increasing X by Y yields X+Y
increasing X by 250% yields X + 2.5*X = 3.5X, not 2.5X
If you said "it was increased to a factor of 250% of the original then you're correct.
no wonder our school system math scores are down the tubes. =P. Eh, I've done some pretty bonehead mistakes in my time too. My math prof walked out infront of the class one day and announced "usually x * 0 = 0, except on the last midterm you guys just wrote, where apparently it could equal just about anything."
I'd rather it did cheat than provide such an easy win.
So what happens when you play a good game for 2 hours, only to run into an opponent that's 3 times your size, bent on destroying you, and you have no hope of winning? I get pissed that I wasted my time to get clobbered by a massively cheating computer AI, I dunno about you. Or to run into the computer player who, despite cheating, is still half my size, and I've already won the game before a single shot is fired. There rarely ever seems to be the middle ground where you run into an opponent that you can beat, but just barely. If that happened, then fine, let the computer cheat to get there.
Although I don't agree with everything you said, you raised some good points.
Why do some people then reverse-engineer their games?
Why do people hack or crack? Why do people break into web pages and servers? (other than the prepubescent teenagers so they can deface disney and proclaim their undying love to their 13 year old girlfriends)... Cuz it's a challenge. Why do I take apart things? To figure out how they work. Why do I build useless robotic gadgets? Because it's interesting to create something. Why do I crack programs? Because it's stimulating and challenging (well, only some of the time). I took apart the alpha centauri save game file to see how it worked, learn about all of the things that it stores so I could better understand the game itself. I did the same with MOO1 and MOO2. Why do people play games? Most of the time it's nothing more than to waste time, really, but at least with strategy games you can exercise you're brain while you're at it. Am I insatiable? Maybe. You can't satisfy me with a static AI that will never change. Once you know the weaknesses, it's not fun anymore. And it's never fun playing against ANYTHING that cheats, even AIs. Sure it's still a bit of a challenge to beat them, but it's highly irregular. If I work for a few hours to build up my empire, then run into another one that's 3 times my size and I have no hope of winning, but it's only that big because they cheated like crazy, where's the fun in that? I know as soon as I meet a race in MOO2 if I'm going to win the game or not, so why do I keep playing? Hmm, I don't actually know, but it's fun anyways. If you created an AI that can change, so once I learn the tricks and go back to exploit them, it pulls my shorts up over my head and ties them in a knot. Now that would please me. And I don't think it's impossible. (I'm fighting back the urge to start the "I have a dream" speech =P)
I think that a successful AI model will attempt to model human thinking
I think you'd need to extrapolate on that one a little bit. In what way? I don't think that we'll be able to do something like that very successfully any time soon. What I had envisioned was designing the game while keeping AI in mind from the start, not designing the game then making the AI. There is only a fixed number of fundamental choices that one can make while playing a game, because one is limited by the versatility of the game itself. I'd like to hear from other people who routinely play strategy games, to see how their strategys have evolved. The rest of that paragraph I agree with entirely. It's evolution through randomization. Eventually you'll find something that'll work.
As for why AIs vs AIs won't work, I don't know. If you play 1 million games with AI vs AI, and sort by the order of success, and replay the top, say, 30% with 70% more random guys for another million rounds, lather, rinse, repeat, you should be able to come up with some stuff that's good fodder to wage against real people. Sure, most of them won't work, but I'll bet that some of them will. Combine this with the ability to say, intelligently "randomize" future chances. For example, compare two settings that failed. If I expanded at rate X here, which was slow, and rate Y here, which was fast, and they both failed, maybe I should try (X+Y)/2, and see how that works. Yes I agree that's it is a huge task, but sit down a group of 20 desktop computers to pile through seemlessly ending "random" AI situations for a few weeks, and I think you'll come out with quite a few good things. Hey, you could actually set up a continuous system. If the computer is constantly finding strategies that seem to work, and analyzing the differences, etc, then human players could jump and play a game just like it's playing itself, the computer system that's testing the AIs treats it like any other chance, and analyzes how successful it was, and integrates it with it's future trials... hmmmm.... mind's a racing...
You know, maybe it's not so much of an intelligent AI that I want, it just one that changes. One that'll throw a curveball into my works and screw me up. Have you ever played games to go on a suicide mission? I hate it when the computer opponents do that, because it's unfair and unrealistic, but as much as I hate to admit it, it's challenging to stop someone bent on killing you without caring about themselves, then turn around and face another 6 opponents not too happy with you either.
