Economics is one of the worst offenders, agreed, and the primary target of The Black Swan. But a number of other fields, including some that are supposedly in mainstream science, are almost as bad. And not necessarily intentionally. The saddest thing is that the researchers themselves are often totally oblivious as to just how biased and/or weakly founded their own results are, because they always get them "using statistics" in a traditional way, not realizing that they are using Statistics 101 (freshman intro stats) where what is required is Statics 404 (a graduate course and not for the faint of heart or incompetent in calculus, ODES, combinatorics, and a certain amount of common sense and experience).
Precisely. Indeed, a major part of the solution is to make scientists their own greatest skeptic, to mistrust our own pet ideas, to hesitate to claim "proof" to a fault even if evidence or model computations seem to support it.
The latter are an entire category in and of themselves. I "do" predictive modelling and moderately advanced statistics on a professional basis, and even have a patent pending in the field. I've done Monte Carlo computations in physics for well over a decade, and know a lot about randomness and hypothesis testing compared to your average scientist in the street, so to speak. I am all too aware that model computations are among the least trustworthy kinds of evidence and usually have far less predictive power than that which is claimed by the modeler. The problem there is subtle and related to complexity and nonlinearity. A highly multivariate, semi-empirical, nonlinear theory implemented as a model is often implemented by "fitting" some or all of its parameters to some (sub)set of data. This in turn is often equivalent in modelspeak to using hill-climbing (gradient search) to find an optimum fit to the data relative to some selected parametric starting point (this is sometimes referred to as making a "Bayesian" choice of the parameters based on some set of data used as priors").
There are many problems with this. One is the problem of omitted variables. In many of the problems where this is done, the choice of parameters (dimensions in the parametric space) is highly model dependent. Heuristics are often used to limit the size of the parametric space simply because doing anything in a really high dimensional space is a lot of work and introduces a substantially higher (but honest) estimate for errors in the final result. Heuristics, of course, is code for "I don't think these variables will significantly contribute", an open opportunity to omit variables that you don't want to be significant because they confound your hypothesis. A second problem is that many models are de facto parametric nonlinear function approximators. This means that -- especially if the data being fit to the parameters is "simple", e.g. monotonic or otherwise simply nonlinear over the range of the fit -- it is often perfectly easy to fit the data with a set of parameters, have the fit be "optimal", have the fit produce a perfectly reasonable chisq, and have the parametric fit be perfectly meaningless. This is all elementary modeling theory 101, but somehow "hiding" the basis by turning it into the solution of a set of coupled ordinary differential equations with nearly e.g. sinusoidal or nearly e.g. polynomial or exponential behavior makes the problem somehow disappear in the minds of the modellers. A third problem is that of complexity -- in many cases (especially for highly multivariate nonlinear models) there may be many local optima and hill-climbing from a selected starting point can easily be yet another form of inadvertent confirmation bias. A global search might find a better optimum, or might reveal that there are several constellations of parameters (especially when omitted variables are included) that can fit the empirical data within its precision, (properly) reducing confidence in the final prediction. A fourth is that even a model built with the variables that were important in the past (where the data being fit resides), that is robust in any parametric/Bayesian search, that uses any of several methods to "validate" the model (using past data) can easily fail in the future because the model simply does not extrapolate. The real space of variables and data is much larger, the model is always being built on some sort of optimistic projection onto a manageable subspace, and ignored stuff eventually becomes important and causes a complete deviation from the model. Chaotic models, stiff differential models -- there is no lack of examples, but somehow this sort of thing doesn't get factor
I disagree about the rarity, on the basis of empirical evidence, for example the recent paper in Science (IIRC, sorry I don't have the reference handy) in which a cancer researcher failed to replicate 46 out of 53 papers published -- all of them with peer review -- prior to embarking on new research in the field. Similar meta-studies have turned up astounding rates of non-reproducible results in other fields (some more than others -- sociology and IIRC social psychology topping the non-medical list).
One problem is that we have constructed a system that rewards the publication of positive results and punishes negative results published or unpublished. Punishes as in makes or breaks the entire career of young researchers, if the negative result occurs when they are up for tenure. Rewards as in ensures research funding and professional advancement as long as positive results keep flowing out.
Another fundamental problem that peer review has a terrible time with is confirmation bias. Science in general has a serious problem with confirmation bias. If one ever embarks on a study where one seeks evidence for some causal linkage associated with some phenomenon in a general population where the phenomenon occurs, one can always find exemplars that support your hypothesis. Lacking actual work to replicate your results using sound methodology (e.g. double blinded and/or conducted using competent statistical analysis, something still as rare as hen's teeth in science in general because to it is difficult to do statistics correctly in a complex problem, not easy, and certainly not easy as in covered in one or two undergrad stats courses which is all that it is probable that the researcher has ever taken) confirmation bias can not only worm its way into the literature, it can come to dominate entire fields as a significant fraction of scientists who do the reviewing for both publication and grants are "descended" from one or two original researchers and their papers. It can take decades for this to be discovered and work out in the wash.
Peer review works better in some disciplines than others. Math it works well, because there is literally nothing up a publisher's sleeve -- fraudulent publication is indeed impossible and even mistaken publication is relatively rare and conditional on involving math so difficult even the reviewers have a hard time following it. Physics and the very hard sciences are also fortunate in that it works decently (although less perfectly), at least where there is competition and the proper critical/skeptical eye applied to results new and old. At least there a mix of laboratory replication and strong requirements of consistency usually keep one out of the worst trouble.
A simple rule of thumb is: The more a result relies on population studies, especially ones conducted with any kind of selection process or worse selection process plus the actual modification of the data according to some heuristic or correction process, where the study itself is conducted from the beginning to confirm some given hypothesis, the more likely it is that the result (when published) is bullshit that will eventually, possibly decades later, turn out to be completely wrong. If you have enough places for a thumb to be subtly placed on the scales and the owner of the thumb has any sort of vested or open interest in the outcome, it is even odds or better that a teensy bit of pressure will be applied, quite possibly without even the intention of the researcher. Confirmation bias is not necessarily "fraud" -- it is just bad science, science poorly done.
There is a move afoot to do something about this. We know that it happens. We know why it happens. We know a number of things that we can do to reduce the probability of it happening -- for example requiring the open publication of all data and methods contemporary with any paper produced from them, permitting absolutely anybody to look at them and see if t
Perhaps it is because the market has been resistant as all hell to moving past Windows XP because a) it worked; b) Vista didn't and cost MS all of its vendor street cred; c) Windows 7 wouldn't run on legacy hardware. At least not unlike a pig; d) legacy software wouldn't always run on Vista or 7 and not all vendors updated their software so that it would (or charged money for the update); the PC market grew suddenly enormously soggy as the desktop market evaporated, laptops became the new desktop (and got very, very cheap) and they had no damn MARGINS left to PAY for "innovation" -- and still pay for all of MS's "innovations" in the high cost of the OS; e) did I mention Android ate their lunch, Google surfaced as a thin appliance who-cares-what-your-OS is layer, and even linux started to come out of the closet?
No?
As I pointed out in the last post on this, this is all just Microsoft engaged in an elaborate form of Seppuku. They screwed Borland, Lotus, Corel, Netscape and got (mostly) away with it. They screwed IBM over OS/2 and got away with it. Now they are basically screwing all of the hardware manufacturers in the world, without a friend left in the software world either. Either they "become" Apple -- an odd goal given that Apple has remained competitive largely because of a handful of highly innovative hardware products of the sort one cannot guarantee finding every five years to keep your company afloat in the long run -- or they die.
I'm betting that they die, slowly and painfully, the way one does in fact die with one's guts spilled out through a self-inflicted wound.
On the good side of things, we have a very good chance of seeing OTC Linux and Android hardware from all of the big manufacturers, far better hardware support, and who knows -- maybe GAME manufacturers will read the handwriting on the wall. Will they still want to write for MS operating systems when MS sells both PC hardware and the Xbox and has a vested interest in eating their lunch? Will they still find a PC market when PC makers turn towards Linux and Android, respectively, to be able to compete with Apple and Microsoft? And actually win -- MS will have to PAY for its OS development on its own hardware platform, where linux development, highly distributed, is much, much cheaper.
Yeah, like what you said, unless you want to make it even pickier and phrase it like Bayes' Theorem itself. Interested parties are also strongly urged to read Richard Cox's monograph: The Algebra of Probable Inference and folks interested in the connection between all of this and both AI computation and real human intelligence might want to look at David Mackay's Information Theory, Pattern Recognition and Neural Networks, available free online for interested readers or for money for interested professionals (I bought it, it's worth it). Even without priors, though, the observation of correlation beyond what is expected from random chance makes all of the various causal hypotheses rather more likely than they were before making the observation because with no causal linkage at all one expect no statistical correlation at all. It's differentiating between them that's tough without priors.
