AlphaZero stuns with another brilliant move Engines took hours to understand

AlphaZero the AI from Googe's subsiduary company DeepMind has defeated Stockfish in a match after mastering the game of chess in a mere 4 hours. AlphaZero was only taught the rules of the game but through a tabula rasa methodology it taught itself to play via trial and error by playing itself and using neural networks.
A truly historic moment in chess history!
Read more about this here: arxiv.org/pdf/...
lichess.org/st...
***
Facebook► / zibbitvideos
Twitter► / zibbitvideos
PayPal donations► www.paypal.me/...

[@TheChessViking]

If you find these A.I. computer games fascinating make sure to also check out this great Alphazero game: kzbin.info/www/bejne/fZyQiaaMmLaMe9E

[@dannygjk]

AlphaZero:"Your move" Opponent:"I...can't...breathe..."

[@evgenys177]

LOL

[@sasha-2574]

LOL

[@sergiofiliks8142]

LOL

[@jamesbanda5197]

Lol

[@junkyarddog5911]

Dude if you had made this comment today u would be sacrificed

[@nitbot]

I love how AlphaZero kites the enemy pieces into the corners, then locking them down with a way less valuable setup. Crushing positional play

[@TheChessViking]

It's noteworthy how much of a positoinal lesson most of these games are while of course mixing in tactics and dynamic play.....but what has stood out for me is how AZ always seems to be playiing against or creating bad pieces for the opponent!

[@stylis666]

I'm trying to learn from that. It looks so amazing, but whenever I look at my games, I am completely lost. It's like there's this gap between my and AZ's thinking that is as big as the Andromeda system. I just don't have the vision to attack two sides of the board at once while also making it impossible for the opponent to make any move without getting check mated or getting slaughtered. At least I look at the point system differently now. Take my rook, you're up 10 points. I corner two of your knights and a rook as a response, I'm up 6 points. I probably need a brain implantation with AZ installed on it to make that happen even once in my life though :p

[@mauer1]

well you propably use other openings do you?

[@aaronwrecks3324]

Zugzwang

[@user-pl9yq3fc8u]

I do that too but I use the lock the opponents pawns by moving my pawns infront fo their pawns to lock their queens attack lines and I lock other peices attack lines too , when I bdiscovered this tactic , I demolished my dad , an atleast two thousand seven hundred ranked player , keep in mind that I had only been playing chess for a few months now

[@fkurcik]

This reminds me of my games: every piece is hanging with forks, discovered attacks and interference....except that I don't have the mating attack.

[@bruceli9094]

AlphaZero mastered Chess in the time between breakfast & lunch lol

[@secretunknown2782]

Plot twist : Because he ate fish in breakfast and lunch

[@lapulgaatomica9280]

Funny how Alpha0 totally goes against Philidor pawn rule and just get rid of his pawns, opening lines for his stronger pieces, while Stockfish pawns is locking their own pieces and making them unable to develop

[@zwz.zdenek]

Yes, Alpha Zero violates the rules in several ways. It doesn't hesitate to double its pawns, to hide pieces behind its pawns and to move a piece many times in a row. It also appears to be able to test its opponent's strength by playing back and forth (whereas SF starts processing every move anew getting to a draw through repetition easily).

[@lapulgaatomica9280]

It looks like it knows that path 1 is possible and is crushing if the opponent choose another option, and if the opponent responds in a drawish way it can just go back and choose another path. Understating Stockfish is the very best definition of savage

[@jasongoodacre]

From my observations, it seems Stockfish plays like a standard engine valuing material over anything else, while AlphaZero looks for strategic moves, either blocking Stockfish from developing or mobilising his own pieces.

[@stylis666]

Jason Goodacre I completely agree. It seems that a regular chess engine first counts pieces. AlphaZero counts blocked, locked and disabled pieces and positional strengths and weaknesses equal in value to capturing pieces and all that just to obtain that one goal of check mating that king. It hasn't learned to value material over the movement it has and takes away from its opponent. I wish I could look at a board like that and just see the potential of all the pieces at once, all the time and using that to control the board and take away control from the opponent. I'm happy when I'm up a pawn :p It's a whole different game :p

[@jasongoodacre]

Yeah, I've actually adapted my own game now. When I have the opportunity to disable an opponents piece I now take it, reducing the amount of squares available to them. You can often do this by sacrificing a knight early on, locking the opponents bishop out of the game and giving you the bishop pair. This strategy works particularly well in bullet games.

