Google's AlphaZero Destroys Stockfish In 100-Game Match
Klein (Mike)
Source: Chess.com, 6 Dec 2017
Paper - Abstract

Paper StatisticsColour-ConventionsDisclaimer


Full Text, including Comments

Emergent Tech → Artificial Intelligence

Google's AlphaZero Destroys Stockfish In 100-Game Match

AlphaZero vs Stockfish / artwork by Chess.com.

Google's AlphaZero Destroys Stockfish In 100-Game Match

FM MikeKlein

Dec 6, 2017, 12:50 PM


Chess changed forever today. And maybe the rest of the world did, too.

A little more than a year after AlphaGo sensationally won against the top Go player, the artificial-intelligence program AlphaZero has obliterated the highest-rated chess engine.

Stockfish, which for most top players is their go-to preparation tool, and which won the 2016 TCEC Championship and the 2017 Chess.com Computer Chess Championship, didn't stand a chance. AlphaZero won the closed-door, 100-game match with 28 wins, 72 draws, and zero losses.

Oh, and it took AlphaZero only four hours to "learn" chess. Sorry humans, you had a good run.

That's right -- the programmers of AlphaZero, housed within the DeepMind division of Google, had it use a type of "machine learning," specifically reinforcement learning. Put more plainly, AlphaZero was not "taught" the game in the traditional sense. That means no opening book, no endgame tables, and apparently no complicated algorithms dissecting minute differences between center pawns and side pawns.

Google Headquarters in London.

Google headquarters in London from inside, with the DeepMind section on the eighth floor. | Photo: Maria Emelianova/Chess.com.

This would be akin to a robot being given access to thousands of metal bits and parts, but no knowledge of a combustion engine, then it experiments numerous times with every combination possible until it builds a Ferrari. That's all in less time that it takes to watch the "Lord of the Rings" trilogy. The program had four hours to play itself many, many times, thereby becoming its own teacher.

For now, the programming team is keeping quiet. They chose not to comment to Chess.com, pointing out the paper "is currently under review" but you can read the full paper here. Part of the research group is Demis Hassabis, a candidate master from England and co-founder of DeepMind (bought by Google in 2014). Hassabis, who played in the ProBiz event of the London Chess Classic, is currently at the Neural Information Processing Systems conference in California where he is a co-author of another paper on a different subject.

Demis Hassabis with Michael Adams in London.

Demis Hassabis playing with Michael Adams at the ProBiz event at Google Headquarters London just a few days ago. | Photo: Maria Emelianova/Chess.com.

One person that did comment to Chess.com has quite a lot of first-hand experience playing chess computers. GM Garry Kasparov is not surprised that DeepMind branched out from Go to chess.

"It's a remarkable achievement, even if we should have expected it after AlphaGo," he told Chess.com. "It approaches the 'Type B,' human-like approach to machine chess dreamt of by Claude Shannon and Alan Turing instead of brute force."

AlphaZero vs. Stockfish 8

AlphaZero vs. Stockfish | Round 1 | 4 Dec 2017 | 1-0

87654321abcdefgh

1. Nf3 Nf6 2. d4 e6 3. c4 b6 4. g3 Bb7 5. Bg2 Be7 6. O-O O-O 7. d5 exd5 8. Nh4 c6 9. cxd5 Nxd5 10. Nf5 Nc7 11. e4 d5 12. exd5 Nxd5 13. Nc3 Nxc3 14. Qg4 g6 15. Nh6+ Kg7 16. bxc3 Bc8 17. Qf4 Qd6 18. Qa4 g5 19. Re1 Kxh6 20. h4 f6 21. Be3 Bf5 22. Rad1 Qa3 23. Qc4 b5 24. hxg5+ fxg5 25. Qh4+ Kg6 26. Qh1 Kg7 27. Be4 Bg6 28. Bxg6 hxg6 29. Qh3 Bf6 30. Kg2 Qxa2 31. Rh1 Qg8 32. c4 Re8 33. Bd4 Bxd4 34. Rxd4 Rd8 35. Rxd8 Qxd8 36. Qe6 Nd7 37. Rd1 Nc5 38. Rxd8 Nxe6 39. Rxa8 Kf6 40. cxb5 cxb5 41. Kf3 Nd4+ 42. Ke4 Nc6 43. Rc8 Ne7 44. Rb8 Nf5 45. g4 Nh6 46. f3 Nf7 47. Ra8 Nd6+ 48. Kd5 Nc4 49. Rxa7 Ne3+ 50. Ke4 Nc4 51. Ra6+ Kg7 52. Rc6 Kf7 53. Rc5 Ke6 54. Rxg5 Kf6 55. Rc5 g5 56. Kd4

One of the 10 selected games given in the paper.

