A Report Card on Current Development (long)
Moderators: hoijui, Moderators
Well, i wont go into a deep argument (these dont seem to work in this forum)...
Chess has not been figured out yet, the best AIs can only do about a dozen moves ahead, and since a lot of games go on for over 50-100 moves, it will never be 100% figured out within our lifetimes unless quantum computing takes a leap.
I believe that chess is more about strategy than TA, and i also think that a simple hand-held calculator has enough processing power to beat an average chess player, and that therefore my 3.5GHz cpu will be able to beat the average TA player, even if TA is thousands of times more complex, especially since the "average" spring gamer plays a very poor game IMO...
What you stated zaph is a fact that i agree with, but keep in mind difficulty is not necessarily proportional to number of states, and it doesnt necessarily mean KAI wont be beating the average spring player in a few months... It already beats all AIs currently released, and id say it is about 10% developed...
But of course i could be wrong. Everyone is entitled to their opinions.
Chess has not been figured out yet, the best AIs can only do about a dozen moves ahead, and since a lot of games go on for over 50-100 moves, it will never be 100% figured out within our lifetimes unless quantum computing takes a leap.
I believe that chess is more about strategy than TA, and i also think that a simple hand-held calculator has enough processing power to beat an average chess player, and that therefore my 3.5GHz cpu will be able to beat the average TA player, even if TA is thousands of times more complex, especially since the "average" spring gamer plays a very poor game IMO...
What you stated zaph is a fact that i agree with, but keep in mind difficulty is not necessarily proportional to number of states, and it doesnt necessarily mean KAI wont be beating the average spring player in a few months... It already beats all AIs currently released, and id say it is about 10% developed...
But of course i could be wrong. Everyone is entitled to their opinions.
lmao, krogothe have you seen how the machiens that play chess actually work?
The chess AI ahs used the same principle since it was created. IBM made a small ciruit that could play a chess move very well. The problemw as that it could play a chess mvoe and had to iterate, and the fact that chess gametrees are very wide and very deep means you need lots and lots and lots of these thgins working together for long times. The best ches machien is a set of circuits with microprocessor versions of the original circuit that run parallel and exist together in trememdndous numbers
Argh, as I said I was tired and I read that and I got annoyed, very annoyed. I think more that i shouldnt have said talkign down on em but rather that you insist what you say is correct and you didnt show any consideration of our ideas untill we made a big point about them at which point only learning came forward and how AAI works took centre stage. That is what i am annoyed about, and I dont like you saying what I have to do, dismissing what I've said and touting your own words as gospel.
You didnt plant the seeds, you just pushed a mass of clouds out the way long enough for them to grow that little bit higher. If anybody ti si zaphod and I who encouraged the mass of AI devs to start up, they werent even ehre when your posts existed and they dismissed your posts when i resurrcted them after the forum got wiped. They took what ideas I'd written, then they took what code zaphod had written and they used them to jump forward.
Eitherway I had great respect for you till you came out of nowhere and immediatly started lecturing us on how we should code our AI's. By doing that i saw it as a personal attack, and when people dismiss my core ideas about projects I'm on as nonexistant and tout new ideas and totally ignore why I'm here in the first palce, I tend to get very annoyed. Heck I was annoyed by people saying that my threat matrix stuff requried a supercomputer, when really what I'd said would have been a single if statement (see AAI JCAI OTAI NTAI and ZcAIn).
btw a cheating skirmish AI is not currently possible without work arounds, and this si for a good reason, because then anybody could write a GlobalAI helper that ran their abse for them and changed things to give them free units and resources and totally mess the game up without the enemy knowing how or cheats being turned on.
The chess AI ahs used the same principle since it was created. IBM made a small ciruit that could play a chess move very well. The problemw as that it could play a chess mvoe and had to iterate, and the fact that chess gametrees are very wide and very deep means you need lots and lots and lots of these thgins working together for long times. The best ches machien is a set of circuits with microprocessor versions of the original circuit that run parallel and exist together in trememdndous numbers
Argh, as I said I was tired and I read that and I got annoyed, very annoyed. I think more that i shouldnt have said talkign down on em but rather that you insist what you say is correct and you didnt show any consideration of our ideas untill we made a big point about them at which point only learning came forward and how AAI works took centre stage. That is what i am annoyed about, and I dont like you saying what I have to do, dismissing what I've said and touting your own words as gospel.
You didnt plant the seeds, you just pushed a mass of clouds out the way long enough for them to grow that little bit higher. If anybody ti si zaphod and I who encouraged the mass of AI devs to start up, they werent even ehre when your posts existed and they dismissed your posts when i resurrcted them after the forum got wiped. They took what ideas I'd written, then they took what code zaphod had written and they used them to jump forward.