I guess all of my stubborness here is based on the idea that when I create my strategy for a game, how do I do it? I play, take guesses, watch the computer play, and I fail. I try again, adjust my strategies, watch, fail, etc until I start winning, and I keep this up until I start clobbering everything. Then when I play against real people, I watch what they do. If they're better than I am, I take what they did and adapt my strategies. It doesn't seem like a very complicated process. Maybe I'm dead wrong.
Also there is defintely more than one way to win a game, more than one successful strategy. In fact, there is probably hundreds of different methods that would all be highly successful. The point here is not to make a single computer game that can learn to outclass the single player, but take all of the games that are connected to the internet, and experiment. Allow the best methods to float to the top, and then the AI can evolve collectively, even if it's with help from people tweaking it. Like someone else said, I think it'd be really cool to be able to tweak settings in my AI, and wage it against someone else's AI to see how well it fares, or even wage it against someone else to see how it fares. I could then see what went wrong, adjust it, and try it again. Even if I'm behind the scenes, the end result is that you'd have hundreds of AI modules out there to plug into your game, and have them whoop your ass a few times before you could figure out how to beat them. It's like taking the best from person-vs-person play that you can do anytime, offline.
okokokok I'll stop rambling now, I've probably tired a lot of you out by now. Actually I wonder if anyone will actually get this far down, I just previewed and wow this is huge. I could go on, but I might hurt someone. -- Telek
(ie. in a 30fps game, even with 10% of cpu available to AI, you have 3.3 milliseconds per frame to do all of your AI, including collision detection and pathfinding
On a 500MHz machine, 3.3milliseconds is 1.65 million instructions. That's a LOT to do some work in, considering the level that needs to be done between frames (i.e. not a LOT). That's 49.5 million instructions per second, and an efficient design can do a lot in that amount of time. (IMHIOOC: In my humble ignorant opinion of course)
I was replying to your "there's no simple fix" statement.
Look at it this way: The game sells 10,000 copies, and probably 50,000 people are using it. Assume they each play for 100 hours in the first month, for 10 games. That's 500,000 games and 5 million man-hours of testing time that the AI has to play with. Multiply that by a possible 7 AI players per game, and you're looking at 3.5 million games. These are conservative numbers, and by no means scientific (ok ok ok I pulled them out of my ass, but they're reasonable). I don't care if you have 500 variables to play with for an AI, or 5000. With 3.5 million games in the first month, your AI can have explored a LOT of ground towards figuring out what's good and what's not. It's like a distributed net concept for AI development. Man, I should patent this idea.;-P
I disagree. I've watched the computer play before, and quite frankly, the only way that they do any good is by cheating, big time.
Take MOO2 for example. Play on impossible, go check out the race stats of your opponents, and add up the pick points. I've seen the psilons with +11 pick points in total over the maximum. Then I decided, hey, let's swap into hotseat and try to play one of the computer opponents. Check out their colonies. For example, when I loaded they were at +51 money, +11 food and 368RP. After clicking on the TAX rates, and not changing it, it dropped to -44. Food fell to -8 and RPs to 268. The computer couldn't even beat me with 7 pick points higher than I was, they had to cheat at every step along the game as well. I know all this, BTW, because I was playing one time, I had *ideal* starting circumstances, a abundant large gaia in my system, and 3 rich planets in the nearest star, and by the time I met the bulrathi, I had 7 colonies, and they had 18. No joke here folks. There was no way possible, even with mild cheating, that they could handle that.
Also, when I captured their colonies, some of the races never even built facilities like Automated Factories, even though they had the tech. This is just plain bad AI programming. So they make up for it by cheating large.
And for SMAC it's the same thing. I disected the save game format for the colonies, and they were putting their production levels WAY above normal, and never built the proper facilities. It was pretty depressing.
With the current trend to get products out the door ASAP, I can understand why this happens. Finising a product at 80% completion, well, the last 20% probably largely contains AI stuff.
