Sure, current autopilots are fairly simple because of the design principle that humans must fly planes. However, an autopilot could potentially be designed to handle far more complex situations. Since it is a machine that reacts the same way in any situation every time it is possible to actually test it and find out how it will react (as opposed to human pilots which are a mixed bag).
Somehow I doubt my using the word "magic" turns me into a sorcerer.:) I'm quite aware of how computers work and what their limitations are. However, humans are basically just machines as well - just with very complex programming. The difference is that with computers you can eventually refine them to a point where mistakes happen extremely rarely - with humans you'll never get beyond a certain point.
In any case, I wasn't suggesting that we cut over to full computer control next Tuesday without further improvements in aircraft design. I'm just saying that putting passengers lives in the hands of a computer isn't any more inherently dangerous than putting them in the hands of a pilot.
I suspect that some of the problem is the need to support a mixed environment - what happens if the guy in front of you jams on the brakes or the car in the other lane swerves towards you. If you start with the premise that NOBODY is manually driving a car then you'd have far fewer surprises potentially. Granted, a design that could still handle the unexpected would be a good thing.
For all the money we spend on various government boondoggles I wonder what the ROI would be if we actually decided to implement something like this. Just think of the potential productivity improvements if the US had an automated roadway system. It would revolutionize transportation. You'd need half the cars on the road since the typical family wouldn't need cars parked all over the place since the cars could drive themselves (empty) to wherever they needed to be. Emissions would be cut dramatically since the cars would travel ideal routes at ideal speeds and wouldn't need to stop much (no need for traffic lights - just interleave crossing traffic). No need for huge parking lots taking up lots of space and all the enviornmental problems that causes - just have a parking garage every 2 miles and cars can drop people off at the door and go park. And, how much do all those accidents cost the economy and the health care system?
Many of these benefits wouldn't appear in a mixed enviornment - you can't interleave cars at intersections if you have human drivers. At what point does somebody's desire to "enjoy driving" outweigh the benefits to society if manual driving were restricted to particular venues such as tracks or maybe the odd road through some woods?
Re:The Mysterious Reoccurrence of Mr. Freckles
on
Most Blogs Now Abandoned
·
· Score: 2, Insightful
Twitter is IRC, except there is only one channel, and by default everybody is on #ignore...
Shamelessly stolen from somebody else - but I have no idea who...
From everything I've read the landing in the hudson seems like a fluke. Apparently landing in water is far more dangerous than landing on land. Granted, with lots of buildings all around him I'm not sure the pilot had a lot of options.
I did take some exception to the term "battle-hardened" - a fair percentage of pilots who go through serious air emergencies end up dead, and since so few emergencies happen few pilots are experienced with them. On the other hand, the flight computer has the experience of every simulated and real emergency any plane has ever been through. Sure, humans can practice in the simulator as well, but the reality is that costs mean that no individual gets that much time in the simulator. Due to the magic of software when one flight computer knows how to handle some situation, they all do.
I suspect the Boeing design reflects the American legal system. If the plane goes down and it is the pilot's fault, you can sue the pilot. Maybe you can even sue the airline who trained him. On the other hand, if the plane goes down and the pilot had no control then you can sue the aircraft manufacturer. Never mind that the design saves lives - better to allow thousands to die at somebody else's hands than one to die at your own. Gotta love the tort system.
For the same reason we'll allow tens of thousands to die every year in auto accidents due to driver error but we'd never consider automating driving because maybe somebody might die every year or two due to a computer error.
Yup, in my times visiting Japan on business I've seen people take three hours to do something that anybody fresh out of college should be able to do in 20 minutes. However, EVERYBODY worked long hours.
It is a true testimony to the incompetence of management in American manfacturing companies that they lost out to this kind of a culture.:)
What about ultra-secret cables designed not to be detected by GPR?
No idea if it were possible, but if you'd go to the trouble to run cables and not map them so that you have to run out and fix them all the time, perhaps you'd go to the trouble to make them hard to find (and therefore tap).
Yup. The reward for giving back is that the software does what you want it to do.
Those who do not give back shouldn't be punished, but they also shouldn't be surprised if the volunteers running the project take it someplace they'd rather not see it go.
Free software exists by and for the pleasure of its developers. Sure, most of them are nice people and are willing to do favors for end users, but in the end if you want to see something fixed there is nothing like a ready-to-go patch.
And in my opinion as a professional systems programmer, this is an oversight in the Posix standard and not a desired behavior
Agreed - there really needs to be a POSIX way of forcing an atomic write but not forcing an fsync. They aren't the same thing, and an atomic write doesn't carry nearly the same performance hit as an fsync. The last thing I want is every app on my system making the write cache useless just to prevent wholesale data loss.
Here is the issue - application wants to make an atomic change to a file. The application doesn't care if the file ends up in the starting state, or the final state - only that the change is atomic.
fsync doesn't do that. Fsync guarantees that the file ends up in the final state quickly (but not atomically). Fsync also degrades system performance.
So, the proposed application change doesn't accomplish what the app writers actually want, and it slows down the system. It does reduce the risk of data loss.
What we really need is transaction support for files - just like we have for databases. Now, I agree that this may not be needed for all file operations (though admins should be able to turn it on by default if they want), but this is really the "right way" of handling this sort of situation.
If anything I find myself patching apps to remove fsyncs. MythTV forces frequent fsyncs of the video stream and it can kill performance and even lead to data loss (buffer overruns - the degraded disk performance can't keep up with recorded video demand). There is no reason a recording needs to be fsynced every 30 seconds. If power goes out I'm going to lose 5 minutes of my recorded show anyway while the system comes back up - losing the previous 30 seconds of unflushed video isn't the end of the world. I'd rather have that then have dropped frames and glitches all over the place from lost video packets.
