Don't get your hopes too high on better AI

The AI will get better as time goes on. We all know it is not going to be as good as a human. But it should at least be able to make more competent decisions.

the question we should all really ask is: Will the team that came up with BBAI will make a similarly better ai for Civ5 after the sdk has been released?

i wanna hear from the foremost experts on AI for civ games.

where are they? i demand a summit meeting of all the AI experts. (by demand, i mean humbly beg)

paging mr. jdog5000... paging mr. jdog5000 to the white courtesy phone please. :D

There may be so called AI experts. But how many grew up playing hex based Avalon Hill games, probably not many. We need, not only someone who can program AI, but also a great armchair general. A virtual kitchen table Napoleon, do you folks know any??? :lol: :D
 
Many thanks for the link, it brought back memories.
While i'm almost sure you already know about this cEvo project too... it proves a point that once a coder goes back to some formal basics with extensive knowledge of a gameplay engine -- any AI can potentially become a source of innovation rather than a pre-compiled secret formula where even the best solutions can fail or appear flawed in certain conditions.
 
The above text forces me to ask: What languages?

C++. Just as a proper set of libraries intervening at masm levels when "Necessary". :)

At issue here is not *ONE* language but rather what features any evolved into... i could have gone alllll the way down to CoBol & still not be far from declarative structures dependent on dispatched assets.
AI simply wraps around stack access & the usual (sadly) loop holes enforced for efficiency.

PS; I can predict you'll want to implicate C# & Java, just to bring up the timeline to this discussion - but, the fundamental scripting issue remains that the IF-THEN-ELSE principle is nothing but a lowest denominator for something much more complex programmatically.
 
There's some really obvious deficiencies in the AI right now that wouldn't take a miracle to fix. Things like sending an unescorted invasion force across an ocean filled with enemy navy. Declaring war on a civ 2 continents away when you have zero navy to speak of. In fact, there doesn't seem to be any AI for navies at all, and it's one of the simplest forms of combat in the game. It's things like this that make a fairly decent AI look like a ******** stepchild.

Aside from that, how's this for a wild idea? A neural network to help with decision making. Normally games aren't very well suited to this kind of thing, but for Civ V it seems perfect. For one thing there's a lot of idle processor time while the human player completes his turn. Unlike IV, combat in V is very situational. The AI can't simply focus on one unit at a time, decisions have to be based on a wide variety of variables all taken together. Really hard to code algorithms for, but the kind of thing neural networks are good at. Learning would be easy, there's thousands of single player games being played that provide a huge amount of information for a neural net to make use of. You could even have clients send back their networks, compare and combine them, then issue updates every now and then.

I doubt Firaxis has the resources for something like this, but it's fun to think about at least. Maybe once the full SDK is released some AI graduate student will turn it into a research paper :D
 
I am thinking of going back to play civ 4 for awhile until the next patch. Honestly and I hate to say it because I mostly have stuck up for this game from day one. The fact is it is getting boring. The biggest issue is the AI and naval combat. I just wiped out a giant English invasion force who came across the bay without naval support. Out of 10 units 1 landed and was decimated immediately. It is just ridiculous, the AI does not try to establish naval superiority at all. It can't even escort embarked units to relative safety. It does not recon properly.

Land combat is not any better. When a war begins the AI sends out these huge armies. After the intial shock is over and these frontline units are destroyed, it gets boring besieging and taking cities one at a time. The AI never seems to send help even if it has 15-20 cities in its empire. I surround a city and bombard it, and finish it with infantry. City after city this goes on and on and the AI civ just lets it happen.

I will keep checking back once in awhile for a new AI patch and try it again then. I remember civ 4 BTS never got boring. The good thing is that I will be rusty as hell so it will probably be even better. :(
 
Aside from that, how's this for a wild idea? A neural network to help with decision making. Normally games aren't very well suited to this kind of thing, but for Civ V it seems perfect. For one thing there's a lot of idle processor time while the human player completes his turn. Unlike IV, combat in V is very situational. The AI can't simply focus on one unit at a time, decisions have to be based on a wide variety of variables all taken together. Really hard to code algorithms for, but the kind of thing neural networks are good at. Learning would be easy, there's thousands of single player games being played that provide a huge amount of information for a neural net to make use of. You could even have clients send back their networks, compare and combine them, then issue updates every now and then.

I doubt Firaxis has the resources for something like this, but it's fun to think about at least. Maybe once the full SDK is released some AI graduate student will turn it into a research paper :D

Or maybe genetic algorithms? Whatever the programming methodology, I think it would be interesting if the program retained some sort of memory across a sequence of games and became aware of your style of play (at the crudest, warmonger versus builder).

Of course that leads to metagames - just alternating play styles might mess up the AI a lot. On the other hand, it may be that we all have unconscious and unavoidable biases (a bit like having a 'tell' when playing poker) that the AI could exploit.
 
