Well, yeah, but then you're getting paid pretty poorly because three years pay has to pay for five year's living expenses. (Three years living while you work for X, and two years while you're unemployed so you don't compete with X.)
The whole idea of intellectual property that one cannot write down is too dangerous and stupid for words. If the IP is not real enough to patent or (gods help us) copyright, then it isn't real.
Research can't be parallelized or managed. But writing up the research sometimes can.
Often, the experiment is done, you've argued it out over a pile of teacups (we're British here) and you think you know pretty much what it means. Then, you can usually do some writing in parallel. And some writing in single-threaded mode.
I've actually had success with editing in parallel, as long as everyone talks to each other. It's a good way to take a rough draft and convert it to something that's almost a final draft. Again, many teacup-long discussions on the way. Then, when it's almost final, you go completely single-threaded and sort out the last problems.
I've used LaTex and CVS or SVN and (more painfully) LaTex and Mercurial. With mercurial, we got lost amid many heads and complicated merges. Distributed source control isn't quite the right thing if you are aiming to produce a single document that everyone agrees to.
No, that's fair enough. But you have to interpret it right. Having a government is a very dangerous thing because it might turn out to be a bad government. Thus, you have to pay attention to make darn sure it doesn't go bad.
And, governments can hurt people by making stupid decisions. The Brits did that after WWII by being overly socialist: they were poor for 20 years because of that. Or, governments can hurt people by not being competent (remember recent American governments and financial regulation?)
So, they are indeed dangerous things and need to be viewed with caution. But, yeah, you can consider the government to be just a tool of the people. It's the same thing.
Darn right! The trouble with doing R&D is the same as the trouble with cleaning the toilet. Everyone hopes that someone else will do it.
However, unlike cleaning toilets, R&D takes a while to turn into a product. We're talking 10 years or more. Consider lasers as an example. How long from the invention until they were making real money? 20-30 years.
Now, there's lots of R&D that pays off faster than lasers, but it does take a while. After all, it takes several years for a radically new version of an OS to come out, so who could expect even the shortest term R&D to pay off faster than that?
Yeah, well, if you read the literature on speech research carefully (and there is a lot of it), you'll find that people have tried for years to get emotional information out of recorded speech. With only minor success. So, it becomes pretty obvious that this company is either (a) much better than the rest of the world put together, or (b) wrong.
Now, researchers have found reasonable correlations between acoustic measurements and excitement. Hesitation, perhaps is obvious (just look for pauses), though what you can measure is that their speech hesitates, which doesn't necessarily they "feel hesitant". It might just mean that they were searching for the right word (concentrating? thinking hard?).
People have done studies with actors, and you'll find papers out there that claim to identify properties of angry speech. Those papers are probably correct, but you have to remember two things: first, they are generally based on acted emotions, and acted emotions are merely social conventions. Unless you collected data from a customer service help-line, it's really hard to find the hours and hours of angry speech that you'd need to train an analysis system.
Second, what researchers have done is to work the problem in the easy direction. Tell an actor to emote, and then look for properties of his speech that differ from "normal", "unemotional" speech. The reverse problem -- that Nemesysco has claimed to solve -- is much harder. Not only do you have to say the speech is not "normal", but you have to disentangle which combination of emotions has caused it.
The other properties have (as far as I am aware) no basis.
Also, the whole idea of "real-time traces" is, as we say in the trade, rather dubious. It suggests that from (for example) a single/s/ sound, you can read the entire emotional state of a person, which is obviously false. Published research has looked at the average properties of large quantities of emotional speech; you need a lot of speech to get even a small amount of emotional information (e.g. excitement). That's because speech is pretty variable on a syllable-to-syllable basis. Some syllables are stressed, others are not. Pitch tends to rise at the end of questions and fall for statements.
In fact, that's a good example of the impossibility of real-time analysis. Consider any emotional measurement that depends on pitch. Person is speaking: pitch goes up. Is it going up because they are excited or because they are asking a question? You can't know that until the sentence is over. Similar arguments apply across the board; extracting emotional information on a scale smaller than a sentence is very dubious, because sentence structure can cause a lot of variability in speech which one cannot understand until you see the entire sentence.
Anyhow, the data it provides will have little to do with any emotional states that can be clearly identified, but doubtless be invaluable for its users to convince themselves of whatever they want to believe.
Freedom of the press refers to journalism, and journalism is more than just running a server that anyone can post to. Journalists take some responsibility for their product, and are supposed to be reporting facts (to the extent possible).
It's like being a scientist in a way. At the end of the day, you want the answer to be right. In pursuit of that goal, scientists and journalists have developed a certain ethic and certain procedures that more or less work.
It's easy to be a pretend scientist: all you have to do is mix pretty solutions in some test tubes. Likewise, it's easy to be a pretend journalist. Easier, these days. But, in either case, the difference between real and pretend is not the web site or the test tubes, but whether or not someone is digging away, really trying to get at the truth.
