| Forum Index > General Forum |
Chess discussion continues here | ||
|
TheRealArtemis
687 Posts
On November 17 2013 08:40 plasmidghost wrote: Just watched through game 6 and I have to say, wow, Carlsen's endgame play is simply stunning to watch. I wonder if the fatigue or nerves is starting to get to Anand. I sure hope not. He looked pretty annoyed during the interview after the match, snarling at reporters  | ||
|
Cel.erity
United States4890 Posts
On November 17 2013 09:43 Sprouter wrote: People who values how often a human player makes the "most optimal" moves don't understand the game at all. Huh? I'm not sure what you're trying to say here. It's always beautiful to me to see someone play a game near-perfectly using nothing but their own judgment. It's very difficult to reach the point where you can make every small decision with great accuracy; so difficult that computers still get it wrong sometimes. | ||
|
peidongyang
Canada2084 Posts
On November 17 2013 05:54 kusto wrote: We're talking about constructing an algorithm for finding a globally optimal decision for EVERY timestamp in the game for it to be ideal. In chess, these are less than 100 in an average game with long breaks in between. In a real-time game (SC or any RTS), you need to perform all these computations at each of these timestamps - i guess that's hardly possible. EDIT: Sure, you can build an AI which executes perfect micro (not the strategic aspect that i'm interested in), therefore beats any human player. ... if that's what you wanted to tell me. actually im pretty sure in theory you could probably have a set of robust formulations which could optimally solve every strategical position in edit sc2, although you'd probably never get the formulation, geometry, a finite reduction or the desired runtime. too bad the OR department at MIT dont run a starcraft 2 research facility, or so I think... | ||
|
Orome
Switzerland11984 Posts
On November 17 2013 06:15 nosliw wrote: What happen if white plays Rook A1 defending the queening square instead of going back to A8? Black simply plays Re6 -> Re1 and he either queens or (in the case of Re6 Kg2 Re1) wins the rook. | ||
Plexa
Aotearoa39261 Posts
On November 17 2013 10:36 Cel.erity wrote: Huh? I'm not sure what you're trying to say here. It's always beautiful to me to see someone play a game near-perfectly using nothing but their own judgment. It's very difficult to reach the point where you can make every small decision with great accuracy; so difficult that computers still get it wrong sometimes. Well, Kf4 in the last game was discarded by computers because it leads to a drawn position with accurate play. But it was a critical move for winning the game for Carlsen. That's the kind of move that reminds you that computer-theoretic chess isn't the same as competition chess (as if that wasn't obvious already) and shows how games can be won through playing far from optimal moves. | ||
|
Cel.erity
United States4890 Posts
On November 17 2013 12:45 Plexa wrote: Well, Kf4 in the last game was discarded by computers because it leads to a drawn position with accurate play. But it was a critical move for winning the game for Carlsen. That's the kind of move that reminds you that computer-theoretic chess isn't the same as competition chess (as if that wasn't obvious already) and shows how games can be won through playing far from optimal moves. All lines were drawn with accurate play, so making the most practical move is optimal in that situation. Regardless, I made no arguments about when it's correct to make a practical choice over a slightly more optimal one, I only said that it's beautiful when players find the best moves despite having no prior knowledge of the position. | ||
|
broz0rs
United States2294 Posts
| ||
|
Hapahauli
United States9305 Posts
On November 17 2013 13:43 broz0rs wrote: Is the Berlin the most popular choice for Ruy Lopez games? It's been trendy in GM play of late, but historically, the most popular choices are the various Closed Ruy systems for black. | ||
|
The_Red_Viper
19533 Posts
On November 17 2013 08:18 Grumbels wrote: If you look at the performance of the strongest human chess player versus the strongest computer chess player then they are still somewhat close with the human player having chances every other game. On the other hand, a minimally competent hypothetical Starcraft 2 AI would be way beyond the capabilities of any human player. It would be more along the lines of matching up Ursain Bolt with a Ferrari and see which one will win the race. This is based on the strength of a computer to execute micro and macro perfectly which is beyond the level of human players. All you need to do is find some sort of safe build that lets you get to mid-game at which point you can send some medivac drops or whatever around and immediately win. kusto being mystified at people finding enjoyment in chess when computers can find the best moves is actually much like being mystified at people enjoying starcraft because while spectating a game you can see the mistakes of the players. It's an ignorant perspective that only a non-chess player could have. For actual chess players it's irrelevant that you can check the best move with a computer much like how it's irrelevant that you can check the replay later on to see what you did wrong while playing starcraft. (of