RE: ai? (Full Version)

All Forums >> [New Releases from Matrix Games] >> Empires in Arms the Napoleonic Wars of 1805 - 1815



Message


Skanvak -> RE: ai? (12/30/2009 8:34:22 PM)

May you are an AI? [:D]




NeverMan -> RE: ai? (1/2/2010 4:51:46 PM)


quote:

ORIGINAL: pzgndr

My unofficial assessment of just those who have posted here on this forum over the past few years is that more players are interested in having a decent computer opponent for some or all of their gaming than there are players who have no interest in AI development or playing against a machine.



I'm pretty sure your "unofficial assessment" is wrong. MY "unofficial" assessment says the exact opposite of your "unofficial" assessment... of course. [;)]




Sorta -> RE: ai? (1/5/2010 4:03:44 PM)

Only speaking for myself I believe PBM issues should be prioritized over AI. I've never played a PC wargame with an AI that approaches anywhere near the challenge of a human and this is a lovely multi-player wargame.




Marshall Ellis -> RE: ai? (1/6/2010 2:30:46 PM)

It just seems to me that the more vocal crowd is the PBEM crowd (At least today LOL)! I intend to fix up the PBEM stuff a bit then move to AI then IP!




Skanvak -> RE: ai? (1/6/2010 5:16:39 PM)

IP crowd! LET's GET VOCAL!

PS : Marshall, don't forget to fix the rules...




Lildorien -> RE: ai? (1/25/2010 12:02:10 AM)

Sorry to dig up this quite old post, but I think it should be feasable to improve the AI on the diplomatic level and possibly quite simply. I've played a few games as lone human against all AI opponents. They are always very passive, almost never issuing any war. I think it could be fixed if the AI was checking it's 'roadplan' : something like 'I need to make another 200 VPs before the end in order to win, There are 20 turns left so I need 10 VPs per economical phase, I'm currently making 8 so I won't win if I don't win some wars/battles first. Furthermore, checking the other players (AI/humans) I see that player A, and B are currently well ahead in VPs while the rest is struggling, so they should be the ones I attack...

This can only be implemented if the VPs gain bug in harder versions of the AI is corrected first or it would ruin it all.

Lildorien, hoping for better AI [&o]




Marshall Ellis -> RE: ai? (1/25/2010 12:40:00 PM)


quote:

ORIGINAL: Lildorien

Sorry to dig up this quite old post, but I think it should be feasable to improve the AI on the diplomatic level and possibly quite simply. I've played a few games as lone human against all AI opponents. They are always very passive, almost never issuing any war. I think it could be fixed if the AI was checking it's 'roadplan' : something like 'I need to make another 200 VPs before the end in order to win, There are 20 turns left so I need 10 VPs per economical phase, I'm currently making 8 so I won't win if I don't win some wars/battles first. Furthermore, checking the other players (AI/humans) I see that player A, and B are currently well ahead in VPs while the rest is struggling, so they should be the ones I attack...

This can only be implemented if the VPs gain bug in harder versions of the AI is corrected first or it would ruin it all.

Lildorien, hoping for better AI [&o]


These are good points and I will try to improve the AI in 1.09 as a focal point. This is on my road map.





Skanvak -> RE: ai? (1/25/2010 6:11:50 PM)

I think that the multiplyer AI and the single player AI should be different. In the single player, it is possible to make a general decision tree for the global situation according to the decision of the player only. Whereas for true MP game you have to give the AI a real possibility to negociate.

That need an interface that does not exist in the present state of the program. I have an idea for negociation interface using diplomacy set publicly offered, then each player will make a bid, the winner will get partial control of the AI country.

Still thinking about both system.




pzgndr -> RE: ai? (1/26/2010 2:21:20 AM)

Skanvak, ideally each AI player should be playing to some established rulesets, like the UMP rules and national objectives for semi-historical play.  Maybe the AI player should know which opponents are AI and human and perhaps act a little differently, but the various different combinations of human/AI games could make this problematic.  But with decent rulesets the AI should play OK as either single or multi-player.

I do not understand human winner getting partial control of an AI country.  Is this like current UMP rules where a human player controls a country with limitations?  The idea is not clear, nor how you would divide control between human and AI.

