I can imagine it's difficult to get the AI to do what you want, even when what you want them to do is both theoretically and practically feasible. Two questions: is the pop up with the information about what the Korath are researching a debug feature or a new espionage interface? And I almost don't even want to ask so as not to cause any more headaches, but any luck in using the same marvelous bit of code to change how the AI evaluates the colonization techs for trade?

Hmm this could be useful. Personally I never had the motivation to dig up what a race is currently researching, as it was buried deep in civilization statistics or some place like that.

With my very limited programming experience I can only imagine whats it like to code a complex system like AI. Now that I have seen couple reacent block buster game releases with terrible AI, I appriciate draginols coding skills even more

It's times like this some forms of weak scripting aren't bad or unrealistic- no matter how much you guys try to stick with your no-scripting rule.

As a player, I don't mind scripting if it fits these two criteria

a) It scripts to avoid something a human player who has played a couple of games would never do

It isn't exploitable by the human

Two areas of particular concern

1) The AI putting the wrong improvement on a bonus tile, when there aren't other non-bonus tiles avaliable.

2) Researching an environmental tech when there's no planet of that type avaliable, or when all known starts are surveyed. Also, in those cases, the value of that tech to the AI should be scripted to 1.

"and on the note sam0t brought up - why the heck don't more big 'blockbuster' games have good AI? is the single player genre just that dead to the populace? *sigh* so much for being anti-social"

In my own pathetically limited experience with AI, I've learned that 'think-ahead' has to be tied to how much the AI knows about how its environment works. In games where the AI isn't learning from scratch, there is level of foresight as to whether the next move will actually produce a desired result. I mean, that's we know better than wasting our time.

Awareness is the key. In systems where the AI is too aware, it's as if they can predict the future. In systems where they aren't aware enough they seem bumbling or random. The amount of awareness an AI has is keyed to how much the AI can sense, and how it senses it, and how it can relate those sensations to what it knows about the task at hand.

...which creates overhead. When you look at a game like, say, Oblivion, that can become staggeringly complex. People complain about pathing issues, etc., but how many billions of years have gone into your ability to maneuver around the environment, and if you codified that subconscious knowledge, what would be the cost of such a resource in terms of system resources?

I doubt it is easier in a game like this, only different. There's a reason that people get years and years of higher education in things like economics, business administration, and even military tactics. Translating that to AI in such a way that the AI performs well without being "God" is a daunting task, no doubt.

Coding AI can be frustrating? It sounds to me like an open admission that coding AI can be frustrating.

All coding *can* be frustrating

Coding AI *is* frustrating.

Especially when some dumbass professor gives you the design... that is bogus... and... ugh... if you code AI you are a better man than I

I think what needs the most work is the diplomacy screen AI. I mean, I've had it where I was offering the AI something like 20,000 credits in exchange for only 2,000 and the AI refused. The AI should *never* refuse any offer like that! I mean, If I offered you 100 dollars in exchange for 10 dollars, there's no way you would refuse. It gets frustrating, because sometimes you really want something that the AI has, like a resource star base, but you can't make them "an offer you can't refuse", you could offer them the grater part of your empire and that stubborn AI would still say something like "do you think I'm stupid?"

I just played my first suicidal game today, and it seemed to me that the only thing that got fixed in the diplomacy screen is that the AI wouldn't let me sell a technology for many thousands of credits, bankrupting them and making myself rich. In fact, there seemed to be a cap of about 1000 credits, past which they wouldn't deal with me.

p with the information about what the Korath are researching a debug feature or a new espionage interface? And I almost don't even want to ask so as not to cause any more headaches, but any luck in using the s

QFT.

I hope it will be in our life time that computers start to code themselves. That would make for some interesting AI.