Libertarianism is about maximizing personal freedom of
everyone not just corporations. Do you see most companies doing this?
Libertarianism is about not allowing the government to restrict
your freedom. Companies are free to do what they want. Presumably,
the customer can choose a different company if he does not like how he
is treated...
Next up, that rate is determine by what people will pay
I assume these people shop around, so it's the competition that
determines the price.
everyone has a budget to stick to - that means the guy paying for a
building has constraints which don't allow new construction at the
rate people want to be paid for it, so it either doesn't get done or
there is some give.
It gets hard to tease out all the competition, but if someone already
owns the land, he's going to try to get as low a price as possible for
the construction even if he can pay more and be profitable.
Basic economics, supply vs demand, but there is a basis to both sides
beyond the theory of it - the supply (labor) wants all they can get
and the demand (client) wants all they can get, for the former that's
a direct translation to wealth and for the latter it's a budgetary
constraint saying how much they can build for x amount.
Maybe one should apply more sophisticated economics. Can this be
modeled with game theory? The workers are in some type of prisoner's
dilemma. They need a job so they end up depressing their wages and
don't benefit from most of the increases in productivity. Of course,
the owners are in a related situation but their are fewer players, and
they have more resources, so perhaps they do better...
But honestly, I find many of these arguments (including my own)
suspect. Economics has little interesting predictive power. It
smells a bit of pseudo-science. The other issue is that economics has
a large engineering component. In other words, just because things
are this way with the current rules doesn't mean they have to be this
way. However, because we have such limited understanding of this
system, it has proven difficult to control.
Assume our main goal is that everyone gets a living wage.
First assume we don't have minimum wage laws. Therefore the
government needs to make up the difference. (This is a bit how it is
now with subsidized housing and food stamps.) This seems like a
reasonable system. If we did have a minimum wage then it might be too
high for full employment and the government would have to fully pay
for this living wage. Therefore the companies are offsetting this
cost. Of course, with this system, what incentive do these workers
have to get a job. If this was a free market then a company would
need to pay above the cost of living to get people to work, probably a
lot above since getting a few extra bucks is not going to entice
someone to work 8 hours a day. Of course, if the company couldn't
afford the cost of living, it can't afford more, so these jobs don't
exist. Therefore, this system doesn't help.
Next assume, we do have minimum wage laws that are somehow pegged at
cost of living. Is this much different? Not really. Still there is
little incentive for people at this wage to work. Perhaps we have
minimum wage at something like twice the cost of living. This is less
clear. A business might not be profitable at the higher wage and
people might be willing to accept the lower wage. (The only upside is
we would be making it harder for companies to not share the wealth and
perhaps decrease inequality for some at the potential cost of pushing
some people back to cost of living and increasing their inequality.
Assuming reducing inequality was a goal, the details could be decided
based on employment data.)
Last assume, we have universal basic income. (Tying it cost of living
is an interesting problem.) This looks like a good solution. There
is no wall to cross for entry level jobs. Any money a job offers you
is money in your pocket. One can think of this as a way to
redistribute wealth. Image a closed system where everyone got an
equal share of 10% of the pie and the rest was distributed based on
value added to the economy. To me this seems like a reasonable
trade-off.
Here is an interesting thought experiment. Assume our main goal is
that everyone gets a living wage.
First assume we don't have minimum wage laws. Therefore the
government needs to make up the difference. (This is a bit how it is
now with subsidized housing and food stamps.) This seems like a
reasonable system. If we did have a minimum wage then it might be too
high for full employment and the government would have to fully pay
for this living wage. Therefore the companies are offsetting this
cost. Of course, with this system, what incentive do these workers
have to get a job. If this was a free market then a company would
need to pay above the cost of living to get people to work, probably a
lot above since getting a few extra bucks is not going to entice
someone to work 8 hours a day. Of course, if the company couldn't
afford the cost of living, it can't afford more, so these jobs don't
exist. Therefore, this system doesn't help.
Next assume, we do have minimum wage laws that are somehow pegged at
cost of living. Is this much different? Not really. Still there is
little incentive for people at this wage to work. Perhaps we have
minimum wage at something like twice the cost of living. This is less
clear. A business might not be profitable at the higher wage and
people might be willing to accept the lower wage. (The only upside is
we would be making it harder for companies to not share the wealth and
perhaps decrease inequality for some at the potential cost of pushing
some people back to cost of living and increasing their inequality.
Assuming reducing inequality was a goal, the details could be decided
based on employment data.)
Last assume, we have universal basic income. (Tying it cost of living
is an interesting problem.) This looks like a good solution. There
is no wall to cross for entry level jobs. Any money a job offers you
is money in your pocket. One can think of this as a way to
redistribute wealth. Image a closed system where everyone got an
equal share of 10% of the pie and the rest was distributed based on
value added to the economy. To me this seems like a reasonable
trade-off.
Sorry I never got back to you. I wanted to look over the material you recommended but then I got busy with other stuff. By the time I got back to the old post they had closed the article for comments. Anyway, send me an email, if you want to hear my response. Thanks.
There are even polynomial-time approximation algorithms that guarantee
to be within a fixed percentage of optimal.
