To Firaxis: Only thing civ4 Needs IMHO

[Brundlefly]

An AI where the 'I' stands for Intelligence. Right now C3C's AI is more like an 'AS' - artificial stupidity.

Give me an AI that knows how to use armies and naval units, knows how to mount an attack, that knows what to build and where, that knows how to use governments and when to switch. An AI that knows what techs to research and what to beeline for. An AI that doesnt hold me responsible for trade route terminations that are not my fault. An AI that learns from the human player.

Nothing on any of the lists in this forum matters to me more than giving me a smarter, craftier, sneakier computer opponent. Spend development time on that instead of new techs, units, terrain. The game is saturated to the point of excess with these things.

If you dont want to spend time on this, Firaxis, consider allowing users to customize AI routines with programmable modules - this idea has been mentioned by several others on this board. Think of the excitement that could be generated by this option. If anyone wants to play me, they dont have to go through the hassle of PBEM or multiplayer, they can just d/l my modules and use them with the Epic game. Many people will find it is more fun to see if they can program an AI that can beat others than to actually play the game!

So, in summary, no more frills Firaxis. I'm bored with C3C. If you make Civ4 just an extension of Civ3 I will not buy it new. Perhaps 3 years after its release when it is sitting in the bargain bin at $9.99 I will pick it up out of curiosity, but I wont buy it new.

[alva]

not gonna happen, way too expensive to devise a real AI.

An AI where the 'I' stands for Intelligence. Right now C3C's AI is more like an 'AS' - artificial stupidity.

Besides, show me one game with a more then decent AI.

[Cmonkey]

Originally posted by alva
not gonna happen, way too expensive to devise a real AI.

An AI where the 'I' stands for Intelligence. Right now C3C's AI is more like an 'AS' - artificial stupidity.

Besides, show me one game with a more then decent AI.

Way too expensive?

Is it way too expensive to know the game you are programming the AI for? Sometimes I get the impression that the people who programmed civ3 never even heard of the game before... or even read a history book for that matter.

Is it way too expensive to buy a notebook and devise massive flowcharts, sheet by sheet, for the decision trees of the FSM? -- That's Finite State Machine - look it up. Some people say they are "dead" and NN (Neural Nets) & GA's (Genetic Algorithims) are the way of the future, but IMO I (and anyone for that matter) can program a much better and much faster AI with a good FSM than with any "fancy" NN or GA.

Keep in mind, AI in games is not intended to be true AI (that being developed by scientists), it is mearly meant to give the illusion of intelligence behind the entity being portrayed by the AI. Civ3 IMO fails horribly in this regard.

I agree with bfg9000, though, I'm more interested with it being a "realistic-behaving" AI than a "sneaky" one. Oh and I second the open AI, bfg9000 refered to, had C3C had this feature I may still be playing it (with my own AI), instead of having slipped it into a warm pile of cow dung the other day.

[Atahualpa]

If you dont want to spend time on this, Firaxis, consider allowing users to customize AI routines with programmable modules - this idea has been mentioned by several others on this board. Think of the excitement that could be generated by this option. If anyone wants to play me, they dont have to go through the hassle of PBEM or multiplayer, they can just d/l my modules and use them with the Epic game. Many people will find it is more fun to see if they can program an AI that can beat others than to actually play the game!

Its not economical to do that. But yes I am all for it.

Anyway you can engage in the CtP2 Source Code Project and implement your own AI there.

[Skanky Burns]

Its called "Multiplayer".

[Kuciwalker]

An AI where the 'I' stands for Intelligence. Right now C3C's AI is more like an 'AS' - artificial stupidity.

1. There are few better AI's out there

2. The reason is not only cost, but processing power

[Skanky Burns]

That's Finite State Machine - look it up. Some people say they are "dead" and NN (Neural Nets) & GA's (Genetic Algorithims) are the way of the future, but IMO I (and anyone for that matter) can program a much better and much faster AI with a good FSM than with any "fancy" NN or GA.

Wonderful.

An AI that is more predictable than the storyline of a soapie. And as entertaining.

