Alpha Zero | Stockfish CRUSHED by Google's neural network Alpha Zero
Alright, guys I am about to show you something that is. Absolutely. Crazy, and. Astonishingly. Beautiful. And. Weird. And it makes you think about the future and where we are going and. Chess. Will, never be the same. Might. Sound, like, a bit, grandiose, but. Wait till you see the game and wait till you hear the story so. This, is a game between. Alpha. 0, and. Stockfish, and you're probably familiar with the stock fish it is the strongest. It used to be the strongest, chess. Program, on the planet, battling. For that position with Komodo. And Houdini and it won the 2016. World Championship, and it's widely. Used as. The top analysis, tool for professional, chess players. Google. Has, a division they. Bought it called, deep mind and they work with something called neural, networks, and, the. Idea, is that it's, a computer, program that can. Learn. Basically. There's. A lot of stuff about. This available. Online just go check it out I will put a link to a. Scientific. Paper, that. Deepmind. Put, out on. On. Alpha, 0. That. Will be in the in, the description so you can read about how they, did this. Basically. Their. First project with deep mind, in. Gaming, was, to defeat. One. Of the best go players in the world go is a game. One. Of the few games that has even, higher complexity, than chess from. A mathematical point. Of view and it has been very very hard to crack and. Until. Very very recently the. Top human players were, able to beat the top. Programs. But. Then. Deepmind. Under. Google they. Programmed, something called alphago. And alphago. Was able to play against, itself and learn from its own mistakes it did that millions, if not billions, of times. Until. It was, strong. Enough to compete and actually beat, a, guy, called Lisa doll I believe, at pronunciation. Is completely off of course. And. Beat. Him convincingly, and now, they have taken that approach and, they, have generalized. It so, what that means is that they have changed. The way the neural network works, so, you can, basically put, in any, game, that has, the. Characteristics. That we're looking for for instance it's. A game that you should be able to win so that it's, you, can evaluate it it's not subjective, and, all of this so like chess. Checkers, go. Shogi. All of these kinds of games you. Can program, the rules in to. The. The, network, the network is called alpha zero and, and. It will then play itself until. It's. Well. Forever. Basically until you say please stop and then you can use. It to play, whatever. Game and they did that and they. They. Played. After. Four. Hours. Four. Hours. Of. Training. They. Played stockfish. They. Played a hundred games they. Won, three. Games with. Black out of 100, and with. White. Alpha. Zero. 125. Games. Against. Stockfish, and they. Lost zero. Zero. Games and. I'm. Going, to take a look in this video of one, of the games, because. This, thing, does not play. Like you're used to see an engine plane, at. All I could. Talk for hours about, the implications, of this this. Is to. Me very, beautiful, and also kind of scary, alright, so let's just. Take, a look at this so. It. Played, yeah, alpha sure has quite here and it played. D4. In a lot of the games it had as white and we. See the French, defense. So. It hasn't invented, something, completely. Original on, the first two moves but just just wait so. Alright. Okay. If you're not familiar with this we are battling for this and later, like. We'll try to undermine this deep, on this is the idea, of blacks play. And. We close the same time this is. Back. In the old days, how. You beat computers. Accepted. Wisdom was, that you made. A very closed position. And then. You maneuvered, very, very, slowly, to. Suffocate the computer, because you, would exploit, what's called The Horizon effect, that if the computer cannot. Search. And find anything concrete. It'll. Just get. Lost in the wilderness of, variations. And, that's how for instance Garry. Kasparov beat. Deep. Blue in 96, he would make all these very. Small moves and just very slowly, positionally. Grind the computer, down and, well. The. Formula, has been updated. Based, on well. What. Is now ancient, with wisdom. So. Alright we, it's. Very typical we reinforce, our Center with this move. And these, poems. Will, be here for a long time and. Their. Idea is to, have. A keep. Black. Box tin, that's. The whole idea, all. Right so, black does not want to be boxed in so he will undermine, the pawn this, is very, very, common.
