On the Ball: Man Versus Machine

 Pages PREV 1 2
 

Nuke_em_05:
Robots can be made to do anything better than a human.

Simply because an artificial superior exists, it does not negate the competition.

Sports are all about human vs human, and the relative skill of the two competitors. We don't enjoy a little league game (in it or as parents) less because we know a Major League team could wipe the floor with us. We don't enjoy a running track race less because cars can go faster.

Very well said, I agree completely.

Xersues:

My logic never said nothing in life can ever be a sport, I just said I can't claim its a sport if a computer can consistently do it better.

You know why? Computers are tools. No one ever got an award for wrenching the fastest. No one gets an off on shaping metal anymore, that's what happens to society and technology advances. Flying planes used to be globally renown, now its simplified and turned into a hobby. It ceases to be amazing. If you think what I said was retarded, then by literal context of course it is. It has little substance, because you gave and so did most of the people on here sans the article that much thought that went into it.

Sports are what, events of enjoyment for spectators, and enjoyment for those that do them. When something no longer becomes an event of wonder its socially downgraded to recreation or hobby. No grand stands dedicated to its monument anymore.

What happens when a tool can consistently outperform its human counterparts? Humans adapt, into a realm the tool or AI cannot perform. That in itself is an awesome sport.

Things are not black and white ladies and gentlemen, and things are not stagnate. The "On a long enough time line everything turns to nothing, this logic sucks!" argument is just silly. Think about things just a little bit more, just like I should have when I mentioned that I can't find that(e-sports) amazing. :)

"Sport" and "pro" are what we make them. Have some one tell you they're a professional lamplighter (a real old job) and you'll laugh at them. That's what a light switch is for.

Anyway, you all know the only thing that really matters is what you want to do and have fun with. Everyone only gets one!

Basic counter-example: fighting, as in UFC. Still regarded as a "sport" of sorts, despite the fact that a well-designed robot could easily knock a man, even a body-builder, out. The fact that there exist tools designed to cause damage to humans doesn't seem to diminish people's excitement at human ability to damage each other.
There are other examples detailed in other posts as well.

Humans care about human achievement, human competition. As many others have said, the fact that a machine can outperform humans doesn't seem to phase our enjoyment of a sport.

Your dismissal of eSports because a perfect AI is hypothetically possible comes across as silly. There are many things which are hypothetically possible, including a perfect batting machine, a perfect golfing machine, a perfect pitching machine, etc. Their hypothetical existence does not invalidate the entertainment value of the sports.

Nowadays, computers can consistently outperform nearly all human players at chess, because of tremendous computational power combined with massive game databases. Does this inherently make chess a worse game? A less impressive display of deep strategic skill?

I think not, though your opinion may be different.

Your examples include things which would probably never have been sports, because there's little complexity or skill involved. "Wrenching" doesn't sound like a sport, not because a robot could wrench better, but because it inherently lacks entertainment value. It's just repetitive work. Yet it could be a competition, and be entertaining in a proper context.
"Lamp-Lighting" doesn't sound like a sport not only because we don't have lamps that need lighting anymore, but because it doesn't sound like it requires sufficient skill. Yet it too could be a sport, in a proper context. Perhaps a Reality TV competition or something like that.

Soo, everyone is pretty worked up about this, eh?

Lets put some things first:
AI can learn. At this point in time not very fast, but that's not the point. In time AI will be as functional as humans are.

When you take stuff down to the core, Deep Blue was nothing more than one big fraud. He saw the board, he ran his database of options, and decided. The database, however, was filled with a lot of matches throughout the years. All it did was chose the statisticaly best one.

It did a pretty good job, but it is not quite AI.

SC, but also Warcraft, Halo or WoW, are games. Games have human factor. Soccer is a game too. It has a human factor. Sure, there's robot soccer, but they don't stand upright, nor technically "kick" the ball.

In time they might.

What I'm trying to say is that there's always a human factor. Being either bluff, luck or sheer intelect.
At this point in time, man has all of those more than the AI. AI can't decide to "bluff" or scare the opponent. It might, but not intentionally. It might in the future, but that's very complicated stuff.

In short, modern day AI is not going to cut it. No matter how hard we try, with the technology of today, we cannot safely say AI>Human in anything. It does things too differently.

FYI there are micro bots out there already. http://www.youtube.com/watch?v=K1HaspqVynw people just don't care for them. Also dynamically responding to a game is not something most AI can do very well.

De Ronneman:
When you take stuff down to the core, Deep Blue was nothing more than one big fraud. He saw the board, he ran his database of options, and decided. The database, however, was filled with a lot of matches throughout the years. All it did was chose the statisticaly best one.

It did a pretty good job, but it is not quite AI.

isn't that what intelligence is? the ability to take past instances into consideration when making decisions?

sounds like Deep Blue wasn't a fraud, just was a very simple form of AI.. still an AI. There wasn't a man behind the curtain pulling levers and making it go.

I think you're being a bit too critical of Deep Blue. I've written very simple programs that can do similar things. A friend of mine and myself made a world of warcraft addon that could tank molten core using a single button being spammed over and over.. (and, admittedly required you to move the character yourself, but I digress)

I'd be hard pressed to call what I made AI, but it acted like AI. It was context sensitive. The program knew the situation that it was in and chose from the options I gave it the one best suited for the situation, from a list of situations I gave it.

