Friday assorted links


4. "Yes, Brexit is a muddle which nobody will be happy with, until the UK decides if it really would rather remain or become a free-market beacon on the edge of the continent. But do not judge democracy on it. Democracy's errors as the mechanism for collective decision-making capacity have been far worse. And then there are the failures of all the other options. "

You betcha!

More insightful comments from Rich "da fuccboi" Berger.

Hi, Neat post. There's a problem with your site in web explorer, might test this?
IE nonetheleds is the market leader and a big part of other people
wiill miss your excellent writing because of ths problem.

4. That was a fun take, but the fact that America is not GDP optimizing it's not something we were supposed to talk about this year, right?

I believe the official theme is build the wall or shut down the government.

What I don't get is Mexico was supposed to pay for it. What happened with that? Why do I have to pay? I'm already paying for it. My lawn boy disappeared.

3. If a coherent theory of quantum gravity emerges, will we be in any better position to determine whether the vast realm of baryonic matter is the colorful stage where sets and costumes change constantly courtesy of the stagehands Dark Energy and Dark Matter--or whether our vast realm of baryonic matter is little more than (or nothing but) an adjunct or appendix (a vaudeville sideshow, to maintain the metaphor briefly) to the much more vast realms of Dark Energy and Dark Matter. (We tend to see black holes and gravitational anomalies as "exits from the universe"--whereas if unknown and actual intelligences operate within the realms of dark energy and dark matter, these portals acquire some other significance. [Why doesn't some philanthropic someone tell us plainly?])

"3. If a coherent theory of quantum gravity emerges ...": whatever happened to Strings?

I once turned on the telly to watch what had been billed as a documentary involving a Cambridge lecture on Strings. It turned out to show a college don and a couple of freshers discussing ... tension in strings.

They're still (after many decades) to show testable predictions. Strings have had some set-backs recently, and no successes. As Kenny said: You've got to know when to hold 'em, know when to fold 'em... For lack of progress, we're at the "Should I hold 'em or fold 'em?" stage for many theoreticians. Or perhaps it's just that we're in the "wandering in the desert" years and soon the Promised Land will appear. no way to know now.

And some of us don’t even believe in black holes. The equations for water swirling down a drain shows that as water speeds up closer to the center and as it appears directly over the center it’s speed does to infinity. So what happens in the real world? There is no water there, instead it’s air.

Where asymptotes appear in math, they do not appear in real life. No reason to think that black holes will not be the same way. My conception is perhaps a shell slightly larger than the event horizon, which radiates heat enough to keep itself from collapsing inward. Much as the sun is a the combination of nuclear outwards and gravity inwards.

I learnt recently (sorry, no citation) that Eddington discussed the idea of a black hole - in writing, I think - and then dismissed it as absurd. So who was the first chap not to dismiss it, and had he known that Eddington had had the idea earlier? And - does it matter? Does history of science ever matter?

4. Cochrane: "A team from Cato and Hoover could probably raise GDP by 20 percent inside a year. If anyone would pay the slightest attention to us." Is Cochrane being ironic? Is he the frustrated child nobody listens to? Why would he write such a thing? A team from Cato and Hoover would just as likely send us into another great depression. One will recall that, during the depths of the great recession, Cochrane was arguing for higher interest rates, claiming that higher interest rates are expansionary. His logic? Higher interest rates would signal that the Fed believed the economy would grow and, thus, induce investors to invest and consumers to consume. Signaling, now that's a recurring theme at this blog. People, including investors, will believe just about anything.

Funny that conservatives worship at the altar of Hoover. He who invented Hoovervilles and the Great Depression. Then again little Bush gave us the Great Recession and bailouts galore.

"... Bush ... recession ..."

Wrong! Dead wrong!

Fannie Mae and Freddie Mac and a bunch of corrupt lenders and ninja borrowers gave us that recession. The main culprits were the Democrats that blocked all attempts to restrain Fannie and Freddie. Jeesh!

Got a link?

I do.

Yeah, because all links are pointers to truth.

Get real.

The financial crisis certainly spawned a lot of he said she said arguments, but the nice thing about that link above is that it deconstructs the lie that you are pushing right here.

Good points. "Why would he write such a thing?"

Arrogance? Hubris?

