The Human Factor

[Ned]

What is the most important thing that we humans do better than the AI that if we could not do it (or if the AI could do it) would make the game much more challenging?

Among the things that come to mind are our use of crawlers, strategic military planning, terraforming, unit design, and a builder strategy. All of these are important. But what I came down to as the most important thing that we humans do that the AI does not and which permits us to easily beat the AI is the use of

RUSH BUILD.

Try beating the AI without it.

Ned

[Quantum]

Interesting thought.
Are you saying the AI NEVER rush builds?

Or are you just saying
Human players use rush build more effectively than AI?

Either way it might be a nice challenge to play without rush building.
At least you would have more cash for bribes!

[WhiteElephants]

Alternately you could mossey on over the Multiplaying section of this forum and get yourself signed up for a game or two.

[Kassiopeia]

I have never ever seen the AI rush build. And the fact that you get warned when the AI is near to completing a project, the poor computer is in quite a trouble.

[Avenoct]

I'd have to go with one of the earlier suppositions...like use of crawlers.

It seems to me, but this is just a, ah, conjecture: The AI does rush build in a couple odd ways. The only cases where I observe this are when you beat the AI to a SP...'Ah, too bad Morgan, I've got the Merchant Exchange and you don't have any other SP's to build! At that point the AI gets to use the way-extra minerals for whatever it wants...sometimes it'll spit out units like a pirate On Golden Pond...the other time I've seen something similar...with extra minerals the AI will build 2 things in one turn...if say you kick out the last defender and it was working on a Hybrid Forest....It'll pop in a new defender and return to the Hybrid Forest with no loss to total minerals?!? Maybe I'm seeing things...but I could almost say it built two units in one turn with that trick once...

Thinking about Crawlers...using them is pretty damn tough when you think about terraforming FOR a crawler, and not for the base production directly...I think it's too much to expect any game in this generation of games to do something like that...but who knows?

Strategically, there do seem to be a number of things the AI could do better. First and foremost would be piling up a ton of units, and THEN attacking, rather than dribbling in a few at a time to the front. The AI does do this occasionally with missiles, but its bad with other units, and horrible with sea units...should it be that hard to say...'no, WAIT till you have 4 cruisers full of invaders before you land on the enemy's island!' Well, it's tough, b/c sometimes the AI doesn't have the strategic time or resources to set up a pre-programmed attack. But a ratio could be used: If I have x sea-capable bases, and y minerals production total at those bases, and if I havn't lost z units or bases this last turn, I should shoot for y/x sea transports to launch an attack ALL at ONCE!

And Ned, speaking of making better building AI's...which you've helped to do by changing their dispositions (glad to see you back here posting), you can also change these figures on the tech-end..simply up the interest in Tree Farm Technology to 80,80,80,80 and every AI will shoot for it. I just WISH that you could change this for the actual facility...that's probably still best handled by making the AI interested in build, and nothing else ([-1, 0, 0, 1, 0] in the factiontext, for those of you who've not fiddled with this yet).

Hop over to the creation threads to see some nifty units that one can force feed the AI.

So many of these things are adressable, but most of all I'd like to see the AI's military strategy tuned up. More important would be crawlers, but it's less likely to happen...anyways, it gives some balance to a human player vs. a 'cheating' AI.

-Smack

[Quantum]

To me the most essential difference between a human player and the AI is that a human
can change strategies to adapt to subtle differences in each situation.
Even if the AI was taught to do all of the
things listed above, the human would still win. Because the fundamental limitation of the AI
is that it has to use an algorithim to make choices, and that algorithim can only consider
a finite number of variables. For some (even most) situations those variables might be the
most important factors. But there will always be exceptions when some other variable is the most important factor. The smart human player can always recognize this and adjust strategy accordingly.
The AI cannot - it is stuck using the same algorithim for every situation

A good example of the effects of this is that the AI can't even recognize when its strategy is
failing miserably. I am currently playing the SP version of AI's revenge (very hard, by the way, but only because of massive AI advantages)
and have been under attack at a single base by Miriam and Lal for about 50 turns. Miriam alone
has lost over 50 units, and Lal probably 25. I have lost ONE defender even though they are both
way ahead in tech. A human player would long ago have come to the conclusion "Gee, this
isn't working. Perhaps I should change my strategy". But the AI is stuck using the same old algorithim,
which evidently tells it to "attack base X".