What we need is for apps to tell the OS what they actually need, and for the OS to figure out how to deliver it. App writers shouldn't care what filesystem you're writing to and what the approved way of modifying files on that filesystem is. They certainly shouldn't care about how the write cache works. Sure, there should be an fsync option, but it should be used to sync disk writes to operations that take place in other media or over the network (such as in a transactional database). There should also be other options like atomic file operatiopns (make the following changes to the following files atomically). Let the app figure out what its requirements are, and let the OS figure out how to deliver it.
I'd think a bigger source of error in practice is missed pulses. The system depends on counting pulses, and at any point if you get out of sync you are going to lose track of your position.
In GPS the satellites transmit the time that each packet is transmitted. So, the receiver doesn't need to be on continuously to not lose track of its location. Obviously the pulsars aren't going to encode anything in their signals so you're dependent on keeping track.
Wouldn't a simpler solution be to just put some transmitters in a few locations in the solar system for now, and in a few solar systems once we get to a point where we actually care about navigating between them?
While I don't care to run ubuntu on my phone (non-ideal UI for the task), what I would like to see is a C API for android.
My phone is SLOW!!! Memory is tight, and applications take forever to switch. The browser is fairly glacial (even when rendering pages that are stored locally in flash so it isn't just the mobile network).
I think that half the problem is Davlik. It is a non-JIT JVM-like implementation (though it isn't really Java).
While I like the app management that android provides, the requirement that everything be written in a completely-interpreted language definintely is slowing things down. Sure, it isn't a supercomputer, but that mobile CPU has more power than any desktop had more than 13 years ago or so. I suspect that if it could spend more of its time actually running instructions it would be a whole lot faster, and compiled code would use far less RAM and would have far fewer cache misses.
Sure, you could lose some platform-independence that way. I'm all for making the API smart - abstract the hardware as much as possible. You might still need different binaries for different CPUs, but you could minimize this to a great degree. If Google came out with some kind of davlik compiler (turns apks into mostly-native code) they could process the apps as they are installed (maybe centrally). That might be the best of both worlds.
It just seems like there is a lot of room to grow. Don't get me wrong - I realize they're just getting started and I'm pretty happy with my phone. It just isn't a finished product yet...
Actually, med schools are a perfect illustration of the flawed reasoning in the GP post. Clearly society needs more doctors, and yet we have a chokehold on their supply.
By all means have standards for doctors. Screen everybody before allowing them to practice medicine. However, anybody who wants to take courses in medicine and attempt to become a doctor should be free to do so as long as they pay their own way.
We really just need a LOT more schools. That would also drop the cost of education. Anybody willing to pay the bills should be able to attend college. Now, I'm fine with colleges having tracked programs and to get into the higher tracks there would be admission standards. I'm also fine with colleges booting out students who are disruptive in class. That said, if somebody who can barely read and write wants to attend a college, pay their own way, and sit quietly at the back of the class, why should we stop them?
Further, there's no reason why anybody who wants to be a rock star and is willing to pay for it shouldn't be able to attain that goal.
But, they can. They can't guarantee that they will be popular, but anybody can buy a guitar and take lessons. And, that is all we're talking about here.
My initial argument was about making an education available to anybody who is willing to pay for it. I don't aruge that people should be guaranteed success or jobs. However, what is wrong with somebody who wants to take 4 years worth of basket weaving classes being able to pay to take them? What is wrong with somebody wanting to pursue a medical degree "for fun" with no intent to enter the workforce at all?
There is no reason to artifically restrict the supply of education.
I have to disagree that this is "unfortunate." Normally I'm pretty libertarian about protecting people from themselves, but I think that labor laws have to be written in this way. Otherwise employers give token attention to the laws but encourage unsafe practices where it will save time/money.
Here are a couple of scenarios - almost all of them probably have happened in the real world (a few are real-world examples) and illustrate the need to ALWAYS hold employers responsible for workplace injuries:
1. HP has a packaging line (putting stuff in boxes) employing minimum-wage workers. When manipulating boxes at such a pace there are serious risks of hand injuries (paper cuts, cuts from knives/etc) - the job requires proper hand protection (such as heavy gloves). However, these gloves do reduce productivity. HP had rules in place requiring these gloves, and they even provided them free. However, they did not fire employees who refused to wear them, but they did fire the slowest few workers on a regular basis. This led to a race for the bottom where everbody who wanted to keep their jobs didn't wear the gloves. When injuries occurred HP pointed to the gloves that were provided and said that it wasn't their fault.
2. I could envision a scenario where employees are provided materials to perform work and then rated on their consumption of these materials. Safety-related items could be charged like any others. Those who reduce consumption of safety-related materials are consequently rewarded.
It isn't enough for employers to set rules and provide safety equipment. They must also ensure that the rules are being followed, and fire people for not being safe just as readily as they would if they were slow or late for work. This really isn't to protect people from themselves. The purpose of this is to protect employees from employers who post one set of rules for lawyers to read, but enforce a different set of rules in practice.
The solution is simple - employer is responsible for any injuries, period. Employers have no trouble with putting in tons of supervision to ensure nobody is goofing off on the job. It isn't any harder to make sure people are following safety rules.
If anything OSHA is underfunded. It also doesn't help that quite often they don't focus on stuff that truly promotes workplace safety but instead on legalistic enforcement of building codes. I realize that most of these rules have a purpose, but when you can't shut down a business that seriously mistreats their employees because you're too busy suing some mom-and-pop shop for having a handrail that is 1" too high, that is a case of misplaced priorities.
I majored in chemistry, and I never had opportunity to work with t-butyl lithium. A typical undergraduate would have no experience with it. However, I wouldn't consider it something that no undergrad should ever touch if they were properly supervised - an advanced undergrad could safely work with it with proper training.