The percentage for an arbitrary graph is still quite poor. It's still
about 1.5 times longer than optimal. Euclidean is much better and this
problem probably qualifies...
And don't forget that NP-complete doesn't mean it's impossible, only
that it takes a long time as the problem size increases. For smaller
problem sizes, solving the problem outright isn't impractical. In this
case, they have a variation on the classic problem where there's a
limit to the number of kids on each bus and a limit to how long (in
time) each bus route can be. But if the stops are fixed and the school
to which they have to deliver the kids is fixed, then it breaks down
the problem to separate problems for each school, and solving each
small NP-complete problem is entirely practical.
Right, there are good algorithms for solving moderately big TSP
problems such as https://en.wikipedia.org/wiki/....
However, all these extra constraints and changes might cause a lot of
problems. I'm sure the MIT guys did a lot of significant work.
Regarding the laser microphone : If this device emits in the inaudible
part of the spectrum, how can it jam speech ? The most obvious
counter-counter-measure, would be to clean the sound spectrum from the
laser, amplifying the frequencies mostly found in human voice, and
masking frequencies out of this frequency-band (thus masking all
frequencies beyond hearing range but still somewhat picked up by the
mic).
I don't know much about this stuff, but maybe the window vibration is
non-linear and can create new frequencies in the audible bands.
Or is the anti-laser counter-measure banking on output so much noise
(out of spectrum sounds), that the signal (speech) is swamped by the
jamming, and the laser mic mostly picks up noise. i.e.: make the
signal/noise ratio suck a lot hoping that the mic won't pick up the
interesting (voice) frequencies against the ultra loud background. so
after frequency isolation, the resulting voice would be completely
distorted due to poor mic response.
So the out of spectrum is so large in some frequencies that the whole
device is screwed up (non-linear)? I guess this could be in the
window or the mic. (The window really is acting as part of the mic.)
(Though signal processing science tells us that if the analog stage -
the lase part of the mic - is able to pick higher frequencies than the
digitizer - which probably could be limited to hearing range - the
signalling will be strongly distorted in the hearing range - which
also the single reason of existance of 192kHz ADC)
One would always use a filter. 44 kHz CDs use a filter to avoid this
problem. I thought the main reason for using higher sampling rates is
because the required filters for 44 kHz have effects in the audible
range. (4 kHz is not enough of a buffer, but most people can't hear
to 20 kHz anyway.)
Regarding the hearing loss : How the hell is hearing loss possible
with a out-of-spectrum noise ? By definition sound outside the
audible spectrum is sound outside the frequency response of the human
auditory system. I.e.: frequencies which physically can't interact
with the inner ear. By which mechanism could they still harm the
hearing if we can't hear them ? (= if they can't manage to interact
with the inner ear) By damaging the middle ear (the small bones that
work as transducers to transmit sound to the inner ear ?)
I think you are right; these hypothetical sounds are going to interact
with some part of the ear. I guess it's possible they damage it.
With many unlimited data plans, one gets fast speeds for around the
first 22 GB. That means at 10 Mbps, one gets full speed for about 5
hours a month. (The contract is probably worse if you use a hotspot.)
That should be enough for anyone.
The real answer is that there is some human who decided
whether a video is extremist, then gave it to both the AI and another
human, and the AI was able to learn how to agree better with the 1st
human. Unfortunately, that doesn't actually tell us if the video is
extremist or if the AI is any good, because the first human isn't more
right than the second.
You're right. The issue is how do we get a more accurate test set
to evaluate the algorithms. Typically the test set has the same
biases and accuracy as the training set. (Often it is generated by a
somewhat random split of the original labeled data.) Even if the
algorithm is more accurate than a single human labeler, we can't know
without something more accurate than a single human.
Fortunately, we can often get a test set that is more accurate
than a human. In this particular case, as others have mentioned, you
might have different levels of humans such as experts who will label
the test set. Another option is that we might combine the labels of
humans to get a consensus. (For many problems, a combination of
labels will be more accurate than a single human.) In the ideal case,
we can get the true label. Typically these problems involve
predicting the future. For example, we know if it rained on Saturday,
but we want to predict that on Friday.
Of course, I've sidestepped the issue of how we know our new test
set is more accurate. This can be hard to prove. For human experts,
we can verify they are making different predictions than normal
humans, but we just assume they are better. Same for a voting
consensus of humans. If we have "true" labels, then I guess it's pretty
clear...
Another issue is why don't we use this more accurate labeling
procedure to create the training data. This is related to cost. It's
expensive to hire human experts. Of course, if it's worth it then we
might spend the money. For example, maybe we get multiple people to
label the same examples to produce better training data. (This becomes
an optimization problem since for a fixed budget, we can either have
more noise but more labeled examples or less noise but fewer labeled
examples.) Another example is based on computation. Perhaps we can
get accurate labels if we use an expensive algorithm. We can use this
to train a classifier that can be designed to be much faster.
While I appreciate an intuitive explanation, I don't think it's as
easy as you suggest, especially for someone practicing in the field as
opposed to someone doing homework.
So to see if the 15 out of 20 is significant, we can compare this
outcome to random chance.