Theory. And in, the French defense and. We protect the pawn with the knight the. Pawn is captured, and. Instead. Of immediately recapturing. We. Make this move, with the night because you would like to recapture with this. Knight. Right a. Little. Check is thrown, this. Is about. The time where you, know stuff. Gets interesting, to reblock the chick, with. A bishop like this and after Lewis Black says oh I was I, was, only joking I'll put. My bishop, back. Here on on c5. And I will protect this poem. And. Alpha0. Says nope. Nope. I will gain space. And. I, will push, you away. So. After. Black. Retreats all the way we. Can alpha. Zero can comfortably, recapture. This poem and this. Might. Look a little, weird. Because, now. We cannot use this poem if it had been here we could use. It to reinforce to play, c4. At some point but, that alpha, sure things. This is not necessary, and it likes the space, it. Has he really. Likes space, it really likes space, and it really understands. The. Advantage of having space. So. Develop. The pieces black develops a piece nothing. Revolutionary. About, that. So. Again, flag. Is trying to undermine the extra. Space and free his pieces and alpha, zero does, not want. Starfish, to have, place. To maneuver, at all so. The. Knight. Is kicked the. Knight is kicked and. We. Have this exchange and now we are back, to. This triangle of pawns in the. Center that. We. Also started, with even. Though at this, poem. Wasn't. It's not the original deboned, it's this the sepal. So. Start. Fish is basic, basically. Being cramped still. So. White has maintained, his or. Her, or its. Positional. Idea, the, strategy, it plays, with a. Strategy. That flows. Throughout, the entire game there's none of this you'll. Notice there's none of this. Random. Moves. Complete, computer. Bullshit, that you'll very. Often see where in. Like, in a closed, position a, computer, will just play, seemingly. Completely random moves because. It. Has no plan. Seemingly. Alpha. Zero has. A, plan. Right. So. The, knight. Is, developed and. We. Just alpha 0 just says you know I like, my space I want. To keep you cramped. Start. Fish goes. On the attack and, now. We put the. Bishop. Agonal. It is the good bishop it is not. Hindered. By its own pawns. And. We. See, just give me a second and start. Fish sis, alright thank you very much I will capture. Your Bishop with my night and I will have the two bishops and. Generally. We say that having two bishops, is an, advantage but, what do we also say we say it's an advantage in open. Positions. This, is not, an open position and, it. Was the bad bishop, of white that we got rid of had it you, know I would be able to have black been able to get rid of this bishop, the. Even evaluation, would be completely different. So. Stockfish clearly, does, not understand. The strategy, that, alpha. Zero is is using, now. We we, see a very. Nice. Move like this is just beautiful so, if I was playing this and I somehow, got this position against like some, human, I would. Catch this. Or, if I was super advanced, somehow I would do some and you were like capture. Here and maybe, try. To get my knight into this square or something like that. But. Alpha zero captures, with the King. Because. The. King only, really, needs to castle. If. There. Are issues with the King safety. But. Let's take a look at these pieces on, the back rank still no. Real. Ways. To get out back. Rank still these, poems, keep. It very very very, much. Under control these, pawns in the center not. Going anywhere fast. So probably, stay on the back rank for some time this. This. Bishop has one check here is. That, going to be a problem. Probably. Not, and. This. Rook also, not. Going. Anywhere soon. So. Stockfish. Tries to get out with the bishop. King. III. You.