Now I'm not saying that I'm the genious and everyone else is stupid, far from it. I'm sure that they have their reasons, and probably good ones, and insights that I don't have, but I don't see why it should be so hard. This is why I've been itching to discuss this, to see what other people think.
oops I think I lied. I just realized that I've seen the computer with artifacts before. I know that they almost never go for pods in the oceans, and they don't make a point to go looking for them is what I probably should have said. As soon as I can get something that floats I send out scouts to find as many of these pods as possible. Always transports too, so that way I can catch artifacts.
Exactly... The AI never goes for the unity pods, and they give you a HUGE advantage. I'm assuming this because I routinely wander into enemy territory and pick up their pods. Also, I tend to save my artifacts for later in the game, when the free tech will really help. Usually right before I start building my army. There's many tips and tricks that I use, and the computer, since it has a static AI, falls, and falls hard all the time. Yeah, that's not a bad idea. Code AI params in a high level language that the users can edit, not bad...
In my (ignorant probably) opinion, we're designing the AIs the wrong way.
We start with a "decision tree", and figure out how to program the AI to do things "smart"
AIs need to be able to change over time. If you make an AI that can learn, then this can happen, and during the design process you can teach it how to play better
Yes, yes, I know. Everyone thinks that an AI that can learn is way too hard to make, but I disagree in these cases. And besides, properly developed once, it can be used for many different types of strategy games, to spread out the development costs.
Consider this, for those of you who have played moo2
What do I build in my build queues, and in what order?
How fast do I start building colony ships?
When do I start building war ships?
What order do I research technology in?
What type of planets do I take, and what do I avoid?
When new tech becomes available (i.e. colony improvement tech), when do I build it, and which ones do I deem important enough to move to the top of the build queue immediately?
How do I handle diplomacy with the other races?
How aggressive am I? Do I ever attack first?
Do I wait until I'm attacked before attacking others?
Do I wait until I have a huge force, or just enough to start attacking others?
Do I try to capture or kill the colonies? Which ones do I keep?
Do I ever try to get out of a war? When? What do I do to get out?
What weapons do I put on my ships, and how many different designs do I employ? Do I have missile boats and beam only ships, or a mixture of both?
In space combat
What ship do I attack first?
Where do I place my ships that are attacking?
Who advances, and who fires from a distance?
I just came up with this list off the top of my head right now in about 5 minutes. Almost all of these questions can be answered with a variable holding a certain numerical value. With weeks to create this structure, I'm sure that I could come up with many many many variables that can take a value, and then all the computer has to do is observe the human player and how effective they are. Combine this with sharing of this information over the internet, and you can have an AI that would learn very quickly, and effectively too. IMHO of course. Again, see my other post, and I welcome constructive comments/criticism. -- Telek
This is only true on action games. For other games that are purely strategy and don't require hardcore graphics capabilities (Moo2/3, Alpha Centauri to name a few) the AI has a lot more potential. See my other post =)
is that they become predictable. Once you learn the exploits and how they work, the game is no longer fun. Take Alpha Centauri or Master of Orion 2, easily 2 of the best, if not the best, strategy games around (IMHO of course). However I can play both of them on impossible levels and win almost every time.
And what really bugs me is that to make up for deficiencies in their AI, as the levels increase in difficulty, the computer just cheats more. I was abhorred when I found out first hand how badly the AIs cheated at the higher levels in the 2 aforementioned games.
So what my question is, is this: How can this be fixed?
I have a few ideas. One is that you need one that learns. Before you flame me about this, let's think about this for a second. We're not talking about an AI here that can learn how to write a novel, we're talking about relatively straightforward strategies and mechanical play in these games. I know that 95% of of my strategy for these games is down to an art, it's just an automated system until I get to the few points at which I need to make a new decision, or something new crops up. So if I can do this by a predefined strategy, then why can't the computer do that? Keep in mind too that the computer can simply try variations on it's current strategies, and see what happens. If I beat the computer 9 out of 10 times, and one time with some wierd method the computer CLOBBERS me, then hey, maybe it should keep that method around. Also the computer can play against itself, with many different strategies, seeing how each one works. Keep in mind here folks that the strategies that I'm talking about have a few variables: how fast do I expand? at what point to I build an army? how big do I build my army? When do I stop expanding? When to I attack, and who? These can be values that can be changed and experimented with, and hence the computer could learn.