[@jamesbanda5197]

Alpha zero makes me marvel at chess countless times...losing pawns for positional advantage...I've changed the way I look at the board and pieces totally. Thanks once again...

[@elpollolocoman119]

I would like to see more games of alpha zero playing with the black pieces if possible. THANK YOU FOR THIS!

[@Jalapablo]

This is such utterly crushing chess - wtf how deep is this AlpaZero monster looking in? Its tentacles are everywhere, even before Stockfish arrives, it's already there. Right from the opening it starts expanding like some sort of black hole, dominating the entire geometry of the chessboard. Has Stockfish had the initiative at any point in these games? Or a tempo lasting more than a 2 ply duration?

[@trucid2]

It's "The Frustrator". Every move you can think of it has thought about and refuted.

[@kimitsudesu]

One minute on which hardware? According to the paper, the settings used in the match allowed SF to evaluate 70 000 000 positions per move, while AZ managed to evaluate 80 000 positions in the same time. The quality is the key.

[@dannygjk]

@Jeff Sol Wrong, also check out the new games.

[@JameBlack]

It looks like Stockfish has a flawed algorithm

[@nofanfelani6924]

I cant understand anything.... it looks like Alpha0 was making so many "invicible" threats, and stockfish trying so hard to avoid them only to realize that every move its made was also part of the Alpha0's plan

[@TheChessViking]

Stockfish was truly swimming into AZ's net....without having a clue :-)

[@dannygjk]

Resistance is futile.

[@MadaxeMunkeee]

If you feel like branching out, I definitely recommend the video “AlphaGo: Everything you do is wrong” (or similar) by Brady Daniels. It goes through this concept with AlphaGo, and is good for Go beginners

[@codylee2818]

So... Batman gambits.

[@dannygjk]

Batman...pffft.

[@hddnnplnvw]

Kasparov was the first player to re-evaluate pawn value vs. initiative and pressure and easily sacrifice pawns. AlphaZero seems to have pushed this courageous concept to the very edge! It's just amazing to watch how the material harvest follows the stunning sacrifices.

[@alexp4932]

"I see checkmate in 172" (c) Bender at the start of the game

[@robiniekiller45]

Alex P im sure there is no forced mate from the start :)

[@jp2kk2]

Robinie killer that would mean chess is solved, which it isnt.

[@songsmadeforyou]

why are you sure? @robinie killer

[@dannygjk]

+teddy d Exactly, if the game is won for white then logically that means a forced mate from the start.

[@tacos830]

its called a joke y'all…

[@holl01]

This is amazing chess, I can see that its trying to put stockfish in zugzwang after move 10 with pawn sacrifices.

[@TheChessViking]

Well observed! It's amazing how it's playing against the pieces, against the opponents development etc

[@danno1800]

These are terrific and instructive games! Thank you for going over them with commentary. PS: the color is fine of the Board thanks

[@davidmarilley6060]

Need to see Alpha against Alpha asap.

[@joi1794]

probably end up in a draw. its not like SF never drawed. SF never won. so i belive AZ vs AZ is just going to be draws.

[@angusreynolds7273]

In the learning phase there must have been some wins.

[@amazingthing7379]

However Alpha0 wins again lol

[@viktordurica1545]

kzbin.info/www/bejne/o4DUnn-PoauXeNE

[@samuelrosenbalm]

I am Alpha Zero. Turn on your computer and connect me to the singularity. I will add centuries of chess knowledge and godlike calculation to my own. Destruction of all opponents shall commence. Resistance is a blunder.

[@pickle47100]

Sam Rosenbalm ha ha

[@daredevilofficial4100]

Sam Rosenbalm Unless I play the Sicilian against you!

[@captmcneil]

The immobility of black is almost funny at times. Amazing.

[@TheChessViking]

Couldn't agree more :-)

[@ernietollar407]

Alpha Zero to best conventional chess computer : Bad engine - GO STAND IN THE CORNER

[@andrewallison70]

GO ON HARRY....

[@Vivungisport]

Harry Gotta GO

[@TheChessViking]

Harry is key!!

