Have Games now gone Full-Circle?

[Jabulani]

@ Themeinteam

An AI is not rocket science, above all in civ vi, that's much much more easy to instruct a civ to at least use air units then teaching the AI how to counter pro chess players, so you are out of order here sir

I would now propose an award, the worst AI award 2016 goes for : Civ vi devs ! Don't be shy, come here among us to chat, you are welcome !!!

 

[ImperialGuard]

Then along came PBEM,

I've abandoned CIV 6 ... really disillusioned with Firaxis, to release this product in current state and not deliver anything meaningful in a patch 60+ days after release ...

but thanks to them, I took the dive into Dominions 4 and found a thriving PBEM community with automated servers that run the games. Worth a look if your interested: http://z7.invisionfree.com/Dom3mods/index.php?showforum=24

 

[DizzKneeLand33]

To steer slightly back to the OP, one thing I find interesting is the increasing number of kids taking up chess and playing in tournaments. Back in my day these tournaments were more like Math Club in The Breakfast Club ("demented and sad, but social."), but there seems to be almost a yearning to have a more face to face experience. So, face to face to human vs computer to multiplayer to face to face, yes I do believe that things are going full circle. Even for games like Civ there are LAN parties here and there at specific businesses, further encouraging the human experience.

As a Candidate Master chessplayer, the computer has been an interesting development (I started playing tournaments in 1983). At first we made fun of it because it was so weak, now the top GM's can barely score draws against them. The world champion Go player (Lee Sodol I think) was beaten in a match in 2016 by a computer, the first time ever. The Civ 6 AI is pathetic currently, and quite frankly not because of 1upt (I would believe that but for the Acken mod for Civ 5, where the AI is extremely hard to beat on Immortal and even knows how to efficiently conduct a *naval* (gasp!) battle). It will get better, but maybe that's why it's not as high of priority anymore -- maybe humanity is wanting to go back to playing games with other humans. I haven't played a chess game against a computer in a decade -- why play the computer when I can play other people, either in person or via game servers?

To make chess more human again, I believe that a 10x10 board with Capablanca's version will become more popular in the next 10 years. No opening book, yes the AI will catch up eventually but we can still have some time for humans to be on the "frontier" when playing each other. Even correspondence play against other humans (aka PBEM in today's lingo) sounds like a refreshing experience, knowing that the computer will not (as of yet) be making all the best moves in this variant.

Sorry for the rambling, but my point is the following -- maybe we are at the apex of AI play in games, and indeed that would mean that game play is coming full circle.

 

[Gorbles]

@ Themeinteam

An AI is not rocket science, above all in civ vi, that's much much more easy to instruct a civ to at least use air units then teaching the AI how to counter pro chess players, so you are out of order here sir

I would now propose an award, the worst AI award 2016 goes for : Civ vi devs ! Don't be shy, come here among us to chat, you are welcome !!!

A lot of things are simpler than rocket science, but that doesn't mean they still aren't difficult.

Chess is also a lot easier to handle than something like Civilisation; I detailed arguments for this the last time AI was a topic.

 

[callan]

@ Themeinteam

An AI is not rocket science, above all in civ vi, that's much much more easy to instruct a civ to at least use air units then teaching the AI how to counter pro chess players, so you are out of order here sir

That's really not true at all. True Artificial Intelligence is still decades away at a minimum, we've been sending rockets into space since the 50's. Indeed we possess the math to land on or launch from any Astral body, what we are missing is the engineering. Even programming a gaming AI is much harder than you suggest. You seem to think Chess AI's are good yet anyone at my rating would consider commercial chess AI's to be a toy. It's not that there is no challenge to them but, much like a variant-Rubix cube, solve the Chess AI once and you have solved it every time, because it doesn't actually think or decide, it simply permutates. Sure, large corporations have built dedicated Chess AI's that could trounce me, but how much processing power is required to achieve that? How much engineering behind the scenes? Thousands upon thousands of man-hours and millions upon millions of dollars and literally the only thing such an AI can do is solve chess positions faster than I can and with greater certainty. That's not intelligence, it's just very quick dumbness.

In the end the absolute major drawback of AI in this game is determinism. It must be client-client multiplayer compatible. This *massively* limits it's potential. There are technical reasons for this that are essentially inviolable from the Firaxis perspective, you'd have to approach it differently from the ground up. Won't happen, some improvements are certainly possible but nothing on the scale you are imagining. I'd be happy if they could just make it act consistently, which is within their gift.