That's the funny thing -- correlation is not causality, but it's all we've got, really. It's putting it to good use to build a network of (Bayesian) correlations that end up with consistency and perhaps even meaning that's the trick.
Let's count the problems with this model. Suppose I'm a maker of tablets or laptops or PCs. So far I've put Windows on them to market them because frankly, I have little choice -- it's that or Linux and besides, in order to remain price competitive I have to get the price breaks that come from Microsoft for using Windows exclusively as a pre-installed OS. However, I have taken comfort in the fact that all of the other manufacturers are in the same boat -- we all have the same product, within hardware choices and tweaks, we all run the same OS preinstalled for pretty much the same price, that OS breaks on our hardware about the same fraction of time and we get enough help fixing it that we can usually release a semi-stable product and not piss off our consumer base.
But now Microsoft is going to play! It will design its own hardware, and will apply its own team of umpty-gazillion semi-unemployed programmers to ensure that its OS works perfectly on that hardware, both optimized and with absolutely functional device drivers. OTOH, Microsoft will have absolutely no incentive to help out third party hardware manufacturers like me. Indeed, they will have a disincentive! If my hardware has a constant list of creeping minor problems, then Microsoft's huge team of sales reps will be able to convince many buyers that my hardware just isn't reliable, where theirs is!
It's not like we haven't seen this before, after all. It is precisely how Microsoft became the monopoly that they are today -- Microsoft branded software always worked when a new version of its OS was released, where non-Microsoft software was usually subtly broken for six months afterwards. This problem was so prevalent and reproducible that the term "FUD" was coined to describe the predictable response of the Microsoft reps in that six month window, while they gradually took over the world from the likes of Lotus, Corel, Borland, all of whom owned a serious piece of the PC software business before Microsoft decided it wanted it all, not just the OS and maybe a reference compiler or two.
So now Microsoft has decided to go one step further and become Apple, even though they at one point took Apple to within coup-de-grace range of bankruptcy and refrained from wiping it out entirely only because they were already in trouble with anti-trust suits and needed a viable competitor to convince the courts that they didn't, actually, need to be broken up. Apple, of course, has succeeded largely recently because they have a certain amount of genuine innovation (on top of a fair bit of righteous anti-innovation, adopting Unix as their basic OS and "inheriting" an enormous base of free/open source software that nevertheless becomes part of their overall offering). So Microsoft is thus committed to out-Jobbing the now deceased Steve Jobs, in spite of the fact that as a corporation it has stolen -- well, "hijacked" is a better term -- almost all of its best ideas using the dual weapon of cloning by the world's largest closed shop of programmers who control both software and OS, and FUD. Worked great with corporations and corporate tools, but how will it play with consumers? How will it play with the vast ocean of hardware makers?
The latter is fairly predictable. The minute Microsoft becomes Apple, and adopts hardware that it either makes itself or outsources from just one specific manufacturer, the incentives that have long given them dominance on the desktop disappear. True, they are already getting hammered by e.g. Android and can read the writing on this particular wall, but it may well be that their only alternative at this point is one form or another of elegant suicide. When all of the hardware manufacturers realize that they are competing with Microsoft as well as Apple, with very likely no room left in between, what alternatives do they have for survival? Anti-trust suits, sure. And support for any viable alte
It is a grave tragedy that people don't understand basic logic and the post hoc ergo propter hoc fallacy, or the statistical version of it as laid out in e.g. Jaynes Probability Theory, the Logic of Science.
Jaynes points out (and proves!) that while the discovery that A and B are correlated does not mean that A causes B or that increasing the prevalence of A in a population will increase the prevalence of B, it does make it more plausible that this is so, compared to the hypothesis that A and B are independent, depending on one's prior beliefs that are themselves statistical knowledge. His specific example, IIRC, is how a policeman observing a man standing in front of a broken window of a jewelry store, his pockets full of gems, is justified in inferring that the man in question is in fact robbing the store; even though there are many other possible explanations for the observed correlation between a broken window in a jewelry store and a man nearby with pockets full of jewelry, those explanations are all rather "special" and hence less probable on the basis of our prior knowledge of e.g. the probability of a completely innocent man walking by a jewelry store with pockets full of jewelry at a time when a construction crew happened to have broken the window.
So to analyze the correlation between belief systems in heaven and hell and crime rates is not easy. It is certainly a mistake to assume that A (belief in hell) is necessarily a control variable of B (tendency to commit a crime). On the other hand, it is perfectly reasonable to assert that it might be as a hypothesis, one made at least plausible by the observed correlation.
Reading TFA (at least) one can easily see a number of problems with the hypothesis even after their efforts to control for various other things. One is that the study does not, in fact, apply to "societies". It applies to countries. Countries are almost never monolithic societies, especially the larger countries, so if a country like the US or Russia or China is rated as "believing in hell more than believing in heaven" according to their criterion it has a disproportionate effect on the outcome at least as far as the population contribution to the study is concerned.
Second, I would argue that it is literally impossible to control for the effect of specific religions in smaller, more monolithic countries with a strong or dominant religious meme in their actual governance, specifically the Muslim countries. Islam has a powerful hell meme, so pretty much all Muslim countries are going to come down on the hellfire side of things. They also have (in many cases if not quite all) draconian punishments for crimes, oppression of religious and personal freedoms, internal and external warfare that doesn't count as a "crime" I'm quite certain even though it arguably causes far more death and mayhem than any police blotter. How does one control for sharia and wealth and draconian punishments and social history and war with only surveyed samples of belief and an outcome of national crime rates (which may or may not even be reported accurately)? Not easily, at the very least. The correlation could easily be an artifact, not causal at all.
Finally, in order to truly be meaningful, one would have to extend the result not to "countries" but to individuals. A country's crime rate is very probably determined by many, many factors, not the least of which is a social history that might well be correlated with prior or current religious belief without being caused by it (example being pretty much the entire New World, settled largely by people seeking religious freedom, social freedom, escape from European tyranny, or involuntary settlers forced here for economic reasons, and indigenous peoples with completely different belief systems that were upended and cast into social disarray). This does not mean that individuals within those countries are more likely to com
OK, sorry, perhaps I needed to encapsulate this SARCASM sufficiently clearly to keep literalists from taking it as anything but a wry but sad observation on the world.
The point still stands, however. It's a lot easier and cheaper to hold a gun to somebody's head (figuratively or literally) than it is to crack any of the halfway decent encryptions. The legal and court system has recourse to direct coercion in the form of contempt of court and jail -- they can compel you to reveal keys or throw you in jail indefinitely until you do. Terrorists obviously have even lower thresholds (and higher levels of coercion) to apply. NSA/CIA/FBI "Homeland security" have a dual track -- the legal one through contempt of court under a warrant, plus the covert ones you object to IN ADDITION to trying to crack the cipher. If the contents of the encrypted message are suspected kiddie porn, there is no harm done keeping the suspect in jail on contempt for eternity -- quite the contrary. If they are the location and time of the nuclear bomb placed by terrorists in downtown Manhattan -- well, that introduces a pretty serious ethical grey area. I won't presume to make this choice on the basis of a theoretical model; it's a judgement call.
(Off topic, of course, but still interesting in the sense that the lose-lose models of psychiatric choice are interesting -- do you sacrifice your own life and that of your child to save the lives of a hundred strangers, that sort of thing...)
Yeah, the public key exchange algorithms I've never felt entirely comfortable with anyway, as they are the kind of thing where a clever theorem (or a clever advance in computing) could significantly reduce the search time even classically without Shor's algorithm (although I think we are still pretty far away from being able to implement Shor's algorithm in a practical, stable, computation on an actual RSA keypair, which is why I discount it as an issue in any system they might build right now). But I was referring to AES without public/private keypairs -- straight up encryption -- with very large keys. From what I've read recently, AES is still "unbroken" in the sense that one cannot really significantly beat a brute force search time, not by the orders of magnitude needed to make an attack feasible (although they do whittle away on it, finding specific weaknesses in specific keyspaces or implementations).
With that said, I'd be interested in how you compute/infer the scaling of AES128 decryption by NSA now. 10^38 is still a pretty big number compared to even petacycles of CPU, and one can hardly check a key in a single CPU cycle with a single machine. Dieharder returns something like 3x10^7 rands per second from AES used straight up as a random number generator, or ballpark 30 nsec per number (close to 100 CPU cycles per number). Even allowing 10 nsec per key (IMO an order of magnitude low) that's only 3x10^15 keys per core per year. Divided into 10^38, "that's a lot of cores" (10^18 core-years of computation). Even if you turned every CPU on the planet to the task, in other words, and they could all test a key every 10 nsec, I think it would take a lot of years, and I think there are orders of magnitude being left out of the compute time per key in reality, especially when testing a key against the actual decryption of the message.