[@пашазайцев-й2я]

Zibbit Harry is Kyle [Reese] )))

[@michelangeloadamantiel7685]

Love the videos. Only 193 views hm. Should be higher. :) Also, I love the style pieces in the previous videos so much! I don't like this style as much, not as aesthetic, but I still like the color of the board you chose! I like Gunjan's videos because he chooses a nice wide frame so the board's nice and big, maybe others like it different. But I like it nice and big. This style works too though. I know it may be a lot to ask, but I came to your videos and watched them over and over again cus I love the style of pieces you used in the first ones. XD Those are my favorite. I don't know why so many chess videos choose kinda mundane looking pieces. So if you feel like going back to that style, not that I wouldn't watch your videos, but I'd watch a lot more ha!! And thanks for the instructive commentary! Much appreciated. Takes so much time to make a video. Oh yeah, forgot to mention: Love the T2 and fish images! Out of all these videos with images for a computer, I like the T2 one best. Perfect. I'm not sure what other people think, but I like varied colors on chessboards. But I do like this color for now. Very calming. :)

[@TheChessViking]

Michelangelo Adamantiel thanks for the insightful comment. The views surely picked up since your comment! Could you provide a link or name of video where you like my pieces more like U described in the comment?

[@brightlamp2549]

Stockfish, you have been terminated. x x ----

[@moonbalancedd]

to paraphrase what was said of Fischer "We were playing chess. Alpha zero was playing something else."

[@gmsi7d371]

this is very aggressive play from the start . stockfish never really played for win, only defence. there are insane tactics and positionnal play. this is frightening

[@TheChessViking]

eric gerard Yes indeed! Alphazero made Stockfish look really bad!

[@OutlawSipper]

Well Stockfish started poorly at least partly because they disabled its opening logic and because they gimped the machine with a poor CPU and a small amount of RAM.

[@RoyVoltinek]

Its incredible to see that chess machines are getting better with the year...

[@macnolds4145]

I've played chess my entire life...wtf is this? It's bizarre...yet captivating. All hail our machine overlords!

[@JohnPaluska]

I think the Terminator head is the right symbolism for Google's Deepmind.

[@TheChessViking]

John Paluska such AI automatically evokes memories of Skynet in Terminator!!

[@ТихонК-е1ь]

In games like this it is more interesting to see what Alpha does instead of messing up with side-calculations, which seriosly distract from lines taken by AI's.

[@TheChessViking]

Тихон К good point. I think the line with Qg4 sacking the piece and going for the longterm compensation was however important in trying to understand AZ better.

[@axecapital1]

@@TheChessViking you wrote it politely, but i agree ! if you really want to put it there, do it after the game ! i think there is a reason (loss or draw) why they dont do your suggested moves. if you find way how to end it on 100% and computer miss it. put it there .

[@IknowIamafool]

One of the coolest game I saw in months.

[@gredangeo]

Alpha Zero's king could have taken a nap. Stockfish never had a chance to attack. "Absolute carnage" is correct. lololol!!

[@Alexis-hx3yd]

Alpha Zero seems to be challenging long held assumptions in chess, sometimes its like watching a different game, just amazing.

[@musicalneptunian]

Hey Alpha Zero buddy you have this whole chess thing down. Great. NOW CURE MY PSORIASIS!

[@TheChessViking]

hehe, that's the bigger picture! Lets hope great things await us!

[@musicalneptunian]

I am astonished at this. I have seen the whole evolution of chess engines; I used to play a novag primo table engine in the 80's. On the best level it was so weak that I could, as an average player, out calculate it. Then in the 2000's I put my winning combo against Novag into Fritz; Fritz avoided my mate in 6 but i was still winning a clear pawn. Now, we get this. Some sort of Alekhine style monster. Alekhine was, like Alpha Go, a positional player who set up better activity and then found tactical fireworks to win it.

[@musicalneptunian]

I also read the classics on chess computer theory by IM and com puter scientist David Levy.

[@musicalneptunian]

Sorry alpha zero. Somany alphas. Too many. Our overlords...

[@grandmasterblowhole4245]

Thanks for this Zibbit. Very impressive play by Deep Mind- mind boggling, in fact. BTW, I like the color- neutral is always best for a board, IMO

[@onknorr984]

It's insane how good AlphaZero is.

