I see AC snuck in a comment again. (Also I just learned that Chrome doesn't consider 'snuck' a word... And here I thought it was more common than 'sneaked', which sounds weird to me.)
"false positives" come out by chance, there is nothing that can be done to prevent that. Now, underestimating the chance of false positives comes from a poor understanding of statistical tools, and there's a lot that can be done to prevent that. Sadly many scientists don't seem to be interested. (And no, I don't think Bayesian methods are the answer to that; if anything they give much more room for incorrect conclusions than frequentist methods do.)
Uh huh, and you can completely get rid of that problem by using Bayesian methods and getting nice clean wrong answers, and then stick your fingers in your ears and go "LALALALALA it's true according to my PRIOR BELIEFS, LALALALA"...
I agree with your cautionary point about the need to understand what one is doing when applying any statistical tool, however I find it hard to imagine any problem (at least in the hard sciences) where a Bayesian method is really called for. The notion of "prior beliefs" can be philosophically seductive, I admit, but in practice it is mathematically waaay too dangerous in my opinion. Yes, I know, there is a lot of work on things such as non-informative priors and such, but why go through all that trouble when you already have a whole theory developed under worst-case loss called frequentist statistics?
Yes, if everyone using statistical tools understood the implications of what they are doing fully then perhaps some people would correctly choose to use Bayesian methods. But given the impracticality of that proposition (seriously, how many biologists or physicists have the time to gain a deep understanding of statistics?), I would personally feel much safer if, when it comes to establishing some kind of scientific "truth" (such as the existence of the Higgs boson, or whether a drug interacts with another or not), people just stuck to frequentist methods. Of course there are many ways to screw that up too, but not nearly as many as for Bayesian tools (in my opinion).
We sometimes joked that there were never problems in the Soviet union, but their successful overcoming.
Hah! I'd forgotten this one, thanks for reminding me. And I wasn't disagreeing with you or anything -- what news should be is similar to max self-information in the Shannon sense.
Still I'm surprised, from what you said I'm guessing you actually hail from those parts, which I don't think is very common for/.. I myself am too young to remember anything before the collapse, but I learned everything I ever want to know about what it was like from my parents...
Wow, you people have no idea what Soviet news was like, do you? "And in national news, the infant mortality rate for calves born to the Soviet farmer has deceased another astonishing 0.03% over the past decade, proving once again the infinite wisdom and foresight of the Fathers of our great nation, V.I. Lenin and I.V. Stalin."
Yeah I agree totally, that's why I was careful to say "all else being equal", which I know is a false premise:). These are the kinds of problems that would be nice if the free market could solve, except that of course in many cases the true cost is either impossible to know, becomes apparent years or decades too late, or is paid by the wrong party... Turns out the world is complicated. Who knew?
I don't think these are really what most opposing posts say, but the rebuttals are pretty obvious:
1) If self driving cars are proven to improve safety, all else being equal it's irresponsible to resist using them purely for the sake of personal enjoyment. Every time someone gets behind the wheel they put their own and, more importantly, other people's health and lives in danger. "But I like driving" isn't really a good enough excuse in my book to do that, at least not habitually (i.e. occasionally taking over in a small, empty road wouldn't be the problem, it's rush hour on the highway that kills people).
2) The question here is whether or not a person who is concerned about privacy would be able to find a car that they can be reasonably sure doesn't store/transmit that kind of information. I can imagine two obstacles to this. One could be federal/state regulations saying that manufacturers must include such a feature (+ possibly operate some kind of database for centralized storage of the information to keep "safe" from destruction by the user). Sadly, it is easy to imagine that such a law could come into existence, in part because as FB has shown most people couldn't give two shits about their privacy, which brings me to the second, more realistic obstacle -- there would probably be some (perhaps tiny) benefit to users from tracking, and it wouldn't make economical sense for manufacturers to cater to those who care about privacy as they would probably basically represent a small niche. The point is, this is more of a general social issue, and as long as the vast majority of people remain ignorant of and apathetic towards the increasing intrusion of corporations and government in their personal lives, things are going to keep getting worse for those who don't like to be constantly tracked and kept tabs on... Self-driving cars or not.