If I had put several thousand variables into it then a hundred or so though, it might have looked a bit more like an AI.. and if I had the resources to give it unlimited variables, it would definitely be a form of AI.

We're not talking about Data on the USS Enterprise here.. we're just talking about the most basic of pattern recognition.

I've your curious about the mod my friend and I made, for context, it's VERY similar to the Gambit/Tactics systems in Final Fantasy 12 and Dragon Age Origins.. But it was much more advanced, and allowed you to use every single variable that world of warcraft could draw up (which, pre-2.0 was quite a bit.. it was patched into obsolescence, but pre-2.0 it was solid)

Altorin:

De Ronneman:
When you take stuff down to the core, Deep Blue was nothing more than one big fraud. He saw the board, he ran his database of options, and decided. The database, however, was filled with a lot of matches throughout the years. All it did was chose the statisticaly best one.

It did a pretty good job, but it is not quite AI.

isn't that what intelligence is? the ability to take past instances into consideration when making decisions?

sounds like Deep Blue wasn't a fraud, just was a very simple form of AI.. still an AI. There wasn't a man behind the curtain pulling levers and making it go.

I think you're being a bit too critical of Deep Blue. I've written very simple programs that can do similar things. A friend of mine and myself made a world of warcraft addon that could tank molten core using a single button being spammed over and over..

I'd be hard pressed to call what I made AI, but it acted like AI. It was context sensitive. The program knew the situation that it was in and chose from the options I gave it the one best suited for the situation, from a list of situations I gave it.

If I had put several thousand variables into it then a hundred or so though, it might have looked a bit more like an AI.. and if I had the resources to give it unlimited variables, it would definitely be a form of AI.

We're not talking about Data on the USS Enterprise here.. we're just talking about the most basic of pattern recognition.

True, but in this thread Deep Blue is being compared to AI that is supossed to be thinking it's way out of situations far deeper than a chessgame. A direct line is drawn between Deep Blue and "learning" AI. Those are not the same.
Deep Blue dates from 1996/1997. That's not comparable to something as complex as an Artificial Neuro Network.

There's no way that an AI could possibly strategise in a game as complex as StarCraft in real-time. Yes, it could micro super fast .. but you can't micro Observers to kill Marines.

De Ronneman:

True, but in this thread Deep Blue is being compared to AI that is supossed to be thinking it's way out of situations far deeper than a chessgame. A direct line is drawn between Deep Blue and "learning" AI. Those are not the same.
Deep Blue dates from 1996/1997. That's not comparable to something as complex as an Artificial Neuro Network.

I wasn't really disagreeing with you. Might have sounded that way but it wasn't. Deep Blue, objectively, was a piece of shit by today's standards.. I was just saying that it was AI, in the sense that most people that aren't writing science fiction describe AI.

A program like Deep Blue would be hard pressed to play Starcraft.

However, I'm pretty sure I saw a video of a Blind kid playing Starcraft, and although I'm not really comparing his mind to a computer, his METHOD of playing would be quite similar to how a computer might play.

Of course, the blind guy loses, but it's only because he shows some of the same drawbacks an AI opponent would have.

It's also interesting to note that we're talking like Blizzard hasn't created at least some sort of AI for their RTS's, and haven't been doing that for years. People complain or laugh about it, but try loading up warcraft 2, loading up the map No Way Out of This Maze with 7 AI opponents, and see what happens.

In Warcraft 2 at least, the relative difficulty of the AI grew exponentially with more and more of them in the game, sort of like the Geth. I'm not sure how it worked exactly, but I do know that with 1 or 2 PC opponents, occasionally they'd get stupid and start building stupid shit, but with 6 or 7.. they were all pro (compared to me and my friends at least)

Another thing I found interesting about the AI with Warcraft 2 is how the gamespeed effected it. It was common practice in my circle to turn the gamespeed up to high.. but if you were playing against humans (the race) AI, they were almost unkillable.. their micro was just so crazy with the paladin heals that it was almost impossible to do anything to them.

If you slowed the game way down, they got retarded.. it wasn't just that they healed slower, they healed less often, even in emergency situations.

Just providing a bit more discussion food. eat at your own peril.

EDIT: Just thought of it, but what seperates humans from AI, I think, in this sense, is Macro. Although AI sometimes cheat and can see things that they player can't to make up for their lack of macro skill, I think that's where the AI fundamentally falters. unable to look at the game as a whole and make informed decisions.. It just knows build orders and priorities.

Altorin:

De Ronneman:

True, but in this thread Deep Blue is being compared to AI that is supossed to be thinking it's way out of situations far deeper than a chessgame. A direct line is drawn between Deep Blue and "learning" AI. Those are not the same.
Deep Blue dates from 1996/1997. That's not comparable to something as complex as an Artificial Neuro Network.

I wasn't really disagreeing with you. Might have sounded that way but it wasn't. Deep Blue, objectively, was a piece of shit by today's standards.. I was just saying that it was AI, in the sense that most people that aren't writing science fiction describe AI.