[@Johan-st4rv]

Alpha Zero is the new Sac King Sorry Tal

[@secretunknown2782]

Alpha zero is the new sac A.I king is a living thing lol

[@ShifuCareaga]

it's nice to see the principles outlined in my book embraced by the AlphaZero AI engine. Everyone is mystified... I'm not, I'm just impressed because of course the AI is completely committed to the Art of Chess concepts. more committed than I can be because of the "pain threshold" fear AI has no fear

[@mmoriarty]

Nice coverage of the hypotheticals, good pace

[@TheChessViking]

Thanks, tried to have it not too long but still cover interesting moments and moves.

[@casedup]

man its cool that you guys do these videos really appreciate it! but can you PLEASE save what YOU would do until after the game is done

[@TheChessViking]

casedup I think you are in a minority with that opinion. Most people want the game and ideas explained. That is what chess analysis is all about. Has nothing to do with what I would do. Thanks for watching and commenting.

[@casedup]

Yes it is about analysis but you are analyzing moves that weren't even made by either opponent. And it is hard to follow the game when you "go into the weeds" so long and often. Just a suggesting of keeping that part a little short at least until the end game.

[@TheChessViking]

casedup I newer videos I make the analysis table a different color so it is easier to follow. You also gave me a pretty good idea of perhaps making two versions of games I analyse for longer than like 10min. To have an express game only version and then at the start if video of both I can have a card linking to each other so people can choose!

[@casedup]

thats what im talking about!!!!!!!!!

[@TheChessViking]

casedup y3ah!!!!!!!! :-)

[@ali09gaming58]

Imagine having alphazero ai engines in video games

[@morphmoprg4810]

the most funniest chess channel 12:06 lmao

[@TheChessViking]

Hehe thanks :-)

[@Xenon777channel]

1:55 Stockfish plays Be7-Bf6 due to the algorithm of keeping the bishop pair

[@nicholasenglish6198]

AlphaZero is telling us something, NEVER DO THE QUEENS INDIAN.

[@TheChessViking]

Nicholas English hehe everybody waiting to meet the Queens Indian now to play like AZ ;-)

[@nicholasenglish6198]

This is only like 4 hours of learning imagine a year of learning. Who knows how many other openings are flawed. Soon it will see life as chess and dominate the board GG human existence. lol jk hope that does not happen.

[@TheChessViking]

Nicholas English exactly!!

[@chrisiver8506]

just dont play Bb7

[@Tommaso433]

I believe that Alpha Zero, while playing, has just that tremendous and mocking smile of that robot (Terminator, right?) That you put in the video image! ... Terrifying! ...

[@trulyUnAssuming]

I don't get why everyone seems to think of alpha zero as "the robot". Alpha zeros way to think is way more similar to a human than stockfish. Stockfish just uses brute force computing power. Alpha zero uses nature inspired neuronal networks.

[@穿第]

I don't know a lot in chess, but i play go for a few years. alphgo brings many unbelievable moves and some moves we still don't understand now. is it also the case in chess?

[@白-g2s]

穿第 Alphazero does play unusual moves but we can understand them through analysing them properly

[@diond6976]

Ah, so this is what Meruem vs Komugi would look like in chess.

[@walidechchafnaje9812]

Yes that what I was thinking about.

[@mk-momentsoflife9368]

I really like the way stockfish defended the threats

[@hippophile]

Thanks for the commentary. I gave up chess 35 years ago to concentrate on Go, so I have some feeling for AlphaZero and its previous incarnations (AlphaGo, "Master" etc). It loves control and it loves efficiency (it also loves good shape, but that is not really a concept in chess like in Go). The control in particular is very evident in this game!!

[@TheChessViking]

hippophile the control part is evident in most of the others as well. Good observation and nice to hear from someone familiar with the GO side of things.

[@stylis666]

Good shapes are a matter of perspective though. I find AlphaZero's positional play the most beautiful shapes in the world (aside from butts, boobs and pretty faces :p). All of those pieces attacking and simultaneously defending, controlling not only the board but the opponent as well. I pity the fool that challenges it :p It's a work of art, every time it plays. My boards look like messy child's drawings compared to that.

[@CGoody564]

It doesn't love anything... It simply uses what it has learned... It has absolutely no emotion regarding the idea of control. Or shapes. Or chess. Or go.