Secondly, one of the things I loved about Alpha Centauri is that just-about all settings were configurable through text files. This was amazing. You could make things easier or harder, change global settings, pollution rates, everything. You could even make new factions and trade them with your friends. If somehow settings for the AI were configurable this way, then people could learn how to tweak the AI to make it a more formidable opponent, and then share this information with others.
Combining those two ideas, throw it on the internet. If you have 5,000 people that are connected (not necessarily at the same time), you can try out hundreds of thousands of strategies for the AI to see what works well, and then upgrade the AI. Actually I think that is a necessity. The AI needs to be easily upgradable, otherwise it'll just get boring as you learn how it works and you can cream the game.
I'd love to hear some (constructive only please) comments about this, as it's been something I've been thinking about for a while.
Want to check out about the new Master of Orion 3? Awesome stuff happening there.
-- Telek
Yes, that was the one I was referring to, and sorry I was too lazy to go searching 'cuz I was on my dialup line at home with a P150 that's slow as molasses...(not to mention paying per minute usage fees, darned europe =P)
I don't believe this... $0.75/pound, say they get what, a ton of it? That's a whopping $1600... If you're going to commit crime, do something worthwhile, geeze. It's like robbing a bank for $1000. Whoopie, lasts you a week until you have to do the same thing again. I always figure that if you're gonna sin, you might as well go for one of the really big ones...;-P
Correct me if I'm wrong, but we seem to live in a world where something is considered done when it's 80% complete and left behind. Money is the top desire, and everything else comes second. If I can get from point A to point B, then why would I spend more of my money to create an alternate path when it's not needed? Management and decision makers have a LARGE tendancy to only look at the bottom line: can we make money of it NOW? No? Seeya!
There was rumours that a internet outage a few weeks ago that affected the @home network was the result of vandals, actually. Unfortunately I have no evidence to confirm or deny this at all, but it's an interesting point. Why would someone want to vandalize an internet line? Then again, why would someone want to vandalize at all? Good thing most criminals are dumb, or we'd be in a whole lotta trouble.
I've seen a whole pile of posts and people thinking that you can "wallpaper" your room or put it on your screen. Think for a second about how sound is made from speakers (or anything for that matter): by vibrations. If things vibrate, it means that they cannot be attached directly to anything except at the edges. You cannot "wallpaper" your room because all of the vibrations would be absorbed by the wall. You could however hang a sheet of this slightly away from the wall, thus giving it room to vibrate. The same thing would happen with a TV screen or your computer monitor. I don't know about you, but the thought of the glass of my screen vibrating at 12,000+ Hz as someone hits the high notes doesn't sound like a very bright idea.
Also, I have here with me a set of Monsoon MM1000s, which are flat panel speakers. Compared to my ($1400, not the best but good IMHO) stereo system at home they sound like crap, however I'm currently in France and I wasn't going to bring my huge towers with me. The flat panels are great for portibility, and considering their size they're really good sounding too. The problem is that you still need an amp for all of this, and the flat panels cannot produce low notes very well at all, so you'll also need a subwoofer.
In any case, speaking as a pseudo-audiophile, in my experience real speakers are still FAR ahead in the game when it comes to raw sound quality. If you want something easy to hook up to your TV or computer to play quake or listen to MP3s, then you can get by on something like this. However if you want quality then you still have to use conventional speakers, and I can't see that changing any time soon, but then again, what do I know?
I saw 21 U.S.Marines, in full dress, with rifles, fire a gun salute to the outgoing president, and every last one of them missed!
Not true at all. I've used normal accelerometers that are very sensitive. Indeed while I was holding it "still" and reading the data realtime, there was quite a bit of movement recorded. I think the only reason why they don't use it is that it's expensive.
Adding to the other comments, it also will most likely use hydrogen canisters to combust, leading to "clean" combusion with an output of water vapour only, so there won't be any bad smelling (not to mention dangerous) odours.
I don't see what the big deal is. Ever notice that when linux has a hole exposed, or netscape has a problem, everyone says "yeah, well it's software, what do you expect?" but when Microsoft has a security hole found everyone is so quick to bash them? Yeah so what, so there's a security hole found... IE6.0 already has this problem fixed, and it's not that big a deal. I don't even use IE/outlook for my email anyways, so I don't care. I never have to worry about any problems, and anything I do use always has any sort of auto-execute options disabled. Simple precaution = no problems. I know that people seem to think that there is a daily quota for MS bashing, but I'm really getting tired of seeing this all the time. Maybe I should start to bash Linux whenever a problem is found there? Don't get me wrong, I use linux as well and have a great deal of respect for it and how far it's come since it began, but can we please stop the immature MS bashing at every opportunity that we get?