After all, if you have a smart phone you take with you (and don't physically remove the battery from) while you drive, you can probably already be tracked by any number of entities, and the cost to do so is probably only going to decrease for them in the future (as carriers and manufacturers give in to advertisers, "law enforcement" and god knows who else more and more).
Well, I guess it's not completely irrational. Automated systems don't deal with highly unlikely exceptions as well as humans sometimes do. Though they also tend to make much fewer mistakes under normal circumstances (or under types of exceptions that were taken into account by the engineers) than humans do... If money wasn't an issue, I'd say that the autopilot + usually passive human operator solution is the best of both worlds. Which is exactly what Google and others have in mind while designing autonomous cars -- it would be silly of them to build/market the technology as completely driverless. And I imagine that the transition will really be very gradual. I don't think the first "self-driving" consumer cars will drive themselves completely from point A to point B. More likely we will first see for example highway-only self-driving cars, which sounds like both a much easier engineering problem as well as having higher potential for saving lives (no falling asleep and drifting, no lane changing into a car in the blind spot, etc). There would have to be a system for smoothly transferring control upon entering an leaving the highway, but again that seems like an easier problem that urban driving with all of its messiness and variety of environment types... As usual, I think the anxiety of people predicting the implications of the technology is highly exaggerated compared to the actual disruption it will cause.
just look at/. discussions about self-driving cars, if you don't believe me
Funny you should bring that up, it's actually been bugging me for a while. Most of the comments on those posts seem to either claim that it's an impossible engineering problem to solve, or that it creates legal/social problems that are impossible to solve (e.g. "just wait until one of these crashes, the resulting lawsuit will bring down all technology-related companies in the world and thrust us back to the bronze age and give everyone AIDS." I'm exaggerating only slightly...). What's up with that? I thought people on/. were supposed to like technology and all that... How long has it been cool for "geeks" to be Luddites?
No problem. One more person will get rich off of scared taxpayers' cash. The best thieves are the ones who know how and when to play nice [e.g. see Wall Street].
Nobody is surprised when a homeless man dies of pneumonia, or when thousands of teenagers commit suicide because of a toxic culture. These things aren't "scary", even though objectively they are much more horrible than the statistically insignificant deaths from terrorist attacks [in the US]. It's all about what is and isn't part of "the plan", to quote The Joker from the greatest part of that movie.
Fuck you, I won't do what you tell me. Fuck you, I won't do what you tell me. Fuck you, I won't do what you tell me...
You see where this is going.
LOL it turns out/. has a "compression filter", where if your comment compresses too much you can't post it, hence I had to decrease the number of repetitions. Maybe if this algorithm was improved to do more semantic compression, comments like "the US is a theocracy", or "1984 wasn't a manual" wouldn't be posted as much... Just a thought.
How do we know that there is an imbalance of matter and antimatter? Perhaps this is only the case locally in the observable universe? Is it at all possible that in the whole of the universe there is in fact no imbalance, and for some reason matter and antimatter formed "pockets" where one dominates the other, and we're just observing one of these pockets?
Heh well I'm Armenian, I've noticed at least one of these... I wasn't disagreeing with you, I think either some weird sense of national pride, unfortunate diplomatic practices, and a bunch of other reasons cause people to ignore any lessons history has to teach us at a great cost to the present and the future...
Really I think a lot of it does, at its core, boil down to pride. Pride in your own little group, in your associations, in your identity... There is a reason why Pride is considered such a deeply vile thing in so many religions. Along with greed, the two emotions are largely responsible for a very large share of the horrors humanity endures.
Now, I often think that Germany has gone too far in the other direction -- I don't like it that expressions of an idea, any idea, can be criminalized anywhere (though I understand why Germans did this). But still, I find that kind of shame much less disturbing than willful ignorance and doublethink as practiced all too often by, well, everyone... How many Americans, for example, do you think would even begin to accept that their government has committed horrible crimes all over the world during the last 50 years? Latin America, SE Asia, etc... And these are widely known facts, anyone who was even a little committed to the truth can read, heck, Wikipedia articles, not to mention the hundreds of books on the subject that have been written. But no, if you do that you "hate America", right? Just like you hate America when you suggest that the US has supported/is supporting dozens of states guilty of violations of international law, not to mention such violations the US has committed itself directly.
