In Ontario it's $45 as a co-payment for any non-hospital originating, health card holding, medically necessary local trip. You can be exempt to this fee if you are in a variety of low-income situations. If, like a lot of us you have supplemental insurance provided by your employer then this is often covered.
It is $0 if it is hospital to hospital (in the same situation above). It's $240 (or more if it's an air-ambulance) if you do not have a valid health card or the trip is considered medically unnecessary.
So yes if you meant we pay a *small portion* of ambulance services in a few *reasonable cases* I'd agree with you but this is completely different that what happens in less communist medical systems:-)
From what you're saying I assume you mean test equipment can fail. Sure but so can ears and all the 'software' in-between in fact it's a lot easier to prove that test equipment is functioning correctly than everything from my ears to my brain. It's because of this I tend to trust equipment - assuming it can be tested and validated for it's function too rather than my ears.
With regard to shorting the 5v line on the SATA cable - actually this is pretty astronomically unlikely. Since the SATA cable can't tell when it's sending data or audio it would either need to be a consistent error rate. Which would would likely cause other problems - unless you think the error tolerance of a piece of executable code is somehow better than audio data.
I appreciate the idea that someone wants to do things right but the issue with relying on ones ears is that there is no diagnostic to ensure proper operation. Attempting to replicate the problem doesn't help unless you are doing so in an environment where you are blinded. Otherwise how does this differ from audiophile nutcases who do A/B listening tests on $400 volume knobs?
I'm not so sure this is 'the other way around'. To me that would mean something that is shown to be good in objective tests but somehow a real problem is found with ones ears.
It also depends on how you define "problem". I mean if you define the term to mean "something I hear" well then it's entirely possible that "fixing" that is as much psychosomatic as anything else.
Let me first say that this is great news - if it turns out to be true however following the addage of most published research is false. It's worth keeping in mind that this has 20 controls, 20 ASD and 19 ADHD - according to the article they could distinguish the ASD diagnoses from the controls and the ADHD but considering that according to the DSM IV autism can have close to 100 unique presentations. I wonder how much this actually demonstrates.
I'm not sure if the brackets were literal or just a typographical convention. However historically it wasn't uncommon to see boxes instead of variables at the younger grades - they may still do today. Also without the paper it's difficult to say if this is the only kind of test they put to students.
I think your argument is kind of bogus though - You could just as easily level the same criticism at word problems...or perhaps you believe that word problem need a highly rigorous and pre-defined format. Math, as I see it anyway isn't just taught so that you can manipulate symbols deterministically. To me a good test of *knowing* math is to be able to recognize and solve mathematical problems without needing it to be spoon fed to you. Otherwise...what's the point? Machines already outstrip our ability to compute - why outside of extremely simple addition and subtraction would you bother ever learning a lot of math EXCEPT to use it so that you can recognize solve problems that are not handed to you in exam question format.
Agreed but I think the reason there's conversation around this point at all is how terms like "tech savvy" are used to describe the younger generations of net users. While Kipling in his day could have been called "car savvy" you wouldn't call the majority of drivers today to be so. Ergo it's just backlash (perhaps well deserved) by those of us who are thoroughly underwhelmed by the lack of understanding espoused by the so-called "tech savvy" group. Sometimes the ignorance is so great it appears to be worse when measured by the standards of other technologies which have been subsumed into common use. i.e. I'd expect, when pressed car users could describe more completely the operation of a car than most kids can describe the operation of the internet. Anyway, this labeling may not even be the fault of the generation it's being applied to. I wouldn't be surprised if it was simply the usage of a group technically incompetent enough not to know the difference between "able to operate" and "understanding how it works".
The concept of 'learning languages' that is - that is there are languages that lend themselves to learning programming in general. I' wrote code as a hobby for about 12 years and after university I did it for 16 years professionally. I now manage a team of developers. Without a specific problem (i.e. writing a web application, writing a driver, etc..) I would expect any of these languages would do.
In my humble opinion, the point of learning to *program* as opposed to learning to operate an application (which is closer to what drag and drop environments are) is that you have learned how to approach a problem solving by using algorithms and in the end you possess a set of "mental tools" which allow you to approach any decidable problem.
Ok I'm going to give a more realistic example...
