positivederivative

[positivederivative]

2018-12-27T05:14:29.000Z

Today I Learned: 1) This one involves a pretty big worldview shift for me. It has to do with healthcare spending in the US, which I'm pretty sure I've written about here before. I'm afraid it's possible I've spread some... well, not *misinformation*, but possible *misinterpretation* of some key data. Here's a widely-known fact: the US spends vastly more on healthcare than countries with similar per-capita GDP. You can see the classic chart showing this below*. This fact is correct. Here's the critical caveat -- GDP is a fairly poor measure of how much money people actually have to spend. GDP is a measure of production. Stuff that gets produced might be bought by households... or by the military... or by other industries... or only by the rich.... Fortunately, there are better measures of household standards of living, and they track with per-capita healthcare expenditures much better than GDP. On these graphs, the US is right in the middle of the trend line. Why, then, do we spend so much per GDP-per-capita? It's because US households consume more (i.e., have quite a bit more disposable income) than the country's GDP would suggest. You can read much, much, much more about healthcare spending trends in the US and other countries here**. His main argument is that healthcare expenditure across the world is driven very, very strongly by how much disposable income people have, and that the US is a standard case of this where people have tons of disposable income. In other words, Americans spend a lot on healthcare because they have a lot of money to spend and they're willing to put a lot of it toward healthcare. That's... a really, really different interpretation from the usual one, which is that there's something pathologically and systematically wrong with the way our healthcare system works that makes it inordinately expensive (usually, that US healthcare is unusually privatized and insurance-driven). An important consequence of the healthcare-is-driven-by-disposable-income theory -- it predicts that we shouldn't expect socialization of healthcare to significantly change healthcare costs, except possibly by denying people the ability to spend as much as they want on healthcare. A note of caution -- the US also has really bad life expectancy for how much we spend***. That implies that there's *something* wrong with the way we're spending our money, or that there's something uniquely unhealthy about living in America. * http://bit.ly/2TenYYp ** http://bit.ly/2CBmkL4 *** See figure 2: http://bit.ly/2TenZLX 2) We have a shingles vaccine. Two of them, actually. They're not the same as the chickenpox vaccine. You might ask... why would there be different vaccines for chickenpox and shingles if they're caused by the same virus? Any antigen in one should also be present in the other one, right? Well, that's true. The chickenpox vaccine, though, is an attenuated (i.e., weakened) live virus. It's perfectly safe to a child or normal adult, but in the elderly or immunocompromised, it carries some risk of infection. The shingles vaccine is dead viral coat. It's not as effective as the chickenpox vaccine, but is safe to use in the elderly and immunocompromised, which is where shingles is most likely. 3) If you mix potassium and sodium metals, you get a room-temperature liquid. The alkali metals continue to surprise me.

[positivederivative]

2018-11-14T02:06:37.000Z

Fun fact -- for a human cell, the energetic cost of replicating a gene is less than the energetic cost of making the histones that pack it.

[positivederivative]

