It might be similar but it's not the same mechanism. When you see an object in static, your brain knows that it's just making a guess so the guess is assigned low confidence. But here they showed that you can actually design a picture that looks random but is assigned very high confidence of being an object.
This type of phenomenon is very well known. It's not news, people have known about this sort of stuff in artificial neural nets since the 80's. I guess they just sort of assumed that deep belief nets would get around this problem, but as far as I know there's no reason to believe that. There's a related phenomenon which is assigning very low confidence to a picture that is very clearly a certain class of object - and then if you add a small bit of noise the confidence goes way up. For the interested, this is a good page which explains why some of these issues happen: http://colah.github.io/posts/2...
Just one thing I want to get off my chest: I wish this deep learning fad would die. I first started using deep belief nets around 2006 or so when Hinton published his now-infamous Science paper. I thought it was cool and used it a lot, but I knew it had limitations. Then around 2012 or so this whole thing just started becoming a hugely-hyped meme that everyone wants to get on board, without any knowledge or wisdom - they just want results. This is going to be a recipe for yet another AI "failure", when people realize that they couldn't live up to their own hype.
JPEG2000 had some serious technical issues. The reference library was itself non-compliant (I know, I used it extensively), which meant that there was no fully reference-compliant implementation available anywhere. The library was cumbersome and hard to integrate into existing image processing/viewing programs. And there wasn't much benefit to using JPEG2000; it had only slightly better compression and the wavelet scheme they used caused a lot of unpleasant artifacts (especially around large, uniformly-colored areas with gradual change of brightness).
BPG doesn't seem to have these problems; the images genuinely look better and it's easy to integrate it into the web. I don't really see a major need for a new image compression format but it would be nice to have an option to use it.
What exactly are you trying to say? That there is a scientific controversy around the idea that humans are the major cause of climate change? Well, no, there's not. Now you can dig up random people saying random things and you can change the goalposts, but you can never prove there's a controversy when there isn't. Have a read here: http://www.ncbi.nlm.nih.gov/pm...
FTA: "97–98% of the climate researchers most actively publishing in the field surveyed here support the tenets of ACC outlined by the Intergovernmental Panel on Climate Change, and (ii) the relative climate expertise and scientific prominence of the researchers unconvinced of ACC are substantially below that of the convinced researchers."
But there really isn't any controversy, no matter how many times you or other people say it or want to believe that it exists.
In this context, controversy would mean a scientific controversy over the main aspects of climate change: that it is happening and it's due to human activities. There is no scientific controversy around this.
Science is full of controversies. You can go to the 'letters' section of any reputable journal and see how much scientists argue over things. But climate change just isn't one of these issues. There might be argument over small details or specifics, but nobody familiar with the science disputes that humans are causing the climate to change.
If you go outside the USA (and a few other anglosphere countries), climate change is pretty much accepted across the political spectrum. But in the USA you have this huge party that's made it it's job to spread FUD, which is exactly what we're seeing here in this thread. It's manufactured controversy. And it works.
It's neither true that everyone is qualified to pass judgement about a scientific field, nor that everyone should just accept what scientists say blindly. The truth is in the middle: It would be great IF everyone could check for themselves, but in practice this just isn't possible, so at some level you do indeed need to trust the consensus of the literature. Of course you put in measures (like peer review and other measures) to make sure the scientists are being honest. Do transgressions happen? Obviously. We are all human. But it's really a stretch to say that all scientists across the world - from wildly differing backgrounds, with different beliefs and supported by different organizations - all came to the same wrong conclusion.
But the issue of global warming is just so simple that you don't even need to go that far. Public temperature records are available. There's all sorts of other public data available. It's really easy to look at the data yourself if you are so inclined.
Again, that doesn't really count as being contested. By that logic, it's also a subject of controversy as to whether the dog did indeed eat little Johnny's homework. Also, it's contested whether the sky is indeed blue - a colorblind person sees it as gray.
It doesn't count as contest if the people contesting it are ignorant and/or are doing it for political purposes.