A program like Deep Blue would be hard pressed to play Starcraft.

However, I'm pretty sure I saw a video of a Blind kid playing Starcraft, and although I'm not really comparing his mind to a computer, his METHOD of playing would be quite similar to how a computer might play.

Of course, the blind guy loses, but it's only because he shows some of the same drawbacks an AI opponent would have.

It's also interesting to note that we're talking like Blizzard hasn't created at least some sort of AI for their RTS's, and haven't been doing that for years. People complain or laugh about it, but try loading up warcraft 2, loading up the map No Way Out of This Maze with 7 AI opponents, and see what happens.

In Warcraft 2 at least, the relative difficulty of the AI grew exponentially with more and more of them in the game, sort of like the Geth. I'm not sure how it worked exactly, but I do know that with 1 or 2 PC opponents, occasionally they'd get stupid and start building stupid shit, but with 6 or 7.. they were all pro (compared to me and my friends at least)

Another thing I found interesting about the AI with Warcraft 2 is how the gamespeed effected it. It was common practice in my circle to turn the gamespeed up to high.. but if you were playing against humans (the race) AI, they were almost unkillable.. their micro was just so crazy with the paladin heals that it was almost impossible to do anything to them.

If you slowed the game way down, they got retarded.. it wasn't just that they healed slower, they healed less often, even in emergency situations.

Just providing a bit more discussion food. eat at your own peril.

EDIT: Just thought of it, but what seperates humans from AI, I think, in this sense, is Macro. Although AI sometimes cheat and can see things that they player can't to make up for their lack of macro skill, I think that's where the AI fundamentally falters. unable to look at the game as a whole and make informed decisions.. It just knows build orders and priorities.

I'll bite, I'll bite:

First of, I don't speak Korean. How did the blind kid navigate and such? How could he see wth was going on? Is he God? I'm really psyched about this now!

On a more discusable note:
I agree on your griefs. A computer could never oversee somethign the same way a human does. If I can see my opponent, I have an advantage. I can see his face, I see his sweat dripping, I can see a grin of laughter.
All of those things give me one thing that is vital for everything, games, sports, job, school:

Information

A computer could never gain the same information a human does. It would only "see" a limited amount of things: The number of enemies, how they move and what surounds them. A human can also think of an enemy holding back or sending in big forces on purpose.

As to your trouble with game spedd and numbers: It's relative.

Because the game is sped up, so is the AI. The AI thinks the same things, just in a speed you, as a human, could never achieve. They might SEEM more stupid, but all they have on you is reactionspeed, and the speed the game provides. You have less time to eract, therefore in a losing position.

The sheer number of opponents is also relative. When they're with 2, they make mistakes that matter. They have noone to fall back on. You can communicate to your friends, consult and talk to them. The AI are 2 seperate forces. They don't work together. That's why a mistake gets noticed: it's vital.
when there's 5 or 6 AI's, the number of mistake they make is the same. It's just unlikely that they all mess up at the same time. When 1 AI fails, the others have it's back. It seems they make less mistakes. That's not true. You just don't notice them, because you still have 4 or 5 functioning AI's to worry about.

Also, this is games. the AI is limited. Of course it's not even as complex as Deep Blue, for god sakes, that thing couldn't even fit in a closet!
Blizzard made AI, but they made limited AI, because they couldn't do it any other way. Doesn't mean it's dumb AI.

Some AI is just stupid, see Mercenaries 2.

Most people in this thread seem to confuse the term "AI" with "SciFi".
AI isn't like in the movies, where a robot functions like a human. That's only because he's played by an actor, and real AI would be boring to watch.

I've seen a show on AI a couple of years back. A man made a program in which he made small AI that could learn. They started of as thick dumbasses. They're boxes who roll around in a shoebox. There's one colour boxes for food, one for male, one for female and one for hazards. As the little critters walked about, they only knew how to mate and what food was. They learned the boundaries of the box, and died from hazards. They saw others die, and ate. As they mated, they gave small pieces of "knowledge" to their offspring. They, in turn, would see more, and give on to their kids. The man explaining stated that in about 50 years, the little critters would be on the same level as humans. The program is sped up and stuff, so evolution skyrockets.
If AI like that ws to be made, that's going to make humans redundant.

Also, the man stated, and I really liked that, use it in games:
"If you ask me, AI is alive if it can ask you not to turn it off."

How awesome would that be in Portal?

The Blind Guy can only play on a single map, and always plays the same race. He has everything set to timed button presses, and he's just very good at playing the exact same game almost everytime. BoxeR was only really able to beat the Blind Guy after his initial handicap (being blindfolded) was lifted (the rules were that he would wear the blindfold for about 3 minutes, then he could take it off.. reading the Youtube comments about that are HILARIOUS, calling him a cheater etc.)

Once BoxeR took off the blindfold, he was able to abuse the blind guy's deficiencies. Hide just out of sight of his units etc, play with units that were able to beat the blind guy's zerglings etc.

As for the Warcraft 2 speed thing.. I really don't know about that. I mean, it makes sense, and that's what I thought at first too, but really it seems to be a lot more then that. The AI actually gets dumber. It doesn't just get slower.