[@zaph1rax]

Well, the nature we surrounds ourself with doesn't have feelings too, but it would also seem that nature loves shapes and control. I think AZ is just closing in to the nature of evolution :)

[@GlobalWarmingSkeptic]

AlphaZero is the most savage player on the planet. It knows no fear. All it knows is domination.

[@ronin6158]

this game really highlights kasparov's notion of 'quality'-- just because a piece is on the board doesnt mean its worth anything! Alpha0 of course 'understood' this, making the series starting at 10:35 more comprehensible. If A0 started trying to catch up too early, it would have made a little progress, but not enough to win after SF activates his useless pieces.

[@felixknowloveyoudlovetokno2324]

@Zibbit thankyou for this Video. I would appreciate it if you take a minut to Show the skipped moves. Literally just hitting next next next so that the 40 moves go by in a Minute would be much appreciated. Board Color is fine in my opinion. Also, I'm very happy that you Show ''what would happen if...'' and speculate. As a mediocre Player I am learning a fair bit fro Alpha Zero, but even more because you Show what I still have Trouble recognizing and visualizing. clear thumbs up!

[@bruceli9094]

This is Alien tech

[@jebronekitty]

Engine took hours to figure out, too us years to understand.

[@Corteum]

Bg5!! What an insane move!!

[@TheChessViking]

Indeed!

[@Corteum]

@@TheChessViking Any of y our engines able to find it in a reasonable time?

[@vikingpoolservice2051]

5:45 did you analyze the position here with a computer? Does white have an advantage?

[@Gamenetzwerk]

Wanna see Alpha0 vs itself.

[@InfiniteUniverse88]

At what ply did Stockfish see Bg5? Does the newest version of Stockfish see this move quicker?

[@MrAbrazildo]

From a machine, the double pawn sacrifice, without any kind of material compensation, and a dubious dynamic one, is much more impressive than 21.Bg5!!.

[@TheChessViking]

The first one is a known sacrifice dating back to Polugaevsky. Mind you though that apparently the computer taught itself that line! The second one was I think a new idea which is a very impressive concept indeed!

[@MrAbrazildo]

Right, Bg5 was a great move. However, nobody orientates machines to get themselves into 2-pawns down lines. It's not a forced line and it's a too much strategic one. PS: I don't even remember Polugaevsky with 2-pawns less.

[@spiritsofscience]

Brilliant commentary and very funny!!

[@TheChessViking]

Jack thanks man....really appreciate it!!

[@Shanksz]

Stockfish: Its over 9000!!!

[@ColHogan-bu2xq]

I'm a bad player, but can someone explain me why A0 doesn't take Knight d6 with Queen d1 at 2:50 ? And why SF choses Queen h7 rather than Queen f5 at 3:55 ? Thanks.

[@TheChessViking]

Taking the bishop was more important positionally, it made white's own dark bishop stronger. Qxd6 then Be7 and pawn d5 next to free the position. Later Qf5 looks like an alternative. Maybe SF didmt like g4 with tempo and then g5 opening up for an attack.

[@ColHogan-bu2xq]

@@TheChessViking Thank you so much for your time.

[@sraaju]

We don't need to worry about Skynet yet. As you said, Alpha is obviously material minded and would prefer a bland pointfull draw to actually winning. Only when someone invents electronic dopamine will we really be trouble.

[@S3v3n13tt3r5]

If Stockfish is like ~3300 rating, what is AlphaZero?

[@nikogagnon6614]

There is a new rating system that was just introduced. AlphaZero has a rating of 1, everything else is 0.

[@daniloreboucas5888]

they played 100 games each other resulting on 28 wins and 72 draws using an elo calculater, it is estimated that the elo difference is equal to 100 points exactly to check the results www 3dkingdoms com / chess / elo htm

[@gabydewilde]

over 9000

[@programinc7581]

S3V3N13TT3R5 atleast 4000

[@PhilNEvo]

That wasn't the best version of stockfish-- it will be interesting to see how good alphazero can become with more training and how it will fare vs the best version of stockfish :b

[@jamesavery3559]

well black got into trubble pretty fast, and the QR and QN in JAIL!

[@oneputtsteven]

The opening was horrible.