The amount of CO2 we release into the atmosphere is easily measurable, and it matches with the observed increase in CO2 in air, water, and biomass. It's about 40 billion tons per year now: http://cdiac.ornl.gov/GCP/carb...
The amount of CO2 naturally emitted by volcanoes and forest fires and such is a bit harder to calculate but you can get reasonable order-of-magnitude estimates. Volcanoes, for instance, emit about 0.3 billion tons per year. There are lots of sources on the US geological survey page: http://volcanoes.usgs.gov/haza...
No matter how you slice it, even the most outlandish estimates for CO2 from natural sources fall 1-2 orders of magnitude short of the amount of CO2 necessary to explain the global increase.
There are natural CO2 absorbing sources but the additional amount they absorb each year is tiny.
FTA: "It is extremely likely that human influence has been the dominant cause of observed warming since 1950, with the level of confidence having increased since the fourth report."
Geoengineering won't solve the problem, it will just replace the problem with another problem. That said, it's possible that the new problem will be easier to deal with than the problem of CO2. In that case I'm all for it, but we need to first figure out more about the effects of climate change before we make any hasty decisions.
Sadly a lot of groups are going to see geoengineering as a way to further their own agendas, so it's possible that that solution too will become corrupted.
Actually the question isn't to what extent humans are responsible. We know that humans are mostly, if not entirely, responsible. This is not controversial. It's also not controversial that over the next century the planet is going to warm by at least a few degrees, regardless of any actions we take (the amount of CO2 in the atmosphere has already 'locked in' a certain amount of warming). In all likelihood, this is going to continue for at least a millenium, again regardless of what we do now. This is unless we figure out a magic way to recapture all the CO2 in the atmosphere faster than nature can absorb it, which is so incredibly hard that you wouldn't be very incorrect in saying that the laws of physics themselves forbid it (removing CO2 from the atmosphere represents a massive decrease in entropy; this needs to be balanced by a proportionately large increase in entropy elsewhere, and this means - ironically - more heat).
The only thing that's controversial right now is what effect global warming will have on ecosystems and human civilization. This is at least a factor we can control, and it's possible it won't be as bad as the alarmists say. Humans are clever and we can overcome the challenges. But we at least need to start doing something and stop pretending that it's a 'controversy'.
The question isn't whether writing verifiable code is hard. It's obviously hard. The question is whether the alternative - the risk of a buggy OS bringing down your system and causing potentially huge losses - is worse. In many cases I'd argue that it is. Maybe not if you're writing an iOS game, but definitely for a lot of other stuff.
It's not necessary to write an OS in C. People have written operating systems in high-level languages, sometimes very high level languages. Hell, someone wrote an OS in haskell. The only time you need to use C to write systems code is when it's some weird hardware with a non-standard C implementation, like for microcontrollers. But even that is slowly disappearing as more and more embedded systems go towards ARM-based (and sometimes x86-based) stuff, which has a very healthy language ecosystem.
A myth about C is that it's a "hardcore" language. C isn't really that complicated a language. There's nothing to be proud of about coding in C. There's a reason C is used in a lot of introductory programming courses. Learn haskell then get back to me. Learning to write good code in C is another matter entirely, but why use a language that makes this hard?
Most important thing, though, is that I've never seen "exceptional results that no other tool can match." The only area where this is even remotely close to being true is speed. I'll concede that C code can be made fast. But speed comes at a huge cost. Sometimes the cost is worth it (like say, a numerical simulation engine). Most of the time it's not. And new technologies are closing the gap every day.
The problem isn't at all coder trust. There are plenty of languages that trust the coder a lot yet are far harder to make mistakes in. C doesn't just let you do bad stuff, it puts the bad stuff at the end of a sparkling, glittery, shiny road lined with flowers and filled with elves playing gentle flutes. Oh and there's also scantily clad mermaids in the distance, seductively beckoning you to join them.
Even if you drill into people's heads that they shouldn't take the easy route, eventually a weary traveler is going to be enticed with the prospect.
