Query re A.I.

[pcelt]

Clearly human opponents are likely to lay down area fire or smoke where they assume the enemy are likely to have observers or units ----without necessarily "Spotting " any evidence.

Does th A.I. possess this type of potential or does it only direct measures against "spotted" enemy activity or visible evidence of their presence.

Thanks

[Seahawk-vfa201]

On my experience the AI does not use smoke in a proactive way, as when advancing on open area toward a possible ambush (Chance Encounter scenario).

[Blackhorse]

While the AI is good, it will never measure up to a human opponent.

[Schugger]

<BLOCKQUOTE>quote:</font><HR>Originally posted by Seahawk-vfa201:

On my experience the AI does not use smoke in a proactive way, as when advancing on open area toward a possible ambush (Chance Encounter scenario).<HR></BLOCKQUOTE>

--------------Spoiler Alert!-----------------

Oh, it does! Those damn AI beast used smoke to toast my concrete MG bunker i've plced in Plomville. The bunker was so placed, that the US forces had no LOS to it until they reached the outskirts of the small village. After my bunker ripped apart some squats the AI layed a very dense smoke screen over the area in question, advanced a flamethrower unit, which miraculously avoided all minefields and barbed wire entrenchments, and celebrated a barbecue party with my hepless bunker crew.

------------------

Es gibt Tage da verliert man und Tage da gewinnen die anderen.

[aka_tom_w]

I think the AI is pretty damn good

the best I have ever played against in a video game war game.

The more you play against the AI the more it will surprise you. Does it make some questionable judgements, yes? Does it use smoke well? sometimes it can use smoke VERY well and surprise you.

I think the AI is very challenging and unpredictable and I could not really ask for much more out of a $45.00 game.

I would like to know what other WWII tactical combat sim games any of you think have BETTER AI?

-tom w

[Schugger]

Originally posted by aka_tom_w:

I think the AI is pretty damn good

the best I have ever played against in a video game war game.

Exactly what i think. Do you remember the first time when you played VoT against the AI blind? It was definitly a challenge, even on the US side without beefing up the germans. You don't know what you are up against in your first attempt and so your style of play differs greatly from the games that follow within the same scenario.

But it must be a much greater challenge to play all scenarios and operations ( if that is possible) against an human opponent double blind It will make me wonder if someone is cool and patient enough to wait for his opponents E-Mails.

------------------

Es gibt Tage da verliert man und Tage da gewinnen die anderen.

[Seahawk-vfa201]

<BLOCKQUOTE>quote:</font><HR>Originally posted by Schugger:

--------------Spoiler Alert!-----------------

Oh, it does! Those damn AI beast used smoke to toast my concrete MG bunker i've plced in Plomville.

<HR></BLOCKQUOTE>

Well in a sense it confirms what I said. The AI used smoke against a *known* target, i.e., the bunker. I have never seen it using smoke as a screen against nothing, as a prevent tactic. And that is understandable: how would you sensibly program the "feeling" that an area could be used to ambush you IF the enemy is hiding there? I did not play much (only 10 times or so) but I have never seen the AI use smoke against area without having a visual on an enemy position.

[Schugger]

<BLOCKQUOTE>quote:</font><HR>Originally posted by Seahawk-vfa201:

Well in a sense it confirms what I said. The AI used smoke against a *known* target, i.e., the bunker. I have never seen it using smoke as a screen against nothing, as a prevent tactic. And that is understandable: how would you sensibly program the "feeling" that an area could be used to ambush you IF the enemy is hiding there? I did not play much (only 10 times or so) but I have never seen the AI use smoke against area without having a visual on an enemy position.<HR></BLOCKQUOTE>

Now i get what you meaned, it takes a while with me sometimes

I never used smoke to cover my advance just because that there may be an ambush waiting for me ( to be honest, i've never thought about it). I think there are "cheaper" options, like the typical half squat ambush triggerers. I like to preserve my shells for bombardement and use smoke only in emergency situations, when a tough german tank appears for example and starts to butcher my rather fragile armour and the AI seems to do just the same.

[pcelt]

Thanks guys for all your helpful comments and observations.