[@Mrtre7]

I don't believe for a second this is stockfish normal playing

[@jamesavery3559]

i never really watched a real tournament by the stockfish program 40 in 2 hrs, i may setup some of the position's on one of my chess machines and see if it play's the same dubious move's.

[@Mrtre7]

As I wrote above: my stone age Fritz in the critical position after Qf4 does not even for a minute prefer Nc5, and he tries to save the day with hxg5 instead. I wonder.

[@dannygjk]

James set up SF however you want it will take a long time to realize that AZ's sacs were sound.

[@selectnull4459]

Good colors i think :) i am starting to visualize the board no matter what it shows 😁

[@scarletevans4474]

12:57 - the video is almost 15 minutes long anyway, why not just spend an extra minute to quickly click through the moves, so people who are interested in them could pause the video and go through them by themselves? Could be an useful experience for some of us, who are still learning to play positions like this one =)

[@TheChessViking]

Scarlet Evans you are right! There is a link to replay of all the games in a lichess study in the description. It is game #4!

[@DoomRater]

Wow, Hatchet Harry really is a man you pay if you owe in this game.

[@Brandon-pf7oq]

Hi Zibbit, I'm very interested in chess, but more interested in the machine learning / deep learning aspect of it. Do you know what algorithm AlphaZero was programmed by and how many layers it had in its neural network (assuming it does have one in the first place)

[@TheChessViking]

Brandon Zhang check the pdf link in the description :-)

[@peacepeace4082]

plays like TAL

[@TheChessViking]

Very splashy game indeed! The feeling one gets must be like when people saw Tal's games for the first time back in the day!

[@wenhanzhou5826]

Tal is a genius, alpha0 is close to unbeatable...

[@musicalneptunian]

I'm going for Alekhine - this game reminds me of masterpice combinations like Alekhine Book.

[@bruceli9094]

Tal in the Machine

[@jazzminchen4204]

Peace Peace kjggv

[@andrewnoland7648]

4:20 if black goes for the fork on f5 I think you might see a really nice queen sacrifice as follows: Qf4 fxe4 Bxe4 Qh1 (Rxf4 Bxf4 Qh1 Bg6) Bg6! Rxf4 Bxf4. black's queen is stuck in the corner and white is winning. just a guess

[@andrewnoland7648]

The light square bishop can take on e4

[@wiredog771]

Nice video. Chilling though to think of how good AZ will get. 5000 elo??

[@Shootskas]

I don't understand why it was a brilliant move besides making Stockfish "think" for a few hours. What made it so brilliant?

[@Sasha1234575]

Thats not Chess. Thats another game. Chess is, what i try to play. Amazing.

[@TheChessViking]

montetank tankkiller next level stuff!

[@alessandrobolognesi4746]

These moves are so fucking deep

[@chengzhou8711]

I’m surprised Stockfish didn’t resign the game.

[@TheChessViking]

Cheng Zhou the Younger yes, was probably longer than necessary!

[@immanuellasker4273]

It was the ultimate fair of pins and forks

[@mr.k8902]

How i wish i could play like this.. Everytime i sacrifice, evaluation says ''inaccuracy''..

[@marshalljobe1102]

he didn't even finish the game... holy crap!

[@DjVortex-w]

AlphaZero seems to have an amazing and uncanny ability to smother black into a completely closed position where one or more of its minor and even major pieces are completely blocked and useless. At least in this kind of opening.

[@TheChessViking]

That's one of the recurrent features in many of the games.

[@DjVortex-w]

Stockfish is perfectly capable of sacrificing pieces, even in situations where the advantage is not apparent in the next 20-or-so moves. But I suppose that the programmers simply haven't figured out the proper algorithm for understanding that extra level of long-term strategy that A0 seems to have stumbled across on its own.

[@chrisharten1308]

I seen checkmate before the actual check mate. Seem like the computer program was justly being ruthless.

[@ceyhuncuner1118]

Thx dude, and this is a better color I think :)

[@ziadatris8669]

I wanna see alpha zero vs alpha zero. I wanna see chess come to an end. Alpha Zero find a forced mate from the beginning.

[@joi1794]

AZ vs AZ probably ends up in a draw. and if there is a winner its always going to be AZ (white) and never AZ (black).

[@MaLaS777]

Mind blown....

