Given further data on the stunning performances of AlphaZero, Charles Murray asked me that on Twitter. And for now the answer surely seems to be yes: just let AlphaZero rip, and keep the human at bay. It’s a bit like the joke about the factory: “The dog is there to keep the man away from the machines, and the man is there to guard the dog.” (Or is it the other way around?)
But here’s the thing: right now there is only one AlphaZero, and AlphaZero does not play like God (I think). At some point there will be more projects of this kind, and they will not always agree as to what is the best chess move. Re-enter the human! Imagine a human turning on AlphaZero and five other such programs, seeing where they disagree, and then querying the programs further to find a better answer. It is at least possible (though not necessary) that a human will be better at doing this than will a machine.
Keep in mind, the original role in the human in Advanced [man-machine] Chess was not to substitute human chess judgment for machine chess judgment in any kind of discretionary fashion. It was to adjudicate disagreements across programs: “Rybka has a slightly better opening book. Fritz is better in closed endgames. Houdini is tops at defense.” And so on. The human then sided with one engine over the others, or simply spent more engine time investigating some options rather than others.
It could possibly run the same way for neural net methods, once we have a general sense of the strengths and weaknesses of different projects. So yes, man-machine cooperation in chess is a loser right now, but it may well come back. And there is a broader economic lesson in that, namely that automation may eliminate jobs, but it does not necessarily eliminate them permanently.
I saw your post about whether the 12th game draw was wise or not, but I haven’t seen this bit so far – I’m curious what you think the 12 draws mean for the future of classical chess? Have we hit the point where the very best in classical will just resign themselves to draws? Should we look to blitz or Chess960 to determine the very best?
It is now 24 world championship games in a row, spread out over two contests, with only two decisive results. Games between top grandmasters don’t end in draws nearly so often, so something is wrong with the incentives! The most common claim you hear is that in a 12-game match it is “too hard to come back from a loss,” so the players don’t take enough chances. That to me seems under-argued from a “maximize expected value backwards induction” point of view (a given move either boosts your expected value from the game or it doesn’t), but in any case there does seem to be a problem. (Too much advance preparation of openings?) On top of that, people are upset that two “classical time control” world championships in a row have been decided by the Rapid tiebreaker.
My first suggestion is to extend the matches to 24 games, but in the event of a tie at the end leave the reigning world champion with the title. That avoids the arbitrariness of any tie-breaking method, places what is to me a justified burden on the challenger, and seems to be enough games to prevent the reigning champion from simply stonewalling with a long series of draws. And there is plenty of precedent in chess history for matches that long, was it not nice when the Soviets paid for everything?
That said, I fear that venue costs are too high, the length of the match too variable (try booking a top hall under such conditions), and the drawing out of play would make the match harder to market to corporate and other sponsors, who are more interested in concentrated media attention (“In the future, every contender will be famous for fifteen chess games.”)
Chess960 games I find ugly, counterintuitive, and hard to follow.
So how about this? Have the openings in each game — say the first eight to ten moves — be chosen randomly, but out of a set of high quality but somewhat riskier than average alternatives (no Petroff!). This would limit the ability of players to choose intrinsically drawish lines with Black. It also would steer the games away from paths where both players know the main lines thirty moves deep or more, which of course is boring and also conducive to lots of draws.
I would note that many computer vs. computer matches already are played with such a method, and it does seem to make those games more dynamic.
I don’t doubt this method might cause top players to invest all the more time in preparing openings, to avoid being caught entirely off guard (everyone would end up knowing at least something about the Poisoned Pawn Sicilian). Still, there are limits to total prep, and the games would end up as more exciting, and probably more decisive, whether the players like it or not.
Let’s do it, and limit the impact of this insane arms race in opening preparation.
From an email I just sent:
My view was this: if you play out the main computer lines for 8-10 moves, Black’s position does not really improve, nor are White’s holding moves hard to find (he just has to shuffle back and forth).
Black does not have the structural advantage to enable a later transposition into a favorable endgame.
So it actually is a draw! (sort of)
Does the agreement[to draw] have to be non-Bayesian? There is a “vague range,” and by Magnus Carlsen offering the draw maybe he, as Kasparov suggested, lowered his own chances for the rapid tiebreak (shows some loss of nerve), until they were in the same “vague range” as the game 12 final setting.
