|
On March 14 2016 02:33 Erugua wrote:
I don't even see how a human could beat a descent AI since it can have 6000 apm and pretty, can manage 5 packs of armies a way human couldn't in different locations while macroing perfectly. Yeah that's probably very hard to make an AI that does that well, but if it exist one day, it'll have 100% win chance vs human, no doubt
For me the real question is " can a machine be powerful enough to realise that goal ", and the awnser is obviously yes on SCBW, and maybe not yet on SC2
How would it have 6000 apm given the mechanical restraints of a keyboard, monitor and a mouse?
Perhaps a better match would be for the only restraint to be a monitor, and allow the human mind to control the game. It would be such an easy win for humans when we don't have to rely on our fat fingers and can just think what need to happens.
So many of you underestimate the mind. No AI, at least in my lifetime, will ever be able to react as intelligently and as fast as the mind to changing environments. SC2 is not a turn based game.
|
United States7483 Posts
On March 15 2016 07:09 BronzeKnee wrote:Show nested quote +On March 14 2016 02:33 Erugua wrote:
I don't even see how a human could beat a descent AI since it can have 6000 apm and pretty, can manage 5 packs of armies a way human couldn't in different locations while macroing perfectly. Yeah that's probably very hard to make an AI that does that well, but if it exist one day, it'll have 100% win chance vs human, no doubt
For me the real question is " can a machine be powerful enough to realise that goal ", and the awnser is obviously yes on SCBW, and maybe not yet on SC2 How would it have 6000 apm given the mechanical restraints of a keyboard, monitor and a mouse? Perhaps a better match would be for the only restraint to be a monitor, and allow the human mind to control the game. It would be such an easy win for humans when we don't have to rely on our fat fingers and can just think what need to happens. So many of you underestimate the mind. No AI, at least in my lifetime, will ever be able to react as intelligently and as fast as the mind to changing environments. SC2 is not a turn based game.
You have it wrong. A computer can calculate much faster than any human can. The difference is that computers suck horribly at dealing with imperfect information, while a human can evaluate and respond more accurately and account for a wider array of potential options.
This is why computers don't lose at chess or go but they can't win at bridge.
|
In a game like SC2 it probably can win because the game is not so much about outthinking and outwitting the opponent as it is of hard countering it with precise numbers and using good micro with an AI can do easily.
SC1 on the other hand is a lot more tactical, a lot more skill based, it requires more thinking, it requires higher level of strategy, etc...
So it won't be too hard for an AI to master SC2, but SC1 on the other hand is a different beast.
User was warned for this post
|
On March 15 2016 09:13 Whitewing wrote:Show nested quote +On March 15 2016 07:09 BronzeKnee wrote:On March 14 2016 02:33 Erugua wrote:
I don't even see how a human could beat a descent AI since it can have 6000 apm and pretty, can manage 5 packs of armies a way human couldn't in different locations while macroing perfectly. Yeah that's probably very hard to make an AI that does that well, but if it exist one day, it'll have 100% win chance vs human, no doubt
For me the real question is " can a machine be powerful enough to realise that goal ", and the awnser is obviously yes on SCBW, and maybe not yet on SC2 How would it have 6000 apm given the mechanical restraints of a keyboard, monitor and a mouse? Perhaps a better match would be for the only restraint to be a monitor, and allow the human mind to control the game. It would be such an easy win for humans when we don't have to rely on our fat fingers and can just think what need to happens. So many of you underestimate the mind. No AI, at least in my lifetime, will ever be able to react as intelligently and as fast as the mind to changing environments. SC2 is not a turn based game. You have it wrong. A computer can calculate much faster than any human can. The difference is that computers suck horribly at dealing with imperfect information, while a human can evaluate and respond more accurately and account for a wider array of potential options. This is why computers don't lose at chess or go but they can't win at bridge.
more precisely, it is not the "CPU" or "computer" rather it is the mathematics of decision-making with partial information that is weak.
as this branch of mathematics grows and improves computers get better at dealing with imperfect
information.
http://www.wired.com/2015/05/humans-play-ai-texas-hold-em-now/
and as Churchill stated in my previous post some heuristics are required in games like Chess and Starcraft. "state spaces are often too large to search completely, so heuristics must be employed to evaluate non-terminal states". the first really great SC AI Bot will need some really smart SC players to craft some top notch heuristic functions.
basically, if a team of AI experts hires Boxer as their only expert player and they end up building a great AI that totally kicks ass.. it'll do so using a Boxer playstyle.
we are many years away from building a Starcraft AI-bot that requires no heuristic functions.
