Your reply is completely orthogonal to what I'm saying. I'm not talking about whether accidents will happen. I'm talking about what happens when they do. It would be very hard for a chernobyl type level of radiation release to ever happen again accidentally.
> Yet more than 98% of radiation from nukes is released during accidents, which the paper ignores.
Citation needed.
Unfortunately a lot of anti-nuke people have used Chernobyl figures as a 'baseline' for nuclear accidents, but Chernobyl was a single freak accident and it's impossible that something like that will ever happen again. The causes of the accident (lack of containment building, embarrassingly lax security procedure, out of date reactor design) simply do not exist anymore anywhere in the world. Actually, Fukushima was also a very outdated design and it's unlikely that anything like that will ever happen again, but still within the realm of possibility.
The total amount of radioactive material put out by a coal power plant is actually larger, per unit of energy produced, than a nuclear power plant. And in a nuclear plant, the radioactivity is confined; even disasters like what happened at Fukushima-Daichi only release a relatively small amount of nuclear material into the environment. Whereas in coal the radioactivity is open to the environment. That's not to mention all the heavy metals that coal produces. Enjoy the mercury in your tuna, courtesy of china's booming economy.
People tend to project their own biases, but it doesn't damage the underlying scientific fact. People associate testosterone with aggressiveness, and that leads to the wrong conclusion that civilization requires people to be less aggressive. But there's no evidence to indicate that testosterone actually causes any aggressiveness at all. The main biological effects of testosterone are muscle mass and body hair (along with some other sex characteristics). So really, it's more valid to say civilization was correlated with less muscle mass. But in all honesty, it's just as valid to say civilization is correlated with less body hair.
Instead of getting more people interested in math, I predict it will wind up getting less people interested in CS. Count on school to turn highly interesting, mentally-stimulating subjects (like math) into boring rote memorization exercises.
I can see a 'meta-compiler' for C, COBOL, C++, and Java being a workable and useful idea (and there are plenty of open-source 'meta-compilers' for these languages), but I can't imagine it possibly being an efficient way to compile in Haskell or Python or Erlang. How would you integrate all the vastly different language features together? Answer: you can't, you have to use ugly workarounds that are slow, brittle, and very, VERY hard to debug.
Until I am convinced otherwise, it seems that just the time wasted on debugging (is the problem in my code or in my codegen?) would warrant learning the languages directly.
Re:"Featuritis" in the whole computing ecosystem
on
Getting Back To Coding
·
· Score: 4, Insightful
Rapidly introducing new tools and deprecating old ones unnecessarily is part of Microsoft's business strategy.
It's the way the Universe works at every level, as far as we know. It doesn't just apply to atoms and electrons, but to rocks, people, and planets. Quantum effects occur at the macroscale. Two important examples are lasers and superfluids.
I know some people are going to say, "But doesn't relativity take over at the scale of galaxies and such? And isn't relativity incompatible with QM" And the answer is: no, they definitely aren't incompatible. In fact most proposals for unified theories have been based on quantizing gravity. It's just that the naive way of combining these two theories (modelling gravity as a traditional QFT) doesn't work.
Well, 'proven correct' assumes that the proof itself is correct (verification), that the proof actually proves the correct thing (validation), and that the underlying system actually conforms to its formal specification (another thing that's hard to verify - formally-verified CPU design anyone?). These people seem to have done their homework so I'll be looking at this more closely, but based on experience it's pretty much guaranteed that there's going to be some unconsidered vulnerability somewhere.
Uhm, do you really think that something as important as assessing the performance of compression algorithms wouldn't have attracted the attention of thousands (or, more likely, hundreds of thousands) of computer scientists over the years? Open up any academic journal that deals with this stuff even tangentially and you find many examples of different metrics for assessing compression performance. And there's nothing new about this 'score'. Dividing ratio by the logarithm of the compression time is a very widely-used theoretical scoring function; I can find references to it from the 90's. This particular form of that score may be new, but gweihir is right; such a score doesn't give much information and has very little use.
This isn't just a silly mistake. NASA bases its budgetary decisions on the price sub-contractors give for various jobs. These sub-contractors often give an intentionally misleading cost underestimate.
