A couple of weeks ago, in The Question of Clones, I used a block diagram taken from a thread on Talkchess.com, and noted,
While I can't claim to understand all of the subtleties present in the diagram, its blocks represent the overall architecture behind Lc0 and are useful for delving into the clone question. In the same thread, three variations of the diagram used color codes to show which portions of other AI/NN engines are unique and which portions are copies of Lc0.
One of the three variations depicted the origin of Lc0 (aka Leela) itself.
My failed attempt to change TCEC NN clone rules
(September 2019)
A comment to the thread by dkappe added,
LCZero and Lc0 are two distinct engines. The first, LCZero, was the initial chess engine used by the Leela Chess project and was derived from a combination of the Leela Go engine and Stockfish chess board logic. The second, Lc0, was a rewrite from scratch.
Dietrich Kappe was the founder of the original LCZero blog -- see that blog's second-ever post for a link to its first incarnation. His comment is expanded in the relevant page on Leela Chess Zero (chessprogramming.org):-
Lc0 • Leela Chess Zero consists of an executable to play or analyze games, initially dubbed LCZero, soon rewritten by a team around Alexander Lyashuk for better performance and then called Lc0. This executable, the actual chess engine, performs the MCTS and reads the self-taught CNN, which weights are persistent in a separate file.
Lc0 is written in C++14 and may be compiled for various platforms and backends. Since deep CNN approaches are best suited to run massively in parallel on GPUs to perform all the floating point dot-products for thousands of neurons, the preferred target platforms are Nvidia GPUs supporting CUDA and CuDNN libraries.
Ankan Banerjee wrote the CuDNN backend code, also shared by Deus X and Allie. None [Non?] CUDA compliant GPUs (AMD) are supported through OpenCL, while much slower pure CPU binaries are possible using BLAS, target systems with or without a graphics card (GPU) are Linux, Mac OS and Windows computers, or BLAS only the Raspberry Pi.
Even though that last sentence doesn't parse correctly, the description helps to understand of blocks 'I', 'K', and 'L' in the block diagram. The diagram labels their aggregate 'Engine' and color codes them as LCZero, the first Leela Chess engine.
Last month, in Oldest Chess Piece? (November 2019), I wrote,
In Yahoo's heyday, stories like [this] would have easily been featured in Yahoo News. Now I have to rely on Google News for this sort of thing, where Google News is more-and-more a rehash of Chess.com News.
Not so long ago, stories like the following would have attracted widespread attention.
Need proof? It's been only eighteen months since June Yahoos (June 2018), which recalled yet another chess story from Iran:-
Let's skip the Yahoo stub and jump directly to the real story. It echoes another Yahoo story that I used twice on my World Chess Championship blog: Hijab Hubbub (October 2016), and Hijab Hubris (ditto).
2018-06-13: Chess player pulls out of championship over Iran's rules. In 2016, the story's protagonist was Nazi Paikidze, the reigning U.S. Women's Champion at the time. This time it's an Indian player. Conscience knows no national boundaries.
The two headlines pictured above are in chronological sequence. The first was the 'before' story:-
Making this the logical 'after' story:-
The main chess news sites ran a report before the 'before' story, for example:-
The 12-24 portion of the story was tentative:-
The Iranian chess star and the country's top player Alireza Firouzja is considering changing his nationality after the Iranian chess federation withdrew its players from the upcoming World Rapid & Blitz in Moscow, according to the Iranian chess federation.
The 12-25 portion was following a moving target:-
Update: Firouzja is now back on the World Rapid & Blitz players list under the FIDE flag.
Recently GM Firouzja also appeared on my chess960 blog in Quarterfinals of FWFRCC Completed (October 2019). That long acronym stands for 'FIDE World Fischer Random Chess Championship', which is longhand for 'World Chess960 Championship'. He was knocked out in the quarterfinals.
Another two weeks have passed since this blog's most recent report on the two most popular engine-vs-engine tournaments. To summarize that previous report, More TCEC Bonuses; CCC11 Semifinal Underway:-
TCEC: The site has progressed to 'Viewer Submitted Openings Bonus 9'. • CCC: The CCC11 Semifinal round is just getting underway.
That wasn't much of a report, but much has happened since then.