IOW : we are many years away from a team of AI specialists building a top notch AIBot with zero input from SC expert world class players.
|
On March 14 2016 19:59 Liquid`Bunny wrote:
Well of course the AI will be able to beat human starcraft players, regardless of it being apm capped or not, as long as they put enough effort into making it.
This is ill-informed as I've seen throughout TL.
This is like saying "of course a human can move a mountain with their hands, they just need to push hard enough" without taking into consideration anything else.
|
Russian Federation34 Posts
On March 15 2016 04:10 ZAiNs wrote:Show nested quote +On March 15 2016 03:29 TheZov wrote:
Oh come on guys, what are we talking about here? The AI of tomorrow is not the Deep Blue archetype that (I might point out) DID beat a professional chess player at his own game. Dynamic thinking machines that have unlimited APM and unlimited ability to process, calculate, analyze and mimic every play in the history of the game (both pro and amateur, including every ladder game ever played in the history of the game), on a micro-second basis, selecting situational solutions in real time with perfect execution and an unlimited potential for multi-prong execution... Does that sound like something anyone else can do? First of all, it's well agreed upon that if the AI exploits infinite APM it will win easily. What's in question is whether an AI will be able to beat a human any time soon, when the AI is subject to physical limitations similar to that of a human. Also your understanding of AI right now isn't very accurate. The AI won't have 'analyzed every play in the history of the game including every pro, amateur and ladder match', it couldn't for Go because there were simply too many game states, and StarCraft has several magnitudes more different game states. Brute force worked for chess but not for Go which needed a neural network AI to 'solve'. For AlphaGo, they fed the AI around 150,000 pro-games (there's no way they will be able to feed a StarCraft AI anywhere close to that, and even if they could I'm pretty sure StarCraft would required waaaay more games than Go to get it to the same place). AlphaGo then played against itself for months using a crazy amount of computing power (which amounts to a completely insane number of games). In StarCraft, every tiny change in any unit's positioning is a different game-state (or even the position of the mouse cursor), so it'll constantly be playing game-states that it's never played before. You also say the AI can perform this super complex analysis 'on a micro-second basis' which isn't the case at all. AlphaGo does not make decisions instantaneously, AlphaGo and Lee Sedol in their ongoing match are both given 2 hours of thinking time per game, and Go is a game with around 200 discrete moves, if you consider every action made in StarCraft there are probably tens of thousands of discrete moves from both sides. Making the AI able to make decisions in 1/60th of a second is going to be a huge challenge, and there is no way it can perform the same level of analysis in 1/60th of a second as it could with even 1 second.
Oh I thought we were talking about the REAL ai, singularity type shit. Of course it won't happen soon but it will happen
|
I will never, ever doubt the will of the technological world that would develop an AI to beat the best players in the world. Will it happen tomorrow? A year from now? It doesn't matter, because technology gets exponentially better that eventually an AI would be smarter, more adaptable than a player and with no emotions.
it could take forever, I don't know if we'll ever see a true thinking AI in our life times but never say never boys, the apocalypse is coming.
http://stream1.gifsoup.com/view7/3949890/sarah-connor-o.gif
|
Every game that has ever been "solved" (in the AI scientific sense of the word) by a computer has had a player that claims a computer won't be able to beat humans in competition. The fact is, as much as I respect Flash, he doesn't really know what he's talking about, so I'm not too impressed.
|
i am really interested in this topic as well, but i would side with the "humans would win" opinion 
Go is a game with "full information", while sc2 is not. AI can't handle this situation.
if you make it a bo5 with random maps (both players don't know the map beforehand) i doubt the current status of the AI could even beat a mid-class master player.
|
i stumbled across this and its really damn frickin' cool.
http://www.gdcvault.com/play/1021848/Building-a-Better-Centaur-AI
the Infinite Axis Utlility System is really frickin' cool and this same guy talked about that subsystem in great depth and detail at GDC 2013.
notice that at its deepest foundation the IAUS requires carefully defined heuristic functions. the response curves are all human created.