Take a look at the OCZ Colossus it's a pure flash 3.5" drive which was produced last year which has a capacity of 1TB. Now mind you this was engineered by tacking together a bunch of flash on a SATA controller, then plugging two of these into another SATA controller to produce a RAID of these two cards. IMHO this this design was not made to create the highest capacity drive possible.
However it might be a better benchmark for figuring out where our limits are for example..
This drive puts about 5mm between each Flash package. Given that Toshiba has a 128 GB Flash chip/controller launching this fall. Considering that the carrier measures 17mm x 22mm x 1.4mm we can probably put a 4 x 5 Array of these on a card like the ones in the Colossus. Either by double-siding the board or adding more boards (but not necessarily by using bulky SATA connectors). Let's say we can put four of these hypothetical boards into a 3.5" chassis. That still clocks us in at 10TB. Which is still 3x what we have going on with spinning disks. So even if we assume that the density increases for Disk vs. Flash are the same. I'd argue that economies of scale will soon make this cost effective.
The No Code Publish approach seems reasonable to me: Publish flaw to everyone - including CERT, Vendors but include no code or exploit for anything publicly readable. Give the vendor your exploit code an a deadline after which you will publish the exploit. If no fix appears by the deadline then you publish.
I agree (I realize that you're being ironic). Overall I think it was an interesting article but his arguments aren't really strong enough for example:
Bit Density Comparisons vs. Cell Density Comparisons: Or that's what I'm going to call the part where he's talking about lithography. I get that when vendors are showing you growth curves they are showing you bit density not cell density. That is the growth we are seeing aligns much more with products being shifted to pre-existing higher density technology - instead of the growth of flash lithography. However for that to be unfair we have to assume that the growth we see in the HD market aligns well with advances in their underlying technology. Fair enough. However that doesn't actually support his argument of "it won't happen soon" unless we look at the current bit densities of both products and the growth rates of their underlying technology. This year we saw the release of a 3TB 3.5" drive - if we use Mr. Newmans argument we can assume this is following reasonably closely with the capacity of the underlying technology (not necessarily limited to theoretical capacity but also production etc...) . Sadly for comparison there isn't a flash drive that pushes the capacity of the 3.5" form factor. However if we assume that the logic for managing multiple drives is small compared to that of the storage itself we might be able to make the following comparison. A 3.5" drive is 147mm x 101mm x 46mm a micro sd flash card is 15mm x 11mm x 1. Which means we could fit at least a 9 x 9 array on a plane the same size as a 3.5" drive. Again I'm going to assume that whatever spacing is required it isn't going to require more than the same thickness above and below each "card". If true we can layer fifteen of these "sheets" of sd cards in the same volume as the 3.5" drive. That gives us about 1200 cards and given that the highest density card is 32GB which means we could fit about 37TB of this kind of flash memory in the 3.5" space. Now I'm not going to try to estimate what the growth rates of each underlying technology is but from my analysis here I'd say it's HD's that need to catch up.
Our one friend who killed herself's account. That would be nice. Having her profile continually show up with the "you haven't talked to X in a while send them a message and reconnect with them" box. Doesn't actually make facebook win any sensitivity awards in my book.
Ok, but despite aesthetically impressive testing facilities...what good did any of this do? No offense to apple but my iPhone is about as good as any other wireless device I've used. I have literally never been in a scenario where it's wireless worked and everything else didn't. There are a few low signal and outright dead areas on my train ride home. You can pretty much watch Blackberry users and iPhone users like groan as their access gets killed. So yes, Apple I'm glad you spent $X on testing equipment and I'll even give you the benefit of the doubt that you used it in the new iPhone but doesn't it seem like you could have achieved a better result by simply increasing the number of field test units/areas? Assuming the problems people are reporting are real, it sure didn't take long to show up. I even understand that you can't cover every usage case but at least I'd like to hear what it is you missed and how you are addressing it.
Also if this is a PR move I'm not so sure it's a good one. Sure the surrealistic test labs will have the fanboys drooling but if the idea is we do awesome testing if the implication isn't "Other people don't" - and considering that nobody else is having quite this problem it casts some doubts of the utility of replicating movie sets for test labs.