The technocrats - I don't want to call the elites anymore, it gives them too much credit - believe everything would be great if we would just let them run things.

Not gonna happen.

I meant, "don't want to call them elites". Grrrrr

#2 That is terrifying.

#4 I believe the same argument he makes for democracy (as in the Churchill statement he quoted) equally applies to capitalism. I've always enjoyed this cartoon which I feel captures the same deal with complex problems with the best government you can.

#6 Now to find out if his use of twitter either proves or denies that he's effectively become a Cuban govt. stooge. Denial doesn't mean the account disappears, it just means that it will suddenly feel ghost-written somehow. Regardless, I'm sure Twitter will assign a blue-check-mark...whatever the case.

Reinforcement Learning is really cool.

The basic setting is trying to find a policy that optimizes reward in the long run.

A policy is just a black box that takes as input a "state" and yields an "action" that should optimize reward.

In technical terms you have a set of states, a set of actions, and a "Transition Function" that takes a state, an action, and yields an immediate reward as well as a subsequent state. You can think of the 'rules of the game' being contained in the transition function.

In the end, what you want is a policy that takes as input a state and gives you your best possible action -- the "optimal policy".

In simple settings, an algorithm can essentially enumerate all steps and keep a table of the average reward it got after visiting that state. This is great for toy problems, but many problems, like chess, have far too many possible states to make any kind of tabular approach tractable. Other problems might have continuous state spaces, which makes keeping tabs even more complicated.

That said, it's not too difficult to find optimal policies in a given setting (supposing the problem is well defined). If I want to make an RL algorithm to play lunar lander, I can use a common approach (Q-learning), define the state as being the x,y position combined with the movement of the lander, and I'll let the policy choose from the three or four valid actions in the game. Things will work, and (thanks to some math magic) in finite time I'll have a self-piloting lunar lander. What won't work is if I try to get that learning architecture to play any other game -- the state, what constitutes winning, what actions exist are "hard coded" into the solution.

What DeepMind did initially was advance RL in two important ways. First, they developed an architecture that could take as input simply an image of a screen -- not an x,y coordinate pair read from memory, but exaclty the same information you have (n.b. I think they did give the machine access to "score") and learn how to play *any* atari game. Instead of imparting the model with "domain knowledge", like "Going down is good in Lunar Lander" they simply let their models learn everything they needed to independently of any human input. This isn't any different from letting a 6 year old loose in an arcade. The second, more technical thing they did was skip the tabular approach and use a Neural Network architecture to approximate the State->Action function. This is nice because in the real world state spaces get obnoxiously large (and aren't really discrete), so keeping a table of the expected value of an action given a state is not really a solution. They also used "memory replay" and some other hacks to essentially "make it work".

So that's what got a company of 50-75 employees bought by Google for 500 million dollars.

Now some months ago DeepMind famously "Solved" Go in a highly publicized match, and they've probably "Solved" chess, and shogi -- all highly complex games with enormous state spaces. What's important and new about AlphaZero is that it represents *one architecture* that can solve all three. No need for code and parameters fine tuned for thinking 15 moves ahead in Go, according to the rules of Go, when you have AlphaZero able to think 15 moves ahead in all of these games -- without any tinkering under the hood. What they're ultimately working toward is algorithms that can understand, in a totally general sense, the concept of choosing actions to bring about good outcomes in practically any setting.

So, ultimately what I'm trying to say is, I for one welcome our robot overlords.

5. Trends in economics. And what do economists actually do?

would someone please summarize those two links/content into plain language....

Kleven measures word counts in economics working papers about government policy. The main topics in 1980 were almost all about taxes. Now it's mainly about evaluating quality of government spending. To do this credibly, economists have to exclude confounding factors by using better techniques to identify true underlying effects rather than, e.g. self-selection. You can't do this with aggregate data, so you need records about individuals, usually before and after the experiment: governments often compile these as administrative data.

Bandiera is presenting a personal view that most economists do what she does: technical evaluation of government policies, mainly for poor people, not money or finance. This is based just on titles of academic articles, suggesting that salesmanship is more important here. It's probably what most academic economists believe, but seems also to be a lack of curiosity.