So am I leaving something out of this estimate? Is there a non-brute-force algorithm that can significantly -- as opposed to marginally, an order of magnitude or so -- beat simply serially testing keys? NSA folks are bright, I'm sure, but they're stuck in the same universe that the rest of us are in, where a computer can only execute order of a billion instructions per second per core, and that is way too few to make 128 good bits of entropy manageable with search algorithms that take as little as nanoseconds per key tested.
With that said, hey, I personally prefer 1024 bit encryption keys. Belt and suspenders, I always say. And I keep my kiddie porn collection under 4096 bits -- can't be too safe. Of course, that leaves me stuck with spending the rest of my life in jail under a contempt of court citation if I don't decrypt it on the demand of some judge anyway... the point being that the real risk isn't that NSA will secretly decrypt 128-bit well-encrypted messages, it is that someone will use physical, economic, or other forms of coercion to force the exposure of the key. Or that some dunderhead will enter a key consisting of his initials and birthday repeated N times -- witness the longstanding success of "crypt" at password guessing. Or that somebody will prove a theorem that reduces the time required to generate the missing member of a prime key pair by ten or twenty orders of magnitude and suddenly expose the soft underbelly of SSL, RSA, and the internet in general to every pimple faced kid on the planet with an ipad to use as "CPU".
Yeah, I'm pretty skeptical about that (and about quantum computing in general). I know a bit more than nothing there, and I agree, they're not exactly ready for prime time (and might be a decade away). Also, their primary benefit is in ENcryption, not DEcryption, and I don't think that requires a supercomputer. Decryption simply requires speed and memory. Enormous, lifetime-of-the-Universe size-of-the-Universe-to solve-the-problem quantities of both for files encrypted with a good algorithm and a non-trivial key -- in fact, I honestly don't think that anybody can brute-force decrypt a properly encrypted document, including the NSA. Or rather, the correct application of brute force is to the individual holding the key -- waterboarding or threatening to remove vital body parts one at a time, that sort of thing. Because without the key, searching the keyspace scales, shall we say, "badly" -- out to lifetime of the Universe sorts of times very quickly, even with massively parallel computers doing the search. 4096 bits is searching 2^4096 = 10^1233 possibilities, which means that if you took every elementary particle in the visible universe (say 10^90 of them) and turned it into a computer, each computer would only have to search 10^1143 possibilities. Given a generous 10^27 nanoseconds in the lifetime of the Universe so far, and assuming that it will live 1000 times longer for 10^30 nanoseconds of compute time in the entire thing, we're down to a mere 10^1113 lifetimes of the Universe in order to brute force search the 4096 bit space, assuming that the encryption algorithm generates that much entropy out of the key and assuming that each elementary particle can do the entire computation required to test a key in a nanosecond. The point being that no, the NSA cannot crack any well-encrypted document without the well-chosen key. Not even if a mere 1024 bit key is used. Not even if a 512 bit key is used, although there one is finally reaching the space where weaknesses in the encryption algorithm and clever tricks might sometimes yield a faster solution. Really, even a 128 bit key is pretty safe (over 10^38 keys to search, so a billion computers searching a key a nanosecond would still take billions of years to search a significant fraction of the keyspace).
Also, YMMV -- any given encryption routine COULD have a hidden weakness, because the random number generator at its heart really does have far, far less entropy than the key suggests, and there may well exist unknown (or known only to the NSA) but true theorems that permit the encryption to be cracked with many orders of magnitude less effort. Witness the WEP versus WPA debacle in wireless -- a weak algorithm can easily be cracked even if the search space is nominally too large.
Actually, the top 500 "competition/list" has been moderately useful for transitioning the world from big iron supercomputers sold by single corporations into the modern commodity/cluster (beowulf) model that costs far less and scales (as one can see) almost indefinitely large/fast for a certain class of linear or embarrassingly parallel problems. It is also the case that some of the problems that are solved using the larger of the computers built (which with the exception of corporate entries aren't really built "just for bragging rights") are both interesting and potentially of some value, either intellectual or monetary, to society. So it isn't, really, just a matter of bigger dicks (although the top ten does have a certain amount of that going on, where for decades some companies were conspicuous by their absence, and only got there eventually by basically building a machine with winning capacity and then giving it to somebody so that they could enter). Sometimes it is a matter of solving problems in nuclear physics or cosmology or cryptography or fluid dynamics that are NP complete or otherwise scale poorly enough that one is always hungry for cycles if one works in the field.
The question of whether or not the answers to those problems are worth the cost is a separate one, and by all means debate it, but be sure to do so in the context of all Big Instrumentation used in science. The LHC is a lot of money to -- maybe -- find the Higgs. Or not, again, to the tune of tens of billions of dollars. NASA routinely spends/spent tens of billions of dollars to lift humans and e.g. the Hubble into orbit -- the Hubble gives us enormous amounts of wonderful science but very little of that science is of direct (as opposed to indirect) benefit to (say) automobile mechanics, lawyers, owners of restaurants, farmers. The cost of a top 500 machine is in comparison cheap, and in some cases may even work on problems with a measurable expectation value that trickles back to the society that ultimately pays for it (outside of the noble cause of supporting the education and research system that has created a truly enormous amount of very concrete wealth by providing work for otherwise unemployed physicists and computer scientists and mathematicians and funding for the many science and math departments that trained them and whose faculty participate). Personally, I think it is well worth it, but I've spent a good fraction of my life attached to that particular teat (although I'm not, currently) and don't pretend to be completely objective here.
I am, OTOH, pretty well informed about cluster computing, while having absolutely no dog in the top 500 race.
In the cases where secrecy is probably preventing you from knowing about it, it probably is optimized for 32-bit precision floating point and/or large storage throughput to fuel data mining.
Or (in the case of NSA) decryption. There isn't a computer large enough to solve really difficult decryption problems, but whatever there is probably lives somewhere in the NSA, and is very likely very, very large.
Maybe not as large as Google's farm, though. Or even Amazon's.
Total lack of data for that statement. I'm willing to check out any support you have, but just as a warning, a 2 C change due to change in bond albedo is basically impossible just based on the temperature data we have.
You mean the Stefan-Boltzmann Law, used to determine the greybody temperature that is the base from which the Greenhouse Effect proceeds to warm the planet? Since the energy influx that has to be in balance with outgoing radiation is TOA insolation less radiation that is directly reflected to space due to albedo, raising albedo directly reduces the greybody temperature. It (as you can see, given a 7% modulation over 15 years as determined by two distinct NASA experiments that track it, one satellite based and the other the Earthlight project) is actually by far the largest direct modulator of expected surface temperatures and an absolutely trivial computation suffices to show that the 7% change translates into a baseline greybody temperature shift of roughly 2 K.
As for the other assertions, obviously we look at different graphs for sea ice -- the SH is over the 30 year mean and has been for a rather long time. The NH has been lower, but this winter meandered up well within a S.D. of the 30 year mean. If you google a bit, you can actually see the variation year by year over the last decade or more, all on one graph. And I'm not mistaking cycles for linear trends -- I'm saying that nobody knows why the albedo has increased, just as nobody knows why it was a minimum during the heating of the 80s and 90s. Oh, and while you're worrying about explaining how you can tell what is a linear trend and what is cyclic in the absence of any sort of serious baseline for data or workable theory, you might think about the NASA report that stratospheric H_2O has (again for unknown reasons) dropped by roughly 10% over the last five years. That has a strong net cooling effect too -- predictions (from the NASA papers) estimate roughly 0.5 K, which is interestingly on the same close order of as the total "warming" observed post 1945. One might be tempted to conclude that warming was correlated strongly with albedo variations and variations of stratospheric water vapor -- or not. But either way the physics of both is perfectly clear, and any halfway decent climate model that includes the measured albedo as a parameter should be showing strong cooling.
But they're not, even though this is bone-simple physics even more fundamental (and prior to) the GHE. I wonder why?
If anything, it has been slowing down over the last decade as global temperatures have stabilized, the net icepack (NH and SH) combined has actually grown, and even the NH ice coverage is within a fingernail's width of the thirty year mean.
People seem to be confusing the order in which science is done. Observations trump theory. When the theory is an elaborate one with many adjustable, essentially unknown parameters and little objective predictive skill, choosing to believe observational evidence instead of theoretical projection is sheer common sense. When (no matter what) the sea level isn't going to suddenly jump ten centimeters in a decade (where at most 1-2 cm is a lot more likely) spending massive amounts of money now to ameliorate what may never emerge as a problem later is again sheer common sense. In the meantime, the measured bond albedo of the Earth has increased by 7% over the last fifteen years, which corresponds to a roughly 2 C temperature drop due to reduced net insolation "off the top" as it were. This dwarfs the entire warming observed since the LIA. Just something to think about.
rgb (sitting on the NC coast, looking out the window at the water in Beaufort NC, where the tidal levels haven't significantly changed for years).
