If the AI were human

[joncnunn]

For the readers amusement:

AI 1: Boy is Jon slow, he takes a full five seconds to make turn 2 is a non-expansionist civ, and only gets slower from there.

AI 2: And sometimes I get the feeling that Jon is a mind reader, I have my stack of doom all ready to march straight to one of his cities to sneak attack it, but does Jon let me? No, and he doesn't even have the curtisty to demand I declare war or leave so I can declare war, but instead he just sends the workers away and sets up a mobile road block and we stay offically at peace.

AI 3: And when there's choke points, Jon always seems to claim more land that he should, he even has the gall to set up mobile roadblocks against my unit-settler pairs to keep me from filling in all those holes with my own cities.

AI 4: And Jon always seems to get Philosphacy first, and to add insult to injury, always picks one of the most expensive techs to get for free instead of the cheapest one.

AI 5: And how come Jon gets to hire a taxmen or scientist to maintaign order and we must hire a entertainer? Human-Cheat!

AI 6: And I think Jon actually adjusts the sliders on each of the last three turns. That's another Human-Cheat!

AI 7: And all the AI in the Metospotainia conquest told me how Jon playing Egypt, a non-religious civ in that conquest, bypassed the Tribal Council government entirely! Cheat! Your supposed to switch to every form of new govt as soon as you discover it to fully experence the anarchy, man!

[ChrisiusMaximus]

Way to go, I say we set up an AI rights comittee

Equal rights for Artificial Intelligence

Funny stuff John and original too, any more to come ?

[joncnunn]

At this point, no. I am kinda of hoping that some of those who end up writing the Civ IV AI read this and write either the core rules or the AI is a way to eliminate all but #1.

#6 would be easist to fix, it doesn't exist in Civ II or SMAC. Just abolish the min & max turn tech limits and let the extra science beakers count towards the next tech like those games while keeping the reduced cost for techs you know that other people have.

Next easyist to fix is #4, greatly increase the AI weight of a tech granting a free tech when it's undiscovered and select max cost tech when it's been granted one for free.

#5 is a little tricker, but still simple. If a city would be in disorder and has no non-entertainer specalists already, select taxmen. If city would still be in disorder change specalist to entertainer. If city would still be in disorder repeat.

#7 is a matter of adding either worst case or average case cost-benifit analysis, including expected future income, future support costs, the zero income but no support of anarchy, and the new worker rate (if there's work to be done) along with the half speed worker rate during anarchy (if there's work to be done). Sugestion to Firaxis: consult with ET whose done extensive matematical analysis for switching for the democracy games here. It's mathematicaly tedious and so I "feel" the right time to switch more than calculate it but you can't very well code an AI to feel the right time.

#2 and #3 are the hardest. Humans know when to try to manevuer to avoid an obsticle, when it's time to try to concentrate a force against a small part to pound thru, and when it's time to give it up.

#2 is occuring because the AI weights the potiental loss of an indivudal unit too highly, especally when odds are overwhelmly in favor of most of the stack surviving unharmed. (The General McC. Syndrome : US Army General early in the US Civil War)

#3 is a manner of the AI doesn't know when to give up. It's settler pair is never going to reach the target against the mobile wall and the single unit isn't enough to pound thru. (The Hitler Syndrome)

They would also need to avoid the General Custer syndrome as well.

[Paddy]

looks interesting

[Mad Bomber]

Originally posted by joncnunn
At this point, no. I am kinda of hoping that some of those who end up writing the Civ IV AI read this and write either the core rules or the AI is a way to eliminate all but #1.

........
#2 and #3 are the hardest. Humans know when to try to manevuer to avoid an obsticle, when it's time to try to concentrate a force against a small part to pound thru, and when it's time to give it up.

#2 is occuring because the AI weights the potiental loss of an indivudal unit too highly, especally when odds are overwhelmly in favor of most of the stack surviving unharmed. (The General McC. Syndrome : US Army General early in the US Civil War)

#2 actually is a matter of the AI valuing undefended targets too highly and thus allowing the human to create killing traps for the AI to fall into or to prevent an offensive that otherwise would have been quite harmful to the human player. For CIV the AI must be developed so that it can recognize the need to conduct assaults against vulnurable (but not undefended) targets. The AI also has no concept of organized defense; it just stacks a few defenders in cities, perhaps with a couple of bombardment units, and calls it a day. It repeatedly fails to organize a defensive line, defense in depth, or utilize fortifications of any type.

[joncnunn]

In addition, I'd say that it's actually sometimes tougher to take a large population AI city on the border when they are caught flat-footed that turn than if you wait a turn for them to screw up their deployment in their front line cities.