Now, when Indymedia posts that kind of personal information that could reasonably imply a threat and isn't relevant to the story, it's the equivalent of a chemist blowing up his/her lab. At the very least, it doesn't give you confidence in their competence.
Of course it's crazy. If they were really wanting to be useful, instead of pretending, they'd spend the money on any one of 100 better schemes.
Sheesh!
Dig a hole and fill it with wood. Invest in CO2 collection technology. Wind farms. Fusion research. All *kinds* of things are better than throwing money at little parasols on graves that pretend to be a solution.
You're not a Real Brit, then. Real Brits turn off the heat on 1 March and don't turn it back
on until 1 November.
Central heating is for wimp and colonials.
Oh. Foo!
It's been said that "Democracy is the worst type of government, except for all the others." The same is true for capitalism.
It's easy enough to point out failings of democracy and capitalism. The hard bit is to suggest something that will work better. And, I mean really work better, even when people try to game the system.
Capitalism has proven to be an impressively successful way to organize people to do things collectively. It needs a certain amount of political control and limitation to avoid the worst side-effects, but what's the alternative?
The essence of capitalism, after all, is that you can get someone to work for you by giving them money. It's a kind of persuasion that nicely fills the gap between saying "please" and threatening violence.
Yup. In the UK, here, the Data Protection Act makes it legally dubious to put anyone else's data onto Google. Here, there's a responsibilty to protect personal data.
The trouble with trusted editors is that any large organization can afford to pay someone to become a trusted editor. All you have to do is hire someone reasonably smart, and tell them to spend a day per week helping Wikipedia. Then, once and a while, you tell them to fix what you want fixed. Some would refuse, but others would not want to risk their job.
Since large organizations spend millions on PR, they would happily spend the small sums it would require for this plan. We're talking about US$40,000, which is not a lot. The only reason this plan would fail is that it would be too tempting to demand a lot of edits.
Ultimately, the problem comes down to anonymity. You really want people to put their reputation on the line, and you need people who care about their reputation. Paying university professors to write articles is one solution, though there may be others.
Alternatively, you just accept that Wikipedia is what it is: good for the stuff that everyone knows, of dubious value for controversial stuff (though often surprisingly good!).
Carbon-13 dating does indeed need corrections for the level of solar activity, but that's a bit of an exception. The corrections don't have *anything* to do with how fast the Carbon-13 decays, though: they relate to how much C-13 is in the atmosphere.
The way it works is like this:
Carbon-13 decays in about 5000 years. Why do we still have some around then? That's because it's constantly being made as cosmic rays hit the upper atmosphere of the earth. Now, as a tree grows, it incorporates carbon from the environment, including C-13. When the tree dies, the incorporation pretty much stops. So, once the tree dies, no more C-13 comes in, and whatever is there continues to decay. Thus, some wood with a lot of C-13 is new because the C-13 hasn't had time to decay. Wood with very little C-13 is old, because it has been dead (i.e. disconnected from the environment) for long enough for most of the C-13 to decay.
So, how does solar activity come into the picture? If the sun is inactive, more cosmic rays come in from outer space and hit the Earth's atmosphere. Thus more C-13 is made. Thus, a tree growing in a time with less solar activity incorporates more C-13, and thus when you do C-13 dating of it, it seems to be newer than it really is. The C-13 always decays at the same rate, but if you start with more, you end up with more.
Vice versa, if the sun is particularly active some decade.
Now, this explanation doesn't apply to most kinds of radioactive dating. Uranium dating, for instance, looks at U-238 that has been there since the beginning of the solar system, and cosmic rays just don't matter.
I see trouble for science, too. Lots of science gets done with computer electronics: speech research and psychology research is done 100% with consumer audio and video cards. Lots of other fields use a mix of consumer andio/video and specialized analog to digital converter cards. They'll all get hurt.
Why? Because you need to be able to trust your data and understand it. Science is hard enough if your tools are trustworthy. If your tools start doing unknown processing on your signals, you're in deep hot water (or something smellier). At best, you'd have $10k of grant money to get the VEIL spec and (hopefully) someone who can tell you if it is important. More likely, you'd waste time testing your I/O devices, and then hope (with fingers crossed) that they'll work in your real experiment.
It's a mess. Many people will pay for this in terms of increased costs, greater complexity, mysterious failures and (in my case) midnight worries that VEIL will somehow screw up my experiments. And all for what? Movies, I suppose. But somehow, I suspect that we'd still have movies to watch even if piracy abounded.
Darn right! The universe does not fall into hard-edged classes, at least not often. Some good classes like "protons" and "neutron stars" exist, of course, but concepts like "words" and "species" are intrinsically fuzzy if you think about them long enough.
