Berkut
Posts: 757
Joined: 5/16/2002 Status: offline
|
I find all this talk about how we hope the AI will be great kind of bemusing. The AI won't be great. Not to be a ass about it, but it is simply foolish to think that the team developing this game has anything like the expertise or resources to create a better AI than all the wargames that came before this one. Hell, programming a really good AI is probably twice as hard, ten times as hard? as programming the entire game to begin with. And really, what is an AI anyway, when we talk about wargames? All it is is a set of heuristics - some rules - that govern what the computer will do. It is a combination, at the end of the day, of some rules that someone sits down and codes, and some scripts. And then it tries to execute those scripts and those rules by evaluating a situation. So here is the thing - assuming that the people doing this are not some insanely intelligent computer science super-brains on the verge of a breakthrough in computer "intelligence", they are going to create the AI in the exact same manner that every AI for a game is made - by hand. And how "good" it is will depend mostly on how much time they can spend coming up with a basic plan. Then coming up with a sub-plan to deal with situation A that is slightly different. And then another sub-plan to deal with situation A1 that is slightly different than A, and A2 that is slightly different than A, and A3, and A4 and A5. Then they can spend some more time coding up situation A1a, and A1b, and A1c. And of course A2a, and A2b, and A2c. And then some playtested will do something they didn't think of, so they will need a set of rules to handle A1a-i, and A1a-ii, and A1s-iii. And on. And on, And on and on. So, the question is - how much time do they really have? And, by the way, a lot of this will end up wasted, since until you get a BUNCH of people playing the game, there are going to be decision trees the people doing the AI have never thought of - the playtesters will get some, but not nearly all of them by any means. So - what am I saying? I am saying it is not even possible to program a competent AI against decent human players. The only way it could be done, is if the rules and scripts were constantly tweaked in response to human play, and it would take a ridiculously large amount of work and would be an ongoing effort, not something they can code up and then call it a day. And even then...well, humans are tricksy beasts, I am still not putting my money on the computer against any decent human player. You guys are asking for them to do something that they could not do even if they had a team of 10 people doing nothing but working on the AI, and they don't have that. The AI, at best, will be competent enough to be a stand in for a human opponent while the human player learns the mechanics of the game, and little more. If you cannot stomp a computer opponent once you learn the mechanics of the game, then either the AI is cheating, or, to be completely blunt, you aren't very bright. I've been playing computer wargames for about as long as there have been computer wargames, and this has always been true, and will continue to be true.
|