One thing I've had to specify with code review is length and scope. I had someone give in 500 lines or a fully functional (but lame) application. I usually ask for a short ( 200 line ) function that solves some specific problem. On occasion I also give a take-home problem where they need to send fully functional code back. Usually ask simple coding style questions in the interview but mostly do what you mentioned and give "situational" questions. So coders might be asked what happens when you notice a security flaw or realize there's a security flaw in code you just released. Oh and make sure that you cross-examine them on their code sample during the interview - especially make or suggest some small change and ask the candidate to explain what that would do to the function - someone who didn't really write the code often can't answer a question like that on the spot. If they can and they didn't write the code then they've spent enough time with it that it suggests they could have.:)
Personally (and I love that someone below mentioned Ahmdals law). The problem isn't as you said about specific language constructs but that there isn't any general solution to parallelism. That is to use Brook's illustration, problems we try to solve with computers aren't like harvesting wheat - they aren't efficiently divisible to an arbitrary degree. We do know of a few problems like this which we call "embarassingly parallel" but these are few and far between. So GPU's are great MD5 crackers, protein folders and I personally *love* writing CUDA code but I don't suffer from the delusion that this is somehow a revolution in software. That the usual day-to-day tasks are going to be affected. So the idea that GPUs are moving into the server room seems optimistic because the majority of stuff in there is pretty mundane.
That said I'd say I wonder if there aren't some architectural limitations on GPUs e.g. memory protection and if we really wanted to use these for general purpose computing and added them would we lose performance? In other words are we just making some kind of cores-to-features tradeoff?
This is a good point. Although people like Orac (a regular scienceblogger) talk about their research they aren't actively promoting it. Similarly I can be assured that while a sciencebloggers employer either ignores, supports or simply tolerates having someone on their staff blog about science - we can be reasonably assured that they aren't being paid to do so. In both of those cases I can't say I'd have the same level of confidence for Professor Pepsi...or whomever ended up there.
Also I can't stand it when people like Adam Bly post what is clearly a thin and minor reason for having a huge company sponsor a blog and avoid the obvious and real reason: Money. Adam - sure perhaps industry is the 'interface to science' for most people (whatever the hell that means) however that's not the reason you are talking to Pepsi, L'Oreal, etc.. instead of the chief scientist of Nobody Inc. it's money and exposure and I wish you had said that up front and clearly laid out the monetary - down to the dollar - benefit you were getting.
I kind of hope you're joking. Otherwise aren't you saying that teachers and educational institutions are responsible for a students inability? or worse yet lazyness? That other than forking over the money (which countries like mine can mean as little as signing the loan forms) they aren't actually responsible for anything else?
I believe that anyone is capable of entering into any occupation with sufficient effort and sufficient funds. I believe that lacking that later the government should provide some mechanism of attaining it. However that's light-years from believing that we are mandated to fulfill anyone's wish to achieve any degree.
I've worked for educational institutions and in one case I recall them attempting to deploy an anti-cheating countermeasures and got shouted down by students. Also given that many public institutions are compensated by degree completion working against cheating costs the institution not just for the price of technology but in the lost tuition and public funding. To me, this seems like an institution who cares about the quality of their student's education.
True but remember when all sorts of devices like printers came with a USB cable? Now most don't because USB is ubiquitous. I'd love to have a P/S standard that would treat laptop power supplies just like that. Assume the user has one, if not it's up to retail to sell them one, creates markets for "specialized" products (high durability, high power, ultra portable). Rather than pay for similar circuitry over and over again. I'm exceptionally bitter that I have to carry two *UNIVERSAL* power adapters to handle the four machines in use around my house (two laptops and two netbooks).
However, I think there are some useful things to think about "crowd sourcing". The first is that in many problems, getting a lot of estimates based on different criteria will end up giving you a better overall estimate. The individual errors will cancel each other out.
It is not very difficult to show a mathematical proof of this.
The problem isn't that some samples can give you a more accurate mean/median/mode whatever than any single sample. It's knowing which set of samples this is true for.
And it is not very difficult to show what kinds of problems this will work well for (something with a numerical solution for which the vast majority of the people in the crowd have a non-random method of obtaining an estimate)
I'll call shenanigans. First of all you've created an epistomological problem for yourself. At best you can only come up with a probabilistic statement as to how likely someone isn't guessing to a problem with a numerical answer. Not to mention how you would establish that fact for the vast majority of... some population. Put another way: How do you know what everyone knows how to know?;-)
However other types of problems are good for "crowd sourcing" too. One of these things is making a FAQ.