For instance, if I select a group of ogres, and set them to attack-move into the center of the enemy's base, and don't even touch the ogres, not microing them at all - if the speed is set to high, my ogres will be utterly destroyed without killing a single enemy (The Paladins will spamheal EVERYTHING and keep everything alive), if the speed is set to low, my ogres will decimate the entire enemy force without taking a single casualty (The Paladins MIGHT throw a single heal or two but nothing close to what they'd throw in high speed). This happened consistently. I really don't have a real explanation for it. This is also not using any input other then sending my ogres into the base.

The other thing, I pretty much gathered that that's what was happening, but it always seemed to be deeper then just "2 out of 5 of them didn't screw up so it's harder". It's hard to describe it better then I did before. I just know that when my friends and I would play with bots, 2 was generally too easy, but 3 would usually dominate us.

I played that game way too much and I have way too many anecdotes about it, lol

Still, it would be awsome if they'd made that AI and after a few years of shame from all of the human kind, someone started winning it. I bet his/her nikcname woud have "Connor" in it.

Calatar:

Your dismissal of eSports because a perfect AI is hypothetically possible comes across as silly. There are many things which are hypothetically possible, including a perfect batting machine, a perfect golfing machine, a perfect pitching machine, etc. Their hypothetical existence does not invalidate the entertainment value of the sports.

It's pretty funny how loose AI is being used in all these posts. AI is not a machine, mostly its a rule based engine currently, a true AI is still being developed because well, we don't understand our own brain well enough to want to create something that replicates it. When the rules can be simplified to the point to where only a handful of strategies are viable and its a grand game of rock,paper,scissors, I find that boring and hardly a spectacle worth calling yourself pro at.

Let me be clear: Lifting 500lbs, 225kg, or running 28mph or 45 kph are amazing feats. But that has NOTHING, absolutely NOTHING to do with game-play mechanics or AI. And everything can become a competition at any moment.

Anyway... apparently saying "when an AI can consistently outperform a human in a simplistic environment of strategy" makes the game not that amazing to me is getting taken out of context. It means I have something against the game, NOT sports or competition.

Definitely agree with De Ronneman: "Most people in this thread seem to confuse the term "AI" with "SciFi"."

Something that can be simplified so much an AI can do it better takes the human element away. In THIS day and age, RIGHT THIS SECOND, making an AI that can outperform a human at a game makes the game not very amazing or worthwhile on a competition level. To me and this opinion is mine alone and to a good handful you probably stinks, which is cool, don't let me rain on your parade.

Xersues:

It's pretty funny how loose AI is being used in all these posts. AI is not a machine, mostly its a rule based engine currently, a true AI is still being developed because well, we don't understand our own brain well enough to want to create something that replicates it. When the rules can be simplified to the point to where only a handful of strategies are viable and its a grand game of rock,paper,scissors, I find that boring and hardly a spectacle worth calling yourself pro at.

Let me be clear: Lifting 500lbs, 225kg, or running 28mph or 45 kph are amazing feats. But that has NOTHING, absolutely NOTHING to do with game-play mechanics or AI. And everything can become a competition at any moment.

Anyway... apparently saying "when an AI can consistently outperform a human in a simplistic environment of strategy" makes the game not that amazing to me is getting taken out of context. It means I have something against the game, NOT sports or competition.

Definitely agree with De Ronneman: "Most people in this thread seem to confuse the term "AI" with "SciFi"."

Something that can be simplified so much an AI can do it better takes the human element away. In THIS day and age, RIGHT THIS SECOND, making an AI that can outperform a human at a game makes the game not very amazing or worthwhile on a competition level. To me and this opinion is mine alone and to a good handful you probably stinks, which is cool, don't let me rain on your parade.

My parade shall not be rained upon then. We have no such Starcraft AI that can outperform humans. The strategy of Starcraft is not limited to Rock-Paper-Scissors complexity. Your claim which irked me is that when an AI can outperform a human at a strategy game, the game cannot be a sport.

Chess is not a simplistic environment of strategy, yet an AI can beat over 99% of human competitors. So is grandmaster-level chess no longer amazing to you?

For an AI to outperform a human at a game doesn't necessitate that the game be simple. It could necessitate that the AI be incredibly complex and well-developed, or that a computational element is capable of vastly exceeding human reaction time.

I still don't see how it fails to be an amazing demonstration of skill and talent merely because a computer is capable of parallel actions while humans must perform serially. Which is the main element of contention with your opinion here.

'Starcraft 2/Boxing'

Anyone?

Calatar:
Chess is not a simplistic environment of strategy, yet an AI can beat over 99% of human competitors. So is grandmaster-level chess no longer amazing to you?

For an AI to outperform a human at a game doesn't necessitate that the game be simple. It could necessitate that the AI be incredibly complex and well-developed, or that a computational element is capable of vastly exceeding human reaction time.

The funny part of all this is that while a computer can beat a human so much of the time these days with chess, the GM level play is dominated not by pure calculating skill, but instead manipulation of high level psychological issues and alligning your play style with that of your own personality.

Computers completely disregard that. Hence comparing a GM vs GM game, to that of a computer vs GM game, is like apples and oranges.