Because it's extremely dangerous and a lot of people are still using it. The 'standard' standard library is so full of security holes it's not even funny, and attempts to 'improve' it over the years have mostly been unsuccessful because the bad coding patterns still exist.
C is a great language, it's just that most humans are incapable of using it safely and securely. It's like a.45 with a downward-pointing barrel. It's all too easy to shoot yourself in the foot.
For full disclosure, I used to be an avid C programmer, until I realized the harm I was causing myself and others. It's like when you drunk drive and think you're just fine. It takes an external perspective to realize how reckless your behaviors are.
> Anyways, I don't understand why "dwarf planet" was not made a subclass of "planets" along with "major planets" (where the others go). But no, it is "planets" and "minor planets", which are by definition not a "planet".
The purpose of technical terminology is to be as clear and efficient as possible. Imagine having to say 'major planet' every time you wanted to talk about Earth or Mars or Jupiter. "Earth is the third major planet from the Sun." It's tedious.
But in informal speech you can say that Pluto is a 'dwarf planet, which is a kind of planet'. Nothing wrong with that. I think that's what a lot of people don't understand about technical vs. informal communication.
Pointing out conflicts of interest isn't ad hominem. An ad hominem is when you respond to someone's argument by attacking their character. But in this case Gladwell's authority is itself the argument. With a past like that, you have to take most of what he says with a grain of salt.
There is nothing noble about 'creating opportunities to work'. The true test of being exploitative or not is how workers are treated. It could be that the workers are going to be treated very well here. If that's the case, great. But the fact that sufferers of mental disorder are being targeted makes it an issue you have to be careful about.
Slashdot Mirror
It might be similar but it's not the same mechanism. When you see an object in static, your brain knows that it's just making a guess so the guess is assigned low confidence. But here they showed that you can actually design a picture that looks random but is assigned very high confidence of being an object.
This type of phenomenon is very well known. It's not news, people have known about this sort of stuff in artificial neural nets since the 80's. I guess they just sort of assumed that deep belief nets would get around this problem, but as far as I know there's no reason to believe that. There's a related phenomenon which is assigning very low confidence to a picture that is very clearly a certain class of object - and then if you add a small bit of noise the confidence goes way up. For the interested, this is a good page which explains why some of these issues happen: http://colah.github.io/posts/2...
Just one thing I want to get off my chest: I wish this deep learning fad would die. I first started using deep belief nets around 2006 or so when Hinton published his now-infamous Science paper. I thought it was cool and used it a lot, but I knew it had limitations. Then around 2012 or so this whole thing just started becoming a hugely-hyped meme that everyone wants to get on board, without any knowledge or wisdom - they just want results. This is going to be a recipe for yet another AI "failure", when people realize that they couldn't live up to their own hype.
JPEG2000 had some serious technical issues. The reference library was itself non-compliant (I know, I used it extensively), which meant that there was no fully reference-compliant implementation available anywhere. The library was cumbersome and hard to integrate into existing image processing/viewing programs. And there wasn't much benefit to using JPEG2000; it had only slightly better compression and the wavelet scheme they used caused a lot of unpleasant artifacts (especially around large, uniformly-colored areas with gradual change of brightness).
BPG doesn't seem to have these problems; the images genuinely look better and it's easy to integrate it into the web. I don't really see a major need for a new image compression format but it would be nice to have an option to use it.
What exactly are you trying to say? That there is a scientific controversy around the idea that humans are the major cause of climate change? Well, no, there's not. Now you can dig up random people saying random things and you can change the goalposts, but you can never prove there's a controversy when there isn't. Have a read here: http://www.ncbi.nlm.nih.gov/pm...
FTA: "97–98% of the climate researchers most actively publishing in the field surveyed here support the tenets of ACC outlined by the Intergovernmental Panel on Climate Change, and (ii) the relative climate expertise and scientific prominence of the researchers unconvinced of ACC are substantially below that of the convinced researchers."