"crowd sourcing" this may be but it is much different than what WoC is about. A FAQ is to represent what is non-obvious to people who are interested in some subject. Not only that FAQ's have maintainers (often someone who is knowledgeable in the subject matter) so there is selection bias happening (which WoC doesn't have). It's also unclear what the comparable benchmark is for FAQs. Crowds are supposed to have more accurate answers than any one of their constituents. How is this applied to FAQs?
I have no doubt that some groups will self-organize effectively but self-organization seems orthogonal to WoC.
Given you did take the time to read the book you could have at least fact checked the Ox study before claiming that there is no data recorded on it.
Given that you read my post you could have at least fact checked your own before spewing nonsense.
I did NOT claim that there "is no data" I claimed that "If the data was not recorded" it's conclusion may not be true. I admit I am surprised that the data was preserved - although unless this experiment was repeated several times with the same crowd (depending on what the hypothesis is) with different cows. It still doesn't lend much support to the idea about the 'wisdom of crowds'.
The concept of the Wisdom of a Crowd isn't that the crowd is some God infallible and immutable. The concept is that the crowd is only better then the individual.
However it's clearly not true in all cases (i.e. it is trivial to construct a crowd where this is not true), seemingly difficult to demonstrate usefully in most cases (i.e. How would you sample 'most crowds'?). Even Surowiecki, who's book you are saying props up this argument doesn't believe it to be true for any crowd since he gives his criteria. Again, I'm not sure if you're responding to my post anymore or just going on about your pet theory. It's not just that crowds don't know everything ('immutable'? Where did that come from), but not every crowd knows better than it's smartest constituent on any subject. Even worse Surowiecki lays down some criteria as to which crowds are 'smart' and which are not. However this criteria seems to only make the question more vague. Ergo even if some crowds are likely to produce a more accurate answer than their smartest constituent on some specific question (which is statistically bound to happen) - there doesn't appear to be a way to derive this reliably. In other words - Surowiecki hasn't told us anything useful.
I find this talk a little like taking the set of all functions that are defined for all the values which can be represented by the concatenated ASCII values of the characters for any well-defined English true/false question. It's easy to assume that somewhere there exists a group of functions which produce an odd value when the question is true and an even value when the question is false for questions with a maximum length of 30 characters. In other words we KNOW there is a function that 'knows' the right answer to every question of this type. However given that the only way we know to demonstrate this is to go through and validate each question. It's not particularly USEFUL!
It is a nice counter to Charles Mackay. It's funny how people like to say crowds are morons and then try to prove it Scientifically like Francis Galton did with his Ox Experiment
Considering both are anecdotal it's hard to say what is a "counter" of what. However as an argument I found that book pretty lame - it's a lot of anecdote. Not to mention that I think people extend the title beyond what Surowiecki intended he doesn't even assert that all or even MOST crowds are smart. Rather that crowds which have some attributes are smart. However most of those attributes are far more vague than the questions posed which makes the problem of determining a smart crowd from a dumb crowd a harder problem than asking the question. I'd add that even given his assumptions are true for some crowd the kind of question is crucially important. It must be limited in scope. i.e. multiple choice or have some generally understood bounds (that is if we asked a bunch of people what the weight of something nobody knew what it was you wouldn't get good answers). You can't "average" the cure for cancer, or the proof for P=NP, etc...
So, to me anyway even if James is correct in his assertions that some crowds are smarter than all of their constituents this information isn't very useful.
Oh and the ox experiment isn't even close to useful since it wasn't repeated and unless the results were recorded somewhere it's lesson might not even be true. Perhaps there was someone with the same guess as the crowd or within one pound of the crowd - a variance you might be able to attribute to chance. Which would mean that the crowd isn't appreciably more intelligent than it's smartest person. This wasn't repeated multiple times so it's difficult to figure out if the crowd vs. constituents is simply a random occurrence. Not to mention that it's possible there was some bias in the crowd (it sure wasn't a random sample), etc... So again it's at best unclear if this is a useful trait.