I would think that most chemistry majors would be aware of the fact that it is pyrophoric. I certainly kenw this. I didn't realize just how impressively dangerous it could be, but I would have known not to treat it lightly.
In any case, grad students should be at least generally supervised in their work and not just sent off to figure it all out on their own. I think that a big problem in grad school is that professors spend almost no time with their students. Now, grad school should teach students autonomy, but not to this extreme.
I think that there are several reasons you don't see this kind of stuff in industry:
1. Liability. As a result of liability everybody has their job and that is all that you do. If an electrical device fails in the lab, the chemist asks for it to be fixed and somebody qualified to perform the repair does so. Nobody just opens up an instrument and starts soldering wires. Liability also means that if an employer fails to provide proper equipment they face serious consequences.
2. Funding. Since there is a profit motive for using the instrument there is money available to get it fixed properly, and right away. Employees don't have to do without, and so they are not tempted to do their own repairs.
3. Enforcement. OSHA requires that employers strictly enforce safety rules. If an inspector comes in and finds 10 employees doing something unsafe, the employer can't just point to the rulebook and training and say that they are "rogue employees." Employers are actually expected to discipline and even terminate employees who do not follow safety rules. The reason is simple - otherwise rules just become legal cover and employers will say that rules should be followed, but fire the slowest people in the operation until everybody figures out that they are expected to cut corners.
4. Inspections. Most industrial labs have routine safety inspections and clear chains of responsibility. Sure, academic labs have occasional inspections, but very little accountability. Where I work every lab has a designated safety officer, who is accountable for any safety violations in their lab. The formalized inspections are essentially designed to make sure they are doing their job. Lab safety officers are expected to police their labs and keep them in order. Safety violations are reported to senior management and there are serious consequences for a lax attitude towards safety. Failure to comply with instructions of a Lab Safety Officer would result in fairly swift discipline, and failure of a safety officer to catch safety problems would subject them to discipline (or at least replacement in the role).
Academic labs should be inspected, and when violations are found the university should be fined - plain and simple. If the university claims that students aren't obeying the rules, then the university should still be fined and advised to start enforcing the rules if they don't want to be fined again. Safety is serious business.
True. But, I think the fundamental issue is supply/demand. There are just more people able to do a job than there are (desirable) jobs that people are willing to pay good money for.
So, if you're applying for a coveted position, and you are one of 50 resumes, and all of them are qualified, but you are the only one that has a record 20 years in the past, what do you think will happen?
Sure, if there were one candidate that stood head and shoulders above the rest stuff like this would get overlooked. But usually that isn't the case.
Same thing applies to standardized testing. Sure, everybody knows that it is imperfect. However, when you have 1000 applicants for a med school opening you need to have some way to throw out applications. Standardized testing is as fair as just about any of them.
The real solution is to find better ways to employ people so that there are more openings.
For school applicants this is easy - build/accredit more universities. There is no reason that everybody who wants a college/medical/legal/whatever education and is willing to pay for it shouldn't be able to obtain one. Sure, if you want a scholarship there can be some evaluation of merit, or if you want to apply to some elite program. However, the problem with schools is one of poor supply management.
For jobs things are harder - you can't just pay people to do things that nobody really wants. Arguably the government can create demand for jobs that truly do benefit the public (infrastructure, blue sky R&D, etc), but there are limits. I think that at some point the whole system of wealth acquisition by working or owning capital will probably need to change (and I'd hardly consider myself a socialist). I'm not sure we're there yet, but at some point automation will get to the point where nobody actually needs to work - and then how do you decide who gets to eat and how large a population is too much? Some resources will still be scarce, but labor will not be - so how do you decide how to allocate the resources that are avaiable. It could be utopia or a nightmare depending on how the world ends up being governed...
If they fire you for obeying a police request, you will have a line of ambulance chasers waiting to "help" you.
Lot of good that will do, for a number of reasons. Sure, you hear about the rare instance where somebody "hits the lottery" suing their employer, but usually it works out badly for you. Here's why:
1. The would never sue the guy for obeying the police request. Instead they would just not say anything at all and a month later they would nitpick his performance until they find a pretense for terminating him. In most states employers don't need to give you a reason to fire you, and they don't, which makes it hard to debate the merits of the action.
2. If you do sue them, what are the damages? The guy probably makes $20k/yr. Courts don't consider the time of just about anybody valuable (except for lawyers of course).
3. Ok, so you get a $50k settlement. You also get branded as a "rat" and nobody ever hires you again for the rest of your life. Good luck retiring on that $50k. Your lawyer gets $15k of that for the modest amount of effort put in, and they get to continue their job.
Sure, it is absolutely unjust. However, unless somebody fixes the system that guy on the phone did what was in his and his family's best interest - he passed the call to his manager and hoped for the best.
Somebody suggested in another comment that the lowly employee should have just paid in $20 out of his pocket to expedite things. Sure, that sounds nice until you realize these guys probably get these kinds of requests all the time, and the police sure aren't going to follow-up on paying the guy back, so he's basically out $20 every time somebody has a hassle. Verizon should certainly be punished for allowing stuff like this to get in the way of life-saving actions, but the poor guy on the phone is as much a victim as anybody else.
I'd think that a lot of this depends on how overt you are about it.
If you had a laundry list of code blocks for stuff that could be disabled and one of them was labeled "disable-save-support" I suspect you'd be OK if the default was that the compliance code was in there by default. If the disable-save-support were the only flag in the file you might have a tougher time of it.
Or, if you want an even grayer example, if you didn't put any IFDEF blocks/etc around the compliance code, but you just put all of it together in a single file where it would be easier to patch you'd probably have almost no problems at all. The code does what it is supposed to and there are no overt instructions on how to disable it. To me that is like a VCR with copy-protection technology where the copy-protect module is easily found inside the case and a simple wiring change would disable it (as opposed to a module embedded in some other vital chip and covered with epoxy). The fact that companies get away with DVD players that simply require some kind of code to bypass region protection indicates that there is a fair amount of leeway. Then again, that is in European countries that might be more sympathetic to region protection if it didn't mean that they had to pay more for a DVD and get it two years after the US market.