|
On March 15 2016 07:09 BronzeKnee wrote:Show nested quote +On March 14 2016 02:33 Erugua wrote:
I don't even see how a human could beat a descent AI since it can have 6000 apm and pretty, can manage 5 packs of armies a way human couldn't in different locations while macroing perfectly. Yeah that's probably very hard to make an AI that does that well, but if it exist one day, it'll have 100% win chance vs human, no doubt
For me the real question is " can a machine be powerful enough to realise that goal ", and the awnser is obviously yes on SCBW, and maybe not yet on SC2 How would it have 6000 apm given the mechanical restraints of a keyboard, monitor and a mouse? Perhaps a better match would be for the only restraint to be a monitor, and allow the human mind to control the game. It would be such an easy win for humans when we don't have to rely on our fat fingers and can just think what need to happens. So many of you underestimate the mind. No AI, at least in my lifetime, will ever be able to react as intelligently and as fast as the mind to changing environments. SC2 is not a turn based game.
Where "at least in my lifetime" is at best the most accurate thing in your post.
Otherwise there is nothing explained why an AI wouldn't beat a human unless there is some magical element in humans that can't be replicated by science.
|
On March 14 2016 23:21 BisuDagger wrote:
Until the AI plays a person, it won't learn the micro tricks that players contain: Stacked lurkers, glitching through mineral lines, observer on top of missile turret. Hell, even ally your opponent so their spider mines don't work could be used.
On the other side, what are the selection limitations of the computer? If a Terran floats barracks over a hatchery then the hatchery is un-selectable. If a barracks is on top of a cluster of tanks that sit on the high ground, then the tanks cannot be target by direct mouse clicks. Does the computer have to abide by the same rules in that sense?
Again not true.
How has humans learnt the micro trics? By some divine power or by testing?
What makes you think AI can't be self learning. Even if the technology isn't there yet for self learning in SC2 there are other games (like GO) where the AI has improved by playing millions and millions of games with itself finding the best moves.
I find it funny that people want to see themselves (humans) that is something completely different from AI. Like we possess some magical or divine power that can't be replicated or exist anywhere else than in or "soul".
Seriously, is it so hard to understand that humans are nothing but advanced biological machines and there is nothing from stopping us from creating AI that one day surprass us?
|
Denmark145 Posts
On March 14 2016 21:01 BeyondCtrL wrote:Show nested quote +On March 14 2016 20:51 Liquid`Bunny wrote:On March 14 2016 20:03 Liquid`Nazgul wrote:On March 14 2016 19:59 Liquid`Bunny wrote:Well of course the AI will be able to beat human starcraft players, regardless of it being apm capped or not, as long as they put enough effort into making it. However it would be boring if players didn't take on the challenge of beating it, i myself would love to experience playing against it, we might learn something! Also it's kind of funny how everyone is viewing the AI winning as humans "losing" I think it would be a great achievement for humanity to make an AI that can learn such a complex task As long as they create some laws restricting AI from taking over the world~~ When we create a program that can make a better program on it's own, that's when the trouble starts. AlphaGo does that already.
AlphaGo doesn't change the way it's programmed, it will always be programmed in a certain way. What it can change is parameters within functions to achieve a better result.
|
Czech Republic12128 Posts
On March 15 2016 12:25 necaremus wrote:i am really interested in this topic as well, but i would side with the "humans would win" opinion Go is a game with "full information", while sc2 is not. AI can't handle this situation. if you make it a bo5 with random maps (both players don't know the map beforehand) i doubt the current status of the AI could even beat a mid-class master player.
That's not necessary. This isn't solving by the traditional method(from A you can go to A1 - A56464984651894) but simulating the learning process of the brain. So the biggest obstacle is to transform all the information in the process to the computer so it understands it. Go is much easier for such translation(you have only X-Y, no ramps, not reachable terrain, blank spaces, bases etc.). But the learning itself works the same as our brain.
In SC2 we have multiple high end replay packs so the PC can learn from the best(not sure about BW).
If they do the job properly and PC can play SC without any problems... then the human person will have really tough enemy. Because then it's all about the time and the learning process. And PC can train 24/7
|
So what map are they gonna program it for? FS?
|
Bisutopia19152 Posts
On March 15 2016 16:16 papaz wrote:Show nested quote +On March 14 2016 23:21 BisuDagger wrote:
Until the AI plays a person, it won't learn the micro tricks that players contain: Stacked lurkers, glitching through mineral lines, observer on top of missile turret. Hell, even ally your opponent so their spider mines don't work could be used.