Slashdot Mirror
In Ontario it's $45 as a co-payment for any non-hospital originating, health card holding, medically necessary local trip. You can be exempt to this fee if you are in a variety of low-income situations. If, like a lot of us you have supplemental insurance provided by your employer then this is often covered. It is $0 if it is hospital to hospital (in the same situation above). It's $240 (or more if it's an air-ambulance) if you do not have a valid health card or the trip is considered medically unnecessary. So yes if you meant we pay a *small portion* of ambulance services in a few *reasonable cases* I'd agree with you but this is completely different that what happens in less communist medical systems :-)
From what you're saying I assume you mean test equipment can fail. Sure but so can ears and all the 'software' in-between in fact it's a lot easier to prove that test equipment is functioning correctly than everything from my ears to my brain. It's because of this I tend to trust equipment - assuming it can be tested and validated for it's function too rather than my ears. With regard to shorting the 5v line on the SATA cable - actually this is pretty astronomically unlikely. Since the SATA cable can't tell when it's sending data or audio it would either need to be a consistent error rate. Which would would likely cause other problems - unless you think the error tolerance of a piece of executable code is somehow better than audio data. I appreciate the idea that someone wants to do things right but the issue with relying on ones ears is that there is no diagnostic to ensure proper operation. Attempting to replicate the problem doesn't help unless you are doing so in an environment where you are blinded. Otherwise how does this differ from audiophile nutcases who do A/B listening tests on $400 volume knobs?
I'm not so sure this is 'the other way around'. To me that would mean something that is shown to be good in objective tests but somehow a real problem is found with ones ears. It also depends on how you define "problem". I mean if you define the term to mean "something I hear" well then it's entirely possible that "fixing" that is as much psychosomatic as anything else.
Audiophiles frequently find differences where none exist...and in other news water is wet.
Let me first say that this is great news - if it turns out to be true however following the addage of most published research is false. It's worth keeping in mind that this has 20 controls, 20 ASD and 19 ADHD - according to the article they could distinguish the ASD diagnoses from the controls and the ADHD but considering that according to the DSM IV autism can have close to 100 unique presentations. I wonder how much this actually demonstrates.
I'm not sure if the brackets were literal or just a typographical convention. However historically it wasn't uncommon to see boxes instead of variables at the younger grades - they may still do today. Also without the paper it's difficult to say if this is the only kind of test they put to students. I think your argument is kind of bogus though - You could just as easily level the same criticism at word problems...or perhaps you believe that word problem need a highly rigorous and pre-defined format. Math, as I see it anyway isn't just taught so that you can manipulate symbols deterministically. To me a good test of *knowing* math is to be able to recognize and solve mathematical problems without needing it to be spoon fed to you. Otherwise...what's the point? Machines already outstrip our ability to compute - why outside of extremely simple addition and subtraction would you bother ever learning a lot of math EXCEPT to use it so that you can recognize solve problems that are not handed to you in exam question format.
Agreed but I think the reason there's conversation around this point at all is how terms like "tech savvy" are used to describe the younger generations of net users. While Kipling in his day could have been called "car savvy" you wouldn't call the majority of drivers today to be so. Ergo it's just backlash (perhaps well deserved) by those of us who are thoroughly underwhelmed by the lack of understanding espoused by the so-called "tech savvy" group. Sometimes the ignorance is so great it appears to be worse when measured by the standards of other technologies which have been subsumed into common use. i.e. I'd expect, when pressed car users could describe more completely the operation of a car than most kids can describe the operation of the internet. Anyway, this labeling may not even be the fault of the generation it's being applied to. I wouldn't be surprised if it was simply the usage of a group technically incompetent enough not to know the difference between "able to operate" and "understanding how it works".
I don't buy it...
The concept of 'learning languages' that is - that is there are languages that lend themselves to learning programming in general. I' wrote code as a hobby for about 12 years and after university I did it for 16 years professionally. I now manage a team of developers. Without a specific problem (i.e. writing a web application, writing a driver, etc..) I would expect any of these languages would do.
In my humble opinion, the point of learning to *program* as opposed to learning to operate an application (which is closer to what drag and drop environments are) is that you have learned how to approach a problem solving by using algorithms and in the end you possess a set of "mental tools" which allow you to approach any decidable problem.
Shouldn't that be 'as' he dies?
Ok I'm going to give a more realistic example...