So the null hypothesis is a 50/50 coin flip with IID Bernoulli
trials. Let's say the result of the experiment is R L L R L L L L L L
R R L L L R L L R L? What is the probability of this data given the
null hypothesis is true? Well the probability is 2^-20. You really
need to justify that the significant result is that I get 15 or more
Ls. I'm not sure of a great way to pick this in general particularly
as the experiment gets more complex.
Also, you to specify what is significant before you run the
experiment. One can bound that area to have p=0.05. Of course, just
because the experiment is not significant does not mean the the data
was generated at random. Assume I ran the experiment and I get R L R
L R L R L R L R L R L R L R L R L. Any human is going to look at this
an think something strange is going on. Perhaps the experiment is not
IID. One probably needs to run more trials and use some of the data
to formulate a model of what is generating the data to construct the
significant region, but even that has issues...
There are other problems if you have a limited amount of
data. Let's say you get 17 lefts. Can you claim a p value of.0013?
I don't think you can change your significance space after the
experiment, otherwise it would be easy to cheat particularly as the
experiments get more complex. Not sure of the best way to resolve
this. Probably need to consider sets of significance sets where each
one has a different p value. This has it's own problems.
Anyway I'm not a statistician. I'm sure these things have good
answers, but I don't think it's so simple...
And of course, if your degree requires you to publish, or your tenure
is based on your publishing history, there are ways to adjust the
results to make the significance more likely.
Even if everyone is completely honest and knowledgeable, because of the
publication bias, the significance is way off. Just keep doing
experiments and eventually you will get a "significant" result. If
we knew the number of experiments done, we could control for this
problem, but in most fields this is impossible. Imagine a field where
all the ideas are crap. If enough people work in the
field, you will still get a lot of significant results. Automation
and computer analysis might be partially driving this problem by
allowing more "experiments" to be performed.
According to conventional wisdom, ideally, scientists should strive
for a power of about 80% (i.e., an 80% chance of detecting an effect
if it truly exists), but very few studies actually achieve power of
this level. In many fields, the power is less than 50% and sometimes
much less.
How does one control this in practice? Let's say I want to compare
means. Do I need to assume the difference is greater than X to lower
bound my power (modulo a bunch of assumptions.) I don't really see
this in practice. Is it implied by other details of the experiment?
Somewhat counterintuitively, underpowered studies are often also more
likely to result in FALSE POSITIVES. This is because, when your power
to detect a true effect is low, and if you test a large number of
effects that are unlikely to be null, most of the hypotheses that you
say are "significantly" non-null will actually be false positives. We
would say that the "false discovery rate" tends to be very high when
the power is low.
I don't understand. Yes if you do a large number of tests than you
are likely to get more false positives. How is this connected to the
tests being underpowered? Are you claiming that we will do more
experiments/tests when things are underpowered? I think the details
would be hard to quantify.
What we should be doing is measuring and reporting effect sizes along
with their credible intervals. While using priors that are based on
our real state of knowledge. In other words, we should be doing
Bayesian statistics.
I do agree that giving things like confidence intervals would be more
informative. I'm not sure about using Bayesian statistics. It does
make the results more intuitive but at the cost of potentially adding
uninformed/convenient/manipulated priors. I assume, if there is
enough data to make the results insensitive to the priors then it's
probably equivalent to a frequency based approach. (It would be nice
to see a counterexample to this claim.) Personally I would be curious
if bootstrapping approaches could be used. This would remove a lot of
bad parametric assumptions, and we've got the computational power to
compute them...
I can make even very small differences in the means appear
statistically significant.
Small differences in the mean can be statistically significant. And
yes you need large sample sizes to show this. Not sure what you mean
by appear...
A better approach is to use that variable as a predictor and have two
data sets -- a training data set and a testing data set. I then
calculate a function to classify storms based on the training data
set, using the variable as a predictor of whether a storm will be
tornadic or not.
I wouldn't say this is a better approach; you are trying to answer
a different (and probably more interesting) question. Your
conditional variable might have a high variance with a lot of overlap
between tornadic or not. This will create a poor predictor. Doesn't
mean that the means of the conditionals are not different.
For showing a difference in the means, giving confidence intervals
is probably a more intuitive presentation of the results. Instead of
saying, with 0.05 confidence the means are different, one should give
the confidence intervals that jointly hold with 0.95 confidence. This
way one can easily see the size of the difference. Of course, one
will lose statistical power with this technique, so it will be harder
to hit the magic 0.05. Also, people will be less impressed if it's
obvious the difference is small. I wonder what trade-off people would
make to create a convincing looking pair of confidence intervals, and
how their inevitable playing with the parameters to create a nice
picture would bias the results...
You need to check your facts. Government per-student expenditures in
the US are near an all-time high, and are higher than in almost all
other developed nations.
I think he was talking about college, and I read some articles
that claimed the same thing. If true, it explains the increase in
cost of state universities. The other factors often mentioned, while
real, are not as significant. (Administration, facilities,...) As
for Europe, I'd be interested in reading evidence about their lower
costs. Do you have any references?
Right or wrong, I think the subsidy is often considered the exploitation of the externalities such as pollution. This is part of the rational for giving cleaner tech subsidies, to balance the playing field. While I think it would be more honest to remove the externality, it's probably easier to give the clean technology the subsidy.