Indeed, much like humans, AlphaZero searches fewer positions that its predecessors. The paper claims that it looks at "only" 80,000 positions per second, compared to Stockfish's 70 million per second.

GM Peter Heine Nielsen, the longtime second of World Champion GM Magnus Carlsen, is now on board with the FIDE president in one way: aliens. As he told Chess.com, "After reading the paper but especially seeing the games I thought, well, I always wondered how it would be if a superior species landed on earth and showed us how they play chess. I feel now I know."

Chess.com's interview with Nielsen on the AlphaZero news.

We also learned, unsurprisingly, that White is indeed the choice, even among the non-sentient. Of AlphaZero's 28 wins, 25 came from the white side (although +3=47-0 as Black against the 3400+ Stockfish isn't too bad either).

The machine also ramped up the frequency of openings it preferred. Sorry, King's Indian practitioners, your baby is not the chosen one. The French also tailed off in the program's enthusiasm over time, while the Queen's Gambit and especially the English Opening were well represented.

null

Frequency of openings over time employed by AlphaZero in its "learning" phase. Image sourced from AlphaZero research paper.

What do you do if you are a thing that never tires and you just mastered a 1400-year-old game? You conquer another one. After the Stockfish match, AlphaZero then "trained" for only two hours and then beat the best Shogi-playing computer program "Elmo."

The ramifications for such an inventive way of learning are of course not limited to games.

"We have always assumed that chess required too much empirical knowledge for a machine to play so well from scratch, with no human knowledge added at all," Kasparov said. "Of course I’ll be fascinated to see what we can learn about chess from AlphaZero, since that is the great promise of machine learning in general—machines figuring out rules that humans cannot detect. But obviously the implications are wonderful far beyond chess and other games. The ability of a machine to replicate1 and surpass centuries of human knowledge in complex closed systems is a world-changing tool."

Garry Kasparov and Demis Hassabis in London.

Garry Kasparov and Demis Hassabis together at the ProBiz event in London. | Photo: Maria Emelianova/Chess.com.

Chess.com interviewed eight of the 10 players participating in the London Chess Classic about their thoughts on the match. A video compilation of their thoughts will be posted on the site later.

The player with most strident objections to the conditions of the match was GM Hikaru Nakamura. While a heated discussion is taking place online about processing power of the two sides, Nakamura thought that was a secondary issue.

The American called the match "dishonest" and pointed out that Stockfish's methodology requires it to have an openings book for optimal performance. While he doesn't think the ultimate winner would have changed, Nakamura thought the size of the winning score would be mitigated.

"I am pretty sure God himself could not beat Stockfish 75 percent of the time with White without certain handicaps," he said about the 25 wins and 25 draws AlphaZero scored with the white pieces.

GM Larry Kaufman, lead chess consultant on the Komodo program, hopes to see the new program's performance on home machines without the benefits of Google's own computers. He also echoed Nakamura's objections to Stockfish's lack of its standard opening knowledge.

"It is of course rather incredible, he said. "Although after I heard about the achievements of AlphaGo Zero in Go I was rather expecting something like this, especially since the team has a chess master, Demis Hassabis. What isn't yet clear is whether AlphaZero could play chess on normal PCs and if so how strong it would be. It may well be that the current dominance of minimax chess engines may be at an end, but it's too soon to say so. It should be pointed out that AlphaZero had effectively built its own opening book, so a fairer run would be against a top engine using a good opening book."

Whatever the merits of the match conditions, Nielsen is eager to see what other disciplines will be refined or mastered by this type of learning.

"[This is] actual artificial intelligence," he said. "It goes from having something that's relevant to chess to something that's gonna win Nobel Prizes or even bigger than Nobel Prizes. I think it's basically cool for us that they also decided to do four hours on chess because we get a lot of knowledge. We feel it's a great day for chess but of course it goes so much further."