course there are some cons to computers and I'm personally not a devoted fan) And well, bringing up the concept of game tree search because technically starcraft 2 is a zero-sum two person turn based game is so beyond irrelevant that I don't even know where to start. Nobody was talking about "perfect play based on theoretical optimal outcomes based on game tree search" because it's totally not relevant when building an AI that can beat human players. I think it is about to beat a top human player EVERY game. Ofc you can give the AI a buildorder and it will execute it better than any human, but what happens if the human knows the buildorder and plays around it? (for example he plays for a basetrade) I mean we had the examples of the BW AI's that are buildt by some of the best CS schools and they still are not able to beat humans, i think it is a little but close minded to say one buildorder (even perfectly executed) is enough to beat a human who has much more options to "learn on the fly". | ||
|
sc4k
United Kingdom5454 Posts
lol that was awesome. | ||
|
kusto
Russian Federation823 Posts
On November 17 2013 08:18 Grumbels wrote: If you look at the performance of the strongest human chess player versus the strongest computer chess player then they are still somewhat close with the human player having chances every other game. On the other hand, a minimally competent hypothetical Starcraft 2 AI would be way beyond the capabilities of any human player. It would be more along the lines of matching up Ursain Bolt with a Ferrari and see which one will win the race. Alright. Nothing relevant has been said in this passage. This is based on the strength of a computer to execute micro and macro perfectly which is beyond the level of human players. All you need to do is find some sort of safe build that lets you get to mid-game at which point you can send some medivac drops or whatever around and immediately win. Sure, i am talking about the strategic aspect of the game, not implementing 10k APM micro which is trivial and does not yield scientific insight - any algorithm-based bunker rush or blink micro all-in would kill any human player immediately. I am talking about AI-decisions like "Do i scout a hidden base right now, or a millisecond later? Or do i not scout at all because i get more minerals from it? How do i evaluate prior data in each timestep efficiently for making certain strategic decisions?" I am not talking about executions in RTS, but about decision making (and tactical moves) obviously, since the comparison to chess only applies there, because there is no execution/micro in chess - only decision making. You are clearly talking about perfect marine splits or perfectly handled medivac-drops. Therein, i would agree that no human opponent would stand a chance. But this is a rather trivial topic that does not cover my concerns. kusto being mystified at people finding enjoyment in chess when computers can find the best moves is actually much like being mystified at people enjoying starcraft because while spectating a game you can see the mistakes of the players. It's an ignorant perspective that only a non-chess player could have. For actual chess players it's irrelevant that you can check the best move with a computer much like how it's irrelevant that you can check the replay later on to see what you did wrong while playing starcraft. (of course there are some cons to computers and I'm personally not a devoted fan) Bolded part is complöetely wrong. In chess, you can stop the game at any point and give winning/losing percentages calculated by a computer, which the in-game engine uses against a human opponent. This can't be done in RTS games. You can't stop the time and give these winning/losing probabilities/statistics for the outcome of the game. Why not? Because it's impossible to compute. For me, chess is demystified by the sole possibility of being able to algorithmically compute such statistics on-line (or any turn-based strategy game with complete information). This was my point. As long as nobody can give me these numbers for tennis, RTS or a racing game, the game remains mystified for me. Chess not anymore. This is a huge motivation factor for me learning and mastering a game. And well, bringing up the concept of game tree search because technically starcraft 2 is a zero-sum two person turn based game is so beyond irrelevant that I don't even know where to start. Nobody was talking about "perfect play based on theoretical optimal outcomes based on game tree search" because it's totally not relevant when building an AI that can beat human players. No need to look down at people using different definitions/notations than you. I don't come from an AI-engineering perspective and have a different look at things. If you are more knowledgeable in an area, you can explain instead of acting like a dick. Turns out you didn't understand what i wanted to say. Oh, and please stop wasting time by acting like a dick. | ||
|
siri
Portugal129 Posts