(Because of the new to conquests zero ranged bombardment with good attack values leaving the cities once the AI moves during war when they would be better off staying inside the city to provide bomb + counter attacking when the enemy is next door.)

Originally posted by Mad Bomber

#2 actually is a matter of the AI valuing undefended targets too highly and thus allowing the human to create killing traps for the AI to fall into or to prevent an offensive that otherwise would have been quite harmful to the human player. For CIV the AI must be developed so that it can recognize the need to conduct assaults against vulnurable (but not undefended) targets. The AI also has no concept of organized defense; it just stacks a few defenders in cities, perhaps with a couple of bombardment units, and calls it a day. It repeatedly fails to organize a defensive line, defense in depth, or utilize fortifications of any type.

[unscratchedfoot]

Wishful thinking gentlemen. Unfortunately Firaxis doesn't share your views that the AI is important so I doubt you'll ever see any indepth defences or strategic maneuvres. I've heard less than 0.01% of their budget is spent on AI development. Better you play chess, shogi or Axis & Allies if you want a tactical challenge. Civ3 is a for when you need a vacation for your brain.

[joncnunn]

Originally posted by unscratchedfoot
Wishful thinking gentlemen. Unfortunately Firaxis doesn't share your views that the AI is important so I doubt you'll ever see any indepth defences or strategic maneuvres. I've heard less than 0.01% of their budget is spent on AI development. Better you play chess, shogi or Axis & Allies if you want a tactical challenge. Civ3 is a for when you need a vacation for your brain.

Human vs AI chess is like going up against a Grand Master. Your probably going to get your head handed to you.

I'm not familiar with Shogi.

Haven't played the computer version of Axis & Allies, and have only played the board game, so I don't know how good it's AI is.

What, Soren's salary is less than 0.01% of Firaxis budget??? If so, he deserves a raise.

[unscratchedfoot]

Originally posted by joncnunn


Human vs AI chess is like going up against a Grand Master. Your probably going to get your head handed to you.

I'm not familiar with Shogi.

Haven't played the computer version of Axis & Allies, and have only played the board game, so I don't know how good it's AI is.

What, Soren's salary is less than 0.01% of Firaxis budget??? If so, he deserves a raise.

Why play against computers? Online chess is the way to go. Very fun and its even more entertaining to see people's reactions after they are crushed. Who is soren? A star trek character? Like I said, I just heard Firaxis hired a part-time worker to write up the AI script but that's just what someone said on the civfanatics forum. Judging by the AI's performance ingame, I'd say that part-time worker is testing his AI scripts against drunken apes.

[joncnunn]

Soren is the lead programmer of the Civ III series AI and is largely responsible for it being smarter than Civ II & SMAC.

[unscratchedfoot]

You are friends with soren? Does he have long pointy ears? You should play Axis & Allies online cause its very nice. Its a bit unstable though and sometimes requires a reload, but then the losing side always runs during reloads. Some people are pretty good while others have no clue and spend all money on tech rolls. I do weird stuff too and get hissed at by opponents for not following the standard way of playing almost everyone follows but same as in chess I guess so they get confused and make bad moves. You need the cd and patch to paly online but hasbro cut off support for it which was a cardinal error on their part considering how popular the game is.

[joncnunn]

Um, Soren doesn't know how am I am. I've read several of the chats he was in when Civ III was new.

I dought your oppoents should have been hissing at you. If your tactic works out great and catches them flat footed, it's great! and you deserve their applause. And if your oppoent hands your head to you right after your weird move, it would probably be your ally(ies) complaining and not your oppoents.

[DarkCloud]

Soren posts here quite often.

[Mad Bomber]

Originally posted by joncnunn


....
Haven't played the computer version of Axis & Allies, and have only played the board game, so I don't know how good it's AI is.....

The Axis and allies AI isn't a complete waste, but it isn't nearly as good as CIV III's AI considering the greater need of code for the CIV III series. The Allies play well by the AI but can be beat even at the 5* level. The Germans are probably the best managed by the AI, but the AI falls on its face with the Japanese. For some reason the Japanese only build naval units, air units, and infantry and the largest percentage of their spending goes to infantry. By the invasion of Japan comes around they usually have 70+ infantry units stacked up in Japan. If they werea little bit more intelligently played, they could have used that spending to put pressure on the Russians making the game much more difficult.

[Shogun Gunner]

Great OP. I it. Thanks for the link to this thread Jon. Yes, an AI conference complaining/conspiring against the humans was in my head!