TCEC: The site has progressed to 'Viewer Submitted Openings Bonus 12'. In TCEC S17 – detailed information (chessdom.com), it announced firm plans for the next season:-
The Top Chess Engine Championship TCEC is going to start in about 10 days time. And it is going to be amazing! TCEC S17 will have a solid hardware upgrade, a new adjusted format and a new structure, a dedicated news website with a merchandising section and more.
That puts the start of the new season around 1 January 2020. In case you've forgotten the previous season, I reported it in Stockfish Wins TCEC S16; CCC11 Underway (October 2019). I reported on the subsequent cup competition in TCEC FRC Leagues; CCC11 R2 Underway (November 2019): 'TCEC Cup 4 finished with Stockfish beating LCZero for 1st place (+1-0=7)'.
The site's '!next' command informs, 'VSOB were all S17 testing and continue until S17 starts.' The acronym 'VSOB' stands for 'Viewer Submitted Openings Bonus'. More information about the new season is available at TCEC Season 17 Rules (wiki.chessdom.org).
CCC: The following chart shows the final crosstable for the CCC11 Semifinal round.
Lc0 and Leelenstein both finished with 37.5 points out of 72 games. The small print on the bottom line explains,
Stockfish and Lc0 advanced to Finals • Lc0 SB tiebreak was 1330.75; Leelenstein SB tiebreak was 1312.75
In their head-to-head semifinal match, Lc0 beat Leelenstein +3-1=20. Before the final match started, the two finalists played a warmup match titled 'CCC11 Finals (Bullet Test : 1+1)', where Lc0 beat Stockfish +24-14=112.
As I write this, the 150-game final match has eight games left to play. Lc0 leads Stockfish by four points.
[For further information from the various stakeholders in the engine-to-engine events, see the tab 'TCEC/CCC Links' at the top of this page. • NB: Leela = LC0 = LCzero]
***
Later: The CCC11 Final match ended with Lc0 maintaining its four point lead over Stockfish. The last eight games were drawn, giving a final score of +13-9=128. This was the second straight loss for Stockfish in a CCC final; see TCEC S16 Final Underway; Leelenstein Wins CCC10 (September 2019) for the result of the previous final.
This month's edition of The Sociology of Chess (November 2016), echoes a post from two years ago, London Chess Conference 2017 (December 2017).
KVDC interview IM Malcolm Pein London Chess Conference 2019 (14:03) • '[Published on] Dec 5, 2019'
KVDC is Karel van Delft. His description says,
IM Malcolm Pein tells about the London Chess Conference 2019, Chess in Schools and research about benefits of chess education.
Other videos on the KVDC channel from the London Conference include:-
A description of the clip about GM Rowson is at 64 Life Lessons (londonchessconference.com):-
In The Moves that Matter the Scottish grandmaster with a PhD in philosophy and a very active Twitter account reflects the life lessons chess taught him in 64 chapters. Stephen Moss from The Guardian, whose book The Rookie was published in 2016, also by Bloomsbury, interviewed him during the lunch break of the first conference day. Jonathan was later on also interviewed by Karel van Delft.
The clip about the journal 'Chess Education and Science' features Jerry Nash of USchess presenting an initiative of the Chess Scientific Research Institute (CSRI), Armenian State Pedagogical University. Anyone who remembers the Journal of Chess Research (May 2014), will hope that the new journal actually sees publication.
For more about the conference, see Chess and Female Empowerment conference concluded in London (fide.com). The web site for the conference itself can be found by following the link for '64 Life Lessons' given above.
When is one chess engine a copy of another chess engine? Engine developers and tournament administrators have been grappling with this question for as long as there have been chess engine competitions. A recent thread on Talkchess.com, titled My failed attempt to change TCEC NN clone rules (September 2019), started,
I get frequent questions why don’t I bring this topic up if I don’t agree. So, I tried to bring this topic up with TCEC administration, with no success.
The author was Alexander Lyashuk (crem; chessprogramming.org), 'primary developer of Leela Chess Zero', and his thoughts attracted a slew of other developers from the expanding universe of AI/NN chess engines. To start the discussion, crem posted a number of block diagrams; a portion of the main diagram is shown below.
While I can't claim to understand all of the subtleties present in the diagram, its blocks represent the overall architecture behind Lc0 and are useful for delving into the clone question. In the same thread, three variations of the diagram used color codes to show which portions of other AI/NN engines are unique and which portions are copies of Lc0. One diagram, for example, explained the relationship between Lc0 and two engines that were the subject of a recent post on this blog, AllieStein and Leelenstein (November 2019).