On the other side, what are the selection limitations of the computer? If a Terran floats barracks over a hatchery then the hatchery is un-selectable. If a barracks is on top of a cluster of tanks that sit on the high ground, then the tanks cannot be target by direct mouse clicks. Does the computer have to abide by the same rules in that sense? Again not true. How has humans learnt the micro trics? By some divine power or by testing? What makes you think AI can't be self learning. Even if the technology isn't there yet for self learning in SC2 there are other games (like GO) where the AI has improved by playing millions and millions of games with itself finding the best moves. I find it funny that people want to see themselves (humans) that is something completely different from AI. Like we possess some magical or divine power that can't be replicated or exist anywhere else than in or "soul". Seriously, is it so hard to understand that humans are nothing but advanced biological machines and there is nothing from stopping us from creating AI that one day surprass us?
I have a background in AI and spent my entire career in simulation so please don't attempt to lecture or patronize me lol. My post was about game rules. Is the AI given the capability to access the ally opponent screen so it can do the spider mine trick? My guess is no. It may however stumble on to hold lurker by seeing the H button is an extra contextual button available to it when a lurker is paired with another unit. On the other hand, Does the computer have the same limitations as a user in terms of mouse clicks to select buildings? This is a big unknown to me. It may perform a raycast and stop at the first object hit or sort through all objects hit by the raycast and if so then be able to select hatcheries covered by Terran buildings.
|
On March 15 2016 17:27 BisuDagger wrote:Show nested quote +On March 15 2016 16:16 papaz wrote:On March 14 2016 23:21 BisuDagger wrote:
Until the AI plays a person, it won't learn the micro tricks that players contain: Stacked lurkers, glitching through mineral lines, observer on top of missile turret. Hell, even ally your opponent so their spider mines don't work could be used.
On the other side, what are the selection limitations of the computer? If a Terran floats barracks over a hatchery then the hatchery is un-selectable. If a barracks is on top of a cluster of tanks that sit on the high ground, then the tanks cannot be target by direct mouse clicks. Does the computer have to abide by the same rules in that sense? Again not true. How has humans learnt the micro trics? By some divine power or by testing? What makes you think AI can't be self learning. Even if the technology isn't there yet for self learning in SC2 there are other games (like GO) where the AI has improved by playing millions and millions of games with itself finding the best moves. I find it funny that people want to see themselves (humans) that is something completely different from AI. Like we possess some magical or divine power that can't be replicated or exist anywhere else than in or "soul". Seriously, is it so hard to understand that humans are nothing but advanced biological machines and there is nothing from stopping us from creating AI that one day surprass us? I have a background in AI and spent my entire career in simulation so please don't attempt to lecture or patronize me lol. My post was about game rules. Is the AI given the capability to access the ally opponent screen so it can do the spider mine trick? My guess is no. It may however stumble on to hold lurker by seeing the H button is an extra contextual button available to it when a lurker is paired with another unit. On the other hand, Does the computer have the same limitations as a user in terms of mouse clicks to select buildings? This is a big unknown to me. It may perform a raycast and stop at the first object hit or sort through all objects hit by the raycast and if so then be able to select hatcheries covered by Terran buildings.
Why would the AI have to obey the same input rules as a human? To make it "fair"? AlphaGo doesn't have an entire set of mechanical limbs and locomotion just so it can drop game pieces on a board. Humans are taking billions of years of evolution for granted, if anything the AI is at a great disadvantage since it's playing games that humans designed for humans to play. You wouldn't make a human play BW by manually inputting onto a circuit board with a power supply and probes the electrical signals that represent in game triggers would you?
As to whether AI's will eventually beat humans at any specific task, in my opinion it's not about whether or not one system has inherent superiority over the other or not, but the amount of energy and resources required to devote to the task to compete. Everything in the universe function in accordance to some set of fundamental laws and rules regardless of whether they are comprehensible to us, if you devote enough energy, time, and ordered structures to a task it will be completed regardless. The real question is whether doing that achieves some conscious goal for some wealth or good. If we devoted all of the planet's most brilliant scientists and engineers, all of the ex-bw pro-players and coaches for strategic input, and all of humanity's manufacturing facilities and raw resources to build a machine just to play starcraft that used all the electricity and fuel sources available to us to power it then it would be a completely trivial task to beat one person. Unless you believe there is something fundamentally non deterministic that violates all computational mathematics and sciences inherent in a human being.
|
Bisutopia19152 Posts
On March 15 2016 18:07 Caihead wrote:Show nested quote +On March 15 2016 17:27 BisuDagger wrote:On March 15 2016 16:16 papaz wrote:On March 14 2016 23:21 BisuDagger wrote:
Until the AI plays a person, it won't learn the micro tricks that players contain: Stacked lurkers, glitching through mineral lines, observer on top of missile turret. Hell, even ally your opponent so their spider mines don't work could be used.