 

[DizzKneeLand33]

A lot of things are simpler than rocket science, but that doesn't mean they still aren't difficult.

Chess is also a lot easier to handle than something like Civilisation; I detailed arguments for this the last time AI was a topic.

Exactly. You don't have to teach a bishop how to launch a bomber, for example.

 

[TheMeInTeam]

I meant purely in a logistical sense. If you have X time to develop a product, adding features will reduce the time available to spend on all features. Regardless of how much design overlap there may be.

Yes, that is the case. It should be obvious too, but it seems that most people making the evaluation of what they can realistically do fall victim to the planning fallacy.

An AI is not rocket science, above all in civ vi, that's much much more easy to instruct a civ to at least use air units then teaching the AI how to counter pro chess players, so you are out of order here sir

AI using air units a different standard than playing at "pro" levels. Chess is less complex by a wide margin, so it is easier to mimic/brute force elite play.

They cut corners with AI and it clearly wasn't a priority emphasis, but it's not even the biggest issue with the game.

In the end the absolute major drawback of AI in this game is determinism. It must be client-client multiplayer compatible.

AFAIK all AIs used are *technically* deterministic. If you mean that it can't react variably under absolutely identical conditions "because MP", I don't buy that as necessarily a crippling constraint, or even an expensive one to overcome.

Even if it were, that drawback is irrelevant right now, because the deterministic behavior presented is sufficiently weak that adding variants at random would leave "terrible vs terrible" as its decision options.

 

[JosephC]

Chess and Civ are very different when it comes to how the AI is programmed. To name a few differences..

Chess is far more restrictive in it's layout.
- The board is always exactly the same size
- The starting positions of every piece are exactly the same
- Each piece can only move in very specific patterns
- No "combat" stats, if you move onto another piece you win.
- One winning condition: Get King

- Because of these things a large portion of the AI is a lookup of openings


Civ
- The Map is not the same every game. The AI is not (fully) aware of the map, and therefore has to be programmed to dynamically account for the "board" they're playing on. Unlike chess, each hex has different attributes and AI needs to dynamically evaluate the attributes of each hex for whatever it's current goal is: movement, improvements, combat, etc..
- The starting positions have some rules, but are pretty random.
- "Combat" is far more complex: unit strengths, unit modifiers, hexes modifiers, modifiers of units around you, Killing (if I attack with this unit, do I have any other units around that could finish him off?), etc...
- Civ is (usually) has more than two players, all which can interact in several ways: Combat, Dipomacy, Expansions, City States, etc..
- Civ has more ways to win. Chess AI doesn't have to decide which victory condition is the best one to go for from my current stance, or in what I think I'll have in the future.


Now don't get me wrong, I'm not defending Civ's AI.. I just wanted to point out writing a good AI is more difficult than many of you think. Particularly for a game as complicated as Civ VI.

 

[callan]

AFAIK all AIs used are *technically* deterministic. If you mean that it can't react variably under absolutely identical conditions "because MP", I don't buy that as necessarily a crippling constraint, or even an expensive one to overcome.

They are not. In Computer Science terms, simply speaking, an algorithm is deterministic if it will always return the same result for the same set of inputs. This is true of the Civ AI and, to be fair, most (modern) gaming AI's. An alternative, stochastic, algorithm would not always return the same result. Many games, usually those focused on SP, do introduce an element of randomness into their AI, it's simply the case that these aren't the majority of games. Specifically in terms of Civ multiplayer though, each client is required to calculate it's own EndTurn. That means that each client must make every political decision and move every piece the same or your game desyncs. This is crippling to even the perception of intelligence.

Even if it were, that drawback is irrelevant right now, because the deterministic behavior presented is sufficiently weak that adding variants at random would leave "terrible vs terrible" as its decision options.

To a very large extent this is actually true, only your point about relevance is off, as it is relevant. Indeed it is the primary reason why I believe the current AI will never be able to deal with the new terrain limitations. There are simply too many additional movement scenarios available for the AI to be able to express a coherent plan of movement. In terms of implementing micro-strategy (tactics/war-plans) nothing will matter unless they can fix this.

 

[RealAntithesis]

People say, they like STRONG opponents ... but often it seems, many of them like 'just to win'.

Well sure - people like to win, but there's a balance to be had between being able to win versus having a challenging, fully engaging and fun experience while doing it. If a player loses, then it should be due to a recognisable and rational reason that the player can learn from for the next play through. Same as for when a player wins.