The more a government interferes with the marketplace the worse
economies become. Evidence for this are the countries such as the
Soviet Union, Cuba, Red China, North Korea, etc.... These countries
absolutely tied the hands of business men. These governments made all
decisions of an economic nature. What was the result? Their citizens
died of starvation by the millions, and when they didn't absolutely
starve they lived in 3rd world conditions.
As I mentioned, the government creates the marketplace, so its got to
do some "interfering". Where does one draw the line? This is where
your first principle against coercion breaks down. You need to add
some other principle or else you become an anarchist. In other words,
the government needs to coerce, so you need some principle to tell it
when to stop. Clearly those dictatorships went too far (though it's
really not one-dimensional and there are other issues besides their
economies). Currently Scandinavian countries have more coercion than
the US and they do well in "happiness" metrics...
The reason for this is that no one person, and no small group of
people, can understand all the economic factors involved, nor the
economic desires and needs of people that they cannot know personally,
nor can they respond to conditions at the speed that millions of
people making their own decisions can respond. That means things get
out of whack in a hurry in a centrally managed economy.
Just because an economy must be engineered does not mean it needs to
be centrally managed. Engineering can include controlling a fiat
currency, patents, monopoly rules,...
The central banks and the governments can't figure out what is going
wrong so they do things like increased deficit spending. huge stimulus
packages based entirely on debt, and manipulation of the currency by
printing money and very loose credit policies.
I agree that economics is a poor science. Currently, it's a difficult
system to model, but this applies to both the removal of controls as
well as the application of controls. Also, in anticipation, I would
disagree to the claim that a "simpler" economic system is easier to
understand. In principle, various controls can be put in place to
stabilize the system (probably at some cost, but this is an expected
trade-off).
A good small book on central banks and the harm they do, in the page
range you were asking for, is "A History of Money and Banking in the
United States". Google it and you can find it for free in ebook form.
This clocks in at 510 pages, so it's not small. It looks good since I
do have some interests in the history of the fractional banking system
and central banks. Still, I doubt he has much in the way of
empirically justified solutions as he seems pretty extreme free market
and those extreme ideas have never been tested.
Also, the countries in my first paragraph were very involved in the
redistribution of wealth in their societies. Think about that for a
while and what kind of economic condition they ended up in. Their
dislike of the wealthy business owner destroyed their own economies,
and plunged their citizens into untold economic misery.
Yes, it is interesting to look at the failures and successes of
various countries and try to determine causes.
Also, notice that when economic freedom disappeared so did civil and
religious liberty. Economic freedom is the foundation upon which all
other freedoms are built, even Karl Marx acknowledged this. What he
didn't understand is that when you remove economic liberty you destroy
the other liberties too.
This gets back to the libertarian book I read. They claimed economic
and social freedom were orthogonal. Typically this is used to
contrast the Republican versus Democratic party. As for your
examples,
There is a role for government to play. It needs to make
sure that the rights of all of its citizens are respected. As an
example, the federal government needs to protect the country from
outside aggression
But doesn't the government have to coerce people to do some of these
things. I think you would agree that it boils down a thoughtful
selection of what the government is allowed to do.
Government redistribution of wealth is the antithesis of liberty.
I can see your point, but you do realize that it's always going to
happen. Some people will benefit more from "essential" government
services then others. Of course, that doesn't mean one can't try to
minimize this redistribution.
I would also like to make a bit of a retraction. I don't think
the current taxation system is fair under any reasonable definition of
the word, but I do think this redistribution might be necessary. Even
a flat tax is probably not fair. Why should people pay a percentage
of their income when they receive a fixed set of benefits? (I can
argue this a bit in terms of the many disproportionate benefits of the
wealthy, but this still doesn't cover many aspects of wealth
redistribution to the poor.) However, I would argue that a fair
system would lead to a very bad social outcome.
I don't think it would be hard to justify a reasonable economic
simulation that would show that true economic freedom would lead to an
extreme concentration of wealth. (Even more extreme then what we are
currently seeing.) Is this good? I would claim it is bad for a range
of reasons. I would also claim that a modern economy is an engineering
problem. Perhaps we could modify the economy is ways that get a
better social outcome without restricting liberty, but I'm OK with the
government imposing some restrictions, just like they do in other
"essential" areas.
To understand my point of view on liberty I would recommend Freiderich
Hayek's book "The Constitution of Liberty", as a format like this is
probably the worst possible place to have a real discussion of such
complex issues.
It's a bit long for me. (Lots of other things in the pipeline.) I do
remember reading a short Libertarian book many years ago that I found
very enlightening. It was around 50 pages. Do you have a reference
in the 20-50 page range?
The healthcare issue is the issue that is at the very heart of the
matter right now, at least for me. Why? Because Obamacare and single
payer healthcare both require government coercion to work. You take
away coercion and it all falls flat on its face.
Yes it does involve coercion, but I'm OK with that in essential areas
where the market fails. For example, the only reasonable type of
health insurance is one that covers chronic conditions. Before
Obamacare, in many states there was no way to get those types of
guarantees, so I consider this a market failure. For this, and other
reasons, the government needs to step in.
I must congratulate you on your attitude.. We disagree on some pretty
big issues, but you have not been disagreeable. This is pretty rare in
my experience online. Thank you for your good attitude.