Same with experiments. Let's take a Linguistic example: deciding whether or not a sentence is gramatically correct. You can do this experiment in several ways:
1) Give the person a sentence, a library, and some paper. Let them take as long as they want.
2) Or, we can make it more like a conversation: read them the sentence, and put a time limit on it. In real speech, you have about a second to understand a sentence, so we only accept a "yes" or "no" if it happens within a second.
3) Make it into a reaction-time experiment. Get them to hit a yes button or a no button and measure how long it takes.
The point is, you can do dozens of variants of any experiment, and any ontology will lump together some things that are different in some important way, or (alternatively) will split apart some experiments that have critical similarities.
Likewise for data analysis.
Personally, I feel that Linguistics has been held back for about two decades by linguist's expectation that everything falls into nice categories. I'd hate for the same thing to happen to other fields.
Just think of the Dewey Decimal system: that's an ontology, and like all ontologies, it puts the dividing lines in the wrong place.
Indeed. Corporations are legally required to attempt to make money for their shareholders (it's called "Fiduciary responsibility"). (This is normally a good thing, because it keeps executives from taking your money and departing for Brazil.)
Corporations are actually forbidden from engaging in activities that do not enhance their business. (This is normally a good thing, because it prevents executives from buying boats for themselves.)
However, anyone with a bit of imagination can see that both of those legal requirements have their dark side. Between them, they prevent corporations from being "nice". They force corporations to be amoral (not immoral): focussed on making money, and little else. Corporate charity, for example, is legally justifiable only from the advertising value it has.
So, Google can perhaps "do no evil", but they cannot legally promise more than that.
Well, I was at AT&T in 1990-1995, and they had a bunch of smart people working on audio and video compression technologies. I suspect that they actually invented something useful.
In fact, AT&T did a video phone in the 1960s (over dedicated co-ax cables) and another around 1990 (over 9600 (!!!) baud modems). There was a long-standing interest and support of this kind of technology.
Patents aren't evil, after all. Not always, anyhow. Sometimes people work hard and spend money and invent something. If that happens, they deserve some rewards. Quite likely (though I don't know for sure) AT&T actually deserves it.
A lot of this carping sounds like sour grapes to me. People who are drunk on the idea of free/open source software, and don't have any historical perspective.
(And, for those who care, I own only a trivial amount of stock in some AT&T spinoffs, like Lucent, which would have more value as a capital loss on my tax return than from the sale itself.)
Most of the phishing (and related!)
problems wouldn't exist if it weren't for those
darn engineers who demanded that a web browser
be able to do everything.
Think about it: The basic mechanism of a phishing attack
is this:
<a href="sleazy.isp">bank.com</a>
That's possible because e-mail is done on HTML clients
these days. Right? Absent HTML, it doesn't work.
Other attacks are done by an advert with an [X] in the corner
of the image or images of [OK] [CANCEL] buttons. They
look real because the modern web browser doesn't get in
the way of any visual display. Imagine an old clunky
browser that put a frame around any image. Would that
kind of attack be so successful? Probably not!
Worse still are attacks that simulate a Windows window on the browser. If the browser insisted on putting a frame and
buttons around pop-ups, those would be obviously
adverts and much less successful.
The problem is that a browser is essentially omnipotent,
so you cannot trivially tell what information comes
from your own computer (and is therefore reasonably trustworthy) versus what comes from some random
criminal on the Internet.
It's the visual equivalent of the Church-Turing thesis: that once a computer's instruction set reaches a certain small complexity, then it can compute anything. Likewise,
once you let the browser be flexible enough, the guy
on the other end can display anything.
So, I lay a lot of the blame on browser writers, including some of our favourite open source projects. In my book, anyone who writes a browser that doesn't always
clearly identify itself as a browser window
is partially responsible for anyone who loses money
by a pfishing attack. Likewise, anyone who writes a
browser that allows content to remove the [X] in the corner
should be legally liable if and when that "feature" is
used to scam someone.
Interestingly enough, actual people count the ballots here in the UK.
I was surprised at first. I thought "How can it scale?" But, if 1 in 1000 of the population volunteers to count votes, then they each need only count about 1000 votes. It takes an evening, and the results are available in the morning.
So, why use a technical solution at all?
I'd contend that if you can't get people to volunteer to count votes, then do you really have an interested electorate?
The actual correlation reported is pretty weak. A correlation of 0.3 (or thereabouts) means that only 10% of the variation of intelligence is explained by the variation in brain size.
Put another way, if you go from a fairly small brained person (1 standard deviation small) to a fairly large brained person (1 standard deviation large), you'd expect the IQ to change by about 3 points.
Someone seems to be trying to manufacture a scandal, or they have an exaggerated view of
the perfection of experiments.
The questions are badly phrased, and cover
ethical activities in addition to unethical
activities. Quotes in italics, comments in roman:
More than 5 percent of scientists answering a confidential questionnaire admitted to having tossed out data because the information contradicted their previous research or said they had circumvented some human research protections.
