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

[Deathmachinept]

There seems to be such an enormous focus on player accessibility and better graphics for immersion.

But A.I. has never been mention except for helping again the player with accessibility...

1 - Is it wrong for me to think, that if HoI 4 devs were proud of their A.I. or it was a important factor for them, it would already been mention at least in the first DD on the original vision of the game?

2 - Was it a factor for delaying the game?

At least when it comes to my personal experience, I prefer a more scripted A.I. that only does the most efficient decisions, ignoring what doesn't make sense or contributes to a faster progression at least on the higher difficulties.

Which at least compared to previous HoI the A.I. gets artificial advantages that feel cheap and too much like cheating.


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.
So being one of the most complex features implementing new mechanics and features ingame should first answer to A.I. while the player experience would be just the final consideration.
(I'm not saying it's would be the main factor but every gameplay feature should first answer to A.I.)

As for the best A.I. or most challenging A.I. that I have encounter it was in Age of Empires 2:
http://ageofempires.wikia.com/wiki/Artificial_Intelligence
(it kicked my ass even in normal difficulty without artificial bonus like in HoI at least until the last difficulty setting)

Thank you,
DeathmachinePT

If anyone has any info pls share...

 

[uberjedi]

Well, considering how bad the AI in hoi3 was I would accept a delay to make it better.
For example:
* US AI invading mainland Europe but doesnt have sense enough to get a good port->Army dies of starvation
* German AI moving units back and forth between northern Finland and Bulgaria through Sweden and thereby not defending anything
* Any AI low on manpower. Still doesnt delete crap units to fill the depleted good ones with more men.
and so on..

 

[Denkt]

Scripted ai is not the way to go and the developers have said that HOI4 will not use a scripted ai unlike HOI3.

 

[novapaddy]

Well actually the AI is most likely going to be VERY SMART!

If you watch the battle planner being used in the published video recently, first the troops move to the first line of battle, but there is then a 2nd line placed by the player.

SO, not only does the AI have to fulfill the first task given to it, it also has to fulfill a subsequent task. I mean there must be an AI and then a sub-set of that AI working for the battle planner to function?

I THINK the AI is going to be very smart?

I have some experience with the darkest hour AI and getting the AI to do anything can be quite a demanding task, naval invasions for instance comes to mind. They work very well, but only when the prerequisites are in place (marines and transports... and a beach).

 

[Deathmachinept]

Scripted ai is not the way to go and the developers have said that HOI4 will not use a scripted ai unlike HOI3.

Well to have the advantages of unscripted A.I. it takes a lot of time to develop, it can be resource intensive and complex to make small changes to account for different features.
I do hope they started on it well before announcing HoI 4 because even 2 years is too optimistic. And I still have to find a not scripted A.I. that is challenging in a RTS or Turn Based strategy game with no artificial bonus.

Still think you can't have a good A.I. without having some scripted decisions..

Well actually the AI is most likely going to be VERY SMART!
If you watch the battle planner being used in the published video recently, first the troops move to the first line of battle, but there is then a 2nd line placed by the player.
SO, not only does the AI have to fulfill the first task given to it, it also has to fulfill a subsequent task. I mean there must be an AI and then a sub-set of that AI working for the battle planner to function?
I THINK the AI is going to be very smart?
I have some experience with the darkest hour AI and getting the AI to do anything can be quite a demanding task, naval invasions for instance comes to mind. They work very well, but only when the prerequisites are in place (marines and transports... and a beach).

Is there even any gameplay footage that shows PC Enemy A.I. ?

In some of the lets play.. it even appears that the A.I. is having some troubles moving units to the lines that were marked, you see quill18 giving multiples orders on the battle planner.


@uberjedi
Did A.I. Managed to make real amphibious invasions with paratroopers and airsupport.. I think initially A.I. wasn't able to do Amphibious invasions and by later patches they could but these didn't have enough power behind them.

 

[Lieftastic]

What do you want them to say about the AI? "HOI4's AI is better than in our past games. You'll see when you play against it. Here is a block of code that 99% of you will find inscrutable. See you guys next week!" You can't really write a DD on AI.

 

[Hans_Schnitzel]

What do you want them to say about the AI? "HOI4's AI is better than in our past games. You'll see when you play against it. Here is a block of code that 99% of you will find inscrutable. See you guys next week!" You can't really write a DD on AI.

