They have a record of doing so. And even if not specifically, many US prisons are at the level of torture by many measures, for example prologed solitary confinement, high levels of prison violence, etc.
Can you rabid Trumpanzees at least get your Hillary hating straight? Seriously this is a Trump administration action and the first two posts are using as an excuse to attack Hillary.
However, the power of 2 seems arbitrary to me, and possibly over-emphasizes outliers.
It's not arbitrary, and yes it does emphasise outliers.
This is going to be hard to explain given slashdot's formatting ability but here goes...
For least squares, the assumption is that the noise is Gaussian, where every datapoint has the same (unknown) noise variance. If you have some linear function of parameters f(a), what you're doing is finding the a that maximises the probability of observing the data.
Assume you have datapoints d_i taken at positions x_i, and a linear function f(x; a) (vector valued a), which in a simple case is f(x;a) = a_1 x + a_2, i.e. the recognisable equation for a straight line.
If your data is gaussian, the probability of observing a point is based on the Gaussian distribution:
where s is the standard deviation and Z is the gaussian normalizing constant which I'm too lazy to write out since it vanishes soon anyway. For all the data, you multiply the probabilities:
P(d|x,a) = P(d_1 | x_1,a) * P(d_2 | x_2,a) *...
You want to find the a that maximizes that. log is monotonic, so if you maximize log(w) you also maximize w, so take the log, making the products sums, and making the exps disappear. Also the position of the maximum is not affected by adding a constant or scaling, so we simply remove all constant multiples and scales. That results in:
argmax_a -(d1 - f(x1;a)^2 - (d2 - f(x2;s)^2 -...
(argmax being meaning return the a that maximizes the expression).
Well, now turn the maximum into a minimum and flip the signs and you have least squares.
You're right about the computational aspect, efficient solutions for linear least squares do exist which make it particularly appealing. It also has a single global minimum which is not only tractable but easy to find. Any power less than 1 does not in general have a tractable, guaranteed solution.
You're also right that you can use other powers, and in fact other functions completely. In general though you can substitute into the expression whatever probability distribution your data does come from. If it's not Gaussian, it will probably have fat tails and therefore the result will be robust to outliers. You can then optimize directly using some numerical method. It's best to take logs and minimize if you don't want to run badly into floating point trouble.
The easiest is to use downhill simplex (fminunc in MATLAB/octave), since it's derivative free. Or you can compute derivatives and use one of the many fine off the shelf general function optimizers. Or you can use iterative reweighted least squares, which is usually pretty efficient for least-something problems.
Now you've also wound up in the whole field of robust statistics and M-estimators.
Anyway your intuition is mostly correct here: if you have outliers you need something robust to outliers, and with the computational power we have now we can (and do) do things more involved thatn least quares, though often related.
One thing though about the best known algorithm: as your algorithms get robust you get local minima, in which case you can find a better solution given a decent starting point, but you (provably in many cases) can't guarantee you'll get the globally optimal solution.
So, to get a decent solution you need a good starting point. For that, you either need a good guess (often really not hard) or failing that, an algorithm like RANSAC is really really good.
Here's a fun addition/side/related point. Take the simplest possible version of what you have above, i.e. a slope of 1 and you're just trying to find the shift. In other words you don't expect the data to vary so you're finding just the central position in some way. With Gaussian noise (least squares) you get the mean dropping out of the squation. With a 2 sided Laplacian (exp(-|x|), rather than exp(-x^2)) you get the median. And as the power goes towards 0, you get the mode.
I submitted the paper over a *year* ago. The necessary changes were minor. But the actual time it took to go through the process was excruciating.
Apart from a few exception, yeah submitting papers takes aaageeess. Basically (and this has nothing to do with you or the quality of your work), no one wants t oread your paper. I mean people accpet they need to read and review papers as part of the academic papers, but no one WANTS to do it.
Half of it is they want to be doing their own research. The other half is that 80% of everything is crap including the papers they have to review. Not only that, some papers are great and it's an easy accept. Some are terrible and it's an easy reject. The rest are kind of shitty and you have to put a lot of effort in reading a shitty paper to decide which side of the line it falls, especially to reject since you should give reasons.
When your paper lands in their inbox, there's an 80% chance it's one of those.
So reading papers is time consuming so the journals give people plenty of time to read them, say, 6 weeks. Then inevitably what happens is the reviewer in question will not get to it during then time and will find a few hours after the second reminder. It's embarrassing after that.
Then the journal must let it compost for a while.