As an addition to 0123456's reply here, there isn't even a concrete plan yet to use the SLS to launch deep-space manned missions. The orion project, as it's currently being developed and funded, will not send humans outside of Earth orbit.
> Given that this would be upwards of 2000x higher density than a lithium battery, does that mean a 60Wh laptop battery has the explosive capacity of half a stick of TNT?
You actually can make a laptop battery explode with impressive force; but it's hard to trigger the detonation. In fact, the comparison with TNT is actually a pretty good one, because if you set fire to TNT it will just smolder and burn. It needs to be set off in a very specific way. But the higher you go on the energy scale, the more precarious the thermodynamic balance becomes, and the harder it becomes to keep a substance's chemical energy on the non-shattering-steel-and-tearing-off-limbs part of the scale.
> TL;DR: Comparing batteries to explosives is a bad idea. They're completely different. A battery isn't going to explode, it's going to burn.
But batteries already can and do explode - even though they are engineered not to. And at some level of energy density, a chain reaction becomes extremely hard to prevent. I agree that it might be pretty premature to talk about the characteristics of batteries made of unobtainium, but with 30x the chemical energy density of the best known sources, I think we can safely assume that you'd be nervous about having such a battery in your laptop or pocket.
If you've got a battery with that kind of power density, it's actually a formidable explosion hazard. With thirty times the energy density of rocket fuel, even a minor internal short would cause a chain reaction that would make a battery pack the size of a laptop battery explode with the power of over one hundred kilograms of TNT.
> Now we're looking at (maybe) 11 years to develop a working rocket to go to an asteroid.
It's worse than that. There will be no deep-space journey to an asteroid. Instead, a near-Earth asteroid would be selected or a small asteroid will be moved to near-Earth orbit using unmanned robotic craft. The 'manned asteroid mission' will not go any further than the Apollo missions did. And it would not do anything other than just take some samples and bring them back to Earth. Little in-situ science, and definitely no in-situ resource extraction. It really raises the question of why we're sending up humans in the first place.
There _may_ be deep-space (i.e. anything outside of Earth orbit) missions in the 'future', but they would need big and complex manned spacecraft that have yet to emerge from the drawing board.
We're not going outside of Earth orbit any time soon, not if we're to rely on NASA.
It's currently being done in a way that makes in inseparable from the SLS rocket, an out-dated and over-budget project enabled by government inertia and congressional pork. Also, the Orion MPCV itself doesn't represent much of an upgrade over existing manned space capsules; if it's to go anywhere outside of Earth orbit it's going to need a much larger and more complex space habitat attachment: http://en.wikipedia.org/wiki/D... which has yet to be developed.
> Computers don't understand handwaving and "you know what I mean" -- yet (while mathematicians are very fond of it)
Yup.
> OTOH it's not so clear-cut in that computer programming is also communication between people
No. If all you wanted to do was tell someone how to, say, divide two numbers, you wouldn't tell them through a computer program. You'd tell them through English. Computer programs are always (except in very rare circumstances) ways of communicating with computers. That they can be written in a way to also be useful for other humans is secondary; you can also write French in a way that English-speaking people would find easier to understand.
Of course, there's pseudo-code, which IS actually intended for human communication, and it's an interesting case study since it 'borrows' from computer languages but has the same kind of ambiguities and nuances as human language. It's also interesting in that lots of high-level languages have been efforts to create 'executable pseudocode' i.e. pseudocode that removes imprecision in intent.
Any sort of ambiguity means there must be a method of resolving the ambiguity. This is true no matter if you're talking about human language, math, or programming. In language and math, the method of resolving ambiguity is context. That is, the person you're communicating the concepts with presumably knows enough about what you're talking about to be able to resolve the ambiguity.
Note that both math formulas and programs are pieces of communication, nothing more. Programs are a way of communicating intent to a computer. Mathematical notation is a way of communicating intent to another human being.
Being skilled in a subject is mostly how much you practice at it. I'm sure that if you put as much time into math as you did programming, your opinion wouldn't be the same. I'm a mathematician-turned programmer and I agree with the article.