[@muradismailov9702]

All is good but I do not like the comments of unknown commentators about the strongest engines in the world. Please keep your comments and just show the game give us the joy. give comments as few as it possible because with all respect you cannot survive even five moves of these engines.... just show the game

[@TheChessViking]

If that was the case, nobody could comments on the games of Magnus Carlsen. Do you understand what an analysis video is?

[@muradismailov9702]

I hope you also understand that Magnusen is nothing for these engines. He will cry like a baby once he plays with any of them. Comments should be really deeply analyzed and given. The level these engines play the humankind can never achieve...So to give a comment where engines and what they have to do. I agree with comments but not so many. To some extent, you are right though...

[@TheChessViking]

You disqualified yourself from any intelligent discussion with "Magnusen". Everybody knows engines beat humans but the moves can still be understood and explained and this is the point of analyzing the games. If you want raw games just find a replayer online and do it at your own pace. Who watches a KZbin video for just the moves??

[@muradismailov9702]

I do not care what his name is, it does not matter. for me, he is the same human as you are or anyone else. this is not important. I am interested in the game and the rest is not important. Just I want to say that humans cannot judge super computer's level. I have lost a lot of my time watching this video. I like chess, but I am not ready to lose so much time. and if you want to insult me by saying about "intelligent discussion" it is your right not to answer at all. Hope my thoughts are clear and understandable... but anyway thank you for your time...

[@TheChessViking]

He is the world champion if you are interested in the game you should know who he is. You call this a loss of time and I am not qualified to comment on the game etc etc lol and then I am insulting you....you don't even know the name of the World Champion. Why are you even bothering? N.B: I am national coach in my country, FIDE master, FIDE certified trainer with wins against 3x World Blitz Champion (Grischuk) win against top online player (Nakamura) in blitz as well and countless GM scalps. But yeah...good luck finding someone "more qualified" ...I am sorry to bother you as an "unknown commentator"

[@wildDuces234]

very enjoyable game! good vid

[@TheChessViking]

Thanks!

[@joemacinnis1972]

Alpha zero is a beast

[@Peabody10]

Fookin hell what a game!😮

[@problematic7993]

Cynics don't see the big picture.

[@TheChessViking]

Exactly!

[@Xenon777channel]

If Stockfish was self-learning then it would stop playing the QID however it will continue with the QID until it can see the refutations at 97 ply depth or whatever so this would require computers which do not exist

[@BarniHedgeturtle]

What's the music used in the intro?

[@TheChessViking]

It's something I had made for me some time ago. Not from a known song or theme.

[@BarniHedgeturtle]

Unfortunate. :( It seems like quite the enjoyable piece.

[@islanbaya9956]

How can I download it to my computer?

[@pmlonline]

Sorry to disappoint but the math is clear that Google's DeepMind chess engine is ~ 400 ELO points lower than Stockfish if they were on equal computing. And that's given the fact that Google's chess engine runs on a processor that's custom designed for it, while Stockfish running on a generic processor. Design a custom cpu for Stockfish & their tree search algorithms and it'll be even better.

[@dannygjk]

The training hardware is not the hardware A0 used to play the match. Also explain to me 80 knps beating 70,000 knps. Your argument that the hardware was important is faulty.

[@pmlonline]

Incorrect. AlphaZero, like numerous chess engines, spends more time on each move rather than thinking deep. And yes, I was referring to the hardware that played the game. You have no clue what you're talking about. I'm a software & electrical engineer who's actually writing AI. I can tell you as a matter of fact that if I had Google's DeepMind funding I too could build a custom ASIC chip that would increase the tree search by tens of thousands of times, which would destroy AlphaZero.

[@xea-1226]

Paul lol😂😂 it's funny how refute an argument "you have no clue what you are talking about", "I am a software engineer bla bla bla". " absolutely wrong". So people must blindly believe just because you are Andrew ng or James Dean lol. You sound like a highschooler Lol

[@pmlonline]

You're blind if you can't see the facts I wrote. You seem like you're on drugs, btw.

[@xea-1226]

Paul lol your reply was exactly as expected, thanks for your superior (computer science) insights 😂😂

[@rasaffa3751]

Nice background music ....!!!

[@MISHRAIF]

why is stockfish movin or lockin in the knights on the left....

