Can AI handle Grand Strategy?

[Edwin P.]

Will the AI be able to coordinate its operational, production, diplomatic and research strategies via scripting?

Example: Assume that the Axis AI wants to launch a battle for the Atlantic. Thru scripting can you via a single trigger/condition jointly activate scripts to have the 1) Tech Script research Advanced Subs AND 2) Production Script build subs AND 3) Diplomatic Script focus on swinging Norway to the Axis AND activate AI script to utilize the subs once they are built.

Example: Assume that the Allied AI wants to launch operation Torch. Can the US and UK scripts be jointly triggered so the forces of both allies act jointly to secure the same objective? As I see it now you can have a 20% that the US launches Operation Torch and a 20% that the UK launches Operation Torch but no way to insure, other than by setting activation chances to 100% that both allies activate activation torch at the same time.

Perhaps by moving a single unit to a garrison tile and saying that if this tile is occupied activate Production Script X, Diplomatic Script Y and AI Offensive Script A.

:confused:

[ March 03, 2006, 08:28 AM: Message edited by: Edwin P. ]

[Night]

Very good point. If the A.I. cannot use all of it's nations for one goal, then it will surley be soundly defeated by any decent player.

[Blashy]

Just wait until you get the game and have a look at the editor and the scripts.

The game is out in a 4-6 weeks, you'll have all your answers then, because you'll be able to try it for yourself.

Bottom line is "we'll I ever find a way to continually beat the AI" The answer is yes.

[Codename Condor]

This game is too complex to have a "decent" AI. SC was much more simple and look how the AI played.

Bottom line is you will have to use the editor and give AI enough things to balance your skill.

You can give the AI all the scripts u want about Op. Torch, OK: A+B+C--> go to OP. TORCH, but this is not like that. maybe that way u get what u wanted (AI to performe Op torch... and have a different game.. whats good) but most prolly this way the AI is doomed cos the script didnt take so MANY things into account.

[TaoJah]

Originally posted by Edwin P.:

Will the AI be able to coordinate its operational, production, diplomatic and research strategies via scripting?

Of course the AI can not do that. They never could. And it will take many years before they can.

Compare it with chess.

In chess, the computers has only 16 pieces at his disposal. Most of these pieces can only go to 1 or 2 fields. That makes the number of possible moves for the Computers about 50.

And after he has choosen ONE UNIT to move, the opponent makes his move.

But in SC it's ALOT ALOT more complicated : even at startup, the Axis has lot more possibilties then in chess : he has like 16 units that can each go to like 20 fields. And what's worse : he has to decide for EACH AND EVERYONE of these units where to place it, not just for one.

This makes it virtual impossible for the computer to do the good old number-crunching (calculate every single possibility, calculate every single counter-move, calculate every single counter-counter move and so on)like in chess.

So, they got to make an AI that really THINKS.

Unfortunatly that kind of AI is years away from now, perhaps we'll see one in our lives, but I won't place my money on it.

Don't forget ; humans are smart ! We got a brain that doesn't have to process all sorts of moves to know instinctly which are bad ones.

So, to answer your question : no, the AI can not handle grand strategy. At least not for the next 50 years or so.

[Retributar]

If an AI needs to calculate evey possibility [Permutations & Combinations],then, yes!,the computer might not be able to handle it!.

However, if the Computer is able to decide or concentrate on only a few possible MAJOR-ACTIONS instead of Every-Possible-Action...then it should be able to play a much better game!.

Also, if the computer can know how & when to shift its priorities based on changing overall circumstances or situations...then, again, it can concentrate its priorities accordingly!.

So, a seemingly intelligent AI, i think, may not have to be that many years away, it just may need a new script or rationalization process to follow to make it act more intelligently!.

[Edwin P.]

Originally posted by Retributar:

if the Computer is able to decide or concentrate on only a few possible MAJOR-ACTIONS instead of Every-Possible-Action...... it should be able to play a much better game!.

Excellent point.

IF going for a battle for the Atlantic then: </font>

• Research Advanced Subs (Research Script)</font>
• Build Subs (Production Script)</font>
• Assemble Fleet (AI Operational Script)</font>

Similarly, if going for a strategic bombing campaign: </font>

• Research Bombing</font>
• Research Long Range</font>
• Build Bombers</font>
• Constantly switch targets to avoid Interception</font>

Ideally, there should be some way to jointly trigger; Production + Research + Diplomatic + Operational, scripts so that the AI executes a coherent strategy for a "FEW MAJOR-ACTIONS."

Its not that complex and certainly doable without requiring the processing power of a super computer.

As Retributar said "it just may need a new script or rationalization process to follow to make it act more intelligently!."

HC has likely addressed this issue, but its not in the advance AI scripting notes he has posted.

[ March 04, 2006, 09:49 AM: Message edited by: Edwin P. ]

[Hubert Cater]

Hey Edwin,

Correct, a few things were found missing in the original AI script notes, I've added an AI level flag now as well , but with a little creativity you might be able to get what you are after in terms of grand strategy.