Yes, one tosses out data occasionally.
One would be an idiot not to, under certain circumstances, because THINGS SOMETIMES GO WRONG IN EXPERIMENTS. Duh!
One should never do it lightly, but there is always a balancing act going on: "Is my data bad, or are the previous publications wrong?"
Ideally, one would not throw out data unless you
could prove that your experiment was flawed,
and knew what was broken,
but sometimes things are just too confusing.
Real experiments are often complicated
things, and they can suffer from subtle problems.
Sometimes you "know" something is wrong, but
simply cannot pin it down.
It would be a dis-service to publish such
data. The world has enough junky papers;
one should not contaminate the literature
further with the results
of broken experiments.
Ten percent admitted they had inappropriately included their names or those of others as authors on published research reports.
There's no excuse for such behavior,
but it doesn't make the results invalid.
It's a human sin, not a scientific sin.
It's often done to steal credit, or
to borrow a reputation.
Personally, I would string up the perpetrators,
but I would not distrust their results.
(There are also grey zones: what do you do
for the guy who lends a willing ear at lunch,
and then spent a couple of volunteer days
on a data analysis algorithm that you
didn't use? Should he be an author or not?)
And more than 15 percent admitted they had changed a study's design or results to satisfy a sponsor, or ignored observations because they had a "gut feeling" they were inaccurate.
Changing results is bad, but changing a study's design can be quite reasonable. There's nothing magical about
the design -- it's just a question one asks
of Nature. Some designs are better than
others, but often there are many reasonable
questions one can ask.
Why not change the question?
One shouldn't change the design to avoid unpleasant
results, of course, but there are certainly
many situations where changing the design
can be quite reasonable and ethical.
As for ignoring results,
well, one cannot condone it,
but (as above) things sometimes go wrong.
If the
gut feeling has some plausible basis,
if one is honestly suspicious about the
data,
it's not necessarily a terrible thing.
It's dangerous, because it is very easy
to fool yourself into dropping correct
but theoretically inconvenient data.
It's bad practice,
because it makes the published results less trustworthy,
but it can be done honestly.
It really boils down to whether or not
one admits it in
the publication: if you admit that you
dropped the data, and if you dropped
it without trying to influence the result,
it's OK.
Other researchers will be warned; they may
take your results with a grain of salt,
but the paper can still be valuable
especially if the amount of data dropped
is small.
Dropping data and not admitting it
in the publication is marginal at best,
unless you are darn sure there was
something wrong with the experiment.
(Bad data points are often better addressed
by using more sophisticated statistical techniques,
but many people don't understand the
necessary techniques or even realize that
they exist.)
None of those failings qualifies as outright scientific misconduct under the strict definition used by federal regulators....
No, people who have to write ten grant
proposals in order to get the opportunity
to work, so they can feed their children
are busy, not sophists.
Despite the reality that academics are not
well paid, the competition is still intense,
and one ends up working much more than the
nominal 40 hours/week, if one wants to maintain
a career.
Consequently, contributing to Wikipedia
or even arguing about it (like I'm doing)
is really a luxury that many academics
cannot afford.
If there aren't references, it isn't scholarship
on
FUD-Based Encyclopedias
·
· Score: 2, Interesting
The trouble with Wikipedia in practice is that
there aren't any (or darn few) references to real
outside sources. That means to documents carefully
written by people who know what they are doing.
Absent that kind of care and detail,
how can you know if it's right or not?
And, if you don't know it's right, why use it?
Call me elitist if you like, but I like my
doctor to have a M.D.; I like the guys who
design my buildings and aircraft to be
Real Engineers, and I like the guys who
write my reference sources to be real scholars.
The basic difference between Wikipedia and
Free Software is that Wikipedia doesn't have
a compiler.
A compiler, you say? Yes!
Compilers do something much more important than
turn C into machine instructions. The do a
critical first step of filtering out people who
are not able to usefully contribute. They
get rid of the people who can't learn to program
and the people who can't be bothered to learn
the details of the program in question.
If someone sends you a patch and it doesn't compile, just toss it out! Sure, you'll lose
a few good ideas, but you'll throw out a lot
of incoherent garbage and save yourself some
valuable time.
Without a compiler as a first filter,
can you imagine actually getting the Linux
kernel to compile, if everyone could add patches?
(If you can imagine it, you should start writing
science fiction...) That's the situation of
Wikipedia.
Does anyone seriously believe that human knowledge
is simpler than the Linux kernel? Seesh!
The problem with the Wikipedia idea is that
all the people who really know and care about
some topic would have to spend their entire
lives guarding it from all kinds of problems:
inveterate fiddlers, guys with axes to grind,
and the many many slightly confused people
in the world. Without that intense and
permanent guardianship, it will simply be
wrong.