Slashdot Mirror
nope!
increasing X by Y yields X+Y
increasing X by 250% yields X + 2.5*X = 3.5X, not 2.5X
If you said "it was increased to a factor of 250% of the original then you're correct.
no wonder our school system math scores are down the tubes. =P. Eh, I've done some pretty bonehead mistakes in my time too. My math prof walked out infront of the class one day and announced "usually x * 0 = 0, except on the last midterm you guys just wrote, where apparently it could equal just about anything."
I'd rather it did cheat than provide such an easy win.
So what happens when you play a good game for 2 hours, only to run into an opponent that's 3 times your size, bent on destroying you, and you have no hope of winning? I get pissed that I wasted my time to get clobbered by a massively cheating computer AI, I dunno about you. Or to run into the computer player who, despite cheating, is still half my size, and I've already won the game before a single shot is fired. There rarely ever seems to be the middle ground where you run into an opponent that you can beat, but just barely. If that happened, then fine, let the computer cheat to get there.
-- Telek
Although I don't agree with everything you said, you raised some good points.
Why do some people then reverse-engineer their games?
Why do people hack or crack? Why do people break into web pages and servers? (other than the prepubescent teenagers so they can deface disney and proclaim their undying love to their 13 year old girlfriends)... Cuz it's a challenge. Why do I take apart things? To figure out how they work. Why do I build useless robotic gadgets? Because it's interesting to create something. Why do I crack programs? Because it's stimulating and challenging (well, only some of the time). I took apart the alpha centauri save game file to see how it worked, learn about all of the things that it stores so I could better understand the game itself. I did the same with MOO1 and MOO2. Why do people play games? Most of the time it's nothing more than to waste time, really, but at least with strategy games you can exercise you're brain while you're at it. Am I insatiable? Maybe. You can't satisfy me with a static AI that will never change. Once you know the weaknesses, it's not fun anymore. And it's never fun playing against ANYTHING that cheats, even AIs. Sure it's still a bit of a challenge to beat them, but it's highly irregular. If I work for a few hours to build up my empire, then run into another one that's 3 times my size and I have no hope of winning, but it's only that big because they cheated like crazy, where's the fun in that? I know as soon as I meet a race in MOO2 if I'm going to win the game or not, so why do I keep playing? Hmm, I don't actually know, but it's fun anyways. If you created an AI that can change, so once I learn the tricks and go back to exploit them, it pulls my shorts up over my head and ties them in a knot. Now that would please me. And I don't think it's impossible. (I'm fighting back the urge to start the "I have a dream" speech =P)
I think that a successful AI model will attempt to model human thinking
I think you'd need to extrapolate on that one a little bit. In what way? I don't think that we'll be able to do something like that very successfully any time soon. What I had envisioned was designing the game while keeping AI in mind from the start, not designing the game then making the AI. There is only a fixed number of fundamental choices that one can make while playing a game, because one is limited by the versatility of the game itself. I'd like to hear from other people who routinely play strategy games, to see how their strategys have evolved. The rest of that paragraph I agree with entirely. It's evolution through randomization. Eventually you'll find something that'll work.
As for why AIs vs AIs won't work, I don't know. If you play 1 million games with AI vs AI, and sort by the order of success, and replay the top, say, 30% with 70% more random guys for another million rounds, lather, rinse, repeat, you should be able to come up with some stuff that's good fodder to wage against real people. Sure, most of them won't work, but I'll bet that some of them will. Combine this with the ability to say, intelligently "randomize" future chances. For example, compare two settings that failed. If I expanded at rate X here, which was slow, and rate Y here, which was fast, and they both failed, maybe I should try (X+Y)/2, and see how that works. Yes I agree that's it is a huge task, but sit down a group of 20 desktop computers to pile through seemlessly ending "random" AI situations for a few weeks, and I think you'll come out with quite a few good things. Hey, you could actually set up a continuous system. If the computer is constantly finding strategies that seem to work, and analyzing the differences, etc, then human players could jump and play a game just like it's playing itself, the computer system that's testing the AIs treats it like any other chance, and analyzes how successful it was, and integrates it with it's future trials... hmmmm.... mind's a racing...