April 16, 2018 at 05:36AM

Today I Learned: 1) Read a few things on space suit tech today. The big one is on space suit thermoregulation. Long explanation of the physics of temperature control in space coming up; numbers two and three are much, much shorter, you can skip to those if you want. How do astronauts stay warm in space? Actually, the better question is, how do they stay *cool*? See, there are basically three ways to cool a human -- conduction, convection, and radiation. Okay, there's also sweating, which isn't really any of those three, so really there's conduction, convection, radiation, and sweating (evaporative cooling). Anyway, conduction is the direct flow of heat between objects touching each other, through the contact interface. That doesn't really happen in space, since there's nothing to touch. Convection is like conduction, but in a fluid media that's moving around and forcing a high temperature gradient to suck out heat, which makes convection much more efficient than conduction. Convection also doesn't really happen in space, because, again, there's nothing to convect *with*. It's space. Evaporative cooling by sweat *could* work in space, but only if you directly exposed that sweat to outside vacuum, which would be a bad idea for a number of reasons. That only leaves radiative cooling, which is loss of heat from spontaneous emission of light. All objects do this; humans emit something like 50 W under normal resting conditions, which represents about 50% of the energy we give off. In space, radiation... still works! But it's not tremendously effective -- only about half of your body heat can escape if all you have is radiation, so if you're stuck in space without a fancy way to actively cool off, then it's going to be hard to dump heat faster than your body naturally produces it. Now, as you heat up, you *will* radiate more efficiently... but the equilibrium temperature you'd reach isn't really human-friendly. Ergo, the problem in space isn't staying warm -- it's staying cool. ...well, that's what I thought. Today I learned that what I just described is... somewhat wrong. It's true that under "normal conditions", humans radiate about 50 W. However, part of "normal conditions" is being on Earth, immersed in air that's slightly cooler than we are, that's *also* radiating heat. That 50 W number is the *net* radiative emission of a human *on Earth*. In space, there's not a ton to radiate back to you (except the sun, which is a HUGE caveat!), so your *net* radiate emission is much, much higher than the usual 50 W -- more like 700 W, which is more than you can continuously generate for long stretches. So yeah, you're going to get pretty cold in space. But wait! What about the sun? Well... at Earth-orbit-distance, the sun provides about 1300 W per square meter (I use that number a lot. I should really memorize it one of these days so I don't have to keep looking it up). The human body, as a number of sources have kindly informed me, has about two square meters of surface area, about half of which can face the sun at once, so if you're floating naked in space, you should get about 1300 W from the sun. Now we're back to overheating. Except on the side that faces away from the sun, that side will cool pretty quickly. So yeah, you freeze on one side and heat to deadly levels on the other, unless you rotate nicely to keep yourself evenly insolated, in which case you just overheat. Oh, and if you happen to fall into Earth's shadow, then you immediately start to freeze instead. So, the truth is, staying safe in space isn't "all about staying warm" or "all about staying cool" -- you have to do *both*, *actively*, because whether or not you're in shadow makes all the difference between freezing solid and heat-stroking. You can see this in space suit design. One common feature of space suits is to provide a thermos-like layer of insulation between the astronaut and the outside of the spacesuit, which buffers them tremendously against *changes* in ambient radiation. Of course, then the astronaut has exactly the opposite problem as a naked person floating in orbit -- they *can't* radiate heat very effectively, and can only conduct and convect and evaporately cool with the rest of the suit. In practice, from what I understand, the astronaut usually needs more help cooling off than staying warm, *within the thermos layer of a space suit*, so space suits come with liquid cooling systems to draw heat from the body to the exterior of the suit, where it can be (radiatively) dumped into space. 2) Here's a smaller space suit fact -- space suits are specifically designed to keep a constant volume no matter how the wearer moves, because changing the volume of a pressurized gas-filled container takes work (in both the physical and colloquial senses of the word). There are a bunch of solutions to the constant-volume problem, but the most interesting nugget of this fact, to me, is that the constant-volume problem is a design consideration at all. 3) There's a horizontal white line in the mouth, right about where the teeth come together. It's called the "linea alba". I didn't read much about it, but apparently it's thought to be a thickening of the skin in response to friction, rubbing, grinding, etc. from the teeth. As far as I can tell, the linea alba provides no fitness benefits. In fact, it is now my go-to example of an evolutionary spandrel -- it's a total side-effect of whatever feedback system makes skin grow tougher where there's more wear and tear.

[positivederivative]

