Does anyone else agree that the AI is the highest priority?

[dexters]

But the "properly" is subjective

you win!

Rephrasing:

If the AI can 'react' to a situation, it is understanding.
How the programmer gets the AI to do that is a detail.

 

[Ryika]

I think the word "understand" still makes sense though. Understanding is what allows us humans to adapt to pretty much any situation - with varying results, but always with a goal that fits the situation.

An AI will only ever be able to react to situations that are pre-programmed, it doesn't try to come up with a goal that fits the situation, it tries to fit the situation into one of its pre-defined goals.

Or in other words: We examine a situation and come up with a solution tailored to the situation. The AI examines a situation and then runs one of the solutions that it "knows" of. That's why its solutions can never compete with human solutions (assuming a general competence in the human player ). No matter how many solutions it has in its repertoire an AI of the sorts that is available now will only ever be able to somewhat mimick us and then needs help by giving them extra bonuses to get to our level.

 

[dexters]

I think the word "understand" still makes sense though. Understanding is what allows us humans to adapt to pretty much any situation - with varying results, but always with a goal that fits the situation.

An AI will only ever be able to react to situations that are pre-programmed, it doesn't try to come up with a goal that fits the situation, it tries to fit the situation into one of its pre-defined goals.

Yeah, I use 'understand' to mean an AI can read a situation. Whether it reads it properly or improperly can be subjective.

So to distil my large post upthread, AI should be able to 'read' the map, read border changes over time, read the score table to see growth or decline in scores, read some/most negative actions done to it.

Their actions in response will largely be pre-programmed, but if an AI can be programmed to read a situation and react with different combinations of pre-programmed action plans, then it can create an emergent system where the AI can pretty well protect it's interests by 'correctly' understanding where the threat is coming from, and not giving human players slack when the humans try to game the mechanics by forward settling to steal land, compete with them for city-state influence, and rope in other people to fight them. etc.

This all sounds academic but it can be as simple as NOT using RNG /proximity to decide who to attack next, but improving their target selection by layering a more thorough threat assessment and ranking of who they need to keep in check. That's all numbers the AI can be programmed to crunch. And the end result is 'better' (subjective term) target selection and less likelihood to be bribed into useless wars.

I mean, their on the ground 1UPT tactics may still and always be outmatched, but in many games, all it takes is an AI declaring war at the right time to make the difference between a epic game where the human player felt like the AI reacted to their rise and stole an easy win from them or made them work for that win that was so close, and one where the AI seemingly throws the game as they are bribed by the humans to attack someone else.

 

[Eagle Pursuit]

Fun is the highest priority and AI is a big part of that.

 

[Stringer1313]

I agree that in THEORY, AI should be the highest priority. In practice, however, I don't think it is useful to make it the highest priority (certainly a high priority, but not the "highest"), because I don't know how much bang for the buck you get out of it.

It is nearly impossible to design a good AI in these kinds of complicated games; playing with humans will always be superior. It is a guarantee that fans, especially expert hard core Civ games (of which I am not), will not be happy with the AI.

Just to illustrate my point with some arbitrary numbers: Out of all available programming hours, I would not want the programmers to spend 50% of their time perfecting the AI, if spending 10% of time gets the AI "good enough", and 40% of the time improves the AI only marginally. I'd rather they spend that time making diplomacy more fun, the leaders more unique, for instance (maybe that's a poor example b/c every Civ says they are improving diplomacy or making leaders unique but it never really happens).

 

[Blkbird]

the most important thing is features, mechanics and balance.

if the game doesn't have working features and mechanics, and isn't well balanced, then it doesn't matter if the AI is good or MP works.

This.

Plus, I measure my performance against my development potential under given circumstance, and don't really compare myself against the AI anyway.

 

[MadDjinn]

Yeah, I use 'understand' to mean an AI can read a situation. Whether it reads it properly or improperly can be subjective.

So to distil my large post upthread, AI should be able to 'read' the map, read border changes over time, read the score table to see growth or decline in scores, read some/most negative actions done to it.

Their actions in response will largely be pre-programmed, but if an AI can be programmed to read a situation and react with different combinations of pre-programmed action plans, then it can create an emergent system where the AI can pretty well protect it's interests by 'correctly' understanding where the threat is coming from, and not giving human players slack when the humans try to game the mechanics by forward settling to steal land, compete with them for city-state influence, and rope in other people to fight them. etc.

This all sounds academic but it can be as simple as NOT using RNG /proximity to decide who to attack next, but improving their target selection by layering a more thorough threat assessment and ranking of who they need to keep in check. That's all numbers the AI can be programmed to crunch. And the end result is 'better' (subjective term) target selection and less likelihood to be bribed into useless wars.

