Why are people not talking about the AI: Artificial Intelligence? This is the really important issue.

[Humble Warrior]

Or programmed routines or whatever you choose to call it.

I believe this is the biggest and most important issue that since at least Empire has been put on the back burner for far too long and only with great effort was it used to fix sieges somewhat in RTW2 (where the AI was dead).

It`s not factions that make TW, or unit colours or religion. It`s not how good the game looks or how great it sounds.

It`s how well the AI handles the whole game from the campaign, to battles to sieges. AI is the net that weaves and works among all of this. CA are very lazy with this because they (whoever organises priority) thinks people aren`t interested in a good AI. People really need to focus more on getting CA to improve the AI in all aspects of the game, from the Campaign map to the ground battles to how it conducts sieges.

Without a good AI there is really no game, just a collection of audio, skins and textures that move around blindly.

We have to let CA know that this is important or they will not make the effort: they`ll just use the last AI and band-aid everything else around it.

[myself]

Agreed. AI is one of the most important aspects of single player campaign. It's what adds replayability to the game (if done correctly).

[Butan]

Ultimate General: Gettysburg, to only cite one example, made it clear that today, in 2014-2015, Artificial Intelligence can be better than this.
The argument of limited technologies and programming barriers doesnt stand. With proper effort and time and investment, the AI can become really something else in the Total War brand. The fact that the vanilla AI has pretty much stagnated for a decade now, can be put down to the fact that other department than AI development took more effort and weighted down on this crucial part of the game, but nothing else. The graphism quality and number of frames attract casuals for a short time, but the AI (and, to replace it, the multiplayer options) is what makes the players stay for a long time.

Does CA just want to have a sale peak at release, be a commercial success and sell as much as possible in very concentrated burts, while 90% of the players just stop playing before even finishing one short campaign, or does it want to be a critic success and give people reasons to play the game they developed to its end?

I would also like CA to give more informations on what the AI can do that it couldnt before, and players to discuss about the shortcoming of previous games. Because I also feel like Humble Warrior that is a very important part of the game.

[Randall C. Reed]

I'm with you. But care must be taken to not OVER-RATE AI. It still must be placed in the context of a total game engine design. While user interface enahnces the player experience, the game engine will hold them, and that includes a carefully crafted AI code set.

[geala]

Totally agreed.

[crucify_ego]

Cause it's the same game as Rome 2

[RedGuard]

Cause it's the same game as Rome 2

this

[alQamar]

Cause it's the same game as Rome 2

this

agreed, but this doensn't mean there is any hope, nor that any propagandist should be allow to state hollow shells like "AI was always an issue in TW". I am quite sad that region exchange isn't in place in Attila too despite all the nice overhauls.

This maybe also a reason to cover, that AI might have serious issues with such a feature and alike senseless would trade regions as they declared war and offered your thousands talents and trade agreement the very next turn. Yes even this exists in patch 16.1 - so a valid post again Humble Warrior and not def. not repeating the same old.


I would like to cite Sun Jetzu in this context:

Antworten
·



Ryan Karlson

vor 1 Tag

So what about the A.I.? This the biggest concern. Pathfinding, unit collision, and A.I.'s ability to disperse and engage in a logical manner takes priority over anything else. Is that not a concern for you, can't believe you didn't mention it once. Your units just stood around while the rest fought one on one battles after engagement. They should spread out a bit and surround the enemy, you know like would happen in real battle. not impressed, doubtful this will change by release. 




Antworten
·
2



Sun Jetzuvor 1 Tag (bearbeitet)



tbh I like the combat in rome 2. And since attila has similar combat, Im a fan. I dont bicker about the one on one thing too much. I did answer this in an interview though. The battle AI seems to be similar to rome 2's for the most part. The only faction that seems to be different is the huns. They literally just rush you with all their cav and break your lines. Watch my hun video to see. The campaign AI is very impressive. They use logical stances, they avoid large forces. If they are outnumbered they will pull back and return with reinforcements. They are far better than in rome 2. Always remember, this is a development in progress. 