Just like the Linux kernel would, if anyone
could add in patches.
Oh, yes. Software has one more advantage
over the Wikipedia. When there's a mistake
inside, sometimes you get lucky and it crashes.
When that happens, people tend to
realize that something is wrong. When an
entry in the Wikipedia is wrong, what happens?
Nothing.
Slashdot Mirror
Well, yeah, but then you're getting paid pretty poorly because three years pay has to pay for five year's living expenses. (Three years living while you work for X, and two years while you're unemployed so you don't compete with X.)
The whole idea of intellectual property that one cannot write down is too dangerous and stupid for words. If the IP is not real enough to patent or (gods help us) copyright, then it isn't real.
Research can't be parallelized or managed. But writing up the research sometimes can.
Often, the experiment is done, you've argued it out over a pile of teacups (we're British here) and you think you know pretty much what it means. Then, you can usually do some writing in parallel. And some writing in single-threaded mode.
I've actually had success with editing in parallel, as long as everyone talks to each other. It's a good way to take a rough draft and convert it to something that's almost a final draft. Again, many teacup-long discussions on the way. Then, when it's almost final, you go completely single-threaded and sort out the last problems.
I've used LaTex and CVS or SVN and (more painfully) LaTex and Mercurial. With mercurial, we got lost amid many heads and complicated merges. Distributed source control isn't quite the right thing if you are aiming to produce a single document that everyone agrees to.
No, that's fair enough. But you have to interpret it right. Having a government is a very dangerous thing because it might turn out to be a bad government. Thus, you have to pay attention to make darn sure it doesn't go bad.
And, governments can hurt people by making stupid decisions. The Brits did that after WWII by being overly socialist: they were poor for 20 years because of that. Or, governments can hurt people by not being competent (remember recent American governments and financial regulation?)
So, they are indeed dangerous things and need to be viewed with caution. But, yeah, you can consider the government to be just a tool of the people. It's the same thing.
Darn right! The trouble with doing R&D is the same as the trouble with cleaning the toilet. Everyone hopes that someone else will do it.
However, unlike cleaning toilets, R&D takes a while to turn into a product. We're talking 10 years or more. Consider lasers as an example. How long from the invention until they were making real money? 20-30 years.
Now, there's lots of R&D that pays off faster than lasers, but it does take a while. After all, it takes several years for a radically new version of an OS to come out, so who could expect even the shortest term R&D to pay off faster than that?
I work in the field. Anders Eriksson is very smart and well respected. (I don't really know Lacerda.)
Yeah, well, if you read the literature on speech research carefully (and there is a lot of it), you'll find that people have tried for years to get emotional information out of recorded speech. With only minor success. So, it becomes pretty obvious that this company is either (a) much better than the rest of the world put together, or (b) wrong.
Now, researchers have found reasonable correlations between acoustic measurements and excitement. Hesitation, perhaps is obvious (just look for pauses), though what you can measure is that their speech hesitates, which doesn't necessarily they "feel hesitant". It might just mean that they were searching for the right word (concentrating? thinking hard?).
People have done studies with actors, and you'll find papers out there that claim to identify properties of angry speech. Those papers are probably correct, but you have to remember two things: first, they are generally based on acted emotions, and acted emotions are merely social conventions. Unless you collected data from a customer service help-line, it's really hard to find the hours and hours of angry speech that you'd need to train an analysis system.
Second, what researchers have done is to work the problem in the easy direction. Tell an actor to emote, and then look for properties of his speech that differ from "normal", "unemotional" speech. The reverse problem -- that Nemesysco has claimed to solve -- is much harder. Not only do you have to say the speech is not "normal", but you have to disentangle which combination of emotions has caused it.
The other properties have (as far as I am aware) no basis.
Also, the whole idea of "real-time traces" is, as we say in the trade, rather dubious. It suggests that from (for example) a single /s/ sound, you can read the entire emotional state of a person, which is obviously false. Published research has looked at the average properties of large quantities of emotional speech; you need a lot of speech to get even a small amount of emotional information (e.g. excitement). That's because speech is pretty variable on a syllable-to-syllable basis. Some syllables are stressed, others are not. Pitch tends to rise at the end of questions and fall for statements.
In fact, that's a good example of the impossibility of real-time analysis. Consider any emotional measurement that depends on pitch. Person is speaking: pitch goes up. Is it going up because they are excited or because they are asking a question? You can't know that until the sentence is over. Similar arguments apply across the board; extracting emotional information on a scale smaller than a sentence is very dubious, because sentence structure can cause a lot of variability in speech which one cannot understand until you see the entire sentence.
Anyhow, the data it provides will have little to do with any emotional states that can be clearly identified, but doubtless be invaluable for its users to convince themselves of whatever they want to believe.
It's like being a scientist in a way. At the end of the day, you want the answer to be right. In pursuit of that goal, scientists and journalists have developed a certain ethic and certain procedures that more or less work.