I mean, their on the ground 1UPT tactics may still and always be outmatched, but in many games, all it takes is an AI declaring war at the right time to make the difference between a epic game where the human player felt like the AI reacted to their rise and stole an easy win from them or made them work for that win that was so close, and one where the AI seemingly throws the game as they are bribed by the humans to attack someone else.

'understand' and 'react' are two different things.

An AI can understand what is going on, but not 'react accordingly' if it wasn't programmed to do so.

An AI can 'react' to situations without actually understanding them.

even if you toss the idea that the AI's 'understanding' is just basic, rather than aiming for knowledge, it's still not the same as pure reactionary systems.

It actually wouldn't be that hard for the AI to be programmed to have some knowledge of what's going on in the game, but then you'd have to toss out the 'AI' programming in favour of the stuff that works ('CI').

 

[dexters]

'understand' and 'react' are two different things.

An AI can understand what is going on, but not 'react accordingly' if it wasn't programmed to do so.

An AI can 'react' to situations without actually understanding them.

Right, my initial use of understanding is closer to your interpretation.
But when it was pointed out the Ai cant really understand situations (like a human can) I rephrased it to react (to a situation), implying it's reacting to pre-determined values with pre-scripted behaviours, which is what I assume what the AI is doing most of the time. The end user can interpret this reaction as understanding the map or situation or combination of those two. This is the illusion and theatrics Civ has always tried to do. Whether getting to those variables is actually considered understanding or not is not something I feel I have to authority to argue. I suppose a game can be programmed to react to certain situations in such a way that a human could deduce is understanding but the game actually doesn't understand. But if it works in the game context reliably, then to the lay person it is a understanding because doing X AI produce Y result.

In casual talk, I would most certainly classify it as the AI 'understanding' something when it reacts appropriately or even so appropriately to a situation.

I also take your point that an AI can 'understand' map variables but react incoherently. That of course is the job of the AI programmer to balance, fix and write.

Going back to my point, the AI 'understanding' the map is the bare minimum I expect. The reactions is what I assume the programmers will spend the rest of their time tweaking, adding to those reactions as the game goes live and they see how people play against it. But I generally agree with your position that the game needs to be balanced with solid features , and that it be a lot of fun.

 

[Thormodr]

Just make it generally compotent. The AI in Civilization 5 was utterly bamboozled by 1UPT and performed very poorly because of it.

Hopefully they'll do a better job with Civ VI.

 

[noto2]

I don't think people are really getting what noto says here, I agree completely that MP is better and more enjoyable and I play it pretty exclusively for CiV but I don't think noto is disagreeing. Rather, the point made in the OP is that it's impossible to make MP both -fully satisfactory, IE a full game with complexity and enjoyability, as well as -completable in a short period of time. So, in those times where you have only 1 hour but you wanna play some Civ, you have to play SP, and then it becomes necessary to make better AI.
I agree with most of what noto says here. The reason I only ever play MP is because the AI is so bad. If the AI was better, I'd still play mostly MP just for personal enjoyment, but it would be nice to then be able to have SP as an alternative option when necessary.

This right here. I would rather play Civ MP. If I were filthy rich and didn't have to work, and didn't have a family, and I had a bunch of friends who were like me and also avid gamers, sure I'd love to live in a fancy hotel with a gaming room and play MP civ all day. In reality, though, Civ is just not MP friendly. Also, making it MP friendly would probably ruin the game. If you made Civ so that it never took more than 2 hours to finish a game, it would probably be very simplistic.

Dota and Warcraft and Starcraft are MP friendly because games take 30-60 mins. A MP game of Civ is going to take 3-6 hours, and most of the time it's just not possible to take a chunk of time like that out of your day without interruptions.

 

[Thormodr]

Fun is the highest priority and AI is a big part of that.

Spot on. Soren Johson did a talk on that subject. Do a search on "Playing to Lose:AI and Civilization" by him. Very insightful message and well worth watching if you can devote an hour.

The AI is participating/facilitating in telling a story. Most of all, it needs to be a fun experience. It needs to be competent at what it's doing, certainly, but we don't play Civ to watch the AI do it's job. We play to have fun.

 

[Acken]

Balance and AI performance go hand in hand in order to create a challenge. You can't have a good AI if you don't have good balance. Or you'd have to force it to take the imbalanced routes.

With fun game features (usable by the AI) these 3 are highest priority for me.

I doubt civ will deliver at release on the balance. But if the 2 others are well made at least mods usually can work on balancing the game.

 

[rastak]

Spot on. Soren Johson did a talk on that subject. Do a search on "Playing to Lose:AI and Civilization" by him. Very insightful message and well worth watching if you can devote an hour.