One thing, though, is that programmers have a lot of tools to help them out. High-level languages like python, interactive development environments, debuggers, and of course being able to run the program and see how it behaves. All of these tools serve to hugely augment one's natural brain capacity (over, say, having to code everything in assembly). But mathematicians have no such tools yet. To be a mathematician, you have to juggle around numerous concepts in your head simultaneously and make sure they all work with each other. Studies have shown that one of the best predictors of mathematical ability is working memory, and it's not surprising why.
By the way, why don't mathematicians adopt tools that let them easily and automatically check that their work is correct? I think it's entirely a cultural issue, not a technological one. Computational tools for mathematics (like Coq and Agda) are still stuck in the 'assembly' period. They're hard to use, because there has never been a strong demand for them from the wider math community and so they never developed as well as computer languages did. And when the math community got involved in computer science, they tended to target more abstract stuff like proof systems and category theory, instead of just sticking to the basics and coming up with a proof checker that's as easy to use as python is (and there's absolutely no reason why such a thing couldn't be made). So I'd say mathematics is a good 5 decades behind computer science. But it has to catch up eventually.
Having a good supervisor is extremely important. The arrangement where your supervisor is a person who is knowledgable, up-to-date, and respected in their field, and draws on his years of experience to guide your through work and train you as a scientist, is the ideal on which the supervisor-student relationship is based on. A person like that more than deserves to have their name on the work you do while under their tutelage.
But going by what I've seen, such a relationship is, sadly, rare. A lot of students are victims of supervisors who either "don't care" or have been effectively outside their field of study for so long (with all the grant-writing) that they have simply no clue about research anymore. Your first experience seems to be the norm.
Slashdot Mirror
If you want to call Pluto a planet, you'll have to call Eris, Haumea, Makemake, Ceres, and a whole bunch of others planets too.
Keep in mind that Ceres is spherical, orbits a star, and was known for 200 years but was NOT considered a planet during this time.
Language can be arbitrary, sure. But why insist on it being self-contradictory?
Your reply is completely orthogonal to what I'm saying. I'm not talking about whether accidents will happen. I'm talking about what happens when they do. It would be very hard for a chernobyl type level of radiation release to ever happen again accidentally.
> Yet more than 98% of radiation from nukes is released during accidents, which the paper ignores.
Citation needed.
Unfortunately a lot of anti-nuke people have used Chernobyl figures as a 'baseline' for nuclear accidents, but Chernobyl was a single freak accident and it's impossible that something like that will ever happen again. The causes of the accident (lack of containment building, embarrassingly lax security procedure, out of date reactor design) simply do not exist anymore anywhere in the world. Actually, Fukushima was also a very outdated design and it's unlikely that anything like that will ever happen again, but still within the realm of possibility.
The total amount of radioactive material put out by a coal power plant is actually larger, per unit of energy produced, than a nuclear power plant. And in a nuclear plant, the radioactivity is confined; even disasters like what happened at Fukushima-Daichi only release a relatively small amount of nuclear material into the environment. Whereas in coal the radioactivity is open to the environment. That's not to mention all the heavy metals that coal produces. Enjoy the mercury in your tuna, courtesy of china's booming economy.
People tend to project their own biases, but it doesn't damage the underlying scientific fact. People associate testosterone with aggressiveness, and that leads to the wrong conclusion that civilization requires people to be less aggressive. But there's no evidence to indicate that testosterone actually causes any aggressiveness at all. The main biological effects of testosterone are muscle mass and body hair (along with some other sex characteristics). So really, it's more valid to say civilization was correlated with less muscle mass. But in all honesty, it's just as valid to say civilization is correlated with less body hair.
Instead of getting more people interested in math, I predict it will wind up getting less people interested in CS. Count on school to turn highly interesting, mentally-stimulating subjects (like math) into boring rote memorization exercises.
I can see a 'meta-compiler' for C, COBOL, C++, and Java being a workable and useful idea (and there are plenty of open-source 'meta-compilers' for these languages), but I can't imagine it possibly being an efficient way to compile in Haskell or Python or Erlang. How would you integrate all the vastly different language features together? Answer: you can't, you have to use ugly workarounds that are slow, brittle, and very, VERY hard to debug.