Actually, if you read TFA, it makes it quite clear that in this case, there is a clear correlation between theism and your choice of stupidity or ignorance (or both). Smart, well educated people (as identified by scholastic accomplishment) are far more likely to be atheist than stupid, poorly educated people. The latter are also more likely to be Republicans. Go figure.
As for "too stupid to vote" -- I generally agree that most cures are worse than the disease, and it is also absolutely true that there are plenty of people who have little formal education who are smart enough not to be suckers, matching well educated people who lack common sense.
In any event, education is the antidote for religion. It isn't a perfect cure (because there are limits on what you can do with the human material at hand, and because a large part of the brainwash- I mean "education" of the young is in the hands of religious parents, but statistically it ups the odds. And atheism has steadily grown over the last 30 years, at the expense of religion in general, just as Christianity in general has lost ground.
As I posted further down, I think I agree. Although I still don't know which problem it is that can't be solved that he solved -- I'm assuming linear drag forces, which should indeed be analytically solvable. It certainly is in one dimension.
But this does not (as I also note) really help, since almost nothing falls according to Stokes drag.
He worked out how to calculate exactly the path of a projectile under gravity and subject to air resistance.
What does this even mean? Linear (Stokes) drag forces (idealized)? Turbulent drag forces? Something in between? For a bluff body? A streamlined body? A tumbling body? In still air of uniform density? In a wind? In actual air that might well vary significantly in density and temperature along the (highly ballistic) trajectory?
I'm assuming that it isn't just Stokes drag, as that never struck me as being unsolvable (it certainly isn't vertically) or quadratic drag (also directly integrable vertically) so it must be one of the "interesting" cases, but TFA doesn't say. Not to take anything away from the young man in question, either -- I'm sure he's very bright and that his solutions are peachy-keen.
I am, however, having a hard time seeing how this will improve ballistics solutions in any case whatsoever compared to numerical solutions; ultimately one has to deal with real nonlinear fluid dynamics to solve almost any sort of less-idealized ballistics problem, the sort involving Navier-Stokes and solution spaces that haven't even been formally proven to exist yet. The idealized problems are good for understanding qualitative behavior, but not so good for quantitative prediction in all but very special cases. Computers really are going to win hands down in almost all problems one can pose in this general arena.
Actually, this is in part a question. Although I could probably read TFA to find out (maybe it answers this) but one issue with rejuvenation is rewinding the telomeres (so to speak) so they aren't too long (cancer) and aren't too short (premature cell death). There are a lot of clocks in cells and tissue, and my general question is: Does skin -> stem -> tissue cell reset them all so that the new tissue really is "like that of a newborn" or whatever with the original DNA complement?
Not good to be able to build the heart of a Chicken Little (props for ref:-) that simply up and dies on you, or gets cancer, for other reasons.
Did I mention that my name is Bob, my first Linuxoid operating system was Slackware, and I'm very definitely an ordained subgenius? Also, I used to smoke a pipe.
So send your money to me.
rgb
(More seriously I'm reading The Black Swan, by Nassim Nicholas Taleb. What a "Black Swan" an Earth-scouring solar flare would be! And one in 2012, too. Those pesky Mayans.
One is also seriously reminded of a Larry Niven short story, but I can't remember the name, am not at home near my bookshelves, and am way too lazy to look it up. But it all starts with the full moon rising and becoming very, very bright, signalling the sequential extinction of land/surface life as the planet rotates. How would you spend your last hours?
The interesting story here is state police cars with built in radioactivity detectors, obviously either checking for dirty radioactive weapons, nuclear weapons or newly arrived aliens hot off the star ships skulking about in their human skin suits;).
Precisely. I did not know that. Not only built with radioactivity detectors, but ones that are on all the time and are damn sensitive if they are picking up the excess flux from a human inside a car from a tracer treatment administered presumably some nontrivial amount of time before from a distance of what -- 7 meters? 10? 20? -- while driving down the road. Tracers are often very short half-life elements -- that's why they use them -- lots or radioactivity but for a very short time. They tend to be produced in the hospital immediately before use and be mostly gone an hour or two later (but with an exponential tail). Clearly they nailed him right after he left the hospital, and he left the hospital rather quickly after the test, probably less than 45 minutes after the production of the tracer.
Are they sensitive enough to pick up a nuclear bomb being transported? Not if it is made with bomb-grade Uranium, which is also the easiest thing to make a bomb out of, but which isn't radioactive, although you might pick up the trigger. Plutonium 239 IS radioactive, producing a 5 MeV or so alpha at a rate sufficient to keep Plutonium warm to the touch, but alpha particles are relatively easy to block. It also typically contains Pu 240, which spontaneously fissions and produces a surplus flux of a few ~10 million neutrons per second from a typical core. Neutrons are more difficult to stop, but the intensity diminishes like 1/(4\pi r^2) so that the intensity at 10 meters is ~10^7/1250 or around 10^4 per meter squared per second. A detector as large as 1cm x 10 cm would then pick up 10 surplus neutrons per second at 10 meters, assuming there was zero attenuation in between and a perfect detector, neither of which is true. MAYBE this would give them signal to noise of a decibel or two, but given detector efficiency probably not until you were much closer. Up close it would be better, of course. Presumably their detectors have some sort of built in discriminator looking for sustained signal to noise above some cut-off.
What the patient was probably emitting is gamma. Gamma radiation has a long range and isn't easily blocked.
Actually, you need to read "The Black Swan" by N. N. Taleb. Science that tries to confirm a theory is already infected with confirmation bias. There are a pile of examples that demonstrate the fallacy of confirmatory inference. Taleb uses a variant of Bertrand Russell's -- a turkey might reasonably infer, based on his daily experience, that humans exist for the sole purpose of feeding him, caring for him, providing for his every need. This might go on for day after day, increasing the turkey's degree of belief in his hypothesis of a good and beneficent humanity filled with love of turkeys, right up to the day that -- ulp -- something unexpected happens.
I'm not a Popperite, rather a Jaynes-Cox-Bayesian, but nevertheless it is important to avoid confounding the relative strength of positive and negative evidence. Absence of evidence is not the same as evidence of absence, yet we almost invariably confound the two.
Taleb damn skippy agrees with you about publicity, however, and the near-criminality of publicity and reporting of science. A newspaper necessarily takes a scientific result or observation and transforms it two ways: First of all, it creates a narrative. It isn't just "a tornado hit Houston", but "a tornado hit Houston, possibly caused by Anthropogenic Global Warming" with the subtext "this isn't an act of nature, random an unpredictable, but is instead our fault". Aztec priests couldn't have come up with a better excuse for ripping the still beating hearts out of a stream of slaves and war captives. Second, it necessarily reduces the complexity of the result to no more than three variables, ideally one. It "Platonifies" it (according to Taleb) -- wraps it up in a pretty, easy to understand package that makes it more predictable, less random than it really was. Global warming is a much simpler "cause" than "A cold front overrunning a warm wet surface layer of air near the ground, creating turbulent rolls that break off and terminate on the ground, sustained and driven by the thermal difference, and it is a better story too.
Sadly, as you point out, real science is all too often (and should be) scientist z looked at something and didn't find much. But what they failed to find and how they looked is actually often as or more important than a study that claims to find something, especially when the latter uses questionable methodology to try to prove something, cherrypicks data (for the same purpose), ignores silent evidence (ditto) etc. Medical science is permeated with this. Nobody gets famous, or rich, or even a job, for looking for a cure for cancer and not finding one. This too is addressed by Taleb. Great book.
Well, yeah, except that all of the bending occurs at the interface surface. So in principle, one could stack 150 lensing surfaces constructed Fresnel-style and bring that right down to a kilometer. Depending on what the "interface surface" is for gamma rays. Or, stack 1500 of them and bring it down to 100 meters. Or stack 15,000 of them, in a 3D structure created with e.g. molecular beam epitaxy, and bring it down to 10 meters (with a lens with a total length of perhaps a meter). At that point it is conceivable that it might be useful, although probably not as an optical grade lens (wouldn't it be uber cool to build a gamma ray telescope with an aperture lens a meter across? Sure it would!)
I think that a lot of these same principles are involved in building x-ray lenses -- the lens is less like a glass lens, more like building an interferometric scattering array that causes a single central primary peak. But not my specialty, just thinking out loud...
