Computer learns to play Civ II using only the game manual

Misterboy

Modern Major General
Joined
Oct 12, 2010
Messages
244
Here's the slashdot story:

http://games.slashdot.org/story/11/07/13/1451246/Computer-Learns-Language-By-Playing-Games?utm_source=feedburner&utm_medium=feed&utm_campaign=Feed%3A+Slashdot%2Fslashdot+%28Slashdot%29&utm_content=Google+Feedfetcher

Civ II related, but I thought since that forum is pretty dead, more people might find it here.

Apparently the main thrust of this work is to teach the AI to understand human language without help. It's cool that it used Civ II and its manual ;)

Edit: Just noticed someone beat me by a minute or two on the Civ II forum. Also, I misclicked and intended to post this in the Civ V forum as that's where I'm most active. Mods, please move this or merge it with the other, or do whatever you see fit. Thanks :)
 
If this gets moved, it'll be my first post in the Civ2 subforum. :D

I mean, I have Civ2 but I got it at the same time I got Civ3, so I never actually played it. :p

ontopic/

I, for one, welcome our new computer overlords. :scan:

Hopefully they'll will be smarter in the near future when the machines take over mankind than they are currently.
Stupid automated workers and city planners, always farming my towns, workshopping my wheat and building nuke plants on my coal healthy cities! :mad:
 
Oh, big woo! I learned to play Civ II by reading the manual too! :p

(Actually, it is pretty neat. I wish I could handle the SDK that well. I'd get some sleep occasionally. :lol: )
 
It's definitely a testament to the quality of the Civ2 manual, and its inclusion of strategy advice. The researchers tried to fool their AI by adding random sentences from the Wall Street Journal, and their algorithm still recognized which sentences were relevant to the current state of the game, *during* the game. They also observed that the relevance decreased as the game progressed, since the manual is mostly focused on how to get your civ started. The manual doesn't talk much about late-game naval invasions, or when/how to use aircraft units.

I've been impressed with the manuals for Civ4, Civ3, and Civ2. Good reads, and useful!
 
The real question on this is how good it plays :devil:

And indeed Civ III and IV manuals are quite good ( 300ish pages each )... the Civ IV one took me up to noble by itself :D
 
I, for one, welcome our new computer overlords. :scan:

Hopefully they'll will be smarter in the near future when the machines take over mankind than they are currently.
Stupid automated workers and city planners, always farming my towns, workshopping my wheat and building nuke plants on my coal healthy cities! :mad:

Without a +40% cheater bonus in research and production, computers will never take over the world. :nuke:

And they'll always suck at poker!
 
I started playing civ 2 without a manuel and I play it as if it was AoE.
 
I bet adaptive pattern recognition (use actions that tend to lead to victory, discard actions that do not, or at least weight things based on expected win variance) would do even better. AI sampling size of actions would have to be enormous though. The cheap approach is to use scripts that mimic good players, for actions that are hard to game against and defeat (IE priority of on appropriate-timed ROI building decisions, surprise raze attacks, algorithms to switch to pillage party when odds of capturing a city are awful, forking algorithm, etc).

Of hell, prevent it from making units move into danger at the start of a turn against play orders. Ohwaiiiit, the game only does that to the HUMAN.

Still, interesting, I'll have to read this...
 
I bet adaptive pattern recognition (use actions that tend to lead to victory, discard actions that do not, or at least weight things based on expected win variance) would do even better. AI sampling size of actions would have to be enormous though. The cheap approach is to use scripts that mimic good players, for actions that are hard to game against and defeat (IE priority of on appropriate-timed ROI building decisions, surprise raze attacks, algorithms to switch to pillage party when odds of capturing a city are awful, forking algorithm, etc).

Of hell, prevent it from making units move into danger at the start of a turn against play orders. Ohwaiiiit, the game only does that to the HUMAN.

