It's 2005, why does the AI still need to cheat?

[smellymummy]

Don't forget that the AI doesn't get to choose the map, can't obuse rules, has to be capable for all maps, not just pangea or continents.

yeah so in that sense ... it's 2005, why do we still need to cheat to beat the AI!

[Capt Dizle]

Well, a couple of points that may or may not be relevant. First, many of us complained that we wanted more stuff to do in the game. We didn't like the simplified game to make a more competive AI concept.

But, by adding stuff to do, we make it harder for the AI to compete. So, I think, rather than be bored with the game because of limited stuff to do, I'd rather see the AI get advantages that translate into making me scramble to overcome. That would be more fun IMO.

And, and this is important, more stuff to do translates into better multiplayer fun.

The other thing is that if the AI were equal to humans we'd only win our fair share. With the possibility of 18 civs in the game, would we be satisfied to win only one in 18 games. Hardly, I think.

I come down on the side of more stuff, as long as the AI can make an entertaining stand, I am going to be satisfied.

[Gaal Dornik]

Originally posted by smellymummy
and on the topic of AI programming, I'm surprised no one's mentioned database driven AI. apparently people have done nice things with that.

Nah thats hardly possible. Way too different environment. Even in chess, with same field and same units everytime, make one unusual move( like pawn b2-b3) and the whole database is doomed. Very helpful vs PC btw

[Capt Dizle]

And one more thing. Just how many of us in a tight squeeze make a mistake and go back and reload. I mean, you've invested hours and make a blunder. Come on, all of us do it. And anyone who doesn't is probably legally insane.

But, does the AI get this option? No.

I'd like to see Firaxis put that in an expansion. You're playing, doing great and then you get a message.. "The AI has decided to reload"... and you are kicked back ten turns or so. With no option but to deal with it.

Hey, the AI ought to have all our cheats! Right?

[alva]

You know what would happen, no?

We would just counter reload!!!

Hmm, wouldn't really be fun as it would mean that in the end every game ends at 4000BC.

[Aileron]

This is from the link that smellymummy provided. Note the "response" is from Sid Meier himself, so if this doesn't answer the OP, then I don't know what does:

"7. By truthsearch (249536):
AI:
I've been a huge fan of Civilization since it first came out. I've always thought the AI of the computer player is relatively good, especially how each has certain characteristics which differentiate them. But AI in strategy games doesn't seemed to have advanced drastically in the last 15 years. What do you imagine the next big advance in game AI will be? When will games really learn how you play? Will we not be able to tell the difference between a human and computer competitor? I probably shouldn't be telling you this, but in my opinion, the goal of AI is not necessarily to simulate a human response. The goal is to generate interest for the player by providing the illusion of a human-like response - or not at all human-like, if that's what it takes to engage the player. I'm not entirely sure that complex games like Civ could ever have true human responses because there is so much complexity that the AI would bring almost any machine to its knees.

Response:
Consider this: we have only recently been able to truly simulate intelligence that can compete with a human in chess. Chess is obviously a complex intellectual game, but it is ultimately fairly easy to define because there are only 64 squares and 6 types of movement. Plus, the rules of engagement are simple - attack and win. Add to that the huge amount of known strategy that has been collected and studied throughout the years and it is even more definable. In a game like Civ, we have over 80 units, all with different movement rates, strengths, special abilities, experience levels, etc. We also have to decide where to place cities, what to build, who to be nice to and who to make war with. We also have to decide what to research, what religion to spread, what Civics to adopt, etc. All in all, I don't expect to see anything close to true human intelligence any time soon, as long as games continue to get more complex."

[Traelin]

Originally posted by GodSpawn
Non-programmers - or even programmers who have never tried or researched it - don't understand how brain crunchingly HARD it is to program decent AI or heuristics for anything other than games with trivial rules (such as chess, backgammon or go).

It's really REALLY hard! And if one was to do a little research, one would realise that AI programming has not gone very far in the last 10 years...

You are dead-on, without a doubt. AI classes were the only ones I had problems with in college, and they just got progressively worse in grad school. Man I hated Lisp, that language messed me up for life.

[piotrr]

While we're on the topic of the sorry state of Artificial Intelligences today:

How come we still use HUMAN leadership in government? Huh? Our computers are really, really, really powerful today, so why the humans? Huh?

/ Per

[piotrr]

They could also invest in a learning AI.

But then that too leads to an extreme amount of trouble, and I don't even mean in the way Arnold Schwartzenegger puts it in Terminator - just imagine trying to perform TESTING on a game that keeps CHANGING while you try to test it.

/ Per

[theNiceOne]

Saying that the number of options to consider for two warriors is 18, is not necessarily wrong, it's just that you get an extremely poor AI by doing so.

Assume the following cIV AI:
It goes through all his units in a pre-determined order.

* For each unit, he considers the current map situation, including the position of all other (known) units, but does not consider where other units may possibly move.

As we all know by now, the number of possible moves for the first warrior will then be 9 (assuming it can do nothing but move one square, or stay put). And since this AI doesn't consider the possible moves of other units, 9 is the number of possible outcomes to consider for this unit.

When doing the next warrior, the AI knows where the first warrior is, so again, it's only 9 possible outcomes to consider, for a total of 18.


Such an AI would be relatively easy to program. The problem with it is that it would totally suck. The first warrior would not consider that the other units could follow up his attack, or that a boat could move him after having been loaded. So when attacking, the warrior would have to consider its own odds for winning a battle, without thinking about whether other units could follow on an mop up.

[LowLife]

IMHO, someday someone will come up with a sort of flickr/delicious model of aggregating human player moves in a services-based game. The exact game situation and move will be stored in a database, along with the downstream results and the estimted skill level of the player. At some point you'll have enough moves aggregated to create a sort of folksplayonomy of potential moves that an AI can draw from. The AI could take a situation - 4 grassland squares away from a scantily defended city with a stack of 2 move units, the best players do this this or this, I'll do one of them. Or - I have a bomber with a chance to bomb A B or C, the most successful players bomb the city with the least defenders, that's what I'll do. Etc. In the beginning the AI would play *straight*, like the out of the box CIV AIs, but gradually the folksplaysonomy would be melded into the AI's decision making routine.

[Node]

Originally posted by Dr Zoidberg
So true. Anyone can ***** and moan about "crappy" AI´s, but very few can actually program a decent one. But there´s no pleasing some people...

believe it or not, you are allowed to complain about things that you aren't an expert about. If that was the case then no one may ever say that a movie sucks becuase you are not a movie director.

[Urban Ranger]

Originally posted by GodSpawn
Maidel, you're wrong. As far as the AI goes, you'd have to consider all the possible moves - and the number of permutations goes up exponentially, as others have already pointed out.

I don't believe he's wrong.

The moves for the three warriors are independent unless they are in close proximity of each other.

Therefore, he's not completely right, but mostly right.

[Urban Ranger]

Originally posted by padillah
So, to make a CIV game without "cheating" or presenting rewards in any way inconsistent with the rules, you would need to create 162 seperate and distinct AI's.

You don't, you just need 162 sets of weighing factors.

[Urban Ranger]

Originally posted by Fried-Psitalon
The CIV AI is new.

Hm, yes, but it "knows" how to play the game. You didn't have n00bs who didn't know how to play the game program the AI, right? Thus, the knowledge of the programmers were infused into the routines.

As a result, comparing the AI to a n00b who hasn't played a game of Civ 4 before makes completely no sense whatsoever.