I think you're misunderstanding the primary complaint about the venture funding bias: 1) Stanford admissions selections, while probabilistic, are dominated by socioeconomic status (this also highly correlates with several often used measures of 'smarts', like the SAT). 2) Stanford students and graduates have privileged access to venture capital funding for their start-ups. 3) This gives incentive for a certain type of highly achieving student to apply to Stanford -- those interested in receiving VC money. 4) That incentive compromises Stanford's ability as a top-tier research institution to attract students who are interested in basic research in proportion to those interested in immediately applicable research topics. 5) Without the broad basic research base, the quality of Stanford alums starts declining because their applied ideas don't use the best current science.
I don't think, even if this cycle perpetuates that it spells death to Stanford or anything, but it sure is non-optimal in terms of technological development, and it will surely also cause a dip in the quality of Stanford's research output, which has generally been extremely high in the past 40 or so years... and given the amount of GDP the Stanford has access to and their research record in the past 40 years, that's bad news not only for the US tax payer but humanity as a whole.
It turns out that humans are really poor at estimating velocity unless they conform to Newtonian accelerations very closely.. While there has been a lot of research on these issues, I'd like to refer to one of my favorite papers, Sverker Runeson's 1975 paper "Constant velocity — Not perceived as such".
I'd first suggest to try reading some John Holland, like Hidden Order for example, which will elucidate the errors of argument better than my brief response.
Suppose we think about evolution and natural selection as a 'fitness landscape' (it's many dimensional, but let's call it two for the moment). Going vertical optimizes for a specific environment, going lateral means being viable in a range of environments. Evolution is in this setting inevitable, transcription errors are made etc and mutations happen. The theory of natural selection claims that those mutations that are beneficial to the organism prosper, whether that movement is an optimization within their own local fitness area or one that allows them to attempt to migrate into areas that were previously not viable.
One primary problem with Behe and Dembski's theories, is that they make contradictory assumptions about the fitness landscape. Behe argues, with irreducible complexity, that the comparatively low entropic (local minima) state of the mammalian eye is evidence of intelligent design since high entropy, intermediate, eyes wouldn't be sustainable (turns out all the intermediate versions exist). Dembski on the other hand seems to argue the position that increasing entropy is demanding, and the growth in complexity is important in the context of the fitness landscape. So those theories are contradictory.
But where they both run into a global (think variable, not earth) problem (human knowledge passing is Lamarckian, not Mendelian on the whole) is that neither consider the lateral aspects of the landscape, and think that everything is either climbing or descending, mostly in opposite directions!
The monarch retains a number of prerogatives (some by law and some by custom), one of which is that no bill can become a law without royal assent. In practice, the last time assent was withheld occurred in 1707. However, if the monarch felt strongly enough about an act, particularly one that would eliminate democracy in England, they could withhold assent and spark a constitutional crisis.
Just to point out: that progress only occurred because our ancestors were smart enough to start living in caves (and cooking their food, which was probably related, since keeping a fire lit was probably a death sentence unless in a protected location). Even Chimps, who from our current understanding of genetics only have 26 allele pair differences in parts of the genome related to neural development haven't gotten that far.
One of my very favorite classes, in algebraic combinatorics, allowed as much or as little student collaboration as anybody wanted to do.. there were two rules: if you worked on the problem with a group at all, everybody had to acknowledge the collaboration on the problem, and second, while you were allowed to discuss the method of proof etc, in detail, everybody had to go home and write up their own proof. I can say from experience that the range of quality in the proof write-up, even from the same methods and framework, was considerable. In fact, there were times that despite the fact that I essentially had a step-by-step outline of the proof that one of the guys in my group had come up with, I wasn't able to create a coherent write-up on my own (it turns out in one of these cases, it was an open problem). Of course, this only works in a hard class -- in this instance an A- was turning in satisfactory proofs to half the problems presented over the course (there were no exams).
Also, at least in a case like the one I described, if you aren't able to contribute on a regular basis, nobody is going to be interested in working with you because the work is hard enough without having to provide remedial explanations constantly. Unless of course it's assigned groups, but then the class has a lot more to do with dealing with colleagues than collaborative problem solving.
Landis is being punished for daring to defy the anti-doping authorities, insist on his rights to a public hearing (no longer allowed), and embarrassing the hell out of the USADA and WADA by absolutely demolishing their scientific credibility with regard to the testosterone case (after they had to dig in their heels because they had already illegally released the preliminary reports, pre-B sample test to the media). I would note that in the original (and appealed) decisions, the panels through out the initial T-E ratio test as being hopelessly compromised. The mass spectrometry tests were allowed to stand, despite being the quality of lab work that would get laughed out of a college chemistry class, because both panels chose to totally disregard the testimony of John Amory. (see: http://rant-your-head-off.com/WordPress/?p=383 or http://trustbut.blogspot.com/2008/12/winnowing-john-amory.html)
Now, as it turned out, Landis later admitted to doping with HGH that season, and testosterone in previous seasons. But I really think that's incidental to this case. He's being punished because he showed the WADA and UCI are just as corrupt as the cyclists, and the Chatenay-Malabry lab technicians are too incompetent to run a mass spectrometer that undergraduates successfully use thousands of times a day in research labs.