You know, maybe it's not so much of an intelligent AI that I want, it just one that changes. One that'll throw a curveball into my works and screw me up. Have you ever played games to go on a suicide mission? I hate it when the computer opponents do that, because it's unfair and unrealistic, but as much as I hate to admit it, it's challenging to stop someone bent on killing you without caring about themselves, then turn around and face another 6 opponents not too happy with you either.
I guess all of my stubborness here is based on the idea that when I create my strategy for a game, how do I do it? I play, take guesses, watch the computer play, and I fail. I try again, adjust my strategies, watch, fail, etc until I start winning, and I keep this up until I start clobbering everything. Then when I play against real people, I watch what they do. If they're better than I am, I take what they did and adapt my strategies. It doesn't seem like a very complicated process. Maybe I'm dead wrong.
Also there is defintely more than one way to win a game, more than one successful strategy. In fact, there is probably hundreds of different methods that would all be highly successful. The point here is not to make a single computer game that can learn to outclass the single player, but take all of the games that are connected to the internet, and experiment. Allow the best methods to float to the top, and then the AI can evolve collectively, even if it's with help from people tweaking it. Like someone else said, I think it'd be really cool to be able to tweak settings in my AI, and wage it against someone else's AI to see how well it fares, or even wage it against someone else to see how it fares. I could then see what went wrong, adjust it, and try it again. Even if I'm behind the scenes, the end result is that you'd have hundreds of AI modules out there to plug into your game, and have them whoop your ass a few times before you could figure out how to beat them. It's like taking the best from person-vs-person play that you can do anytime, offline.
okokokok I'll stop rambling now, I've probably tired a lot of you out by now. Actually I wonder if anyone will actually get this far down, I just previewed and wow this is huge. I could go on, but I might hurt someone. -- Telek
(ie. in a 30fps game, even with 10% of cpu available to AI, you have 3.3 milliseconds per frame to do all of your AI, including collision detection and pathfinding
On a 500MHz machine, 3.3milliseconds is 1.65 million instructions. That's a LOT to do some work in, considering the level that needs to be done between frames (i.e. not a LOT). That's 49.5 million instructions per second, and an efficient design can do a lot in that amount of time. (IMHIOOC: In my humble ignorant opinion of course)
I was replying to your "there's no simple fix" statement.
;-P
Look at it this way: The game sells 10,000 copies, and probably 50,000 people are using it. Assume they each play for 100 hours in the first month, for 10 games. That's 500,000 games and 5 million man-hours of testing time that the AI has to play with. Multiply that by a possible 7 AI players per game, and you're looking at 3.5 million games. These are conservative numbers, and by no means scientific (ok ok ok I pulled them out of my ass, but they're reasonable). I don't care if you have 500 variables to play with for an AI, or 5000. With 3.5 million games in the first month, your AI can have explored a LOT of ground towards figuring out what's good and what's not. It's like a distributed net concept for AI development. Man, I should patent this idea.
Really? Like which games for example? I'd love to find some more, good, strategy games.
I disagree. I've watched the computer play before, and quite frankly, the only way that they do any good is by cheating, big time.
Take MOO2 for example. Play on impossible, go check out the race stats of your opponents, and add up the pick points. I've seen the psilons with +11 pick points in total over the maximum. Then I decided, hey, let's swap into hotseat and try to play one of the computer opponents. Check out their colonies. For example, when I loaded they were at +51 money, +11 food and 368RP. After clicking on the TAX rates, and not changing it, it dropped to -44. Food fell to -8 and RPs to 268. The computer couldn't even beat me with 7 pick points higher than I was, they had to cheat at every step along the game as well. I know all this, BTW, because I was playing one time, I had *ideal* starting circumstances, a abundant large gaia in my system, and 3 rich planets in the nearest star, and by the time I met the bulrathi, I had 7 colonies, and they had 18. No joke here folks. There was no way possible, even with mild cheating, that they could handle that.
Also, when I captured their colonies, some of the races never even built facilities like Automated Factories, even though they had the tech. This is just plain bad AI programming. So they make up for it by cheating large.
And for SMAC it's the same thing. I disected the save game format for the colonies, and they were putting their production levels WAY above normal, and never built the proper facilities. It was pretty depressing.