April 05, 2018 at 01:21AM

Today I learned: 1) More details in the still-unfolding picture of bird orienteering. Everyone knows that songbirds migrate north and south depending on the season. What nobody knows is how they know which way is north and which way is south. Well, nobody knows *for sure* and *in complete detail*, but an incomplete picture has been emerging in the last few decades. This year, we got a few new pieces in the puzzle. Something I didn't know about before, but has been known by bird orientation scientists for a while, is that bird orientation is closely associated with a couple of cryptochrome proteins. Cryptocrhomes are a class of proteins found in a bunch of animals and plants that sense blue light. In most organisms, cryptochrome proteins have something to do with detecting daytime and maintaining circadian rhythms. In the 1970s, one Klaus Schulten postulated that cryptochrome proteins might also be sensitive to magnetic fields, and might be responsible for magnetoperception, which is pretty well understood to be critical for bird orienteering. Since then, a bunch of ancillary evidence has been gathered that points to cryptochromes actually, in fact, being important for bird orientation (for example, bird orienteering requires blue light to function). Here's some new information as of this year -- birds have three cryptochrome protiens, called Cry1, Cry2, and Cry4. Sorry, that bit isn't new. What's new is that we think we know which one is responsible for magnetoperception. Spoiler alert -- it's Cry4. To figure this out, some scientists at Lund University and the Carl von Ossietzky University Oldenburg looked at Cry1, Cry2, and Cry4 gene expression over time in the brains of a bunch of birds. The idea is that if a bird uses a particular cryptoprotein for sensing day/night, then its expression should oscillate in a roughly day/night way; if, on the other hand, a bird uses a particular cryptoprotein for sensing magnetic fields, then it would probably be expressed constantly throughout the day. If a bird used a cryptoprotein for *both*, then the data would be difficult to interpret and some grad student would have to call on more sophisticated analysis methods. As it turns out, Cry4 is the odd one out -- not only is it expressed more evenly across day and night than Cry1 and Cry2, but it also is expressed more during migratory seasons. The same researchers also identified *where* Cry4 is expressed -- turns out there's a bunch of it in the retina, right where it would pick up the most blue light. It's not a knockdown argument for Cry4 being The Protein for bird navigation, but it's quite suggestive. I like this particular story because it's a good reminder that Science is Still Happening. It's easy to forget that science is littered with the passed-over husks of questions that once seemed intractable or even impossible to answer scientifically, like "where did the moon come from and how does it stay in the sky?" or "how do animals move?" or "what is light?" or "why do mothers love their children?". This is the shape of science in action: we don't understand something, sometimes profound (why is there something instead of nothing?), sometimes mundane (why do we yawn?); somebody (or a bunch of somebodies) decides to investigate the issue carefully and thoroughly using experimentation as their guide; science grinds away the question until the answer is mundane enough that it's obvious, in some sense, if you've seen all the data. I think there's a lesson in this story -- if you ever hear a claim that "science can't explain ", remember that being totally inexplicable is just the first stage in the life cycle of a question, just as many now-answered questions have been. Eventually, the lens of science will turn to and it will gestate for a time as an active field -- like the question of how birds navigate -- before hatching as a fully-fledged fact. 2) There's a (free-living!) Eukaryote smaller than an E. coli. It's name is Ostreococcus tauri. O. tauri is a photosynthetic plankton, found commonly anywhere in the ocean you can find plankton. O. tauri is *small* -- morphologically, it's a simple, unadorned sphere about 0.8 microns across. For reference, E. coli is about the same size (which is typical for a bacteria); typical human cells are about 10 microns across; a human hair is about 100 microns across. O. tauri is small in a couple of other cool ways, too. For one thing, it has a pretty small genome at only about 12.5 million bases. That's not quite *tiny*. E. coli is a third that size (though it's a bacteria), and baker's yeast (S. ceriviciae) has a slightly *smaller* genome. However, the O. tauri genome is incredibly *compact* -- it fits almost twice as many genes into its genome as S. ceriviciae. It seems to have done this largely be removing introns and intergenic regions. I imagine there must also be a pretty serious amount of gene overlapping to make it all fit. O. tauri also has a rather compacted, uh, organellosome? If a quick scan of the literature is to be believed, each O. tauri cell comes with a single nucleus, a single (pretty big) chloroplast, and a single mitochondria, along with a few other attendant organelles and a bunch of ribosomal granules. If I, say, wanted to port mitochondria into a bacteria, I'd start with O. tauri's mitochondria -- they're already small enough to fit, and I'll be they'd be a good simple model for mitochondrial reproduction. 3) Got my wisdom teeth taken out today, so I learned what that was like. The whole thing was done under local anaesthesia, so I was fully conscious and aware through the whole thing, and more-or-less fully functional as soon as it was done (aside from not being able to talk or tell whether my lips were in the right place). It was a remarkably cool experience, while the actual surgery was happening. It kind of felt like I was a skeleton. There was no pain, nor sensation in the gums or other soft tissue (except in the back of the throat, which got sprayed with enough water and dust to be pretty uncomfortable), but I could feel everything happening to the bone. Both lower wisdom teeth were impacted, so the surgeon had to go through plenty of gums, but to me it felt just like a dentist poking and prodding any other tooth (but less painful), as though there were no gums in the way at all. Now I'm at the part where it feels like I got punched in the face a couple of times. At least I didn't have to deal with the aftermath of sedation.

[positivederivative]

March 17, 2018 at 05:40AM

Today I learned: 1) The PDF for the Gaussian distribution was discovered something like 80 years before Guass by de Moivre, who wrote it out as part of his treatise on probability "The Doctrine of Chances" (which was apparently written primarily for gamblers?). From what I gather, de Moivre's paper on the normal distribution (which he used as an approximation of a binomial distribution for large numbers of trials(!)) languished in obscurity until well after Gauss formalized the distribution and popularized it. 2) Related: Stigler's Law of Eponymy states that any scientific or mathematical discovery named after a person was not discovered by the person it was named after. This includes Stigler's Law (attributed by Stigler to one Robert K. Merton). 3) On a more sober note, today I learned that lynchings in the US were much, much worse than I thought. I *thought* lynchings were basically public killings with a big crowd. In fact, lynchings were more akin to medieval executions -- that is, the murder part would be preceeded by hours of torture involving things like open flames and removal of body parts.

[positivederivative]