So at the moment I'm playing through a Japan campaign in Victoria 2, which is Paradox's pseudo-realtime complex conquest and development game simulating from 1836-1936. The Brits obviously start the game with a huge advantage (as do the other European powers) and indeed, Japan starts as an uncivilized nation, with major penalties to research and the inability to industrialize among other things. There is however, a path to becoming civilized (which Japan can do through the Meiji restoration decision) and indeed by 1878, I'm in a war with my allies France and (uncivilized) China against Great Britain and the North German federation in an attempt to take Northern India. In this war, about 2.5 million men are fighting on either side, and there will be about 3 million dead (mostly through the British and Chinese armies marching over the Himalayas) by the time the war is over.
In Victoria 2, each soldier is a member of an individual 'pop' living in a certain province, and having its own needs, incomes and political positions. In this war, there are many conscripted regiments who belong to specific 'poor strata' pops of jobs such as farmers, laborers and craftsmen (which I will note, separate men of working/fighting age from women and other parts of the population). Each death on the battle field decreases the size of the 'pop' by the same number of men. Also, I've enacted policies of minority building restrictions, and a discriminatory schooling system to speed assimilation.
By this rationale, aren't I doing worse (in both war and peace) in a single playing session than all the Call of Duty players can do in a similar session combined?
Looking at the following source: http://www.sturmvogel.orbat.com/SovLendLease.html which may or may not be accurate, but does jive with my memory (the most recent essay I've written about a related topic was about 12 years ago and I no longer have the library books), about 75% of the (total) rail stock was delivered pre-December 1941, as well as with a big chunk, 60% or so of the (available) aviation fuel.. which would have been vital to sustaining defensive supply lines. The best thing I found on trucks is http://www.armchairgeneral.com/forums/showthread.php?t=74473 which, if true, would indicate that lend-lease trucks were vital to sustaining the Soviet advances in 1943 and 1944.
In both WWI and WWII the German army managed to critically damage Russia's railstock at the outset of the conflict (albeit, for different reasons), to say that several hundred locomotives weren't vital to sustaining the Soviet defensive positions in late 1941 requires affirming evidence.
And how did the Soviet weapons get deployed to the front? Using 2000 Lend-Lease locomotives, 11,000 Lend-Lease rail cars, and 70%+ Lend-Lease truck strength, and about 18,700 aircraft, about 80% of which were transport aircraft. It may not be glamorous, but that's what allowed the USSR to counter-attack.
S&P had such a major effect when they downgraded Japan's debt in 2002!!!!!! I'm sure S&P etc are still hoping for another 4 or 5 disasters to make that come true. Let's of course, not forget that S&P rated the rags that we previously knew as collective mortgage assets AAA up unto the last day... and I'm supposed to invest? Heh.
.. who works with primates... I do so because I'm convinced there is no other way of collecting data that is important to our health and understanding about how our minds work. Food.. there are other sources.. but neuronal data, we're limited. I'm a big fan of the Reduce, Refine, and Replace idea, and if this is confirmed it's a big step, for 2 R's, and that's exciting.
There were also times in our country's history when such grand investments were politically out of reach. The Land Grant College Act and the Transcontinental railroad were only passed after the departure of the southern states during the Civil War. Rural electrification, despite being shown feasible in 1923, was part of the New Deal.
But even knowing that, it really does annoy me is that we currently have a lot of unemployed from the construction sector. Our infrastructure, according to the American Assc of Civil Engineers is a D. We can borrow money long term (30 year bonds) at about 4.7%, which is about where it was when we decided to build the Interstate system (debt/GDP is also similar). This is the most fiscally favorable time to get our infrastructure back into shape in the last 50 years.
And instead of taking the opportunity to catch up on the repair bill, we have a Republican congress screaming bloody murder about rounding errors in 12% of the budget weeks after blessing a tax cut of similar size, while Obama proposes cutting Pell grants and loans to students pursuing graduate degrees weeks after declaring we're in an education race.
Think about this from a Hollywood perspective for a moment (and I don't mean the snorting middle-manager suits). Hollywood wields the modern day pen with considerable acumen, and they understand the making of a political 'plot' and what's required to be plausible all too well. And they already make their money at the high end of the risk spectrum (I don't know of any studio bailouts).
Shutting down Google is not an objective, it's the denouement of the first act. The misinformed people, the 'other' and even the gods are against them, but down on their luck they still are fighting, and you, the intrepid state senator who would really like a movie-middleman job can help!
The first part of Act 2 will be when they lobby most of the state legislatures into passing draconian infringement or usage laws, while we're busy chortling over how Google stomped on them.
I don't even know how we get through Act 2, much less Act 26.
To put it another way: Would you rather have $100m and the ability to invest it with 2% better return than the S&P, or the ability to talk anybody you meet into giving you 1% of their wealth?
