|
On August 07 2017 21:30 aerlinss wrote:
The AI enters ur base with a scout.
From what it sees, the amout of drones, the amout of minerals mined on every single patches, it will exactly know which of all the possible every played openings are possible, and which are not. It will know if there is an scv/probe out on the map building a proxy, or not.
It will see your gas minded, ur units and buildings built and will instantly know what s the earliest possible time when a cloaked banshee or dark templar can arrive at his base.
It will detect certain openings and know exactly how many zerglings he has to build to stop this push, he will not build a single zerling too much.
You don t even need a really smart AI, all you need is enough data and a programm able to read this data fast enough.
The more data you have, the less smart the AI can be, it doesn t need to understand things, it just needs to copy things that worked out in the past, and since AI will have perfect micro and macro, this will be more than enough.
lol
if you ever tried playing lower leagues you would know that build orders of non-pros don't make sense. That's cause they make strategical and macro mistakes that can lead even a human misread the situation.
Same here. If you screw up with timings/BO you will essentially feed false data into AI to analyze and make incorrect analysis 
|
On August 06 2017 19:26 Simberto wrote:Show nested quote +On August 06 2017 18:58 sabas123 wrote:On July 13 2017 01:23 BronzeKnee wrote:It is going to be a long time before AI conquers the human mind. I don't think I will be alive to see it, and even the youngest among us probably won't. But many of you assume the AI will have an unfair advantage without actually realizing that it is an unfair advantage.On July 12 2017 19:34 Grumbels wrote:
While it seems very difficult to create an AI capable of outthinking humans in SC2, I don't think it will take them 10 years to beat humans. If you have perfect micro there must be an infinite number of one-trick build orders that are unstoppable by conventional means. Remember SC2 is played with a mouse, keyboard, speakers and a monitor. The AI should have to play with those too. Or the mind shouldn't have to.
I have no idea what your trying to imply, but the AI won't use anything physical, but will have access to a virtual version of those. And he implies that that is an advantage. The AI can basically control the game "with its mind", while for a human, there are a bunch of hoops to jump through. I actually hope that we will eventually get to control games "with our minds". UI design is basically just that, trying to make controlling games feels more natural and not require extra thoughts. For me this is often one of the problems with games, especially strategy games that require a lot of different inputs quickly. I know what needs to be done, but i just can't get that information from my mind to the units on screen quickly enough. Of course, more practice would make this easier and work better, but there is a big hurdle in between my mind and the game, and an AI does not have this problem.
In the grand scheme of all of this, the advantage gained will still be very negligible.
Besides, the ai still has to work with the same UI and pros are good enough that they don't have to think about an action in order to execute it.
|
This is so interesting. All this AI evolution really gets excited and somewhat scared. But as someone already said before, AI Libratus already beat top poker heads up players, and by quite a margin. And this is relevant because its a game with incomplete information, such as Starcraft. So adding the real time factor, micro, APM, and all of that, I just don't see how humans could beat this AI, when its ready to play. Regardless, I will be waiting eagerly to watch these matches.
|
On August 07 2017 18:02 Grumbels wrote:Show nested quote +On August 06 2017 17:39 Modesty00 wrote:
Ofcourse AI will be better in terms ot makro and micro. Mechanical task are easy for a machine , humans dont want to compete in that area and you dont have to. Once game reaches late middle game and position is equal the machine will take over easy.
The thing is can the machine outsmart and predict human in strategy, build order and early game plan. Can a machine make a massive drop while setting a trap on other side of the map? Can machine change game plan and readjust accoriding to new information that receives about opposition? Can hide his intention? Cancel building? Make a pylon in secret area? Those and more of this type of questino are importhant.
