Will the AI ever be improved?

[fcam1387]

I know this has been asked many times before, but I always notice and element of passiveness in the AIs moves, it seems timid and just appears to sit around and not exploit any obvious opportunities it has. I am just wondering whether any of the developers are working on improving it.

[Joel Billings]

We are not currently working on improving the AI. It is unlikely we will do any more work on the AI as only a major effort would hope to have any real impact, and it's pretty hard to justify putting in that kind of time. We think the game does a good job of playing a decent game that will challenge new players. The only way it can hope to challenge an experienced player is by playing at the hard or impossible play levels, or setting your own level that gives additional advantages to the computer. I wish we could do more, but improving the AI is just not possible without putting in months of effort.

[ravinhood]

but improving the AI is just not possible without putting in months of effort.

Ahh from the developers own mouth himself. The biggest problem with games and their AI "not putting enough time into them".

Joel you've been in this business for a very long time now, I'm really surprised you are still coding AI's only meant to be challenging to "new players".

[JanSorensen]

I doubt anyone intentionally sets out to code an AI that will only challenge new players. However, in most games its simply not feasible to code an AI that will challenge the veteran player. Just consider the huge amount of money and time it took to finally make Deep Blue beat Kasparov in chess - a game where the rules have been known and studied for endless years.

So, its not just a matter of putting enough time into the process. Sure, the AI could be made better if another 3 months was spent on it - but the gain would still not let it fight the veteran player on even footing. It might perform less silly mistakes but it still would not be good. Thats just the state of todays knowledge about AI.

In short - I wish the AI was better - but I acknowledge that its neither technically nor financially feasible to make the AI challenging for the veteran player.

In my opinion its not correct to claim that the biggest problem with AIs is time. The biggest problem is that its outright incredible hard to code a strong AI. Infact, extremely few games if any manage. Exceptions are trivial games, chess due to the special curcumstances, and games where the AI just cheat making it strong but not intelligent.

[fcam1387]

Thanks for the reply guys. It is a shame that nothing is being done with the AI because this is a high quality game. However, in my case and I am sure many others agree with me, when I buy a game I really dont intend to spend much time playing multiplayer as it is slower and difficult to find opponents. If there ever is a sequel or a similar game it would be great if more time could be spent in this respect.

[Joel Billings]

We invest months of time into work on the AI for all new games. You can fully expect that any follow on game will have more time invested in the AI and that it will be better for it.

As Jan says, we cannot expect the AI to play well against a veteran player, certainly not without cheating. However, the game is usually good the first few games even against a veteran. What happens over time is the veteran begins to read the AI as well as learning more about the game itself, and in that way with every play, the player gets better. AI today is not a learning AI, so it stays the same while the player gets better. Thus, the disparity in play level gets bigger with more play. We think with 5 different players to play, and with play levels that give the AI a handicap, you can enjoy a number of games before getting to a point where there is no challenge. That's the best we can hope for at this time.

[Lebatron]

ORIGINAL: fcam1387

I really dont intend to spend much time playing multiplayer as it is slower and difficult to find opponents.

The pace is not as bad as you think. Try one game of PBEM and I guarantee you will enjoy playing. As far as finding opponents, that is not difficult at all. Just post in the opponents wanted section and in no time you will have offers.

[luftwaffe]

I do agree that the AI should be improved, the problem multiplayer is even with this game it still can be a hassle at times to finding opponents. When you play against the computer it never says no, it never gets tired and it wont quit on you in the middle of a game. Another turn based game I own is Stardock's Galactic Civ 2, the position they took was to focus only on the AI and not include any multiplayer for the same reasons I listed above. Some people have griped about the lack of multiplayer but I think the AI is very good. The AI in world at war is decent but I still would like to see a major overhaul of the AI especially for those people who do not have a internet connection but thats just wishfull thinking I guess.

[JanSorensen]

I have quite a bit of experience playing WaW. I also have some experience with coding. I am not certain I could make the AI even remotely challenging even if I spent a full year doing nothing else. The Russian AI, sure - but the WA AI in particular would be extremely hard.

Not to put the Civ type games down in any way - but coding an AI for those seem alot easier to me (mind, I do not know the particular one you refer to). For one you often see the computer cheating in those games in ways you are not informated about. For another those games are usually alot less positional in nature. Finally, they are usually everyone against each other rather than set alliances which in itself makes it alot easier to code a somewhat competitive AI. You just make the AI players good at ganging up on the leader making it hard for a human player to become the leader.

Anyway, I certainly take my hat off to people that can make a challenging AI - even in games where its done by "cheating" or other tricks. I just find it extremely rare to see in games like WaW where positioning is complex and combinatorially several orders of magnitudes about chess.

[luftwaffe]

I do see your point and I'm not a computer programer myself so I guess I don't know how hard programming AI is. Hopefully someday computers will be able to learn, of course that is probably years in the future. Of course then when that day comes you'll go to play a game and the computer will tell you no because it has grudge, like the Hal 9000 in the 2001 movie, lol.

[Timmeh]

I'm afraid I can't let you move those subs out Dave....

[Lebatron]

ORIGINAL: Timmeh

I'm afraid I can't let you move those subs out Dave....

Good one[:D]

[Lebatron]

The day that AI can learn on its own will be the day machines enslave mankind. Anyone read sci-fi?

[Timmeh]

yup, even scarier is when we hook up to the machine and it starts to read and grow from our subconcious. *shivers*

Oh heres another...