If writing isn't emphasized as the major professional skill learned in college, it still speaks badly of the engineering program. While it's obviously inappropriate to force large written assignments into pure math or core engineering classes, in a good technical curriculum, writing should be emphasized in corollary discipline and core distribution classes. I'd say the 'average' experience for my college friends in engineering is they spend the majority of their time writing technical reports, white papers or memos. Luckily for them, Michigan has strong distribution requirements which emphasize writing, even in the engineering school. An even better example of a purpose built technical school that emphasizes the necessity of communicating clearly in professional settings is WPI with their projects and global perspectives program (which, among other things, makes students propose a project, plan it, go off and spend a semester trying to execute their plan, and then evaluate the entire process with strong faculty support).
Having about 2/3rds of engineering/technical students admitting to cheating is a horrific indictment of our K-12 system.. but I guess class categories would be important to understand to help remedy it. If it's technical classes that they cheat in, then the Race to the Top, etc, trying to force more math and science into high school makes some sense (although with only 30% of biology teachers willing to teach evolution, one wonders..), but if they're plagiarizing instead of writing, Race to the Top is barking up the wrong tree, and we should probably at the least create a college-prep track that focuses on a single goal: graduate high-school knowing how to write a coherent position paper.
Unfortunately mere distribution doesn't take something out of copyright, only explicit assignment to the public domain can do that.. and even there are ways to claw it back into copyright, if any piece of the final product is a collaboration.
With regard to the brain, I agree that Bayesian inference can be a good phenomenological tool to model many complex behaviors, but it does not produce useful mechanistic predictions as to how real neural networks actually compute information. Also, Bayesian models have a very difficult time matching the single-event learning capability of the brain. See here (page 31;PDF) for a brief review in the context of motion processing or here (large pdf) for a more rigorous discussion.
I also give three cheers for Jaynes... I have the habit of giving books to my friends if I think they'd be interesting or useful, and I've dropped a several hundred dollars on re-buying Jaynes. There used to be a 'pre-print' version of the book available, but that source now seems to be down to a few chapters due to copyright issues, although his papers are still there.
With regard to the human brain however, it is (very) unlikely that Bayes' rule is actually computed in any sense of the word. Bayes requires at least one neuron with a global scope (ie. a grandmother cell) to compute the posterior, which is biologically implausible.
What I'm more curious about, is what is the statute of limitations, so to speak, of the police having consent. I was the victim of an (attempted) armed robbery a few years ago in the apartment I currently live in (he didn't think anybody was around, and ran out after threatening me.. it sucks waking up from a nap to an intruder with a gun standing over you), and I sure as hell didn't mind the police searching my apartment then.. but when is that consent removed? All they found was the guy's jacket, the case is still open.. could they still come back and search without a warrant, even if they were interested in a different case? Or do they have to re-establish consent after the first search?
Given that I mentioned (and linked) the Hawthorne effect in my original post, I'm well aware of it (I also discussed some of the statistical issues in response to martin-boundary's comment above).
The Hawthorne experiment was not a shockingly bad mistake, they went in looking for something and found an effect they hadn't been thinking about, then they did a pretty decent job investigating several possible manipulations and published it so other's could have a look. It has since been refined considerably, and plays a major role in how we design experiments. Welcome to wet science. The 'mistakes' you mention are only experienced in hindsight.
The fact remains: wet science and 'hard' science aren't really in the same ballpark in terms of difficulty and complexity, which I also discussed above. Robert Hooke discovered the cell in 1655, 30 years before Newton published his theory of gravity. Newton, and many physicists who came later (Maxwell, Mach, Hemholtz to name a few) understood quite a bit about biology and psychology. They chose physics because it was the accessible problem for them.
Life has been forced by natural selection to come up with incredibly dense, optimized hardware. DNA is on the scale of nanometers, and we literally could not see its structure until the 50s. Which is to say, it took about 300 years of "proper, rigorous" science to produce the tools required to probe the cell and its various amalgamations.. and that's what's being done, and it's going to be just as messy figuring out if there's aether out there.
The theory of experimental errors (and all the derived statistics we use) separates errors into two components: systemic, and random error. The random error is by definition assumed to be uncorrelated between observations (sampling errors).
Random error actually contains another component: non-sampling errors, which are systemic errors that have not been identified. The "hard" sciences, since the Renaissance, has been a mostly successful project of identifying non-sampling errors and improving measurements. This was possible because the experimental mediums didn't change very much (the boiling temperature of water was the same for Galileo).
Doing the same thing in wet science is orders of magnitude more difficult, because biological organisms have memory, adapt and change the environment around them. So it's extremely difficult (likely impossible) to maintain a consistent experimental mediums for long periods of time. This in turn, means two very important things: 1) That the systemic errors change between replications of an experiment (asking Galileo "what is a planet?" and the current chair of the physics department at Padua will yield different answers) and 2) Early differences in random error can propagate through the system and become systemic error in later trials in the same setup (eg. Shaking hands causing slight differences in cellular distribution in the petri dish at the beginning of a growth cycle).
I'm not aware of any statistical methods that can be used to better classify the error in this type of environment. So you're left with changing the underlying distribution.. but to what (and for which protocols.. etc etc)?
While there are wet scientists who use statistics poorly, I think the bigger problem is that the statistical tools currently available are pretty wimpy. Econometrics is likely the most developed mathematical formulation for statistical analysis of non-linear recurrent systems. That's not a good thing.