A.I did this many years ago with chess. Not only showed perfect tactical vision, traps, sacrifices, incredible defense and deep calculation predicting future, (the easy part) but was able to spot deep strategic ideas and plans. And i think this eventually will happen. The Brute force of CPU is already enough. We don't need faster CPU. The SC is limited in terms of possibilities. Chess and GO had infinity positions and still the machine won not by seeing everything because it can't (CPU not fast enough even all super computers put togheter). It will do it with starcraft too, but it might take longer, because is completely different type of game with inconpete information but defintely possible. It will be expontential growth. For example 3 month AI practice will be playing like crap. 1 year later still bad. 3 years later still making silly mistakes and playing worse than C- broodwar player. but 2 weeks more and is beyond super human level out of nowhere. Just like they expected Alpha Go to beat best GO player in 2025.. but it happened in 2016 huge surprise. So we just have to wait and will happen.
AI to be smart as bug - a lot of years, to be smart as dog - many many years, to reach chimpase level - super long. To go from chimpase to ultra smarter many times more than all humans brain put togheter - just a year. Technical Singularity is close, but that does not mean it will happen, but is getting closer.
Is like building a puzzle with 100 pieces. It's extremely hard to make sense which piece have to go with which piece at beginning or even at middle. So one by one, with test you link them togheter. At the end you have just 5 pieces left. So much work and so much time has passed. But still nothing - is incomplete. Then BUM.. last 5 pieces are easier than first 80 and from nothing to complete picture is way shorter time. I think that's a good point. If you look at the Encephalization quotient (roughly brain to body mass ratio) then humans outpace other great apes by only a factor of three. Assuming that the main difference between humans and others is in brain size, which translates to computing power, that means that if you are capable of simulating a chimpansee then simulating human intelligence is only a question of linking another supercomputer. This is obviously a rather simplified of looking at it, since brain differences between mammals are qualitative as well as quantitative, but it is useful. If you compare Starcraft players at the top level, then there are primarily quantitative differences. All players are capable of doing everything, but some are better at it. The leap from awful to godlike execution for an AI is fairly trivial, the more difficult question is whether you can bring an AI to care about scouting or micro or build orders etc. at all. Once it is capable of competing with a human on any level, then it's only a few more months of practice or another upgrade in hardware for it to vastly outpace any human. Show nested quote +On August 06 2017 19:26 Simberto wrote:On August 06 2017 18:58 sabas123 wrote:On July 13 2017 01:23 BronzeKnee wrote:It is going to be a long time before AI conquers the human mind. I don't think I will be alive to see it, and even the youngest among us probably won't. But many of you assume the AI will have an unfair advantage without actually realizing that it is an unfair advantage.On July 12 2017 19:34 Grumbels wrote:
While it seems very difficult to create an AI capable of outthinking humans in SC2, I don't think it will take them 10 years to beat humans. If you have perfect micro there must be an infinite number of one-trick build orders that are unstoppable by conventional means. Remember SC2 is played with a mouse, keyboard, speakers and a monitor. The AI should have to play with those too. Or the mind shouldn't have to.
I have no idea what your trying to imply, but the AI won't use anything physical, but will have access to a virtual version of those. And he implies that that is an advantage. The AI can basically control the game "with its mind", while for a human, there are a bunch of hoops to jump through. I actually hope that we will eventually get to control games "with our minds". UI design is basically just that, trying to make controlling games feels more natural and not require extra thoughts. For me this is often one of the problems with games, especially strategy games that require a lot of different inputs quickly. I know what needs to be done, but i just can't get that information from my mind to the units on screen quickly enough. Of course, more practice would make this easier and work better, but there is a big hurdle in between my mind and the game, and an AI does not have this problem. This seems fun for FPS games. Everyone always has perfect headshots because that's how they envisioned it mentally.
I envision this boiling down to "whoever's brain can process information the fastest wins" because you would never need to be more accurate or precise than anyone else, you would simply need to process faster than anyone else. I doubt that would be very fun, especially since the less overall brain power someone has, the larger percentage of brain power they do have would need to be spent on movement "controls" (walking, crouching, etc.) rather than processing headshots.