One of the main problems I'd see with this is actually getting the damn thing to make a decision. Lets take a stroll down memory lane, back to late 1998 and Jurassic Park: Trespasser. As y'all know it was one of the great flops of gaming history - one of the most ambitous projects our hobby has seen and one of the most complete failures (But I still think that a remake on a modern engine would be awesome. The voice work, story & feature list all already exist. The first two are easily on a par with anything coming out these days I don't think any game has yet succeded in what Trespasser set out for all those years ago. Hell, the infamous arm would be the ideal thing to show off all this new fangled motion sensing technology. Trespasser Wii/Natal/Move anyone?).

Now, the dino AI scripting in Trespasser is (or at least was for the time), very complex - dinos had numerous different attributes that affected thier behaviour. Hunger, thirst, fear, agression, etc (And the back of my Trespasser box, even though it was from a "collector" reissue, still listed these features) but because of all these options the AI simply couldn't make a decision - it was stuck trying to prioritise which of its behaviors was the most important. Now, of course things have moved on, but even so an "AI" system with every strategy and every trigger may simply not be able to decide on a course of action when faced with constanly changing tactical environment.

randommaster:
You are right that you can "train"AI, but I was indeed thinking only about conventional desktops, so I didn't mention the crazy supercomputers that scientists are so fond of.

I'm a computer scientist, and I've trained an algorithm on my laptop. AI is remarkably powerful and flexible at many levels of computing power. Please refrain from making assumptions about computer science without doing some research.

A Starcraft "trained" AI would need thousands of input cases to begin to make a smart decision. When confronted with a scenario outside its training, it cannot be creative, imaginative, or inspired; it can only fumble and "guess" based on what it knows so far. While an AI could implement the Nuke Rush strategy, it could not intuit it itself without much further advanced understanding of the game rules, the opponent's records, and both conventional Protoss and Terran strategies and counter-strategies, as well as some kind of programmed "incentive" to experiment and try alternative approaches. Things like masking a nuke silo with a floating Engineering Bay to prevent the opponent identifying your strategy is something completely beyond the comprehension of conventional AI, without being taught to do it first, and even then it likely has little to no comprehension of the value of doing so without a kind of "meta-mind" for storing assumptions about the opponent's thought patterns.

AI is an incredibly complex field. Building a Starcraft AI to compete at professional level is only just within the grasp of plausibility; remember that the rules of chess lead to a game of open information on an 8x8 board. Starcraft has many more pieces, non-symmetry, closed information, and much more besides. It took decades of funding and computing power to make a chess AI that could convincingly challenge a pro, and we still haven't "solved" chess as a game. What hope is there for a convincing Starcraft AI? Human players will always be better, until true Artifical General Intelligence exists and decides to learn to play old video games. Which will be a while.

BobisOnlyBob:

randommaster:
You are right that you can "train"AI, but I was indeed thinking only about conventional desktops, so I didn't mention the crazy supercomputers that scientists are so fond of.

I'm a computer scientist, and I've trained an algorithm on my laptop. AI is remarkably powerful and flexible at many levels of computing power. Please refrain from making assumptions about computer science without doing some research.

A Starcraft "trained" AI would need thousands of input cases to begin to make a smart decision. When confronted with a scenario outside its training, it cannot be creative, imaginative, or inspired; it can only fumble and "guess" based on what it knows so far. While an AI could implement the Nuke Rush strategy, it could not intuit it itself without much further advanced understanding of the game rules, the opponent's records, and both conventional Protoss and Terran strategies and counter-strategies, as well as some kind of programmed "incentive" to experiment and try alternative approaches. Things like masking a nuke silo with a floating Engineering Bay to prevent the opponent identifying your strategy is something completely beyond the comprehension of conventional AI, without being taught to do it first, and even then it likely has little to no comprehension of the value of doing so without a kind of "meta-mind" for storing assumptions about the opponent's thought patterns.

AI is an incredibly complex field. Building a Starcraft AI to compete at professional level is only just within the grasp of plausibility; remember that the rules of chess lead to a game of open information on an 8x8 board. Starcraft has many more pieces, non-symmetry, closed information, and much more besides. It took decades of funding and computing power to make a chess AI that could convincingly challenge a pro, and we still haven't "solved" chess as a game. What hope is there for a convincing Starcraft AI? Human players will always be better, until true Artifical General Intelligence exists and decides to learn to play old video games. Which will be a while.

I meant to agree with you, but it seems like I didn't do that very well.

I have done some research about Deep Blue and how complex a game of chess is even though you have complete information. A game of Starcraft would be much more complex and hidden information makes things harder.

The whole thing about supercomputers was meant to be a joke, but apparently I failed. I know a bit about computers myself, but apparently I can't mix it with humor too well.

I ask that a new forum award/shame be made for this occasion:

"Asshole On Accident - Get a newsletter response from an Escapist staff member on a knee-jerk opinion and continue to argue your point, but fail to see how it matters anyway, because all you said was pure opinion. You defiantly got everyone riled up, but did not get yourself banned. You saucy bitch."

I clicked through to the comments frothing from the delusional, high-handed article. I see Xersues has already had most of the salient points raised for him and is engaged in some serious ivory towering. So I'll just ask a question.