I don't know that anybody actually posted a link to the actual study, but were subjects randomly-assigned to the diet vs regular groups, or was there an element of self-selection?
While there is clearly an interesting interaction between diet soda and obesity, the data you present do not in any way show what that relationship is. It could be that people who are genetically prone to becoming obsese are also prone to liking diet soda. It could be that diet soda causes weight gain. It could be that getting lots of calories from sugar somehow suppresses hunger and you eat less of other stuff. It could be that people who drink diet soda think they can get away with eating other things.
In order to be valid you must control for the placebo effect. What happens if you give both groups the same thing but tell half of them that it is diet soda and the other half that it is regular? Better still - don't let anybody know what they're getting (which can be hard to do with food), and randomly assign individuals to the groups.
It isn't a valid experiment unless there is a true control. It is surprisingly easy to mess things up that ruin a controlled experiment. When your test subjects are humans it is almost impossible to get it right. If it were, chances are we'd have cured cancer by now...
Except that it doesn't work that way. The vast majority of criminal cases are settled in a plea-bargain. It works like this:
There is a 20% chance that you killed your wife. The prosecutor charges you with first-degree murder (life sentence, plus risk of execution). Then the presecutor offers you a plea for manslaughter that carries a 2-year prison term. Your choices are to go to trial and risk a life sentence, or just take the 2 years.
Now, sure, if you go to trial the jury SHOULD find you innoent and you get out. Of course, you might sit in jail for three months before your trial even comes up if you don't get bail. However, if you do get convicted you are REALLY up the creek.
Most people just take the deal, and end up being punished for a crime they didn't commit.
Also - merely having to go through the trial is essentially punishment. You could end up in prison for months if you don't make bail. You end up missing work even if you do make bail. While you are entitled to free representation, if you were facing life in prison chances are you'd spend money on the best lawyer you could find.
All of this makes the current system unjust. It is about following a process, not doing what is right.
Don't get me wrong - I'm as tough on crime as anybody is. However, to punish the innocent is not just, and I think that as a society we bear the consequences of doing so. In fact, when a criminal does something wrong they're the only one with blood on their hands, but when as a society we punish the innocent we all end up with blood on our hands.
By the same token, if a prosecutor loses many cases, you have to wonder about his/her skill.
Frankly - I think this is a major problem in criminal justice. Prosecutors are interested in convictions - not prosecuting those who have actually committed crimes. A prosecutor with a weak case doesn't wonder if the guy might be innocent - he just hedges his bets by levelling a nasty charge and then bargaining for a minor one. Sounds nice on paper - a person with a 20% chance of being guilty gets 20% of the punishment, but justice would be 20% of those people getting 100% of the punishment and 80% getting released as expediently as possible.
Btw, NASCAR does this already on some tracks for safety reasons. You don't see any of those cars going 200+ mph. Even though they are completely capable of it.
Pray tell, where can a research biochemist go to do research where such questionable issues aren't a concern? The pharma industry employs literally tens of thousands of scientists and pretty-much every pharma company has these kinds of practices (I'd say Merck is maybe even slightly better than average in some regards - they're just the company most subject to discovery right now due to Vioxx). Just about every pharma company is laying off scientists, so even if you wanted to switch companies you couldn't (and again it wouldn't be any nicer from an ethical perspective).
Unless you're lobbying for the government to create 50k jobs at the NIH that pay pharma industry wages I'm not qutie sure what you intend for these folks to do. The government can't figure out to do with works in the auto makers - and their skills are a whole lot more generic than some guy doing studies on calcium channels or whatever.
Makes sense - nations have a responsibility to police their own borders. If some Chinese hacker crashes Wall Street the US isn't going to be happy if the Chinese government simply states that it isn't their problem since they didn't sanction the attack.
If some guy set up a big gun in New York state and took to shelling random towns in Canada the Canadians are going to expect the US Government to do something about it. Such a person would be HEAVILY prosecuted and very swift actions would be taken to stop the attack. In the very bizarre case that the US governemnt didn't act swiftly to stop the shelling the Canadians would certainly be within their rights to launch a military strike against the attacker. Certainly such an incident would have huge negative effects on US-Candian relations, and of course it would never be allowed to happen in real life.
Cyber attacks should be no different. Sure, proactively preventing any hacker attack from any rogue actor is going to be just about impossible (though if anybody can do it the Chinese can with their firewalls). However, if the Pentagon calls up the Chinese ambassador and points out that such-and-such an IP is being used to hack into some military installation they would expect the Chinese to halt the attacks and to see somebody punished for it.
Likewise NO government limits retaliation to a tit-for-tat response. If you steal a car on the streets of France and are caught, the police do not merely take it back, and maybe just take your bicycle as well. In order to deter crime the fear of the penalties for being caught when weighed with the liklihood of getting caught needs to be large. So, if you rob a car you get captured in a roadblock, tossed around a bit, and end up in prison for five years. If some Iran-sponsored terrorist cell blows up a UK airplane the UK government doesn't blow up an Iranian plane in retaliation or ask for a million dollars in restitution - they take measures guaranteed to at least cost the Iranians billions, if not outright invading.
A Cyberattack that causes real damages should be responded to in ways that cause real harm to the real people who allowed it to happen. Of course, if a foreign government is generally cooperative in prosecuting the crime then diplomacy should be employed - nobody is going to bomb Germany over a German hacker breaking into some bank since everybody expects the German government to be genuinely horrified by the incident. However, you can't turn a genuinely blind eye towards real harm being done to real people and claim innocence when confronted with the results.