Well, they could show us examples of what the AI does now that it didn't do in HoI3, like coordinate better or use strategies like pockets and encirclements more often, what do I know.

 

[Bugnr01]

The AI is always difficould, it is nearly impossible to programm a unskripted AI for a strategygame that is smart enouth for all the players.
I hope it will be realy smart, but you cant add with the AI. It will always be too weak for some players, to scripted or to crazy for others, maybe too difficould for new players.
It dont matters how you anounce the AI at the ende there will be a lot of unhappy people. So just be quiet and work on it, improve it as much as possible.

Dont forget, we need a AI that makes misstakes (misstakes are also variety)! Or would you like to play aganst an AI which knows the optimal way to play the USSR (and plays it all the time in the same way because its the most gamy way).

 

[GhengisKhan]

In some of the lets play.. it even appears that the A.I. is having some troubles moving units to the lines that were marked, you see quill18 giving multiples orders on the battle planner.

He kept forgetting to execute the new plans.

 

[Deathmachinept]

The AI is always difficould, it is nearly impossible to programm a unskripted AI for a strategygame that is smart enouth for all the players.
I hope it will be realy smart, but you cant add with the AI. It will always be too weak for some players, to scripted or to crazy for others, maybe too difficould for new players.
It dont matters how you anounce the AI at the ende there will be a lot of unhappy people. So just be quiet and work on it, improve it as much as possible.

Dont forget, we need a AI that makes misstakes (misstakes are also variety)! Or would you like to play aganst an AI which knows the optimal way to play the USSR (and plays it all the time in the same way because its the most gamy way).

If you check that Age of Empire AI link you will see on hard it will play the optimal way, while on lower difficult sometimes it waits for players to advance. But I would like that on normal it had a chance to make a mistake, some chances to play the optimal way but if I choose hard I really want just the optimal way.

 

[mdw1985]

If i read the last dd from johan, it seems, that the game crashed quite often. He said nothing about the AI, which could mean nothing or a lot. So the AI thing seems to be speculation at this point. But as the devs have plenty of time now to build the game and tune it, i expect a very good AI, when the game is released. I guess we can read a dd on this matter before (i guess it will be one of the last before the release).

The AI if AoE2 was awesome...no...the whole game was awesome. spend over 200 with it and still play it with friends on multiplayer

 

[Deathmachinept]

If i read the last dd from johan, it seems, that the game crashed quite often. He said nothing about the AI, which could mean nothing or a lot. So the AI thing seems to be speculation at this point. But as the devs have plenty of time now to build the game and tune it, i expect a very good AI, when the game is released. I guess we can read a dd on this matter before (i guess it will be one of the last before the release).
The AI if AoE2 was awesome...no...the whole game was awesome. spend over 200 with it and still play it with friends on multiplayer

What worries me is that on the 1st DD about the vision nothing is said about the enemy A.I. maybe they didn't want to hype since they weren't 100% on the launch day it would be better than the older A.I. Opponents.

If indeed HoI 4 will have less or no scripted A.I. this would have been a big step forward and still no mention?...

Total War Rome 2 was also trying to implement a new A.I. system but just see what it got them even with 60% more budget for A.I. it still ended in all that mess. And this was always one of the main problems about Total War Series and I recalled never hearing any real information about those features that were the most important to nail it right, actually AJ was the only one to question about Diplomacy and A.I. of all interviews I saw..

Still it would be cool to know if this was the main reason behind the delay.

 

[ringhloth]

Doing a DD on the computer opponents isn't really possible, AFAIK. It's all somewhat abstract. DD's show off features of the game. You can't really show off the quality of the computer opponents. Definitely not in text. It's difficult to do over the course of an entire game. However, EU4 has one of the (if not the) best computer opponents in the genre. PDS has steadily improved the quality of the computer opponents, and I don't think HoI4 will be any different.

 

[Gunnarr]

Biggest test for AI:

Can it do a naval invasion effectively

That is what they need to prove.

 

[xtfoster]

Without even bothering to read the rest of the thread I will just say this.

1) They just started the Alpha, so while everything is there, it isn't complete.
2) Any AI is going to require almost everything else to be mostly complete before it can be finished.
3) We are 6+ months from Beta testing (IIRC), so they have plenty of time to finish everything that needs to be done before the AI can be finalized.
4) Paradox DD policy is (and has always been) that they will not do a DD on a feature that isn't complete (although they have done some where features were changed anyway (transport planes anyone).
5) So expect any DD going into detail on the AI (if any) to be very late in the schedule. They might do one on the basics of the AI sooner, but don't count on any detail.