So Magnus is saying “I don’t have a way of pressing that is better than my chances in the rapid tiebreak,” and Caruana is agreeing and knows that Magnus knows this.
Maybe not “strictly” Bayesian, but it doesn’t seem crazy to me either.
I thank S. for a relevant conversation on these points.
Those are the topics of my latest Bloomberg column, here is one excerpt:
It turns out that chess is oddly well-suited for a high-tech world. Chess does not make for gripping television, but the option of live viewing online, supplemented by computer analysis or personal commentary, has driven a renaissance of the game.
For one thing, computer evaluations have made watching more intelligible. Even if you barely understand chess, you can quickly get a sense of the state of play with the frequently changing numerical evaluations (“+ 2.00,” for instance, means white has a decisive advantage, whereas “0.00” signals an even position). You also can see, with each move, whether the player will choose what the computer finds best.
In essence, some of the suspenseful stupidities of low-level video games have been infused into eggheady chess. You can indulge your inner Pac Man without feeling guilty about it.
At first it was thought that online viewers would favor rapid and blitz chess, which are (as you might expect) more fast-paced. In fact, the slower games, including contests of five hours or more, have not put viewers off. If you are sitting at your office desk, you might wish to glance at the position every few minutes or so. A slower game means you can do that without missing much of the action, and yet still most of your work will get done. If the game is heading to a climax, you can pay full attention for that short period.
Fortunately, the software programs that evaluate the games and players are not yet infallible. So if Stockfish (one such program) indicates that your favorite player is far behind, you can hold out a slim hope that the software is wrong. “Creating artificial suspense” is one of the killer apps of the internet.
There is much more, including a discussion of basketball and trash talking, do read the whole thing.
I am arrived in Baku! Here goes:
1. Chess player: Garry Kasparov. Maybe the greatest player of all time? He is not ethnic Azerbaijani, but grew up in Baku.
Teimour Radjabov. It is amazing for how long he has gotten away with playing the King’s Indian Defense at the highest levels of chess competition.
Shakhriyar Mamedyarov. Over the last year, he has had the best results of anyone in the chess world, including Carlsen. His forcing style resembles that of Kasparov.
Vugar Gashimov. He was pretty good too, passed away prematurely in 2014.
Cellist and conductor: Mstislav Rostropovich, born in Baku. His Bach Cello Suites are perhaps my favorite of all extant recordings. Here is one (different) YouTube version. As a conductor he was uneven, but capable of spectacular live performances of Shostakovich.
Philosopher: Max Black, also born in Baku. He edited Frege and worked on problems from Leibniz, such as the identity of indiscernibles.
Note that numbers 1, 5, 6, and 7 on this list were Jews who emigrated to America.
Namely the fear of owing other people, or institutions, a favor, or maybe just the possible perception of such?:
The researchers believe reciprocity anxiety is likely to be greater the bigger a favour and the more public its receipt. They think it’s a trait that companies should take an interest in – while loyalty schemes, vouchers and other freebies have obvious appeal to many customers, results from two initial studies suggested that these marketing strategies are actually likely to deter others…
In a follow-up study, volunteers imagined a shop attendant offering them a free drink and plate full of snacks. Afterwards, high scorers in reciprocity anxiety scored lower for customer satisfaction and they said they would be less willing to visit the store again and less willing to spread a good word about the shop.
“Reciprocity works to establish a psychological bond” between customer and firm, the researchers said, but the discomfort it causes can backfire among those high in reciprocity anxiety, especially if they feel the benefits reflect badly on them or that they will struggle to reciprocate (around 18 per cent of people tested in these new studies scored highly in the trait; age and gender were unrelated).
…I wonder how it might impact the ways that people manage their friendships and other relationships – perhaps high scorers in reciprocity anxiety are inclined to turn down invitations, seek help or receive other friendly favours, putting them at risk of loneliness and isolation.
I have just spent a week in Beijing talking to officials and intellectuals, many of whom are awed by his skill as a strategist and tactician…He [Yafei] worries that strategic competition has become the new normal and says that “trade wars are just the tip of the iceberg”.