Slashdot Mirror
Sure, current autopilots are fairly simple because of the design principle that humans must fly planes. However, an autopilot could potentially be designed to handle far more complex situations. Since it is a machine that reacts the same way in any situation every time it is possible to actually test it and find out how it will react (as opposed to human pilots which are a mixed bag).
Somehow I doubt my using the word "magic" turns me into a sorcerer. :) I'm quite aware of how computers work and what their limitations are. However, humans are basically just machines as well - just with very complex programming. The difference is that with computers you can eventually refine them to a point where mistakes happen extremely rarely - with humans you'll never get beyond a certain point.
In any case, I wasn't suggesting that we cut over to full computer control next Tuesday without further improvements in aircraft design. I'm just saying that putting passengers lives in the hands of a computer isn't any more inherently dangerous than putting them in the hands of a pilot.
I suspect that some of the problem is the need to support a mixed environment - what happens if the guy in front of you jams on the brakes or the car in the other lane swerves towards you. If you start with the premise that NOBODY is manually driving a car then you'd have far fewer surprises potentially. Granted, a design that could still handle the unexpected would be a good thing.
For all the money we spend on various government boondoggles I wonder what the ROI would be if we actually decided to implement something like this. Just think of the potential productivity improvements if the US had an automated roadway system. It would revolutionize transportation. You'd need half the cars on the road since the typical family wouldn't need cars parked all over the place since the cars could drive themselves (empty) to wherever they needed to be. Emissions would be cut dramatically since the cars would travel ideal routes at ideal speeds and wouldn't need to stop much (no need for traffic lights - just interleave crossing traffic). No need for huge parking lots taking up lots of space and all the enviornmental problems that causes - just have a parking garage every 2 miles and cars can drop people off at the door and go park. And, how much do all those accidents cost the economy and the health care system?
Many of these benefits wouldn't appear in a mixed enviornment - you can't interleave cars at intersections if you have human drivers. At what point does somebody's desire to "enjoy driving" outweigh the benefits to society if manual driving were restricted to particular venues such as tracks or maybe the odd road through some woods?
Twitter is IRC, except there is only one channel, and by default everybody is on #ignore...
Shamelessly stolen from somebody else - but I have no idea who...
From everything I've read the landing in the hudson seems like a fluke. Apparently landing in water is far more dangerous than landing on land. Granted, with lots of buildings all around him I'm not sure the pilot had a lot of options.
I did take some exception to the term "battle-hardened" - a fair percentage of pilots who go through serious air emergencies end up dead, and since so few emergencies happen few pilots are experienced with them. On the other hand, the flight computer has the experience of every simulated and real emergency any plane has ever been through. Sure, humans can practice in the simulator as well, but the reality is that costs mean that no individual gets that much time in the simulator. Due to the magic of software when one flight computer knows how to handle some situation, they all do.
I suspect the Boeing design reflects the American legal system. If the plane goes down and it is the pilot's fault, you can sue the pilot. Maybe you can even sue the airline who trained him. On the other hand, if the plane goes down and the pilot had no control then you can sue the aircraft manufacturer. Never mind that the design saves lives - better to allow thousands to die at somebody else's hands than one to die at your own. Gotta love the tort system.
For the same reason we'll allow tens of thousands to die every year in auto accidents due to driver error but we'd never consider automating driving because maybe somebody might die every year or two due to a computer error.
Yup, in my times visiting Japan on business I've seen people take three hours to do something that anybody fresh out of college should be able to do in 20 minutes. However, EVERYBODY worked long hours.
It is a true testimony to the incompetence of management in American manfacturing companies that they lost out to this kind of a culture. :)
What about ultra-secret cables designed not to be detected by GPR?
No idea if it were possible, but if you'd go to the trouble to run cables and not map them so that you have to run out and fix them all the time, perhaps you'd go to the trouble to make them hard to find (and therefore tap).
Yup. The reward for giving back is that the software does what you want it to do.
Those who do not give back shouldn't be punished, but they also shouldn't be surprised if the volunteers running the project take it someplace they'd rather not see it go.
Free software exists by and for the pleasure of its developers. Sure, most of them are nice people and are willing to do favors for end users, but in the end if you want to see something fixed there is nothing like a ready-to-go patch.
And in my opinion as a professional systems programmer, this is an oversight in the Posix standard and not a desired behavior
Agreed - there really needs to be a POSIX way of forcing an atomic write but not forcing an fsync. They aren't the same thing, and an atomic write doesn't carry nearly the same performance hit as an fsync. The last thing I want is every app on my system making the write cache useless just to prevent wholesale data loss.
Define bug.
Here is the issue - application wants to make an atomic change to a file. The application doesn't care if the file ends up in the starting state, or the final state - only that the change is atomic.
fsync doesn't do that. Fsync guarantees that the file ends up in the final state quickly (but not atomically). Fsync also degrades system performance.
So, the proposed application change doesn't accomplish what the app writers actually want, and it slows down the system. It does reduce the risk of data loss.
What we really need is transaction support for files - just like we have for databases. Now, I agree that this may not be needed for all file operations (though admins should be able to turn it on by default if they want), but this is really the "right way" of handling this sort of situation.
If anything I find myself patching apps to remove fsyncs. MythTV forces frequent fsyncs of the video stream and it can kill performance and even lead to data loss (buffer overruns - the degraded disk performance can't keep up with recorded video demand). There is no reason a recording needs to be fsynced every 30 seconds. If power goes out I'm going to lose 5 minutes of my recorded show anyway while the system comes back up - losing the previous 30 seconds of unflushed video isn't the end of the world. I'd rather have that then have dropped frames and glitches all over the place from lost video packets.
