It's funny sometimes it seems no matter the principle there's someone out there who wants to vastly inflate it's importance. Tyranny of the normal distribution I guess. You very rarely ever attempt to determine a value from a graph (Such as in the OECD science example). For example if I want to know the dose of anesthetic a patient gets I don't trot out a graph and take a ruler and do some back of the envelope calculation. With regard to your ideas about fraudulent information in graphs. This is exactly what these exams are teaching people. That graphs are a trusted source of information. IMHO if you taught students the exact opposite lesson - graphs are generally lies - you would probably end up no worse off and possibly better.

Interestingly as a maths student nobody ever needed to teach us to "read" graphs. Make graphs yes, understand how to annotate experimental error, sure but there's no need to do mindless exercises all concerned with deriving data from a graph which is probably beyond it's accuracy anyway.

Given that the article is somewhat focused on the ScienceDebate questions and with the notable exception of the one about climate change. The third option I'd see is that those questions are pretty unfocused and in one case - education a little deceptive.

For example are they referring to the OECD exam results? I downloaded and wrote that math exam and I found it to be weird. In some cases I'm not sure the questions were even about math and very often not the kind of math you use in science (there were huge numbers of graph reading questions). Not to mention that the purveyors of the exam themselves only recognized three statistically significant groups (those working at, beneath and above the median). Not to mention when people start throwing ordinal values around it often makes me wonder how much they actually know about science or math. Ordinals provide zero information about the distance between ranks which is far more important than being 1st or 17th.

Thanks for the link to the current exam. I wrote the 58 question 2008 one in about an hour. This one seems harder and I agree it is oddly worded and while I'm not in a position to comment on what children are capable of. It seems at least plausible that it's obscure enough to be testing something other than the ability to solve mathematical problems.

That said, I think the idea of the article is one of bounding. i.e. If an adult can't do this then... or if an adult can be successful and not know a single question then... which is probably my greatest objection. I would expect that any person who does some degree of quantitative research to be able to pass this kind of test. Likewise I would expect someone with an undergraduate degree in science to be able to do most of these things. I also expect these people to be able to write an exam without having anxiety problems.

The underlying misunderstanding with the argument presented is that while it may be possible for person X to succeed while being terrible at math may in fact be true but that doesn't make it likely nor does it mean that school isn't supposed to prepare people for a wide variety of careers not just ones where they need to do more math than just make change.

I think the general idea then would be that these offices are not necessarily linked to a single business - in most cases. You have co-workers but they may not be from your business. Clearly this raises some IP/Trade Secret questions but probably not insurmountable ones.

We use take-home problems. We also do some verbal testing on concepts - to see how well they can talk about things developers tend to discuss. We have used written tests but they weren't that good.

I'm actually working on trying out a "live" requirements gathering test but as hiring is only a small part of my job it's not yet been validated.

Depends on what you need from ZFS. I've been using it under Linux for two years now. I don't have to use PPS for anything so I can't comment on that but considering that the context of the OP was "on the desktop" I'd think this is out of scope. Anyway Linux while doing a good job of being most things to most people. It's unlikely that anyone here is claiming it's everything to everyone.

Just to let you know I'm not disagreeing with you but I do think what you are saying needs to be highlighted given the number of times I have to deal with "BSD is X" nonsense from people who have absolutely abysmal math skills. Not only are you correct about the variability of the data but the netcraft chart is a weighted ranking of a number of metrics and while these may be perfectly reasonable for evaluating a hosting provider. None of them are intrinsically OS related and only two have to do with reliability. One of those two is the same for every system down to #30 (on the Feb 2011 data)! For example in Feb 2011 # 6 serverbeach (a Linux based provider) outperforms BSD (and everyone above it) in connect latency, first byte, total and kb/s. The only reason this is in #6 is because it's failed requests differs by 0.008%. While someone might argue that this difference is important in some context as we can see from the two samples you gave it's not entirely clear how much each one of these OS's vary. While I wouldn't necessarily expect the "best" os to always be in the top. I would expect that - over time - the variance in the data to be smaller. Without this, it's impossible to derive anything about the OS from the "failed req %" (assuming it isn't simply a function of completely irrelevant things like their upstream ISP's)

I don't really see that point of the implication that Android Linux isn't optimized everywhere. It isn't, nothing is, there's no point really.