I've read the Feynman essay, and it's a useful point as far as it goes, but it doesn't address the issues that I brought up at all: people (and animals) are a lot harder to experiment on than inert objects (marbles don't punch back). It has a lot more to do with the fundamental difference in complexity than it has to do with sloppiness.
But you do bring up a much more general point about experimental protocol that you overlooked in your description: a measurement is meaningless unless you also provide the resolution at which it was assessed. The famous example of this problem is trying to measure the length of the coast of England with various size rulers, which led to the development of fractal geometry etc (see also: scale space).
With regards to your replication of F=ma, to bring that experiment up to the level of Mr. Young's work, you would have found that F=ma broke down at some scales, and discovered quantum theory and relativity. Of course, that rigor took physics over 300 years of further effort!
Just a note on Freud: I happen to research neuroscience, and while I wouldn't call my educational path the most typical, my only encounter with Freud in a class was that some of his later work pioneered the study of neural networks.
Well, passing over for the moment the likes of determinism and ecological psychology, I think you're mistaken in direct studies of human behavior. There are a number of very subtle effects, that when run through the non-linear recurrent processes of the brain can lead to significant behavioral changes (ie. the demand effect). While some of these were touched on lightly in the New Yorker article (about blinding protocols and so on) there are second order effects that are impossible to control. A psychologist who does a large subject pool experiment needs funding, which to get generally requires pilot results. These results are exciting to the psychologist and their lab, they're more motivated, have higher energy, probably are interacting better with the subjects (or the technicians running the subjects, if they have enough money to blind it that deeply), more motivated people are going to produce different behaviors than less motivated people. If the blinded study is positive and appears significant, it may become a big thing.. but by the nth repetition of the original experiment by another lab to verify they understand the protocol, the lab might be completely bored with the initial testing and the result disappears, essentially a variation of the Hawthorne effect (which has itself been disappearing). That may well mean that the effect exists in certain environments but not others, which is an ultimately frustrating thing to classify in systems as complex as human society.
It essentially boils down to the fact that we're all fallible, social beings that interact with the environment, rather than merely observing it. Whether you want to say that this adds correlated but unpredictable noise to any data analysis that is not being appropriately controlled for (but can be), or is fundamental limit on our ability to understand certain aspects of the world, at our current level of understanding it does rather seem that there is a class of experiments in which a scientist's mental state affects the (objective) results.
Hell, George RR Martin's "A Song of Ice and Fire" series has plenty of incest, to the point one might call it a theme and The Silmarillion also features the incest of Turin and Nienor (I think Asimov and Le Guin also wrote stories containing incest). But why stop there? Gotta ban Oedipus, any stories about the Ptolemies (probably where Martin got the idea for the Targaryen sibling marriages), Nabakov's Ada or Ardor and Gabriel Garcia Marquez' One Hundred Years of Solitude, not to mention the the bible, just to name a few off the top of my head.
Censorship of art is by its nature arbitrary, capricious and dumb. But that doesn't mean people won't try to continue to perpetrate it, most often today under feeble rationale that it isn't the government censoring, but a corporation (whose existence relies on a special government charter) and this somehow makes it all okay. Guess what, it doesn't. I mean the fact that WBAI wouldn't broadcast a reading of Ginsberg's Howl on the 50th anniversary of the censorship trial that declared it a non-obscene work... it's not ironic, it's just fucked up, and it shows we're in the middle of a dangerous backslide on free speech.
Upper management and idea people who 'just need a programmer' suffer from the same problem: they cannot articulate their idea(s) in a useful form. An idea is only as powerful as the explanatory power and effort behind it (obligatory xkcd). This however, has nothing to do with ivory tower issues--the ivory tower problem is that people who can clearly describe ideas to each other have a problem describing those ideas to non-experts (or that the underlying assumptions in the field are hard to translate or inapplicable). The problem described above is that they may have an idea, that they are convinced is novel, useful, profitable etc, but they cannot articulate the specifics of the idea to anybody to the point where development can proceed. Rather than recognize their failure to research and specify the idea sufficiently clearly for others to buy in, they instead blame the problem on outside forces for which programmers provide one of the myriad of scapegoats.
(Disclaimer: I am a member of the department, but do not work on the project)
The macro-connectivity of the brain is quite well understood. What is lacking is a rigorous and realistic treatment of laminar cortical circuits, and equivalents for sub-cortical areas of the brain. The founder of CNS, Steve Grossberg, has spent his entire career showing that a few fundamental circuits and learning rules can be configured in different ways to simulate many brain behaviors. However, these models are limited by the current hardware whose processing strengths are orthogonal to the complexity of neural circuits. So physical instantiation is incredibly important when you consider the details of the hardware, and the system you're trying to model. Think about how you might go about modeling the basic circuit differences between say, eulaminate and granular cortices. The complexity of these circuits are far beyond our current understanding of systems of non-linear differential equations, so you have to simulate them. And since current computers are really limited in their ability to simulate large-scale neural networks, moving forward on understanding the details of cortical circuits and their formation is going to require a hardware platform whose strengths parallel those of the brain (which is to say, highly parallel computation with super-dense memory).
Slashdot Mirror
I think you're misunderstanding the primary complaint about the venture funding bias:
1) Stanford admissions selections, while probabilistic, are dominated by socioeconomic status (this also highly correlates with several often used measures of 'smarts', like the SAT).