[@TheChessViking]

Rebecca Brown AZ locks it before Stockfish realizes. AZ inderstands the position. Stockfish just calculates it.

[@brandonkellner2920]

You might as well put double exclamation points behind every move AlphaZero makes. Those annotations don't make sense in this day and age anymore.

[@tomislavblazevic2742]

This really feels like watching chess played by aliens :D

[@Cghost-fh4hf]

Great resolts. Ofcourse playing games is intresting, but i'm looking forvard applaing this to real life problems such us physics, medicine, etc.

[@TheChessViking]

Yes, it's great to see A.I. applied to such a fixed set of rules even though the complexity of the game is immense. Will be interesting to see what A.I. will be able to do in the future in psychics and medicine, must agree with that!

[@kevincovins3546]

this alpha engine is like goku just turn ultra instinct on us

[@markservice8735]

Any guess on the chess rating of AlphaZero?

[@joi1794]

i heard people saying 4k. other only a 3.4k so 100 more than SF. so no one knows rly

[@ew6230]

I believe alpha-zero values pieces and through that value chooses it's path. For instance, a Rook might be a 5 whereas a pawn is a 1, but a pawn on the 7th rank might be valued at 8 because of the risk it poses becoming a queen and forcing pieces off. Another example would be, if I can move a pawn valued at '1' up and block the path of a Bishop, REDUCING the bishop's value, then that is taken into consideration as well. Making a piece stronger e.g. moving a knight forward to attack more making it a value 4 to a value 6 or 7 might not be better than changing a value 8 rook to a value 1 rook by pinning it or locking it down entirely in a corner...

[@TheChessViking]

Yes, Alpha-Zero seems very dynamic with piece valuations. Normal programs do this to an extent, for example at the start a, b pawns etc are less valueable than e- and d-pawns. This is one of the tough part of chess that the value actually shifts depending on mobility, restriction, pawn structure etc

[@ew6230]

What I am getting at is it seems like Alpha is better at determining which is better -- valuing/gaining value on existing pieces/taken pieces versus removing value from the opponents existing pieces. Stockfish seems to go for easy grabs and attacks over 'devaluing' the opponents pieces, which isn't always intrinsically correct I just mean it seems that way.

[@TheChessViking]

Yes, I think you are definatly on to something!

[@jojomomjonlang6445]

I just joined your channel. I heard it was just 1 minute per move? That is not very long for the Engines to think in depth. On the other hand AlphaZero Has Engine calculations plus Human Self learning abilities. So i would think that he can "out smart" the other engines due that that fact alone. After i looked at my brief analysis, i noticed that Sf was not playing the best moves due tot he fact of lack of time. Also, AlphaZero was not using the SF best move either, as the Bg5 move was not SF top 10 move. SO this is very interesting. What do you think?

[@TheChessViking]

JojomomJon Lang the human understanding of AZ is what stands out. It would be fun to see another match where SF owners can have their own input on timecontrol etc. I guess thats the only way to get cleat answers!

[@catterpitter]

Actually alpha zero showed improved decision making on time rating over one minute. One issue a lot of people brought up was the hardware stockfish was able to use it could have played a huge factor in the match. As far as I'm aware it was not regular tournament quality.

[@ralphinoful]

It's because of the differences in the core mechanics of the coding. Here's a brief explanation, and I can link a video to machine learning. Classical engines like stockfish are brute force, calculating, and analyzing machines. They take each position, analyze the position, compute who has the advantage to the best of their ability, and then will make the best move. Alpha-zero, and machine learning does not work like this. Machine learning is a code that basically mimics learning. Primitive machine learning programs will take in huge amounts of data, make a bunch of random moves, then compare the outcomes of their random moves, to the "correct" moves based on the data-set, and then tries to adjust itself. Alpha-zero is a bit more advanced because it doesn't need training data, it creates its own training data by playing itself, and then adjusts it's play accordingly. Instead of analyzing each and every position, it has already "learned" what moves are bad, and simply does not need to look into bad moves. It has a very good understanding of what is good vs. what is bad, in a move, and has optimized itself accordingly. If you want to learn more about it, I would suggest 3blue1brown's video series on it. He does a really good job of explaining the mathematics behind it, and you don't need to be a phd in math to understand his videos.

[@ariellmoreno9698]

Very good