So.... A.I. why hasn't been any info, focus or DD about it?

[aruon]

it's highly likely that the AI will either have a DD very near release, or simply not at all. like others have said, just how do you show that ingame? plus it will be an even evolving thing as the game gets more advanced, and as HOI 3 shows, creating even a half-decent AI with a game like HOI with all its different focuses is no easy task.

 

[justin6477]

How will they show that?

A good AI won't do those things on command. They'll have to play a bunch of games and get a scrapbook of the AI being smart and then explain the smartness when they show the screenshot.

Remember that time Ethiopia was able to push its way into Italy and annex most of the peninsula as well as a random enclave in the north? I'd say that's some pretty decent maneuvering on the AI's part.... well, at least Ethipia's.

 

[SteelVolt]

As the primary AI coder on HoI 4 I will try to shed some light on this to try to reduce the ammount of nervous speculation.

Some various responses to topics in the thread:
* The AI was not the main reason for the delay, but rather one of several things that needed more work. The game was simply not ready when we had initially hoped it would be.
* Comparing the AI of two different games will never be fully fair.
* It is most assuredly easier to build the AI around a more or less complete feature than guessing based on design. Features may change over iterations and that means the AI will have to change along with them. As such, the AI will (by necessity) be in development as long as the game is, but it is folly to start working on it too early.
* So far it has been stated that the AI will use the same battle plan system as is available to the player, so there is that piece of information
* Would you really like to know the details of the AI? Wouldn't that kind of be like having a human opponent write down how they intended to play the game before hand?
* "Optimal" decisions is not such a clear cut concept as it may seem in a complex game like HoI. If the AI consistently makes the same decision percieved as optimal, it will become predictable (and would thus no longer be optimal.)
* Of course we will do what we can to make the AI as good as possible, do we really need to state that? Currently we have one coder dedicated only to this, and the legendary Wiz as overseer.

And indeed, if there will be an AI DD, it would be reasonable to put this rather late in development, especially if it were to relate in any way to the features of the game (which it would be making decisions about.) I admit, though, that I have not seen the intended DD topics list myself, so I guess Johan or Podcat might surprise even me on this.

At any rate, it feels good to see that people care about the AI. Rest assured that we do as well

Cheers.

 

[Gamer_1745]

@SteelVolt
Thanks for the info!

I would like it if you can comment on this idea to if it would be do able with HOI IV at some time for release or later in a DLC:

What I thought would be a nice feature is that players could make Battle Plans for a country and save them to a file. Things like 'The Defense of Russia 01', 'What France Should have done!' or 'Japan's Seaborne invasion of Southern China'. These would then be loaded into the game, and they would give options for the AI. As in Soviet AI (not a player playing the Soviet Union) would check the situation (threat levels on which border, size of his army, other things) and look at it's filed Battle Plans for ones matching the situation and say it has 5 plans that fit the situation and randomly (some plans may have a higher % than others) picks one. When the objective is achieved or the situation fails apart the AI goes back and looks for the best plan(s) that match the current needs and picks one. This way you have Humans making the plans and with the random chance the player doesn't know which one. If the AI has 10 different plans for the defense of France, the player does know which one the AI picked.

As time goes on players post their plans for others to download to their game so eventually there would be many plans covering many different situations. At this point I am thinking this could be a DLC (not in the game as released). What do you all think?

Yep, that is why I said thousands of plans!

SU has X number of division, what are the current borders, what countries are in the Axis, all these things.

With the current system somethings the current system just doesn't see.

Let me give you an example. Germany is Axis, Poland is partitioned between the Soviets & Germany and Lithuania is neutral and no threat to the Soviets. The Soviet AI would not defend the border with Lithuania so the Germans just march though and outflank them. But with this system and the above conditions (yes I know there would be many others as well) the AI might find 5 plans that meet the conditions and picks one randomly. Three of the five might include units along the border with Lithuania, the AI wouldn't understand why, it would just do it.

I fully understand if it would not work with HOI IV! I just don't want to keep pushing an idea that just wouldn't work!

 

[SteelVolt]

I think it is an interesting idea, at least to some extent. Since the plans would require rather specific conditions to be valid at all I am affraid it would not fit very well with how dynamic HoI 4 is, though.