Slashdot Mirror
Economics is one of the worst offenders, agreed, and the primary target of The Black Swan. But a number of other fields, including some that are supposedly in mainstream science, are almost as bad. And not necessarily intentionally. The saddest thing is that the researchers themselves are often totally oblivious as to just how biased and/or weakly founded their own results are, because they always get them "using statistics" in a traditional way, not realizing that they are using Statistics 101 (freshman intro stats) where what is required is Statics 404 (a graduate course and not for the faint of heart or incompetent in calculus, ODES, combinatorics, and a certain amount of common sense and experience).
Precisely. Indeed, a major part of the solution is to make scientists their own greatest skeptic, to mistrust our own pet ideas, to hesitate to claim "proof" to a fault even if evidence or model computations seem to support it.
The latter are an entire category in and of themselves. I "do" predictive modelling and moderately advanced statistics on a professional basis, and even have a patent pending in the field. I've done Monte Carlo computations in physics for well over a decade, and know a lot about randomness and hypothesis testing compared to your average scientist in the street, so to speak. I am all too aware that model computations are among the least trustworthy kinds of evidence and usually have far less predictive power than that which is claimed by the modeler. The problem there is subtle and related to complexity and nonlinearity. A highly multivariate, semi-empirical, nonlinear theory implemented as a model is often implemented by "fitting" some or all of its parameters to some (sub)set of data. This in turn is often equivalent in modelspeak to using hill-climbing (gradient search) to find an optimum fit to the data relative to some selected parametric starting point (this is sometimes referred to as making a "Bayesian" choice of the parameters based on some set of data used as priors").
There are many problems with this. One is the problem of omitted variables. In many of the problems where this is done, the choice of parameters (dimensions in the parametric space) is highly model dependent. Heuristics are often used to limit the size of the parametric space simply because doing anything in a really high dimensional space is a lot of work and introduces a substantially higher (but honest) estimate for errors in the final result. Heuristics, of course, is code for "I don't think these variables will significantly contribute", an open opportunity to omit variables that you don't want to be significant because they confound your hypothesis. A second problem is that many models are de facto parametric nonlinear function approximators. This means that -- especially if the data being fit to the parameters is "simple", e.g. monotonic or otherwise simply nonlinear over the range of the fit -- it is often perfectly easy to fit the data with a set of parameters, have the fit be "optimal", have the fit produce a perfectly reasonable chisq, and have the parametric fit be perfectly meaningless. This is all elementary modeling theory 101, but somehow "hiding" the basis by turning it into the solution of a set of coupled ordinary differential equations with nearly e.g. sinusoidal or nearly e.g. polynomial or exponential behavior makes the problem somehow disappear in the minds of the modellers. A third problem is that of complexity -- in many cases (especially for highly multivariate nonlinear models) there may be many local optima and hill-climbing from a selected starting point can easily be yet another form of inadvertent confirmation bias. A global search might find a better optimum, or might reveal that there are several constellations of parameters (especially when omitted variables are included) that can fit the empirical data within its precision, (properly) reducing confidence in the final prediction. A fourth is that even a model built with the variables that were important in the past (where the data being fit resides), that is robust in any parametric/Bayesian search, that uses any of several methods to "validate" the model (using past data) can easily fail in the future because the model simply does not extrapolate. The real space of variables and data is much larger, the model is always being built on some sort of optimistic projection onto a manageable subspace, and ignored stuff eventually becomes important and causes a complete deviation from the model. Chaotic models, stiff differential models -- there is no lack of examples, but somehow this sort of thing doesn't get factor
I disagree about the rarity, on the basis of empirical evidence, for example the recent paper in Science (IIRC, sorry I don't have the reference handy) in which a cancer researcher failed to replicate 46 out of 53 papers published -- all of them with peer review -- prior to embarking on new research in the field. Similar meta-studies have turned up astounding rates of non-reproducible results in other fields (some more than others -- sociology and IIRC social psychology topping the non-medical list).
One problem is that we have constructed a system that rewards the publication of positive results and punishes negative results published or unpublished. Punishes as in makes or breaks the entire career of young researchers, if the negative result occurs when they are up for tenure. Rewards as in ensures research funding and professional advancement as long as positive results keep flowing out.
Another fundamental problem that peer review has a terrible time with is confirmation bias. Science in general has a serious problem with confirmation bias. If one ever embarks on a study where one seeks evidence for some causal linkage associated with some phenomenon in a general population where the phenomenon occurs, one can always find exemplars that support your hypothesis. Lacking actual work to replicate your results using sound methodology (e.g. double blinded and/or conducted using competent statistical analysis, something still as rare as hen's teeth in science in general because to it is difficult to do statistics correctly in a complex problem, not easy, and certainly not easy as in covered in one or two undergrad stats courses which is all that it is probable that the researcher has ever taken) confirmation bias can not only worm its way into the literature, it can come to dominate entire fields as a significant fraction of scientists who do the reviewing for both publication and grants are "descended" from one or two original researchers and their papers. It can take decades for this to be discovered and work out in the wash.
Peer review works better in some disciplines than others. Math it works well, because there is literally nothing up a publisher's sleeve -- fraudulent publication is indeed impossible and even mistaken publication is relatively rare and conditional on involving math so difficult even the reviewers have a hard time following it. Physics and the very hard sciences are also fortunate in that it works decently (although less perfectly), at least where there is competition and the proper critical/skeptical eye applied to results new and old. At least there a mix of laboratory replication and strong requirements of consistency usually keep one out of the worst trouble.
A simple rule of thumb is: The more a result relies on population studies, especially ones conducted with any kind of selection process or worse selection process plus the actual modification of the data according to some heuristic or correction process, where the study itself is conducted from the beginning to confirm some given hypothesis, the more likely it is that the result (when published) is bullshit that will eventually, possibly decades later, turn out to be completely wrong. If you have enough places for a thumb to be subtly placed on the scales and the owner of the thumb has any sort of vested or open interest in the outcome, it is even odds or better that a teensy bit of pressure will be applied, quite possibly without even the intention of the researcher. Confirmation bias is not necessarily "fraud" -- it is just bad science, science poorly done.
There is a move afoot to do something about this. We know that it happens. We know why it happens. We know a number of things that we can do to reduce the probability of it happening -- for example requiring the open publication of all data and methods contemporary with any paper produced from them, permitting absolutely anybody to look at them and see if t
Raccoons get rabes? Who knew? Now I have to wear thick padded pants on top of my combination faraday cage/sharkproof mail to go to the store.
Damn you, AC!
rgb
Perhaps it is because the market has been resistant as all hell to moving past Windows XP because a) it worked; b) Vista didn't and cost MS all of its vendor street cred; c) Windows 7 wouldn't run on legacy hardware. At least not unlike a pig; d) legacy software wouldn't always run on Vista or 7 and not all vendors updated their software so that it would (or charged money for the update); the PC market grew suddenly enormously soggy as the desktop market evaporated, laptops became the new desktop (and got very, very cheap) and they had no damn MARGINS left to PAY for "innovation" -- and still pay for all of MS's "innovations" in the high cost of the OS; e) did I mention Android ate their lunch, Google surfaced as a thin appliance who-cares-what-your-OS is layer, and even linux started to come out of the closet?
No?
As I pointed out in the last post on this, this is all just Microsoft engaged in an elaborate form of Seppuku. They screwed Borland, Lotus, Corel, Netscape and got (mostly) away with it. They screwed IBM over OS/2 and got away with it. Now they are basically screwing all of the hardware manufacturers in the world, without a friend left in the software world either. Either they "become" Apple -- an odd goal given that Apple has remained competitive largely because of a handful of highly innovative hardware products of the sort one cannot guarantee finding every five years to keep your company afloat in the long run -- or they die.
I'm betting that they die, slowly and painfully, the way one does in fact die with one's guts spilled out through a self-inflicted wound.
On the good side of things, we have a very good chance of seeing OTC Linux and Android hardware from all of the big manufacturers, far better hardware support, and who knows -- maybe GAME manufacturers will read the handwriting on the wall. Will they still want to write for MS operating systems when MS sells both PC hardware and the Xbox and has a vested interest in eating their lunch? Will they still find a PC market when PC makers turn towards Linux and Android, respectively, to be able to compete with Apple and Microsoft? And actually win -- MS will have to PAY for its OS development on its own hardware platform, where linux development, highly distributed, is much, much cheaper.
rgb
Yeah, like what you said, unless you want to make it even pickier and phrase it like Bayes' Theorem itself. Interested parties are also strongly urged to read Richard Cox's monograph: The Algebra of Probable Inference and folks interested in the connection between all of this and both AI computation and real human intelligence might want to look at David Mackay's Information Theory, Pattern Recognition and Neural Networks , available free online for interested readers or for money for interested professionals (I bought it, it's worth it). Even without priors, though, the observation of correlation beyond what is expected from random chance makes all of the various causal hypotheses rather more likely than they were before making the observation because with no causal linkage at all one expect no statistical correlation at all. It's differentiating between them that's tough without priors.