Then the unpaid editors have to process that, check the reviews, and decide what to do etc etc.
Why would convicting someone of something without bothering to look into the facts be doing him a favor?
The law is written that way precisely to stop dickheads from pulling this shit then weaseling out of it. When you get behind the wheelof a car you are a danger to others. Take some responsibility.
I am sure also apple music Germany is censoring selling pro nazi song sales. We may not like the law of *their* country, but apple has to respect local laws *where ever* it sells.
Yeah but Germany is basically a democracy with free citizens. China is a repressive autocracy that harvests organs from political prisoners.
Supporting one is not the same as supporting the other.
If they don't want to respect the law of a country, the solution is simply : don't sell to anybody in that country.
That does seem to be what people are indeed advocating. Don't lend support to the repressive organ-harvesting regime even if there's lots of money to be made.
Well this is a substantially different claim from the original one, and I agree more with it.
But for a majority of tasks that programmers used C++ for earlier, Java has proven to be an effective and easier replacement.
Yesbut...
I mean yes, for plenty of stuff Java is a more appropriate choice than C++. On the other hand if anything C++ is growing now, and a lot of migration was a long time ago. The world of C++ now is nothing like the world of C++ in 1996 when Java first stormed onto the scene.
Obviously, no one actually wants to use something like Hot Java
Wow that brings back memories. I've not heard that name in a VERY long time!
Some of my colleagues work with image processing (research). They all work in MATLAB, most of the time. Obviously, it is glacial as one would expect a dynamically typed language to be. But that isn't the point when you are exploring transformations. This is the expensive part (human time). Once the algorithms stabilize, they are handed over to C++ engineers for final implementations.
I use both. I've made prototypes in both MATLAB and C++, and continue to do so. The thing that sets C++ aside is it's a much more flexible language than MATLAB, it's sometimes referred to a a libary writer's language.
Writing good libraries is hard, very hard, and few people can do it. But those libraries can be truly excellent nd act almost like an EDSL. MATLAB doesn't have that. It's basically an array processing language with an emphasis on linear algebra, and it always lookslike that.
For example despite it's mathematical pedigree, MATLAB still doesnt have autodiff. If your image processing step involves any optimization it's much quicker and easier if you don't have to compute gradients.
Likewise the lack of strong typing is a bit of a pain. It means I have tofind bugs at runtime.
Performance ("Web browsers and compilers"), not expressiveness.
I think you're stretching the point somewhat. It's not easier to write a web browser in Java is the resulting web browser is not useable. What's the point in doing three quarters of a task?
It would generally be more elegant and easier to express the Eigen, Ceres, Armadillo code in MATLAB, Mathematica or a similar language.
I know MATLAB and no it wouldn't. MATLAB for example doesn't give you static typing on, well, anything. Eigen and similar libraries give you static typing on matrix sizes, eliminating a whole slew of things at comile time.
And with matlab,to get anything other than truly glacial performance, you haveto use MATLAB idioms which involves cramming eveything into an array expession which is generally harder to read, write and debug than for loops.
MATLAB has it's place and I like it a great deal, however, for many tasks I reach for C++ first.
Java is generally much easier than C++, unless you need memory control.
Depends on the task. I've worked on image procesing code in both C++ and Java. It's a lot easier in C++.
When you need that flexibility in Java, you pull out Scala
That's not Java. That's a language running on the JVM but it certainly isn't Java.
I translated "mind-numbingly" to "utter garbage", an exaggeration, I agree.
So you translated one thing to something completely different. It would make sense if the ISO's job was to entertain. But it's not. Boring/exciting is not within the remit of the ISO and frankly is not an issue for standards.
Poor technology documentation is a HUGE problem. "Mind-numbingly" is just one of the problems with poor documentation.
Lack of precision for for a standard is a bigger problem then being boring. Much bigger.
You're lumping in user guides, getting started tutorials with standards, which is frankly facile. The goal of a standard is not to entertain, it's so that two gery techincal groups can without talking to each other make two implementations of something which behave in a consistent manner.
A standards documentation is needed so that I can determine if struct {int a;b;} a; &(a.a)+1; is valid.
Also, personal attacks are unhealthy for the attacker.
So why did you do it then? You started off on the attack about a group of people some of whom I've worked with.
Unfortunately, the C++ standards committee is philosophically opposed to subsetting,
Not when it comes to the library they're not. There's currently non hosted (no library) and hosted (with the library). There are proposals making their way through to make the non hosted one have some useful libraries.
but it is much easier to write Java than C++ for all but the simplest programs.