Until I am convinced otherwise, it seems that just the time wasted on debugging (is the problem in my code or in my codegen?) would warrant learning the languages directly.
Rapidly introducing new tools and deprecating old ones unnecessarily is part of Microsoft's business strategy.
From the point of view of an electron, what is it like to pass through both slits at the same time?
It's the way the Universe works at every level, as far as we know. It doesn't just apply to atoms and electrons, but to rocks, people, and planets. Quantum effects occur at the macroscale. Two important examples are lasers and superfluids.
I know some people are going to say, "But doesn't relativity take over at the scale of galaxies and such? And isn't relativity incompatible with QM" And the answer is: no, they definitely aren't incompatible. In fact most proposals for unified theories have been based on quantizing gravity. It's just that the naive way of combining these two theories (modelling gravity as a traditional QFT) doesn't work.
Well, 'proven correct' assumes that the proof itself is correct (verification), that the proof actually proves the correct thing (validation), and that the underlying system actually conforms to its formal specification (another thing that's hard to verify - formally-verified CPU design anyone?). These people seem to have done their homework so I'll be looking at this more closely, but based on experience it's pretty much guaranteed that there's going to be some unconsidered vulnerability somewhere.
Who else thinks that it's only a matter of time before the first vulnerability is found?
Uhm, do you really think that something as important as assessing the performance of compression algorithms wouldn't have attracted the attention of thousands (or, more likely, hundreds of thousands) of computer scientists over the years? Open up any academic journal that deals with this stuff even tangentially and you find many examples of different metrics for assessing compression performance. And there's nothing new about this 'score'. Dividing ratio by the logarithm of the compression time is a very widely-used theoretical scoring function; I can find references to it from the 90's. This particular form of that score may be new, but gweihir is right; such a score doesn't give much information and has very little use.
This isn't just a silly mistake. NASA bases its budgetary decisions on the price sub-contractors give for various jobs. These sub-contractors often give an intentionally misleading cost underestimate.
> Private industry already designs and builds basically everything they do
Indeed, and dishonest cost estimates from private companies are usually the main reason for things going way over-budget.
And this is exactly where it needs to restructure itself. Dismantle the SLS program, pour more funds into research and facilities like JPL.
As an addition to 0123456's reply here, there isn't even a concrete plan yet to use the SLS to launch deep-space manned missions. The orion project, as it's currently being developed and funded, will not send humans outside of Earth orbit.
> Given that this would be upwards of 2000x higher density than a lithium battery, does that mean a 60Wh laptop battery has the explosive capacity of half a stick of TNT?
You actually can make a laptop battery explode with impressive force; but it's hard to trigger the detonation. In fact, the comparison with TNT is actually a pretty good one, because if you set fire to TNT it will just smolder and burn. It needs to be set off in a very specific way. But the higher you go on the energy scale, the more precarious the thermodynamic balance becomes, and the harder it becomes to keep a substance's chemical energy on the non-shattering-steel-and-tearing-off-limbs part of the scale.
> TL;DR: Comparing batteries to explosives is a bad idea. They're completely different. A battery isn't going to explode, it's going to burn.
But batteries already can and do explode - even though they are engineered not to. And at some level of energy density, a chain reaction becomes extremely hard to prevent. I agree that it might be pretty premature to talk about the characteristics of batteries made of unobtainium, but with 30x the chemical energy density of the best known sources, I think we can safely assume that you'd be nervous about having such a battery in your laptop or pocket.
If you've got a battery with that kind of power density, it's actually a formidable explosion hazard. With thirty times the energy density of rocket fuel, even a minor internal short would cause a chain reaction that would make a battery pack the size of a laptop battery explode with the power of over one hundred kilograms of TNT.
> Now we're looking at (maybe) 11 years to develop a working rocket to go to an asteroid.