March 15, 2018 at 03:50AM

Today I learned: 1) You know how gravity falls off with the square of the distance between two objects? As does electromagnetism? That's because we live in three spatial dimensions (for the same reason that the surface area of a 3D sphere increases with the square of its radius). If we lived in 4D space, then forces would fall off with the cube of distance; if we lived in N-D space, forces would fall off to the power of 1/(distance^N). That also means that the inverse-R-squared law is strong evidence that we do, indeed, live in a 3D space, and not, for example, a 3D slice of a higher-dimensional space. Wouldn't that immediately falsify string theory, which posits lots and lots of dimensions? Well... no. There's a caveat to the "forces fall off to the power of 1/(distance^N)" law, which is that it only holds as long as all of the spatial dimensions have the same characteristic length scale. Now, I must admit, I don't fully understand what the "length scale" of a dimension is. Nevertheless, if a dimension is "small", then not much force will leak out into it, and the force falloff will remain very close to the 1/R^2 law. As of 2005, gravity had not been measured to a high enough precision to distinguish between a 1/R^2 falloff and an almost-1/R^2 falloff, leaving room for the possibility of other, smaller dimensions. As far as I know, that fact hasn't changed in the last decade. 2) ...how to return Amazon packages. It's ridiculuosly easy. First, you go to your orders on Amazon, find the thing you want to return, and click some relatively obvious button that says something about returning the item. Follow the instructions. If you can get it to a Kohls or an Amazon locker, they'll package it and label it and ship it to you for free. Otherwise, if you get it to a UPS store, they'll package it, label it, and ship it for some (not always outrageous) application of money. Also, to keep things snappy, if you ask for an item replacement from Amazon, they will immediately ship you the new thing. If you don't return the original item postmarked before some date, they'll automatically charge you for it again. I do wonder if you could abuse this somehow by, say, buying a ton of things at once, ordering up replacements for all of them, then shutting down your Amazon account (or your credit card) before the due date. 3) I somehow got it into my head that you could decompose any linear transformation a rotation and a scaling, possibly with a reflection. I was wrong -- I'm pretty sure scalings *don't* account for shearings, and even including shearings doesn't give you all of the linear transformations. As a side note, I *did* find the conditions under which you *can* write a linear transformation as a rotation plus a scaling -- for a linear transformation with matrix [[a, b], [c, d]], you can do that decomposition iff b = -ac/d. You're welcome? I guess?

[positivederivative]

March 13, 2018 at 11:36PM

Today I learned: 1) The number line can be thought of as a limit of a circle with increasingly large radius. This is useful for proofs involving the number line, and can get you a factor of pi in sums that otherwise isn't obvious. For more, see https://www.youtube.com/watch?v=d-o3eB9sfls. 2) Synthetic biological circuits break. A lot. Almost all synthetic circuits are bad for the cell they're in, so most mutations that break the circuit will be selected for pretty quickly. One of the most common types of circuit-breaking mutations is accidental insertion by Insertional Sequence (IS) elements, which are small DNA fragments that make transposons, which flip the IS out of whatever DNA it's in and move it somewhere else. Bacterial genomes have lots of IS elements (E. coli has a few dozen, depending on the strain), so any engineering in bacteria will run afoul of them eventually. So. IS elements are common in bacterial genomes, and they mess up biocircuits. Why not remove them from the genome? Well, somebody has -- Scarab Genomics LLC sells a variety of IS-free cell strains for a variety of cloning needs, for the low low price of... actually, I don't know how much they cost. You have to make an account with them and sign in to see prices. More importantly, though, Scarab Genomics has some pretty nasty licencing restrictions on their strains -- they get a cut on any IP you develop using their lines, for example. Not nice. 3) Strong ribosomal binding sites (RBSs) can protect mRNAs against degradation. Probably. This is a pretty new finding, but it looks like if an RNA is covered in ribosomes, the ribosomes can physically block RNAses from binding and degrading the RNA. Strong RBS -> lots of attached ribosomes -> less degradation. The effect is modest, but measurable.

[positivederivative]

March 13, 2018 at 01:10AM

Today I learned: 1) ...how to replace snap-on buttons! Technically, that's a lie. I looked up how to replace snap-on buttons late last week. But! Today I actually *did* some snap-button replacing, which honestly felt a lot more like learning than watching a video. For the record, the hardest part was removing the old buttons, which wasn't terribly difficult once Andrey Shur provided me with the right tools. 2) One of the most clear-cut problems with the US healthcare system is the use of opt-in organ donation. It's well-known that many, many people who would be fine with donating their organs on death never bother to sign up for organ donation, and lo! We have a chronic, severe shortage of organs for transplantation. It would be SO EASY to switch to an opt-out system where everyone is, by default, signed up for organ donation, and you can fill out a form to not donate if you really want to. This would vastly increase rates of donation, while still giving people the opportunity to not donate if they have any objections serious enough to warrant filling out some paperwork. ...except that it isn't *quite* that clear-cut after all. Opt-out policies *do* correlate with higher rates of donation, but there are opt-out countries with very low donation rates, and Spain, the most widely-cited success story for opt-out policy, has the Very Large Caveat that they implemented a bunch of organ-transplantation-related reforms around the same time as opt-out, which seem to be responsible for a good part of the increased donation rates there. There are also some specific failure modes for opt-out. For example, when Wales switched to an opt-out system, they saw rates of organ donation *decrease* (albeit slightly). A possible reason is that in Wales (and many other countries with opt-out organ donation), the family can still decide to deny access to a deceased organ, so opt-out isn't quite so opt-out as it sounds. Before they switched to opt-out, the family could *also* decide to allow transplantation if the family member wasn't signed up. So really, in both systems, anybody who doesn't bother filling out paperwork to make a decision one way or another is actually at the mercy of their family. In opt-in, there's a way to guarantee that you will donate your organs; in opt-out, there is only a way to guarantee you will *not* donate your organs. Therefore, it's possible that the opt-out system actually creates an extra group of people that don't donate. 3) There is a belief running around that chewing gum is bad for pregnant women. I'm quite skeptical. Some quick googling brought up a bunch of results (with a fairly wide spread of claims), but none from sources I would consider reliable. The only plausible-looking claim, as far as I can tell, is that some gums are sweetened with sorbitol, which is a diuretic and can cause intestinal distress if taken in large quantities. Is that likely to actually be a problem? I'm guessing not.