So I don't really get your point.
There's plenty of stuff written in C++. If it's much easier to write programs in Java, then why are major bitsof infrastructure (like Web browsers and compilers) generally written in C++? Not to mention games, numerical computation, embedded stuff, and so on.
So I'm not sure what your point is. There are plenty of nontrivial programs for which C++ is the correct choice.
but it is much easier to write Java than C++ for all but the simplest programs.
I strongly disagree with that.
I'm not a fan of java, but I'm not a hater of it either, from my point of view, it's solidly OK. But there's no way java is better for all programs than C++. C++ allows for much more expressive libraries than Java and if someone's put in the work that makes code much much easier to write.
Take Eigen for example. Not a library you want to delve into, but as a user, you can write code that looks like maths. Or even moreso, take Ceres. You write maths just like normal and it automatically differentiates the expressions.
For certain applications there are nice fixed point libraries.
Image processing is miserable in java. There are many reasons but the lack of unsigned bytes is really super annoying
That's all on the computation end, but that's my field at the moment. Java's just not a good fit, and it's simply easier in C++.
clang and VS have imlemented it, albeit at the moment two different versions of the draft standards. GCC has a WIP implementation that's woking but not complete.
"The designers were primarily motivated by their shared dislike of C++."
Euch. You could insert "irrational" or "misinformed" into that sentnce. I use C++ a lot. I could give you a long list of complaints about it. A very long list. There are things I'd love to see fixed, or a "better" language replace it.
Basically degenerates into a whiny rant about go is too awesome for C++ programmers and that's why C++ programmers have been pretty uninterested in go. It seems he can't really cope with the possibility that go isn't that amazing and certainly doesn't cover as many use cases as C++.
"Why ESR Hates C++, Respects Java, and Thinks Go (But Not Rust) Will Replace C"
Yeah well, ESR is a bit of a plonker. What are you going to post next? His guide to sex? (no realy this is a thing http://www.catb.org/esr/writin...)
You do see the flaw with that, I would hope. As long as the political narrative is acceptable, or you didn't give someone booboo feelings you would get paid. Deviate from the straight and narrow, and you don't.
That's one of the most warped and entitled things I've ever read. No, nothing at all gives you a RIGHT to get paid for your speech.
I thought you far left people thought that companies are pretty amoral.
Yes well you would think that wouldn't you. You have no idea wht "left" even is, so you ascribe some mishmash bag of things to it then convince yourself that anyone who doesn't agree with you on everything must therefore believe all the things in your bag.
They need to use demographics, just like regular Television has so successfully. Takes a bit of work, but not that hard.
No you don't get it. It's their money, they don't need to do anything.
Why do you keep telling other people what to do with their money? Are you a communist or something?
This everybody should code fad is one of the dumbest things I've ever seen.
why stop there? I think teaching reading and writing is a fad.
maths too.
I think what we should do is thoroughly analyze every 4 year old, determine their vocation then train them exclusively for that ignoring everything else.
And do you know what else is a dream job? Being a financial advisor to someone with a crapload of money and a 9th grade education
*blink*
um
I don't think "financial adviser" is ANYONE's dream job. Well it might be but then you have the dullest dreams in the world. You could couple your dream job with a dream house of a moderately sized place in suburbia with upper-end Ikea furniture (not the cheap stuff).
I'll keep working on it for now, but if they screw me over again I might have to bail out.
I don't get it. Before they moved the goal posts on monteization, you racked up 43 cents in 2 years. It sounds like you're in it for the fun of making videos not the money so why do you care if they move the monetisation goalposts?
Slashdot Mirror
Some people have the opinion I should be murdered. That's pretty toxic.
You don't get a free pass from being an asshole just because you have opinions.
Why would they torture him?
They have a record of doing so. And even if not specifically, many US prisons are at the level of torture by many measures, for example prologed solitary confinement, high levels of prison violence, etc.
On the other hand, he humiliated the government and caused it to waste many millions on policing outside the embassy.
That's a bit of a stretch. They chose to do that; they wouldn't normally put anything like those resources on a crime of that level.
Can you rabid Trumpanzees at least get your Hillary hating straight? Seriously this is a Trump administration action and the first two posts are using as an excuse to attack Hillary.
You won, get over it.
However, the power of 2 seems arbitrary to me, and possibly over-emphasizes outliers.
It's not arbitrary, and yes it does emphasise outliers.
This is going to be hard to explain given slashdot's formatting ability but here goes...