I do think you have a reasonable position based on your main
principle, and you seem to live by that principle. A reasonable reading
of the Constitution might also be on your side. However, I'm pretty
stubborn, and I'm going to stick with an, as yet undefined, notion of
social good over individual liberty.
What I do right now is get my insulin from the manufacturer on their
low income plan. If your income is above medicaid elgible yet below
$40K for a married couple they will supply you for free.
So you survive based on the generosity of a corporation. Not a
position I would want to be in.
When I suceed I don't want the government taking from me to give to
someone else who may or may not give a rip about standing on their own
two feet and supporting themselves.
While you have many legitimate grievances, I would say this is your
core issue. Personally, I want the government to do some wealth
redistribution. For a range of reasons, I believe our current system
is unfair. Of course this is a democracy, and I respect your right to
fight for your beliefs and argue for the government you want.
It's a soul destroying idea that everyone else, the government, is
responsible for the success or failure of the individual. To make
people dependent upon government destroys ambition, drive,
self-respect, and much more.
Almost everyone is dependent on the government. At a minimum, the
government needs to do things like create a market, setup a system of
courts, protect the nation from hostile governments,... Of course,
this is not what you mean, but I want to stress that the debate is not
about the government being involved it's about where and how much.
But, and it is a huge but, I would not help the guy that didn't want
to work but could. He was on his own. If all he wanted to do was get
drunk and get high, let him get that way, but I was never going to
help him finance his self-destruction. I would feed a hungry man but I
wouldn't just stick money in his hand to go waste on something other
than food.
Again, I think this is the key issue. I see right leaning people
who feel so strongly about this that they are willing to pay more to
punish people than the cost of "successfully" subsidizing their
behavior. At some level this make sense; you do not want your money
to be used in a way that you feel is harming these people.
This is a complex issue, but I don't think it strongly relates to
healthcare. Affordable access to healthcare is not going to finance
someone's self-destruction. Perhaps it will not make him get a job,
but for an individual in this position, lack of healthcare is also
unlikely to make him get a job. As in the rest of the first world, I
would suspect it will just become part of society and not have much
effect on ambition or drive. In fact, it might have the opposite
effect. If workers were not reliant on jobs for health insurance then
they might be more willing to go into business for themselves and
become true capitalists.
Still I have to think your cost would be even higher without
Obamacare. Do you have a death wish?
Even a gold plan would leave me with more than $500/month
in insulin costs.
Don't you hit your maximum out of pocket expense limit.
the VA is a typical US government intrusion into
healthcare. It is inefficient, corrupt, and next to impossible to
change.
As can be seen in the above article, private healthcare also has
problems. I believe that single payer works in many countries. I
don't believe we can dismiss all US government run programs.
He worked for the VA for more than 15 years and never once did he
actually earn his wages.
I've heard this about some government jobs. Is it also true for
contractors? What about medicare? (Yes there is fraud, but you can't
compare against Utopia.)
Since the federal government already borrows $4 out of ever $10 it spends just where do you see the money to pay for a single payer system coming from? And just how sustainable are the federal government's current spending habits, let alone with a 75% increase in federal spending?
So what happens to all the money that is currently spent on health care? Maybe we use that money...
So why can the governments of other countries pay for single payer? (The British are pretty happy with their health care, and your anecdotes on the British health care system make you sound like an astroturfer.) Also, why don't the ACA subsidies lower your costs? A simple search online suggest you should be paying around $100 a month for a silver plan.
In our last election, rural voters preferred Trump and that is why the
rural voter trumped the urban voter to override the popular vote.
So you're saying if we got rid of the extra 2 electoral votes per
state then Trump would have lost. This is false. Hillary won about
20 states to Trump's 30. Subtracting out the extra two votes gives
Trump 246 and Hillary 192.
The real issue was the winner take all part of the system. With
winner take all it is mathematically possible to win 25% of the
popular vote and still win the election. With the extra 2 votes, it's
even worse.
Slashdot Mirror
There is some evidence that neurons are not important as people generally believe. https://www.scientificamerican...
Libertarianism is about not allowing the government to restrict your freedom. Companies are free to do what they want. Presumably, the customer can choose a different company if he does not like how he is treated...
Citations needed.
I assume these people shop around, so it's the competition that determines the price.
It gets hard to tease out all the competition, but if someone already owns the land, he's going to try to get as low a price as possible for the construction even if he can pay more and be profitable.
Maybe one should apply more sophisticated economics. Can this be modeled with game theory? The workers are in some type of prisoner's dilemma. They need a job so they end up depressing their wages and don't benefit from most of the increases in productivity. Of course, the owners are in a related situation but their are fewer players, and they have more resources, so perhaps they do better...
But honestly, I find many of these arguments (including my own) suspect. Economics has little interesting predictive power. It smells a bit of pseudo-science. The other issue is that economics has a large engineering component. In other words, just because things are this way with the current rules doesn't mean they have to be this way. However, because we have such limited understanding of this system, it has proven difficult to control.
Assume our main goal is that everyone gets a living wage.
