Pessimism of the Intellect, Optimism of the Will     Archive   Favorite posts   Twitter: @steve_hsu

Tuesday, January 29, 2013

US income distribution

Where is the middle class in this figure? If we eliminate young and old people, does the peak shift to the $30k range, or are things bleaker than I had thought? [ See links in comments for better figures. ]

Wikipedia: ... Household income in the United States varies substantially with the age of the person who heads the household. Overall, the median household income increased with the age of householder until retirement age when household income started to decline.[25] The highest median household income was found among households headed by working baby-boomers.[25]

Households headed by persons between the ages of 45 and 54 had a median household income of $61,111 and a mean household income of $77,634. The median income per member of household for this particular group was $27,924. The highest median income per member of household was among those between the ages of 54 and 64 with $30,544 [The reason this figure is lower than the next group is because Pensions and Social Security add to income while a portion of older individuals also have work-related income.]).[25]

The group with the second highest median household income, were households headed by persons between the ages 35 and 44 with a median income of $56,785, followed by those in the age group between 55 and 64 with $50,400. Not surprisingly the lowest income group was composed of those households headed by individuals younger than 24, followed by those headed by persons over the age of 75. Overall, households headed by persons above the age of seventy-five had a median household income of $20,467 with the median household income per member of household being $18,645. These figures support the general assumption that median household income as well as the median income per member of household peaked among those households headed by middle aged persons, increasing with the age of the householder and the size of the household until the householder reaches the age of 64.

Iceland: Let banks go bankrupt

A misallocation of human capital: @2m4s.

See also The illusion of skill.

Sunday, January 27, 2013

Big data from a big eye in the sky

From NOVA Rise of the Drones.

ARGUS ("Wide Area Persistent Stare"): 1.8 Gigapixels, built using off the shelf components: 368 cellphone cameras, 5 MP each. Can surveil 15 square miles simultaneously from 17,000 ft, with 6 inch resolution of objects on the ground. Generates 1E6 terabytes of data per day. This system has been available for a couple of years already.

Watch What Drones Can See on PBS. See more from NOVA.

Thursday, January 24, 2013

Learn to solve every problem that has been solved

Feynman had that on his final blackboard. Crazy? Even for Feynman? An admirable ambition, nonetheless.

At what point did this become impossible for even the smartest human alive? What if we amend it to Learn to solve every important problem that has been solved? (For some threshold of importance...)

Feynman's TO LEARN list:

Bethe Ansatz, Kondo Effect, 2-D Hall Effect, "accel. temp" = Unruh Effect?, Non-linear classical Hydrodynamics

Do I know anyone well-acquainted with all of these topics? I can think of a few people who come close ...

While it may be impossible to achieve Feynman's goal, I'm surprised that more people don't attempt the importance threshold-modified version. Suppose we set the importance bar really, really high: what are the most important results that everyone should try to understand? Here's a very biased partial list: basic physics and mathematics (e.g., to the level of the Feynman Lectures); quantitative theory of genetics and evolution; information, entropy and probability; basic ideas about logic and computation (Godel and Turing?); ... What else? Dynamics of markets? Complex Systems? Psychometrics? Descriptive biology? Organic chemistry?

Wednesday, January 23, 2013

Wonders of science

More iPhone snapshots in aid of my long term memory :-)

Jeff Hawkins at a recent MSU workshop Cognitive, Evolutionary, and Computational Models of the Mind.

URA (Universities Research Association; oversees the operation of Fermilab) meeting in Washington, held in the Washington Post building.

Bill Foster, one of two physicists in Congress, addresses the meeting.

William Brinkman, Director of the Office of Science at DOE.

Pier Oddone, Fermilab Director.

The White House, day after the inaugural.

Robert Wilson

In 1969, when Wilson was in the hot seat testifying before the Congressional Joint Committee on Atomic Energy, Sen. John Pastore demanded to know how a multimillion-dollar particle accelerator [Fermilab] improved the security of the country. Wilson said the experimental physics machine had "nothing at all" to do with security, and the senator persisted.

"It has only to do," Wilson told the lawmakers, "with the respect with which we regard one another, the dignity of men, our love of culture. It has to do with: Are we good painters, good sculptors, great poets? I mean all the things we really venerate in our country and are patriotic about. It has nothing to do directly with defending our country except to make it worth defending."