It's easy to be a pretend scientist: all you have to do is mix pretty solutions in some test tubes. Likewise, it's easy to be a pretend journalist. Easier, these days. But, in either case, the difference between real and pretend is not the web site or the test tubes, but whether or not someone is digging away, really trying to get at the truth.
Now, when Indymedia posts that kind of personal information that could reasonably imply a threat and isn't relevant to the story, it's the equivalent of a chemist blowing up his/her lab. At the very least, it doesn't give you confidence in their competence.
Of course it's crazy. If they were really wanting to be useful, instead of pretending, they'd spend the money on any one of 100 better schemes.
Sheesh!
Dig a hole and fill it with wood. Invest in CO2 collection technology. Wind farms. Fusion research. All *kinds* of things are better than throwing money at little parasols on graves that pretend to be a solution.
You're not a Real Brit, then. Real Brits turn off the heat on 1 March and don't turn it back on until 1 November. Central heating is for wimp and colonials.
It's easy enough to point out failings of democracy and capitalism. The hard bit is to suggest something that will work better. And, I mean really work better, even when people try to game the system.
Capitalism has proven to be an impressively successful way to organize people to do things collectively. It needs a certain amount of political control and limitation to avoid the worst side-effects, but what's the alternative?
The essence of capitalism, after all, is that you can get someone to work for you by giving them money. It's a kind of persuasion that nicely fills the gap between saying "please" and threatening violence.
Yup. In the UK, here, the Data Protection Act makes it legally dubious to put anyone else's data onto Google. Here, there's a responsibilty to protect personal data.
The trouble with trusted editors is that any large organization can afford to pay someone to become a trusted editor. All you have to do is hire someone reasonably smart, and tell them to spend a day per week helping Wikipedia. Then, once and a while, you tell them to fix what you want fixed. Some would refuse, but others would not want to risk their job.
Since large organizations spend millions on PR, they would happily spend the small sums it would require for this plan. We're talking about US$40,000, which is not a lot. The only reason this plan would fail is that it would be too tempting to demand a lot of edits.
Ultimately, the problem comes down to anonymity. You really want people to put their reputation on the line, and you need people who care about their reputation. Paying university professors to write articles is one solution, though there may be others.
Alternatively, you just accept that Wikipedia is what it is: good for the stuff that everyone knows, of dubious value for controversial stuff (though often surprisingly good!).
Uh, you're a bit confused.
Carbon-13 dating does indeed need corrections for the level of solar activity,
but that's a bit of an exception. The corrections don't have *anything*
to do with how fast the Carbon-13 decays, though: they relate to how much
C-13 is in the atmosphere.
The way it works is like this:
Carbon-13 decays in about 5000 years. Why do we still have some around then?
That's because it's constantly being made as cosmic rays hit the upper atmosphere
of the earth. Now, as a tree grows, it incorporates carbon from the environment,
including C-13. When the tree dies, the incorporation pretty much stops.
So, once the tree dies, no more C-13 comes in, and whatever is there continues
to decay. Thus, some wood with a lot of C-13 is new because the C-13 hasn't
had time to decay. Wood with very little C-13 is old, because it has been
dead (i.e. disconnected from the environment) for long enough for most of the
C-13 to decay.
So, how does solar activity come into the picture? If the sun is inactive,
more cosmic rays come in from outer space and hit the Earth's atmosphere.
Thus more C-13 is made. Thus, a tree growing in a time with less solar
activity incorporates more C-13, and thus when you do C-13 dating of it,
it seems to be newer than it really is. The C-13 always decays at the same
rate, but if you start with more, you end up with more.
Vice versa, if the sun is particularly active some decade.
Now, this explanation doesn't apply to most kinds of radioactive dating.
Uranium dating, for instance, looks at U-238 that has been there since
the beginning of the solar system, and cosmic rays just don't matter.
I see trouble for science, too. Lots of science gets done with computer electronics:
speech research and psychology research is done 100% with consumer audio and video cards.
Lots of other fields use a mix of consumer andio/video and specialized analog to digital
converter cards. They'll all get hurt.
Why? Because you need to be able to trust your data and understand it.
Science is hard enough if your tools are trustworthy. If your tools start doing
unknown processing on your signals, you're in deep hot water (or something smellier).
At best, you'd have $10k of grant money to get the VEIL spec and (hopefully)
someone who can tell you if it is important. More likely, you'd waste time
testing your I/O devices, and then hope (with fingers crossed) that they'll
work in your real experiment.
It's a mess. Many people will pay for this in terms of increased costs,
greater complexity, mysterious failures and (in my case) midnight worries
that VEIL will somehow screw up my experiments. And all for what?
Movies, I suppose. But somehow, I suspect that we'd still have movies to
watch even if piracy abounded.