 

[BordZ]

This ai talk is all very interesting but if you want to see ai done right look at gal civ 2. Easily the best ai ever done, we should be comparing HOI4 to this. As stardock found paradox will find the ai development an iterative processes as us humans find and exploit holes in the underlying logic.

On that note I expect the ai to be okay when the game comes out, but in a couple of months it will be lack luster because we will have worked out how to game the system. At this point paradox either has the option of patching the exploits out or alternatively as stardock did with gal civ improving the ai to enable it to identify and move to counter these strats.

 

[Taciturn Scot]

You won't really have much of an AI, never mind be able to talk about it, until you have all the systems working more or less properly and the game has just entered Alpha which means that it is far from feature-complete. If I understand correctly, Johan doesn't expect the game to go into Beta much before the end of the year so I'm not expecting to hear much about the AI until then. What's in the game right now is likely rudimentary and is able to pass off against newcomers playing on high speeds.

I don't have high expectations of a good AI military opponent from Paradox anyway. Their games are incredibly complex as is and while the AI does a remarkable job of handling al these various systems, their combat AI is merely passable at best and I see no reason whatsoever to expect it to be better than it currently is in EU4. In a combat-centric game like HOI, the AI opponent isn't going to offer a human much of a challenge unless you allow the AI to manage your forces for you. This game is going to be more about putting together a good plan for me that HOI3 was.

And you might want to try 'Flashpoint Campaigns: Red Storm' or 'Command Ops' for good AI combat opponents.

 

[Deathmachinept]

This ai talk is all very interesting but if you want to see ai done right look at gal civ 2. Easily the best ai ever done, we should be comparing HOI4 to this. As stardock found paradox will find the ai development an iterative processes as us humans find and exploit holes in the underlying logic.
On that note I expect the ai to be okay when the game comes out, but in a couple of months it will be lack luster because we will have worked out how to game the system. At this point paradox either has the option of patching the exploits out or alternatively as stardock did with gal civ improving the ai to enable it to identify and move to counter these strats.

Really ? pretty sure GC2 A.I. has artificial buffs and I remember the AI would usually choose one type of weapon in all ships like just going railguns while I go lasers and I would own them, I also don't remember seeing them retrofit their ships.(still wasn't bad but not AoE 2 level.
But the diplomacy was pretty good and transparent.

@Taciturn Scot @xtfoster
Well current DD shows more about their vision then complete features and A.I. hasn't been mention yet.

Working on A.I. when every mechanic is done or at the end of alpha stage is what I think makes bad A.I. and I doubt Paradox does such thing, if you left for late stage of development you could have your entire studio halted because the A.I. wasn't finished especially when it comes to a grand strategy that is complex. I think for each new mechanic finish you ought to have the A.I. handle it already.

I never intended to refer as to when it is developed but changing the paradigm of developing game mechanics from player -> A.I. to A.I -> Player.

 

[Taciturn Scot]

@Taciturn Scot @xtfoster
Well current DD shows more about their vision then complete features and A.I. hasn't been mention yet.

Working on A.I. when every mechanic is done or at the end of alpha stage is what I think makes bad A.I. and I doubt Paradox does such thing, if you left for late stage of development you could have your entire studio halted because the A.I. wasn't finished especially when it comes to a grand strategy that is complex. I think for each new mechanic finish you ought to have the A.I. handle it already.

I never intended to refer as to when it is developed but changing the paradigm of developing game mechanics from player -> A.I. to A.I -> Player.

Well, since people were actually playing it, there obviously is a rudimentary (at least) AI in the game already that can manage the systems that are already up and running so they don't leave it all to the last minute. However, the AI opponent isn't going to worth talking about until after all the features have been finished on as you agreed. And that won't likely happen until much later this year.

Besides, what is there really to talk about with an AI opponent? Of course they hope that it will be able to play the game and handle all the systems with a degree of competence and that's about as much as they can say. I'm really NOT expecting an AI any better than we have in any other Paradox title at the moment. EU4 is likely a good model of what we can expect to see and I'm sure that they are very proud of their AI. These games are massively complex

 

[Porkman]

Well, they could show us examples of what the AI does now that it didn't do in HoI3, like coordinate better or use strategies like pockets and encirclements more often, what do I know.

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.