On the other side, what are the selection limitations of the computer? If a Terran floats barracks over a hatchery then the hatchery is un-selectable. If a barracks is on top of a cluster of tanks that sit on the high ground, then the tanks cannot be target by direct mouse clicks. Does the computer have to abide by the same rules in that sense? Again not true. How has humans learnt the micro trics? By some divine power or by testing? What makes you think AI can't be self learning. Even if the technology isn't there yet for self learning in SC2 there are other games (like GO) where the AI has improved by playing millions and millions of games with itself finding the best moves. I find it funny that people want to see themselves (humans) that is something completely different from AI. Like we possess some magical or divine power that can't be replicated or exist anywhere else than in or "soul". Seriously, is it so hard to understand that humans are nothing but advanced biological machines and there is nothing from stopping us from creating AI that one day surprass us? I have a background in AI and spent my entire career in simulation so please don't attempt to lecture or patronize me lol. My post was about game rules. Is the AI given the capability to access the ally opponent screen so it can do the spider mine trick? My guess is no. It may however stumble on to hold lurker by seeing the H button is an extra contextual button available to it when a lurker is paired with another unit. On the other hand, Does the computer have the same limitations as a user in terms of mouse clicks to select buildings? This is a big unknown to me. It may perform a raycast and stop at the first object hit or sort through all objects hit by the raycast and if so then be able to select hatcheries covered by Terran buildings. Why would the AI have to obey the same input rules as a human? To make it "fair"? AlphaGo doesn't have an entire set of mechanical limbs and locomotion just so it can drop game pieces on a board. Humans are taking billions of years of evolution for granted, if anything the AI is at a great disadvantage since it's playing games that humans designed for humans to play. You wouldn't make a human play BW by manually inputting onto a circuit board with a power supply and probes the electrical signals that represent in game triggers would you? As to whether AI's will eventually beat humans at any specific task, in my opinion it's not about whether or not one system has inherent superiority over the other or not, but the amount of energy and resources required to devote to the task to compete. Everything in the universe function in accordance to some set of fundamental laws and rules regardless of whether they are comprehensible to us, if you devote enough energy, time, and ordered structures to a task it will be completed regardless. The real question is whether doing that achieves some conscious goal for some wealth or good. If we devoted all of the planet's most brilliant scientists and engineers, all of the ex-bw pro-players and coaches for strategic input, and all of humanity's manufacturing facilities and raw resources to build a machine just to play starcraft that used all the electricity and fuel sources available to us to power it then it would be a completely trivial task to beat one person. Unless you believe there is something fundamentally non deterministic that violates all computational mathematics and sciences inherent in a human being.
This has nothing to do with mechanical input. I am talking completely on software level. A raycast is a software term. I guess people who don't program shouldn't be relying to me. Anyone who has a full understanding of the SC1 API can feel free to lecture me on its limitations compared to a human player.
|
Russian Federation93 Posts
They can do it in 4 different ways:
1. Computer can read SC2 process memory and send commands directly into the game process.
2. Computer can read SC2 process memory but can only send mouse and keyboard commands.
3. Computer has no access to SC2 process memory. It needs to observe the picture on the monitor to gather information and can only send mouse and keyboard commands.
4. Computer needs to observe the picture on the monitor to gather information and can only physically control mouse and keyboard.
AI will win in cases №1 and №2 and won't even consider playing in conditions №3 and 4 and that's why:
Starcraft wasn't designed to be played perfectly. Because it is impossible to do it by humans and that's why whole game design and balance only works with limitation of human motility and reaction.
Can you calculate which if your roaches siege tank will shoot right now by looking on his weapon position? Absolutely not. But even if you do, there is no way that you can properly react on this information by microing roaches properly to receive as minimum damage as possible.
We all saw these Automaton 2000 micro videos which proves that 5 siege tanks with medevacs controlled by AI can destroy a 200 limit zerg ground army without taking any damage. And this Automaton 2000 was created by one or several enthusiasts. It is nothing compared to what team from Google can create.