2) Stanford students and graduates have privileged access to venture capital funding for their start-ups.
3) This gives incentive for a certain type of highly achieving student to apply to Stanford -- those interested in receiving VC money.
4) That incentive compromises Stanford's ability as a top-tier research institution to attract students who are interested in basic research in proportion to those interested in immediately applicable research topics.
5) Without the broad basic research base, the quality of Stanford alums starts declining because their applied ideas don't use the best current science.
I don't think, even if this cycle perpetuates that it spells death to Stanford or anything, but it sure is non-optimal in terms of technological development, and it will surely also cause a dip in the quality of Stanford's research output, which has generally been extremely high in the past 40 or so years... and given the amount of GDP the Stanford has access to and their research record in the past 40 years, that's bad news not only for the US tax payer but humanity as a whole.
It turns out that humans are really poor at estimating velocity unless they conform to Newtonian accelerations very closely.. While there has been a lot of research on these issues, I'd like to refer to one of my favorite papers, Sverker Runeson's 1975 paper "Constant velocity — Not perceived as such".
http://www.springerlink.com/content/nt61hh074k7123q5/
I'd first suggest to try reading some John Holland, like Hidden Order for example, which will elucidate the errors of argument better than my brief response.
Suppose we think about evolution and natural selection as a 'fitness landscape' (it's many dimensional, but let's call it two for the moment). Going vertical optimizes for a specific environment, going lateral means being viable in a range of environments. Evolution is in this setting inevitable, transcription errors are made etc and mutations happen. The theory of natural selection claims that those mutations that are beneficial to the organism prosper, whether that movement is an optimization within their own local fitness area or one that allows them to attempt to migrate into areas that were previously not viable.
One primary problem with Behe and Dembski's theories, is that they make contradictory assumptions about the fitness landscape. Behe argues, with irreducible complexity, that the comparatively low entropic (local minima) state of the mammalian eye is evidence of intelligent design since high entropy, intermediate, eyes wouldn't be sustainable (turns out all the intermediate versions exist). Dembski on the other hand seems to argue the position that increasing entropy is demanding, and the growth in complexity is important in the context of the fitness landscape. So those theories are contradictory.
But where they both run into a global (think variable, not earth) problem (human knowledge passing is Lamarckian, not Mendelian on the whole) is that neither consider the lateral aspects of the landscape, and think that everything is either climbing or descending, mostly in opposite directions!
The monarch retains a number of prerogatives (some by law and some by custom), one of which is that no bill can become a law without royal assent. In practice, the last time assent was withheld occurred in 1707. However, if the monarch felt strongly enough about an act, particularly one that would eliminate democracy in England, they could withhold assent and spark a constitutional crisis.
Just to point out: that progress only occurred because our ancestors were smart enough to start living in caves (and cooking their food, which was probably related, since keeping a fire lit was probably a death sentence unless in a protected location). Even Chimps, who from our current understanding of genetics only have 26 allele pair differences in parts of the genome related to neural development haven't gotten that far.
One of my very favorite classes, in algebraic combinatorics, allowed as much or as little student collaboration as anybody wanted to do.. there were two rules: if you worked on the problem with a group at all, everybody had to acknowledge the collaboration on the problem, and second, while you were allowed to discuss the method of proof etc, in detail, everybody had to go home and write up their own proof. I can say from experience that the range of quality in the proof write-up, even from the same methods and framework, was considerable. In fact, there were times that despite the fact that I essentially had a step-by-step outline of the proof that one of the guys in my group had come up with, I wasn't able to create a coherent write-up on my own (it turns out in one of these cases, it was an open problem). Of course, this only works in a hard class -- in this instance an A- was turning in satisfactory proofs to half the problems presented over the course (there were no exams).
Also, at least in a case like the one I described, if you aren't able to contribute on a regular basis, nobody is going to be interested in working with you because the work is hard enough without having to provide remedial explanations constantly. Unless of course it's assigned groups, but then the class has a lot more to do with dealing with colleagues than collaborative problem solving.
Landis is being punished for daring to defy the anti-doping authorities, insist on his rights to a public hearing (no longer allowed), and embarrassing the hell out of the USADA and WADA by absolutely demolishing their scientific credibility with regard to the testosterone case (after they had to dig in their heels because they had already illegally released the preliminary reports, pre-B sample test to the media). I would note that in the original (and appealed) decisions, the panels through out the initial T-E ratio test as being hopelessly compromised. The mass spectrometry tests were allowed to stand, despite being the quality of lab work that would get laughed out of a college chemistry class, because both panels chose to totally disregard the testimony of John Amory. (see: http://rant-your-head-off.com/WordPress/?p=383 or http://trustbut.blogspot.com/2008/12/winnowing-john-amory.html)
Now, as it turned out, Landis later admitted to doping with HGH that season, and testosterone in previous seasons. But I really think that's incidental to this case. He's being punished because he showed the WADA and UCI are just as corrupt as the cyclists, and the Chatenay-Malabry lab technicians are too incompetent to run a mass spectrometer that undergraduates successfully use thousands of times a day in research labs.