(Bandiera is making these assertions about applied micro, by the way, which is largely the study of individual transactions. The arguments get more tendentious as one goes on. We are asked to believe that poverty is not about money, and that so-called behavioral approaches to economics are dominant, where you program people's behaviour by silly rules instead of attempts to improve their position. Neither is true!)

5. Today most economists are basically programmers who run computer models about social phenomena (that is, most economists are computerized social scientists) and they can be divided into two groups: R and Stata people, usually with degrees from low ranked PhDs and Python and Julia people, who got their degree from higher ranked PhD.

People like Adam Smith, Keynes, Hayek and even Buchanan wouldn't be regarded as economists today. I guess they would be considered some kind of social philosophers that like to talk about stuff involving money.

#1: Wonder where this ability to learn games "in a distinctive, unorthodox, yet creative and dynamic playing style" that are often considered surrogates for war is eventually going?


#4: Democracy does not give us speedy technocratically optimal solutions to complex questions revolving around 2 percentage points of GDP. Democracy, and US democracy in particular, serves one great purpose -- to guard against tyranny.

Well said, JC.

Then you would no doubt oppose a "democratic" leader who is "undisciplined, doesn’t like to read’ and tries to do illegal things?

Not the biggest Trump fan, no.

Oh. My. God. It descends beyond parody.

Mike Pompeo is doing a great job, I am very proud of him. His predecessor, Rex Tillerson, didn’t have the mental capacity needed. He was dumb as a rock and I couldn’t get rid of him fast enough. He was lazy as hell. Now it is a whole new ballgame, great spirit at State!

That, ladies and gentlemen is an authentic communication from the President of the United States.

Either that or the President's Twitter has finally been hacked.

And if you use this service, it will look exactly as what it is, according to the Trump Administration -

A nice citizen doing the work that the White House apparently is no longer capable of doing on its.

... own.

I can only say, WTF? Mr. Tillerson is as much as conservative and pro-business as you can be. So, what's Trump plan beyond being a sea cucumber?

'a "democratic" leader who is undisciplined, doesn’t like to read and tries to do illegal things?' Doesn't a fair part of that apply to FDR?

I feel a need to pat your little head.

"When Einstein visited the White House, his English was not all that good, so FDR switched to German and, according to Einstein, spoke the language rather well. FDR also once held his own in a shouting match in French with de Gaulle. FDR, whilst at Groton, had won the school’s Latin prize.."

Sure, sure, we are in that boat again. We have a President who only excelled at Latin, speaks French and German, demonstrating that he doesn't read enough English!

What a fascinating exercise in answering a different question. Do you have a natural gift for the irrelevant, or have you had to work at it?

(Pats head again.)

A Latin prize should answer both studiousness and discipline. On the crime, I don't remember that 30 indictments were brought there or then.

Jeff R said, " Well said, JC".

JC didn't say that, Cochrane did!

Here's what Cochrane said:

"Democracy does not give us speedy technocratically optimal solutions to complex questions revolving around 2 percentage points of GDP. Democracy, and US democracy in particular, serves one great purpose -- to guard against tyranny. That's what the US colonists were upset about, not the fine points of tariff treaties. US and UK Democracy, when paired with the complex web of checks and balances and rule of law protections and constitutions and so forth, has been pretty good at throwing the bums out before they get too big for their britches. At least it has done so better than any other system. "

Roger that!

#1. Wow. I feel like the characters in a sci-fi movie when they're going about their daily business and suddenly a huge black shadow blots out the sky. Looking up they see it's a spaceship of enormous diameter. I'd love to see Stockfish be replaced with another alphazero after the first couple of moves. Is the 3 pawn (!!!) sacrifice really a good idea playing an equally skilled opponent? And then there's all the "wasted" moves as the white king scampers into the corner. Hmm. Makes me want to build the engine just to game out that one game's choices.
#3. One red flag (although in this case I find it difficult to believe the author is this unqualified, but I didn't check her credentials) is her use of the term black hole as being different from the volume of space-time inside the event horizon. I see no reason to distinguish between the volume bordered by the event horizon and the black hole's "interior", they are one and the same. My theory of black hole interiors is that that's where the fairies and hobgoblins live. Prove me wrong. (Point is, the interior of a black hole (the volume inside the EH) is not "part" of our Observable Universe, i.e. it is not "physically real" because theoretically it can not ever be observed. Some physicists, like Susskind, have swallowed the Kool-Aid when they publish fanciful speculations on BH interiors - unless they can establish that these internal workings can have some exterior effects, but they can't. (Although I should never say never, it's a good bet that any demonstration of exterior effects from interior phenomena would be Nobel Prize worthy (and possibly be the "next" revolution in our understanding of our reality).