Know Obviously. So. That that's how this alpha zero plays beautiful. So. The. King is active, it will be nice. For the end game if. You if. Alpha zero ever wants to trade into some endgame if the King will be close to the sensor it can defend, the pawns in the, words of Wilhelm Stein it's the king is an, active, piece use. It and that is what. Alpha. Zero is doing, here. Seems. A bit like a computer move to me this. One p6, I don't. Understand that completely. And. Now. It's time you know and. Now, alpha, zero has everything. Under control, put. Very very. Nice pawns in the center reticulum. With, the king of course so what do we do we expand, we gain more space. Like. This g4. And. Starfish. Wants to undermine. Of course and free. Itself. So it can use the beautiful, bishops, and. Now. At least. The. Rook is in, the game. This. Is how. Few. Moves. Stockfish. Has available, by the way stockfish, when i run this on my computer, has four cores like an i7, machine. The. New version of starfish starfish says zero point zero zero it says this is a draw and. It's. Comfortable, that it can hold this and. It. Moves, the, bishop. Back to protect here and so, that's, you. Just see how. Alpha. Zero is just. Keeping. Black. Completely. Boxed in there, is no, there is no way to get, out this. Rug is on an open fire but where can it go we are, covering all, of. This you know you could go, here and then what would you do it's. No use for, it here we even even. Have this one so. All. Right gain, more space we. Know that we know the plan we know how it plays now game on space. Kind. Of a random Queen move so, the. Queen it's, not completely random the Queen, wants. To come in here and maybe. Threaten. Some stuff but if you look at how. This. King is. It. Has made like it's it's castling, 2.0, this this is the next level of castle, it's, Castille in the middle of the board so. The, Queen can get in to the position, but everything. Is protected it has no. It's. Just one queen you know it you, need you. Need more backup and all the rest of the, pieces are still at home, alright. So. We. Came to space over here rook, goes over ready to infiltrate. So. Starfish. Does not want to keep this pawn, protected, forever but, as soon, as it plays this move. It. Also. Says, I'm not going. To have a pawn break any time soon because. With. This pawn back here we could more. We. Could dream of pushing. F6. At some point trying to do a pawn break. So. Starfish, made it even harder for itself, not. That I know any, I like. I can easily say all those mistake, by stockfish but I you know it's not that I can suggest anything, better. Roxy. - just trying to double up, alpha. Zero has all the time in the world and this. Move by the King does, absolutely nothing, and why does it by, the starfish and then do it well it because it has to make a move and now. No other moves gives it any. Sort of advantage, whatsoever so it just doesn't, move that does nothing and. We. Double, up so. Basically. This, game looks a bit like, a very very strong, positional. Player playing, against. Some. Kind of like a master, or something that was just. And. He was just or she was just being slowly. Grinded. Down. Except. This, is stockfish. The, silicon, monster, that makes no technical. Mistakes at all ever, apparently. Apparently. Not so. The. Tactical, mistakes are just. Extremely. Deep. So. Alpha zero. Doubles. Up the rooks, and, stock. Is just shuffling. The pieces around and, now. Alpha zero puts. Rook. On the, 7th so. A lot of the stuff that it does is. Old-school. Kind. Of human. Ideas. Double. Up the rooks put a rock on the seventh that's kind of nice, which wrote about this and every. Grandmaster. Knows about this and every club player, knows about this the thing is the thing is that alpha zero is doing this, against. Stock, fish. Stock. Fish wants to. Challenge. These almighty, rooks and. Get. Rid of some material, and. Alpha. Zero is very very happy to oblige because. I. Look. At these pieces all on, the back rank and. No. Future whatsoever. Disgusting. Infiltrate. With a rock in number two. There's. A little threat, so, we just go back and says now it's no problem. And. We, see these ridiculous. King moves buy stock fish. Now. We are propping a little bit over. Here, on the. King side and. This. Is about the time where the game goes, into the. Next tier, where. It just goes from being, beautiful positional. Chest just, being. Absolutely. Out. There, beautiful. And, scary stock. Fish still convinced. It's doing okay. It. Says I will threaten, the night here. We have a little surprise. How. About that for a move. How. About that, for. A. Okay. So let's just get, rid of the tactical, miss, here so we can like clean, up the position and see what this is about, capture. Like this. Recapture. With the Queen keeping. Our dangerous. Outside pawn, and now recapture. The bishop this is a bitch, this is a piece. Sacrifice. And. There's no. Immediate. Tactical. Justification. So, I'm not going to show you some amazing way, where, flag.