With the current trend to get products out the door ASAP, I can understand why this happens. Finising a product at 80% completion, well, the last 20% probably largely contains AI stuff.
Now I'm not saying that I'm the genious and everyone else is stupid, far from it. I'm sure that they have their reasons, and probably good ones, and insights that I don't have, but I don't see why it should be so hard. This is why I've been itching to discuss this, to see what other people think.
oops I think I lied. I just realized that I've seen the computer with artifacts before. I know that they almost never go for pods in the oceans, and they don't make a point to go looking for them is what I probably should have said. As soon as I can get something that floats I send out scouts to find as many of these pods as possible. Always transports too, so that way I can catch artifacts.
Exactly... The AI never goes for the unity pods, and they give you a HUGE advantage. I'm assuming this because I routinely wander into enemy territory and pick up their pods. Also, I tend to save my artifacts for later in the game, when the free tech will really help. Usually right before I start building my army. There's many tips and tricks that I use, and the computer, since it has a static AI, falls, and falls hard all the time. Yeah, that's not a bad idea. Code AI params in a high level language that the users can edit, not bad...
In my (ignorant probably) opinion, we're designing the AIs the wrong way.
We start with a "decision tree", and figure out how to program the AI to do things "smart"
AIs need to be able to change over time. If you make an AI that can learn, then this can happen, and during the design process you can teach it how to play better
Yes, yes, I know. Everyone thinks that an AI that can learn is way too hard to make, but I disagree in these cases. And besides, properly developed once, it can be used for many different types of strategy games, to spread out the development costs.
Consider this, for those of you who have played moo2
I just came up with this list off the top of my head right now in about 5 minutes. Almost all of these questions can be answered with a variable holding a certain numerical value. With weeks to create this structure, I'm sure that I could come up with many many many variables that can take a value, and then all the computer has to do is observe the human player and how effective they are. Combine this with sharing of this information over the internet, and you can have an AI that would learn very quickly, and effectively too. IMHO of course. Again, see my other post, and I welcome constructive comments/criticism. -- Telek
This is only true on action games. For other games that are purely strategy and don't require hardcore graphics capabilities (Moo2/3, Alpha Centauri to name a few) the AI has a lot more potential. See my other post =)
is that they become predictable. Once you learn the exploits and how they work, the game is no longer fun. Take Alpha Centauri or Master of Orion 2, easily 2 of the best, if not the best, strategy games around (IMHO of course). However I can play both of them on impossible levels and win almost every time.
And what really bugs me is that to make up for deficiencies in their AI, as the levels increase in difficulty, the computer just cheats more. I was abhorred when I found out first hand how badly the AIs cheated at the higher levels in the 2 aforementioned games.
So what my question is, is this: How can this be fixed?
I have a few ideas. One is that you need one that learns. Before you flame me about this, let's think about this for a second. We're not talking about an AI here that can learn how to write a novel, we're talking about relatively straightforward strategies and mechanical play in these games. I know that 95% of of my strategy for these games is down to an art, it's just an automated system until I get to the few points at which I need to make a new decision, or something new crops up. So if I can do this by a predefined strategy, then why can't the computer do that? Keep in mind too that the computer can simply try variations on it's current strategies, and see what happens. If I beat the computer 9 out of 10 times, and one time with some wierd method the computer CLOBBERS me, then hey, maybe it should keep that method around. Also the computer can play against itself, with many different strategies, seeing how each one works. Keep in mind here folks that the strategies that I'm talking about have a few variables: how fast do I expand? at what point to I build an army? how big do I build my army? When do I stop expanding? When to I attack, and who? These can be values that can be changed and experimented with, and hence the computer could learn.
Secondly, one of the things I loved about Alpha Centauri is that just-about all settings were configurable through text files. This was amazing. You could make things easier or harder, change global settings, pollution rates, everything. You could even make new factions and trade them with your friends. If somehow settings for the AI were configurable this way, then people could learn how to tweak the AI to make it a more formidable opponent, and then share this information with others.
Combining those two ideas, throw it on the internet. If you have 5,000 people that are connected (not necessarily at the same time), you can try out hundreds of thousands of strategies for the AI to see what works well, and then upgrade the AI. Actually I think that is a necessity. The AI needs to be easily upgradable, otherwise it'll just get boring as you learn how it works and you can cream the game.