For least squares, the assumption is that the noise is Gaussian, where every datapoint has the same (unknown) noise variance. If you have some linear function of parameters f(a), what you're doing is finding the a that maximises the probability of observing the data.
Assume you have datapoints d_i taken at positions x_i, and a linear function f(x; a) (vector valued a), which in a simple case is f(x;a) = a_1 x + a_2, i.e. the recognisable equation for a straight line.
If your data is gaussian, the probability of observing a point is based on the Gaussian distribution:
P(d_i | x_i,a) = exp(-(d_i - f(x_i;a))^2 / 2s^2) / Z
where s is the standard deviation and Z is the gaussian normalizing constant which I'm too lazy to write out since it vanishes soon anyway. For all the data, you multiply the probabilities:
P(d|x,a) = P(d_1 | x_1,a) * P(d_2 | x_2,a) * ...
You want to find the a that maximizes that. log is monotonic, so if you maximize log(w) you also maximize w, so take the log, making the products sums, and making the exps disappear. Also the position of the maximum is not affected by adding a constant or scaling, so we simply remove all constant multiples and scales. That results in:
argmax_a -(d1 - f(x1;a)^2 - (d2 - f(x2;s)^2 - ...
(argmax being meaning return the a that maximizes the expression).
Well, now turn the maximum into a minimum and flip the signs and you have least squares.
You're right about the computational aspect, efficient solutions for linear least squares do exist which make it particularly appealing. It also has a single global minimum which is not only tractable but easy to find. Any power less than 1 does not in general have a tractable, guaranteed solution.
You're also right that you can use other powers, and in fact other functions completely. In general though you can substitute into the expression whatever probability distribution your data does come from. If it's not Gaussian, it will probably have fat tails and therefore the result will be robust to outliers. You can then optimize directly using some numerical method. It's best to take logs and minimize if you don't want to run badly into floating point trouble.
The easiest is to use downhill simplex (fminunc in MATLAB/octave), since it's derivative free. Or you can compute derivatives and use one of the many fine off the shelf general function optimizers. Or you can use iterative reweighted least squares, which is usually pretty efficient for least-something problems.
Now you've also wound up in the whole field of robust statistics and M-estimators.
Anyway your intuition is mostly correct here: if you have outliers you need something robust to outliers, and with the computational power we have now we can (and do) do things more involved thatn least quares, though often related.
One thing though about the best known algorithm: as your algorithms get robust you get local minima, in which case you can find a better solution given a decent starting point, but you (provably in many cases) can't guarantee you'll get the globally optimal solution.
So, to get a decent solution you need a good starting point. For that, you either need a good guess (often really not hard) or failing that, an algorithm like RANSAC is really really good.
Here's a fun addition/side/related point. Take the simplest possible version of what you have above, i.e. a slope of 1 and you're just trying to find the shift. In other words you don't expect the data to vary so you're finding just the central position in some way. With Gaussian noise (least squares) you get the mean dropping out of the squation. With a 2 sided Laplacian (exp(-|x|), rather than exp(-x^2)) you get the median. And as the power goes towards 0, you get the mode.
I submitted the paper over a *year* ago. The necessary changes were minor. But the actual time it took to go through the process was excruciating.
Apart from a few exception, yeah submitting papers takes aaageeess. Basically (and this has nothing to do with you or the quality of your work), no one wants t oread your paper. I mean people accpet they need to read and review papers as part of the academic papers, but no one WANTS to do it.
Half of it is they want to be doing their own research. The other half is that 80% of everything is crap including the papers they have to review. Not only that, some papers are great and it's an easy accept. Some are terrible and it's an easy reject. The rest are kind of shitty and you have to put a lot of effort in reading a shitty paper to decide which side of the line it falls, especially to reject since you should give reasons.
When your paper lands in their inbox, there's an 80% chance it's one of those.
So reading papers is time consuming so the journals give people plenty of time to read them, say, 6 weeks. Then inevitably what happens is the reviewer in question will not get to it during then time and will find a few hours after the second reminder. It's embarrassing after that.
Then the journal must let it compost for a while.
Then the unpaid editors have to process that, check the reviews, and decide what to do etc etc.
Why would convicting someone of something without bothering to look into the facts be doing him a favor?
The law is written that way precisely to stop dickheads from pulling this shit then weaseling out of it. When you get behind the wheelof a car you are a danger to others. Take some responsibility.
More important, be deaf and it is legal. How can not being able to hear be a crime if not being able to hear is optional to get a driver's license?
Given your posting history I'm pretty sure this is a rhetoical question, not a genuine one.
