AAI 0.63 released (win + linux version included)

[esteroth12]

Sounds great, Acid :)

NTai isn't tough enough, and Kai cheats, so I'm looking to a better skirmish AI.

"Did you upgrade to NTai 8.1 RC21 or newer? The one in the installer, XE7.5, is a half of a year old. And if so, did you do .cheat .aicheat?"

and who cares if the AI cheats? OTA's ai resource cheated...

[AF]

Tthe OTA ai did nto use resource cheats, onyl the modified AI configs released by certain OTA members made the OTA cheat by giving it access to AI only buildings.

There's also a lot of people who would prefer if the AI cheated if they wanted it to cheat else for it to play fairly.

I dont like how people are making comments on AI's and not suggesting improvements, they're just making sweeping comments, a lot of NTai comments are actualyl amde by people using the 6 month old XE7.5, and a lot fo KAI comments disregard what krogothe said about no gameplay comments and expect its a final version and not an early beta version.

So pelase, if your AAI doesnt work, your KAI is doing strange things, and your NTai's a pansy, make sure you have the latest version, then look for the answer before making comments and asking questions.

[KDR_11k]

How many people here can really compile an AI on their own? Most here wouldn't even know what a compiler is. What's not in binary form doesn't exist to the majority of players.

[unpossible]

So pelase, if your AAI doesnt work, your KAI is doing strange things, and your NTai's a pansy, make sure you have the latest version, then look for the answer before making comments and asking questions.

too many children with attention spans that are far to short. myAIdusn'tworkhelpmeplzthx

[hollowsoul]

How many people here can really compile an AI on their own? Most here wouldn't even know what a compiler is. What's not in binary form doesn't exist to the majority of players.

Damn windows users forgetting about us linux users again... j/k

[submarine]

Does AAI use some system to determine where the enemy is, what areas are important and where the enemy placed defence? If so, the builders could use this as well. Or invent some system Would make the AI much smarter!

well infact aai does already something similiar:

on the one hand it should not send its builders to sectors where it has scouted enemy defences (however since aai doenst use any maphack it might not always detect defenses properly)

second it remembers in which sectors it lost which kind of units, so if aai is loosing some builders in the same sector within a short period of time, it will stop to send further builders to that sector

[Vassago]

Most of us are not AI programmers, so how could we possibly offer anything BUT sweeping comments on the matter?

Having the OPTION to have the AI cheat is one thing, but to have it cheat all the time by default? That's honestly quite stupid.

[Erom]

That's honestly quite stupid.

I dunno, most commercial games have cheating AI's. I'd go so far as to say that AI's produced by companies are always concerned about giving you the best game they can, rather than some sort of "fairness" concern, and most companies answer that by allowing the AI's little "cheats" to make up for blatant human superiority.

And come on, map hacks. It's not like it's doing NTai's .aicheat routine automatically.

[Vassago]

You make AI SKILLED, not a hacker.

You don't let it exploit or cheat.

AI in real video games is vastly differant. You don't make the AI cheat or exploit whatsoever. You CAN'T.
You make it smarter or more agile, things like that. Make a weak unit run from a strong unit. Every unit seeks cover, etc etc.

Exploits are not forgivable. Like I said, if I feel like cheating, then it's nice to have the option to let the AI cheat. But it should not default that way.

[rattle]

Just imagine KAI would actually build the heavy artillery units/buildings in EE Epic... though it'd be nice to turn such features on manually.

[BvDorp]

You make AI SKILLED, not a hacker.

You don't let it exploit or cheat.

AI in real video games is vastly differant. You don't make the AI cheat or exploit whatsoever. You CAN'T.
You make it smarter or more agile, things like that. Make a weak unit run from a strong unit. Every unit seeks cover, etc etc.

Exploits are not forgivable. Like I said, if I feel like cheating, then it's nice to have the option to let the AI cheat. But it should not default that way.

Dude, what are you complaining about? Your thoughts about what AI's should and shouldn't do aren't that important, now are they? And IMO this _is_ a real videogame, thnx for the disrespect. Forgivable.. c'mon

Then again, this discussion has been done over and over agan. Leave it to the AI makers, they'll do what they're good at: make the best AI possible.