The AI is participating/facilitating in telling a story. Most of all, it needs to be a fun experience. It needs to be competent at what it's doing, certainly, but we don't play Civ to watch the AI do it's job. We play to have fun.

Amen.

Engaging gameplay, a challenge for the average man at normal difficulty. I normally play a couple of levels below max and hope I'm not walking over everything. If you stomp everything it gets dull. See the game Acendency many years ago......

I do get that people have fun in different ways. For those with furrowed brow and sweat dripping over the next grand stroke based on advanced mathematical analysis I suggest chess......

 

[Acken]

For those with furrowed brow and sweat dripping over the next grand stroke based on advanced mathematical analysis I suggest chess......[/B]

No. I suggest making an AI competent enough that getting double production and double growth make it challenging.

 

[Rusty Edge]

This is where I mention that a good A.I. is one that makes the human react to it, rather than one that understands what the human is doing and react to that.

In civ IV humans thrived on stealing A.I. workers.
So the patch made the workers run & hide safely in the cities.
So the humans would park an axe or warrior on a wooded hill and scare the worker indoors, crippling it's growth.
Etc. There's no end to reacting to the human.


There used to be a multitude of multiplayer forms before STEAM- PBEM, succession games, LAN, hotseat, etc. With that, people could allocate their available time. That keeps the humans reacting to the other humans, which makes up for a lot.

I do adore an improved A.I. But my top priority?

A game full of meaningful decisions in an historical context. Seeing what happens next as a result is what keeps me playing "One...More...Turn!"

 

[spfun]

I just want an AI that's competent and doesn't need huge bonuses to be competitive. The AI in Civ 5 had no clue how to play 1UPT and easily gets wrecked even with huge bonuses and this repeated with BERT. I don't want another repeat with Civ 6 and I'm hoping the changes to support units and armies allow the AI to wipe me off the map if they catch me napping.

I spend 99.8% of my time in SP as its more fun.

 

[drivebyslow]

the most important thing is features, mechanics and balance.

if the game doesn't have working features and mechanics, and isn't well balanced, then it doesn't matter if the AI is good or MP works.

Agreed. Balance is what it's all about. Even if the graphics aren't there. That's why some games remain classic and entertaining. Remember King's Quest 6? Hell yeah!

 

[Biz_]

yeah, but i gave up on firaxis actually caring about the AI for non-newbie players

it's a sad state of affairs, but i'm realizing that the only realistic way to get a good non-cheating 4X AI is to program one myself. nobody else seems to care about it enough to make one
i should be done in a few years, but it's not for a civ game

 

[Talcove]

I've pondered the idea of making Civ more MP friendly, but this runs into many problems. I play Dota and enjoy it, but the fact that I can only play it online gives it certain limitations. I can't play Dota if I only have 45 minutes to spare. I can't play it if I'm expecting a call, or anticipating the need to get up and do something in the near future. I can't save a game and pick it up later. Dota works because most games only last about 45 minutes, on average, or less.

I don't see how Civ could be enjoyable as a game that only lasts an hour per game. So I don't see any way to make it practical as a MP experience for most people. If everyone had to clear their schedule for 8-10 hours to play a game it would be something that most of us could barely do once per week, if that.

You're missing a key element, Civ is turn based. In DotA the action continues until one side wins/loses, whether you're at the keyboard or not. There's no pauses and there's no saving. But in Civ you can very easily just complete your turn and get up and go get a sandwich, or tell the people you're playing with that you need to take a break for a phone call, or whatever. If it takes particularly long you can just save the current game and come back to it later when you have more time. That's what allows it to work well as a multiplayer game.

But yes, even with that the length of an average game necessitates that singleplayer should still be the main focus. You can take breaks, you can save and come back later, but in something like an 8 person game that can be VERY hard to schedule, impossibly so if you're playing with randoms. For the game to be challenging and enjoyable while playing by yourself is paramount, and a competent AI is one of the main components of that. However I have to disagree with the previous posts, the AI should absolutely play to win. They should have their own unique personalities, traits, quirks, and be balanced around the player and difficulty, but at the end of the day playing off against an opponent or opponents that don't try to stop you from winning and win themselves is just dull.

 

[noto2]

I'd like it if there was an option to make the AI play to win. I get it that there are people out there who want the AI to roleplay instead of play to win, but I'm not one of them. In civ4, Karadoc made a mod where the AI really did play to win, it played almost exactly like a human. It was completely ruthless and cut-throat. It would rush you in 3000 BC with 7 archers if that meant it could take a city from you. I absolutely loved it. It also meant I went from being an immortal player back down to an emperor player and it was still 3x more difficult. If it wasn't for that mod I would have grown bored of Civ4 about 1500 hours before I did.