That's the funny thing -- correlation is not causality, but it's all we've got, really. It's putting it to good use to build a network of (Bayesian) correlations that end up with consistency and perhaps even meaning that's the trick.
rgb
...right down to the bottom of the sea.
Let's count the problems with this model. Suppose I'm a maker of tablets or laptops or PCs. So far I've put Windows on them to market them because frankly, I have little choice -- it's that or Linux and besides, in order to remain price competitive I have to get the price breaks that come from Microsoft for using Windows exclusively as a pre-installed OS. However, I have taken comfort in the fact that all of the other manufacturers are in the same boat -- we all have the same product, within hardware choices and tweaks, we all run the same OS preinstalled for pretty much the same price, that OS breaks on our hardware about the same fraction of time and we get enough help fixing it that we can usually release a semi-stable product and not piss off our consumer base.
But now Microsoft is going to play! It will design its own hardware, and will apply its own team of umpty-gazillion semi-unemployed programmers to ensure that its OS works perfectly on that hardware, both optimized and with absolutely functional device drivers. OTOH, Microsoft will have absolutely no incentive to help out third party hardware manufacturers like me. Indeed, they will have a disincentive! If my hardware has a constant list of creeping minor problems, then Microsoft's huge team of sales reps will be able to convince many buyers that my hardware just isn't reliable, where theirs is!
It's not like we haven't seen this before, after all. It is precisely how Microsoft became the monopoly that they are today -- Microsoft branded software always worked when a new version of its OS was released, where non-Microsoft software was usually subtly broken for six months afterwards. This problem was so prevalent and reproducible that the term "FUD" was coined to describe the predictable response of the Microsoft reps in that six month window, while they gradually took over the world from the likes of Lotus, Corel, Borland, all of whom owned a serious piece of the PC software business before Microsoft decided it wanted it all, not just the OS and maybe a reference compiler or two.
So now Microsoft has decided to go one step further and become Apple, even though they at one point took Apple to within coup-de-grace range of bankruptcy and refrained from wiping it out entirely only because they were already in trouble with anti-trust suits and needed a viable competitor to convince the courts that they didn't, actually, need to be broken up. Apple, of course, has succeeded largely recently because they have a certain amount of genuine innovation (on top of a fair bit of righteous anti-innovation, adopting Unix as their basic OS and "inheriting" an enormous base of free/open source software that nevertheless becomes part of their overall offering). So Microsoft is thus committed to out-Jobbing the now deceased Steve Jobs, in spite of the fact that as a corporation it has stolen -- well, "hijacked" is a better term -- almost all of its best ideas using the dual weapon of cloning by the world's largest closed shop of programmers who control both software and OS, and FUD. Worked great with corporations and corporate tools, but how will it play with consumers? How will it play with the vast ocean of hardware makers?
The latter is fairly predictable. The minute Microsoft becomes Apple, and adopts hardware that it either makes itself or outsources from just one specific manufacturer, the incentives that have long given them dominance on the desktop disappear. True, they are already getting hammered by e.g. Android and can read the writing on this particular wall, but it may well be that their only alternative at this point is one form or another of elegant suicide. When all of the hardware manufacturers realize that they are competing with Microsoft as well as Apple, with very likely no room left in between, what alternatives do they have for survival? Anti-trust suits, sure. And support for any viable alte
It is a grave tragedy that people don't understand basic logic and the post hoc ergo propter hoc fallacy, or the statistical version of it as laid out in e.g. Jaynes Probability Theory, the Logic of Science . Jaynes points out (and proves!) that while the discovery that A and B are correlated does not mean that A causes B or that increasing the prevalence of A in a population will increase the prevalence of B, it does make it more plausible that this is so, compared to the hypothesis that A and B are independent, depending on one's prior beliefs that are themselves statistical knowledge. His specific example, IIRC, is how a policeman observing a man standing in front of a broken window of a jewelry store, his pockets full of gems, is justified in inferring that the man in question is in fact robbing the store; even though there are many other possible explanations for the observed correlation between a broken window in a jewelry store and a man nearby with pockets full of jewelry, those explanations are all rather "special" and hence less probable on the basis of our prior knowledge of e.g. the probability of a completely innocent man walking by a jewelry store with pockets full of jewelry at a time when a construction crew happened to have broken the window.
So to analyze the correlation between belief systems in heaven and hell and crime rates is not easy. It is certainly a mistake to assume that A (belief in hell) is necessarily a control variable of B (tendency to commit a crime). On the other hand, it is perfectly reasonable to assert that it might be as a hypothesis, one made at least plausible by the observed correlation.
Reading TFA (at least) one can easily see a number of problems with the hypothesis even after their efforts to control for various other things. One is that the study does not, in fact, apply to "societies". It applies to countries. Countries are almost never monolithic societies, especially the larger countries, so if a country like the US or Russia or China is rated as "believing in hell more than believing in heaven" according to their criterion it has a disproportionate effect on the outcome at least as far as the population contribution to the study is concerned.
Second, I would argue that it is literally impossible to control for the effect of specific religions in smaller, more monolithic countries with a strong or dominant religious meme in their actual governance, specifically the Muslim countries. Islam has a powerful hell meme, so pretty much all Muslim countries are going to come down on the hellfire side of things. They also have (in many cases if not quite all) draconian punishments for crimes, oppression of religious and personal freedoms, internal and external warfare that doesn't count as a "crime" I'm quite certain even though it arguably causes far more death and mayhem than any police blotter. How does one control for sharia and wealth and draconian punishments and social history and war with only surveyed samples of belief and an outcome of national crime rates (which may or may not even be reported accurately)? Not easily, at the very least. The correlation could easily be an artifact, not causal at all.
Finally, in order to truly be meaningful, one would have to extend the result not to "countries" but to individuals. A country's crime rate is very probably determined by many, many factors, not the least of which is a social history that might well be correlated with prior or current religious belief without being caused by it (example being pretty much the entire New World, settled largely by people seeking religious freedom, social freedom, escape from European tyranny, or involuntary settlers forced here for economic reasons, and indigenous peoples with completely different belief systems that were upended and cast into social disarray). This does not mean that individuals within those countries are more likely to com
OK, sorry, perhaps I needed to encapsulate this SARCASM sufficiently clearly to keep literalists from taking it as anything but a wry but sad observation on the world.
The point still stands, however. It's a lot easier and cheaper to hold a gun to somebody's head (figuratively or literally) than it is to crack any of the halfway decent encryptions. The legal and court system has recourse to direct coercion in the form of contempt of court and jail -- they can compel you to reveal keys or throw you in jail indefinitely until you do. Terrorists obviously have even lower thresholds (and higher levels of coercion) to apply. NSA/CIA/FBI "Homeland security" have a dual track -- the legal one through contempt of court under a warrant, plus the covert ones you object to IN ADDITION to trying to crack the cipher. If the contents of the encrypted message are suspected kiddie porn, there is no harm done keeping the suspect in jail on contempt for eternity -- quite the contrary. If they are the location and time of the nuclear bomb placed by terrorists in downtown Manhattan -- well, that introduces a pretty serious ethical grey area. I won't presume to make this choice on the basis of a theoretical model; it's a judgement call.
(Off topic, of course, but still interesting in the sense that the lose-lose models of psychiatric choice are interesting -- do you sacrifice your own life and that of your child to save the lives of a hundred strangers, that sort of thing...)
rgb
Yeah, the public key exchange algorithms I've never felt entirely comfortable with anyway, as they are the kind of thing where a clever theorem (or a clever advance in computing) could significantly reduce the search time even classically without Shor's algorithm (although I think we are still pretty far away from being able to implement Shor's algorithm in a practical, stable, computation on an actual RSA keypair, which is why I discount it as an issue in any system they might build right now). But I was referring to AES without public/private keypairs -- straight up encryption -- with very large keys. From what I've read recently, AES is still "unbroken" in the sense that one cannot really significantly beat a brute force search time, not by the orders of magnitude needed to make an attack feasible (although they do whittle away on it, finding specific weaknesses in specific keyspaces or implementations).
With that said, I'd be interested in how you compute/infer the scaling of AES128 decryption by NSA now. 10^38 is still a pretty big number compared to even petacycles of CPU, and one can hardly check a key in a single CPU cycle with a single machine. Dieharder returns something like 3x10^7 rands per second from AES used straight up as a random number generator, or ballpark 30 nsec per number (close to 100 CPU cycles per number). Even allowing 10 nsec per key (IMO an order of magnitude low) that's only 3x10^15 keys per core per year. Divided into 10^38, "that's a lot of cores" (10^18 core-years of computation). Even if you turned every CPU on the planet to the task, in other words, and they could all test a key every 10 nsec, I think it would take a lot of years, and I think there are orders of magnitude being left out of the compute time per key in reality, especially when testing a key against the actual decryption of the message.