What we need is for apps to tell the OS what they actually need, and for the OS to figure out how to deliver it. App writers shouldn't care what filesystem you're writing to and what the approved way of modifying files on that filesystem is. They certainly shouldn't care about how the write cache works. Sure, there should be an fsync option, but it should be used to sync disk writes to operations that take place in other media or over the network (such as in a transactional database). There should also be other options like atomic file operatiopns (make the following changes to the following files atomically). Let the app figure out what its requirements are, and let the OS figure out how to deliver it.
I'd think a bigger source of error in practice is missed pulses. The system depends on counting pulses, and at any point if you get out of sync you are going to lose track of your position.
In GPS the satellites transmit the time that each packet is transmitted. So, the receiver doesn't need to be on continuously to not lose track of its location. Obviously the pulsars aren't going to encode anything in their signals so you're dependent on keeping track.
Wouldn't a simpler solution be to just put some transmitters in a few locations in the solar system for now, and in a few solar systems once we get to a point where we actually care about navigating between them?
While I don't care to run ubuntu on my phone (non-ideal UI for the task), what I would like to see is a C API for android.
My phone is SLOW!!! Memory is tight, and applications take forever to switch. The browser is fairly glacial (even when rendering pages that are stored locally in flash so it isn't just the mobile network).
I think that half the problem is Davlik. It is a non-JIT JVM-like implementation (though it isn't really Java).
While I like the app management that android provides, the requirement that everything be written in a completely-interpreted language definintely is slowing things down. Sure, it isn't a supercomputer, but that mobile CPU has more power than any desktop had more than 13 years ago or so. I suspect that if it could spend more of its time actually running instructions it would be a whole lot faster, and compiled code would use far less RAM and would have far fewer cache misses.
Sure, you could lose some platform-independence that way. I'm all for making the API smart - abstract the hardware as much as possible. You might still need different binaries for different CPUs, but you could minimize this to a great degree. If Google came out with some kind of davlik compiler (turns apks into mostly-native code) they could process the apps as they are installed (maybe centrally). That might be the best of both worlds.
It just seems like there is a lot of room to grow. Don't get me wrong - I realize they're just getting started and I'm pretty happy with my phone. It just isn't a finished product yet...
Actually, med schools are a perfect illustration of the flawed reasoning in the GP post. Clearly society needs more doctors, and yet we have a chokehold on their supply.
By all means have standards for doctors. Screen everybody before allowing them to practice medicine. However, anybody who wants to take courses in medicine and attempt to become a doctor should be free to do so as long as they pay their own way.
We really just need a LOT more schools. That would also drop the cost of education. Anybody willing to pay the bills should be able to attend college. Now, I'm fine with colleges having tracked programs and to get into the higher tracks there would be admission standards. I'm also fine with colleges booting out students who are disruptive in class. That said, if somebody who can barely read and write wants to attend a college, pay their own way, and sit quietly at the back of the class, why should we stop them?
Further, there's no reason why anybody who wants to be a rock star and is willing to pay for it shouldn't be able to attain that goal.
But, they can. They can't guarantee that they will be popular, but anybody can buy a guitar and take lessons. And, that is all we're talking about here.
My initial argument was about making an education available to anybody who is willing to pay for it. I don't aruge that people should be guaranteed success or jobs. However, what is wrong with somebody who wants to take 4 years worth of basket weaving classes being able to pay to take them? What is wrong with somebody wanting to pursue a medical degree "for fun" with no intent to enter the workforce at all?
There is no reason to artifically restrict the supply of education.
Unfortunately, employment law disagrees with you.
I have to disagree that this is "unfortunate." Normally I'm pretty libertarian about protecting people from themselves, but I think that labor laws have to be written in this way. Otherwise employers give token attention to the laws but encourage unsafe practices where it will save time/money.
Here are a couple of scenarios - almost all of them probably have happened in the real world (a few are real-world examples) and illustrate the need to ALWAYS hold employers responsible for workplace injuries:
1. HP has a packaging line (putting stuff in boxes) employing minimum-wage workers. When manipulating boxes at such a pace there are serious risks of hand injuries (paper cuts, cuts from knives/etc) - the job requires proper hand protection (such as heavy gloves). However, these gloves do reduce productivity. HP had rules in place requiring these gloves, and they even provided them free. However, they did not fire employees who refused to wear them, but they did fire the slowest few workers on a regular basis. This led to a race for the bottom where everbody who wanted to keep their jobs didn't wear the gloves. When injuries occurred HP pointed to the gloves that were provided and said that it wasn't their fault.
2. I could envision a scenario where employees are provided materials to perform work and then rated on their consumption of these materials. Safety-related items could be charged like any others. Those who reduce consumption of safety-related materials are consequently rewarded.
It isn't enough for employers to set rules and provide safety equipment. They must also ensure that the rules are being followed, and fire people for not being safe just as readily as they would if they were slow or late for work. This really isn't to protect people from themselves. The purpose of this is to protect employees from employers who post one set of rules for lawyers to read, but enforce a different set of rules in practice.
The solution is simple - employer is responsible for any injuries, period. Employers have no trouble with putting in tons of supervision to ensure nobody is goofing off on the job. It isn't any harder to make sure people are following safety rules.
If anything OSHA is underfunded. It also doesn't help that quite often they don't focus on stuff that truly promotes workplace safety but instead on legalistic enforcement of building codes. I realize that most of these rules have a purpose, but when you can't shut down a business that seriously mistreats their employees because you're too busy suing some mom-and-pop shop for having a handrail that is 1" too high, that is a case of misplaced priorities.
I majored in chemistry, and I never had opportunity to work with t-butyl lithium. A typical undergraduate would have no experience with it. However, I wouldn't consider it something that no undergrad should ever touch if they were properly supervised - an advanced undergrad could safely work with it with proper training.
I would think that most chemistry majors would be aware of the fact that it is pyrophoric. I certainly kenw this. I didn't realize just how impressively dangerous it could be, but I would have known not to treat it lightly.