Still, interesting, I'll have to read this...
You know, I would really like to play a game that you have designed, and I'm not being sarcastic. I really would enjoy it. I think that most of us here would like a challenging, well thought out game that works properly.

It's too bad that you weren't the project manager on Civ5. It might be worth playing if you were. ;)
 
Well, their AI improves by 78% just by reading the official manual.
It will become unbeatable once it has read the strategy manuals from CFC !!! :)

But seriously, reading texts from the forum could be a next step in their research. So their AI would tap into the vast experience from the best human players.
 
Well, their AI improves by 78% just by reading the official manual.
It will become unbeatable once it has read the strategy manuals from CFC !!! :)

But seriously, reading texts from the forum could be a next step in their research. So their AI would tap into the vast experience from the best human players.

But it's just so many info and different people have different opinions!! Besides, we discuss about so many different themes!!

Damn, it will eventually happen, won't it?? Having computers learning on their own and beating human players, I mean. It has always been fairly harmless always losing in chess versus a computer on very hard, but on something as complex as Civilization, damn...
 
This is pretty awesome and reminds me of the computer playing Jeopardy. The interesting I find is that the computer was playing against AIs in the game, not humans. If the computer was playing against humans instead of AIs, I wonder what the results would have been?
 
You know, I would really like to play a game that you have designed, and I'm not being sarcastic. I really would enjoy it. I think that most of us here would like a challenging, well thought out game that works properly.

It's too bad that you weren't the project manager on Civ5. It might be worth playing if you were. ;)

I've thought of doing it, but I have no programming experience so my ability to do these things (or earn a position which can manage people who do so) is curtailed. There are lots of places game design can go wrong though:

1. Game rules/objectives/etc ----> historically civ has been very good here. Before anything else the concept of the game has to be viable or it's very hard to make a good title. The answer to what makes a title "fun" is a very difficult one to answer, even for experts and the market itself (for fun, make a poll and ask people to say why this game is fun, you'll get a lot of different answers...but not all games adapt well to multiple approaches like that!) For civ, I feel like the amount of choices available, the sense of scale, and the long term/short term strategy dynamic coupled with a connection to something people can readily identify (history, empire building and the like are common enough themes in entertainment). There's a good bit of (ability scale-able) challenge and a variety of approaches...these things are good enough to help civ overcome a lot of flaws.

2. Actual Function -----> This is, too often, lost when pushing the above. Early NES games suffer this the most, as some are not playable (same as civ V mp). This also includes bugs, broken controls, and everything else that inhibits one from interacting with the gameplay itself. Civ has historically been terrible here. The ability for players to seamlessly connect to the gaming experience is very important.

3. Project Management ------> I've had a few courses on this in grad school. Simply put, this is a rather difficult field to do well in as a manager and the pitfalls can be immense. Simply building a project plan without major conflicts and proper utilization of physical and human assets is a chore. Planning for contingencies (to the extent possible) on a reasonable cost/benefit basis, choosing the right amount of actual hirings, hiring the right people, tightening the process of design such that it's structured enough to give people a chance to do well, and somehow juggling budget ranges from a big challenge to a nightmare. I'm an arrogant person to some extent, so I think I can do it better than most :p (I handle stress, particularly but not limited to time pressure, very well and tend to be very pragmatic when it comes to getting things done...and project management needs a heaping helping of pragmatism). I'm not stupid though, I'd need some heavy actual experience or a tremendous amount of luck to not screw up such a large undertaking as game design at the level of a major flagship title, such as civ V was (is? I still feel it's in a beta state :lol:).

Programming adds another layer too. Similar to engineering projects, do you hire a project manager that is really good at programming, or one that is experienced with the reality that is project management? Non engineers have botched construction projects to the point where people have died in the past (a better manager would defer to expert experience and not try to rush the project at the risk of human safety, but I digress....). However, people put in that role lacking project management often get their projects abandoned or wind up ridiculously over-budget and taking forever because they have unrealistic estimation of how long things will take, how many people they need, that things will go perfectly, etc. Finding a project manager that has in-field experience but is ALSO personable and competent at project management itself is of course ideal. Good luck though...those people aren't common :p.