At least it has a certain efficiency to it - they can write the strategy guide for the game before the AI is even programmed, as every state will have an obvious counter.

[Atahualpa]

umm Skanky Burns you can do pretty unpredictable stuff with FSMs. Take some MLFSR's (maximum length feedback shift register; edit: some sources mention them as modified/modular linear feedback shift register; I dont know the exact term...) and combine them through a function and you got a pretty good (depending on the function) Pseudo Random Number Generator.
Its possible to use it in Cryptography.

[Skanky Burns]

While true, I hardly think bfg9000 had the term "moving about at random" in mind when he though of a challenging AI. An AI to be even worthy of the name will need direction and purpose. The point is, you can achieve a much better AI by weighing the options available to persue and picking the one that gives you the most advantage/your opponent(s) the least advantage. Which, funnily enough, is what Civ3's AI does.

[Dauphin]

Originally posted by Skanky Burns
Its called "Multiplayer".

Indeed. I don't play against the computer alone, the biases given to it don't make up for its inherent stupidity.

The action of barbs can be quite intelligent at times.

[alva]

BTW, the CIV3 AI is pretty good, except for people on these boards I'm sure many, if not most, struggle on regent.

The thing is the the programmer is not playing just one player here, but a whole community of intelligent lunatics.

[GePap]

There will always be gamers that learn how to reach full optimization and thus overcome any AI-these types of gamers won;t ever be happy unless we get an AI that is as obsesive about winning.

I think of it this way: the human has one empire to run, with a few things automated.

The AI runs up to 30 empires, all thier automations, and the human player's as well.

[okblacke]

Originally posted by bfg9000
An AI where the 'I' stands for Intelligence. Right now C3C's AI is more like an 'AS' - artificial stupidity.

Not really fair. C3's AI is way better than previous Civs. In fact, I think here on Apolyton there were many raves about the AI when C3 first came out.

Now it's 2 1/2 years later, and we've had plenty of time to map out its weaknesses. Of course it looks stupid. If Civ 4 makes about the same leap AI-wise that Civ 3 did, I'll be happy with it.

It's not going to do anything revolutionary AI wise. The investment's just too steep.

[Aqualung71]

But I think the point here is that because civ3 is now nearly 3 years old, players have been able to work out different successful strategies tailored to specific situations. I would bet that when civ3 was first written, those doing the programming were not strategy experts and in fact could nowhere near match the better players you find on these forums, particularly cis-a-vis early build strategies, tech choics, terrain improvement choices, etc. So, they weren't able to programme the AI with what we now accept as sensible and logical strategies.

So, with the benefit of experience and hindsight, perhaps the programmers will now be able to programme better strategies into the AI sequences based on what are now well researched and accepted "truths" about the game. I don't think this is too much to ask, do you?

[okblacke]

Originally posted by aqualung
So, with the benefit of experience and hindsight, perhaps the programmers will now be able to programme better strategies into the AI sequences based on what are now well researched and accepted "truths" about the game. I don't think this is too much to ask, do you?

I would expect that, to a degree.

But the new game will be different and old rules won't apply.

So, we'll be back to square one. (Almost. Not quite.) That's why Civ 3's AI was pretty impressive.

[okblacke]

Originally posted by Cmonkey
Is it way too expensive to buy a notebook and devise massive flowcharts, sheet by sheet, for the decision trees of the FSM? -- That's Finite State Machine - look it up.

Personally, I don't need to look it up. I've done more than a few.

Take a crack at defining that FSM, would you? I'm not suggesting you describe the whole machine, just show us in broad, basic strokes how you would define, let's say, three states of this machine.

I've actually given this topic a fair amount of thought. I know how it couldn't be done which has given me a theoretical basis for how it might be done. But I'd be intrigued in hearing your thoughts. Perhaps you see a simplification that I don't.

[OPD]

The current AI is very good, and doing the AI is a really hard job.

I wouldn't like to see less gameplay to allow a better AI or the designing of the game is such a way as to make the AI better. As realistically, IMO, civ4 will need to be all about multi player.
In fact I think the whole game should be designed with MP in mind, even if the first release is non-mp.