The time when such a model would be interesting would be when wars start, before countries or borders have had a time to change. After that so much can happen so differently that it would require an unreasonable ammount of scripted plans. I would judge it to be more worth the time making an AI that can think on its feet dynamically rather than going down the road of scripted plans.
But I guess if we go down the road of more detailed campaigns it would be worth looking at, but unless it turns out scripted plans are necessary for a tutorial I do not see this happening in a foreseeable future. The step from that to matching a good selection with a certain situation is kind of a big one though.

Thanks for the input though, and sorry for having to bring bad news.

 

[Vonboe]

I think it is an interesting idea, at least to some extent. Since the plans would require rather specific conditions to be valid at all I am affraid it would not fit very well with how dynamic HoI 4 is, though.

The time when such a model would be interesting would be when wars start, before countries or borders have had a time to change. After that so much can happen so differently that it would require an unreasonable ammount of scripted plans. I would judge it to be more worth the time making an AI that can think on its feet dynamically rather than going down the road of scripted plans.
But I guess if we go down the road of more detailed campaigns it would be worth looking at, but unless it turns out scripted plans are necessary for a tutorial I do not see this happening in a foreseeable future. The step from that to matching a good selection with a certain situation is kind of a big one though.

Thanks for the input though, and sorry for having to bring bad news.

Just wanted to say it's awesome to see a dev actively responding to a constructive idea like this, looking forward to see more posts from you in the future SteelVolt

 

[Cpack]

Some things I would like to see the AI doing is reacting differently on Players behaviour.

F.e. example Germany is going to build massive Bombers, AI will react in building more fighters, Germany is pushing out Subs, AI builds more DD's.
Also the opposite, Germany is building up a big fleet of CV's and BB's, AI will react in skipping DD's and build more naval Bombers and surfave fleet.

Or british spys found out Germany is building massive Landing ships and transports after the fall of France, UK AI starts massively in defending the british coasts......

These things didn't work very well in previous HOI games

 

[DINOSAURS1986]

I feel one of the reasons I think A.I. is usually not that good in games is that instead of developing the game around the A.I. it's always developed around the player and the player experience.

That's the impression I get too, though I think designing a game around A.I. might end up dull since most A.I. seems pretty basic. I find the A.I. in Paradox games less jarring than something like a Total War game though. Starcraft II A.I. is pretty good, but that's because it can follow strict build orders in fairly tight environments; and even there it fails once it gets to tactical or dynamic situations.

C-Evo is a freeware 4X game that aims to have pretty smart A.I. that might suit you: http://www.c-evo.org/text.html

Principle 3: AI Liberation. Empire building games are typically asymmetric. They are built around the human player as their center, with some pseudo-AI mainly having the job to keep him amused and to make the whole thing a realistic simulation. C-evo, in contrast, is a competition of equals. AI has no jobs, because that would reduce its strength. AI just has a goal, which is the same as the player's goal: to win. All are playing by the same symmetric rules, no matter if human or AI. Frequently made suggestions show that many players do not fully realize the consequences of this principle (which is forgivable because the games they're used to are far away from it.) Particularly, there is no way to direct the behavior of the AI - the AI is as free as the player is. Rules and AI are strictly separated.

I find it hard to go back to turn-based systems since I found Paradox games so I haven't dived into, but the A.I. beat me the couple of times I ran it.

 

[SteelVolt]

Just wanted to say it's awesome to see a dev actively responding to a constructive idea like this, looking forward to see more posts from you in the future SteelVolt

I do try to pop my head into constructive threads that I have relevant input to. Nice to know it is appreciated.

Principle 3: AI Liberation. Empire building games are typically asymmetric. They are built around the human player as their center, with some pseudo-AI mainly having the job to keep him amused and to make the whole thing a realistic simulation. C-evo, in contrast, is a competition of equals. AI has no jobs, because that would reduce its strength. AI just has a goal, which is the same as the player's goal: to win. All are playing by the same symmetric rules, no matter if human or AI. Frequently made suggestions show that many players do not fully realize the consequences of this principle (which is forgivable because the games they're used to are far away from it.) Particularly, there is no way to direct the behavior of the AI - the AI is as free as the player is. Rules and AI are strictly separated.