…In Chinese eyes, Mr Trump’s response is a form of “creative destruction”. He is systematically destroying the existing institutions — from the World Trade Organization and the North American Free Trade Agreement to Nato and the Iran nuclear deal — as a first step towards renegotiating the world order on terms more favourable to Washington. Once the order is destroyed, the Chinese elite believes, Mr Trump will move to stage two: renegotiating America’s relationship with other powers. Because the US is still the most powerful country in the world, it will be able to negotiate with other countries from a position of strength if it deals with them one at a time rather than through multilateral institutions that empower the weak at the expense of the strong…
My interlocutors say that Mr Trump is the US first president for more than 40 years to bash China on three fronts simultaneously: trade, military and ideology. They describe him as a master tactician, focusing on one issue at a time, and extracting as many concessions as he can. They speak of the skillful way Mr Trump has treated President Xi Jinping. “Look at how he handled North Korea,” one says. “He got Xi Jinping to agree to UN sanctions [half a dozen] times, creating an economic stranglehold on the country. China almost turned North Korea into a sworn enemy of the country.” But they also see him as a strategist, willing to declare a truce in each area when there are no more concessions to be had, and then start again with a new front.
That is highly speculative, to say the least. And perhaps you should not be happy if China sees your strategy as strong, since China itself generally does a poor job cultivating allies and also undervalues them. In any case, that is from Mark Leonard at the FT.
For the time being, we have turned off comments on MR posts. Is not a higher gdp a good thing?
A strata on Vancouver Island is experiencing a backlash after passing a bylaw last week banning outdoor play — a rule that is not unusual but goes further than most, according to the Condominium Home Owners’ Association of B.C.
In B.C., ownership in condominiums, apartments or townhouses sharing common areas is often purchased through an owners’ corporation under a strata title. The owners elect a council that sets policy for the strata.
Artisan Gardens, a neighbourhood development in Chemainus, about 80 kilometres north of Victoria, voted 15-4 in favour of adopting a bylaw that prohibits using the roadway “for play, including hockey, baseball, basketball, skateboarding, chalk artistry, bicycling or other sports and recreational activities.”
I gave a talk yesterday, and did not have time to get to this question, from Eric S., which we discussed during the dinner hour:
Who could launch themselves higher on a trampoline? LeBron James or Simone Biles?
She is a world class female gymnast, and much lighter (and less strong) than LeBron. The question is assuming that both parties are motivated to win the competition, and have sufficient time to train to achieve their maximum potential in the contest.
Ultimately I settled on Simone as the better answer, mumbling something about small ants being very powerful for their size, and that magnification and extension of muscle spans ends up producing problematic results. The power gain from the extra weight might be more than offset by the “drag” loss on the way up. But I genuinely do not know. Your view?
Addendum: This is an interesting article on animals and elastic springs. And Jason Kottke adds comment, amazing photo too.
No, they are negotiating, read their latest statement, it is full of “ifs.” If you are negotiating, especially in a fraught situation, often you will feel the need to walk away from the talks, or at least threaten to do so. (Of course, many people suggest Obama should have done more of this leading up to the Iran deal.) And Kim doesn’t want to enter the talks with Trump having had an unbroken string of PR successes.
Now, there is a perfectly reasonable argument for being pessimistic about the North Korean nuclear talks, namely that some of the demands of the two sides may prove incompatible. The good news, if you would call it that, is that we are not actually calling for complete denuclearization of North Korea, though nonetheless we may require more than they are willing to cede. Most of all, we want them to start acting like a normal evil government, rather than like a crazy evil government. Maybe that is too hard for Kim to pull off and still feel stable.
Still, the new news isn’t really bad news at all. It is how an evil tyrannical government negotiates. It is also how some non-evil tyrannical governments negotiate as well, not to mention non-evil, non-tyrannical governments too.
The survival rates of GMs at 30 and 60 years since GM title achievement were 87% and 15%, respectively. The life expectancy of GMs at the age of 30 years (which is near the average age when they attained a GM title) was 53.6 ([95% CI]: 47.7–58.5) years, which is significantly greater than the overall weighted mean life expectancy of 45.9 years for the general population. Compared to Eastern Europe, GMs in North America (HR [95% CI]: 0.51 [0.29–0.88]) and Western Europe (HR [95% CI]: 0.53 [0.34–0.83]) had a longer lifespan. The RS analysis showed that both GMs and OMs had a significant survival advantage over the general population, and there was no statistically significant difference in the RS of GMs (RS [95% CI]: 1.14 [1.08–1.20]) compared to OMs: (RS [95% CI]: 1.09 [1.07–1.11]) at 30 years.