The democracy of 2018 AD isn't the same as the democracy of 425 BC. Maybe the newer version needs a different name so they can be told apart.

Good idea, we'll just call that one "the better one".

"direct democracy" and "indirect (or representative) democracy" already exist. What's wrong with them?

Democracies in 425 BC were all run by pedophiles. Can't have that today although the Alabama GOP almost put one into office.

#2 is just above a story about cow vigilantes, which makes me think there are potholes in the road ahead, for uniculturalism. And I don't even particularly like hamburgers.
#Remember, Remember
# Make them pay!

#2 A very unexpected result coming out of a black box -- strongly consider whether there's an error in the algorithm or an error in the data being analyzed or a change in the way data is being reported...

3. New claims about black holes.

I had this theory down pat for years ago. We are not expanding, we are compacting in place. The new cosmology theory.

My construction of the theory: Avogradros's number is mathematically defined according to the rank of the complexity graph. Five color universes inside black holes get larger avogadro when packing sphere. Outside we are three color as in the quarks. Most of the constants of physics are derivable from the mathematics of compaction under quantization. I remember talking about this right here on this blog. The fix is simle, under a maximizing entropy theory, any theory of operation is finite and can be warped into any of the theory via a unique translation, dimensionality of complexity of the theory is preserved, there is only one form of the theory when entropy is maximized. So there must be a pure mathematical solution, a theory of compaction according to complexity. If blackholes are doing what they do, expand apparently, then the theory of operation inside must be increasing in rank, they must therefore have a larger avagadros' which should be derivable. Krugman's trade theory is of the same sort, maximum entropy packing of the trade channel as a constricted, quantized flow. Hawking's nailed is, no flow no quantization. The same theory holds for the singularity, it will compact the semantic graph as more data pressure is incurred, jumping in complexity rank; the singularity effect. That is why the term is useful in trade, AI and physicists.

#4..."A team from Cato and Hoover could probably raise GDP by 20 percent inside a year. If anyone would pay the slightest attention to us."

Or raise the dead if paying really close attention. Sadly for you, the Mercatus Center has assured me of a 30% return and I don't have to pay attention at all.

Cochran is persuasive on this one. That column wasn't one of Tyler's better ones.

#2 the faster they grow they get closer to sustainable equilibrium and stop growing. great news =)

1. I was super-enthusiast for alpha zero when it did its first games one year ago -- and now I ams till enthusiast but a little skeptical as well. Sure the linked game is beautiful, though I think not a much as the ones I had seen one year ago. But more importantly, can we honestly say that alphazero is, at this time, better than Stockfish 8? In the video, it is explained that stockfish 8 was deprived of its opening library -- alpha zero has not, by design. But Stockfish 8 is designed to work with an opening library (and am end-game library, etc.) while alpha zero is not.

In a tournament, organized with the usual rules of computer-chess tournament, will alpha zero be the first? I'd like to know either your informed opinions or your best guesses..

#4, Cochrane's is an argument for democracy, but far from the only one or the best one. And the arguments are voluminous; natural right, democratic peace theory, peaceful concession and succession as an alternative to destructive conflict, tax morale; all rather plausible, and more.

Seeing Cochrane launch into a defense of democracy, I'm left wondering what the point of TC's original column was though. How exactly is Brexit a problem that is particularly structurally like "climate change", "boosting innovation" or "improving education"? How exactly have we in detail actually got from discussing how to implement a referendum result (which is frankly rather simple, were it not for the Irish Border), to questioning democracy itself?

We have to keep everything in perspective. The lady who was dumb enough to think she could vote with a green card should get eight years in prison. The president who was dumb enough to collude with Russian spies should just walk, to a fully taxpayer funded retirement.

You can tell just by looking at them.

#2 Predictions are difficult, especially ...

Comments for this post are closed