ARCHIVED POSTS
< Earlier Kibitzing · PAGE 1041 OF 1118 ·
Later Kibitzing> |
| Dec-06-17 | | Tiggler: <WannaBe: <Domdaniel> Too late, we are actually not real, but> ... living in a reinforcement paradise. |
|
Dec-06-17
 | | Domdaniel: Here's something that isn't quite clear to me: does AlphaZero have the processor strength to see further and reach a greater ply depth than dedicated engines like Stockfish? Or are the algorithms that it evolves so superior that depth barely matters? Anyone? |
|
Dec-06-17
 | | Domdaniel: <Wannabe> But that would make you the Final Wabbit ... which I can't believe. There must be an ultimate uber-wabbit somewhere out there ... Otherwise, why would carrots have evolved? |
|
| Dec-06-17 | | Tiggler: <Domdaniel> I don't think it works that way at all. It just sees the current position and chooses the best move. So, zero ply. It just knows. |
|
Dec-06-17
 | | WannaBe: <Domdaniel> From what my limited learning of computers and how it works, depth and ply will only be limited by: 1. RAM, how much it can store in its memory for immediate access. 2. Hard drive capacity. Because for a computer, when it runs out of RAM, it writes information to disk. Imagine, if you will, someone who is forgetful, (low RAM) who is constantly writing things down. For computers this will slow down performance, hence they always recommend more RAM. From what I have read on A0, it is a specially designed/proprietary chip by Google (TPU) see my post on Stumpers page, but no information on the RAM or what OS it runs on. Is it Windows? Java (android phones run on Java)? Linux? I, too would love to know more, but I doubt our 'new' master Google will divulge/reveal. |
|
| Dec-07-17 | | zanzibar: <chessgames> You can find the PGN for the AlphaZero--Stockfish match from a link here: Stockfish (Computer) (kibitz #239) Ready for uploading. |
|
Dec-07-17
 | | jessicafischerqueen:
<gOOGLE> wants to kill us. You heard it here first. |
|
Dec-07-17
 | | chessgames.com: Note *quite* ready for uploading:
[Event "PGN Import"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "*"] |
|
Dec-07-17
 | | Stonehenge: Ready for uploading are games that are in the queue. Very ready if I may say so. |
|
| Dec-07-17 | | zanzibar: <chessgames> I rechecked the PGN, doing a download from Google drive and reading it into SCID, and it checked out for me. Did you really follow the links and try to download it? There no views in my records - nobody downloaded it (yet, I hope). * * * * *
Could somebody try to download the PGN using this link https://zanchess.wordpress.com/2017... and let me know it's all OK?
* * * * *
<<Stonehenge:> Ready for uploading are games that are in the queue. Very ready if I may say so.> <Stonehenge> To be explicit, did you upload the AlphaZero games? If so, might I suggest that in the future you provide a brief note in the forums so that your colleagues can avoid needlessly duplicating work? E.g. this forum would work for the match games:
Stockfish (Computer) Of course <chessgames> would have the games in the queue, but even he might overlook them without public notice. And the rest of us still don't have the ability to see the submit queue status for submitted games(*). (*) A feature we'd all like to see <CG> adopt, I suspect. |
|
Dec-07-17
 | | WannaBe: <Z> I downloaded one here is a snippet: ;; <AlphaZero // Stockfish * (R1) 67 1-0>
;; <Stockfish -- AlphaZero (R1) 67 0-1> [Event "AlphaZero -- Stockfish 8 match"]
[Site "London ENG"]
[Date "2017.12.04"]
[Round "1"]
[White "Stockfish"]
[Black "AlphaZero"]
[Result "0-1"]
[EventDate "2017.12.04"]
[PlyCount "134"]
1.e4 e5 2.Nf3 Nc6 3.Bb5 Nf6...... |
|
| Dec-07-17 | | zanzibar: Great <WannaBe>, thank you. |
|
Dec-07-17
 | | Stonehenge: <did you upload the AlphaZero games?> No, I was talking about other games. CG should upload the games that are in the queue at least once a week but one may as well talk to the walls here. |
|
Dec-07-17
 | | Domdaniel: <Jess> - <gOOGLE wants to kill us> ... "Go Google, go ogle googols of go-go gals, go glue goals..." |
|
| Dec-07-17 | | zanzibar: <Stonehenge> ah, sorry on my part for being pedantic then. Agree with your sentiment otherwise... |
|
Dec-07-17
 | | chessgames.com: <Stonehenge> Working on it now. Just FYI only one of yours was rejected, because the final move makes no sense: [Event "Antwerp"]
[Site "Antwerp BEL"]
[Date "1931.07.30"]
[EventDate "?"]