Darn right! The universe does not fall into
hard-edged classes, at least not often.
Some good classes like "protons" and
"neutron stars" exist, of course, but
concepts like "words" and "species" are
intrinsically fuzzy if you think about them
long enough.
Same with experiments. Let's take a Linguistic
example: deciding whether or not a sentence is
gramatically correct. You can do this experiment
in several ways:
1) Give the person a sentence, a library, and
some paper. Let them take as long as they want.
2) Or, we can make it more like a conversation:
read them the sentence, and put a time limit on
it. In real speech, you have about a second
to understand a sentence, so we only accept
a "yes" or "no" if it happens within a second.
3) Make it into a reaction-time experiment.
Get them to hit a yes button or a no button
and measure how long it takes.
The point is, you can do dozens of variants of any
experiment, and any ontology will lump together
some things that are different in some important
way, or (alternatively) will split apart some
experiments that have critical similarities.
Likewise for data analysis.
Personally, I feel that Linguistics has been held
back for about two decades by linguist's expectation
that everything falls into nice categories.
I'd hate for the same thing to happen to other fields.
Just think of the Dewey Decimal system: that's an
ontology, and like all ontologies, it puts the
dividing lines in the wrong place.
Indeed. Corporations are legally required to attempt to make money for their
shareholders (it's called "Fiduciary responsibility").
(This is normally a good thing, because it keeps executives from
taking your money and departing for Brazil.)
Corporations
are actually forbidden from engaging in activities that do not
enhance their business. (This is normally a good thing,
because it prevents executives from buying boats for themselves.)
However, anyone with a bit of imagination can see that both
of those legal requirements have their dark side. Between
them, they prevent corporations from being "nice". They force
corporations to be amoral (not immoral): focussed on making money,
and little else. Corporate charity, for example, is legally
justifiable only from the advertising value it has.
So, Google can perhaps "do no evil", but they cannot legally
promise more than that.
Well, I was at AT&T in 1990-1995, and they had a bunch of smart
people working on audio and video compression technologies.
I suspect that they actually invented something useful.
In fact, AT&T did a video phone in the 1960s (over
dedicated co-ax cables) and another around 1990 (over
9600 (!!!) baud modems). There was a long-standing
interest and support of this kind of technology.
Patents aren't evil, after all. Not always, anyhow.
Sometimes people work hard and spend money and
invent something. If that happens, they deserve some
rewards. Quite likely (though I don't know for sure)
AT&T actually deserves it.
A lot of this carping sounds like sour grapes to me.
People who are drunk on the idea of free/open source
software, and don't have any historical perspective.
(And, for those who care, I own only a trivial amount
of stock in some AT&T spinoffs, like Lucent, which
would have more value as a capital loss on my tax return
than from the sale itself.)
Think about it: The basic mechanism of a phishing attack is this:
<a href="sleazy.isp">bank.com</a>
That's possible because e-mail is done on HTML clients these days. Right? Absent HTML, it doesn't work.
Other attacks are done by an advert with an [X] in the corner of the image or images of [OK] [CANCEL] buttons. They look real because the modern web browser doesn't get in the way of any visual display. Imagine an old clunky browser that put a frame around any image. Would that kind of attack be so successful? Probably not!
Worse still are attacks that simulate a Windows window on the browser. If the browser insisted on putting a frame and buttons around pop-ups, those would be obviously adverts and much less successful.
The problem is that a browser is essentially omnipotent, so you cannot trivially tell what information comes from your own computer (and is therefore reasonably trustworthy) versus what comes from some random criminal on the Internet.
It's the visual equivalent of the Church-Turing thesis: that once a computer's instruction set reaches a certain small complexity, then it can compute anything. Likewise, once you let the browser be flexible enough, the guy on the other end can display anything.
So, I lay a lot of the blame on browser writers, including some of our favourite open source projects. In my book, anyone who writes a browser that doesn't always clearly identify itself as a browser window is partially responsible for anyone who loses money by a pfishing attack. Likewise, anyone who writes a browser that allows content to remove the [X] in the corner should be legally liable if and when that "feature" is used to scam someone.
Interestingly enough, actual people count the ballots here in the UK.
I was surprised at first.
I thought "How can it scale?" But, if 1 in 1000 of
the population volunteers to count votes, then
they each need only count about 1000 votes.
It takes an evening, and the results are available
in the morning.
So, why use a technical solution at all?
I'd contend that if you can't get people to
volunteer to count votes, then do you really
have an interested electorate?
The actual correlation reported is pretty weak.
A correlation of 0.3 (or thereabouts) means
that only 10% of the variation of intelligence
is explained by the variation in brain size.
Put another way, if you go from a fairly small
brained person (1 standard deviation small) to
a fairly large brained person (1 standard deviation
large), you'd expect the IQ to change by about
3 points.
So, even if it's true, it's no big deal.