Elite chess players live longer than the general population and have a similar survival advantage to elite competitors in physical sports.
That is from An Tran-Duy, David C. Smerdon, and Philip M. Clarke, via a loyal MR reader.
That is the topic of my latest Bloomberg column, here is one excerpt:
First: The North Korean regime has never been bureaucratized in the modern sense of that term. While we don’t have comprehensive information, it seems that until recently Kim as leader had not been going abroad, nor had he been receiving many visits from other heads of state. His position and perhaps his mood has been one of extreme isolation, and he is not surrounded by anything resembling the U.S. State Department or even the old-style Soviet bureaucracies that managed foreign policy for the USSR. The rest of his regime is probably poorly informed about the extent of American military superiority, should a conflict come to pass.
By meeting with other foreign leaders, the North Korean regime would be forced to build up its basic processes for dealing with the rest of the world. That in turn creates interest groups and flows of information (some of which invariably leak out). The North Korean populace responds by thinking more about the outside world, making it harder to control by propaganda. In turn the North Korean leadership may decide to continue economic liberalization.
One need not count on an “End of History” story culminating in liberalism and democratization. The more modest hope would be for the North Korean leadership to become more decentralized, more bureaucratic, better informed and harder to marshal behind crazy military measures.
The unspoken goal of engagement would be to encourage North Korea to evolve into a more banal and more predictable form. That is the natural flow of most bureaucratic organizations, so in this regard American negotiators actually have time on their side. The North Koreans are going to change a lot more than the U.S. is likely to.
And the concluding sentence:
Think of any diplomatic talks with North Korea as a big act of theater — designed not to fool him, but to teach him that theater itself can be fun.
Do read the whole thing.
That is the topic of my latest Bloomberg column, here is one excerpt:
Some historical events are relatively easy to model with game theory: the Cuban Missile crisis, many of the Cold War proxy wars, the crisis over North Korean nuclear weapons. In those conflicts, the number of relevant parties is small and each typically has some degree of internal cohesion.
To find a situation comparable to the Middle East today, with so many involved countries, and so many interrelationships between internal and external political issues, one has to go back to the First World War, not an entirely comforting thought.
The situation right before that war had many distinct yet related moving parts, including the dissolution of the Ottoman Empire, the imperialist scramble for colonies, the prior Balkan Wars, a rising Germany seeking parity or superiority with Great Britain, an unstable alliance system, an unworkable Austro-Hungarian Empire, and the complex internal politics of Russia, which eventually led to the Bolshevik Revolution.
What do we learn from the history of that time? Well, even if the chance of war was high by early 1914, it was far from obvious that the Central Powers attack on France, Belgium and Russia would be set off by a political assassination in the Balkans.
Nonetheless, in sufficiently complex situations, chain reactions can cause small events to cascade into big changes. In World War I, one goal behind the assassination of Archduke Franz Ferdinand was to break off parts of the Austro-Hungarian Empire into a new Yugoslavia. The empire responded by making some demands on Serbia, which were not heeded, a declaration of war followed, and the alliance system activated broader conflicts across Europe.
If you don’t quite follow how a single assassination, which was not even seen as so important the day it occurred, triggered the death of so many millions, and the destruction of so much of Europe, that is exactly the point. When there is no clear way for observers to model the situation, a single bad event can take on a very large significance and for reasons that are not entirely explicable.
Do read the whole thing.
That is the topic of my latest Bloomberg column, they chose a nice photo. Here is an excerpt:
If North Korea and the U.S. simply talk, and nothing comes of it, that raises the status of Kim Jong Un, who then would keep improving his missiles anyway. So if the U.S. goes ahead with the talks, you might rationally infer that the risk of war has gone up. Furthermore, there is the risk that Trump or Kim could feel humiliated by a summit that yielded nothing, again raising the chance of war or miscalculation leading to disaster.
In chess there is a concept known as “zugzwang,” or “compulsion to move.” It’s used to describe the position of a player with no good options who would prefer to do nothing at all. That’s not a possibility in chess, so the unfortunate player faces a situation in which all roads involve a deterioration of his position.
There is much more at the link, only one scenario being even somewhat optimistic.