So am I leaving something out of this estimate? Is there a non-brute-force algorithm that can significantly -- as opposed to marginally, an order of magnitude or so -- beat simply serially testing keys? NSA folks are bright, I'm sure, but they're stuck in the same universe that the rest of us are in, where a computer can only execute order of a billion instructions per second per core, and that is way too few to make 128 good bits of entropy manageable with search algorithms that take as little as nanoseconds per key tested.
With that said, hey, I personally prefer 1024 bit encryption keys. Belt and suspenders, I always say. And I keep my kiddie porn collection under 4096 bits -- can't be too safe. Of course, that leaves me stuck with spending the rest of my life in jail under a contempt of court citation if I don't decrypt it on the demand of some judge anyway... the point being that the real risk isn't that NSA will secretly decrypt 128-bit well-encrypted messages, it is that someone will use physical, economic, or other forms of coercion to force the exposure of the key. Or that some dunderhead will enter a key consisting of his initials and birthday repeated N times -- witness the longstanding success of "crypt" at password guessing. Or that somebody will prove a theorem that reduces the time required to generate the missing member of a prime key pair by ten or twenty orders of magnitude and suddenly expose the soft underbelly of SSL, RSA, and the internet in general to every pimple faced kid on the planet with an ipad to use as "CPU".
rgb
Yeah, I'm pretty skeptical about that (and about quantum computing in general). I know a bit more than nothing there, and I agree, they're not exactly ready for prime time (and might be a decade away). Also, their primary benefit is in ENcryption, not DEcryption, and I don't think that requires a supercomputer. Decryption simply requires speed and memory. Enormous, lifetime-of-the-Universe size-of-the-Universe-to solve-the-problem quantities of both for files encrypted with a good algorithm and a non-trivial key -- in fact, I honestly don't think that anybody can brute-force decrypt a properly encrypted document, including the NSA. Or rather, the correct application of brute force is to the individual holding the key -- waterboarding or threatening to remove vital body parts one at a time, that sort of thing. Because without the key, searching the keyspace scales, shall we say, "badly" -- out to lifetime of the Universe sorts of times very quickly, even with massively parallel computers doing the search. 4096 bits is searching 2^4096 = 10^1233 possibilities, which means that if you took every elementary particle in the visible universe (say 10^90 of them) and turned it into a computer, each computer would only have to search 10^1143 possibilities. Given a generous 10^27 nanoseconds in the lifetime of the Universe so far, and assuming that it will live 1000 times longer for 10^30 nanoseconds of compute time in the entire thing, we're down to a mere 10^1113 lifetimes of the Universe in order to brute force search the 4096 bit space, assuming that the encryption algorithm generates that much entropy out of the key and assuming that each elementary particle can do the entire computation required to test a key in a nanosecond. The point being that no, the NSA cannot crack any well-encrypted document without the well-chosen key. Not even if a mere 1024 bit key is used. Not even if a 512 bit key is used, although there one is finally reaching the space where weaknesses in the encryption algorithm and clever tricks might sometimes yield a faster solution. Really, even a 128 bit key is pretty safe (over 10^38 keys to search, so a billion computers searching a key a nanosecond would still take billions of years to search a significant fraction of the keyspace).
Also, YMMV -- any given encryption routine COULD have a hidden weakness, because the random number generator at its heart really does have far, far less entropy than the key suggests, and there may well exist unknown (or known only to the NSA) but true theorems that permit the encryption to be cracked with many orders of magnitude less effort. Witness the WEP versus WPA debacle in wireless -- a weak algorithm can easily be cracked even if the search space is nominally too large.
rgb
Actually, the top 500 "competition/list" has been moderately useful for transitioning the world from big iron supercomputers sold by single corporations into the modern commodity/cluster (beowulf) model that costs far less and scales (as one can see) almost indefinitely large/fast for a certain class of linear or embarrassingly parallel problems. It is also the case that some of the problems that are solved using the larger of the computers built (which with the exception of corporate entries aren't really built "just for bragging rights") are both interesting and potentially of some value, either intellectual or monetary, to society. So it isn't, really, just a matter of bigger dicks (although the top ten does have a certain amount of that going on, where for decades some companies were conspicuous by their absence, and only got there eventually by basically building a machine with winning capacity and then giving it to somebody so that they could enter). Sometimes it is a matter of solving problems in nuclear physics or cosmology or cryptography or fluid dynamics that are NP complete or otherwise scale poorly enough that one is always hungry for cycles if one works in the field.
The question of whether or not the answers to those problems are worth the cost is a separate one, and by all means debate it, but be sure to do so in the context of all Big Instrumentation used in science. The LHC is a lot of money to -- maybe -- find the Higgs. Or not, again, to the tune of tens of billions of dollars. NASA routinely spends/spent tens of billions of dollars to lift humans and e.g. the Hubble into orbit -- the Hubble gives us enormous amounts of wonderful science but very little of that science is of direct (as opposed to indirect) benefit to (say) automobile mechanics, lawyers, owners of restaurants, farmers. The cost of a top 500 machine is in comparison cheap, and in some cases may even work on problems with a measurable expectation value that trickles back to the society that ultimately pays for it (outside of the noble cause of supporting the education and research system that has created a truly enormous amount of very concrete wealth by providing work for otherwise unemployed physicists and computer scientists and mathematicians and funding for the many science and math departments that trained them and whose faculty participate). Personally, I think it is well worth it, but I've spent a good fraction of my life attached to that particular teat (although I'm not, currently) and don't pretend to be completely objective here.
I am, OTOH, pretty well informed about cluster computing, while having absolutely no dog in the top 500 race.
rgb
In the cases where secrecy is probably preventing you from knowing about it, it probably is optimized for 32-bit precision floating point and/or large storage throughput to fuel data mining.
Or (in the case of NSA) decryption. There isn't a computer large enough to solve really difficult decryption problems, but whatever there is probably lives somewhere in the NSA, and is very likely very, very large.
Maybe not as large as Google's farm, though. Or even Amazon's.
rgb
Total lack of data for that statement. I'm willing to check out any support you have, but just as a warning, a 2 C change due to change in bond albedo is basically impossible just based on the temperature data we have.
You mean the Stefan-Boltzmann Law, used to determine the greybody temperature that is the base from which the Greenhouse Effect proceeds to warm the planet? Since the energy influx that has to be in balance with outgoing radiation is TOA insolation less radiation that is directly reflected to space due to albedo, raising albedo directly reduces the greybody temperature. It (as you can see, given a 7% modulation over 15 years as determined by two distinct NASA experiments that track it, one satellite based and the other the Earthlight project) is actually by far the largest direct modulator of expected surface temperatures and an absolutely trivial computation suffices to show that the 7% change translates into a baseline greybody temperature shift of roughly 2 K.
As for the other assertions, obviously we look at different graphs for sea ice -- the SH is over the 30 year mean and has been for a rather long time. The NH has been lower, but this winter meandered up well within a S.D. of the 30 year mean. If you google a bit, you can actually see the variation year by year over the last decade or more, all on one graph. And I'm not mistaking cycles for linear trends -- I'm saying that nobody knows why the albedo has increased, just as nobody knows why it was a minimum during the heating of the 80s and 90s. Oh, and while you're worrying about explaining how you can tell what is a linear trend and what is cyclic in the absence of any sort of serious baseline for data or workable theory, you might think about the NASA report that stratospheric H_2O has (again for unknown reasons) dropped by roughly 10% over the last five years. That has a strong net cooling effect too -- predictions (from the NASA papers) estimate roughly 0.5 K, which is interestingly on the same close order of as the total "warming" observed post 1945. One might be tempted to conclude that warming was correlated strongly with albedo variations and variations of stratospheric water vapor -- or not. But either way the physics of both is perfectly clear, and any halfway decent climate model that includes the measured albedo as a parameter should be showing strong cooling.
But they're not, even though this is bone-simple physics even more fundamental (and prior to) the GHE. I wonder why?
rgb
Accelerating? Are we speaking of the same ocean?
If anything, it has been slowing down over the last decade as global temperatures have stabilized, the net icepack (NH and SH) combined has actually grown, and even the NH ice coverage is within a fingernail's width of the thirty year mean.
People seem to be confusing the order in which science is done. Observations trump theory. When the theory is an elaborate one with many adjustable, essentially unknown parameters and little objective predictive skill, choosing to believe observational evidence instead of theoretical projection is sheer common sense. When (no matter what) the sea level isn't going to suddenly jump ten centimeters in a decade (where at most 1-2 cm is a lot more likely) spending massive amounts of money now to ameliorate what may never emerge as a problem later is again sheer common sense. In the meantime, the measured bond albedo of the Earth has increased by 7% over the last fifteen years, which corresponds to a roughly 2 C temperature drop due to reduced net insolation "off the top" as it were. This dwarfs the entire warming observed since the LIA. Just something to think about.
rgb (sitting on the NC coast, looking out the window at the water in Beaufort NC, where the tidal levels haven't significantly changed for years).