[Round "2"]
[Result "0-1"]
[White "Georges Koltanowski"]
[Black "Fred Dewhirst Yates"]
[ECO "?"]
[WhiteElo "?"]
[BlackElo "?"]
[PlyCount "46"]
1. d4 d5 2. c4 e6 3. Nc3 Nf6 4. Bg5 Nbd7 5. e3 c6 6. a3 Be7
7. Qc2 O-O 8. Rd1 b6 9. Nf3 Bb7 10. Bd3 c5 11. dxc5 bxc5
12. Be2 Ne4 13. Nxe4 dxe4 14. Ne5 Bxg5 15. Nxd7 Qa5+ 16. Qd2
Qxd2+ 17. Rxd2 Rfc8 18. b4 cxb4 19. axb4 Be7 20. c5 Rc7
21. O-O Bd5 22. Bb5 Rd8 23. Ne5 Rb5 0-1
<AlphaBeta> I have usable PGN now, I just have to infer that the Site should read "London ENG" according to the paper published here: https://arxiv.org/pdf/1712.01815.pdf. Look for the games soon. |
|
| Dec-07-17 | | zanzibar: <chessgames> yes, I used "London ENG" in the PGN because of that. |
|
| Dec-07-17 | | zanzibar: <There no views in my records - nobody downloaded it (yet, I hope).> I was wrong about this, btw... not that it matters anymore (user error, i.e. me). OK, thanks. |
|
Dec-07-17
 | | chessgames.com: We've got it now, and all I can say is, wow. See for yourself: AlphaZero - Stockfish (2017) Biographers: feel free to improve upon my slapdash article. Hopefully Google will divulge more information in the near future. |
|
Dec-07-17
 | | Stonehenge: Sorry, I got that game here:
http://www.belgianchesshistory.be/t... It says "<23. Ne5 Het Schaakleven gives here Rb5 and white resigns. I don't know what happened. 0-1>" I guess either you don't upload it or put < 23. ♘e5 ...more moves 0-1 > in the PGN. |
|
Dec-07-17
 | | chessgames.com: Ah, 23.Ne5 makes much more sense. I inserted it with {...} at the end: Koltanowski vs Yates, 1931 even though it's not clear if more moves were played. Perhaps 23...Rb8 was played with the threat of Rxb5, and somehow that got conflated with the game score. Anyhow, nice game to have among giants of the day. |
|
Dec-07-17
 | | Domdaniel: <CG> Nice work, Daniel. |
|
| Dec-07-17 | | Alien Math: <Domdaniel> A paper about the AlphaZero processor https://arxiv.org/abs/1704.04760 and about their algorithm,
<Instead of an alpha-beta search with domain-specific enhancements, AlphaZero uses a general-purpose Monte-Carlo tree search (MCTS) algorithm.> as found at the link shared earlier in the chessgames forum https://arxiv.org/pdf/1712.01815.pdf |
|
Dec-07-17
 | | chessgames.com: A brave new world we're living in.
I see that some members have already started to use the Analysis Laboratory to analyze the match. So we have Stockfish analyzing its own moves — as well as the moves of the computer that beat it. Not quite sure what that will teach us, but it should be interesting. |
|
| Dec-07-17 | | Travis Bickle: Daniel, I'm still getting ads on the Kibitzer Cafe on the right side of the screen. What gives? |
|
 |
 |
ARCHIVED POSTS
< Earlier Kibitzing · PAGE 1041 OF 1118 ·
Later Kibitzing> |