The conclusion, from these examples, would seem to be that ,good and variable though the A.I is----probably the best around,it does not behave proactively like a human opponent,judging a likely threat without actual visible evidence and setting out to neutralise it BEFORE it reveals itself.It reacts intelligently, sensibly variably and imaginatively to threats which SHOW themselves.

The lack of evidence of genuinely proactive behaviour such as smoking or area- targeting a potential threat , however, puzzles me.It puzzles me because it would seem the A.I. should be capable of this." The A.I.Improvement" article at CMHQ states ,in the course of its discussion of the role of the "Operational A.I. and the "Strategic A.I. "The Operational A.I. informs the Strategic A.I. where it BELIEVES the enemy will defend...."

If this does occur then presumably the Strategic A.I. could inform the Tactical A.I. to proactively fire smoke or area- target any of those locations. In other words it could behave in this respect just like a human player hitting possible threats BEFORE they reveal themselves.

Does anyone have any solid examples or illustrations of such purely proactive measures by the A.I.before your troops reveal their presence with movement or fire etc.

[Guest Big Time Software]

Ah, but it isn't that simple Remember that smoke cuts both ways; it blocks LOS to the enemy as well as from. For example, I would not put down a smoke barrage in something like Chance Encounter before seeing the enemy was. It actually limits my range of actions and will in fact reduce my ability to assess what the enemy has. I could be walking right into a big trap. And this is why, historically speaking, forces did not deploy smoke willy nilly.

Now, something like VoT is a different matter. The Americans are going up against a fortified line. In this type of a situation smoke was historically used as a planned element of the assault. The AI does this to some extent, but not nearly as well as a human might.

There is certainly room for improvement here, but the logic coding is not easy because the AI doesn't know it is assaulting a fortified line. It simply knows it is attacking and the other side is defending. So first we would have to "teach" the AI to figure out what kind of battle (more specifically) it was engaged in, the estimated relationship between its forces and the enemy's, and a host of other variables. Then, and only then, would the AI have the necessary information to know if it should put down a proactive smoke screen. Hehe... and then we would have to teach it how to know WHERE and (more importantly) WHEN to put it down, as well as WHEN to stop.

This is why AI work is soooooooo very time consuming and difficult to do. For every one thing you want the strategic level AI to do about a dozen or more things usually have to be in place for it to work with any degree of "intelligence"

Steve

[pcelt]

Steve----many thanks for your helpful and constructive reply . I am sure that coding the A.I is a mammouth and incredibly complex task.

Just one clarification please----the article about the A.I. at CMHQ seemed to suggest that the A.I. was currently able to identify places where the enemy were likely to be. Your reply here seemed to imply that identifying WHERE to be proactive was still not within the capacity of the A.I. Was the article perhaps a shade optimistic re the A.I.'s ability here.

Thanks again for all the clarification and help.

[Jimmy 4 Eyes]

I am not sure that the AI acting proatively and the AI having an idea where the unseen enemy is are at all the same thing. Asking the AI to read a map and see the likely avenues of approach and strategic attack/defense positions is one thing, but having code written it to tell the AI how to react to each of these situations is entirely another.

I think even in human opponents this is such a variable thing that it would be nearly impossible to have the AI act to the unseen, but almost any human with limited experience and training can at least start to recognize the set up and I think the AI mimics this quite well.

I guess in short I dont think that this was overstated at all I think this game is exactly as advertised if not a whole lot more.

[Guest Big Time Software]

Jimmy 4 Eyes nailed that one on the head. Humans can make a near instant leap from A to B, but the computer is dumber than a rock so you have to specifically tell it that A = B, C = D, etc. So guessing where the enemy might be doesn't have any bearing on the enemy's intentions, likely force composition, friendly force capabilities vs. the enemy, etc. So the article probably isn't off the mark (been a while since I read it!).

Steve

[pcelt]

Jimmy-------re your comment--I agree they're not exactly the same thing.But I cant see the point in developing an A.I. which is sophisticated enough to "read a map" and to "identify likely defense positions" but then cannot tell the tactical A.I. to take any proactive measures at all in relation to that identification.

Dont get me wrong here, This is not intended to be any sort of carping criticism of the A.I. I am mightily impressed by the whole game and the A.I. involved---by far the best there is in my opinion and I cannot wait to receive my full version. My only aim is to try and clarify where the current A.I. boundaries or grey areas lie.