[positivederivative]

March 12, 2018 at 03:28AM

Today I Learned: 1) Medieval executioners were almost completely separated from larger society. Firslty, executioning was, like most trades in medieval Europe, familial. Executioners were born into the job, whether they liked it or not. Secondly, executioners were considered kind of magical, kind of blessed, and kind of cursed. Importantly, they had the special ability to remove a person's honor with a touch, like a morbid, adult form of cooties. You're in a marketplace and you accidentally brush up against an executioner? Whoops. You just got infected with the executioner's aura. You are no longer fit for polite society. One side effect of executioner hygeine was that executioners were a bit of an inbred breed -- only the children of executioners were fit to marry an executioner, so their lineages remained separate from most of society. This fact-set courtesy of Dan Carlin's Hardcore History. 2) Henna tattoos are traditional for brides. Didn't know that. 3) The Viking probe we sent to Mars found some evidence for life, although it didn't pan out in the long-term. Viking carried four tests for carbon-based life: a mass spec to directly look for organic compounds, and three experiments that looked for release of compounds of various kinds when nutrients were added to soil samples. One of the release experiments came up strongly positive... but it turned out that the release (in this case, radiolabeled CO2) could be explained quite well by gamma ray irradiation of the sample.

[positivederivative]

March 11, 2018 at 05:13AM

Today I learned: 1) I'm a big fan of Beethoven, espeically his piano sonatas. I love how beastly they are, how passionate, how ridiculously ahead of their time they are (did you know Beethoven invented boogie-woogie? https://youtu.be/ccyHT1sFmsg?t=1032). Today I learned that Beethoven's sonata no. 23 ("Appassionata") was *so* ahead of its time that it was never performed in public until after Beethoven's death. He played it for some of his colleagues and students (including Czerny), nobody wanted to play it. One critic (often-quoted online, but without citation) called it "incomprehensibly abrupt and dark". From the sound of it, Beethoven's contemporaries couldn't, for the most part, parse it. On a related note, apparently Beethoven never technically bought a piano. All of his were loaned, rented, or gifted by piano manufacturers. Beethoven was famously frustrated with the piano of his time. Pianos of the late 18th and early 19th century were sickly cousins of their modern descendants in a lot of ways. They were smaller both in soundboard size and range, they were more limited in their ability to play repeated notes, and they generally sounded much weaker (compare harpsichords to a modern concert piano). The biggest single advancement in piano technology was the cast-iron frame, which lets modern pianos be strung with absolutely immense tension, and lets you put a ton of kinetic energy into a performance in a quite literal way. Unfortunately, the cast-iron frame was only invented in the last couple of years of Beethoven's life, and he was constantly frustrated by the lack of his pianos' abilities to express what he wanted. He pushed what he had to the limit, though -- the Appassionata, for example, goes all the way to the highest and lowest notes available on his piano at the time. 2) If you're eating a Thai curry and you bite into a chunk of something that looks, tastes, and feels like ginger, odds are it's not ginger. It's probably one of the four varieties of galangal, a ginger-like root used as one of the main ingredients in Thai curry. 3) You can buy oreos in eastern Asia, but they're packaged a little differently -- each oreo is individually wrapped. In fact, single-bite individually-wrapped packages inside a larger package seems to be a common motif of east Asian snack foods.

[positivederivative]

