Last week I posted that the TCEC Season 8 was essentially over even though Stockfish had (on paper) a chance to win. Now with 98 games played and only two games left the match is mathematically over as well. The smell of fried Stockfish permeates the air--Komodo has dominated! The current score is 2 wins for Stockfish, 9 wins for Komodo with 87 draws. A lopsided match like this is remarkable. Remember Stockfish handily dispatched Nakamura last year. That match saw two games where Nakamura had computer help and two games where Nakamura had pawn odds plus the white pieces. Even though accounts have him playing for the win instead of taking a draw you still have to come to terms that one of the best players in the world had white plus a pawn advantage and couldn't beat Stockfish. Stockfish is that good.
But as strong as Stockfish is to mere carbon based life forms that's just not good enough against last year's winner Komodo. Komodo showed that Stockfish is far from perfect. Sometimes the difference between their evaluation was glaring; here Komodo evaluates the position as -250 while Stockfish has it at -6.18:
Chess For all Ages has a post on the second week of TCEC which mentioned some issues in the evaluation of positions. Specifically it said, citing an earlier post, "What happened in game 22 of the final? Stockfish's evaluation was steeply raising, until at move 62 it evaluated its position as +26.13. Komodo agreed to a degree by evaluating it at +4.22. Then two moves later both evals had dropped to 0.00, and it ended in draw. Rather unusual, I would say.". Rather unusual, indeed.
GM Ben Finegold used the term PMG over 25 years ago to refer to a (theoretical) perfect move generator; as in "How would Kasparov do in a 10 game match against a PMG?". Komodo's domination of Stockfish shows that although Komodo isn't a PMG (it lost 2 games and there might be an evaluation issue) but it is getting so very, very close. While Stockfish's weaknesses aren't exploited by humans, Komodo's play has been another notch higher; even Stockfish is having trouble beating Komodo. That exceptional play seems to have caught the attention of other people as well. Godel's Lost Letter and P=NP blog notes, "If Komodo could hold at least two draws every ten games against any player, then by the rules of the rating system, no player could be rated more than 366 Elo points above it. Thus a rating estimated near 3250 for it now would translate to an absolute ceiling of about 3600, which is the high end of where various of my chess model’s regression lines cross the -axis at meaning perfect play.". I will be curious how much Komodo's rating changes after this match. PMG might be reality in about 10 years.
Here are some stories that caught my eye over the last week:
- In these times, ZeroHege gives a history lesson: "This is likely to be a little tough to swallow for some Americans but The Statue of Liberty - symbolically welcoming the world's tired, poor, and huddled masses - was originally conceived as a Muslim peasant woman.". It goes on to cite Smithsonian magazine, "But the sculptor Frederric Auguste Bartholdi of France, proved unable to sell the idea to the khedive of Egypt, Ishma’il Pasha. Bartholdi remained determined to erect a colossus on the scale of the one in ancient Rhodes. He sailed to America with drawings of the Muslim woman transformed to the personification of Liberty."
- Quanta Magazine with the story of how a 50 year old math problem was solved. For those people who don't get why pure math, it's another example of how mathematics which seems to lack practical usefulness can turn out to be very important. "Network sparsification has applications in data compression and efficient computation, but Spielman’s particular problem suggested something different to Kalai. It seemed connected to the famous Kadison-Singer problem, a question about the foundations of quantum physics that had remained unsolved for almost 50 years....As a computer scientist, Spielman knew little of quantum mechanics or the Kadison-Singer problem’s allied mathematical field, called C*-algebras....C*-algebras are an esoteric subject — “the most abstract nonsense that exists in mathematics,” in Casazza’s words. “Nobody outside the area knows much about it.” For the first two decades of the Kadison-Singer problem’s existence, it remained ensconced in this impenetrable realm." But not anymore. The solution of the problem has applications to the Traveling Salesman Problem and "Now, using ideas from the proof of the Kadison-Singer problem, Nima Anari, of the University of California, Berkeley, and Shayan Oveis Gharan, of the University of Washington in Seattle, have shown that this algorithm performs exponentially better than people had realized. The new result is “major, major progress,” Naor said."
