural Vermont
    Keep in mind that the developers of VP have had the benefit of modding a base game that has not changed in four years (last Civ V patch I recall was Oct. 2014). Those who work in AI (certainly not me :D) have repeatedly made the point that it is well-nigh impossible to fine-tune AI for a game whose systems remain in rapid development.
     
    Gorbles and steveg700 like this.
  18. SupremacyKing2

    SupremacyKing2 Deity

    Joined:
    Oct 25, 2014
    Messages:
    4,495
    Location:
    Indiana
    I disagree. Obviously, that's the kind of game that YOU want and I have no doubt that there are lots of other players like you but there are also a ton of players who would be completely turned off if the game were that hard right out of the gate. A lot of players will take the path of least resistance. If the game is too hard right out of the game, they'll jump ship and play something else. Now, I know you say that the game should be intuitive and easy to learn but hard at the same time. That's a really tough balance to find.

    Having civs that start at different levels is a very interesting idea. Although Beyond Earth kinda tried this, and maybe it was just not implemented correctly, but I don't think it worked at preventing the snowball effect. I think the cause of the snowball problem is different. The reason the human player tends to snowball relate to your points 2 and 3 below.

    Agreed. There are no internal (or external) threats to your civ that left unchecked could reverse your ever growing power. Historically, big empires faced internar problems and/or external threats which eventually caused them to collapse. Now, I am not suggesting that the game should destroy your entire civ completely. But if the game had internal threats which, if left unchecked, could shrink your civ at least, that would definitely prevent a snowball effect. So I do think the game needs a real "rubber band" mechanism that would cause the runaway civ to lose power and shrink. I agree that loyalty does not do this but I think it has the potential that could. The game needs more things that add or subtract loyalty from individual cities to reflect how your population feels about things. For example, maybe a city that is facing barbarians attacks could incur a loyalty penalty until you build city walls (you would get a message explaining that the people demand protection). Or cities with high science would incur a loyalty penalty if you remove a pro science policy card (the people are upset that you cancelled a policy that they liked). Or a lack of amenities or a lack of fresh water could cause a loss of loyalty. Things like this could add empire management and maybe cause you to even lose cities if you ignored the loss of loyalty for too long. The game could also add a loyalty penalty to cities that are far away from your capital and also based on number of total cities. So wide bloated empires would lose loyalty on their borders. This would be a rubber band against never ending snowballing.
     
    WorldWarIV, 679x, PeterChu and 2 others like this.
  19. steveg700

    steveg700 Deity

    Joined:
    Feb 9, 2012
    Messages:
    3,624
    Okay, but aircraft hasn't really changed, has it?

    And screening an artillery unit behind melee units rather marching it out to be destroyed, that's pretty much an out-of-the box expectation that hasn't changed.

    And not trying to build every wonder just because it can, and even while their empire burns....Hate to be reductive, but that could have percolated in the interim between a first or second expansion.
     
    Civrinn likes this.
  20. Malachi256

    Malachi256 Prince

    Joined:
    Feb 16, 2008
    Messages:
    539
    If you want a good AI you'll have to look elsewhere. That is clearly not the priority, nor will it ever be the priority, for these devs. Your opponents are just an additional resource for you to exploit in your kingdom-building simulation.
     

Share This Page