Mr. Deam, were you being intentionally obtuse and provocative when you suggested that human players would be "relegated to the minor leagues" if sufficiently advanced AI were created? You were right in pointing out that humans are more fun to follow (as a fan) because they have identities and personalities and styles, and perhaps emotions play a part here, but you forgot a fundamental maxim of voyeuristic human existence: the enjoyment found in appreciating human endeavour.

Creating AIs and pitting them against one another is definitely an impressive form of endeavour, but it's quite different from a person gaining competence at an activity and competing in a ranked environment. And tournament organizers and spectators alike know whence cometh the awe inspired by these gamers: they are normal humans and have achieved incredible excellence through practice, perseverance or some spark of ingenuity not attributable to grunt computational bandwidth.

No amount of very advanced AI will change this fact, just as it didn't for chess. There will always be human only leagues, and I suspect they will remain the most popular. As long as normal humans are the audience, practiced normal humans will be the only competitors taken seriously in mainstream competitive play. If you truly think otherwise then I submit that you are experiencing a very curious disconnect with the reality of things.

(Disclosure: I'm a hardcore CS nerd and will watch more AI-related material than most fans. I'll still watch more human-only games than AI, since, for now, that's the sole source of innovation and creative strategy.)

Emlyn:
In addition people don't seem to realize just how difficult it is to program good AI that doesn't cheat. The human player has the advantage of being able to adapt and more importantly do new things. No AI is capable of doing something its not coded to do.

This. Truth. A thousand times. The AI is only as good as the programmer writing it. AI cannot adapt and survive (at least in its current state as a VI - Virtual Intelligence, which is what computer games actually use, as opposed to Artificial Intelligence, which is something from science fiction that can learn on its own).

Human brains can adapt to situations and draw upon years and years of learned knowledge and instinct. We have computers more than capable of flying fighter jets, yet we still have a human in ultimate control. Why? Because that cliche bullshit they feed us at the cinemas (Stealth, I'm looking at you!) is actually true.

Without the human element, VI cannot exist, and to a certain extent, it is true for AI. This will not change for a very long time.

Altorin:

Nimbus:
@Xersues

A software bot can beat people at Starcraft easily, so therefore, it's not a sport? Bullshit. Even in a human body, a robot mind could easily beat and human mind at any traditional sport. Computers do shit better than humans do. Deal with it.

that's not exactly true.

Human brains are still more complex then any computer system we've ever created (even the new Nvidia chip with its 3 billion transistors pales in comparison).

What Deam's saying is that computers have the edge MECHANICALLY in a game like starcraft.

When your brain thinks of something, it sends an impulse to your fingers, and then your fingers act, and then something happens on the computer, its sent to a screen, the screen broadcasts it to your eyes, your brain reinterprets the data it's fed from the screen, your brain responds and sends another impulse to your fingers.

The computer AI merely has to be aware that something is happening and it can react instantaneously. It's sort of like if your bloodsugar spikes, your body will send out more insulin to bring it under control. Think of bloodsugar as a zerg rush, and insulin as a squad of marines and that's basically how a computer AI of the Deep Blue variety would act. With no real thought, just reactions.

In that case, your interface with a computer IS slower then the computer's interface with itself.

But your brain still beats the computer almost every single time.

You made me think of an ai which does insanely fast zerg rushes. Could be a nice and hard thing to beat. Coz ai can have a lot of units fast. Faster then any human

BobisOnlyBob:

randommaster:
You are right that you can "train"AI, but I was indeed thinking only about conventional desktops, so I didn't mention the crazy supercomputers that scientists are so fond of.

I'm a computer scientist, and I've trained an algorithm on my laptop. AI is remarkably powerful and flexible at many levels of computing power. Please refrain from making assumptions about computer science without doing some research.

A Starcraft "trained" AI would need thousands of input cases to begin to make a smart decision. When confronted with a scenario outside its training, it cannot be creative, imaginative, or inspired; it can only fumble and "guess" based on what it knows so far. While an AI could implement the Nuke Rush strategy, it could not intuit it itself without much further advanced understanding of the game rules, the opponent's records, and both conventional Protoss and Terran strategies and counter-strategies, as well as some kind of programmed "incentive" to experiment and try alternative approaches. Things like masking a nuke silo with a floating Engineering Bay to prevent the opponent identifying your strategy is something completely beyond the comprehension of conventional AI, without being taught to do it first, and even then it likely has little to no comprehension of the value of doing so without a kind of "meta-mind" for storing assumptions about the opponent's thought patterns.

AI is an incredibly complex field. Building a Starcraft AI to compete at professional level is only just within the grasp of plausibility; remember that the rules of chess lead to a game of open information on an 8x8 board. Starcraft has many more pieces, non-symmetry, closed information, and much more besides. It took decades of funding and computing power to make a chess AI that could convincingly challenge a pro, and we still haven't "solved" chess as a game. What hope is there for a convincing Starcraft AI? Human players will always be better, until true Artifical General Intelligence exists and decides to learn to play old video games. Which will be a while.

I agree.
And while I'm not a computer scientist, programming is a major part of my job. This post is the first one that seems to have any real understanding of programming. The AI required to play starcraft strategically against a professional player is miles outside our abilities at the moment.