Saturday, January 19, 2013

As flies to wanton boys are we to the gods

An earlier post, Discrete genetic modules can control complex behavior, described genetic control of burrowing behavior in deer mice. A reader commented that the results were entirely unsurprising. I wasn't aware of similar results in mammals, but of course this sort of thing has long been known in drosophila, thanks to Seymour Benzer and collaborators.

WSJ: ... When the great California Institute of Technology geneticist Seymour Benzer set out in the mid-1960s to find mutations in fruit flies that affected behavior, rather than mere anatomy, he was ridiculed for challenging the consensus that all behavior must be learned.

Benzer told the geneticist Max Delbrück about the plan to find behavioral mutants; Delbrück said it was impossible. To which Benzer replied: "But, Max, we found the gene, we've already done it!" (Benzer's mother was more succinct: "From this, you can make a living?") He was soon able to identify mutations related to hyperexcitability, learning, homosexuality and unusual circadian rhythms, like his own: Benzer was almost wholly nocturnal.

Since then, thanks to studies of human twins and a rash of genetic investigations in animals, it has become routinely accepted that most things, including personality, sexual orientation and intelligence, are to some degree affected by genes. The University of Virginia's Eric Turkheimer has declared what he calls the "first law of behavior genetics": that all human behavioral traits are heritable.

Benzer started in solid state physics, migrated to molecular biology, and then to neuroscience.
Caltech Oral History:  ... I had my nose on the transistor. It’s like Max Delbrück [professor of biology at Caltech; d. 1981] failed to discover fission, and he had it under his nose. [Laughter] I failed to discover the transistor, because I had three electrodes in there, and I was measuring things—using one to measure what the other one was doing—but I never had the idea of trying to use that arrangement as an amplifier. Instead, I had a different idea; I had the idea of making a crystal amplifier, but it was too sophisticated. It was based on putting a metal layer on top of a semiconductor and using a tunnel effect to control the current that’s passing through, but I never got it to work. Instead, the Bell Labs guy did the most simpleminded thing, which was to have just these two wires next to each other and have one influence the other. It escaped me, and it was under my nose. Some time later, there was a big demonstration of it at Bell Labs. These guys grabbed me and said, “You should have done this.” [Laughter] And they were right. But, you know, maybe to some extent, because I was already into biology at that time, I wasn’t really focused on that problem. Of course, being a graduate student and not being all that able or having big resources [played a role]. But by the time I got my Ph.D. in 1947, I was already interested in biology.

Aspaturian: What had happened?

Benzer: I was always interested in biology. But two things happened. One of the guys in the lab — his name was Lou [Louis L.] Boyarsky—told me about mapping genes on chromosomes, the work that had been done here at Caltech by [Alfred H.] Sturtevant and [Thomas Hunt] Morgan and their group. I thought that was very exciting. And then I read this book by [Erwin] Schrödinger, written around 1944, called What Is Life?, which inspired a number of other people as well—Francis Crick, for one. Max Delbrück was in the book—he had been at Caltech in the thirties, switching from physics to biology—and there’s a chapter in there on Delbrück’s model of mutation. ...

Aspaturian: What brought you to Caltech, the first time you came?

Benzer: During the sixties, I was getting more and more interested in behavior. One reason was my two children. I have two daughters with very different personalities. If you have one daughter, you don’t notice anything, but if you have a second one, you begin to wonder, “Are we doing things differently, or is it genetic?” So I got interested in this general problem of personality and behavior—how much is genetics and how much is environment? And how do you study such a problem? I had actually begun to be interested even before that time. There was a meeting about ’63, I think, at Cold Spring Harbor, where I remember having a conversation with Marshall Nirenberg. We had this feeling that all the molecular biology problems were on the verge of being solved. It was a little bit like the physicists at the end of the nineteenth century saying, “All we have left to do is one more decimal place.” Little did we anticipate all the recombinant DNA technology. So that was another part of it, the fact that molecular biology was going so well, becoming rather crowded. When things get to that stage, you wonder why you should be doing something somebody else is already doing. It’s just redundant. ...