97% of scientists rarely agree on anything.
But there really isn't any controversy, no matter how many times you or other people say it or want to believe that it exists.
In this context, controversy would mean a scientific controversy over the main aspects of climate change: that it is happening and it's due to human activities. There is no scientific controversy around this.
Science is full of controversies. You can go to the 'letters' section of any reputable journal and see how much scientists argue over things. But climate change just isn't one of these issues. There might be argument over small details or specifics, but nobody familiar with the science disputes that humans are causing the climate to change.
If you go outside the USA (and a few other anglosphere countries), climate change is pretty much accepted across the political spectrum. But in the USA you have this huge party that's made it it's job to spread FUD, which is exactly what we're seeing here in this thread. It's manufactured controversy. And it works.
No they don't.
Not climate scientists, anyway.
It's neither true that everyone is qualified to pass judgement about a scientific field, nor that everyone should just accept what scientists say blindly. The truth is in the middle: It would be great IF everyone could check for themselves, but in practice this just isn't possible, so at some level you do indeed need to trust the consensus of the literature. Of course you put in measures (like peer review and other measures) to make sure the scientists are being honest. Do transgressions happen? Obviously. We are all human. But it's really a stretch to say that all scientists across the world - from wildly differing backgrounds, with different beliefs and supported by different organizations - all came to the same wrong conclusion.
But the issue of global warming is just so simple that you don't even need to go that far. Public temperature records are available. There's all sorts of other public data available. It's really easy to look at the data yourself if you are so inclined.
I'd never seen that list before. It's pretty illuminating. Thanks!
Again, that doesn't really count as being contested. By that logic, it's also a subject of controversy as to whether the dog did indeed eat little Johnny's homework. Also, it's contested whether the sky is indeed blue - a colorblind person sees it as gray.
It doesn't count as contest if the people contesting it are ignorant and/or are doing it for political purposes.
The amount of CO2 we release into the atmosphere is easily measurable, and it matches with the observed increase in CO2 in air, water, and biomass. It's about 40 billion tons per year now: http://cdiac.ornl.gov/GCP/carb...
The amount of CO2 naturally emitted by volcanoes and forest fires and such is a bit harder to calculate but you can get reasonable order-of-magnitude estimates. Volcanoes, for instance, emit about 0.3 billion tons per year. There are lots of sources on the US geological survey page: http://volcanoes.usgs.gov/haza...
No matter how you slice it, even the most outlandish estimates for CO2 from natural sources fall 1-2 orders of magnitude short of the amount of CO2 necessary to explain the global increase.
There are natural CO2 absorbing sources but the additional amount they absorb each year is tiny.
It's not 'contested a lot'. The only people who 'contest' it are US republicans.
http://en.wikipedia.org/wiki/I...
FTA: "It is extremely likely that human influence has been the dominant cause of observed warming since 1950, with the level of confidence having increased since the fourth report."
Geoengineering won't solve the problem, it will just replace the problem with another problem. That said, it's possible that the new problem will be easier to deal with than the problem of CO2. In that case I'm all for it, but we need to first figure out more about the effects of climate change before we make any hasty decisions.
Sadly a lot of groups are going to see geoengineering as a way to further their own agendas, so it's possible that that solution too will become corrupted.
Actually the question isn't to what extent humans are responsible. We know that humans are mostly, if not entirely, responsible. This is not controversial. It's also not controversial that over the next century the planet is going to warm by at least a few degrees, regardless of any actions we take (the amount of CO2 in the atmosphere has already 'locked in' a certain amount of warming). In all likelihood, this is going to continue for at least a millenium, again regardless of what we do now. This is unless we figure out a magic way to recapture all the CO2 in the atmosphere faster than nature can absorb it, which is so incredibly hard that you wouldn't be very incorrect in saying that the laws of physics themselves forbid it (removing CO2 from the atmosphere represents a massive decrease in entropy; this needs to be balanced by a proportionately large increase in entropy elsewhere, and this means - ironically - more heat).