So at the moment I'm playing through a Japan campaign in Victoria 2, which is Paradox's pseudo-realtime complex conquest and development game simulating from 1836-1936. The Brits obviously start the game with a huge advantage (as do the other European powers) and indeed, Japan starts as an uncivilized nation, with major penalties to research and the inability to industrialize among other things. There is however, a path to becoming civilized (which Japan can do through the Meiji restoration decision) and indeed by 1878, I'm in a war with my allies France and (uncivilized) China against Great Britain and the North German federation in an attempt to take Northern India. In this war, about 2.5 million men are fighting on either side, and there will be about 3 million dead (mostly through the British and Chinese armies marching over the Himalayas) by the time the war is over.
In Victoria 2, each soldier is a member of an individual 'pop' living in a certain province, and having its own needs, incomes and political positions. In this war, there are many conscripted regiments who belong to specific 'poor strata' pops of jobs such as farmers, laborers and craftsmen (which I will note, separate men of working/fighting age from women and other parts of the population). Each death on the battle field decreases the size of the 'pop' by the same number of men. Also, I've enacted policies of minority building restrictions, and a discriminatory schooling system to speed assimilation.
By this rationale, aren't I doing worse (in both war and peace) in a single playing session than all the Call of Duty players can do in a similar session combined?
Looking at the following source: http://www.sturmvogel.orbat.com/SovLendLease.html which may or may not be accurate, but does jive with my memory (the most recent essay I've written about a related topic was about 12 years ago and I no longer have the library books), about 75% of the (total) rail stock was delivered pre-December 1941, as well as with a big chunk, 60% or so of the (available) aviation fuel.. which would have been vital to sustaining defensive supply lines. The best thing I found on trucks is http://www.armchairgeneral.com/forums/showthread.php?t=74473 which, if true, would indicate that lend-lease trucks were vital to sustaining the Soviet advances in 1943 and 1944.
In both WWI and WWII the German army managed to critically damage Russia's railstock at the outset of the conflict (albeit, for different reasons), to say that several hundred locomotives weren't vital to sustaining the Soviet defensive positions in late 1941 requires affirming evidence.
And how did the Soviet weapons get deployed to the front? Using 2000 Lend-Lease locomotives, 11,000 Lend-Lease rail cars, and 70%+ Lend-Lease truck strength, and about 18,700 aircraft, about 80% of which were transport aircraft. It may not be glamorous, but that's what allowed the USSR to counter-attack.
http://en.wikipedia.org/wiki/Lend-Lease
S&P had such a major effect when they downgraded Japan's debt in 2002!!!!!! I'm sure S&P etc are still hoping for another 4 or 5 disasters to make that come true. Let's of course, not forget that S&P rated the rags that we previously knew as collective mortgage assets AAA up unto the last day... and I'm supposed to invest? Heh.
.. who works with primates... I do so because I'm convinced there is no other way of collecting data that is important to our health and understanding about how our minds work. Food.. there are other sources.. but neuronal data, we're limited. I'm a big fan of the Reduce, Refine, and Replace idea, and if this is confirmed it's a big step, for 2 R's, and that's exciting.
There were also times in our country's history when such grand investments were politically out of reach. The Land Grant College Act and the Transcontinental railroad were only passed after the departure of the southern states during the Civil War. Rural electrification, despite being shown feasible in 1923, was part of the New Deal.
But even knowing that, it really does annoy me is that we currently have a lot of unemployed from the construction sector. Our infrastructure, according to the American Assc of Civil Engineers is a D. We can borrow money long term (30 year bonds) at about 4.7%, which is about where it was when we decided to build the Interstate system (debt/GDP is also similar). This is the most fiscally favorable time to get our infrastructure back into shape in the last 50 years.
And instead of taking the opportunity to catch up on the repair bill, we have a Republican congress screaming bloody murder about rounding errors in 12% of the budget weeks after blessing a tax cut of similar size, while Obama proposes cutting Pell grants and loans to students pursuing graduate degrees weeks after declaring we're in an education race.
Think about this from a Hollywood perspective for a moment (and I don't mean the snorting middle-manager suits). Hollywood wields the modern day pen with considerable acumen, and they understand the making of a political 'plot' and what's required to be plausible all too well. And they already make their money at the high end of the risk spectrum (I don't know of any studio bailouts).
Shutting down Google is not an objective, it's the denouement of the first act. The misinformed people, the 'other' and even the gods are against them, but down on their luck they still are fighting, and you, the intrepid state senator who would really like a movie-middleman job can help!
The first part of Act 2 will be when they lobby most of the state legislatures into passing draconian infringement or usage laws, while we're busy chortling over how Google stomped on them.
I don't even know how we get through Act 2, much less Act 26.
To put it another way: Would you rather have $100m and the ability to invest it with 2% better return than the S&P, or the ability to talk anybody you meet into giving you 1% of their wealth?
If writing isn't emphasized as the major professional skill learned in college, it still speaks badly of the engineering program. While it's obviously inappropriate to force large written assignments into pure math or core engineering classes, in a good technical curriculum, writing should be emphasized in corollary discipline and core distribution classes. I'd say the 'average' experience for my college friends in engineering is they spend the majority of their time writing technical reports, white papers or memos. Luckily for them, Michigan has strong distribution requirements which emphasize writing, even in the engineering school. An even better example of a purpose built technical school that emphasizes the necessity of communicating clearly in professional settings is WPI with their projects and global perspectives program (which, among other things, makes students propose a project, plan it, go off and spend a semester trying to execute their plan, and then evaluate the entire process with strong faculty support).