This wasn't meant to devolve into a rant about the US actually, though I disagree with a lot of US policies I still think it has been a much more ethical "empire" than any other in the past (seriously, think about it, what state with even one tenth as much power, relative to its time in history, hasn't committed much more horrible crimes? China? USSR? The Persian empire?)...
Yeah I need to step down on the caffeine today too.
Holy shit man, lighten up. It's a Family Guy reference. GP probably wasn't trying to make a point, s/he just thought it would be a relevant joke to GP's post.
Yeah, I'm just repeating (poorly) what a very smart and well spoken physicist who actually understands this stuff said. Not a physicist myself... (And I regret the string of decisions that led to this reality increasingly often.)
Perhaps it's time you took a course in manners -- I am very well aware of the equivalence of mass and energy, I was referring to some as of yet highly theoretical, but aesthetically pleasing mathematical models of the universe that suggest that the universe in fact has zero total energy in the sense that it is the result of a so-called "quantum fluctuation". This is aesthetically pleasing because it means there needn't be any "origin" for the energy in the universe, and in fact it suggests that new universes could be coming into existence all around us all the time that we will simply never be able to measure (i.e. will always be of zero size from our perspective) due to expansion and such.
FYI, I am a statistician, not a physicist, so I have only the vaguest intuition as to what all this actually means in a precise mathematical sense.
Perhaps there is too much politics on/., but this topic is highly relevant to a large portion of the user base here who own/operate web businesses, so I think your rant is misplaced.
Slashdot Mirror
Definitely not as important as making sarcastic comments on slashdot . Whoa, this is some meta shit right here!
I see AC snuck in a comment again. (Also I just learned that Chrome doesn't consider 'snuck' a word... And here I thought it was more common than 'sneaked', which sounds weird to me.)
"false positives" come out by chance, there is nothing that can be done to prevent that. Now, underestimating the chance of false positives comes from a poor understanding of statistical tools, and there's a lot that can be done to prevent that. Sadly many scientists don't seem to be interested. (And no, I don't think Bayesian methods are the answer to that; if anything they give much more room for incorrect conclusions than frequentist methods do.)
Uh huh, and you can completely get rid of that problem by using Bayesian methods and getting nice clean wrong answers, and then stick your fingers in your ears and go "LALALALALA it's true according to my PRIOR BELIEFS, LALALALA"...
I agree with your cautionary point about the need to understand what one is doing when applying any statistical tool, however I find it hard to imagine any problem (at least in the hard sciences) where a Bayesian method is really called for. The notion of "prior beliefs" can be philosophically seductive, I admit, but in practice it is mathematically waaay too dangerous in my opinion. Yes, I know, there is a lot of work on things such as non-informative priors and such, but why go through all that trouble when you already have a whole theory developed under worst-case loss called frequentist statistics?
Yes, if everyone using statistical tools understood the implications of what they are doing fully then perhaps some people would correctly choose to use Bayesian methods. But given the impracticality of that proposition (seriously, how many biologists or physicists have the time to gain a deep understanding of statistics?), I would personally feel much safer if, when it comes to establishing some kind of scientific "truth" (such as the existence of the Higgs boson, or whether a drug interacts with another or not), people just stuck to frequentist methods. Of course there are many ways to screw that up too, but not nearly as many as for Bayesian tools (in my opinion).
We sometimes joked that there were never problems in the Soviet union, but their successful overcoming.
Hah! I'd forgotten this one, thanks for reminding me. And I wasn't disagreeing with you or anything -- what news should be is similar to max self-information in the Shannon sense.
/.. I myself am too young to remember anything before the collapse, but I learned everything I ever want to know about what it was like from my parents...