Here's a non rhetorical question: do you really not understand the difference?
Oh wait, there's about a bazillion others, and the Democrats are on the wrong side of most of them.
Like what? Interesting that you didn't bother to mention any of them...
This article can be summed up with "company respects sovereignty of a country, and I'm outraged!"
This article can be summed up as "company makes lots of money supporting the censorship of a repressive, autocratic and deeply unpleasant regime."
"Obey our laws or get out of China." It's just that simple.
Yes and the ethical thing is to just get out.
TFA makes it sound like they have some magical 3rd choice and they're being evil for not "doing the right thing".
Not really. They're pointing out the activity. You can make whatever judgement you like about that.
a very small price to pay for them to stay in the market
Yes you can make lots of money doing deeply unethical things.
I am sure also apple music Germany is censoring selling pro nazi song sales. We may not like the law of *their* country, but apple has to respect local laws *where ever* it sells.
Yeah but Germany is basically a democracy with free citizens. China is a repressive autocracy that harvests organs from political prisoners.
Supporting one is not the same as supporting the other.
If they don't want to respect the law of a country, the solution is simply : don't sell to anybody in that country.
That does seem to be what people are indeed advocating. Don't lend support to the repressive organ-harvesting regime even if there's lots of money to be made.
Well this is a substantially different claim from the original one, and I agree more with it.
But for a majority of tasks that programmers used C++ for earlier, Java has proven to be an effective and easier replacement.
Yesbut...
I mean yes, for plenty of stuff Java is a more appropriate choice than C++. On the other hand if anything C++ is growing now, and a lot of migration was a long time ago. The world of C++ now is nothing like the world of C++ in 1996 when Java first stormed onto the scene.
Obviously, no one actually wants to use something like Hot Java
Wow that brings back memories. I've not heard that name in a VERY long time!
Some of my colleagues work with image processing (research). They all work in MATLAB, most of the time. Obviously, it is glacial as one would expect a dynamically typed language to be. But that isn't the point when you are exploring transformations. This is the expensive part (human time). Once the algorithms stabilize, they are handed over to C++ engineers for final implementations.
I use both. I've made prototypes in both MATLAB and C++, and continue to do so. The thing that sets C++ aside is it's a much more flexible language than MATLAB, it's sometimes referred to a a libary writer's language.
Writing good libraries is hard, very hard, and few people can do it. But those libraries can be truly excellent nd act almost like an EDSL. MATLAB doesn't have that. It's basically an array processing language with an emphasis on linear algebra, and it always lookslike that.
For example despite it's mathematical pedigree, MATLAB still doesnt have autodiff. If your image processing step involves any optimization it's much quicker and easier if you don't have to compute gradients.
Likewise the lack of strong typing is a bit of a pain. It means I have tofind bugs at runtime.
Performance ("Web browsers and compilers"), not expressiveness.
I think you're stretching the point somewhat. It's not easier to write a web browser in Java is the resulting web browser is not useable. What's the point in doing three quarters of a task?
It would generally be more elegant and easier to express the Eigen, Ceres, Armadillo code in MATLAB, Mathematica or a similar language.
I know MATLAB and no it wouldn't. MATLAB for example doesn't give you static typing on, well, anything. Eigen and similar libraries give you static typing on matrix sizes, eliminating a whole slew of things at comile time.
And with matlab,to get anything other than truly glacial performance, you haveto use MATLAB idioms which involves cramming eveything into an array expession which is generally harder to read, write and debug than for loops.
MATLAB has it's place and I like it a great deal, however, for many tasks I reach for C++ first.
Java is generally much easier than C++, unless you need memory control.
Depends on the task. I've worked on image procesing code in both C++ and Java. It's a lot easier in C++.
When you need that flexibility in Java, you pull out Scala
That's not Java. That's a language running on the JVM but it certainly isn't Java.
I translated "mind-numbingly" to "utter garbage", an exaggeration, I agree.
So you translated one thing to something completely different. It would make sense if the ISO's job was to entertain. But it's not. Boring/exciting is not within the remit of the ISO and frankly is not an issue for standards.
Poor technology documentation is a HUGE problem. "Mind-numbingly" is just one of the problems with poor documentation.
Lack of precision for for a standard is a bigger problem then being boring. Much bigger.
You're lumping in user guides, getting started tutorials with standards, which is frankly facile. The goal of a standard is not to entertain, it's so that two gery techincal groups can without talking to each other make two implementations of something which behave in a consistent manner.