If the "mind game" was made in such a way that you had to mentally move the gun in the virtual space so that accuracy and precision were still necessary, then it could be more fun. But in that case I imagine a game where your mind is placed in a full virtual body and you control the soldier's actions with your mind, which is very different from current controller or keyboard and mouse FPS.
|
double post - dang edit/quote buttons
|
On August 07 2017 22:05 LDaVinci wrote:Show nested quote +On August 07 2017 21:30 aerlinss wrote:
The AI enters ur base with a scout.
From what it sees, the amout of drones, the amout of minerals mined on every single patches, it will exactly know which of all the possible every played openings are possible, and which are not. It will know if there is an scv/probe out on the map building a proxy, or not.
It will see your gas minded, ur units and buildings built and will instantly know what s the earliest possible time when a cloaked banshee or dark templar can arrive at his base.
It will detect certain openings and know exactly how many zerglings he has to build to stop this push, he will not build a single zerling too much.
You don t even need a really smart AI, all you need is enough data and a programm able to read this data fast enough.
The more data you have, the less smart the AI can be, it doesn t need to understand things, it just needs to copy things that worked out in the past, and since AI will have perfect micro and macro, this will be more than enough. Then that is not AI, and that is very far from what is developping with deepmind. And though this might be enough for chess, It was not enough for GO. I guess it wouldn't work with Starcraft also.
OP is talking about the future, so am i, i know that AI isn t like that right now.
The question is what AI is in the end. Is it a programm that you will just tell "learn sc2" and then it will automaticly do it, that would be REAL AI.
Or are we talking about some sort of superbot that combines perfect micro with an AI like decission making databes request out of megadatabases with preset informations about every single element in the game, because as far as i understand that is what the AI playing GO is doing.
If you break down the game in very little tiny parts and steps every decission, every action is just a small thinking process considering a small amout of factors.
The more your break this process down into little tiny parts, the better it gets. Instead of the normal KI just building a missle turret in TvT, the KI now starts a process comparing thousands of games with a similar opening into:
Whats the earliest possible time a missle turret was needed, you ll have to break down this one in further requests obviosly. Like search all games u have on this map for TvT for when was the earliest hit from a flying unit to a target withing range x from spawning position.
Then you will start to also take into consideration the scouting information, like oh no gas first, so remove all games from the database with gas first builds. Oh cc first, remove all games that are not cc first etc. etc.
That are requests a human has to categorize by hand. But the more requests like that one are added, the stronger the computer gets. And after that, the more data is available, the stronger it gets.
You can add endless amounts of little improvements, you can also add requets like "do i have to take a risk, or should i play safe because im ahead etc." into the request for every single action.
Combine that with a "perfect KI marine micro against banelings" like KI and it will be unbeatable after a few weeks or months of development, depending on the amount of efford / money someone is willing to put into it.
|
The only way to win when AlphaSC is playing will be to bet money on it.
|
lol
if you ever tried playing lower leagues you would know that build orders of non-pros don't make sense. That's cause they make strategical and macro mistakes that can lead even a human misread the situation.
Same here. If you screw up with timings/BO you will essentially feed false data into AI to analyze and make incorrect analysis
----------------
Oh come on. I can t descibe EVERY single step, but its obvios that you will give the data from pros way more value then the data from bronze leage players in your database......if KI plays against masterleauge, the database should only use data from hig master and gm for example.....thats easy to solve.
|
How on cloud is the AI supposed to be limited in terms of handling 'seeing' burrowed units?
The current AI is not allowed to see burrowed units, whereas humans currently can. I disagree with this, but is there an alternative?
There are many factors in the function of SCII that actually depend on the other competitor having imperfect knowledge compared to what he is actually "able" to have.
For example, in a typical scout, you may see some building going up, click on a geyser to determine remaining, etc.
But an AI?
Example AI scout goals:
Click on every mineral patch, both vespene geysers Deduce how much mining.
Count probes everywhere. If anything wrong, determine farthest that a missing probe could be on map.