Eitherway I had great respect for you till you came out of nowhere and immediatly started lecturing us on how we should code our AI's. By doing that i saw it as a personal attack, and when people dismiss my core ideas about projects I'm on as nonexistant and tout new ideas and totally ignore why I'm here in the first palce, I tend to get very annoyed. Heck I was annoyed by people saying that my threat matrix stuff requried a supercomputer, when really what I'd said would have been a single if statement (see AAI JCAI OTAI NTAI and ZcAIn).
btw a cheating skirmish AI is not currently possible without work arounds, and this si for a good reason, because then anybody could write a GlobalAI helper that ran their abse for them and changed things to give them free units and resources and totally mess the game up without the enemy knowing how or cheats being turned on.
I'll make NTAI set itself a bigger handicap when it sees .cheat ahs been typed then.
Otherwise the same effect could be gotten by updating the build tree since I am nto an expert strategist at XTA and AA, infact doing that would give a much better overrall improovement in everythignt he AI does than makign ti gt extra resources...
Otherwise the same effect could be gotten by updating the build tree since I am nto an expert strategist at XTA and AA, infact doing that would give a much better overrall improovement in everythignt he AI does than makign ti gt extra resources...
It takes about 400-600 lines of code to make that AI that beat NTAI on CC then both JCAI and NTAI on the cold place on the replays you watched sub. It would beat a beginner that doesnt rush too.
Now to make a chess engine that will beat a beginner... how many lines of code would it take? All the simple engines ive seen are in the 2k+ line range...
If you guys compare the way to make the computer play a chess game to the way an AI should play spring, it wont work, you cant computate any useful number of iterations and plan the next move. If thats the way you think then of course chess is easier but you cant obviously make RTS AIs...
But you can set simple rules like a buildtree, attack coordinates and a mex class and already have an AI play decently (current state of my AI) against other AIs and beginners...
I found making my AI so far easier than making a chess engine, so what the hell is wrong with thinking that?
grill me all you want, i wont go away
Now to make a chess engine that will beat a beginner... how many lines of code would it take? All the simple engines ive seen are in the 2k+ line range...
If you guys compare the way to make the computer play a chess game to the way an AI should play spring, it wont work, you cant computate any useful number of iterations and plan the next move. If thats the way you think then of course chess is easier but you cant obviously make RTS AIs...
But you can set simple rules like a buildtree, attack coordinates and a mex class and already have an AI play decently (current state of my AI) against other AIs and beginners...
I found making my AI so far easier than making a chess engine, so what the hell is wrong with thinking that?
grill me all you want, i wont go away
? You've lost me there, and I think we've lost you too by the looks of it...
Eitherway, just remember that those 400-600 lines balloon when you hardcode your build trees and attempt to support mroe than just XTA Core. Also factoring in attacking too and not just resourcing. Also all line counts of NTAI include header includes comments, white space, #defines, and a number of files that keep getting commented and uncommented every few days. All in all the line count is somewhere around 6000 but it should be smaller I'd say around 2700-3000 if I removed all the comments and the build tree.
But overrall we tried to compare to show it wasnt the same and you cant make a fair compairson. A sort of taking the logic to extremes that ahsnt been fullfilled to the right extent.
And my last post was saying if things continue you mgiht get fed up and bugger off. Not that that would be good because you're making headway in palces others arent and it'd be unfair, but I found it somewhat amusing to see subs reply..
Granted once you've figured out the basic system for a chess AI all you ahve to do is iterate it through a game tree whcih is the simple bit. And as we all know thats a totally different kettle of fish to our AI projects.
Eitherway, just remember that those 400-600 lines balloon when you hardcode your build trees and attempt to support mroe than just XTA Core. Also factoring in attacking too and not just resourcing. Also all line counts of NTAI include header includes comments, white space, #defines, and a number of files that keep getting commented and uncommented every few days. All in all the line count is somewhere around 6000 but it should be smaller I'd say around 2700-3000 if I removed all the comments and the build tree.
But overrall we tried to compare to show it wasnt the same and you cant make a fair compairson. A sort of taking the logic to extremes that ahsnt been fullfilled to the right extent.
And my last post was saying if things continue you mgiht get fed up and bugger off. Not that that would be good because you're making headway in palces others arent and it'd be unfair, but I found it somewhat amusing to see subs reply..
Granted once you've figured out the basic system for a chess AI all you ahve to do is iterate it through a game tree whcih is the simple bit. And as we all know thats a totally different kettle of fish to our AI projects.
I dont get upset and leave that easily (although sometimes i have a short temper ), like i said, grill me, internet flames cant hurt!
I never really used forums much so im not used to writing large amounts of text in a short time, probably the reason why my posts might be misleading, so you guys wont understand me all of the time...
Rest assured, i know what im talking about, even if it doesnt seem so. Its a bit like argh. He knows what hes talking about but the way he presented his ideas made people attack him.
As i get more used to posting hopefully it will all make more sense.
), like i said, grill me, internet flames cant hurt!I never really used forums much so im not used to writing large amounts of text in a short time, probably the reason why my posts might be misleading, so you guys wont understand me all of the time...
Rest assured, i know what im talking about, even if it doesnt seem so. Its a bit like argh. He knows what hes talking about but the way he presented his ideas made people attack him.