[Cope]

All you need to know is that torches are still in the game and the AI is still incapable of defeating double layered walls (despite the fact that they are in the game). This pretty much indicates that the AI is still going to be as incompetent as it was in Rome II...because as someone has already said, Attila is the same game as Rome II.

[Butan]

Always remember, this is a development in progress.

I hate to see that typed everywhere, 1 month away from the release

[DeliCiousTZM]

The AI will be as good as they can make it, so unless you have any hidden coding skills this thread is useless.

[Symphony]

The AI will be as good as they can make it, so unless you have any hidden coding skills this thread is useless.

This is an entirely ludicrous statement.

I am not an automotive engineer. However, I can expect a car I purchase to have adequate safety features, usable convenience features, and to last more than 20,000 miles. I am not a professional-grade chef. However, when I go to a restaurant, I can reasonably expect the food to taste good. I am neither a fashion designer nor a tailor, but I can expect the clothes I buy to look decent and wear serviceably.

We're consumers....our entire role is to purchase products that we don't have the skills to make for ourselves. The role of the producers of those products is to provide quality goods that we want to purchase. I'm not entirely certain where and when video game production received an exemption from that concept.

[crucify_ego]

Really? Why does shogun 2 have better A.I.?

[DeliCiousTZM]

Really? Why does shogun 2 have better A.I.?

It doesn't, Rome II AI is way better then Shogun 2, you just need a trip down memory lane. And Shogun 2 is a super simple game, Tetris compared to Rome II.

[Butan]

It doesn't, Rome II AI is way better then Shogun 2, you just need a trip down memory lane. And Shogun 2 is a super simple game, Tetris compared to Rome II.

[Fredrin]

It doesn't, Rome II AI is way better then Shogun 2, you just need a trip down memory lane. And Shogun 2 is a super simple game, Tetris compared to Rome II.

This is gold! And this:

The AI will be as good as they can make it, so unless you have any hidden coding skills this thread is useless.

There is a bright career for you in the niche world of TW-related comedy, my friend!

[Mithridates II the Great]

It doesn't, Rome II AI is way better then Shogun 2, you just need a trip down memory lane. And Shogun 2 is a super simple game, Tetris compared to Rome II.

It does ,

In Shogun 2 the AI attacked from 3 and sometimes 4 directions which was never seen in Rome II .

In Shogun 2 the enemy attacked you from behind and rear but after 16 updates we barely see them in Rome II .

CA had trust on Shogun 2 that gave people a demo but you can't see that in Attila or Rome II cause they know what this game is .

[PointOfViewGun]

If anyone at CA could write an AI code that is capable of impressing over a decade long raised expectations of this game's fans then they'd be working for DoD. Nonetheless, I would say that the AI of Total War games have been getting more complex over the years. Whether you like it or not is a different issue. The AI of older games might look better on screen, not really because they were better, but because they were simpler, less variables, less options, less outcomes. Less chances exists for the AI to screw up.

Plus, a lot of points expressed here have to do with game design and nothing to do with the AI in the firs place.

[Symphony]

If anyone at CA could write an AI code that is capable of impressing over a decade long raised expectations of this game's fans then they'd be working for DoD.

Oh, rubbish.

Look, I'll agree that coding AI isn't as easy as some people make out. In particular, bug-hunting in AI code is generally more onerous than in other coding exercises; more variables that interact with each other, and multiplication of those interactions as they pile up the further down the decision tree you get, make it that way. So, it generally takes more trial-and-error iterations (and therefore more time and more failures) to fix unintended behavior that gets introduced. This is why I generally don't jump on the bandwagon that screams when an AI bug persists through several patches, or when things are still broken after patch notes say it should be fixed; I get that it takes more time.

....but it's not rocket brain science surgery, either. At the end of the day, code is just code; AI code is just "bigger" (it stacks in a more complex way). TW has always had a sub-par AI, and its quality (in a way that is noticeable and enhances the quality of the gameplay experience) hasn't progressed significantly since R:TW. What was understandable in 2003 becomes less so in 2015. Worse, it has regressed in some areas since then (pathfinding sticks out as the best example). Some of that is undoubtedly due to new feature implementation (CA trying to get the AI to do new and different things), and that's good of them to try, but not only can that account for all of it, CA would have been better off leaving those features out if the cost is a degradation in AI performance....I agree with Humble when he calls it the core of the game. If a clumsy workaround like torches are the only workable way to solve siege pathfinding issues, new AI features should have been stripped out until the workaround was no longer necessary.