I'd love to hear some (constructive only please) comments about this, as it's been something I've been thinking about for a while.
Want to check out about the new Master of Orion 3? Awesome stuff happening there. -- Telek
Yes, that was the one I was referring to, and sorry I was too lazy to go searching 'cuz I was on my dialup line at home with a P150 that's slow as molasses...(not to mention paying per minute usage fees, darned europe =P)
;-P
I don't believe this... $0.75/pound, say they get what, a ton of it? That's a whopping $1600... If you're going to commit crime, do something worthwhile, geeze. It's like robbing a bank for $1000. Whoopie, lasts you a week until you have to do the same thing again. I always figure that if you're gonna sin, you might as well go for one of the really big ones...
-- Telek
Correct me if I'm wrong, but we seem to live in a world where something is considered done when it's 80% complete and left behind. Money is the top desire, and everything else comes second. If I can get from point A to point B, then why would I spend more of my money to create an alternate path when it's not needed? Management and decision makers have a LARGE tendancy to only look at the bottom line: can we make money of it NOW? No? Seeya!
There was rumours that a internet outage a few weeks ago that affected the @home network was the result of vandals, actually. Unfortunately I have no evidence to confirm or deny this at all, but it's an interesting point. Why would someone want to vandalize an internet line? Then again, why would someone want to vandalize at all? Good thing most criminals are dumb, or we'd be in a whole lotta trouble.
I've seen a whole pile of posts and people thinking that you can "wallpaper" your room or put it on your screen. Think for a second about how sound is made from speakers (or anything for that matter): by vibrations. If things vibrate, it means that they cannot be attached directly to anything except at the edges. You cannot "wallpaper" your room because all of the vibrations would be absorbed by the wall. You could however hang a sheet of this slightly away from the wall, thus giving it room to vibrate. The same thing would happen with a TV screen or your computer monitor. I don't know about you, but the thought of the glass of my screen vibrating at 12,000+ Hz as someone hits the high notes doesn't sound like a very bright idea.
Also, I have here with me a set of Monsoon MM1000s, which are flat panel speakers. Compared to my ($1400, not the best but good IMHO) stereo system at home they sound like crap, however I'm currently in France and I wasn't going to bring my huge towers with me. The flat panels are great for portibility, and considering their size they're really good sounding too. The problem is that you still need an amp for all of this, and the flat panels cannot produce low notes very well at all, so you'll also need a subwoofer.
In any case, speaking as a pseudo-audiophile, in my experience real speakers are still FAR ahead in the game when it comes to raw sound quality. If you want something easy to hook up to your TV or computer to play quake or listen to MP3s, then you can get by on something like this. However if you want quality then you still have to use conventional speakers, and I can't see that changing any time soon, but then again, what do I know?
I saw 21 U.S.Marines, in full dress, with rifles, fire a gun salute to the outgoing president, and every last one of them missed!
Not true at all. I've used normal accelerometers that are very sensitive. Indeed while I was holding it "still" and reading the data realtime, there was quite a bit of movement recorded. I think the only reason why they don't use it is that it's expensive.
Adding to the other comments, it also will most likely use hydrogen canisters to combust, leading to "clean" combusion with an output of water vapour only, so there won't be any bad smelling (not to mention dangerous) odours.
Hello??
their batteries have 435Wh/L, compared to normal Li-ions that have 250-300Wh/L. That's plastered all over their site.
Warp drive here we come! That is, if there is such a thing as negative energy...
I don't see what the big deal is. Ever notice that when linux has a hole exposed, or netscape has a problem, everyone says "yeah, well it's software, what do you expect?" but when Microsoft has a security hole found everyone is so quick to bash them? Yeah so what, so there's a security hole found... IE6.0 already has this problem fixed, and it's not that big a deal. I don't even use IE/outlook for my email anyways, so I don't care. I never have to worry about any problems, and anything I do use always has any sort of auto-execute options disabled. Simple precaution = no problems. I know that people seem to think that there is a daily quota for MS bashing, but I'm really getting tired of seeing this all the time. Maybe I should start to bash Linux whenever a problem is found there? Don't get me wrong, I use linux as well and have a great deal of respect for it and how far it's come since it began, but can we please stop the immature MS bashing at every opportunity that we get?