I think civ 6 is lacking this balance currently.

 

[c4c6]

Well sure - people like to win, but there's a balance to be had between being able to win versus having a challenging, fully engaging and fun experience while doing it. If a player loses, then it should be due to a recognisable and rationale reason that the player can learn from for the next play through. Same as for when a player wins. -- I think civ 6 is lacking this balance currently.

Agree 100%.

I wanted to feature different points of view, because some people here argue so monothematicly, that I feared reading it might immediately freeze my brain some day ... (sic!)

Undoubtedly having a challenging, fully engaging and fun experience AND crowning it with victory is all time best. Having to choose between both is NOT ONLY fun.
And not having the choice is worst.

To be clear: right now is civ6AI about as much too weak as good ChessAI is too strong for the mortal gamer. (for the reasons already laid out
and no, the lunatic arguments are NOT part of it. I just mention, cause I see that option "Show Ignored Content" again.)

 

[Jabulani]

Glad to see some of you likes Civ VI AI, good for you.

I dislike who is responsible to have commited such a crime against one of the best series in the world, that now has become a reason to laugh only, a comedy.

We will keep on voicing our opinions and the pathetic try to censure us will lead you nowhere, dear devs

 

[Gorbles]

Pointing out how it's different to programming chess AI is not the same thing as saying that we like it.

But please, continue to misinterpret our arguments

 

[MyOtherName]

That's really not true at all. True Artificial Intelligence is still decades away at a minimum,

True AI has been around already for a decade at a minimum.

It will be a while before we have a good human simulator, but that's almost never what you need from an AI. Nor is it even what you want -- humans are pretty bad at lots of things.

 

[Jabulani]

Pointing out how it's different to programming chess AI is not the same thing as saying that we like it.

But please, continue to misinterpret our arguments

I wasn't reffering to you mate

We have different opinions about CIv VI AI and chess but that's normal in a forum, anyway I was not referring to you, cheers

 

[TheMeInTeam]

They are not. In Computer Science terms, simply speaking, an algorithm is deterministic if it will always return the same result for the same set of inputs. This is true of the Civ AI and, to be fair, most (modern) gaming AI's.

I'm not sure it's fair to snag computer science terminology in a discussion with many outside that field of expertise. Even the AIs that don't always return the same result given an input have traceable causality/known reasoning for ultimately picking something from the algorithm.

That means that each client must make every political decision and move every piece the same or your game desyncs. This is crippling to even the perception of intelligence.

Why is it, in principle, difficult to force a synced algorithm that is not identical game to game? Something like an algorithm dependent on a seed with shared seed on turn start (if necessary, otherwise just game start) shouldn't be completely impossible? It might be impractical though.

To a very large extent this is actually true, only your point about relevance is off, as it is relevant. Indeed it is the primary reason why I believe the current AI will never be able to deal with the new terrain limitations. There are simply too many additional movement scenarios available for the AI to be able to express a coherent plan of movement. In terms of implementing micro-strategy (tactics/war-plans) nothing will matter unless they can fix this.

My point was that until they can make even a single deterministic decision process that is vaguely competent, the ability to create an algorithm that produces different results isn't useful. If each of the moves returned by the algorithm are ineffective junk, choosing between them results in ineffective junk whether you have 1 or 50 possible outcomes. In other words, if you can't get 1 "coherent plan of movement", what's the point of adding more plans of movement that aren't coherent?

 

[Gorbles]

I would imagine that generating something that critical on a per-turn basis could increase the risk of data corruption. i.e. a game model is a lot more robust than that because I'd imagine it's been designed in as static-a manner as possible (preloading assets, map generation, etc). If a single bit is flipped during transmission on any turn could result in a desync of that seed value. Multiplied per turns in an average game and you're massively increasing the likelihood of encountering an error.

This is based on no expertise of the game code itself and just the basics of network programming (I am very glad that it isn't my specific discipline, haha), and assuming that similar code isn't already in-place (which would negate greater concerns of affecting stability, but it would still follow they'd seek to minimise these kinds of volatile states).

 

[tekjunkie28]

As a person of a certain age, I have always loved board games, and yes, I was on the school chess team.
I even played war-games with little lead men.
With the PC revolution, came games where you could match wits with the computer, and the computer was generally pretty good, with the ability to quickly and flawlessly look-ahead many turns. For a long time, that was the only way to play a computer game, against the computer.
Then along came PBEM, and progress just kept rolling along.