It's worse than that. There will be no deep-space journey to an asteroid. Instead, a near-Earth asteroid would be selected or a small asteroid will be moved to near-Earth orbit using unmanned robotic craft. The 'manned asteroid mission' will not go any further than the Apollo missions did. And it would not do anything other than just take some samples and bring them back to Earth. Little in-situ science, and definitely no in-situ resource extraction. It really raises the question of why we're sending up humans in the first place.
There _may_ be deep-space (i.e. anything outside of Earth orbit) missions in the 'future', but they would need big and complex manned spacecraft that have yet to emerge from the drawing board.
We're not going outside of Earth orbit any time soon, not if we're to rely on NASA.
It's currently being done in a way that makes in inseparable from the SLS rocket, an out-dated and over-budget project enabled by government inertia and congressional pork. Also, the Orion MPCV itself doesn't represent much of an upgrade over existing manned space capsules; if it's to go anywhere outside of Earth orbit it's going to need a much larger and more complex space habitat attachment: http://en.wikipedia.org/wiki/D... which has yet to be developed.
> Computers don't understand handwaving and "you know what I mean" -- yet (while mathematicians are very fond of it)
Yup.
> OTOH it's not so clear-cut in that computer programming is also communication between people
No. If all you wanted to do was tell someone how to, say, divide two numbers, you wouldn't tell them through a computer program. You'd tell them through English. Computer programs are always (except in very rare circumstances) ways of communicating with computers. That they can be written in a way to also be useful for other humans is secondary; you can also write French in a way that English-speaking people would find easier to understand.
Of course, there's pseudo-code, which IS actually intended for human communication, and it's an interesting case study since it 'borrows' from computer languages but has the same kind of ambiguities and nuances as human language. It's also interesting in that lots of high-level languages have been efforts to create 'executable pseudocode' i.e. pseudocode that removes imprecision in intent.
Any sort of ambiguity means there must be a method of resolving the ambiguity. This is true no matter if you're talking about human language, math, or programming. In language and math, the method of resolving ambiguity is context. That is, the person you're communicating the concepts with presumably knows enough about what you're talking about to be able to resolve the ambiguity.
Note that both math formulas and programs are pieces of communication, nothing more. Programs are a way of communicating intent to a computer. Mathematical notation is a way of communicating intent to another human being.
Being skilled in a subject is mostly how much you practice at it. I'm sure that if you put as much time into math as you did programming, your opinion wouldn't be the same. I'm a mathematician-turned programmer and I agree with the article.
One thing, though, is that programmers have a lot of tools to help them out. High-level languages like python, interactive development environments, debuggers, and of course being able to run the program and see how it behaves. All of these tools serve to hugely augment one's natural brain capacity (over, say, having to code everything in assembly). But mathematicians have no such tools yet. To be a mathematician, you have to juggle around numerous concepts in your head simultaneously and make sure they all work with each other. Studies have shown that one of the best predictors of mathematical ability is working memory, and it's not surprising why.
By the way, why don't mathematicians adopt tools that let them easily and automatically check that their work is correct? I think it's entirely a cultural issue, not a technological one. Computational tools for mathematics (like Coq and Agda) are still stuck in the 'assembly' period. They're hard to use, because there has never been a strong demand for them from the wider math community and so they never developed as well as computer languages did. And when the math community got involved in computer science, they tended to target more abstract stuff like proof systems and category theory, instead of just sticking to the basics and coming up with a proof checker that's as easy to use as python is (and there's absolutely no reason why such a thing couldn't be made). So I'd say mathematics is a good 5 decades behind computer science. But it has to catch up eventually.
Funding is a necessary, but not sufficient, condition for research to take place. It's only the first step.
Having a good supervisor is extremely important. The arrangement where your supervisor is a person who is knowledgable, up-to-date, and respected in their field, and draws on his years of experience to guide your through work and train you as a scientist, is the ideal on which the supervisor-student relationship is based on. A person like that more than deserves to have their name on the work you do while under their tutelage.
But going by what I've seen, such a relationship is, sadly, rare. A lot of students are victims of supervisors who either "don't care" or have been effectively outside their field of study for so long (with all the grant-writing) that they have simply no clue about research anymore. Your first experience seems to be the norm.