Take a look at the OCZ Colossus it's a pure flash 3.5" drive which was produced last year which has a capacity of 1TB. Now mind you this was engineered by tacking together a bunch of flash on a SATA controller, then plugging two of these into another SATA controller to produce a RAID of these two cards. IMHO this this design was not made to create the highest capacity drive possible.
However it might be a better benchmark for figuring out where our limits are for example..
This drive puts about 5mm between each Flash package. Given that Toshiba has a 128 GB Flash chip/controller launching this fall. Considering that the carrier measures 17mm x 22mm x 1.4mm we can probably put a 4 x 5 Array of these on a card like the ones in the Colossus. Either by double-siding the board or adding more boards (but not necessarily by using bulky SATA connectors). Let's say we can put four of these hypothetical boards into a 3.5" chassis. That still clocks us in at 10TB. Which is still 3x what we have going on with spinning disks. So even if we assume that the density increases for Disk vs. Flash are the same. I'd argue that economies of scale will soon make this cost effective.
The No Code Publish approach seems reasonable to me: Publish flaw to everyone - including CERT, Vendors but include no code or exploit for anything publicly readable. Give the vendor your exploit code an a deadline after which you will publish the exploit. If no fix appears by the deadline then you publish.
I agree (I realize that you're being ironic). Overall I think it was an interesting article but his arguments aren't really strong enough for example: Bit Density Comparisons vs. Cell Density Comparisons: Or that's what I'm going to call the part where he's talking about lithography. I get that when vendors are showing you growth curves they are showing you bit density not cell density. That is the growth we are seeing aligns much more with products being shifted to pre-existing higher density technology - instead of the growth of flash lithography. However for that to be unfair we have to assume that the growth we see in the HD market aligns well with advances in their underlying technology. Fair enough. However that doesn't actually support his argument of "it won't happen soon" unless we look at the current bit densities of both products and the growth rates of their underlying technology. This year we saw the release of a 3TB 3.5" drive - if we use Mr. Newmans argument we can assume this is following reasonably closely with the capacity of the underlying technology (not necessarily limited to theoretical capacity but also production etc...) . Sadly for comparison there isn't a flash drive that pushes the capacity of the 3.5" form factor. However if we assume that the logic for managing multiple drives is small compared to that of the storage itself we might be able to make the following comparison. A 3.5" drive is 147mm x 101mm x 46mm a micro sd flash card is 15mm x 11mm x 1. Which means we could fit at least a 9 x 9 array on a plane the same size as a 3.5" drive. Again I'm going to assume that whatever spacing is required it isn't going to require more than the same thickness above and below each "card". If true we can layer fifteen of these "sheets" of sd cards in the same volume as the 3.5" drive. That gives us about 1200 cards and given that the highest density card is 32GB which means we could fit about 37TB of this kind of flash memory in the 3.5" space. Now I'm not going to try to estimate what the growth rates of each underlying technology is but from my analysis here I'd say it's HD's that need to catch up.
Our one friend who killed herself's account. That would be nice. Having her profile continually show up with the "you haven't talked to X in a while send them a message and reconnect with them" box. Doesn't actually make facebook win any sensitivity awards in my book.
...ironically it was on my way to the hotel that Google had reserved for me prior to my interview.
Ok, but despite aesthetically impressive testing facilities...what good did any of this do? No offense to apple but my iPhone is about as good as any other wireless device I've used. I have literally never been in a scenario where it's wireless worked and everything else didn't. There are a few low signal and outright dead areas on my train ride home. You can pretty much watch Blackberry users and iPhone users like groan as their access gets killed. So yes, Apple I'm glad you spent $X on testing equipment and I'll even give you the benefit of the doubt that you used it in the new iPhone but doesn't it seem like you could have achieved a better result by simply increasing the number of field test units/areas? Assuming the problems people are reporting are real, it sure didn't take long to show up. I even understand that you can't cover every usage case but at least I'd like to hear what it is you missed and how you are addressing it.
Also if this is a PR move I'm not so sure it's a good one. Sure the surrealistic test labs will have the fanboys drooling but if the idea is we do awesome testing if the implication isn't "Other people don't" - and considering that nobody else is having quite this problem it casts some doubts of the utility of replicating movie sets for test labs.