This game is not turn based. And it will be harder for AI. Correct. But it also so much harder for humans. Imagine if in GO you can play stones as fast as you can. Board would be filled with stones by computer even before Lee could touch his first stone. Turn based concept doesn't give any advantage to computers. It is much more favorable for humans and still computer wins 4-1 in the most difficult turn based game vs the best human possible.
In first two cases where AI can read SC2 process memory (limited by fog of war of cause) to gather information for processing it will beat any pro gamer just by simply outmicroing him in a way that game wasn't designed for. Micro in case №2 will be 1000 times slower than in №1 but it will still be 100000 faster than any human which is enough.
You could say that micro only just isn't working. That's why current Starcraft AI can't beat humans. And you are totally right. But current Starcraft AI is nothing compared to what google have created. It is like comparing a medusa to Stephen Hawking at who can do math better. Google AI will be 1000 times better than any human in both micro and macro.
Because this beast can learn.
And he can do it really fast. He can play 10 million games per day vs himself to understand how the game works and what move or tactic or strategy is better at any possible game position. He can analyse all pro replays in the internet to study how humans play and what is better to use against them. It might sound really hard to believe but that is how it works. He doesn't just calculate all possible moves and chose the best. Google AI thinks much more like human does than you think. And he does it better.
Cases 3 and 4 isn't about AI. It is about AI, but totally different AI. It is about image analysis and recognition. Which is totally not what this particular program is supposed to do.
|
On March 15 2016 19:24 sh1RoKen wrote:
They can do it in 4 different ways:
1. Computer can read SC2 process memory and send commands directly into the game process.
2. Computer can read SC2 process memory but can only send mouse and keyboard commands.
3. Computer has no access to SC2 process memory. It needs to observe the picture on the monitor to gather information and can only send mouse and keyboard commands.
4. Computer needs to observe the picture on the monitor to gather information and can only physically control mouse and keyboard.
AI will win in cases №1 and №2 and won't even consider playing in conditions №3 and 4 and that's why:
Starcraft wasn't designed to be played perfectly. Because it is impossible to do it by humans and that's why whole game design and balance only works with limitation of human motility and reaction.
Can you calculate which if your roaches siege tank will shoot right now by looking on his weapon position? Absolutely not. But even if you do, there is no way that you can properly react on this information by microing roaches properly to receive as minimum damage as possible.
We all saw these Automaton 2000 micro videos which proves that 5 siege tanks with medevacs controlled by AI can destroy a 200 limit zerg ground army without taking any damage. And this Automaton 2000 was created by one or several enthusiasts. It is nothing compared to what team from Google can create.
This game is not turn based. And it will be harder for AI. Correct. But it also so much harder for humans. Imagine if in GO you can play stones as fast as you can. Board would be filled with stones by computer even before Lee could touch his first stone. Turn based concept doesn't give any advantage to computers. It is much more favorable for humans and still computer wins 4-1 in the most difficult turn based game vs the best human possible.
In first two cases where AI can read SC2 process memory (limited by fog of war of cause) to gather information for processing it will beat any pro gamer just by simply outmicroing him in a way that game wasn't designed for. Micro in case №2 will be 1000 times slower than in №1 but it will still be 100000 faster than any human which is enough.
You could say that micro only just isn't working. That's why current Starcraft AI can't beat humans. And you are totally right. But current Starcraft AI is nothing compared to what google have created. It is like comparing a medusa to Stephen Hawking at who can do math better. Google AI will be 1000 times better than any human in both micro and macro.
Because this beast can learn.
And he can do it really fast. He can play 10 million games per day vs himself to understand how the game works and what move or tactic or strategy is better at any possible game position. He can analyse all pro replays in the internet to study how humans play and what is better to use against them. It might sound really hard to believe but that is how it works. He doesn't just calculate all possible moves and chose the best. Google AI thinks much more like human does than you think. And he does it better.
Cases 3 and 4 isn't about AI. It is about AI, but totally different AI. It is about image analysis and recognition. Which is totally not what this particular program is supposed to do.
Great post.
In any case, I'm not sure about what are Google goals concerning AI, but as an external observer cases 3 and 4 are more interesting to me. If everything else is equal (apm, visual recognition, physical constraints, etc)... can AI beat the best decision makers just by learning playing itself?
Btw, 10 million games per day vs himself seems a lot to me :D
|
|
|
|
|
|
EPT