The relative quality of AI in other games shows us that it doesn't require a military-grade understanding of AI to code something competent. I don't expect a flawless AI. I don't even really expect to be pushed by an AI; I'm human, and it's not....I'm always going to be smarter. But we can certainly expect more than what we got with Rome 2. This is a strategy game; laughable, easily exploitable AI shouldn't even be something we have to talk about.

I would have been happy with the AI in R:2 if it could have reliably replicated the actions of actual classical armies. This is not hard to do; history provides the model, and you simply have to make your AI do what somebody else did before. That would allow me, as a live human being, to react to that behavior and try to beat it. I'm not even expecting the AI to be so smart that it counter-reacts intelligently....just script the initial maneuvers to reliably simulate classical tactics, and I wouldn't be writing this post.

To the OP: People aren't talking about it because this is a TW project, and my personal expectation is that we'll see more of the same sub-par AI we've grown accustomed to.

[PointOfViewGun]

Oh, rubbish.

Look, I'll agree that coding AI isn't as easy as some people make out. In particular, bug-hunting in AI code is generally more onerous than in other coding exercises; more variables that interact with each other, and multiplication of those interactions as they pile up the further down the decision tree you get, make it that way. So, it generally takes more trial-and-error iterations (and therefore more time and more failures) to fix unintended behavior that gets introduced. This is why I generally don't jump on the bandwagon that screams when an AI bug persists through several patches, or when things are still broken after patch notes say it should be fixed; I get that it takes more time.

....but it's not rocket brain science surgery, either. At the end of the day, code is just code; AI code is just "bigger" (it stacks in a more complex way). TW has always had a sub-par AI, and its quality (in a way that is noticeable and enhances the quality of the gameplay experience) hasn't progressed significantly since R:TW. What was understandable in 2003 becomes less so in 2015. Worse, it has regressed in some areas since then (pathfinding sticks out as the best example). Some of that is undoubtedly due to new feature implementation (CA trying to get the AI to do new and different things), and that's good of them to try, but not only can that account for all of it, CA would have been better off leaving those features out if the cost is a degradation in AI performance....I agree with Humble when he calls it the core of the game. If a clumsy workaround like torches are the only workable way to solve siege pathfinding issues, new AI features should have been stripped out until the workaround was no longer necessary.

The relative quality of AI in other games shows us that it doesn't require a military-grade understanding of AI to code something competent. I don't expect a flawless AI. I don't even really expect to be pushed by an AI; I'm human, and it's not....I'm always going to be smarter. But we can certainly expect more than what we got with Rome 2. This is a strategy game; laughable, easily exploitable AI shouldn't even be something we have to talk about.

I would have been happy with the AI in R:2 if it could have reliably replicated the actions of actual classical armies. This is not hard to do; history provides the model, and you simply have to make your AI do what somebody else did before. That would allow me, as a live human being, to react to that behavior and try to beat it. I'm not even expecting the AI to be so smart that it counter-reacts intelligently....just script the initial maneuvers to reliably simulate classical tactics, and I wouldn't be writing this post.

To the OP: People aren't talking about it because this is a TW project, and my personal expectation is that we'll see more of the same sub-par AI we've grown accustomed to.

It's exactly rocket science or brain surgery. We're not talking about a calculator level AI. We're talking about an AI that has to consider location and proximity to other allied units, type, strength, morale and fatigue of any enemy or allied unit. Then the AI has to consider a lot of scenarios to act upon. Most games out there, if not all of them, utilizes much fewer variables when their AI operates. The most advanced AI, based on articles from 2013, ConceptNet4, has an IQ equal to that of a 4 year old.

Sub-par means that something is below the standard. What standard is Total War's AI below?