In any case, grad students should be at least generally supervised in their work and not just sent off to figure it all out on their own. I think that a big problem in grad school is that professors spend almost no time with their students. Now, grad school should teach students autonomy, but not to this extreme.
I think that there are several reasons you don't see this kind of stuff in industry:
1. Liability. As a result of liability everybody has their job and that is all that you do. If an electrical device fails in the lab, the chemist asks for it to be fixed and somebody qualified to perform the repair does so. Nobody just opens up an instrument and starts soldering wires. Liability also means that if an employer fails to provide proper equipment they face serious consequences.
2. Funding. Since there is a profit motive for using the instrument there is money available to get it fixed properly, and right away. Employees don't have to do without, and so they are not tempted to do their own repairs.
3. Enforcement. OSHA requires that employers strictly enforce safety rules. If an inspector comes in and finds 10 employees doing something unsafe, the employer can't just point to the rulebook and training and say that they are "rogue employees." Employers are actually expected to discipline and even terminate employees who do not follow safety rules. The reason is simple - otherwise rules just become legal cover and employers will say that rules should be followed, but fire the slowest people in the operation until everybody figures out that they are expected to cut corners.
4. Inspections. Most industrial labs have routine safety inspections and clear chains of responsibility. Sure, academic labs have occasional inspections, but very little accountability. Where I work every lab has a designated safety officer, who is accountable for any safety violations in their lab. The formalized inspections are essentially designed to make sure they are doing their job. Lab safety officers are expected to police their labs and keep them in order. Safety violations are reported to senior management and there are serious consequences for a lax attitude towards safety. Failure to comply with instructions of a Lab Safety Officer would result in fairly swift discipline, and failure of a safety officer to catch safety problems would subject them to discipline (or at least replacement in the role).
Academic labs should be inspected, and when violations are found the university should be fined - plain and simple. If the university claims that students aren't obeying the rules, then the university should still be fined and advised to start enforcing the rules if they don't want to be fined again. Safety is serious business.
True. But, I think the fundamental issue is supply/demand. There are just more people able to do a job than there are (desirable) jobs that people are willing to pay good money for.
So, if you're applying for a coveted position, and you are one of 50 resumes, and all of them are qualified, but you are the only one that has a record 20 years in the past, what do you think will happen?
Sure, if there were one candidate that stood head and shoulders above the rest stuff like this would get overlooked. But usually that isn't the case.
Same thing applies to standardized testing. Sure, everybody knows that it is imperfect. However, when you have 1000 applicants for a med school opening you need to have some way to throw out applications. Standardized testing is as fair as just about any of them.
The real solution is to find better ways to employ people so that there are more openings.
For school applicants this is easy - build/accredit more universities. There is no reason that everybody who wants a college/medical/legal/whatever education and is willing to pay for it shouldn't be able to obtain one. Sure, if you want a scholarship there can be some evaluation of merit, or if you want to apply to some elite program. However, the problem with schools is one of poor supply management.
For jobs things are harder - you can't just pay people to do things that nobody really wants. Arguably the government can create demand for jobs that truly do benefit the public (infrastructure, blue sky R&D, etc), but there are limits. I think that at some point the whole system of wealth acquisition by working or owning capital will probably need to change (and I'd hardly consider myself a socialist). I'm not sure we're there yet, but at some point automation will get to the point where nobody actually needs to work - and then how do you decide who gets to eat and how large a population is too much? Some resources will still be scarce, but labor will not be - so how do you decide how to allocate the resources that are avaiable. It could be utopia or a nightmare depending on how the world ends up being governed...
If they fire you for obeying a police request, you will have a line of ambulance chasers waiting to "help" you.
Lot of good that will do, for a number of reasons. Sure, you hear about the rare instance where somebody "hits the lottery" suing their employer, but usually it works out badly for you. Here's why:
1. The would never sue the guy for obeying the police request. Instead they would just not say anything at all and a month later they would nitpick his performance until they find a pretense for terminating him. In most states employers don't need to give you a reason to fire you, and they don't, which makes it hard to debate the merits of the action.
2. If you do sue them, what are the damages? The guy probably makes $20k/yr. Courts don't consider the time of just about anybody valuable (except for lawyers of course).
3. Ok, so you get a $50k settlement. You also get branded as a "rat" and nobody ever hires you again for the rest of your life. Good luck retiring on that $50k. Your lawyer gets $15k of that for the modest amount of effort put in, and they get to continue their job.
Sure, it is absolutely unjust. However, unless somebody fixes the system that guy on the phone did what was in his and his family's best interest - he passed the call to his manager and hoped for the best.
Somebody suggested in another comment that the lowly employee should have just paid in $20 out of his pocket to expedite things. Sure, that sounds nice until you realize these guys probably get these kinds of requests all the time, and the police sure aren't going to follow-up on paying the guy back, so he's basically out $20 every time somebody has a hassle. Verizon should certainly be punished for allowing stuff like this to get in the way of life-saving actions, but the poor guy on the phone is as much a victim as anybody else.
I'd think that a lot of this depends on how overt you are about it.
If you had a laundry list of code blocks for stuff that could be disabled and one of them was labeled "disable-save-support" I suspect you'd be OK if the default was that the compliance code was in there by default. If the disable-save-support were the only flag in the file you might have a tougher time of it.
Or, if you want an even grayer example, if you didn't put any IFDEF blocks/etc around the compliance code, but you just put all of it together in a single file where it would be easier to patch you'd probably have almost no problems at all. The code does what it is supposed to and there are no overt instructions on how to disable it. To me that is like a VCR with copy-protection technology where the copy-protect module is easily found inside the case and a simple wiring change would disable it (as opposed to a module embedded in some other vital chip and covered with epoxy). The fact that companies get away with DVD players that simply require some kind of code to bypass region protection indicates that there is a fair amount of leeway. Then again, that is in European countries that might be more sympathetic to region protection if it didn't mean that they had to pay more for a DVD and get it two years after the US market.