One thing I've had to specify with code review is length and scope. I had someone give in 500 lines or a fully functional (but lame) application. I usually ask for a short ( 200 line ) function that solves some specific problem. On occasion I also give a take-home problem where they need to send fully functional code back. Usually ask simple coding style questions in the interview but mostly do what you mentioned and give "situational" questions. So coders might be asked what happens when you notice a security flaw or realize there's a security flaw in code you just released. Oh and make sure that you cross-examine them on their code sample during the interview - especially make or suggest some small change and ask the candidate to explain what that would do to the function - someone who didn't really write the code often can't answer a question like that on the spot. If they can and they didn't write the code then they've spent enough time with it that it suggests they could have. :)
Personally (and I love that someone below mentioned Ahmdals law). The problem isn't as you said about specific language constructs but that there isn't any general solution to parallelism. That is to use Brook's illustration, problems we try to solve with computers aren't like harvesting wheat - they aren't efficiently divisible to an arbitrary degree. We do know of a few problems like this which we call "embarassingly parallel" but these are few and far between. So GPU's are great MD5 crackers, protein folders and I personally *love* writing CUDA code but I don't suffer from the delusion that this is somehow a revolution in software. That the usual day-to-day tasks are going to be affected. So the idea that GPUs are moving into the server room seems optimistic because the majority of stuff in there is pretty mundane.
That said I'd say I wonder if there aren't some architectural limitations on GPUs e.g. memory protection and if we really wanted to use these for general purpose computing and added them would we lose performance? In other words are we just making some kind of cores-to-features tradeoff?
This is a good point. Although people like Orac (a regular scienceblogger) talk about their research they aren't actively promoting it. Similarly I can be assured that while a sciencebloggers employer either ignores, supports or simply tolerates having someone on their staff blog about science - we can be reasonably assured that they aren't being paid to do so. In both of those cases I can't say I'd have the same level of confidence for Professor Pepsi...or whomever ended up there. Also I can't stand it when people like Adam Bly post what is clearly a thin and minor reason for having a huge company sponsor a blog and avoid the obvious and real reason: Money. Adam - sure perhaps industry is the 'interface to science' for most people (whatever the hell that means) however that's not the reason you are talking to Pepsi, L'Oreal, etc.. instead of the chief scientist of Nobody Inc. it's money and exposure and I wish you had said that up front and clearly laid out the monetary - down to the dollar - benefit you were getting.
I kind of hope you're joking. Otherwise aren't you saying that teachers and educational institutions are responsible for a students inability? or worse yet lazyness? That other than forking over the money (which countries like mine can mean as little as signing the loan forms) they aren't actually responsible for anything else?
I believe that anyone is capable of entering into any occupation with sufficient effort and sufficient funds. I believe that lacking that later the government should provide some mechanism of attaining it. However that's light-years from believing that we are mandated to fulfill anyone's wish to achieve any degree.
I've worked for educational institutions and in one case I recall them attempting to deploy an anti-cheating countermeasures and got shouted down by students. Also given that many public institutions are compensated by degree completion working against cheating costs the institution not just for the price of technology but in the lost tuition and public funding. To me, this seems like an institution who cares about the quality of their student's education.
True but remember when all sorts of devices like printers came with a USB cable? Now most don't because USB is ubiquitous. I'd love to have a P/S standard that would treat laptop power supplies just like that. Assume the user has one, if not it's up to retail to sell them one, creates markets for "specialized" products (high durability, high power, ultra portable). Rather than pay for similar circuitry over and over again. I'm exceptionally bitter that I have to carry two *UNIVERSAL* power adapters to handle the four machines in use around my house (two laptops and two netbooks).
However, I think there are some useful things to think about "crowd sourcing". The first is that in many problems, getting a lot of estimates based on different criteria will end up giving you a better overall estimate. The individual errors will cancel each other out. It is not very difficult to show a mathematical proof of this.
... some population. Put another way: How do you know what everyone knows how to know? ;-)
The problem isn't that some samples can give you a more accurate mean/median/mode whatever than any single sample. It's knowing which set of samples this is true for.
And it is not very difficult to show what kinds of problems this will work well for (something with a numerical solution for which the vast majority of the people in the crowd have a non-random method of obtaining an estimate)
I'll call shenanigans. First of all you've created an epistomological problem for yourself. At best you can only come up with a probabilistic statement as to how likely someone isn't guessing to a problem with a numerical answer. Not to mention how you would establish that fact for the vast majority of
However other types of problems are good for "crowd sourcing" too. One of these things is making a FAQ.