Count pylons.
Check hp of warping in building to determine when it was started compared to normal.
etc, etc., etc.
The list could go on in many other areas, but basically this kind of information SHOULD make it strategically impossible to trick the AI. Just imagine the perfect information available to a zerg AI with sufficient pneumatized carapace overseers. And imagine how many buildings would get covered in slime and stop funcitoning whenever you looked away?
Therefore, ultimately the developers will have to decide how the AI is limited. I don't think APM is a good restriction. Maybe add some sort of "virtual mouse" physics? Like add a virtual mouse that mimics the cursor control of a real mouse/keyboard? To give a delay between desired item (info, action, etc.) and result, which is very much the human element of the game that keeps it fun.
Here's an example of where the deepmind will improve the current overmind:
Playing right now against hard AI, they don't avoid reaper grenades, ever. Playing against elite AI, they avoid grenades every time. Either way every time is just a stupid reaction. You need a mix. That kind of thing is what alphaSCII can handle.
|
On August 08 2017 05:19 sabas123 wrote:Show nested quote +On August 06 2017 19:26 Simberto wrote:On August 06 2017 18:58 sabas123 wrote:On July 13 2017 01:23 BronzeKnee wrote:It is going to be a long time before AI conquers the human mind. I don't think I will be alive to see it, and even the youngest among us probably won't. But many of you assume the AI will have an unfair advantage without actually realizing that it is an unfair advantage.On July 12 2017 19:34 Grumbels wrote:
While it seems very difficult to create an AI capable of outthinking humans in SC2, I don't think it will take them 10 years to beat humans. If you have perfect micro there must be an infinite number of one-trick build orders that are unstoppable by conventional means. Remember SC2 is played with a mouse, keyboard, speakers and a monitor. The AI should have to play with those too. Or the mind shouldn't have to.
I have no idea what your trying to imply, but the AI won't use anything physical, but will have access to a virtual version of those. And he implies that that is an advantage. The AI can basically control the game "with its mind", while for a human, there are a bunch of hoops to jump through. I actually hope that we will eventually get to control games "with our minds". UI design is basically just that, trying to make controlling games feels more natural and not require extra thoughts. For me this is often one of the problems with games, especially strategy games that require a lot of different inputs quickly. I know what needs to be done, but i just can't get that information from my mind to the units on screen quickly enough. Of course, more practice would make this easier and work better, but there is a big hurdle in between my mind and the game, and an AI does not have this problem. In the grand scheme of all of this, the advantage gained will still be very negligible. Besides, the ai still has to work with the same UI and pros are good enough that they don't have to think about an action in order to execute it.
But in terms of what the brain is capable of, it's actually not negligible. Being able to coordinate all your muscles and stuff to be able to play sc2 at 300+ apm is incredibly complicated. In terms of conscious and active decision making, yeah to an extent we can make sc2 mechanics muscle memory, but having to do this with eyes and hands still puts a lot of overhead.
There was one video of rat neurons directly connected to a computer, playing a flight simulator in the highest difficulty scenario, and it flew through without crashing. This was like 20k neurons I think. That's an absolutely tiny amount and they could do a task we humans take a lot of training to do. The one post talking about the AI being able to record everything about gamestate is what I think really is gonna put the AI above humans. It could essentially have almost perfect reads on most scenarios.
edit: and we do all have perfect headshots in a sense, our eyes can lock on to something really quickly :D
|
On August 11 2017 21:12 Jasper_Ty wrote:Show nested quote +On August 08 2017 05:19 sabas123 wrote:On August 06 2017 19:26 Simberto wrote:On August 06 2017 18:58 sabas123 wrote:On July 13 2017 01:23 BronzeKnee wrote:It is going to be a long time before AI conquers the human mind. I don't think I will be alive to see it, and even the youngest among us probably won't. But many of you assume the AI will have an unfair advantage without actually realizing that it is an unfair advantage.On July 12 2017 19:34 Grumbels wrote:
While it seems very difficult to create an AI capable of outthinking humans in SC2, I don't think it will take them 10 years to beat humans. If you have perfect micro there must be an infinite number of one-trick build orders that are unstoppable by conventional means. Remember SC2 is played with a mouse, keyboard, speakers and a monitor. The AI should have to play with those too. Or the mind shouldn't have to.