I would agree the negotiation interface needs improvements.  Even amongst human players the conditional/unconditional surrender conditions need to be more flexible and offer a better interface for negotiating terms.  For other diplomacy phase negotiations there is an interface to talk with AI allies, and this could be enhanced with more options and perhaps add options to talk with AI neutrals/opponents.  But really, what negotiating are you after?  For requests like we currently have, you can expect either a yes or no response, or maybe something conditional like the AI will do something for PPs if you agree.  Again, this could be enhanced some but the bottom line is you ask for something and you either get it or you don't.  Same with a human.  Human may offer a long-winded rationalization for his decision whereas the computer only calculates probabilities, but the end result is the same.  So what exactly is a "real possibility" to negotiate??




Skanvak -> RE: ai? (1/26/2010 6:38:39 PM)

fast reply :

The winner of the negociation auction get a partial control of the AI. It is a bit like the actual UMP system but gradual and compel the player to uphold its promise to the AI otherwise the AI will withdraw its authorization.

As to how it is quite simple. Ai Give control of it DOW and part of its troops increasing number if truth increase until the AI become a controlled minor.

I see AI as a minor player not the equal of a human player.

For the diplomatic interface. we have 2 options : one is to do something like the Storytron (please look at it, I can't explain it in a few words). Or an oversimplified system that could work for a boardgame.

I think you miss the point of a negociation. Human can always negociate as long as they can communicate, so an interface is not needed. For the computer you need one. Something that allow to chat with the computer in a way understandable for both party. So you need not only to tell the AI how to evaluate the situation, but to actually speak (use a language). Establish a psychological profile of its opposant ...etc... It is a challenge in itself.

There are too much possibility in a discution for the AI to be efficient for negociating with several player. They are machines, they think like machines. It is not a question of probability.




pzgndr -> RE: ai? (1/26/2010 10:24:56 PM)

Skanvak, you're joking about Storytron, yes?

quote:

Storyplay (also known as interactive storytelling) is an art form that combines the strengths of games with those of traditional narrative. Storyworlds, which realize this new art form, are highly interactive, like games. Like stories, though, events center on the people. The computer-generated characters whom the player meets are emotionally aware, each with their own personalities, relationships, moods, and personal agendas. They are in essence primitive “AIs.”


Maybe the term AI is confusing in the context of a computer game? If you're thinking about advanced AI that is emotionally aware, has a personality and psycological profile, has voice recognition software and speech capability, etc., then you've got a very unrealistically high expectation. A decent programmable (or scripted) computer opponent is more reasonable for a $60 game. I speak of reasonable AI enhancements and improvements; you seem to speak of unattainable SciFi. We are talking about two different things.

quote:

They are machines, they think like machines. It is not a question of probability.


Machines do not think. They are programmed with fuzzy logic and probabilities to produce variance. Rulesets and national objectives can be programmed to produce reasonable outcomes, with some variance for replayability. Personalities can be introduced to cause the computer opponent to be more aggressive or more passive when probablities are used to generate a decision. Ultimately a game decision is a decision and, as I said, the end result is the same whether your opponent is human or AI-controlled.




Skanvak -> RE: ai? (1/27/2010 6:08:31 PM)

I see that you have seen the project but misunderstood it. The program is quite simple and has been used for a game called "Balance of power 2". The emotionnaly aware AI has a language (written one is enough...sometime I wonder if your irony is not offensice...) So I think this engine is fine for a 60$ game. That why I think if we want to have AI player conducting diplomacy, we should enlist M. Crawford on this project to use its existing engine.

Otherwise, we should focus on an improve UMP and nothing more.

You should understand that negociation is about telling the AI to do thing together, building trust, planning stratégy. All those thing require a common language. Other wise we return to the improved UMP rule.




pzgndr -> RE: ai? (1/27/2010 10:39:53 PM)

quote:

we should focus on an improve UMP


Skanvak, it's nice to see that you now support some continued AI development as worthwhile. I'm optimistic that the computer opponent can become a tad bit more feisty than just a bare-bones UMP, so we shall see. [:)]




NeverMan -> RE: ai? (1/28/2010 5:48:58 PM)

Just implement the UMP rules from Empires in Arms.




Marshall Ellis -> RE: ai? (1/29/2010 2:40:46 PM)

I do tend to like those rules. They have good controls against abuse and allow an ally to really be helpful.




Page: <<   < prev  1 [2]

Valid CSS!




Forum Software © ASPPlayground.NET Advanced Edition 2.4.5 ANSI
3.341797