First assume we don't have minimum wage laws. Therefore the government needs to make up the difference. (This is a bit how it is now with subsidized housing and food stamps.) This seems like a reasonable system. If we did have a minimum wage then it might be too high for full employment and the government would have to fully pay for this living wage. Therefore the companies are offsetting this cost. Of course, with this system, what incentive do these workers have to get a job. If this was a free market then a company would need to pay above the cost of living to get people to work, probably a lot above since getting a few extra bucks is not going to entice someone to work 8 hours a day. Of course, if the company couldn't afford the cost of living, it can't afford more, so these jobs don't exist. Therefore, this system doesn't help.
Next assume, we do have minimum wage laws that are somehow pegged at cost of living. Is this much different? Not really. Still there is little incentive for people at this wage to work. Perhaps we have minimum wage at something like twice the cost of living. This is less clear. A business might not be profitable at the higher wage and people might be willing to accept the lower wage. (The only upside is we would be making it harder for companies to not share the wealth and perhaps decrease inequality for some at the potential cost of pushing some people back to cost of living and increasing their inequality. Assuming reducing inequality was a goal, the details could be decided based on employment data.)
Last assume, we have universal basic income. (Tying it cost of living is an interesting problem.) This looks like a good solution. There is no wall to cross for entry level jobs. Any money a job offers you is money in your pocket. One can think of this as a way to redistribute wealth. Image a closed system where everyone got an equal share of 10% of the pie and the rest was distributed based on value added to the economy. To me this seems like a reasonable trade-off.
Here is an interesting thought experiment. Assume our main goal is that everyone gets a living wage.
First assume we don't have minimum wage laws. Therefore the government needs to make up the difference. (This is a bit how it is now with subsidized housing and food stamps.) This seems like a reasonable system. If we did have a minimum wage then it might be too high for full employment and the government would have to fully pay for this living wage. Therefore the companies are offsetting this cost. Of course, with this system, what incentive do these workers have to get a job. If this was a free market then a company would need to pay above the cost of living to get people to work, probably a lot above since getting a few extra bucks is not going to entice someone to work 8 hours a day. Of course, if the company couldn't afford the cost of living, it can't afford more, so these jobs don't exist. Therefore, this system doesn't help.
Next assume, we do have minimum wage laws that are somehow pegged at cost of living. Is this much different? Not really. Still there is little incentive for people at this wage to work. Perhaps we have minimum wage at something like twice the cost of living. This is less clear. A business might not be profitable at the higher wage and people might be willing to accept the lower wage. (The only upside is we would be making it harder for companies to not share the wealth and perhaps decrease inequality for some at the potential cost of pushing some people back to cost of living and increasing their inequality. Assuming reducing inequality was a goal, the details could be decided based on employment data.)
Last assume, we have universal basic income. (Tying it cost of living is an interesting problem.) This looks like a good solution. There is no wall to cross for entry level jobs. Any money a job offers you is money in your pocket. One can think of this as a way to redistribute wealth. Image a closed system where everyone got an equal share of 10% of the pie and the rest was distributed based on value added to the economy. To me this seems like a reasonable trade-off.
Sorry I never got back to you. I wanted to look over the material you recommended but then I got busy with other stuff. By the time I got back to the old post they had closed the article for comments. Anyway, send me an email, if you want to hear my response. Thanks.
The percentage for an arbitrary graph is still quite poor. It's still about 1.5 times longer than optimal. Euclidean is much better and this problem probably qualifies...
Right, there are good algorithms for solving moderately big TSP problems such as https://en.wikipedia.org/wiki/.... However, all these extra constraints and changes might cause a lot of problems. I'm sure the MIT guys did a lot of significant work.
I don't know much about this stuff, but maybe the window vibration is non-linear and can create new frequencies in the audible bands.
So the out of spectrum is so large in some frequencies that the whole device is screwed up (non-linear)? I guess this could be in the window or the mic. (The window really is acting as part of the mic.)
One would always use a filter. 44 kHz CDs use a filter to avoid this problem. I thought the main reason for using higher sampling rates is because the required filters for 44 kHz have effects in the audible range. (4 kHz is not enough of a buffer, but most people can't hear to 20 kHz anyway.)
I think you are right; these hypothetical sounds are going to interact with some part of the ear. I guess it's possible they damage it.
With many unlimited data plans, one gets fast speeds for around the first 22 GB. That means at 10 Mbps, one gets full speed for about 5 hours a month. (The contract is probably worse if you use a hotspot.) That should be enough for anyone.
You're right. The issue is how do we get a more accurate test set to evaluate the algorithms. Typically the test set has the same biases and accuracy as the training set. (Often it is generated by a somewhat random split of the original labeled data.) Even if the algorithm is more accurate than a single human labeler, we can't know without something more accurate than a single human.
Fortunately, we can often get a test set that is more accurate than a human. In this particular case, as others have mentioned, you might have different levels of humans such as experts who will label the test set. Another option is that we might combine the labels of humans to get a consensus. (For many problems, a combination of labels will be more accurate than a single human.) In the ideal case, we can get the true label. Typically these problems involve predicting the future. For example, we know if it rained on Saturday, but we want to predict that on Friday.
Of course, I've sidestepped the issue of how we know our new test set is more accurate. This can be hard to prove. For human experts, we can verify they are making different predictions than normal humans, but we just assume they are better. Same for a voting consensus of humans. If we have "true" labels, then I guess it's pretty clear...