Not too surprisingly, even the experts couldn't come up with an agreed definition of an AI/NN chess clone. What hope is there for the rest of us? Even so, I'm sure I'll be referrring to the diagram in the future.
'Jackpot!', I told myself, when Flickr invited me to 'View all 265' images available since last month's Doesn't Look Like a Photo (November 2019). My enthusiasm quickly evaporated when I realized that 95% of the images were copies of a handful of chess magazine++ covers all pointing to the same shopping page. Did the page have anything to do with chess? I'll never know, because I rarely click through to an obvious spam scam.
Fortunately, I had another handful of real chess photos to choose from for this post. This was the best of the bunch.
Cirque Du Soleil
© Flickr user R.I. Pienaar under
Creative Commons.
The full title of the photo -- and, yes, it's a photo -- could also serve as a description:-
Pretty good show last night at Cirque Du Soleil. I would say this has been a hugely challenging situation for the pedant in me to go to a chess themed show with a 7 x 7 board!
What's Cirque Du Soleil? Wikipedia's page on the subject, Cirque du Soleil, says,
Cirque du Soleil (French, "Circus of the Sun" or "Sun Circus") is a Canadian (Quebecois) entertainment company and the largest contemporary circus producer in the world. Based in Montreal, Quebec, Canada, and located in the inner-city area of Saint-Michel, it was founded in Baie-Saint-Paul on 7 July 1984, by two former street performers.
The photo's home page didn't offer any clues to where it was taken, but many of the photographer's other photos pointed to the Mediterranean island of Malta. Indeed, I found another page dated December 2019, titled Acrobats wow audience at MCC (timesofmalta.com), and subtitled 'Cirque du Soleil brings original show inspired by the game of chess to Malta'. MCC doesn't mean Malta Chess Club; it means Mediterranean Conference Centre, located in Valletta, which happens to be the capital of Malta. The plot goes something like this:-
This is not a story about war. When the Golden King falls in love with the Black Rook, he embarks on reckless moves to capture her heart, throwing this orderly world into disarray. As he tries to get closer to his love, he faces a pretentious, narcissistic monarch and his fearsome Queen.
That doesn't fit at all into my world view of chess, but I would still take advantage of an opportunity to see it.
Time to get up-to-date on the world's premier, ongoing engine vs. competitions. To summarize the previous post, TCEC/CCC Bonus Events, from two weeks ago:-
TCEC: The site is running a series of bonus tournaments; the current event is titled 'Viewer Submitted Openings Bonus 4', with eight engines. • CCC: CCC11 R2 finished with Lc0 3.0 points ahead of second place Stockfish. After R2, the CCC ran two matches; Fat Fritz lost to both Stockfish and Lc0 by convincing margins.
TCEC is still in a holding pattern, while CCC has advanced another stage. Let's have some specifics.
TCEC: The site has progressed to 'Viewer Submitted Openings Bonus 9'. As for the next season (S17), the schedule is nearly identical to what was announced two weeks ago; it says:-
!s17 • Tentatively set mid-December 2019. Delay due to a potential sponsored hardware upgrade.
The plans for the various stages of S17 are still unannounced.
CCC: CCC11 R3 finished with AI/NN engines Leelenstein and Lc0 tied for 1st/2nd, a half-point ahead of Stockfish. Komodo placed fourth, 6.5 points behind Stockfish. The following chart shows the full R3 crosstable. There was only one decisive game among the top three engines, and the two AI/NN engines lost no games.
The top four engines qualified into the CCC11 Semifinal round, which is just getting underway. The tournament is projected to run for somewhat less than a week. The top two engines in the semifinal will advance to a 75-game final.
[For further information from the various stakeholders in the engine-to-engine events, see the tab 'TCEC/CCC Links' at the top of this page. • NB: Leela = LC0 = LCzero]
When was the last time you had chess instruction with comic asides?
Paul Morphy's 5 Most Brilliant Chess Moves (9:29) • '[Published on] Nov 13, 2019'
The description says,
Join GM Simon Williams as he breaks down the best moves from Paul Morphy, the first great American chess player and considered by many to be an unofficial world champion.