Aspaturian: Would you say that Drosophila is about the most complex organism with which you can get really rigorous results in this kind of research?

Benzer: Well, I don’t know. It depends on what you want to study. You can get rigorous results with humans now. Modern technology makes it almost as easy to work with humans as with flies, and that’s why I have the courage to get into the human business now.

Aspaturian: But there are so many more behaviors to look at in humans.

Benzer: Humans are wonderful. There’s a book on viewing disorders of man, containing 4,000 hereditary disorders in humans, one or two thousand of which have been actually mapped on the chromosome. Many of these have behavioral components, and hundreds affect the eye. There’s a similar book on Drosophila. And we’re finding that more and more of the genes correspond to one another.
See also this video interview of Benzer. Among other things, he discusses specific mutations that control sexual behavior in drosophila (e.g., length of copulation, courtship), learning ability, memory, etc. Of course, these are just flies  ;-)

For more on Max Delbruck, see For the historians and the ladies; for more on physicists and early molecular biology, see The Eighth Day of Creation.

Thursday, January 17, 2013

US-China software arbitrage

Who says outsourcing doesn't work?  :-)

This is just a single anecdote, but it suggests that US software developers cost many times more than coders of similar quality in China ...
Verizon RISK team security blog: ... As it turns out, Bob had simply outsourced his own job to a Chinese consulting firm. Bob spent less that one fifth of his six-figure salary for a Chinese firm to do his job for him. Authentication was no problem, he physically FedExed his RSA token to China so that the third-party contractor could log-in under his credentials during the workday. It would appear that he was working an average 9 to 5 work day. Investigators checked his web browsing history, and that told the whole story.

A typical ‘work day’ for Bob looked like this:

9:00 a.m. – Arrive and surf Reddit for a couple of hours. Watch cat videos
11:30 a.m. – Take lunch
1:00 p.m. – Ebay time.
2:00 – ish p.m Facebook updates – LinkedIn
4:30 p.m. – End of day update e-mail to management.
5:00 p.m. – Go home

Evidence even suggested he had the same scam going across multiple companies in the area. All told, it looked like he earned several hundred thousand dollars a year, and only had to pay the Chinese consulting firm about fifty grand annually. The best part? Investigators had the opportunity to read through his performance reviews while working alongside HR. For the last several years in a row he received excellent remarks. His code was clean, well written, and submitted in a timely fashion. Quarter after quarter, his performance review noted him as the best developer in the building.

Wednesday, January 16, 2013

Discrete genetic modules can control complex behavior

"Horses ain't like people, man, they can't make themselves better than they're born. See, with a horse, it's all in the gene. It's the f#cking gene that does the running. The horse has got absolutely nothing to do with it." --- Paulie (Eric Roberts) in The Pope of Greenwich Village.

NYTimes: ... Dr. Hoekstra started with a species called the oldfield mouse (Peromyscus polionotus), the smallest of the deer mice. For 80 years or more, field scientists have documented its behavior, including excavating characteristically long burrows with an escape tunnel, which the mice will dig even after generations of breeding in cages in a laboratory.

Dr. Hoekstra treated tunnel length and architecture as a physical, measurable trait, much like tail length or weight, by filling burrows with foam that would produce a mold easily measured and catalogued – behavior made solid.

She and her students did this in the field and repeated it in the laboratory by putting the mice in large, sandbox-like enclosures, letting them burrow and then making molds of the burrows. They did the same with another deer mouse species, Peromyscus maniculatus, that digs short burrows without escape tunnels.

The team bred the two species together (they are close enough to interbreed) and measured the burrows of the offspring. Their tunnels showed a blend of parental characteristics, varying in length and with and without escape tunnels. Further breeding crosses between the hybrids and the original short-burrow species were conducted and the tunnels measured again.

Then the scientists matched variations in tunnel architecture to variations in DNA. What they found were three areas of DNA that contributed to determining tunnel length, and one area affecting whether or not the crossbred mice dug an escape tunnel. That was a separate behavior inherited on its own, so that the mice could produce tunnels of any length, with or without escape tunnels.