I don't know that anybody actually posted a link to the actual study, but were subjects randomly-assigned to the diet vs regular groups, or was there an element of self-selection?
While there is clearly an interesting interaction between diet soda and obesity, the data you present do not in any way show what that relationship is. It could be that people who are genetically prone to becoming obsese are also prone to liking diet soda. It could be that diet soda causes weight gain. It could be that getting lots of calories from sugar somehow suppresses hunger and you eat less of other stuff. It could be that people who drink diet soda think they can get away with eating other things.
In order to be valid you must control for the placebo effect. What happens if you give both groups the same thing but tell half of them that it is diet soda and the other half that it is regular? Better still - don't let anybody know what they're getting (which can be hard to do with food), and randomly assign individuals to the groups.
It isn't a valid experiment unless there is a true control. It is surprisingly easy to mess things up that ruin a controlled experiment. When your test subjects are humans it is almost impossible to get it right. If it were, chances are we'd have cured cancer by now...
Except that it doesn't work that way. The vast majority of criminal cases are settled in a plea-bargain. It works like this:
There is a 20% chance that you killed your wife. The prosecutor charges you with first-degree murder (life sentence, plus risk of execution). Then the presecutor offers you a plea for manslaughter that carries a 2-year prison term. Your choices are to go to trial and risk a life sentence, or just take the 2 years.
Now, sure, if you go to trial the jury SHOULD find you innoent and you get out. Of course, you might sit in jail for three months before your trial even comes up if you don't get bail. However, if you do get convicted you are REALLY up the creek.
Most people just take the deal, and end up being punished for a crime they didn't commit.
Also - merely having to go through the trial is essentially punishment. You could end up in prison for months if you don't make bail. You end up missing work even if you do make bail. While you are entitled to free representation, if you were facing life in prison chances are you'd spend money on the best lawyer you could find.
All of this makes the current system unjust. It is about following a process, not doing what is right.
Don't get me wrong - I'm as tough on crime as anybody is. However, to punish the innocent is not just, and I think that as a society we bear the consequences of doing so. In fact, when a criminal does something wrong they're the only one with blood on their hands, but when as a society we punish the innocent we all end up with blood on our hands.
By the same token, if a prosecutor loses many cases, you have to wonder about his/her skill.
Frankly - I think this is a major problem in criminal justice. Prosecutors are interested in convictions - not prosecuting those who have actually committed crimes. A prosecutor with a weak case doesn't wonder if the guy might be innocent - he just hedges his bets by levelling a nasty charge and then bargaining for a minor one. Sounds nice on paper - a person with a 20% chance of being guilty gets 20% of the punishment, but justice would be 20% of those people getting 100% of the punishment and 80% getting released as expediently as possible.
Btw, NASCAR does this already on some tracks for safety reasons. You don't see any of those cars going 200+ mph. Even though they are completely capable of it.
Yeah, and restrictor plates certainly have done wonders for safety!
Pray tell, where can a research biochemist go to do research where such questionable issues aren't a concern? The pharma industry employs literally tens of thousands of scientists and pretty-much every pharma company has these kinds of practices (I'd say Merck is maybe even slightly better than average in some regards - they're just the company most subject to discovery right now due to Vioxx). Just about every pharma company is laying off scientists, so even if you wanted to switch companies you couldn't (and again it wouldn't be any nicer from an ethical perspective).
Unless you're lobbying for the government to create 50k jobs at the NIH that pay pharma industry wages I'm not qutie sure what you intend for these folks to do. The government can't figure out to do with works in the auto makers - and their skills are a whole lot more generic than some guy doing studies on calcium channels or whatever.
Makes sense - nations have a responsibility to police their own borders. If some Chinese hacker crashes Wall Street the US isn't going to be happy if the Chinese government simply states that it isn't their problem since they didn't sanction the attack.
If some guy set up a big gun in New York state and took to shelling random towns in Canada the Canadians are going to expect the US Government to do something about it. Such a person would be HEAVILY prosecuted and very swift actions would be taken to stop the attack. In the very bizarre case that the US governemnt didn't act swiftly to stop the shelling the Canadians would certainly be within their rights to launch a military strike against the attacker. Certainly such an incident would have huge negative effects on US-Candian relations, and of course it would never be allowed to happen in real life.
Cyber attacks should be no different. Sure, proactively preventing any hacker attack from any rogue actor is going to be just about impossible (though if anybody can do it the Chinese can with their firewalls). However, if the Pentagon calls up the Chinese ambassador and points out that such-and-such an IP is being used to hack into some military installation they would expect the Chinese to halt the attacks and to see somebody punished for it.
Likewise NO government limits retaliation to a tit-for-tat response. If you steal a car on the streets of France and are caught, the police do not merely take it back, and maybe just take your bicycle as well. In order to deter crime the fear of the penalties for being caught when weighed with the liklihood of getting caught needs to be large. So, if you rob a car you get captured in a roadblock, tossed around a bit, and end up in prison for five years. If some Iran-sponsored terrorist cell blows up a UK airplane the UK government doesn't blow up an Iranian plane in retaliation or ask for a million dollars in restitution - they take measures guaranteed to at least cost the Iranians billions, if not outright invading.
A Cyberattack that causes real damages should be responded to in ways that cause real harm to the real people who allowed it to happen. Of course, if a foreign government is generally cooperative in prosecuting the crime then diplomacy should be employed - nobody is going to bomb Germany over a German hacker breaking into some bank since everybody expects the German government to be genuinely horrified by the incident. However, you can't turn a genuinely blind eye towards real harm being done to real people and claim innocence when confronted with the results.