So, all of the suggestions above regarding teaching the AI to use crawlers, rush build, or imrove
attack strategy all boil down to improving the AI's algorithims. But as long as the AI has
to use a set algorithim, the human will always win.

[Ned]

Quantum, This has long been an AI weakness - since CIV. By the way, haven't Miriam or Lal offered a truce? After a while, the AI in CIV and CIV II would. Ned

[Vale Lurk]

I have to agree that Rush Building really gives the human a big advantage. But, I'd like to air a couple of opinions on this subject because I haven't seen anyone else say these things:

Rush building adds considerably to dull micro-management.

Can it really be justified "realistically"? Minerals are needed to make things - isn't converting energy to minerals a bit of a cheat?
I know E=MC2, but converting pure energy into plasma steel is surely no easy thing.

Does it add to the game experience? Being able to freely convert one resource to another detracts from the game IMHO. (I've read that in Civ3, they are going to have difficult-to-get resources which presumably arent convertible. Seems a good idea and a move in the right direction). Just build up lots of energy, and hey presto! your mineral costs are reduced.

What do you guys think?

[D4everman]

I always thought that by rush building you were paying the costs of labor to build something. Kinda like the base needs a recycling tank now and we can't wait until next month when its scheduled to be done so the governor (or basically the player) says "fine. We have some extra money in the budget. Pay the laborers extra to work triple shifts and get it finished."

Its not converting energy into materials....energy is used as currency. Then thats just my opinion.

D4

[Ari Rahikkala]

Besides, you can just weld things together instead of using nails .

[Blake]

I would actually say the thing which lets the AI down the most is it's terraforming.

Basically, given a patch of land the AI would be lucky to make half the raw resources as a human player, they make so little energy that usually about 3/4 of the energy is lost to upkeep. This gets progressivly worse - as the human player creates more and more high energy specialists, and makes full use of forests by building the TF, HF. (meanwhile the AI builds more and more roads, and sets half the population to doctors). This, I believe is the single greatest flaw in the AI, and also one of the most easily remidied (by the programmers, but smart mod design can also force the AI to do good terraforming, altough not good specialist managment).

The AI attacks suck, but this is not a thing easily fixed, this is one flaw I can forgive. Altough I get the distinct impression a lot more effort was put into AI military managment than AI economy managment (AI in the generic sense).

Another big failing of the AI is it's inability to do good diplomacy, trade income is almost invaluable for AI's (which are always on the brink of going bankrupt), yet crazily the higher the difficulty the LESS the AI's try to get treaties and pacts, this is simply self destructive behaivour, born of the misguided notion than a madly agressive enemy is an effective and challenging opponent.

I suspect AI's actually do rushbuy stuff (I've witnessed unit upgrades for sure). The problem is their messed up economy means they usually cant afford to rushbuy stuff.

[Vytae]

I must disagree on the Ai's terraforming habits. They have some pretty decent terraforming its just that it never makes a beeline to the tech and never make its a priority. Ive had Miriam asking for centauri ecology when i was building the ascetic virtues! And i dont think rush build gives humans a huge advantage,its just that they rarely stock up any amount of energy (although ive seen an Ai Morgan stock up to 20k)

I have to agree with Quantum that its the fact the Ai is predictable and cant change strategies. If force/resources/mil etc were all equal in a 1v1 human vs AI,if the Human has played for any amount of time he cant lose. But if it was 1v1 human vs human,you couldnt tell till near the very end who was gonna win =o

[Earwicker]

The AI's use of terraforming is erratic. Occasionally it will make the right choice, particularly with condensors, or mines in rocky squares. However, for the most part it will go for the little-bang for a lotta-buck. Mine/farm combinations in particular: so often I have seen moist squares with extensive terraforming to get a 1-2-1 production, instead of a forest. Sea-forming is woefully underutilized by many factions. And how many times have you invaded the territory of the AI that's had the WP for 100 years and seen not one borehole, despite the dozens of mines and 25 formers parked in New Jerusalem?! Plain silly.