All complicated behaviors are affected by many things, Dr. Hoekstra said, so these regions of DNA do not determine tunnel architecture and length by themselves. But tunnel length is about 30 percent inherited, she said, and the three locations account for about half of that variation. The rest is determined by many tiny genetic effects. As for the one location that affected whether or not mice dug an escape tunnel, if a short-burrow mouse had the long-burrow DNA region, it was 40 percent more likely to dig a complete escape tunnel.

Both Dr. Anholt and Dr. Bargmann said that for complex behaviors, which can be affected in ways too small to measure by many other genes, the effects of these DNA locations were very significant.

Title and abstract of the Nature article:
Discrete genetic modules are responsible for complex burrow evolution in Peromyscus mice

Relative to morphological traits, we know little about how genetics influence the evolution of complex behavioural differences in nature1. It is unclear how the environment influences natural variation in heritable behaviour2, and whether complex behavioural differences evolve through few genetic changes, each affecting many aspects of behaviour, or through the accumulation of several genetic changes that, when combined, give rise to behavioural complexity3. Here we show that in nature, oldfield mice (Peromyscus polionotus) build complex burrows with long entrance and escape tunnels, and that burrow length is consistent across populations, although burrow depth varies with soil composition. This burrow architecture is in contrast with the small, simple burrows of its sister species, deer mice (P. maniculatus). When investigated under laboratory conditions, both species recapitulate their natural burrowing behaviour. Genetic crosses between the two species reveal that the derived burrows of oldfield mice are dominant and evolved through the addition of multiple genetic changes. In burrows built by first-generation backcross mice, entrance-tunnel length and the presence of an escape tunnel can be uncoupled, suggesting that these traits are modular. Quantitative trait locus analysis also indicates that tunnel length segregates as a complex trait, affected by at least three independent genetic regions, whereas the presence of an escape tunnel is associated with only a single locus. Together, these results suggest that complex behaviours—in this case, a classic ‘extended phenotype’4—can evolve through multiple genetic changes each affecting distinct behaviour modules.
Note the QTLs for tunnel length interact linearly (additively) -- no sign of epistasis.

Saturday, January 12, 2013

The good books

I did some unpacking today and assembled some favorite technical books (mostly physics) on one set of shelves. There are more that are still lost in boxes, but I think this is a decent collection. Surely no man can call himself educated without being familiar with the contents of a few of these books ;-)

In response to comments I'm posting some shelves of less technical books. The best way to characterize the collection below is that all of these are good, and some are among my favorites.

Click for larger images.

Low hanging fruit and technological innovation

Have we picked all the low hanging fruit? GDP growth may not be the same as growth in "utils" (units of utility, as in happiness or utility function), but it's a reasonable proxy. Click graph below for larger version.

The util return per unit of technological effort is probably decreasing as the problems left to be solved become more challenging. But it's hard to put a util value on some things that are in the foreseeable future, like machine intelligence and genetic engineering. A GDP value will be assigned by definition of commerce (the market), but actual utility is harder to understand, as we may be altering ourselves and our civilization along the way! Singularitarians would have you believe that the graph below will reach a point of divergence in the near future ...

Economist: ... For most of human history, growth in output and overall economic welfare has been slow and halting. Over the past two centuries, first in Britain, Europe and America, then elsewhere, it took off. In the 19th century growth in output per person—a useful general measure of an economy’s productivity, and a good guide to growth in incomes—accelerated steadily in Britain. By 1906 it was more than 1% a year. By the middle of the 20th century, real output per person in America was growing at a scorching 2.5% a year, a pace at which productivity and incomes double once a generation (see chart 2). More than a century of increasingly powerful and sophisticated machines were obviously a part of that story, as was the rising amount of fossil-fuel energy available to drive them.

But in the 1970s America’s growth in real output per person dropped from its post-second-world-war peak of over 3% a year to just over 2% a year. In the 2000s it tumbled below 1%. Output per worker per hour shows a similar pattern, according to Robert Gordon, an economist at Northwestern University: it is pretty good for most of the 20th century, then slumps in the 1970s. It bounced back between 1996 and 2004, but since 2004 the annual rate has fallen to 1.33%, which is as low as it was from 1972 to 1996. Mr Gordon muses that the past two centuries of economic growth might actually amount to just “one big wave” of dramatic change rather than a new era of uninterrupted progress, and that the world is returning to a regime in which growth is mostly of the extensive sort (see chart 3).