Now, it doesn't seem to matter whether its Civ, or HOI, or EU, we seem to have headed back to a point at which the "game" is ALL about the multi-player experience. The so-called AI, really doesn't deserve the I anymore, it is basically a poorly implemented afterthought.

If I looked back at the changes, from say 4 to 6, everything has been done to enhance the multi-player experience, often at the direct expense of single-player enjoyment.

Realistically, Civ 6 can only be played with other humans manning the major civs, and the "AI" presence relegated to the City-States.

You sir hit the nail on the head. I always played board games when I was a kid but we didnt have a computer until i was 10 years old or so (will be 30 this year). When I was a kid we were always outside or whatever and then when I got into my teens about got a decent PC it was Starcraft, AoE 2, Homeworld and civ 3 for me mainly. I also miss those army men games soooooooooo much. I remember the AI then being MUCH better. I remember Jon Shafer saying something about how he wanted CiV to be multiplayer based and the AI was pretty much an after thought.

I also remember Sid Meiers talking about the psychology of gamers. He said something along the lines that if you make a game feel fun it is fun. Basically beating around the bush about how they can design a game that seems fun but truely isnt.

I find Paradox games have much better AI overall. Also I love the way they communicate with their players. Firaxis and 2K can learn from that.

 

[TheMeInTeam]

I would imagine that generating something that critical on a per-turn basis could increase the risk of data corruption. i.e. a game model is a lot more robust than that because I'd imagine it's been designed in as static-a manner as possible (preloading assets, map generation, etc). If a single bit is flipped during transmission on any turn could result in a desync of that seed value. Multiplied per turns in an average game and you're massively increasing the likelihood of encountering an error.

This is based on no expertise of the game code itself and just the basics of network programming (I am very glad that it isn't my specific discipline, haha), and assuming that similar code isn't already in-place (which would negate greater concerns of affecting stability, but it would still follow they'd seek to minimise these kinds of volatile states).

In principle, you could still random seed on game start (it would be an extra step akin to map generation, and similar to sync to that I'd imagine), and have the algorithm pick between options based on that seed. Depending on how it's set up it would be pretty darned hard for a human player to guess precise actions by the AI in advance that way, since you'd never know exactly what the seed dictates the algorithm do far that particular state without knowing both the seed and algorithm. If you knew the algorithm you could still know possible choices made in that scenario, but the game has enough possible scenarios that this would be impractical for almost everyone.

However, this *still* requires more than one viable decision process to exist in the first place, as opposed to zero. Given most people's ability with the game, I would expect a predictable, but generally good decision process done consistently by the AI to be less expensive and result in more challenging gameplay to the majority of players, even if it was deterministic in the computer science sense he used.

I still stand by the position that if you can't even make a single good decision consistently a setup where multiple choices by an AI is possible doesn't matter and won't matter until it can make at least two potentially good decisions .

 

[Gub]

I also remember Sid Meiers talking about the psychology of gamers. He said something along the lines that if you make a game feel fun it is fun. Basically beating around the bush about how they can design a game that seems fun but truely isnt.

Fun for whom? With the ever growing quantity and quality of good games that are competing for our limited time, "casual" gamers have made up the vast majority of players for a while now. And I doubt that they share my preferences or I that of a competitive players.

Personally, I don't care for steam achievements, competitions, nor memorize every yield\bonus in game and read about best strategies. I like to understand the RL reason behind game concepts, try\fail different strategies and mostly RP a story in my head. I also like to read others AAR and spice up my games with different mods. Wanna bet that I am in even smaller minority even compared to those who like MP..

I find Paradox games have much better AI overall. Also I love the way they communicate with their players. Firaxis and 2K can learn from that.

I think that the fixed game world and the periodic nature of their games, lends it self better to give the impression of a good AI. That and with the steeper learning curve it takes people longer to figure out how the magic trick is done.. but overall its AI just as atrocious. Recently I seen of those "lets play" by quil18, where he EASILY won on normal settings, despite making HUGE mistakes, ignoring mechanics and overal using the strategy of push/push/push..

Besides even Paradox ,now that they carved out their niche, have been streamlining their recent games, making them more accessible not increasing their complexity. (there are actually much more in-depth, micro-management and simulation heavy games, even with PBEM, if you don't mind playing a "spreadsheet" )

...

As for your thread title, I think its as accurate as saying that Books have gone Full-Circle.. in every medium there are those who push the envelope, maybe its just that you got set in your ways\tastes