Still I'm surprised, from what you said I'm guessing you actually hail from those parts, which I don't think is very common for
Wow, you people have no idea what Soviet news was like, do you? "And in national news, the infant mortality rate for calves born to the Soviet farmer has deceased another astonishing 0.03% over the past decade, proving once again the infinite wisdom and foresight of the Fathers of our great nation, V.I. Lenin and I.V. Stalin."
Yeah I agree totally, that's why I was careful to say "all else being equal", which I know is a false premise :). These are the kinds of problems that would be nice if the free market could solve, except that of course in many cases the true cost is either impossible to know, becomes apparent years or decades too late, or is paid by the wrong party... Turns out the world is complicated. Who knew?
I don't think these are really what most opposing posts say, but the rebuttals are pretty obvious:
1) If self driving cars are proven to improve safety, all else being equal it's irresponsible to resist using them purely for the sake of personal enjoyment. Every time someone gets behind the wheel they put their own and, more importantly, other people's health and lives in danger. "But I like driving" isn't really a good enough excuse in my book to do that, at least not habitually (i.e. occasionally taking over in a small, empty road wouldn't be the problem, it's rush hour on the highway that kills people).
2) The question here is whether or not a person who is concerned about privacy would be able to find a car that they can be reasonably sure doesn't store/transmit that kind of information. I can imagine two obstacles to this. One could be federal/state regulations saying that manufacturers must include such a feature (+ possibly operate some kind of database for centralized storage of the information to keep "safe" from destruction by the user). Sadly, it is easy to imagine that such a law could come into existence, in part because as FB has shown most people couldn't give two shits about their privacy, which brings me to the second, more realistic obstacle -- there would probably be some (perhaps tiny) benefit to users from tracking, and it wouldn't make economical sense for manufacturers to cater to those who care about privacy as they would probably basically represent a small niche. The point is, this is more of a general social issue, and as long as the vast majority of people remain ignorant of and apathetic towards the increasing intrusion of corporations and government in their personal lives, things are going to keep getting worse for those who don't like to be constantly tracked and kept tabs on... Self-driving cars or not.
After all, if you have a smart phone you take with you (and don't physically remove the battery from) while you drive, you can probably already be tracked by any number of entities, and the cost to do so is probably only going to decrease for them in the future (as carriers and manufacturers give in to advertisers, "law enforcement" and god knows who else more and more).
Well, I guess it's not completely irrational. Automated systems don't deal with highly unlikely exceptions as well as humans sometimes do. Though they also tend to make much fewer mistakes under normal circumstances (or under types of exceptions that were taken into account by the engineers) than humans do... If money wasn't an issue, I'd say that the autopilot + usually passive human operator solution is the best of both worlds. Which is exactly what Google and others have in mind while designing autonomous cars -- it would be silly of them to build/market the technology as completely driverless. And I imagine that the transition will really be very gradual. I don't think the first "self-driving" consumer cars will drive themselves completely from point A to point B. More likely we will first see for example highway-only self-driving cars, which sounds like both a much easier engineering problem as well as having higher potential for saving lives (no falling asleep and drifting, no lane changing into a car in the blind spot, etc). There would have to be a system for smoothly transferring control upon entering an leaving the highway, but again that seems like an easier problem that urban driving with all of its messiness and variety of environment types... As usual, I think the anxiety of people predicting the implications of the technology is highly exaggerated compared to the actual disruption it will cause.
just look at /. discussions about self-driving cars, if you don't believe me
Funny you should bring that up, it's actually been bugging me for a while. Most of the comments on those posts seem to either claim that it's an impossible engineering problem to solve, or that it creates legal/social problems that are impossible to solve (e.g. "just wait until one of these crashes, the resulting lawsuit will bring down all technology-related companies in the world and thrust us back to the bronze age and give everyone AIDS." I'm exaggerating only slightly...). What's up with that? I thought people on /. were supposed to like technology and all that... How long has it been cool for "geeks" to be Luddites?
No problem. One more person will get rich off of scared taxpayers' cash. The best thieves are the ones who know how and when to play nice [e.g. see Wall Street].