In general, it's the adaptability of the human player that is her chief advantage. One thing I haven't seen mentioned is SE choice. Running FM econommics, it will instigate countless vain vendettas, ship off troops, and construct attack aircraft, all without taking appropriate safeguards. Half its bases in drone riots is no way to win a war. A human player would know that it might be time to go with another combination. Generally, not the AI.

[Quantum]

Originally posted by Ned
Quantum, This has long been an AI weakness - since CIV. By the way, haven't Miriam or Lal offered a truce? After a while, the AI in CIV and CIV II would. Ned

Ned, I would agree, and add that in fact it has been an AI weakness
since the invention of AI, and will continue to be an AI weakness
until someone comes up with a radical innovation in AI design (and
I'm sure that person will become rich doing something much more
important than TBS games )

I guess I interpreted your question a little differently than you intended it.
You were asking what single thing could we teach the AI to make it more
challenging. Your answer - rush building - implies you were not really asking on
a metagame level. My response was more on a metagame level. I was just trying
to make the point that it doesn't really matter what we teach the AI to do, it will
still be inferior to a smart human.

And, by the way, Lal and Miriam have not offered truce becuase in the AI's Revenge
scenario they are set at permenant Vendetta against the human. But it doesn't detract
from my point - they are prohibited from declaring truce but they are not prohibited from
attacking a different base, temporarily suspending the attack to research a better
weapon, etc, yet the algorithim keeps telling them to "attack base X".

[Blake]

The AI terraforming habits cant be defended. most of the time the AI terraforms all of it's territory with the following terraforming:
Farm/Mine/Road
Farm/Solar/Road
Mine (rocky)/Road
Forest
With every rocky tile being mined, (occasionly condensed...), and forest planted only erratically.

2nd. The AI almost never makes use of a mine, because very rarely can it's farmed terraforming make more than 2 food per tile. This means that about 1/5 of it's terrain is not utilised at all. Not used, period. Complete waste. It would be better off programmed to level rocky and do ANYTHING else with them.*
3rd: The solar panels are often on low land, meaning that on average the AI would be lucky to get 2 energy per tile.
4th: The AI is slow to build forest enhancing facilities, meaning it's scant forests are of more limited use.
5th: If the AI uses Free Market, it switches a lot of it's population into doctors, meaning it harvests even less resources, sort of defeating the purpose of FM. If instead it uses planned much of the energy gets lost to ineffic.
6th: Upkeep costs, the AI struggles to pay it's upkeep. SMAC is intentially fairly difficult to make a decent profit in, iow from average terraforming energy income is close to facility upkeep, this makes any high-energy terraforming doubly valuable, because it becomes profit.

*Well, actually in a crazy twist AI crawlers KNOW to crawl rocky tiles, it's just too stupid to build them, and if it does they get cashed in for a SP or prototype. Leaving all these hundreds of mines unutilised. ARG.

What I think is the SMAC AI economy managment got half finished, then they ran out of time/went on to do other things (half finished things, can use crawlers, but wont build them, builds mines, but wont use them, runs FM but cant benefit from it, so on). Prehaps I can excuse this for SMAC, because they didn't fully understand the game dynamics, and ran out of time. But neither excuse is really appliciable for SMAX, some jaw-dropping AI improvments could have been implemented relativly easily, and lack of time is hardly an excuse either, couldn't they just have employed some monkey** to tweak the AI while they the sound/graphics people made the new SMAX factions, techs etc?

**Someone who knows how to play SMAC, and has some experience in Coding.