March 09, 2018 at 08:28PM

Today I learned: 1) According to Harvard professor Johan Paulsson, if you fuse most fluorescent proteins to a protein that naturally multimerizes (that is, a protein that forms a complex of two or more copies of itself), the fluorescent protein attachment will cause the whole protein to clump dramatically. I'd read this before, but misread it slightly and thought that this applied to *any* fluorescent protein fusion. The thinking is that fluorescent proteins normally bind to each other, but only very weakly; however, when two or three or ten of them are all fused to one complex, it acts as a nucleation site for larger-scale aggregation. Citation: http://ift.tt/2FwbNQi 2) You know the weird thing in quantum mechanics where particles kind of appear and disappear at random in a kind of quantom froth? Well, there's a way of viewing that phenomena as a consequence of simple formulations of quantum mechanics married to special relativity -- essentially, quantum systems get random spikes in energy, and special relativity says that you can interconvert between matter an energy, so there must be occasional production of particles from random QM fluctuations. According to Anthony Zee, the "marriage of QM with special relativity" is also one of the primary motivations for developing quantum field theory. I don't yet understand this claim, nor why it should be true. 3) Say you're a cell, and you want to make some specific amount of a protein. There are, roughly, two variables you can (dependently) vary to get the right amount of output protein -- transcription speed and translation speed*. For a fixed amount of output, you could have TONS of transcription and a little bit of translation on each of the many mRNAs you make, or extremely little transcription and TONS of translation on every mRNA you make, or anywhere in between. Today I learned that, across a wide variety of organisms, cells overwhelmingly choose to have low transcription and high translation over high transcription and low translation. Citation: http://ift.tt/2oIM3K9. This flies in the face of experimental data I've seen before that tells us that translation is much, much more energetically expensive for a cell than transcription and, in particular, that high-strength RBSs are *even more* energetically expensive than you'd expect. So... I don't know why cells would do this. The Alon paper has some arguments that don't make sense to me. One possibility, though, is that expressing low transcriptional rates makes gene expression *noisier*, which could be better for some processes. Still, I'd be surprised if the *vast majority* of genes are better expressed noisily than consistently. * Okay, you can also modify degradation speed and a couple other things, but I'm going to neglect those, because they turn out not to be too interesting in this story.

[positivederivative]

March 07, 2018 at 02:36AM

Today I learned: 1) ...there seems to be some disagreement online about whether non-profit and not-for-profit organizations are the same thing. Both non-profits and not-for-profits are organizations whose purpose is not to make money for shareholders, and both are tax exempt. But are they the *same* thing? I'm not sure -- I've found multiple sources claiming that they are the same thing... and multiple sources claiming that there are subtle differences between the two terms (e.g., not-for-profits can pay employees and members, non-profits can't). I'm not going to link any of those sources, because frankly I don't trust any of them overly much. Anybody know a good authoritative source that might define non-profits vs. not-for-profits? 2) Intuitively enough, there is such a thing as low treason. Sort of. It's actually called "petty treason", and it doesn't exist in most modern legal systems to my knowledge, but it *used* to be for treason against a lawful institution that was below the state -- say, a slave trying to murder his or her master or mistress. 3) Somewhere I picked up the idea that basically nothing in the ocean eats jellyfish, with the exception of sea turtles. Boy was I wrong -- plenty of sea life eats jellyfish, which is consistent with jellyfish essentially being crunchy, pulsating vegetables. Fish, crabs*, sea anemones, and even sea stars will eat jellyfish, given the chance. *Some of them, at least. Some crabs eat some jellyfish. Some jellyfish eat some crabs. I think both groups are "if I can stick it in my mouth, I can eat it" kinds of creatures.

[positivederivative]

March 06, 2018 at 02:10AM

Today I Learned: 1) English monarchs used to have a servant specifically for assisting with defecation and bathing, officially called the Groom of the Stool. The Groom of the Stool had to be someone in whom the king (always king -- queens got a First Lady of the Bedchamber instead) had total trust, and so usually also served as confidant, and ranked rather highly in the court. 2) ...a new way of visualizing how a Fourier transform works, courtesy of youtuber 3Blue1Brown. This one's really much better seen than described, so I recommend checking it out yourself. Also worth a watch if you really like spirographs. https://www.youtube.com/watch?v=spUNpyF58BY 3) The Romantic composer Felix Mendelssohn had a sister, Fanny Mendelssohn, who also wrote hundreds of piano pieces. I am no expert in either composer, but from a brief listen to some of her music, she and Felix shared a lot of musical sensibilities....

[positivederivative]

March 04, 2018 at 04:58AM

Today I learned: 1) Can you guess the most-played race in D&D? Today I learned that it's humans, by a pretty wide margin, according to Wizards of the Coast's polling. 2) Here's a potentially useful narrative trick -- the main character of a story doesn't have to be as interesting as the other characters, especially if it's a first-person story. In general, you can get a reader to go along with a main character by sheer dint of them being a main character. Learned this from a fellow student and part-time writer who had a problematic character that needed to exist for plot/connective reasons, but wasn't very interesting. So they flipped the story to first-person around that character. Instant fix. 3) Romans... ancient Romans were a special people. Incredible, but also so, so terrible. Case in point on the "Romans are the worst people ever" side of the ledger -- fatal charades. This was a Roman practice of performing plays using condemned criminals as actors. The only catch was that the plays involved the deaths of main characters, which were performed for real, live, on stage. Like... imagine Hamlet, but all the actors are on death row, and they actually die on stage. The Romans did that. I'm a bit curious how they convinced the prisoners to go along with it. Also, they must have had awful rehersals.

