I am not.

I don't know of any A.I. that is capable of keeping up with a human player after a few games, in any game. the more complex the game the harder the A.I. fails in my experience.

Nope not disappointed at all.  I believe Paul said in the stream that the goal for the AI at "normal" would basically be performing to it's full potential.  That means playing as smart as they can make it play, with the skills and resources the devs have.  With a game having as many variables as Gal Civ 3 has, it would be an incredibly daunting time consuming and expensive task to create an AI that could come anywhere close to a moderately intelligent and experienced gamer.  I mean we are talking about a game that far  far far surpasses chess in it's complexity and number of variables, so it would take a computer AI vastly more powerful than the one's that have been made that can beat human chess players.

So when the devs say that "normal" difficulty is going to be about as smart as they can get the AI, that means this is about the best they can do for an AI with the amount of time and resources they have available.  I won't fault anybody, or be disappointed with them, for giving me their best effort.  And they know that their best effort is probably going to fall short of providing a long term challenge for some (or perhaps many) of the people that will be playing their game.  So they offer levels beyond "normal" difficulty where yes indeed the game does cheat a bit, just to add that extra bit of challenge for those that want it.

So nope, I am not disappointed at all.

In answer to your question Blue, No I am not disappointed at all. See the dev post above for the reason why. I have played GC II with multi core use set to maximum and it will out play you several turns ahead of you in planning.

Hello Math King,

I don't think I'm confused. You're using "Normal" as a marketing term. New improved! Normal is now Maximum!  (That's supposed to sound like a laundry detergent commercial. It's not meant to sound mocking.) I said I am disappointed, because the highest level of play requires giving the computer advantages. I wondered if anyone else was also disappointed. This is interesting to me from a psycho-sociological perspective.

Correct me if I'm wrong, please, but you are saying that you always expect a game to have higher levels of play where the computer is given advantages. I want to thank you for answering my question.

Hello admiralWillyWilber,

Thank you for answering my question, it doesn't necessarily imply that you're disappointed, but I get your drift.

Hello eviator,

Thank you for answering the question, I am grateful.

There will be different opinions on this but I'm going to have to disagree with OP. I do have programming experience with C++/Java/Delphi but I don't do AI. I think trying to scale the AI with difficulty is going to be a huge waste of development time that could be better spent making the AI as intelligent as possible. 

Hello kryo,

First, thank you for your response. I like the way you put it, "the AI is fully enabled at 'normal'", if you'll forgive the out-of-context quote, please.  My question is trying to get at something I think is interesting at a social level.  It only dawned on me when I wondered if, and now using your superior phrasing, and my own lessons learned, , if people expect the higher levels of computer game play above where the AI is fully enabled to require certain advantages.

As a society, we've experienced quite a growth in what are labelled "smart" computer activities in the real world. From computers beating world-class chess players, to public autonomous vehicles in a few years (maybe), and so forth, a lot of rapid progress was made recently.  Only some of it comes from faster chips and cheaper memory and special hardware architecture. A substantial part of that progress relies on algorithms that actually work better the more data thrown at them, or which work well in cooperative, or parallel, distributed environments.

Yet, at the same time it seems people have lower expectations for certain systems. Games are certainly complex, and the difficulty in playing them well mustn't be downplayed. Game playing was a rich source for a lot of computer science work. And one reason is because a game provides a set of rules and a controlled environment. Computers have worked their way up the chain. No human can be the best checkers program.  Eventually, chess fell. And then poker.

A car that can tool around town and on the freeways is an amazing accomplishment. The real-world is a dangerous, unpredictable, complicated place. Poker, a game where one plays the person across from them as much as the cards dealt, is quite a test of what a computer can do, even though it takes place in an environment controlled by rules.

That sparked the question. Do people expect the higher levels of computer game play (those above where the AI is fully enabled) to require certain advantages?

I'm not trying to get at anything else with this question. It isn't an indictment of any company or development team or quality of a game, nor does it imply anything about the level of play of the fully enabled AI, either terrific or poor.

It's simply about what the the GalCivIII gestalt expects.

Hello erischild,

Thank you for your answer. It is much appreciated. One thing I've learned in the answers is that an individual's experience with all-things-computer, including playing other games, helps shape their expectation for GalCivIII. One specific area is the individual's assessment of the state of AI. It would seem this influences expectations for the game, and in both ways. That's a wonderful revelation.