Nobody will die in programming hopefully, but when you have cross-company project pressure (especially with ignorant fools like mr 2k "turn based strategy is old and unpopular, we're making xcom a shooter now and gamers need to adapt") it's easy to see how even small unforeseen complications can result in disaster. The right answer here isn't easy. As a player, I'm annoyed that firaxis/2k chose to stain their name by releasing a beta, but I'm cynical and I've been trained to think in business terms:

1. Despite it being a shoddy product and far from being complete, a large portion of the market accepted civ V beta public sale release with open arms and still do so.
2. My guess is that they (accurately) estimated that the title would do very well in sales regardless, and projects do cost money to continue to refine...
3. 2k's recent public statement make me wonder very much their intention of releasing or even bothering to develop another flagship TBS. If for whatever reason management decides tbs are not going to be viable long-term, deliberately releasing an unfinished product and milking max money before dumping the studio or changing the series entirely makes business sense...especially when you can get a mixed or even somewhat positive public response in the past (or pin the blame on developers with their hands tied!). I'm not saying this is what happened because I have no idea of the relationship between firaxis/2k or either's long term plans, but my point is that there are viable, if cynical, arguments for firaxis/2k to behave in ways that are overtly unfavorable to gamers who want a complete and fluid experience.
4. The specific project management design goal could therefore have been on a profitable game rather than a great one. In business terms this trades long-term reputation effects and customer loyalty (which *is* profitable) for short term income...very viable if you don't think long-term prospects good. Could I make a good game in that environment? Could anyone? It would be very difficult to do it. Ironically, this is not unlike some decision making in civ itself...you get loyal allies and milk them until it looks like they can't be useful anymore...and then they'd better watch their backs :lol:.

Nobody in business will ever word it that way, but actions reflect the reality. Time will tell 2k/firaxis choice very well.

TL : DR

Could I make a good game? I believe I could, but it would be difficult to do it alone. Even games that were essentially start-ups like minecraft have continued to add things over time with increasing help/suggestions/implementations from others. It's an interesting model though, one I might have considered if I'd have gone the programming route in schooling...but then I'd be lacking a lot of the skills I currently possess. Don't tempt me to go back to school X_X.
 
Don't tempt me to go back to school X_X.

I hear Notch is now really wealthy. :p

Besides, if Civ players fully dedicated into a project like making a Civ type game, the results could be incredible.

Just by looking at some Civ mods, there are in this very forum alone a lot of people with great skills in making fun yet challenging games, and these people don't get nearly half as much recognition as they deserve.

Maybe I'm dreaming too much (I often dream about my Civ game, and it would be so cool) and am already setting my hopes too high for something that will probably never happen, but isn't all life just a big dream?
 
I hear Notch is now really wealthy. :p

Besides, if Civ players fully dedicated into a project like making a Civ type game, the results could be incredible.

Just by looking at some Civ mods, there are in this very forum alone a lot of people with great skills in making fun yet challenging games, and these people don't get nearly half as much recognition as they deserve.

Maybe I'm dreaming too much (I often dream about my Civ game, and it would be so cool) and am already setting my hopes too high for something that will probably never happen, but isn't all life just a big dream?

What are the reasons you yourself don't do it? You don't have to give the answer here, but remember it when dreaming of such a product from others ;). We all have our reasons in life, and things we care about more or less. Human beings govern themselves short of their potential routinely, but only if something like this project were to be a true passion of someone with requisite skills could such a thing reach the lofty level of your/civ player dreams...and it's not exactly like Firaxis would just hand the rights over ;).

Also, even if people can break the barriers that usually hold ambition of this level back, are they willing to pay the time, focus, and effort costs on a project like this? In my own personal life, I've come upon the dual tug of pursuing one field greatness vs more experiences but less notoriety several times.