Having about 2/3rds of engineering/technical students admitting to cheating is a horrific indictment of our K-12 system.. but I guess class categories would be important to understand to help remedy it. If it's technical classes that they cheat in, then the Race to the Top, etc, trying to force more math and science into high school makes some sense (although with only 30% of biology teachers willing to teach evolution, one wonders..), but if they're plagiarizing instead of writing, Race to the Top is barking up the wrong tree, and we should probably at the least create a college-prep track that focuses on a single goal: graduate high-school knowing how to write a coherent position paper.
Unfortunately mere distribution doesn't take something out of copyright, only explicit assignment to the public domain can do that.. and even there are ways to claw it back into copyright, if any piece of the final product is a collaboration.
With regard to the brain, I agree that Bayesian inference can be a good phenomenological tool to model many complex behaviors, but it does not produce useful mechanistic predictions as to how real neural networks actually compute information. Also, Bayesian models have a very difficult time matching the single-event learning capability of the brain. See here (page 31;PDF) for a brief review in the context of motion processing or here (large pdf) for a more rigorous discussion.
I also give three cheers for Jaynes... I have the habit of giving books to my friends if I think they'd be interesting or useful, and I've dropped a several hundred dollars on re-buying Jaynes. There used to be a 'pre-print' version of the book available, but that source now seems to be down to a few chapters due to copyright issues, although his papers are still there.
With regard to the human brain however, it is (very) unlikely that Bayes' rule is actually computed in any sense of the word. Bayes requires at least one neuron with a global scope (ie. a grandmother cell) to compute the posterior, which is biologically implausible.
What I'm more curious about, is what is the statute of limitations, so to speak, of the police having consent. I was the victim of an (attempted) armed robbery a few years ago in the apartment I currently live in (he didn't think anybody was around, and ran out after threatening me.. it sucks waking up from a nap to an intruder with a gun standing over you), and I sure as hell didn't mind the police searching my apartment then.. but when is that consent removed? All they found was the guy's jacket, the case is still open.. could they still come back and search without a warrant, even if they were interested in a different case? Or do they have to re-establish consent after the first search?
Given that I mentioned (and linked) the Hawthorne effect in my original post, I'm well aware of it (I also discussed some of the statistical issues in response to martin-boundary's comment above).
The Hawthorne experiment was not a shockingly bad mistake, they went in looking for something and found an effect they hadn't been thinking about, then they did a pretty decent job investigating several possible manipulations and published it so other's could have a look. It has since been refined considerably, and plays a major role in how we design experiments. Welcome to wet science. The 'mistakes' you mention are only experienced in hindsight.
The fact remains: wet science and 'hard' science aren't really in the same ballpark in terms of difficulty and complexity, which I also discussed above. Robert Hooke discovered the cell in 1655, 30 years before Newton published his theory of gravity. Newton, and many physicists who came later (Maxwell, Mach, Hemholtz to name a few) understood quite a bit about biology and psychology. They chose physics because it was the accessible problem for them.
Life has been forced by natural selection to come up with incredibly dense, optimized hardware. DNA is on the scale of nanometers, and we literally could not see its structure until the 50s. Which is to say, it took about 300 years of "proper, rigorous" science to produce the tools required to probe the cell and its various amalgamations.. and that's what's being done, and it's going to be just as messy figuring out if there's aether out there.
I wrote a similar post below, but make a slightly different criticism. The difference between wet science and "hard" science boils down to one issue: non-linear recurrence (here is a nice history of the split between physics and psychology [big pdf]).
The theory of experimental errors (and all the derived statistics we use) separates errors into two components: systemic, and random error. The random error is by definition assumed to be uncorrelated between observations (sampling errors).
Random error actually contains another component: non-sampling errors, which are systemic errors that have not been identified. The "hard" sciences, since the Renaissance, has been a mostly successful project of identifying non-sampling errors and improving measurements. This was possible because the experimental mediums didn't change very much (the boiling temperature of water was the same for Galileo).
Doing the same thing in wet science is orders of magnitude more difficult, because biological organisms have memory, adapt and change the environment around them. So it's extremely difficult (likely impossible) to maintain a consistent experimental mediums for long periods of time. This in turn, means two very important things: 1) That the systemic errors change between replications of an experiment (asking Galileo "what is a planet?" and the current chair of the physics department at Padua will yield different answers) and 2) Early differences in random error can propagate through the system and become systemic error in later trials in the same setup (eg. Shaking hands causing slight differences in cellular distribution in the petri dish at the beginning of a growth cycle).
I'm not aware of any statistical methods that can be used to better classify the error in this type of environment. So you're left with changing the underlying distribution.. but to what (and for which protocols.. etc etc)?
While there are wet scientists who use statistics poorly, I think the bigger problem is that the statistical tools currently available are pretty wimpy. Econometrics is likely the most developed mathematical formulation for statistical analysis of non-linear recurrent systems. That's not a good thing.