However as mentioned elsewhere, the programs micro abilities would give it a major advantage to compensate it's stratgic weakness. It could organize formation and pull back wounded units to maximize survival, focus fire precisily so no shot is wasted and never be taken by suprise ( tacticly, not stratgically ), it would never leave a drone unassigned or miss-time an upgrade, it might not be smart, but it will be fast. I belive this advantage could prove so powerfull that the programs lack of stratgic depth will not show. But even this is not nearly as trivial as you would think. The problems of formation, movment, path-finding and spreading units with regards to each other are amazingly complex even without taking into account the battle those units are in and the fact that they have to react to enemy units.

How complex are these issues? Go to any modren RTS, create a tight group of mixed units, randomly pick half the units as group 1, other half as group 2, Tell group 1 to go left of the mixed group location and group 2 to the right. All hell then breaks loose as the units collide, mismanuver and do all kinds of stupid shit. Pathfinding is hard, pathfinding in a dynamic terrain is very hard, pathfinding in a dynamic terrain with other movable obstacles that some are trying to get pass you and other are trying to get to the same spot you are is INSANE. We think that a program will be able to turn every unit into a being as smart as we are, solving the problem of movement fulidly and smoothly as human crowd walking across a cross-walk. Sorry guys, we're not even close.

Epoetker:
-snip-

there's no call for being so rude to anyone here.

Emlyn:
In addition people don't seem to realize just how difficult it is to program good AI that doesn't cheat. The human player has the advantage of being able to adapt and more importantly do new things. No AI is capable of doing something its not coded to do.

Agreed, it annoys the hell out of me that AI's have to be given obvious advantages but on the other hand I can understand why they need it. Just do this for example: Remove the fog of war. Now you are a little more on par with a computer and guess what, that alone is normally enough to kick it's little cheating arse all around a map.

Epoetker:
/snip

Goddess, you had to bring up Othello didn't you? When I feel really self loathing I boot up a Othello program just to have my arse handed to me again and again by the computer. Managed to beat it on simple once, ONCE, and many times on harder settings it takes the whole damn board many a time. There just isn't a way to beat a half decent AI in Othello, period.

Epoetker:
RAAAAAAAGE

The difference is, neither Othello nor Chess are in real-time, where you need to coordinate perfect optimal production whilst micromanaging an army of units on the field, using their abilities correctly, focus-firing, moving in and out of danger, etc.

The mark of a truly lightning-fingered StarCraft 1 player is that he can drop a Siege Tank out of a Dropship, have it fire at a Sunken Colony, and then pick it up in the split second before the Sunken Colony's attack connects. To be able to do that is to be ultra-hardcore.

A computer could do that without any effort. That's where Jordan is saying that computers could outpace human SC players; not in strategy but execution.

And regardless, calling contributors - let alone staff - "dickheads" and insulting them so rudely isn't something we tolerate here. You're going to take a little break and learn how to be more mature about expressing criticism :) Good-bye.

I hope you enjoyed your stay at the Escapist, because there's no call for being so rude to anyone here.

Alty, rudeness is a state of mind. Rudeness is failure to consider the feelings and experiences of others. Rudeness is assuming that you know everything, and that points against that thought need to be ignored, not addressed. Rudeness is making snide comments about how you were really just baiting everyone when they call you out on your bullshit. Rudeness is lamely retreating to 'this is just my opinion' when you were actually trying to put out a position.

Rudeness is failure to apologize.

Epoetker:
Ragesnip

I don't think anyone has any reason to apologize. Starcraft comes with AI. There's an entire single player campaign dedicated to it.

Hell, did you fail to realize that almost ALL the game has some AI pathing to it? You DON'T get full control of the behavior of units, they follow some sort of A* or pathing algorithm to flock to where you tell them to.

I happen to be a computer scientist, and I happen to read a lot about AI. Like I said, AI is mostly a rule based engine for now (which isn't real AI), if it knows what you're building, why couldn't it just build a counter, especially a prescripted one, which isn't TRUE AI, but who cares. You rolled rock, so its going to make paper. How annoying would it be that you've been bested not by real AI, but by a rule engine, the same thing that manages your packages at UPS/Fed ex? :)

The game is not that easy however, as we all know, but it is possible to rank all possible units as threats, and have the counters listed for each, and make a rule engine that can systematically and perfectly destroy what you're bringing up. Just like John said:

John Funk:

The mark of a truly lightning-fingered StarCraft 1 player is that he can drop a Siege Tank out of a Dropship, have it fire at a Sunken Colony, and then pick it up in the split second before the Sunken Colony's attack connects. To be able to do that is to be ultra-hardcore.

A computer could do that without any effort. That's where Jordan is saying that computers could outpace human SC players; not in strategy but execution.

That can be added to a rule engine. Zerg unit A has a priority of 10 (highest) and it's counter is Protoss unit B. The rule engine finds out you're making A, and starts making B, effectively (by whatever tested means on a per map basis that might be) destroying your units because its doing 500+ actions per minute.

It would be controlling its base/building, unit production, and fighting at the same time. That's freaking HARD to do. The little bastard AI can rebuild while you're fighting and trying to micro the battle.