Comments

south_pawn south_pawn 1 day ago

augusto_jc schreef:

I understand what you say, but I don't fully agree, the hardware has to be the same, otherwise we are testing the software and the efficiency of the hardware at the same time. Then it is hard to get conclusions. I think they should call the creators of Stockfish and set up a proper tournament with prosecutors from both sides. This being said, still I think this is really great, not only for chess. The number of possible applications is huge. Regards.

The same hardware for stockfish and alpha zero just doesn't make sense. Neural networks can only run efficiently on specialized hardware, that can do lots of tensor calculations in parallel.

leoultimater leoultimater 1 day ago

Paper here: Link

The ten games in above paper also found here: Link

Follow Demis Hassabis on twitter here: Link

65Squares 65Squares 1 day ago

1200 games played by AlphaZero against Stockfish :

Quite interesting analysis of win rates for the Openings as W and B:

Link

augusto_jc augusto_jc 4 days ago

south_pawn wrote:

iRio schreef:

Alpha used for playing 4 TPU processors. It is maybe at least 1000 times faster than hardware for Stockfish. Try create match where will be play Stockfish with minute for move vs. Stockfish with 1000 minutes per move. What will be result ? 99 : 1 or 100 : 0 ??

That doesn't mean anything.. just an average desktop GPU also has 1000 cores.. What counts is the wattage that was used for both machines.. and that was the same AFAIK. So if the wattage was the same, the test was fair IMO.

I understand what you say, but I don't fully agree, the hardware has to be the same, otherwise we are testing the software and the efficiency of the hardware at the same time. Then it is hard to get conclusions. I think they should call the creators of Stockfish and set up a proper tournament with prosecutors from both sides. This being said, still I think this is really great, not only for chess. The number of possible applications is huge. Regards.

mcris mcris 9 days ago

@Eseles: Yes, see my reply here: Link

mcris mcris 12 days ago

alexanderla

mcris wrote:

alexanderla

mcris wrote:

I am beggining to think that SF8 was allowed only 1 sec/move, because only then can blunders like 35.Nc4 in the first game and 34...Rd8 in the game in this article can be explained.

What is your alternative to 34...Rd8 ?

It is bxc4

ok

35. g4 and black is lost

Just have seen on Youtube an IM explaining how SF lost because it didn't stop rather obvious attack plans of AZ. Maybe you can explain why 35.g4 is so good move?

mcris mcris 13 days ago

This would work like that: Blank screen "Computer is thinking". Next normal screen "Your move".

No engine-engine matches.

south_pawn south_pawn 13 days ago

mcris schreef:

GPU on desktop doesn't calculate chess moves.

not yet.. because brute force engines can't use them.. next generation chess engines like Alpha Zero will run on your graphics card. happy.png

mcris mcris 13 days ago

GPU on desktop doesn't calculate chess moves.

south_pawn south_pawn 13 days ago

iRio schreef:

Alpha used for playing 4 TPU processors. It is maybe at least 1000 times faster than hardware for Stockfish. Try create match where will be play Stockfish with minute for move vs. Stockfish with 1000 minutes per move. What will be result ? 99 : 1 or 100 : 0 ??

That doesn't mean anything.. just an average desktop GPU also has 1000 cores.. What counts is the wattage that was used for both machines.. and that was the same AFAIK. So if the wattage was the same, the test was fair IMO.

Eseles Eseles 13 days ago

coaches.png ?!?

iRio iRio 13 days ago

south_pawn wrote:

These brute force type of engines like Stockfish don't really get much better with more computing power.

Alpha used for playing 4 TPU processors. It is maybe at least 1000 times faster than hardware for Stockfish. Try create match where will be play Stockfish with minute for move vs. Stockfish with 1000 minutes per move. What will be result ? 99 : 1 or 100 : 0 ??

south_pawn south_pawn 13 days ago

kroksis schreef:

Reading more and more about this so-called match I am becoming more and more convinced that is this is real fake news case. There was no real match on equal conditions. Even the publicly revealed conditions of this pseudomatch are so very biased.

Oh, yes, sure I can beat Mike Tyson. If both of his hands are tied behind his back, and his legs are tied as well, and I am given a machine-gun.