The question isn't when will Intel create a power efficient CPU (as you could argue that they have - such as the Atom Z series) but what will it take them to match ARM's performance per watt. In other words once we have a 1W Atom - which is probably pretty close to the consumption of the A8. What did Intel give up to close that gap? Die shrink? Die size? (aka drop functionality) Or will some other factor make it all moot? (battery tech, market change)

So far you haven't made much of an argument for ARM being more incompatible than various versions of x86. - ie. I can't run AES instructions on anything but the 2nd Gen Nehelem. Not only that (and this may be true for ARM but I don't code in that much) but there are lots of contradictory optimizations in x86. For example just taking block moves what is optimal for 286 (Unrolled loop if move can fit in cache or REP MOV), is suboptimal for 386-486 (unrolled loop for in cache move otherwise REP MOVS - for blocks aligned on word or double word boundaries) which is suboptimal for Pentium (MMX), etc.. this isn't even counting dealing with the wide variety of cache sizes.

I await your better definitions. In the mean time, I'll use Turing's.
An awesome display of missing the point. Congratulations!

To quote Turing: 'The original question, "Can machines think?" I believe to be too meaningless to deserve discussion. Nevertheless I believe that at the end of the century the use of words and general educated opinion will have altered so much that one will be able to speak of machines thinking without expecting to be contradicted.'

So as you can see what Turing is talking about is almost the exact opposite of what you are asserting. You appear to say that this test (regardless if you do it on a computer or a person) tells you that the object you are speaking to has understanding of the subject and is intelligent (but spend a fair amount of time trying to worm your way out of defining what either of those things actually are). Whereas Turing seems to be saying that the term "think" will - at least colloquially - come to mean the results of this kind of test. At least Turing understood that the terms are poorly defined. Too bad you don't.

Why "specific contexts" ? The topics that you can talk about in a Turing test are endless, as well as the duration of the test.
Wrong in both senses. Anyway "contexts" meant the kind of social situation. i.e. Chatting over an IM as opposed to talking in person.

The whole point about the Turing test is that the test itself provides the definition,
Moronic and incorrect. So in your world someone can produce a test that defines intelligence as being able to play a perfect game of tic-tac-toe and that would be a reasonable definition of the term. Yeah, good luck with that. I said it better earlier, the test borrows from OUR implicit tools for judging intelligence. However the term "intelligence" (other than in your world of tic-tac-toe players) doesn't have a very good definition. So we don't really know how accurate it is.

If I discuss a problem with a co-worker, I also get a good idea of how intelligent they are, and if they understand the problem.
How do you know? What standard are you comparing against? How are you recording the data? What sort of tests have you done to determine the significance of the data you are collecting? Whoops...you're probably don't none of those things in any useful way.

There is no reason why the same approach doesn't work with a computer program.
A test that has some key commonalities with some other test is good reason to believe that passing one test means passing the other. However confusing a test for what it's testing for is a classic mistake.

It's actually a case study which, you're right is low grade evidence. You establish your control by having sham sessions.

That bit should have read "on the researchers web site".

You can read the study here:

Oddly it's the only one not featured on the

It's a little interesting but it is only one subject and I'm a little skeptical about the descriptions used by the subject for the symptoms.

No offense but just listen to yourself. "it should give a good indication of true intelligence and understanding".

Why? I mean while I agree that it is plausible that someone might create a machine that is indistinguishable from a person - in specific contexts - why does that necessitate a good indication of two things that are pretty hard to define to begin with?

..says something. The point of a Turing test is a gedankenexperiment. The idea that our idea of "human" comes from the fact that we don't have any formal criteria for defining it. Instead we assume that the people we meet (one the phone or online). Are human because we can't distinguish them from being so. In truth this isn't any more a real test than going out an buying and gassing a cat is somehow a real experiment proving superposition.

That said, even if we were to formulate an experiment from what Turing talked about I don't see how 59% qualifies as "indistinguishable".