Is Going to lose the Queen in the next couple of moves at, all this is a long-term. Positional. Sacrifice. And it's under. The same, theme, that the entire game has been under. This. Crap. Bishop, this. Crap, rook and this. Crap. Queen and, also, scrap. Queen really, none. Of these are doing anything the, pawns are immobile. All, frozen. And. Now. There will never ever ever, be an f6, pawn break so. There's. Nothing, at all that like can, really do with the extra piece this. Is kind of a kind, of peace sacrifice, where like if you saw, Aronian. Played, you would go oh that's a very beautiful practical. Idea. Of. Course the computer will refute, it in a second but it will give him practical, chances or whatever but but no this. Is. Not. Going. To breathe with you did by starfish. Alright. Let's. Start. Some fireworks. We. Play. F5. Here, because. You. Cannot capture like. This. Because. This, check. Is going to pick up the, bishop, also. You cannot and if you want to capture like like, this the. Queen will be able to infiltrate so let's take, away that you capture, here you go here you have, a thread, of some. Checkmate here. Working, with the rook also, threatening this poem and there's very very long tactical. Line. Where. You can see that this, winds. You can also kind of feel it but. Stockfish. Agrees that this. Winds. I'm. Not going to show the entire line line is quite long to, justify, this completely but starfish agreed so it didn't capture the pawn. Instead. Of plate wrote g8 and. Now, we see. The Queen coming in slowly still. You. Cannot. Capture. Any of these pawns we're just trying to orchestrate, a. Defense. Like. It's a piece of butt, defending. And. Now we'll play f6. Ourselves, so this. Pawn is super, strong it's. Going. To be a queen maybe someday and. It's. There's, no way to undermine it this. Pawn chain is rock, rock solid right. I'm. Ridiculous. King move again, doesn't. Do, too, much and, now. We see. Mysterious. King move by alpha zero that I don't, really understand. Again. King. Move by. Black and. This. Move, also bit, mysterious maybe. Maybe. This, superhuman. Supercomputer. Computer, was. Even. Getting, a little bit lost in the variations, here or maybe this is just way too deep for. For. My. Pathetic. Little mind to understand, place. Just going back and forth with the, king and now now. We see the Queen coming in and. Of. Course we. Are hoping to get a queen traits and we do get a queen trait. And. Of. Course you know normally, the Queen trade should benefit. The, defending. Side because, now. There is you. Know we have less threats. To defend. Against and a threat of being checkmate checkmate it it's, not that high but there's. It's. Going to fall these. Two pawns will be able to push, and maybe break, three break, break free and you, still cannot use your extra piece and your work is still, crap. We. Make, sure that the rook can't, go anywhere, keep. It chained, excuse. Me keep, it change, to, the pawn and. I. Don't know if this is just pure desperation because. Of course we can we are easily able to. Catch this pawn and. This. Is about the time where. Stockfish. Really realizes, that it's.
Losing, Actually. And. Even. On my, machine and this was run on, optimal. Settings, so like 64 threats, on, some, really, badass hardware, supplied by Google. When. They played the games so it wasn't it wasn't playing like a weak, version, of starfish or anything like that but even on my measly four calls it you can actually see that it's losing now. And. We. Just captured, the poem and get ready to use the king in, the endgame -. And, the rook to. Break. Free over here. And. Now we have a passed pawn we have to actually, this, one and. This, one. And, we, just. Infiltrate. Just. Go over this kind of quickly because this. Is. This. Is. Kind. Of understandable. Our. Place and now. We. Threaten to. Queen. This. Move, is also is just typical, computer, move like I mean any, human. Player would probably, resign, here so. This. Point, is just captured, there's no tactical, justification. Or anything like that and this is the final position. Because. At. This point stockfish. Resigned, I don't know if computer. Can be depressed, or whatever but it. Resigned. And obviously. You. Know these pawns just, 100%. Completely unstoppable and there is no, sort. Of. Perpetual. Check or anything at, all the king is super active and. It's. Just it's. Completely over in this point so. Yeah. I mean. Wow. What. A game this, let. Me repeat the. Computer. Alpha. Zero it ran on some. Insane. Google, hardware. But. It only trained, for four hours. What. This is going to mean for. Artificial. Intelligence. I'm. Not an expert but. Hey. I. Think. I, think, the future is. It's. Going to be very soon, science. Fiction times coming very very soon so. I hope you enjoyed the video um I, do. Some, other videos, also that. Are, more like just in this video because I just needed to share this game with, you guys but I do some really, serious and. Also kind of fun chess videos also on different, channel called outrageous. I'll just, play. A little bit and. I'll. Put a link on the description to, like.
2017-12-16 21:07