[AF]

You cant make sweeping comments

Here are two examples:

[IVP]1v0ry_k1ng suggested I allow con units to do AAI style defences in order to help spread out defences.

That was a specific constructive criticism

Vassago said: NTai isn't tough enough

This was a sweeping comment

BvDorp said: Or invent some system Wink Would make the AI much smarter!

This isn't very helpful

Lindir said: Here is how I would make an anti-stall algorithm: *followed by pseudo code*

This was very helpful

These 4 people have never compiled an AI to the best of my knowledge, they've never written an AI although they may have given suggestions and made ltos fo comments and written pseudo code, but they're not AI devs, and they show 2 opposite ends. Lindir and [IVP]1v0ry_k1ng made useful comments that gave an idea, an insight or information, and they weren't ambiguous. for example [IVP]1v0ry_k1ng's comment was self explanatory, it was you should make ti build defences near mexes, it was you should build a defence after every mex like AAI, which is much more understandable because it fits into the construction system and has a set out routine.

Whether lindirs new stall algorithm worked or [IVP]1v0ry_k1ng's suggestion failed utterly is besides the point, they had a use and they where simple and constructive, much more constructive than "NTai isn't tough", why isn't it though? Is there anything specific? How would you solve that? What do you think it should do?

Don't tell submarine AAI isn't good enough at xxxxxx. Tell him why you think its failing and what would solve it, and the same goes for other AI's

[Vassago]

Dude, what are you complaining about? Your thoughts about what AI's should and shouldn't do aren't that important, now are they? And IMO this _is_ a real videogame, thnx for the disrespect. Forgivable.. c'mon.

If our comments and thoughts were not important, this thread would have 1 thing in it - a link to the executable. It would be locked, and not accepting comments.
Without the public's feedback, a game gets nowhere. It's called testing, ya know? We test, then give feedback. Simple premise.

And by "real" game, I'm talking about retail. Not open source freebies. Spring is an amazing project, but by no means is it a real game. It's a hobby :)

And I meant no disrespect about it.

[AF]

Actually I find a threads purpose in AI has shifted more to reporting errors and crashbugs rather than giving general AI feedback and suggestions.

Infact if you follow KAI the whole point of that thread is error reports, he wont look at anything else.

[Vassago]

Okay very well.

I'll never offer improvements or advice to anyone.

[AF]

Dont twist ym words. i was just saying dotn make generic statements, try to make them useful rather than vague

[ILMTitan]

Vassago, first off I only know of one 'real' game in which the AI does not cheat as you increase its difficulty level (although there are some AIs that don't cheat but don't scale in difficulty, one being TA).

Second, I believe AF and all of the other AI devs are trying to make a smart, crafty AI, but that is a lot of work and a long way off, so they use hacks to create a challenging AI, which is what I believe the majority of the community want first and foremost.

Finally, AF gave you a good example of how to offer advice or improvements. Everyone agrees a skilled AI is a good thing. The only question is how to get one (or several).

[rattle]

The AI in KKND2 didn't cheat iirc and it was a tough nut on higher levels. Though the game didn't have that much of an economy system, which makes AI developement difficult at last. The least you should do is flame each other, work together and accomplish more

Just accept the fact that there are AIs which cheat and others which don't. Adds variety. Wouldn't it be best if this was carried on in a seperated general AI discussion thread rather than in submarine's AAI thread? :)

[AF]

KKND2 is a specific game, the AI can be tailor made to fit the game specifically.

Spring skirmish AI's have to handle a lot of designs and gameplay variations with wide ranging maps and unit types aswell as a wide range of different mods, and a resource system unlike any other.

[Sgt Doom]

Counter Strike AI cheats, you take a turn where the AI doesn't see you, it takes the shortest route to you and gets a cheap headshot.
HL2 AI cheats, they nade you even if they never saw or heard you,
UT AI cheats, turn it to godlike and it's all cheap headshots, even if you are completely invisible or in a vehicle (ut2k4)

All AI cheats, in one way or another, otherwise they might as well kill themselves, saving the player the monotomy of headshotting them for the 500th time.

I honestly don't care if it cheats as long as it's CHALLENGING, not "OMFGH4X!" beats you every round kind AI.
I can whup NTAI, AAI when it worked and KAI (with difficulty)