Another issue is why don't we use this more accurate labeling procedure to create the training data. This is related to cost. It's expensive to hire human experts. Of course, if it's worth it then we might spend the money. For example, maybe we get multiple people to label the same examples to produce better training data. (This becomes an optimization problem since for a fixed budget, we can either have more noise but more labeled examples or less noise but fewer labeled examples.) Another example is based on computation. Perhaps we can get accurate labels if we use an expensive algorithm. We can use this to train a classifier that can be designed to be much faster.
While I appreciate an intuitive explanation, I don't think it's as easy as you suggest, especially for someone practicing in the field as opposed to someone doing homework.
So the null hypothesis is a 50/50 coin flip with IID Bernoulli trials. Let's say the result of the experiment is R L L R L L L L L L R R L L L R L L R L? What is the probability of this data given the null hypothesis is true? Well the probability is 2^-20. You really need to justify that the significant result is that I get 15 or more Ls. I'm not sure of a great way to pick this in general particularly as the experiment gets more complex.
Also, you to specify what is significant before you run the experiment. One can bound that area to have p=0.05. Of course, just because the experiment is not significant does not mean the the data was generated at random. Assume I ran the experiment and I get R L R L R L R L R L R L R L R L R L R L. Any human is going to look at this an think something strange is going on. Perhaps the experiment is not IID. One probably needs to run more trials and use some of the data to formulate a model of what is generating the data to construct the significant region, but even that has issues...
There are other problems if you have a limited amount of data. Let's say you get 17 lefts. Can you claim a p value of .0013?
I don't think you can change your significance space after the
experiment, otherwise it would be easy to cheat particularly as the
experiments get more complex. Not sure of the best way to resolve
this. Probably need to consider sets of significance sets where each
one has a different p value. This has it's own problems.
Anyway I'm not a statistician. I'm sure these things have good answers, but I don't think it's so simple...
Even if everyone is completely honest and knowledgeable, because of the publication bias, the significance is way off. Just keep doing experiments and eventually you will get a "significant" result. If we knew the number of experiments done, we could control for this problem, but in most fields this is impossible. Imagine a field where all the ideas are crap. If enough people work in the field, you will still get a lot of significant results. Automation and computer analysis might be partially driving this problem by allowing more "experiments" to be performed.
How does one control this in practice? Let's say I want to compare means. Do I need to assume the difference is greater than X to lower bound my power (modulo a bunch of assumptions.) I don't really see this in practice. Is it implied by other details of the experiment?
I don't understand. Yes if you do a large number of tests than you are likely to get more false positives. How is this connected to the tests being underpowered? Are you claiming that we will do more experiments/tests when things are underpowered? I think the details would be hard to quantify.
I do agree that giving things like confidence intervals would be more informative. I'm not sure about using Bayesian statistics. It does make the results more intuitive but at the cost of potentially adding uninformed/convenient/manipulated priors. I assume, if there is enough data to make the results insensitive to the priors then it's probably equivalent to a frequency based approach. (It would be nice to see a counterexample to this claim.) Personally I would be curious if bootstrapping approaches could be used. This would remove a lot of bad parametric assumptions, and we've got the computational power to compute them...
Small differences in the mean can be statistically significant. And yes you need large sample sizes to show this. Not sure what you mean by appear...
I wouldn't say this is a better approach; you are trying to answer a different (and probably more interesting) question. Your conditional variable might have a high variance with a lot of overlap between tornadic or not. This will create a poor predictor. Doesn't mean that the means of the conditionals are not different.
For showing a difference in the means, giving confidence intervals is probably a more intuitive presentation of the results. Instead of saying, with 0.05 confidence the means are different, one should give the confidence intervals that jointly hold with 0.95 confidence. This way one can easily see the size of the difference. Of course, one will lose statistical power with this technique, so it will be harder to hit the magic 0.05. Also, people will be less impressed if it's obvious the difference is small. I wonder what trade-off people would make to create a convincing looking pair of confidence intervals, and how their inevitable playing with the parameters to create a nice picture would bias the results...
I think he was talking about college, and I read some articles that claimed the same thing. If true, it explains the increase in cost of state universities. The other factors often mentioned, while real, are not as significant. (Administration, facilities, ...) As
for Europe, I'd be interested in reading evidence about their lower
costs. Do you have any references?
Right or wrong, I think the subsidy is often considered the exploitation of the externalities such as pollution. This is part of the rational for giving cleaner tech subsidies, to balance the playing field. While I think it would be more honest to remove the externality, it's probably easier to give the clean technology the subsidy.
As I mentioned, the government creates the marketplace, so its got to do some "interfering". Where does one draw the line? This is where your first principle against coercion breaks down. You need to add some other principle or else you become an anarchist. In other words, the government needs to coerce, so you need some principle to tell it when to stop. Clearly those dictatorships went too far (though it's really not one-dimensional and there are other issues besides their economies). Currently Scandinavian countries have more coercion than the US and they do well in "happiness" metrics...
Just because an economy must be engineered does not mean it needs to be centrally managed. Engineering can include controlling a fiat currency, patents, monopoly rules, ...