On November 17 2013 17:53 kusto wrote: bla bla bla your mindset is dumb and ignorant. I already told you to google automaton 2000 to see how easy AI would beat humans. Discussing AI that isnt allow to have 10k apm because humans cant do it, is like saying chess programs should not be allow to calculate thousands of moves per seconds because humans cant. You want to discuss strategic part of the AI implying that the strategic part of chess programs is what makes them stronger than humans and not the ability to calculate +1000/s moves. Your whole argument is flawed Stop derailing and dumbing down this thread. Go make your own one. And btw, when someone stops enjoying things because something/someone does it better, he/she has clearly something wrong. Go see a therapist | ||
|
Yurie
12148 Posts
On November 17 2013 18:52 siri wrote: I already told you to google automaton 2000 to see how easy AI would beat humans. I googled it, didn't find anywhere where it played a game against a human. Just a few good micro tricks. | ||
|
Elroi
Sweden5617 Posts
On November 17 2013 07:38 undyinglight wrote: I am rooting for Anand as he and Morphy are likely my two favorite players of all time (Unless you count Deep Blue as a player). Though he is not a GM Josh Waitzkin is a player I love as I learned so much from him in Chessmaster 4400 back in the day. Though I am still holding out hope for Anand to take a victory I have been very impressed by Carlsen's play going up two games over Anand. If Anand pulls it back this is going to be a mighty comeback! he's IM. I love him too though  | ||
|
urboss
Austria1223 Posts
Imagine that the computer would have to base its decisions solely on the path a stalker has taken over the last 60 seconds. A stalker can move into (at least) 9 different directions every millisecond. The engine would have to calculate its decision based on all the paths the stalker could have taken. The number of possible paths the stalker could have taken exceeds the number of possible positions of a chess game. Add blink to that and you would get close to an infinite number of paths for the stalker. I think the beauty of chess comes exactly from the fact that it is solvable by computers. There are absolutely no random elements in the game. | ||
|
siri
Portugal129 Posts
On November 17 2013 19:02 Yurie wrote: I googled it, didn't find anywhere where it played a game against a human. Just a few good micro tricks. So you are blind to the potential of this micro "tricks"? Cant you brain go any futher? Put this micro "tricks" in a protoss work rush and it would beat any human. Just like that | ||
|
hypercube
Hungary2735 Posts
On November 17 2013 19:05 urboss wrote: As mentioned before, the information you have during a StarCraft game is incomplete because of the fog of war. That makes StarCraft theoretically not solvable in the same way that chess would be solvable using a brute-force method. There are ways to deal with incomplete information. It's called Game Theory. Rock-Paper-Scissors has been 'solved', even though it has incomplete information. The best RSP player could not beat the computer in the long run. Imagine that the computer would have to base its decisions solely on the path a stalker has taken over the last 60 seconds. A stalker can move into (at least) 9 different directions every millisecond. The engine would have to calculate its decision based on all the paths the stalker could have taken. Yes, and a good program would likely deal with it the same way a human does: assume that most of these paths are fundamentally the same. So it would only need to look maybe 5-10 different paths (one from each class of 'essentially same' paths). Then come up with with a number of different responses and find the Nash Equilibrium between its responses and those 5-10 different paths. | ||
|
kusto
Russian Federation823 Posts
On November 17 2013 18:52 siri wrote: Discussing AI that isnt allow to have 10k apm because humans cant do it, is like saying chess programs should not be allow to calculate thousands of moves per seconds because humans cant. No, it's not. micro is not decision making. To compare chess and Starcraft, i only consider decision making. You want to discuss strategic part of the AI implying that the strategic part of chess programs is what makes them stronger than humans and not the ability to calculate +1000/s moves. The ability to calculate +1000 moves is correlated to strategic decision making obviously. Considerate as i am, i have omitted the rest of your garbage english. | ||
|
Arnstein
Norway3381 Posts
| ||
|
urboss
Austria1223 Posts
On November 17 2013 19:23 hypercube wrote: There are ways to deal with incomplete information. It's called Game Theory. Rock-Paper-Scissors has been 'solved', even though it has incomplete information. The best RSP player could not beat the computer in the long run. Yes, and a good program would likely deal with it the same way a human does: assume that most of these paths are fundamentally the same. So it would only need to look maybe 5-10 different paths (one from each class of 'essentially same' paths). Then come up with with a number of different responses and find the Nash Equilibrium between its responses and those 5-10 different paths. Yes, all of what you mention would enable computers to beat humans. But none of what you mention makes StarCraft solvable in the same way that chess is solvable. | ||
| ||
[ Submit Event ] |
|
|
|
WardiTV Weekly
The Patches Monday
Afreeca Starleague
Sharp vs Shinee
Action vs Shine
GSL
PiGosaur Cup
Replay Cast
Afreeca Starleague
BeSt vs Paralyze
Jaedong vs Speed
Kung Fu Cup
Replay Cast
The PondCast
[ Show More ] |
|
|
EPT