These brute force type of engines like Stockfish don't really get much better with more computing power. The Stockfish on a smartphone can probably beat Carlsen with ease. That's because the search tree gets exponentially bigger.. for each extra depth, there are on average 20 more moves to consider. That's 20 times more computing power for each extra move. What Alpha Zero does is completely different. It uses (an approximation of) infinite depth for all the moves it considers. It plays entire games against itself and determines what percentage of games are winning, losing, or drawing. That's how it finds those really deep positional ideas that brute force engines will never find.

Eseles Eseles 13 days ago

But really, alL this reaffirms is how productive playing with your self for a few hours can be...

kroksis kroksis 14 days ago

Reading more and more about this so-called match I am becoming more and more convinced that is this is real fake news case. There was no real match on equal conditions. Even the publicly revealed conditions of this pseudomatch are so very biased.

Oh, yes, sure I can beat Mike Tyson. If both of his hands are tied behind his back, and his legs are tied as well, and I am given a machine-gun.

SirFlintstone SirFlintstone 14 days ago

It was easier for google than that. Any one from google who follows chess and sees how Stockfish's analysis often changed from bad moves, to avg moves, to ok ones, and so on could have easily seen at the one minute point SF was switching to better moves.

Thomas_Levi Thomas_Levi 14 days ago

MickinMD wrote:

Just as a machine can kick or hit a ball farther than a human and solve math problems faster than humans, it takes nothing away from the admiration we should have for how for human chess players, just as we admire football, etc. players/

What I want out of a chess engine is for it to tell me something like, "You should NOT have pushed your h-Pawn to launch a K-side attack. You should have posted your Knight on the great outpost at c5 and pushed your d and e Pawns right up the middle because your pieces are aimed there and Black is weak on the White Squares."

The goal of the research is not to give advice to humans. Would you rather have a super intelligent AI perform surgery on you with perfect precision and knowledge, or a super human AI give advice to the surgeon after the surgery is over? The goal here was not to improve our knowledge of chess but demonstrate what the AI is capable of through self-learning and then move on to real world tasks that are far more important then the game of chess, like medicine or discovering new cures. The fact that we can learn some things from the games is a bonus.

admkoz admkoz 14 days ago

prusswan wrote:

admkoz wrote:

"But going deeper the relationships between the pieces on the various squares comes into play as weighted nodes. The number of potential nodes is practically endless, and they aren't telling us what nodes are and aren't tracked."

If that is how it works, then it seems that the decision as to what relationships to track would be a backdoor way of smuggling in human knowledge.

They can certainly input human knowledge as training data, but what would be the value of doing so? By stripping away human input, they already achieved a higher level of Go-playing than ever before: Link

That's a different question. My question is what they actually did, not so much why. I guess I am trying to figure out something along the lines of: you fire up alpha zero for the first time and it plays a random game. Say Black wins. What does it "learn"? Just a) that exactly those random moves were a bad idea, or b) does it generalize as some point? It seems like option a would imply that AZ was brute force and it seems like that would in turn take forever. Option b on the other hand raises the issue of how the generalization works without human advice.

inselschaker inselschaker 15 days ago

Aizen89 wrote:

Is anybody else surprised at how little the Sicilian was played as Black and how much the English was played as White? I found that to be shocking.

It merely indicates that AlphaZero settled on a positional style - while still spotting tactics when they arise. No more "shocking" than some (not all) human players having a similar preferred opening repertoire.

For example, Aronian rarely plays the Sicilian (in the current decade just three times in rapid/blitz). Similar story for Adams (two games 2012/2013 at the Bunratty Masters against weak opponents). Kramnik played the Sicilian regularly until about 2005, then only in three must-win situations (last game of his WCh match against Anand, last round of Amber 2009 against Leko, last round of Dortmund 2013 against Adams).

Aizen89 Aizen89 15 days ago

Is anybody else surprised at how little the Sicilian was played as Black and how much the English was played as White? I found that to be shocking.

Comment:

Text Colour Conventions (see disclaimer)

  1. Blue: Text by me; © Theo Todman, 2018
  2. Mauve: Text by correspondent(s) or other author(s); © the author(s)



© Theo Todman, June 2007 - Sept 2018. Please address any comments on this page to theo@theotodman.com. File output:
Website Maintenance Dashboard
Return to Top of this Page Return to Theo Todman's Philosophy Page Return to Theo Todman's Home Page