"crowd sourcing" this may be but it is much different than what WoC is about. A FAQ is to represent what is non-obvious to people who are interested in some subject. Not only that FAQ's have maintainers (often someone who is knowledgeable in the subject matter) so there is selection bias happening (which WoC doesn't have). It's also unclear what the comparable benchmark is for FAQs. Crowds are supposed to have more accurate answers than any one of their constituents. How is this applied to FAQs?
I have no doubt that some groups will self-organize effectively but self-organization seems orthogonal to WoC.
Given you did take the time to read the book you could have at least fact checked the Ox study before claiming that there is no data recorded on it.
Given that you read my post you could have at least fact checked your own before spewing nonsense.
I did NOT claim that there "is no data" I claimed that "If the data was not recorded" it's conclusion may not be true. I admit I am surprised that the data was preserved - although unless this experiment was repeated several times with the same crowd (depending on what the hypothesis is) with different cows. It still doesn't lend much support to the idea about the 'wisdom of crowds'.
The concept of the Wisdom of a Crowd isn't that the crowd is some God infallible and immutable. The concept is that the crowd is only better then the individual.
However it's clearly not true in all cases (i.e. it is trivial to construct a crowd where this is not true), seemingly difficult to demonstrate usefully in most cases (i.e. How would you sample 'most crowds'?). Even Surowiecki, who's book you are saying props up this argument doesn't believe it to be true for any crowd since he gives his criteria. Again, I'm not sure if you're responding to my post anymore or just going on about your pet theory. It's not just that crowds don't know everything ('immutable'? Where did that come from), but not every crowd knows better than it's smartest constituent on any subject. Even worse Surowiecki lays down some criteria as to which crowds are 'smart' and which are not. However this criteria seems to only make the question more vague. Ergo even if some crowds are likely to produce a more accurate answer than their smartest constituent on some specific question (which is statistically bound to happen) - there doesn't appear to be a way to derive this reliably. In other words - Surowiecki hasn't told us anything useful.
I find this talk a little like taking the set of all functions that are defined for all the values which can be represented by the concatenated ASCII values of the characters for any well-defined English true/false question. It's easy to assume that somewhere there exists a group of functions which produce an odd value when the question is true and an even value when the question is false for questions with a maximum length of 30 characters. In other words we KNOW there is a function that 'knows' the right answer to every question of this type. However given that the only way we know to demonstrate this is to go through and validate each question. It's not particularly USEFUL!
It is a nice counter to Charles Mackay. It's funny how people like to say crowds are morons and then try to prove it Scientifically like Francis Galton did with his Ox Experiment
Considering both are anecdotal it's hard to say what is a "counter" of what. However as an argument I found that book pretty lame - it's a lot of anecdote. Not to mention that I think people extend the title beyond what Surowiecki intended he doesn't even assert that all or even MOST crowds are smart. Rather that crowds which have some attributes are smart. However most of those attributes are far more vague than the questions posed which makes the problem of determining a smart crowd from a dumb crowd a harder problem than asking the question. I'd add that even given his assumptions are true for some crowd the kind of question is crucially important. It must be limited in scope. i.e. multiple choice or have some generally understood bounds (that is if we asked a bunch of people what the weight of something nobody knew what it was you wouldn't get good answers). You can't "average" the cure for cancer, or the proof for P=NP, etc...
So, to me anyway even if James is correct in his assertions that some crowds are smarter than all of their constituents this information isn't very useful.
Oh and the ox experiment isn't even close to useful since it wasn't repeated and unless the results were recorded somewhere it's lesson might not even be true. Perhaps there was someone with the same guess as the crowd or within one pound of the crowd - a variance you might be able to attribute to chance. Which would mean that the crowd isn't appreciably more intelligent than it's smartest person. This wasn't repeated multiple times so it's difficult to figure out if the crowd vs. constituents is simply a random occurrence. Not to mention that it's possible there was some bias in the crowd (it sure wasn't a random sample), etc... So again it's at best unclear if this is a useful trait.
Yeah, a lot of us did...then some of us realized that there's more to it than IF...THEN and we went to university and learned what that was.