[positivederivative]

March 03, 2018 at 04:26AM

Today I learned: 1) Cave bacteria! There are bacteria that live exclusively on the inside surfaces of cave rock. They form visible sheets, and sometimes excrete really pretty minerals. I don't know much about them, but they can somehow live off of the rock face -- I'm not sure whether they're filter-feeding stuff that comes by, or if they're actually reacting the rock for energy. Whatever they're eating, it's not a particularly *accessible* source of food -- they are VERY SLOW GROWING, taking decades to fill small gaps in their mats. Some kind scientists at University of New Mexico compared the community compositions of cave bacteria and soil bacteria (http://ift.tt/2F8VxVa). Their conclusion was that the broad distribution of taxa was quite similar in the caves and in the soil above, but that there was quite a bit of divergence between *species* in the two climes. In other words, it looks like all of the usual inhabitants of soil got into the caves and colonized with roughly the same success, but then they evolved to look pretty different from their above-ground ancestors. Thanks to Patricia Prewitt for tipping me off to the existence of these critters! 2) RAND corporation has a new meta-study on the effects of gun policy. I've only read their summaries, but it looks pretty comprehensive. The big take-away is that there isn't enough data to make strong conclusions about most questions around the effets of gun control. There isn't a ton of data on gun violence, and what data there is isn't sufficient to reliably detect small effect sizes (which could still add to thousands of deaths per year). None of this is surprising, since the US government is BARRED BY LAW FROM FUNDING RESEARCH ON GUN CONTROL. It's usually not a good sign if someone feels they have to ban research on a topic, especially a poitically-charged one. Frankly, I don't take much of a hard stand on gun control one way or another, but IMHO the current research climate on gun control reeks of Lysenkoism. ANYWAY, the second-biggest take-away from the RAND meta-study is that there are *some* policies that appear to be effective at reducing violent crime, accidental deaths, and suicides (especially suicides). The first figure at this link is a good summary (http://ift.tt/2HYu9uF). For those who prefer a written digest, here it is: background checks and child access prevention laws help reduce suicides pretty dramatically; child-access prevention laws also almost certainly reduce accidental gun deaths; background checks and mental health screening probably decrease violent crime, and (perhaps surprisingly) stand-your-ground laws probably *increase* violent crime; concealed-carry laws might increase both accidental and homicidal deaths from guns, but the evidence is weak; most surprisingly to me, there is not evidence that bans on assault weapons and high-capacity magazines have much effect at all on anything. You can read the rest of the report here, if you're super-interested in gun control research and you have a lot of time on your hands: http://ift.tt/2t87FE9 3) Cells in a developing embryo can detect their position within the embryo within about 1% error, we think only using the concentrations of four transcription factors that are distributed in specific ways around the embryo. We don't know exactly how they do it, and that's quite close to the inforamtion-theretical limit of precision for the amount of information available in those transcription factor signals (unless, of course, there are channels they're exploiting that we don't know about). Thanks to Andy Halleran for this one!

[positivederivative]

March 02, 2018 at 02:47AM

Today I learned: 1) The determinant of a matrix is the amount by which the matrix stretches or squishes space (when that matrix is viewed as a linear transformation). I've seen the determinant of a matrix expressed in a lot of ways... but this one makes by far the most sense. Good lord, this took to long to run across. 2) Today I analyzed the simplest possible molecular feedback system -- one where a molecule is produced catalytically (in this case, protein from a gene) that inhibits its own production*. I'll call the molecule a "repressor", since it represses its own production. The basic gist of this system is simple enough -- if there's a lot of the repressor around, it will bind up its gene and stop production and its concentration will fall, eventually bringing total concentration back into balance. If there isn't enough of the repressor, it won't bind much of its gene, and the gene will pump out tons of product until its goes back to equilibrium. Simplest-possible negative feedback system, right? Here's my question about this system -- as you add more gene, do you get more repressor? From the cartoon-level view, it's not immediately obvious. Maybe the repressor will just compensate for the extra gene and push production down until you have the same level as before... or maybe it will push it down, but not quite to the same equilibrium, and you'll get more repressor. But how much more? Is it a linear relationship? Less than linear? *More* than linear? So, I crunched out the ODEs for the system and got the following result -- repressor concentration *does* increase as you add more gene. Specifically, it increases with the square root of the amount of gene. So adding more gene does push up the repressor concentration, but in a (big-O) "modest" way. * More technically, this is for a gene that is produced, can degrade, and monomerically binds to its own promoter, completely preventing production while bound. 3) There's a common form of problem in Bayesian learning where you are trying to figure out the parameters of some model, and you're willing to use a uniform distribution to describe your prior information about the parameters (i.e., any parameter value is equally likely), and the probability of any particular piece of data is Gaussian-distributed for a particular choice of parameter (i.e., your model predicts that an output is something-determined-by-input-and-parameters plus some Gaussian noise). If you write out Bayes' Rule for that particular scenario and drop the annoying mathy normalization constant, you get something like P(parameter) = Gaussian(f(x, parameter), σ) * Uniform(parameter) where f(x, parameter) is whatever the model predicts based on an input x. For example, if you think that, say, temperature in some place drops linearly with height, (H) but you don't know how *quickly* temperature drops off (M) or what temperature it is at ground level (T0), then you might have a model that looks like T = f(x, M, T0) = T0 - MH If you additionally think that any temperature measurement will actually have some noise in it (which is pretty reasonable), then you might add some Gaussian noise with some spread σ: T = Gaussian(f(x, M, T0), σ) = Gaussian(T0 - MH, σ) Then, if you observe the temperature at some height h, you can use (a stripped-down) Bayes' Rule as written above to figure out the probability of any particular values of M and T0: P(M, T0) = Gaussian(T0 - Mh, σ) * Uniform(M, T0) Today I learned that the probability that you get out of that particular piece of updating *isn't Gaussian*, despite the fact that it looks so clearly like it's "a gaussian times one". This particular fact makes life somewhat more annoying, from a Bayesian perspective... but I'm pretty sure this explains some bad inferences I've crunched through in the past.