You also raised an interesting point about a past game that "simulated clever play".  Another poster also mentioned that there is more to a good computer AI than winning.  Simulating a human playing GalCivIII, or just winning at GalCivIII?  I would agree, those are two different things but with huge overlap.  I think the former is probably more difficult.

Hello DARCA1213,

Thanks, the laugh is appreciated.

Hello Bamdorf,

Thank you for your answer, I am indebted.

I'll bet you'd like Claude Shannon's chess papers from the fifties and early sixties. There are more board positions in chess than there are atoms in the universe. One interesting note for you Bamdorf, I hope: did you know that chess has a single optimal path? Pretty mind-blowing, considering.

Human players at the highest level look ahead 11 to 20 plies. "Brute force" search is required but is not sufficient for a computer World Champion. As for cheating, both humans and computers "memorize" many chess "book" openings. After that, it's well understood computers play chess differently than a human, but similarities must exist of course. For example, it is cheating if either a human or computer player gets outside help. A human can evaluate about 2 nodes in a move tree, whereas Deep Blue did about 200 million. All computers know the optimum way to evaluate a move tree, yet not all computers play top level chess. Clearly there is more going on than brute force search.

Many (non-chess) solutions today use a combination of techniques to achieve what they do. And many of the most important algorithms in use today actually work better at predicting human behavior the more data points there are. And it can be done without taking a performance hit. Brute force is dead as a dodo. Lot's of data is actually a good thing. Which is why we get to use so much free software and phone apps.

And I agree with you that Civ 4 is/was a great game.

Hi back,

Let me be a little clearer if I can.    I am a bit under the weather today so who knows (better half says what's the difference, lol).    Please understand that I am aware that you probably know most of this, but whatever.

A computer plays chess by building move trees and assigning values to positions.   The important development over the last, oh, say 15 years or is in the position evaluation and the methods of deciding when to abandon a particular line.    Human grandmasters play by looking, sometimes, quite deep into a move tree, but only for a very few lines (1 or 2 maybe).  They have a algorithm also that is quite good for deciding what lines to think about.  That algorithm doesn't work for computers.  They can look at a lot more lines, however.

But what I was trying to get at is the very method of making a chess program play well won't work at all for a 4x game.    It is true that the number of positions to evaluate grows incredibly fast (in chess) as one looks more moves into the possibilities.   So you need a fast computer to look several moves deep.   But the number of possibilites for a 4x game is enormously larger, and trying to assign position values would be hopeless.   So the approach is qualitatively different perforce.  The fact that a computer program can pay world class chess isn't necessarily applicable. That's what I was trying to get at.  It would be of great interest to know more about how this problem is approached in a 4x game, but I think that this is an area where we might find game companies not in a sharing mood.   Don't know for sure, though.  I have read many game company representatives talking about AI but generally in a very vague manner.   If you have any links I would be grateful.   Understandable that details might not be available, I think;  it's a competitive marketplace.  Maybe we should put the question in a live stream.

I was around in the 60s when I was a fair (Not good!) tournament player and most chess players figured they would never see a computer beat a grandmaster.  Shows what we knew.   It was precisely this problem of having to evaluate a gazillion positions that was one of two problems I saw.   The second is that so much of what a good player does to evaluate a position didn't seem like it would translate to a computer.   Oh well.  The computer has to have a game or situation where it can feasibly roll out positions to evaluate.    The problem is drastically more difficult here, it seems to me.   A chess player might be pondering which of 20 possible moves to make (and ruling out 17 or 18 in a flash); in a Galciv3 program I might have hundreds of pieces (ships, colonies, etc.) to work with, most of which have quite a lot of possible moves.  Course, I could be wrong; I was about chess.

Now there is a different question:  would I LIKE to see the AI "not cheat" and play at a level that would challenge experts, if that were possible.   Have to think about that one. It's interesting that chess programs can beat world champions but it has nothing much to do with me, as my player rating was roughly 800 points lower that these people/machines (translation --- I would be road kill).    Sigh.  I am never going to be an "expert" GalCiv3 player.  I am probably going to be normal.   Hmmm, better half might be interested to know that!

Anyway, it is nice to think a little bit about chess after so many years (thanks for mentioning it).   I didn't have the time to keep playing seriously (family, job) and now that I do, I don't have the stamina.   But turn based 4x games work beautifully; no time pressure stress.