Land combat is not any better. When a war begins the AI sends out these huge armies. After the intial shock is over and these frontline units are destroyed, it gets boring besieging and taking cities one at a time. The AI never seems to send help even if it has 15-20 cities in its empire. I surround a city and bombard it, and finish it with infantry. City after city this goes on and on and the AI civ just lets it happen.

well this really sounds boring
in civ4, unless you are completely stearolling wither a weak civ, or playing way belowe your difficulty level, the war can turn at any moment, if AI gets new tech, or it's new stack takes your by surprise, or some other civs joins in

but when you have a large advantage, at least it is fast, and you don't have to besiege any city.....
this really seems like artificial mechanism, a poor excuse for not being able to garrison more units in the city.....
 
After the intial shock is over and these frontline units are destroyed, it gets boring besieging and taking cities one at a time.

If I remember correctly this very thing used to drive me crazy in civ4 pre BTS, and I basically hated the game because of it.
Except that the "initial shock" was a lot lamer in civ4 than it is in civ5.

There are different values in the XML folder that can be changed to at least help the AI reinforce a little better.

For example, someone changed the number of defensive units per city from 1 to 4; and giving the AI a higher unit production ratio than you also helps a little.
 
Whatever the problem, all I know is this - AI didn't start nuking me till I won the game on Science Victory. Human Player winning and choosing to continue apparently pushes the line too far!
 
I don't know much about coding, but EU3 is similar to Civ, yet it's AI is pretty good! How come?
 
I don't know much about coding, but EU3 is similar to Civ, yet it's AI is pretty good! How come?

I don't know but I am going to order that game via amazon. I need a change of pace for awhile. I have Shogun 2 Total War as well. However, Total War is not interesting me at the moment.

Ok, I ordered Europa Universalis III complete, does anyone know if it is compatible with Windows 7? It mentions vista, usually games that play on vista usually work on 7. Let me know, thanks! I'll laugh if it has a better AI, the game came out in 08'. I remember the first one went from 1492-1792 AD.

In the mean time I am going to adjust the CiV AI and bring units per city up from 1 to 4, and adjust AI unit production. I still believe that if it has more men in the field the more it will lose, but maybe it will overwhelm me.
 
I don't know but I am going to order that game via amazon. I need a change of pace for awhile. I have Shogun 2 Total War as well. However, Total War is not interesting me at the moment.

Ok, I ordered Europa Universalis III complete, does anyone know if it is compatible with Windows 7? It mentions vista, usually games that play on vista usually work on 7. Let me know, thanks! I'll laugh if it has a better AI, the game came out in 08'. I remember the first one went from 1492-1792 AD.

In the mean time I am going to adjust the CiV AI and bring units per city up from 1 to 4, and adjust AI unit production. I still believe that if it has more men in the field the more it will lose, but maybe it will overwhelm me.

EU3 runs quite well with Windows 7. It seems to crash on me when I exit the game sometimes (speeds up things actually. :P) but that's about it.

I'd say the AI is pretty good and is certainly challenging. At least for me anyway. ;)
 
EU3 runs quite well with Windows 7. It seems to crash on me when I exit the game sometimes (speeds up things actually. :P) but that's about it.

I'd say the AI is pretty good and is certainly challenging. At least for me anyway. ;)

Sounds good its on the way. I can't wait to try it.
 
Sounds good its on the way. I can't wait to try it.

It really is quite fun.

No game is perfect certainly but it really is an elegant design the way it plays. The UI could improved IMHO though. (Hopefully in EU4)


There are so many different routes to victory. (Victory of course being what you want it to be. Setting your own goals is fun.) You can try to be the papal controller, control the HRE, be a small trading nation, be a great colonizer, be a warmonger, play as a horde and raid and pillage etc. The replayability is excellent, especially considering you can play over 200 different nations.

Playing as Norway right now. Colonizing North America and crushing the Swedes. :viking:
The reformation is starting and I'm going to have to make a big decision soon and then the ****'s really going to hit the fan. Lol.
Maybe religious wars for 100+ years.
 
It really is quite fun.

No game is perfect certainly but it really is an elegant design the way it plays. The UI could improved IMHO though. (Hopefully in EU4)


There are so many different routes to victory. (Victory of course being what you want it to be. Setting your own goals is fun.) You can try to be the papal controller, control the HRE, be a small trading nation, be a great colonizer, be a warmonger, play as a horde and raid and pillage etc. The replayability is excellent, especially considering you can play over 200 different nations.

Playing as Norway right now. Colonizing North America and crushing the Swedes. :viking:
The reformation is starting and I'm going to have to make a big decision soon and then the ****'s really going to hit the fan. Lol.
Maybe religious wars for 100+ years.

This sounds very interesting. I checked and amazon has shipped. Perhaps we can start a discussion thread about this in other games thread. I was wondering do you play a mod or the standard game? I see gametime goes from 1399-1820 AD extended from the first version which was 300 years, I like that? Also are there any patches I need to be aware of? I will let you know when I start playing the game should be here in a few days.

Welcome to the club. ;)
This should be fun and thanks! :D
 
Back
Top Bottom