This is all theoretical. Blizzard made an AI, in a game that's partially controlled by AI. I don't see why it can't be attempted. It might not always be perfect, but it might have a strat to roll out on some maps that would just destroy.

I apologize to everyone that takes this entire thing too seriously. Because it's your blood pressure, not mine. Less reading opinions and more playing starcraft, right?

John Funk:
And regardless, calling contributors - let alone staff - "dickheads" and insulting them so rudely isn't something we tolerate here. You're going to take a little break and learn how to be more mature about expressing criticism :) Good-bye.

I was literally gobsmacked when I read what they said Funk. Was all that to Deam? Seriously? I thought it was to Xerseus, and it was bad enough at that, but calling a staffer (especially an editor) a dickhead?

*laughs and shakes his head*

I think I've said my piece in this thread.. it was interesting discussing my experiences with Warcraft 2's AI.. it's not a game I get to talk about much anymore.

asiepshtain:
I agree.
And while I'm not a computer scientist, programming is a major part of my job. This post is the first one that seems to have any real understanding of programming. The AI required to play starcraft strategically against a professional player is miles outside our abilities at the moment.

I'm just a struggling undergrad myself, so you likely have a much greater understanding of the actual applications of code and their limitations. But thanks all the same.

asiepshtain:
Pathfinding is hard, pathfinding in a dynamic terrain is very hard, pathfinding in a dynamic terrain with other movable obstacles that some are trying to get pass you and other are trying to get to the same spot you are is INSANE. We think that a program will be able to turn every unit into a being as smart as we are, solving the problem of movement fulidly and smoothly as human crowd walking across a cross-walk. Sorry guys, we're not even close.

Pathfinding is my personal obsession within AI. The simple elegance of A* compared to the horrendous hacks you have to apply to it to get it to work dynamically fascinate me.

Xersues:
I happen to be a computer scientist, and I happen to read a lot about AI. Like I said, AI is mostly a rule based engine for now (which isn't real AI), if it knows what you're building, why couldn't it just build a counter, especially a prescripted one, which isn't TRUE AI, but who cares. You rolled rock, so its going to make paper. How annoying would it be that you've been bested not by real AI, but by a rule engine, the same thing that manages your packages at UPS/Fed ex? :)

The game is not that easy however, as we all know, but it is possible to rank all possible units as threats, and have the counters listed for each, and make a rule engine that can systematically and perfectly destroy what you're bringing up. Just like John said:

John Funk:

The mark of a truly lightning-fingered StarCraft 1 player is that he can drop a Siege Tank out of a Dropship, have it fire at a Sunken Colony, and then pick it up in the split second before the Sunken Colony's attack connects. To be able to do that is to be ultra-hardcore.

A computer could do that without any effort. That's where Jordan is saying that computers could outpace human SC players; not in strategy but execution.

That can be added to a rule engine. Zerg unit A has a priority of 10 (highest) and it's counter is Protoss unit B. The rule engine finds out you're making A, and starts making B, effectively (by whatever tested means on a per map basis that might be) destroying your units because its doing 500+ actions per minute.

It would be controlling its base/building, unit production, and fighting at the same time. That's freaking HARD to do. The little bastard AI can rebuild while you're fighting and trying to micro the battle.

This is all theoretical. Blizzard made an AI, in a game that's partially controlled by AI. I don't see why it can't be attempted. It might not always be perfect, but it might have a strat to roll out on some maps that would just destroy.

I mostly agree with you that it's plausible, but the majority of the existing AI is very much "conventional". It's not designed to use "cheap" tricks like dropping and reclaiming a unit so it can fire without being fired on; as much as that's a valid strategy, it doesn't make any sense within the game lore. A tank that can be picked up and dropped before an enemy can react, from a dropship that's in mid-air the entire time? It's absurd, and only emerged because of vicious competition and one-upmanship between humans. I think even with sufficient data and resources, a trained AI would never reach the point of requiring such a strategy unless demonstrated it first, and even then its execution may be flawless but the chances to use it would be so narrow as to never be attempted by a logical AI. It's so risky no probabilistic engine would ever try it!

What I'm getting at here, is that human players take queues from outside the game. They know their opponents, they know how much the High Templar lightning freaks out Zerg players, they know their opponents won't think to click past a floating engineering bay and might not notice a scanner sweep going off. To factor all these external behaviours into an AI is a task of such insane complexity, above and beyond merely getting pathfinding right on a map with fog of war - and even that's still not quite right. There's a reason micro exists, and it's because the AI doesn't.

Sport: an active diversion requiring physical exertion and competition

playing SC competitively sure sounds like a sport to me ^-^ if you argue otherwise than please justify why Golf is a sport. Machines can do them both way better and consistently as well. other than competition neither sport require much physical exertion :P

edit:: mind u im not talking about the skill required to do either well :P

 Pages PREV 1 2

Reply to Thread

Log in or Register to Comment
Have an account? Login below:
With Facebook:Login With Facebook
or
Username:  
Password:  
  
Not registered? To sign up for an account with The Escapist:
Register With Facebook
Register With Facebook
or
Register for a free account here