[positivederivative]

February 08, 2018 at 03:38AM

Today I Learned: 1) A couple of data facts for you, brought you to in part by Andy Halleran. Current total data of the human race is estimated at around 10^24 bytes. That's a trillion terabytes, or a yottabyte. I know some pretty big metric prefixes, but I had to look that one up. The estimated total data size of all (unique) genomes on the planet is 10^36 bytes. That's... a truly ludicrus amount of data. But we're catching up. For a (three-year-old) overview of some of our biggest data generators, see this (open-access!) review article: http://ift.tt/1HcOjIi. A few highlights: a) The Australian Square Kilometre Array Pathfinder project acquires 7.5 terabytes *every second*. b) Twitter's data storage needs are estimated at around 500 terabytes per year. Not massive by big data standards, but it does put into perspective why Twitter doesn't, in general, make their Tweets algorithmically searchable*, and makes it all the more impressive that they can serve up selected data as quickly as they do. c) We're predicting to store somewhere on the order of 2 to 40 *exabytes* (~8.5 billion terabytes) of human genome sequences alone. * You can collect random tweets from a couple of Twitter APIs, known as the twitter firehoses, and you're welcome to make your own mini-database. 2) A couple of rocket facts for you today, courtesy of Andrey Shur. Liquid-fuel rocket engines use (effectively) a turbojet to pressurize their fuel just before it's burned. To rocket good, you have to make really, really high-pressure air. One way to pressurize air really efficiently is to burn it; that's why explosions are useful in rockets. Another way to pressurize air is to use a pump; that's why turbojets are useful in rockets. The "turbojet" in a rocket is actually called a "turbopump", and its job is to pressurize incoming liquid fuel (as it's aerosolized coming out of the fuel tanks, I gather) so it can be exploded really efficiently. Now, pumping fuel requires a lot of energy. You *can* run turbopumps electrically, if you want, but a much more common strategy is to *pre*burn a bit of your fuel and use *that* explosion to spin up the turbopump. It is not intuitive to me that this should be efficient, but apparently it is. I guess pressurizing fuel before burning it must give you a disproportionate boost in thrust post-burn. There are (at least) two main kinds of turbopumps. An "axial flow" turbopump looks pretty much like a turbojet, which is the big spinning engine on the wing of a commercial plane. Basically, it's a bunch of high-density propellors that force air through a narrowing chamber, compressing them. The "centrifugal" turbopump is more common in rockets, and it's the one that makes me giggle -- it works by spinning really fast and *flinging fluid to the outside* where it's collected at high speed. So now you know -- these things (https://www.youtube.com/watch?v=PfHu-UJaK0Q) are, when you get right down to it, centrifugal squirrel pumps. Not quite a fact, but here's a gem from the Wikipedia page on rocket engines, on the topic of the dangers of liquid propellants: "With liquid propellants (but not gaseous), failure to ignite within milliseconds usually causes too much liquid propellant to be inside the chamber, and if/when ignition occurs the amount of hot gas created can exceed the maximum design pressure of the chamber, causing a catastrophic failure of the pressure vessel. This is sometimes called a hard start or a rapid unscheduled disassembly (RUD)." 3) You know how you're not supposed to daisy-chain together power strips? If you don't already know this, you're not supposed to plug power strips into other power strips. Ever wonder why? Like, why would that be dangerous? I've been told these terrible stories of daisy-chained power strips sparking and causing fires, but nobody ever gave me a reason that they would do that. Well, today I got fed up and googled the answer. It turns out that daisy-chained power strips are dangerous mostly because they make it much, much easier to accidentally draw too much total power at once. Power strips (and, for that matter, house circuits) are only rated for so much current draw, and if you exceed it, you risk overheating or shorting the strip. Practically speaking, most power strips have fuses and will commit suicide before they do any real damage, but it's still a hazard. The upshot -- you should be safe daisy-chaining power strips *if* you are quite sure you're not going to exceed the max draw for your power strips (say, if you fill every plug with cell phone chargers).