- What's behind the sudden rise in "White Student Unions"? Southern Valley News takes a look: "At least 30 social media profiles of so-called “white student unions” have been set up in the last week, in response to nationwide student protests demanding action to address campus racism. ...It appears that the first of these WSUs was the “Illini White Student Union”, created at the University of Illinois just hours after students gathered for a black solidarity event on campus last Wednesday. Its page has since been taken down by Facebook at the request of campus administrators....Given the rapid speed with which they have emerged, some have questioned whether the groups are really created by students. In a Medium post by the author “Bears for Equity”, it was noted late Sunday that well known white-supremacist and Neo-Nazi blogger Andrew Anglin called for his followers to “[m]ake more of these White Student Union pages on Facebook for various universities. You don’t have to go there. Make one for Dartmouth, Princeton, etc.”".
- Check out this piece from WHBL.com; it starts out reasonably enough making the analogy between math and cooking. The video goes horribly astray however. It tries to justify "supplementing" the standard algorithms with the Common Core way. The idea is that the new methods give students, according to the video, a better understanding of what they're doing as well as illustrating multiple ways to find the answer.This seems like the reasoning of overzealous math ed majors who lack math understanding forcing a bad idea onto their students. 1. Better understanding how? There's more confusion than before and the other methods really are irrelevant as they go deeper into math. 2. Force feeding multiple methods down their throats takes away time from other math topics, including the standard method which students are weak in. 3. These alternate methods might be appropriate with respect to mental math or remedial math. It has no place on tests of mathematical understanding.
- The many facets of technology: EAGNews reports on, "Numerous iPads issued to students and staff in Columbia Public Schools are now evidence in criminal investigations, and police are using them to determine who did what, and when...“Each case is different, but for the majority it was sexual exploitation of a child in some form or fashion,” Boone County detective Tracy Perkins told the news site.".
- Yahoo News has a piece on Massachusetts abandoning Common Core. "The Massachusetts State Board of Education has voted to forego Common Core testing in favor of redesigning its own state exam, an influential move from a national education leader that may hasten the end of a national high-stakes testing era, while challenging education experts to come up with a better alternative.". The Charlotte Observer writes "The Common Core isn’t dead yet, but it will be. The multi-state education collaborative suffered its most profound blow last week, when the Massachusetts Board of Education decided to leave the Common Core and develop its own state tests. ...Now, Massachusetts joins more than 15 states that have left the effort or, like North Carolina, are taking steps to do so. Each departure dilutes the case that Common Core makes – that it helps states measure how well they’re educating children by comparing test results with other states."
- bcrnews makes some lame excuses for poor PARCC results. "Princeton Elementary School (PES) Superintendent Tim Smith said the PARCC test results for the district were “every bit as bad as we thought they would be,” yet he isn’t worried, seeing as how Illinois State Superintendent of Education Tony Smith recently told administrators not to take the test results too seriously...“An eighth-grader’s taking the PARCC test as if he’s been getting Common Core from pre-kindergarten on, so how in the world that can be considered a useful test is beyond me,” Smith said....“The PARCC scores are so low we just know what we’re seeing is not realistic,” he said, indicating constant monitoring of the students will be a surefire way to pinpoint where they are and are not progressing..". So poor scores aren't can't be trusted as realistic (but believe me, good scores would have been trumpeted as proof of what a good job was being done) and the realistic scores won't be meaningful until the test has been in place for many years. But don't worry, there will be a new test before that happens, see the stories above, or Mr Smith will have moved on by then so he won't have to account for the failure. Where is the accountability? Oh wait...there is none.
- All those patriotic American who fought and died to defend our freedoms might be surprised to hear "40% Of Millennials Would Censor Offensive Speech". From the ZeroHedge piece, "Of those aged 18-34, 40 percent support censoring offensive speech.
"We asked whether people believe that citizens should be able to make public statements that are offensive to minority groups, or whether the government should be able to prevent people from saying these things. Four-in-ten Millennials say the government should be able to prevent people publicly making statements that are offensive to minority groups.""
- Liberty Blitzkrieg has a video of the day for the "crybully" millenials that is an "...excellent lecture ... by Jonathan Haidt, Professor of Ethical Leadership at New York University’s Stern School of Business."
- EAGNews with some input from students who went through a "white privilege" retreat. The University of Vermont hosted some voluntary retreats so white people could understand "how they have unfairly benefited from “white privilege,”...The program is hosted by the ALANA Student Center, which is composed of “proud people of color.” The message is clearly that it’s bad to be white because society automatically favors you, but a point of pride to be a racial minority. And that approach is somehow supposed to bring people closer together." One student who went to the retreat opined "“Being white means that when you’re just as poor as the next minority who (lives) paycheck to paycheck, you have the added bonus of being told you get to do it while subconsciously being racist. That’s white privilege. All of the blood, sweat, and tears that got me to where I’m at was actually privilege pouring out of me.”".