Someone seems to be trying to manufacture a scandal, or they have an exaggerated view of the perfection of experiments. The questions are badly phrased, and cover ethical activities in addition to unethical activities. Quotes in italics, comments in roman:
More than 5 percent of scientists answering a confidential questionnaire admitted to having tossed out data because the information contradicted their previous research or said they had circumvented some human research protections.
Yes, one tosses out data occasionally. One would be an idiot not to, under certain circumstances, because THINGS SOMETIMES GO WRONG IN EXPERIMENTS. Duh!
One should never do it lightly, but there is always a balancing act going on: "Is my data bad, or are the previous publications wrong?" Ideally, one would not throw out data unless you could prove that your experiment was flawed, and knew what was broken, but sometimes things are just too confusing. Real experiments are often complicated things, and they can suffer from subtle problems. Sometimes you "know" something is wrong, but simply cannot pin it down. It would be a dis-service to publish such data. The world has enough junky papers; one should not contaminate the literature further with the results of broken experiments.
Ten percent admitted they had inappropriately included their names or those of others as authors on published research reports.
There's no excuse for such behavior, but it doesn't make the results invalid. It's a human sin, not a scientific sin. It's often done to steal credit, or to borrow a reputation. Personally, I would string up the perpetrators, but I would not distrust their results. (There are also grey zones: what do you do for the guy who lends a willing ear at lunch, and then spent a couple of volunteer days on a data analysis algorithm that you didn't use? Should he be an author or not?)
And more than 15 percent admitted they had changed a study's design or results to satisfy a sponsor, or ignored observations because they had a "gut feeling" they were inaccurate.
Changing results is bad, but changing a study's design can be quite reasonable. There's nothing magical about the design -- it's just a question one asks of Nature. Some designs are better than others, but often there are many reasonable questions one can ask. Why not change the question? One shouldn't change the design to avoid unpleasant results, of course, but there are certainly many situations where changing the design can be quite reasonable and ethical.
As for ignoring results, well, one cannot condone it, but (as above) things sometimes go wrong. If the gut feeling has some plausible basis, if one is honestly suspicious about the data, it's not necessarily a terrible thing. It's dangerous, because it is very easy to fool yourself into dropping correct but theoretically inconvenient data. It's bad practice, because it makes the published results less trustworthy, but it can be done honestly. It really boils down to whether or not one admits it in the publication: if you admit that you dropped the data, and if you dropped it without trying to influence the result, it's OK. Other researchers will be warned; they may take your results with a grain of salt, but the paper can still be valuable especially if the amount of data dropped is small. Dropping data and not admitting it in the publication is marginal at best, unless you are darn sure there was something wrong with the experiment. (Bad data points are often better addressed by using more sophisticated statistical techniques, but many people don't understand the necessary techniques or even realize that they exist.)
None of those failings qualifies as outright scientific misconduct under the strict definition used by federal regulators. ...
Because they are either
Despite the reality that academics are not well paid, the competition is still intense, and one ends up working much more than the nominal 40 hours/week, if one wants to maintain a career.
Consequently, contributing to Wikipedia or even arguing about it (like I'm doing) is really a luxury that many academics cannot afford.
Call me elitist if you like, but I like my doctor to have a M.D.; I like the guys who design my buildings and aircraft to be Real Engineers, and I like the guys who write my reference sources to be real scholars.
The basic difference between Wikipedia and Free Software is that Wikipedia doesn't have a compiler. A compiler, you say? Yes!
Compilers do something much more important than turn C into machine instructions. The do a critical first step of filtering out people who are not able to usefully contribute. They get rid of the people who can't learn to program and the people who can't be bothered to learn the details of the program in question. If someone sends you a patch and it doesn't compile, just toss it out! Sure, you'll lose a few good ideas, but you'll throw out a lot of incoherent garbage and save yourself some valuable time.
Without a compiler as a first filter, can you imagine actually getting the Linux kernel to compile, if everyone could add patches? (If you can imagine it, you should start writing science fiction...) That's the situation of Wikipedia.
Does anyone seriously believe that human knowledge is simpler than the Linux kernel? Seesh!
The problem with the Wikipedia idea is that all the people who really know and care about some topic would have to spend their entire lives guarding it from all kinds of problems: inveterate fiddlers, guys with axes to grind, and the many many slightly confused people in the world. Without that intense and permanent guardianship, it will simply be wrong. Just like the Linux kernel would, if anyone could add in patches.
Oh, yes. Software has one more advantage over the Wikipedia. When there's a mistake inside, sometimes you get lucky and it crashes. When that happens, people tend to realize that something is wrong. When an entry in the Wikipedia is wrong, what happens? Nothing.
I don't see why people need to restrict the
bad idea of a unique EULA for each object
that you own to software.
See http://bunop.com for pencils with EULAs.
That'll fix the guy up above who recommended
using pencil and paper for security...