For example, here is a sample script of an Allied invasion of France:

; UK prepares to attack France:

{

#NAME= UK Build Up Amphibious - Brest (D-Day)

#POPUP=

#FLAG= 1

#TYPE= 1

#COUNTRY_ID= 1

#TRIGGER= 100

#LEVEL= 0

#PLAN_ID= 2

#SIZE= 4

#LENGTH= 1

; Brest

#GOAL_POSITION= 59,18

#DATE= 1943/01/01

#STEAL= 0

; Set friendly positions:

; 1st Line - London

#FRIENDLY_POSITION= 64,15

; Set variable conditions:

; 1st Line - UK politically aligned with Allies and not surrendered

#VARIABLE_CONDITION= 1 [2] [100] [0]

; Set tactical conditions:

; 1st Line - London not tactically threatened (dummy condition)

#TACTICAL_CONDITION= 64,15 [3]

; Set activate position:

; 1st Line - 20 Allied units in England

#ACTIVATE_POSITION= 64,15 [7,7] [20,20] [2]

; Set dummy cancel position (single neutral unit at position 0,0). This is not possible as no

; unit can occupy tile 0,0 so event will not be cancelled due to #CONDITION_POSITION

#CANCEL_POSITION= 0,0 [0,0] [1,1] [0]

}

A similar script exists but tailored for the US and so in this way both scripts will be ACTIVATED whenever there are 20 Allied units in the UK. Subsequent scripts also exist that check for the success of the invasion, i.e. capture of key cities and ports and handle the Sea Transport of follow up units.

In general, the use of #ACTIVATE_CONDITION can be extended to coordinate between various fronts as well. For example, you could write related scripts for the USSR to do something specific whenever the Allies launch D-Day or even add in scripts for the Allies to do an earlier limited size invasion depending on the situation on the Soviet front, i.e. relieve pressure etc. You can also tailor the events with the #CANCEL_CONDITION so in the end the good news is that there are many options

[Edwin P.]

Hubert,

Many, many thanks for the AI level flag.

Also, thanks for the tips on how to use the Activate_Position flag to coordinate actions of two nations, and the position tag to co-ordinate the Soviet response with Allied actions elsewhere.

When I receive the game in 3 to 4 weeks I am sure I will be spending quite a bit of my spare time analyzing the standard AI scripts that you wrote.

On question though, can you coordinate production and research? Example: If the USA AI adopts a strategic bombing strategy I want the USA AI to produce 2 bombers and purchase 2 tech chits in strategic bombing. Is there a way to link the triggering of both plans - the production plan and the research plan?

[SeaMonkey]

Hubert, I was wondering, is there a possibility that later we can get an addition to the scripts that would allow us to link them?

The simplest way would be through the #TRIGGER entry. If we can have a unique definition/designation for each script, whether AI or not, like the number "$2" defines the second script. This would be in the #NAME entry line. #NAME=..............,$2

For the AI it could be AI$2. Anyway we could then make an entry into the script dialogue line, like #TRIGGER=[$2],100.

This would mean that the trigger to activate this script was dependent on the second(defined "$2") script being activated and this would have a 100% chance of occurring.

[ March 04, 2006, 03:28 PM: Message edited by: SeaMonkey ]

[Hubert Cater]

Good ideas guys and while any extras are not likely to make the final cut before release, as the good news is that what we've got now is working quite nicely, I can always add stuff in as needed/applicable/desirable down the road

[TaoJah]

Originally posted by Hubert Cater:

Good ideas guys and while any extras are not likely to make the final cut before release, as the good news is that what we've got now is working quite nicely, I can always add stuff in as needed/applicable/desirable down the road

Does that mean you're not tired of the game-developing now ? I know I would after three years !

Anyway... At one hand the quality of a game should not depend on patches, if it doesn't run well the first release there will be alot of demo-players that will never try it again and buy the game. But on the other hand I can imagine that this is something that is very, very, very hard to do. So if the 1.0 AI is decent, any extra's should not push the deliverable date IMHO.

[Hubert Cater]

Not to worry it won't. Truth is, and as I've mentioned in the past, the bulk of development time, and I do mean 99.9%, has simply been to just complete the game based on the original design document.

Everyone has to remember that other than some excellent campaign work and beta testing, there is only person coding/developing/producing this game and so things will take time. This is of course on top of all the other day to day business requirements that need my attention and well you get the picture.

Am I tired of developing... that might be an understatement at this point ... but if and when any fixes are required post release I usually add in some additional touches to improve game play (as needed) overall. If you look at the history of SC1 there were many additions that were not typical of *fixes* but rather updates to make the game that much better.

In general this is all I am referring to here in the sense that if I am already working on some aspect of the game (post release) I can easily add in a few additions (as well) that the SC2 community has reached a general consensus on... but in no way does this delay release and/or indicate some sort of fault in the game as is.

[Konigs]

Sounds great Hubert, I will wait patiently like so many other fans.