For me, I much prefer the role of entertainer, and have for a long time. My youtube series is a hobby, no quitting the day job for that ;). However, if I found a way to do so, I would. I've put together over 800 videos (I'll break 1000 this year) and it's never stopped being fun. This despite a skillset that, by training, is better suited to management, accounting, etc. Fortunately, my work currently involves a good bit of human interaction/communication also. I'm not rich, but I do enjoy that aspect.

Can someone reading this say it? That they'd love the process of putting something like the "ultimate civ game" (probably multiple times :lol:) out enough to commit their all to it and never burn out? Does that person have the management/delegation/training/financial to make it reality? There's a risk there with a big chasm. I could run such a project for a while, but I wouldn't enjoy doing that for long. I'd be more than happy to support it though.
 
Making mods is already a quite undertaking task, and most of the biggest and famed mods in civ IV are actually concretions of code from various sources and made in the course of half decade. Try to follow the development history of a mod like FfH... Just for a quick example, FfH has some code from BUG, including some sniplets that come straight from some very early vanilla mods. This is a very diferent enviroment from a development studio, where the thing that it is assured that the devs don't have is time :D

Speaking of development enviroments ...Well, in terms of Civ, IMHO 2K management has been shooting in the foot left and right since they bought Firaxis. Civ IV was basically a game with good foundations but with a terrible x-pak policy ( Warlords and BtS were good expansions, I'm not debating that ( kudos to the devs ), but they piled issue after issue due to unfinished features ... the vassal system and the AP are good examples ) and a quite iffy patching...

But IMHO the worst that 2k did was the management of civ IV spinoffs ( yup, even worse than the fact that Civ V being in the sorry state it still is after almost a year ... ). CivCol was a complete mess gamewise and was never properly fixed, a rip-off of a 20 years title and letting FfF II standalone go Elemental :p was pretty much throwing a good modder's work and ideas to the competition...

Anyway, getting back to the point, OFC that in theory anyone that has the necessary computer skills can make the "ultimate civ game". But let's be honest, not everyone has the stomach or inclination to be like Dwarf fortress Toady and to devote years of work to a game alone while living off donations :D
 
I don't mind spending a good deal of my free time modding. I mostly do it as a way to enhance my own game, but I also enjoy the feeling of contributing to the community of players. I could never be a project manager, because I'd probably kill all of the staff. :lol:

I think it would be great to make a Civ version that combined some of the best ideas from the community. Not a total conversion mod, but an actual new game version.

Unfortunately, we'll need Firaxis to release the game engine to do that. :(
 
TMIT, I can only dream, I'm 17 years old and have no experience whatsoever. :lol:

I have some limited knowledge about XML from what I've learned here. Python or SDK, for example, I only know how to copy-paste! :p

Besides, I'm about to enter college to study Economics. It's not like I'll have the time or chance to stop and learn Programming 101 any time soon. I'd just like to support the Civ series I so much enjoyed in the past to prevent it from getting ruined, as it is atm. ;)

Anyway, if some attempt is made by Civ players to rescue or salvage what is left of the Civ series, I'd be most happy to be able to help wherever I can.
I spent a lot of time of late childhood and early teenage years playing Civ3, and even today I play some considerable amount of Civ4. It's not something I'm willing to let go off easily. :)
 
I don't mind spending a good deal of my free time modding. I mostly do it as a way to enhance my own game, but I also enjoy the feeling of contributing to the community of players. I could never be a project manager, because I'd probably kill all of the staff. :lol:

I think it would be great to make a Civ version that combined some of the best ideas from the community. Not a total conversion mod, but an actual new game version.

Unfortunately, we'll need Firaxis to release the game engine to do that. :(

Dennis Shirk said that they were working on getting the source code released. I imagine that they'll need to get the game finalized first so it could take 6+ months before we see that.

I do think that there are many great members on these forums that could make a great game. It's a great community here. :)
 
Back
Top Bottom