As i get more used to posting hopefully it will all make more sense.
But if we are just talking super basic vs a noob, then the chess engine could be super basic too, just use the blitzkrieg opening, win in a few moves.krogothe wrote:It would beat a beginner that doesnt rush too.
Now to make a chess engine that will beat a beginner... how many lines of code would it take? All the simple engines ive seen are in the 2k+ line range...
Im surprised how often that works vs non-chess people.
that would be only a few lines of code.
...but TA can be played without thinking about those complexities, unlike chess...
A good chess player will consider hundreds of moves much alike an optimized machine but a good TA player will use strategies and the overall tactical situtation, instead of counting every unit and their exact coordinates. They wont need minutes or hours to plan their next move, they receive information real-time and process it real-time thinking about their overall game-plan and small details at hand, much unlike chess, where you cant just forget a pawn or two for a few moves or not worry if a bishop is on a white or its adjancent black square.
As for the blitzkrieg opening, it works differently. It will beat a percentages of players of varying levels (mostly low), while my AI will always win if the player is below a certain standard. It is also a lot more adaptable than a chess engine that only does one move and if that fails it simply loses 100% of the time. Blow 99% of my simple AI up and the single builder will eventually rebuild everything it lost and attack you again...
You guys are taking a completely literal comparison between chess and TA, stating the absolute obvious (eg chess has less discrete states than TA) which no one needs to hear about. If you play TA like chess then of course you will find TA harder than chess and suck at it too. I find TA compared to chess requires a lot less deep thought from me and a lot more reflexes and the ability to keep my attention spread around different places, now which of those is a computer better at?
If you guys still dont understand and agree, I dont blame you guys, most (not all, i hope) of you dont even play spring against humans and therefore wouldnt know!I wont repeat myself once more. Make your AIs that play TA like chess and ill make mine that plays TA like TA. We will see who's AI wins out in the end wont we?
A good chess player will consider hundreds of moves much alike an optimized machine but a good TA player will use strategies and the overall tactical situtation, instead of counting every unit and their exact coordinates. They wont need minutes or hours to plan their next move, they receive information real-time and process it real-time thinking about their overall game-plan and small details at hand, much unlike chess, where you cant just forget a pawn or two for a few moves or not worry if a bishop is on a white or its adjancent black square.
As for the blitzkrieg opening, it works differently. It will beat a percentages of players of varying levels (mostly low), while my AI will always win if the player is below a certain standard. It is also a lot more adaptable than a chess engine that only does one move and if that fails it simply loses 100% of the time. Blow 99% of my simple AI up and the single builder will eventually rebuild everything it lost and attack you again...
You guys are taking a completely literal comparison between chess and TA, stating the absolute obvious (eg chess has less discrete states than TA) which no one needs to hear about. If you play TA like chess then of course you will find TA harder than chess and suck at it too. I find TA compared to chess requires a lot less deep thought from me and a lot more reflexes and the ability to keep my attention spread around different places, now which of those is a computer better at?
If you guys still dont understand and agree, I dont blame you guys, most (not all, i hope) of you dont even play spring against humans and therefore wouldnt know!I wont repeat myself once more. Make your AIs that play TA like chess and ill make mine that plays TA like TA. We will see who's AI wins out in the end wont we?
OMG the logic here.... first you compare spring AI with chess, then people disagree with that, then you post that people shouldn't play spring like chess, so neither should an AI...If you guys still dont understand and agree, I dont blame you guys, most (not all, i hope) of you dont even play spring against humans and therefore wouldnt know!I wont repeat myself once more. Make your AIs that play TA like chess and ill make mine that plays TA like TA. We will see who's AI wins out in the end wont we?
Seriously make up your mind... and stop bragging about an AI that only builds resources and does not attack nor do recon, you're just making a fool of yourself.
Fine i wont argue about chess anymore. I am sorry i ever brought it up, really. I made contradictory and stupid statements and so i ask for redemption....
Once you are done with grilling me, you might want to take a look at my infant steps post and see that my AI not only attacks but has beaten JCAI already in its simple state. Its not the king AI yet but its not as crap as you think it is.
Gosh i need a break from these forum arguments for good
Once you are done with grilling me, you might want to take a look at my infant steps post and see that my AI not only attacks but has beaten JCAI already in its simple state. Its not the king AI yet but its not as crap as you think it is.
Gosh i need a break from these forum arguments for good
And I've pointed out that after seeing that i saw the ened to improve NTAI's build tree immediatly, and saw flaws in NTAI vs KAI.
KAI is not to be udnerestimated and I am not taking the threat to NTAI's superiority lightly at all.
Please can we censor the phrase "chess AI" in the phpbb config zaphod? Any mention fo chess AI's in my experience leads to an arguement that ends badly.
KAI is not to be udnerestimated and I am not taking the threat to NTAI's superiority lightly at all.
Please can we censor the phrase "chess AI" in the phpbb config zaphod? Any mention fo chess AI's in my experience leads to an arguement that ends badly.