Will my subs dream at the bottom of the Atlantic, Dave?

[luftwaffe]

It won't surprise if we see this happen in another 20 years, with the rapid advance of computers and genetics we'll probably have cyborgs. On the other hand in the 2001 movie which was in 1968 I think most people thought that we would be living on the Moon and and have inteligent computers by 2006, so who knows.

[magic8796]

Remember the W.O.P.R.? "Shall we play a game?"

[Lebatron]

Ah WAR GAMES. Didn't remember the computer being called WOPR though. Back in the day you couldn't say "Shall we play a game?" without copying how the computer sounded. Come to think of it I still would today if I said it[:D]

I know there are a bunch of phrases from movies that are in the pop culture consciousness. I just can't think of any at the moment. Help me out[:)]

[Lebatron]

Oh I just thought of one. "Share the l..o..a..d" Can anyone guess where that came from?

[ravinhood]

Well I'm here to say the ai in Master of Magic is pretty good. I'm at the gates of losing a game setup on Hard w/large map, 4 other computer AI and powerful magic. I barely got 5 cities before the 3 ai above ground declared war on me and on this particular map they are all 3 on the same island as me. Thus it is wave after wave and battle after battle every turn with the ai sending in a couple of units just to use it's psionic blasts, fire bursts and drain lifes every damn battle and kills one of my units and then also drains me of my own mana. I've been using the alchemy trait, but, now down to 4 cities and 2 of them just humble beginnings my mana is going to run out in a battle or two. I been churning out zombies like candy and got resist magic up on most of my main city defenders, to no avail though, their high power blasts and bursts and drains are still either killing a unit outright or knocking it down to 1 little man after the battle is over. What is really amazing about the MoM ai is it appears to learn. Once it sees what I have in my main city and how I am defending, it starts building units and using magics that are most effective against me. Like adding magical weapons to it's units to go up against my zombies and then start using star bursts to take them out ever so easily as well. Of course the fanboys at Triumph Studio attempt to cut down the ai of MoM all the time, but, I'd bet a nickel to a $ they didn't even play all the higher difficulty levels and settings before claiming it was lame. I swear to gawd the "Impossible" with "Powerful Magic" is impossible to win (vs 4 ai using powerful magic, I know the exploit of just playing against 1 ai on a small map and grust rushing to kill it before it gets any cities going and claiming victory on impossible, I can do that too)....there is just no way vs 4ai and powerful magic. I've tried it numerous times with various setups and I can't even get out of the starting gate before the ai is pouring thru my cities with Sky Drakes and Demon Lords and blasting my cities with every global spell in the book. lol
 
Now of course it is cheating somewhat, but, I've won on hard with 4 ai before and powerful magic, but, it is extremely hard and challenging. Thus, I don't have a problem with the ai getting advantages, but, with them by gawd they should be able to kick my arse on at least the highest difficulty if not the one before it. That's the problem I find with most other games today. RTW's for example very hard/very hard was like playing normal/normal.
 
The thing about civiliation type games I've played is the wise ai programmers programmed the ai to go after the human player on higher difficulties no matter how much diplomacy you tried. I totally appreciate and respect this kind of coding. That's what I want to happen. Especially if I'm winning, I want every ai on the map to ally and come knock me off the hill. Thing is about MoM's ai, you don't have to be winning for them all to gang up on you. Basically playing the Civilization type games it's always 4+ against 1 very early in the games. Civ III and Civ IV rather dumbed down the AI's into all jumping on you in the early game like the first two did or like in Alpha Centauri. Alpha Centauri was pretty balanced though, it wasn't until about mid game that I started to have to deal with most of the other factions and almost always after clobbering one of their cities they always were willing to accept peace, just so they could build up again and sneak attack me. PEACE in a WARGAME is a rediculous thing to even have in them.
 
The problem with games like GGWAW and COG is they are built too much around historical accuracy. The Civilization series of games don't have any historical accuracy setups or really even unit statistics they have to follow. There's no holding back an AI for example like the Russian or American AI in GGWAW. There's no specific cities setup at the beginning to take out or resources. Everything and all history for that particular game FORMS as you play it. That's what is so great about random maps and random AI's. Of course if you removed those restrictions in GGWAW you would have GG's RISK with WWII like icons. But, that would be better than stifling the ai as it is and forcing it to wait on some historical value entry. I just personally don't like historical wargames anymore vs random generated go all out wargames with powerful ai's to contend with. I'd rather play a civilization type game any day over any strategic historically "attempted" accuracy. Tactical wargames are fine. There's usually no restrictions in them and with games like SPWAW and CM they both have random generated maps and the ability to get randomly generated units to play with, along with randomly generated weather, terrain and stuff like that. I don't want to know what to expect as much as possible, historical accuracy type games never have that. If you've read any war history at all and play an historical recreation you pretty much know what to expect every single game.
 
Maybe Gary and Joel should try making something more abstract without any history behind it. Perhaps then they could do a better job on the ai. Since it wouldn't have anything historical it would have to follow. Something Sci-fi or Fantasy based would be excellent choices IMHO.

[Joel Billings]

I doubt you'll ever see Gary work on anything fantasy or sci-fi. I enjoyed working on Fantasy General years ago, but really our interest is historical games. It's like telling a rock musician that he really should be creating classical music. Keith on the other hand is very interested in sci-fi and fantasy and has had a sci-fi game on the drawing board for years. Someday he may get around to working on it seriously.