I agree that economics is a poor science. Currently, it's a difficult system to model, but this applies to both the removal of controls as well as the application of controls. Also, in anticipation, I would disagree to the claim that a "simpler" economic system is easier to understand. In principle, various controls can be put in place to stabilize the system (probably at some cost, but this is an expected trade-off).
This clocks in at 510 pages, so it's not small. It looks good since I do have some interests in the history of the fractional banking system and central banks. Still, I doubt he has much in the way of empirically justified solutions as he seems pretty extreme free market and those extreme ideas have never been tested.
Yes, it is interesting to look at the failures and successes of various countries and try to determine causes.
This gets back to the libertarian book I read. They claimed economic and social freedom were orthogonal. Typically this is used to contrast the Republican versus Democratic party. As for your examples,
That seems a reasonable first principle.
But doesn't the government have to coerce people to do some of these things. I think you would agree that it boils down a thoughtful selection of what the government is allowed to do.
I can see your point, but you do realize that it's always going to happen. Some people will benefit more from "essential" government services then others. Of course, that doesn't mean one can't try to minimize this redistribution.
I would also like to make a bit of a retraction. I don't think the current taxation system is fair under any reasonable definition of the word, but I do think this redistribution might be necessary. Even a flat tax is probably not fair. Why should people pay a percentage of their income when they receive a fixed set of benefits? (I can argue this a bit in terms of the many disproportionate benefits of the wealthy, but this still doesn't cover many aspects of wealth redistribution to the poor.) However, I would argue that a fair system would lead to a very bad social outcome.
I don't think it would be hard to justify a reasonable economic simulation that would show that true economic freedom would lead to an extreme concentration of wealth. (Even more extreme then what we are currently seeing.) Is this good? I would claim it is bad for a range of reasons. I would also claim that a modern economy is an engineering problem. Perhaps we could modify the economy is ways that get a better social outcome without restricting liberty, but I'm OK with the government imposing some restrictions, just like they do in other "essential" areas.
It's a bit long for me. (Lots of other things in the pipeline.) I do remember reading a short Libertarian book many years ago that I found very enlightening. It was around 50 pages. Do you have a reference in the 20-50 page range?
Yes it does involve coercion, but I'm OK with that in essential areas where the market fails. For example, the only reasonable type of health insurance is one that covers chronic conditions. Before Obamacare, in many states there was no way to get those types of guarantees, so I consider this a market failure. For this, and other reasons, the government needs to step in.
I do think you have a reasonable position based on your main principle, and you seem to live by that principle. A reasonable reading of the Constitution might also be on your side. However, I'm pretty stubborn, and I'm going to stick with an, as yet undefined, notion of social good over individual liberty.
So you survive based on the generosity of a corporation. Not a position I would want to be in.
While you have many legitimate grievances, I would say this is your core issue. Personally, I want the government to do some wealth redistribution. For a range of reasons, I believe our current system is unfair. Of course this is a democracy, and I respect your right to fight for your beliefs and argue for the government you want.
Almost everyone is dependent on the government. At a minimum, the government needs to do things like create a market, setup a system of courts, protect the nation from hostile governments, ... Of course,
this is not what you mean, but I want to stress that the debate is not
about the government being involved it's about where and how much.
Again, I think this is the key issue. I see right leaning people who feel so strongly about this that they are willing to pay more to punish people than the cost of "successfully" subsidizing their behavior. At some level this make sense; you do not want your money to be used in a way that you feel is harming these people.
This is a complex issue, but I don't think it strongly relates to healthcare. Affordable access to healthcare is not going to finance someone's self-destruction. Perhaps it will not make him get a job, but for an individual in this position, lack of healthcare is also unlikely to make him get a job. As in the rest of the first world, I would suspect it will just become part of society and not have much effect on ambition or drive. In fact, it might have the opposite effect. If workers were not reliant on jobs for health insurance then they might be more willing to go into business for themselves and become true capitalists.
Agreed. Here is an interesting article: http://www.cnn.com/2017/06/27/...
Still I have to think your cost would be even higher without Obamacare. Do you have a death wish?
Don't you hit your maximum out of pocket expense limit.
As can be seen in the above article, private healthcare also has problems. I believe that single payer works in many countries. I don't believe we can dismiss all US government run programs.
I've heard this about some government jobs. Is it also true for contractors? What about medicare? (Yes there is fraud, but you can't compare against Utopia.)
So what happens to all the money that is currently spent on health care? Maybe we use that money...
So why can the governments of other countries pay for single payer? (The British are pretty happy with their health care, and your anecdotes on the British health care system make you sound like an astroturfer.) Also, why don't the ACA subsidies lower your costs? A simple search online suggest you should be paying around $100 a month for a silver plan.
I read a Scientific America special issue that covered this topic. Here's a review of the main topics. http://www.creatifik.com/fivep...
So you're saying if we got rid of the extra 2 electoral votes per state then Trump would have lost. This is false. Hillary won about 20 states to Trump's 30. Subtracting out the extra two votes gives Trump 246 and Hillary 192.
The real issue was the winner take all part of the system. With winner take all it is mathematically possible to win 25% of the popular vote and still win the election. With the extra 2 votes, it's even worse.