Actually, if you read TFA, it makes it quite clear that in this case, there is a clear correlation between theism and your choice of stupidity or ignorance (or both). Smart, well educated people (as identified by scholastic accomplishment) are far more likely to be atheist than stupid, poorly educated people. The latter are also more likely to be Republicans. Go figure.
As for "too stupid to vote" -- I generally agree that most cures are worse than the disease, and it is also absolutely true that there are plenty of people who have little formal education who are smart enough not to be suckers, matching well educated people who lack common sense.
In any event, education is the antidote for religion. It isn't a perfect cure (because there are limits on what you can do with the human material at hand, and because a large part of the brainwash- I mean "education" of the young is in the hands of religious parents, but statistically it ups the odds. And atheism has steadily grown over the last 30 years, at the expense of religion in general, just as Christianity in general has lost ground.
rgb
As I posted further down, I think I agree. Although I still don't know which problem it is that can't be solved that he solved -- I'm assuming linear drag forces, which should indeed be analytically solvable. It certainly is in one dimension.
But this does not (as I also note) really help, since almost nothing falls according to Stokes drag.
rgb
He worked out how to calculate exactly the path of a projectile under gravity and subject to air resistance.
What does this even mean? Linear (Stokes) drag forces (idealized)? Turbulent drag forces? Something in between? For a bluff body? A streamlined body? A tumbling body? In still air of uniform density? In a wind? In actual air that might well vary significantly in density and temperature along the (highly ballistic) trajectory?
I'm assuming that it isn't just Stokes drag, as that never struck me as being unsolvable (it certainly isn't vertically) or quadratic drag (also directly integrable vertically) so it must be one of the "interesting" cases, but TFA doesn't say. Not to take anything away from the young man in question, either -- I'm sure he's very bright and that his solutions are peachy-keen.
I am, however, having a hard time seeing how this will improve ballistics solutions in any case whatsoever compared to numerical solutions; ultimately one has to deal with real nonlinear fluid dynamics to solve almost any sort of less-idealized ballistics problem, the sort involving Navier-Stokes and solution spaces that haven't even been formally proven to exist yet. The idealized problems are good for understanding qualitative behavior, but not so good for quantitative prediction in all but very special cases. Computers really are going to win hands down in almost all problems one can pose in this general arena.
rgb
Actually, this is in part a question. Although I could probably read TFA to find out (maybe it answers this) but one issue with rejuvenation is rewinding the telomeres (so to speak) so they aren't too long (cancer) and aren't too short (premature cell death). There are a lot of clocks in cells and tissue, and my general question is: Does skin -> stem -> tissue cell reset them all so that the new tissue really is "like that of a newborn" or whatever with the original DNA complement?
Not good to be able to build the heart of a Chicken Little (props for ref:-) that simply up and dies on you, or gets cancer, for other reasons.
rgb
Did I mention that my name is Bob, my first Linuxoid operating system was Slackware, and I'm very definitely an ordained subgenius? Also, I used to smoke a pipe.
So send your money to me.
rgb
(More seriously I'm reading The Black Swan, by Nassim Nicholas Taleb. What a "Black Swan" an Earth-scouring solar flare would be! And one in 2012, too. Those pesky Mayans.
One is also seriously reminded of a Larry Niven short story, but I can't remember the name, am not at home near my bookshelves, and am way too lazy to look it up. But it all starts with the full moon rising and becoming very, very bright, signalling the sequential extinction of land/surface life as the planet rotates. How would you spend your last hours?
The interesting story here is state police cars with built in radioactivity detectors, obviously either checking for dirty radioactive weapons, nuclear weapons or newly arrived aliens hot off the star ships skulking about in their human skin suits ;).
Precisely. I did not know that. Not only built with radioactivity detectors, but ones that are on all the time and are damn sensitive if they are picking up the excess flux from a human inside a car from a tracer treatment administered presumably some nontrivial amount of time before from a distance of what -- 7 meters? 10? 20? -- while driving down the road. Tracers are often very short half-life elements -- that's why they use them -- lots or radioactivity but for a very short time. They tend to be produced in the hospital immediately before use and be mostly gone an hour or two later (but with an exponential tail). Clearly they nailed him right after he left the hospital, and he left the hospital rather quickly after the test, probably less than 45 minutes after the production of the tracer.
Are they sensitive enough to pick up a nuclear bomb being transported? Not if it is made with bomb-grade Uranium, which is also the easiest thing to make a bomb out of, but which isn't radioactive, although you might pick up the trigger. Plutonium 239 IS radioactive, producing a 5 MeV or so alpha at a rate sufficient to keep Plutonium warm to the touch, but alpha particles are relatively easy to block. It also typically contains Pu 240, which spontaneously fissions and produces a surplus flux of a few ~10 million neutrons per second from a typical core. Neutrons are more difficult to stop, but the intensity diminishes like 1/(4\pi r^2) so that the intensity at 10 meters is ~10^7/1250 or around 10^4 per meter squared per second. A detector as large as 1cm x 10 cm would then pick up 10 surplus neutrons per second at 10 meters, assuming there was zero attenuation in between and a perfect detector, neither of which is true. MAYBE this would give them signal to noise of a decibel or two, but given detector efficiency probably not until you were much closer. Up close it would be better, of course. Presumably their detectors have some sort of built in discriminator looking for sustained signal to noise above some cut-off.
What the patient was probably emitting is gamma. Gamma radiation has a long range and isn't easily blocked.
rgb
Well said and yes.
Actually, you need to read "The Black Swan" by N. N. Taleb. Science that tries to confirm a theory is already infected with confirmation bias. There are a pile of examples that demonstrate the fallacy of confirmatory inference. Taleb uses a variant of Bertrand Russell's -- a turkey might reasonably infer, based on his daily experience, that humans exist for the sole purpose of feeding him, caring for him, providing for his every need. This might go on for day after day, increasing the turkey's degree of belief in his hypothesis of a good and beneficent humanity filled with love of turkeys, right up to the day that -- ulp -- something unexpected happens.
I'm not a Popperite, rather a Jaynes-Cox-Bayesian, but nevertheless it is important to avoid confounding the relative strength of positive and negative evidence. Absence of evidence is not the same as evidence of absence, yet we almost invariably confound the two.
Taleb damn skippy agrees with you about publicity, however, and the near-criminality of publicity and reporting of science. A newspaper necessarily takes a scientific result or observation and transforms it two ways: First of all, it creates a narrative. It isn't just "a tornado hit Houston", but "a tornado hit Houston, possibly caused by Anthropogenic Global Warming" with the subtext "this isn't an act of nature, random an unpredictable, but is instead our fault". Aztec priests couldn't have come up with a better excuse for ripping the still beating hearts out of a stream of slaves and war captives. Second, it necessarily reduces the complexity of the result to no more than three variables, ideally one. It "Platonifies" it (according to Taleb) -- wraps it up in a pretty, easy to understand package that makes it more predictable, less random than it really was. Global warming is a much simpler "cause" than "A cold front overrunning a warm wet surface layer of air near the ground, creating turbulent rolls that break off and terminate on the ground, sustained and driven by the thermal difference, and it is a better story too.
Sadly, as you point out, real science is all too often (and should be) scientist z looked at something and didn't find much. But what they failed to find and how they looked is actually often as or more important than a study that claims to find something, especially when the latter uses questionable methodology to try to prove something, cherrypicks data (for the same purpose), ignores silent evidence (ditto) etc. Medical science is permeated with this. Nobody gets famous, or rich, or even a job, for looking for a cure for cancer and not finding one. This too is addressed by Taleb. Great book.
rgb
Well, yeah, except that all of the bending occurs at the interface surface. So in principle, one could stack 150 lensing surfaces constructed Fresnel-style and bring that right down to a kilometer. Depending on what the "interface surface" is for gamma rays. Or, stack 1500 of them and bring it down to 100 meters. Or stack 15,000 of them, in a 3D structure created with e.g. molecular beam epitaxy, and bring it down to 10 meters (with a lens with a total length of perhaps a meter). At that point it is conceivable that it might be useful, although probably not as an optical grade lens (wouldn't it be uber cool to build a gamma ray telescope with an aperture lens a meter across? Sure it would!)
I think that a lot of these same principles are involved in building x-ray lenses -- the lens is less like a glass lens, more like building an interferometric scattering array that causes a single central primary peak. But not my specialty, just thinking out loud...
rgb
Who knew...