This video is one of a series by GM Williams on the same theme:-
Chess instruction and entertainment at the same time -- yes, it's possible, if you're Simon Williams.
Of the three chess personalities featured in the previous post, December 1969 & 1994 'On the Cover', the most photogenic was GM Pia Cramling of Sweden. By coincidence, the last time I looked at chess items on eBay I noticed several photos of her that sold a few months ago.
The full title of the photo on the bottom right was 'Vintage photograph of Pia Cramling playing chess with Hasse Alfredsson'. According to my page World Chess Championship : Index of Women Players, GM Cramling has been competing in the Women's World Championship since the 1985 Havana Interzonal Tournament, where she tied for 3rd through 5th place, thereby qualifying for the 1986 Malmo Candidates Tournament.
Last month -- in the most recent post for an ongoing series, November 1969 & 1994 'On the Cover' -- I introduced a small change in direction:-
Let's continue the 'On the Cover' series with CL&R from 50 years ago, together with CL from 25 years ago.
A second post using the same idea confirms the start of a new series.
Left: 'International Master Norman T. Whitaker'
Right: 'Happy Holidays from Sweden!'
Chess Life & Review (50 Years Ago)
Sixty-five Years in American Chess by International Master Norman T. Whitaker • I learned the moves of chess from my father (a mathematician) when I was fourteen; we lived then in Philadelphia. The next year, 1905, I learned of its beauties by watching Harry Nelson Pillsbury on a pier in Atlantic City, New Jersey. Late in the day he would play all comers for a fee, but I would visit him before noon. [...]
We've already seen IM Whitaker on this blog in a couple of previous posts: Smythe, Not Smith (May 2017, photo) and Masters Emeritus in the 1950s (August 2017, links to relevant pages offsite; Wikipedia: 'several terms in prison').
Chess Life (25 Years Ago)
Season's greetings: Actually, it was August. But there is a fair amount of holiday coloring in the photograph taken by Peter Walz at the opening ceremony of a tournament in Osterskars [Osterskar], a suburb of Stockholm. We thank Nick deFirmian [de Firmian] for bringing "King Karpov" and "Queen Pia Cramling" to our attention. Oh, yes - Nick tied for second place in the Swedish event, behind Kveinys of Lithuania.
For more about "Queen Pia Cramling", see Pia Cramling (wikipedia.org).
The 'CCC PGN Archive' (see the tab 'TCEC/CCC Links' at the top of this page) hasn't been updated in over six months, and only shows events through CCC8. For this post I set out to find more recent PGN files and found links for them via #tournament-pgns (discordapp.com). The Discord messages are of the type...
CCC Archive 9/21
https://pastebin.com/ZKLFBFfU
... and look like this:-
Download links are all of the form:
https://cccfiles.chess.com/archive/tournament-TOURID.pgn
Tourid Title
40044 CCC10 Finals
39682 CCC10 Semifinals
[...]
With this data it should be trivial to download the relevant files. I'll add the link given in the first paragraph to 'TCEC/CCC Links'.
The first and last time we saw jewelry on Top eBay Chess Items by Price (started March 2010), was The Value of Jewelry (March 2016). The items shown in the image below appeared a week apart on eBay from different sellers.
Top: 'Rare Trifari Sterling Silver White King Queen Chess Pin Clips'; sold for US $406 after three bids from two bidders.
Bottom: 'Rare Find! 1940s TRIFARI CHESS King Queen STERLING Set Figurals Pins Brooches'; US $475 Buy-It-Now.
Although both photos show the Queen on the left, I bet many sellers would have shown the King on the left. Isn't that some sort of protocol for royals?
The description of the item on the top only repeated the title of the auction and added 3/4" x 1-5/8". The description on the bottom was more informative:-
A set of the 1945 Trifari Sterling Chess Figural Brooches. Designed by Alfred Philippe. You get the King and Queen!
Featuring black enameled faces, cream enameled accents, ruby and crystal chaton rhinestones and small turquoise beads in their crowns, all set in vermeil plated sterling silver, with double prong fur/pin clip mechanisms on reverse. A royal treat! Love these figurals!
Hallmarks: Trifari with Crown over the T, Sterling, des. pat. no. 140843 and 140844 (see patent drawings). Measurement: 3/4" x 1-5/8".
For more about Trifari, see Vintage Trifari Jewelry (collectorsweekly.com).
Posts
All Comments