A standards documentation is needed so that I can determine if struct {int a;b;} a; &(a.a)+1; is valid.
Also, personal attacks are unhealthy for the attacker.
So why did you do it then? You started off on the attack about a group of people some of whom I've worked with.
The C++ standard is UTTER GARBAGE, says ISO CPP
You're a fuckwit says you.
I mean that statement has about as much logic as your post.
Unfortunately, the C++ standards committee is philosophically opposed to subsetting,
Not when it comes to the library they're not. There's currently non hosted (no library) and hosted (with the library). There are proposals making their way through to make the non hosted one have some useful libraries.
I don't think any one would suggest that.
You got awfully close. You said:
but it is much easier to write Java than C++ for all but the simplest programs.
So I don't really get your point.
There's plenty of stuff written in C++. If it's much easier to write programs in Java, then why are major bitsof infrastructure (like Web browsers and compilers) generally written in C++? Not to mention games, numerical computation, embedded stuff, and so on.
So I'm not sure what your point is. There are plenty of nontrivial programs for which C++ is the correct choice.
This sounds like a parody except I know you're a fuckwit and actually believe this shit.
but it is much easier to write Java than C++ for all but the simplest programs.
I strongly disagree with that.
I'm not a fan of java, but I'm not a hater of it either, from my point of view, it's solidly OK. But there's no way java is better for all programs than C++. C++ allows for much more expressive libraries than Java and if someone's put in the work that makes code much much easier to write.
Take Eigen for example. Not a library you want to delve into, but as a user, you can write code that looks like maths. Or even moreso, take Ceres. You write maths just like normal and it automatically differentiates the expressions.
For certain applications there are nice fixed point libraries.
Image processing is miserable in java. There are many reasons but the lack of unsigned bytes is really super annoying
That's all on the computation end, but that's my field at the moment. Java's just not a good fit, and it's simply easier in C++.
clang and VS have imlemented it, albeit at the moment two different versions of the draft standards. GCC has a WIP implementation that's woking but not complete.
C++'s days are numbered.
Techincally everything's days are numbered.
"The designers were primarily motivated by their shared dislike of C++."
Euch. You could insert "irrational" or "misinformed" into that sentnce. I use C++ a lot. I could give you a long list of complaints about it. A very long list. There are things I'd love to see fixed, or a "better" language replace it.
That language isn't golang.
In fact that article you posted:
"Less is exponentially more": http://commandcenter.blogspot....
Basically degenerates into a whiny rant about go is too awesome for C++ programmers and that's why C++ programmers have been pretty uninterested in go. It seems he can't really cope with the possibility that go isn't that amazing and certainly doesn't cover as many use cases as C++.
"Why ESR Hates C++, Respects Java, and Thinks Go (But Not Rust) Will Replace C"
Yeah well, ESR is a bit of a plonker. What are you going to post next? His guide to sex? (no realy this is a thing http://www.catb.org/esr/writin...)
You do see the flaw with that, I would hope. As long as the political narrative is acceptable, or you didn't give someone booboo feelings you would get paid. Deviate from the straight and narrow, and you don't.
That's one of the most warped and entitled things I've ever read. No, nothing at all gives you a RIGHT to get paid for your speech.
I thought you far left people thought that companies are pretty amoral.
Yes well you would think that wouldn't you. You have no idea wht "left" even is, so you ascribe some mishmash bag of things to it then convince yourself that anyone who doesn't agree with you on everything must therefore believe all the things in your bag.
They need to use demographics, just like regular Television has so successfully. Takes a bit of work, but not that hard.
No you don't get it. It's their money, they don't need to do anything.
Why do you keep telling other people what to do with their money? Are you a communist or something?
This everybody should code fad is one of the dumbest things I've ever seen.
why stop there? I think teaching reading and writing is a fad.
maths too.
I think what we should do is thoroughly analyze every 4 year old, determine their vocation then train them exclusively for that ignoring everything else.
And do you know what else is a dream job? Being a financial advisor to someone with a crapload of money and a 9th grade education
*blink*
um
I don't think "financial adviser" is ANYONE's dream job. Well it might be but then you have the dullest dreams in the world. You could couple your dream job with a dream house of a moderately sized place in suburbia with upper-end Ikea furniture (not the cheap stuff).
I'll keep working on it for now, but if they screw me over again I might have to bail out.
I don't get it. Before they moved the goal posts on monteization, you racked up 43 cents in 2 years. It sounds like you're in it for the fun of making videos not the money so why do you care if they move the monetisation goalposts?