I have no idea what your trying to imply, but the AI won't use anything physical, but will have access to a virtual version of those. And he implies that that is an advantage. The AI can basically control the game "with its mind", while for a human, there are a bunch of hoops to jump through. I actually hope that we will eventually get to control games "with our minds". UI design is basically just that, trying to make controlling games feels more natural and not require extra thoughts. For me this is often one of the problems with games, especially strategy games that require a lot of different inputs quickly. I know what needs to be done, but i just can't get that information from my mind to the units on screen quickly enough. Of course, more practice would make this easier and work better, but there is a big hurdle in between my mind and the game, and an AI does not have this problem. In the grand scheme of all of this, the advantage gained will still be very negligible. Besides, the ai still has to work with the same UI and pros are good enough that they don't have to think about an action in order to execute it. But in terms of what the brain is capable of, it's actually not negligible. Being able to coordinate all your muscles and stuff to be able to play sc2 at 300+ apm is incredibly complicated. In terms of conscious and active decision making, yeah to an extent we can make sc2 mechanics muscle memory, but having to do this with eyes and hands still puts a lot of overhead. There was one video of rat neurons directly connected to a computer, playing a flight simulator in the highest difficulty scenario, and it flew through without crashing. This was like 20k neurons I think. That's an absolutely tiny amount and they could do a task we humans take a lot of training to do. The one post talking about the AI being able to record everything about gamestate is what I think really is gonna put the AI above humans. It could essentially have almost perfect reads on most scenarios.
An article about the rat neuron thing, because it is incredibly cool
This is amazing, how did i not know about this before? Everyone needs to know this.
Also, it opens up two very interesting questions:
A) What could you do if you actually link up a human brain to a computer,
B) What kind of amazing computer could we build out of rat neurons.
Both of them are amazing and cool and incredibly SciFi. We shouldn't try to bring the computer down to human levels artificially, we should hook the humans up directly, and then play whatever amazing game is challenging in that state.
|
On August 11 2017 21:20 Simberto wrote:Show nested quote +On August 11 2017 21:12 Jasper_Ty wrote:On August 08 2017 05:19 sabas123 wrote:On August 06 2017 19:26 Simberto wrote:On August 06 2017 18:58 sabas123 wrote:On July 13 2017 01:23 BronzeKnee wrote:It is going to be a long time before AI conquers the human mind. I don't think I will be alive to see it, and even the youngest among us probably won't. But many of you assume the AI will have an unfair advantage without actually realizing that it is an unfair advantage.On July 12 2017 19:34 Grumbels wrote:
While it seems very difficult to create an AI capable of outthinking humans in SC2, I don't think it will take them 10 years to beat humans. If you have perfect micro there must be an infinite number of one-trick build orders that are unstoppable by conventional means. Remember SC2 is played with a mouse, keyboard, speakers and a monitor. The AI should have to play with those too. Or the mind shouldn't have to.