Tuesday, January 08, 2013

Cognitive, Evolutionary, and Computational Models of the Mind

Video of talks from the MSU workshop New Frontiers in Cognitive, Evolutionary, and Computational Models of the Mind. Compliments of BEACON:
The BEACON Center for the Study of Evolution in Action approaches evolution in an innovative way, bringing together biologists, computer scientists, and engineers to study evolution as it happens and apply this knowledge to solve real-world problems. BEACON is an NSF Science and Technology Center, headquartered at Michigan State University with partners at North Carolina A&T State University, University of Idaho, University of Texas at Austin, and University of Washington.
Slides from my opening remarks.

The next workshop in this series is January 15, and will include speakers such as:
Giulio Tononi (Wisconsin), Daniel Wagenaar (Caltech), Ken Stanley (Central Florida), Richard Lewis (Michigan), Jeff Hawkins (Numenta), Michael Hawrylycz (Allen Institute)
The videos from these presentation should also appear online.

Sunday, January 06, 2013

Sprints, interval training and energy expenditure

I've read studies in the past that found jogging or running at a moderate pace burns calories at a rate of about 100 calories per mile. This rate of energy expenditure depends on bodyweight, but only weakly on the actual running speed. Thus if you run, e.g., 2 miles you probably burned about 200 calories (depending on how big you are), whether you ran at 7 minute pace or 11 minute pace (i.e., you covered the distance in 14 minutes or 22 minutes).

However, from personal experience it seems that sprinting increases the rate of calorie consumption per unit distance significantly. The study below is the first I've seen showing this kind of nonlinear dependence of energy consumption as a function of level of exertion. Note, some of the 200 calories resulting from 2.5 minutes of sprinting may be consumed during post-exercise recovery, due to elevated metabolism.

See also Tabata or High Intensity Interval Training (HIIT).
American Physiological Society : ... the men then checked in to a research facility at the University of Colorado Anschultz Medical Campus that was outfitted much like a typical hospital room. However, this room was completely enclosed, with air intake and exhaust regulated and equipment installed to analyze oxygen, carbon dioxide, and water content. Based on the results of this analysis, the researchers could determine how many calories the volunteers burned while each stayed in the room.

For two days, each volunteer lived in the room, continuing to eat the prescribed diet and spending the majority of their time in sedentary activities, such as watching movies or using a computer. However, on one of the days, they engaged in a sprint interval workout that involved pedaling as fast as possible on a stationary bicycle in the room that was set at a high resistance for five 30-second periods, each separated by four-minute periods of recovery in which they pedaled slowly with very little resistance. During the intense, 30-second bouts, the researchers coached the volunteers over an intercom system, encouraging them to give 100 percent effort.

Analyzing results from the room calorimeter system showed that the volunteers burned an average of an extra 200 calories on the sprint interval workout day, despite spending just 2.5 minutes engaged in hard exercise. ...

Saturday, January 05, 2013

Annals of brainpower: Oregon football

Q: How does Oregon football compete against the top teams in the country (four BCS appearances in the last four years) with recruiting classes that almost never break the top 10, and are usually ranked below 20? (In today's 2013 rankings, they're at 42!)

A: Great coaching by Chip Kelly and his staff. This video explains some of the basic concepts behind the Oregon spread offense. See also The zone read option game for Kelly's extremely well-written explanation of the Oregon running attack. Even a casual investigation reveals that football is by far the most complex sport in terms of coaching and game strategy. Too bad its days are numbered.

One of the big adjustments I had to make in coming to Michigan State was to Big 10 football. The offensive execution reminds me of high school play ;-)  Does any team squander more athletic talent year after year than Michigan? (As usual: currently #5 recruiting class, but will probably have another mediocre season next year!)

Thursday, January 03, 2013

Scientific publications by country

Click for larger image.

Judging by the numbers alone, China needs only 10 or 20 years to catch up with the US. But of course it takes a long time to build a high quality scientific tradition and infrastructure, so this is probably an underestimate.

The collapse of Russia is very sad.

Anglosphere still dominant, both in quantity and quality, normalized to population. Highly cited papers metric probably a bit biased toward these countries, but even correcting for that the previous comment stands.

Blog Archive


Web Statistics