I agree with this sentiment, though I think the kind of "pseudo-AI mainly having the job to keep him amused" it talks about is relatively rare these days (the project is rather old, after all).

 

[henzington]

The statement is mostly true but computers are quite bad at handling situations with incomplete information where as a person can guess.

 

[Art1985]

Interesting to know how friendly will be for modding community? (What scripts will be available for moders? I know it's too early to say for sure, but any thoughts on this(plans)?).

 

[SteelVolt]

The statement is mostly true but computers are quite bad at handling situations with incomplete information where as a person can guess.

Indeed. Which is why it is so common for AI to cheat by having access to more information than the player, even if it plays the game by the same rules beyond that (rather than moving around setpieces just to give the player a certain experience). I would say it is more of a handicap to compensate for inferior pattern recognition rather than "psudo AI".

Interesting to know how friendly will be for modding community? (What scripts will be available for moders? I know it's too early to say for sure, but any thoughts on this(plans)?).

It is indeed too early to say, but it is certainly not 100% either way (code vs script).

 

[Had a dad]

Thanks SteelVolt for the information, and look forward to hearing much much more from you

 

[Deathmachinept]

@SteelVolt
Thank you for commenting... Having no insight in such a crucial feature felt bad, appreciate the clarification.

@SteelVolt
But are you significantly departing from how previously the A.I. was coded or implementing a new A.I. technology has is being said or are you doing more polish and smaller improvements over the previous A.I. systems?

1- As for a DD about the A.I. I would like to see some light if there are any significant improvements or new approaches you are trying in HoI 4?

Will this make A.I. better at combine arms and more complex battles like doing an invasion with paratroopers etc..

Not sure if you have them but sometimes there are some paper/documents of A.I. designs that show the complex behavior and relations on what the A.I. is suppose to have, sometimes people use some tools that allow you to see a visual representation of code and how it is related more with lines clouds etc..

2 - I would also enjoy some background on how you develop as in tools, design decisions cooperation with coders... etc something that would be useful for people who are trying to get into the industry and would like some more information what goes behind the development.
(like it was done with modelling the soldiers in High Poly then putting into a texture and using as a skin to make a low poly model look as a high poly)

Thank you,
DeathmachinePT

 

[podcat]

We'll have some kind of AI dev diary for sure (to go over how it can be modded etc for one). But this is really one of those things that needs to come late for the reasons SteelVolt mentioned.

 

[Deathmachinept]

We'll have some kind of AI dev diary for sure (to go over how it can be modded etc for one). But this is really one of those things that needs to come late for the reasons SteelVolt mentioned.

Thanks for the info,
But it would be cool having an early DD more about what goes behind the scenes on how the A.I. is developed, then the actual A.I. that will be on game, like the DD about 3D models? That would be cool, good A.I. is the holy grail of Singleplayer games.
I also enjoy and like to learn how programmers and different companies differ in development methodologies.

 

[gnome109]

Thanks for the information

 

[Gamer_1745]

@SteelVolt

Thanks so much for the response! I don't look at is as bad news, it is just what it is a clear statement of what is likely to be done with HOI IV. I don't want to push on ideas that can't be done within reason with HOI IV. If you do consider something like this I could have much more input than I made in those posts!

I just hope with HOI IV the days of the Poles defending every border province with Germany are gone! I hope they will make a sensible defensive plan with shorter defensive lines & on better terrain (rivers & things).

 

[ray4ever]

Thank you for all these answers!

As the primary AI coder on HoI 4 I will try to shed some light on this to try to reduce the ammount of nervous speculation.
* Would you really like to know the details of the AI? Wouldn't that kind of be like having a human opponent write down how they intended to play the game before hand?
Cheers.

Maybe not, but something in between. Wiz did one great post about the eu4 ai where he stated what the ai knows and what bonuses it gets (where it "cheats") and in which areas the ai and human players are on the same "information level".

 

[Krafty]

AI is bad. AI has always been bad. AI will always be bad.

Empires Deluxe, after hundreds of people have gone at the open sourced title (AI programming experts) over the past 15 years...the AI is still woefully inadequate.

HOI4s AI will be bad. Its just a fact of life. The AI that controls your units, and the AI that controls the other nations units, will be terrible. (terrible on a scale of a rock, to a human, all AI will fall closer to the rock, than the human)