I've read the Feynman essay, and it's a useful point as far as it goes, but it doesn't address the issues that I brought up at all: people (and animals) are a lot harder to experiment on than inert objects (marbles don't punch back). It has a lot more to do with the fundamental difference in complexity than it has to do with sloppiness.
But you do bring up a much more general point about experimental protocol that you overlooked in your description: a measurement is meaningless unless you also provide the resolution at which it was assessed. The famous example of this problem is trying to measure the length of the coast of England with various size rulers, which led to the development of fractal geometry etc (see also: scale space).
With regards to your replication of F=ma, to bring that experiment up to the level of Mr. Young's work, you would have found that F=ma broke down at some scales, and discovered quantum theory and relativity. Of course, that rigor took physics over 300 years of further effort!
Just a note on Freud: I happen to research neuroscience, and while I wouldn't call my educational path the most typical, my only encounter with Freud in a class was that some of his later work pioneered the study of neural networks.
Well, passing over for the moment the likes of determinism and ecological psychology, I think you're mistaken in direct studies of human behavior. There are a number of very subtle effects, that when run through the non-linear recurrent processes of the brain can lead to significant behavioral changes (ie. the demand effect). While some of these were touched on lightly in the New Yorker article (about blinding protocols and so on) there are second order effects that are impossible to control. A psychologist who does a large subject pool experiment needs funding, which to get generally requires pilot results. These results are exciting to the psychologist and their lab, they're more motivated, have higher energy, probably are interacting better with the subjects (or the technicians running the subjects, if they have enough money to blind it that deeply), more motivated people are going to produce different behaviors than less motivated people. If the blinded study is positive and appears significant, it may become a big thing.. but by the nth repetition of the original experiment by another lab to verify they understand the protocol, the lab might be completely bored with the initial testing and the result disappears, essentially a variation of the Hawthorne effect (which has itself been disappearing). That may well mean that the effect exists in certain environments but not others, which is an ultimately frustrating thing to classify in systems as complex as human society.
It essentially boils down to the fact that we're all fallible, social beings that interact with the environment, rather than merely observing it. Whether you want to say that this adds correlated but unpredictable noise to any data analysis that is not being appropriately controlled for (but can be), or is fundamental limit on our ability to understand certain aspects of the world, at our current level of understanding it does rather seem that there is a class of experiments in which a scientist's mental state affects the (objective) results.
Hell, George RR Martin's "A Song of Ice and Fire" series has plenty of incest, to the point one might call it a theme and The Silmarillion also features the incest of Turin and Nienor (I think Asimov and Le Guin also wrote stories containing incest). But why stop there? Gotta ban Oedipus, any stories about the Ptolemies (probably where Martin got the idea for the Targaryen sibling marriages), Nabakov's Ada or Ardor and Gabriel Garcia Marquez' One Hundred Years of Solitude, not to mention the the bible, just to name a few off the top of my head.
Censorship of art is by its nature arbitrary, capricious and dumb. But that doesn't mean people won't try to continue to perpetrate it, most often today under feeble rationale that it isn't the government censoring, but a corporation (whose existence relies on a special government charter) and this somehow makes it all okay. Guess what, it doesn't. I mean the fact that WBAI wouldn't broadcast a reading of Ginsberg's Howl on the 50th anniversary of the censorship trial that declared it a non-obscene work... it's not ironic, it's just fucked up, and it shows we're in the middle of a dangerous backslide on free speech.
Upper management and idea people who 'just need a programmer' suffer from the same problem: they cannot articulate their idea(s) in a useful form. An idea is only as powerful as the explanatory power and effort behind it (obligatory xkcd). This however, has nothing to do with ivory tower issues--the ivory tower problem is that people who can clearly describe ideas to each other have a problem describing those ideas to non-experts (or that the underlying assumptions in the field are hard to translate or inapplicable). The problem described above is that they may have an idea, that they are convinced is novel, useful, profitable etc, but they cannot articulate the specifics of the idea to anybody to the point where development can proceed. Rather than recognize their failure to research and specify the idea sufficiently clearly for others to buy in, they instead blame the problem on outside forces for which programmers provide one of the myriad of scapegoats.
(Disclaimer: I am a member of the department, but do not work on the project)
The macro-connectivity of the brain is quite well understood. What is lacking is a rigorous and realistic treatment of laminar cortical circuits, and equivalents for sub-cortical areas of the brain. The founder of CNS, Steve Grossberg, has spent his entire career showing that a few fundamental circuits and learning rules can be configured in different ways to simulate many brain behaviors. However, these models are limited by the current hardware whose processing strengths are orthogonal to the complexity of neural circuits. So physical instantiation is incredibly important when you consider the details of the hardware, and the system you're trying to model. Think about how you might go about modeling the basic circuit differences between say, eulaminate and granular cortices. The complexity of these circuits are far beyond our current understanding of systems of non-linear differential equations, so you have to simulate them. And since current computers are really limited in their ability to simulate large-scale neural networks, moving forward on understanding the details of cortical circuits and their formation is going to require a hardware platform whose strengths parallel those of the brain (which is to say, highly parallel computation with super-dense memory).