Nobody is surprised when a homeless man dies of pneumonia, or when thousands of teenagers commit suicide because of a toxic culture. These things aren't "scary", even though objectively they are much more horrible than the statistically insignificant deaths from terrorist attacks [in the US]. It's all about what is and isn't part of "the plan", to quote The Joker from the greatest part of that movie.
shit, slashdot is turning into reddit... ABANDON SHIP!
Fuck you, I won't do what you tell me. Fuck you, I won't do what you tell me. Fuck you, I won't do what you tell me...
/. has a "compression filter", where if your comment compresses too much you can't post it, hence I had to decrease the number of repetitions. Maybe if this algorithm was improved to do more semantic compression, comments like "the US is a theocracy", or "1984 wasn't a manual" wouldn't be posted as much... Just a thought.
You see where this is going.
LOL it turns out
How do we know that there is an imbalance of matter and antimatter? Perhaps this is only the case locally in the observable universe? Is it at all possible that in the whole of the universe there is in fact no imbalance, and for some reason matter and antimatter formed "pockets" where one dominates the other, and we're just observing one of these pockets?
Heh well I'm Armenian, I've noticed at least one of these... I wasn't disagreeing with you, I think either some weird sense of national pride, unfortunate diplomatic practices, and a bunch of other reasons cause people to ignore any lessons history has to teach us at a great cost to the present and the future...
Really I think a lot of it does, at its core, boil down to pride. Pride in your own little group, in your associations, in your identity... There is a reason why Pride is considered such a deeply vile thing in so many religions. Along with greed, the two emotions are largely responsible for a very large share of the horrors humanity endures.
Now, I often think that Germany has gone too far in the other direction -- I don't like it that expressions of an idea, any idea, can be criminalized anywhere (though I understand why Germans did this). But still, I find that kind of shame much less disturbing than willful ignorance and doublethink as practiced all too often by, well, everyone... How many Americans, for example, do you think would even begin to accept that their government has committed horrible crimes all over the world during the last 50 years? Latin America, SE Asia, etc... And these are widely known facts, anyone who was even a little committed to the truth can read, heck, Wikipedia articles, not to mention the hundreds of books on the subject that have been written. But no, if you do that you "hate America", right? Just like you hate America when you suggest that the US has supported/is supporting dozens of states guilty of violations of international law, not to mention such violations the US has committed itself directly.
This wasn't meant to devolve into a rant about the US actually, though I disagree with a lot of US policies I still think it has been a much more ethical "empire" than any other in the past (seriously, think about it, what state with even one tenth as much power, relative to its time in history, hasn't committed much more horrible crimes? China? USSR? The Persian empire?)...
Yeah I need to step down on the caffeine today too.
Holy shit man, lighten up. It's a Family Guy reference. GP probably wasn't trying to make a point, s/he just thought it would be a relevant joke to GP's post.
Yeah, I'm just repeating (poorly) what a very smart and well spoken physicist who actually understands this stuff said. Not a physicist myself... (And I regret the string of decisions that led to this reality increasingly often.)
Perhaps it's time you took a course in manners -- I am very well aware of the equivalence of mass and energy, I was referring to some as of yet highly theoretical, but aesthetically pleasing mathematical models of the universe that suggest that the universe in fact has zero total energy in the sense that it is the result of a so-called "quantum fluctuation". This is aesthetically pleasing because it means there needn't be any "origin" for the energy in the universe, and in fact it suggests that new universes could be coming into existence all around us all the time that we will simply never be able to measure (i.e. will always be of zero size from our perspective) due to expansion and such.
FYI, I am a statistician, not a physicist, so I have only the vaguest intuition as to what all this actually means in a precise mathematical sense.
Actually, the total sum of the energy of the universe is probably 0, so the mass didn't "come from" anywhere...
Perhaps there is too much politics on /., but this topic is highly relevant to a large portion of the user base here who own/operate web businesses, so I think your rant is misplaced.
Expansion of the universe, duh!
And this, ladies and gentlemen, is the Holy Mother of sampling biases (and a 5-digit-uid /. reader who fell for it... for shame).
I have no clue what you're trying to say.