I have no idea what your trying to imply, but the AI won't use anything physical, but will have access to a virtual version of those. And he implies that that is an advantage. The AI can basically control the game "with its mind", while for a human, there are a bunch of hoops to jump through. I actually hope that we will eventually get to control games "with our minds". UI design is basically just that, trying to make controlling games feels more natural and not require extra thoughts. For me this is often one of the problems with games, especially strategy games that require a lot of different inputs quickly. I know what needs to be done, but i just can't get that information from my mind to the units on screen quickly enough. Of course, more practice would make this easier and work better, but there is a big hurdle in between my mind and the game, and an AI does not have this problem. In the grand scheme of all of this, the advantage gained will still be very negligible. Besides, the ai still has to work with the same UI and pros are good enough that they don't have to think about an action in order to execute it. But in terms of what the brain is capable of, it's actually not negligible. Being able to coordinate all your muscles and stuff to be able to play sc2 at 300+ apm is incredibly complicated. In terms of conscious and active decision making, yeah to an extent we can make sc2 mechanics muscle memory, but having to do this with eyes and hands still puts a lot of overhead. There was one video of rat neurons directly connected to a computer, playing a flight simulator in the highest difficulty scenario, and it flew through without crashing. This was like 20k neurons I think. That's an absolutely tiny amount and they could do a task we humans take a lot of training to do. The one post talking about the AI being able to record everything about gamestate is what I think really is gonna put the AI above humans. It could essentially have almost perfect reads on most scenarios. An article about the rat neuron thing, because it is incredibly coolThis is amazing, how did i not know about this before? Everyone needs to know this. Also, it opens up two very interesting questions: A) What could you do if you actually link up a human brain to a computer, B) What kind of amazing computer could we build out of rat neurons. Both of them are amazing and cool and incredibly SciFi. We shouldn't try to bring the computer down to human levels artificially, we should hook the humans up directly, and then play whatever amazing game is challenging in that state.
I think you would be interested in this: https://www.theverge.com/2017/3/27/15077864/elon-musk-neuralink-brain-computer-interface-ai-cyborgs
|
On August 08 2017 20:57 aerlinss wrote:
Oh come on. I can t descibe EVERY single step, but its obvios that you will give the data from pros way more value then the data from bronze leage players in your database......if KI plays against masterleauge, the database should only use data from hig master and gm for example.....thats easy to solve.
exactly, it will learn on data from PROs, but in the game itself i would be interested to know how it will handle deviations from the norm based on human mistakes/fake actions.
machine learning works by the means of grouping events and making cross-links between them. AI uses this in decision making. When we observe pros playing sometimes we see they make decisions based on mind-games and not necessarily what would be the best strategy in this particular game. I wonder how AI is going to process that.
|
On August 12 2017 00:54 fLyiNgDroNe wrote:Show nested quote +On August 08 2017 20:57 aerlinss wrote:
Oh come on. I can t descibe EVERY single step, but its obvios that you will give the data from pros way more value then the data from bronze leage players in your database......if KI plays against masterleauge, the database should only use data from hig master and gm for example.....thats easy to solve. exactly, it will learn on data from PROs, but in the game itself i would be interested to know how it will handle deviations from the norm based on human mistakes/fake actions. machine learning works by the means of grouping events and making cross-links between them. AI uses this in decision making. When we observe pros playing sometimes we see they make decisions based on mind-games and not necessarily what would be the best strategy in this particular game. I wonder how AI is going to process that.
What you wrote indicates that you are familiar with the neural network pop-literature but do not understand how the system actually works. I'm not going to explain it to you either, but I will explain one of the first implementations with huge success.
Take 1,000,000 passport photos. Make a RGB map based on a finite pixellation spatial frequency, and then assign a ranking of some probability amplitude to select a ~35 state eigenbasis. Then, you can proceed decompose every one of 7 billion people's photographs into a 35-unit "vector" (a few bytes) plus a kilobyte-size analysis of the main features of the remaining error. With this, you can reproduce the photo from scratch. The deal is, if your neural network algorithm finds that 35 bases don't cut it by analysis of the error, it just adds the next most likely thing to the basis. It won't "adjust" the previous bases based on new (oin your case, mis-) information. Instead, this type of procedure is used to "learn".
Oh, I forgot. You don't need passport-photos. Just resolve the 3D Fourier-transform with 2 cameras to make holographic standardization in position between eyes, mouth, etc. You can do it live for facial recognition with 2 webcams and a dedicated ~4GB VRAM machine.
|
AI will never beat the true machine A.K.A. Innovation
|
|
|
|
|
|
EPT