The only thing that's controversial right now is what effect global warming will have on ecosystems and human civilization. This is at least a factor we can control, and it's possible it won't be as bad as the alarmists say. Humans are clever and we can overcome the challenges. But we at least need to start doing something and stop pretending that it's a 'controversy'.
The question isn't whether writing verifiable code is hard. It's obviously hard. The question is whether the alternative - the risk of a buggy OS bringing down your system and causing potentially huge losses - is worse. In many cases I'd argue that it is. Maybe not if you're writing an iOS game, but definitely for a lot of other stuff.
Well I was attempting to be humorous. It's true - it's NOT a great language.
> And provably correct code is still a pipe dream.
http://sel4.systems/
It's not necessary to write an OS in C. People have written operating systems in high-level languages, sometimes very high level languages. Hell, someone wrote an OS in haskell. The only time you need to use C to write systems code is when it's some weird hardware with a non-standard C implementation, like for microcontrollers. But even that is slowly disappearing as more and more embedded systems go towards ARM-based (and sometimes x86-based) stuff, which has a very healthy language ecosystem.
A myth about C is that it's a "hardcore" language. C isn't really that complicated a language. There's nothing to be proud of about coding in C. There's a reason C is used in a lot of introductory programming courses. Learn haskell then get back to me. Learning to write good code in C is another matter entirely, but why use a language that makes this hard?
Most important thing, though, is that I've never seen "exceptional results that no other tool can match." The only area where this is even remotely close to being true is speed. I'll concede that C code can be made fast. But speed comes at a huge cost. Sometimes the cost is worth it (like say, a numerical simulation engine). Most of the time it's not. And new technologies are closing the gap every day.
The problem isn't at all coder trust. There are plenty of languages that trust the coder a lot yet are far harder to make mistakes in. C doesn't just let you do bad stuff, it puts the bad stuff at the end of a sparkling, glittery, shiny road lined with flowers and filled with elves playing gentle flutes. Oh and there's also scantily clad mermaids in the distance, seductively beckoning you to join them.
Even if you drill into people's heads that they shouldn't take the easy route, eventually a weary traveler is going to be enticed with the prospect.
Because it's extremely dangerous and a lot of people are still using it. The 'standard' standard library is so full of security holes it's not even funny, and attempts to 'improve' it over the years have mostly been unsuccessful because the bad coding patterns still exist.
C is a great language, it's just that most humans are incapable of using it safely and securely. It's like a .45 with a downward-pointing barrel. It's all too easy to shoot yourself in the foot.
For full disclosure, I used to be an avid C programmer, until I realized the harm I was causing myself and others. It's like when you drunk drive and think you're just fine. It takes an external perspective to realize how reckless your behaviors are.
> Anyways, I don't understand why "dwarf planet" was not made a subclass of "planets" along with "major planets" (where the others go). But no, it is "planets" and "minor planets", which are by definition not a "planet".
The purpose of technical terminology is to be as clear and efficient as possible. Imagine having to say 'major planet' every time you wanted to talk about Earth or Mars or Jupiter. "Earth is the third major planet from the Sun." It's tedious.
But in informal speech you can say that Pluto is a 'dwarf planet, which is a kind of planet'. Nothing wrong with that. I think that's what a lot of people don't understand about technical vs. informal communication.
Pointing out conflicts of interest isn't ad hominem. An ad hominem is when you respond to someone's argument by attacking their character. But in this case Gladwell's authority is itself the argument. With a past like that, you have to take most of what he says with a grain of salt.
http://shameproject.com/report...
There is nothing noble about 'creating opportunities to work'. The true test of being exploitative or not is how workers are treated. It could be that the workers are going to be treated very well here. If that's the case, great. But the fact that sufferers of mental disorder are being targeted makes it an issue you have to be careful about.
The charity could have all the best intentions in the world, doesn't mean the IT companies are going to look at it that way.
Exploiting the vulnerable.
But to some degree, I guess that describes all jobs.