So ultimately, all the UBI does is raise the prices on
everything for everyone.
I can also weave a story. Now that potential landlords can make
twice as much money on apartments, there is a boom in apartment
construction. Prices go lower and everyone is happy.
My naive economics seems just as valid as yours. Of course, you
or I could make it more complex, but they would both still be crap.
What one needs are empirical results which is why people are now doing
experiments with UBI.
However, some simple arguments can't be refuted. Guy has no
money. UBI gives him money. Now he can buy stuff. Unless you are
saying the price of everything goes up by the amount UBI gives (or
more if you're a pessimist.)
Thanks for replying with the citations. I hate to complain, since
you did go to the effort to send the citations, which were
interesting, but do you have any page numbers? You cite a 126 page
report and a 336 page book which would take a long time to digest
without more information. The short article you did post is
interesting, but doesn't support your argument.
You can search up to power plants very easily as Australia found
out. Not only that but you can build a 1GW NG plant for less then
$400m.
My rough calculations show it would take about 10 billion to run
the plant for 30 years at peak capacity. A rough cost estimate for a
1GW Tesla battery is about 1 billion which should (hopefully) cover
most of the energy and waste used in construction. Of course, this is
apple to oranges in many ways...
Corporatism is destroying our biosphere, which will kill us. And it is
predicated solely upon greed.
As you know, it's an unfortunate consequence of capitalism and
corporations inherit desire to exploit externalities. An ethical but
more expensive way to extract energy will always lose in an
unregulated market. I'm not sure, I would attribute it directly to
greed. It's more a consequence of a system that allows guilt reducing
deniability.
But here's the thing, your idea of storage is built around batteries for the most part. It takes more energy and creates more waste to build them, then it does to build a natural gas power plant of comparable size and run it for 30 years.
Companies that didn't do everything everyone is dreaming
up before the FCC started controlling the internet won't suddenly
start doing it when they stop.
No, they are using a standardised dataset, just like everybody else.
That's good, but I hope everyone else isn't doing that. It would
be a great result to improve performance by using outside datasets.
In fact many of these neural nets use pretrained nets as building
blocks.
And of course the downside of everyone using the same dataset is
the statistical bias. The contest have value because the test data is
hidden. After the contest is over, people can beat up on that data as
much they want. I'm not saying they are cheating by using the test
data to train, but enough people testing their algorithms and
reporting their positive results will bias things up.
The parent NN sets "hyper-parameters", such as the number
of layers, the size of each layer, the activation function, the
convolution size, dropout rate, the learning rate damping factor, the
batch size, etc. Then it turns the children NNs loose on the image
dataset. It then sees which hyper-parameters lead to better/faster
performance, and then applies ML techniques to learn better
hyper-parameters.
None of this is new. What is new, is that Google is now
applying this recursively, and using AutoML to design a better
AutoML. This is another step toward the singularity.
Seems they are exposing/solving a dirty secret of deep learning.
Yes, one can claim the network is learning features, but to get the
best performance takes a human to design the structure to capture
those features. However, I wonder if they are exploiting the massive
amounts of data Google can access. One of the hard aspects of ML is
the statistical bias given a limited amount of data relative to the
set of relevant concepts. If they are coming up with ways of helping
these networks with their own data, it's not surprising the AutoML
technique works. Seems they are doing something sophisticated since
they are using reinforcement learning, but their webpage is light on
details. I'd like to see if they are beating humans given a equal set
of training data. (Of course, that training data yields more biased
results over time...)
But I doubt that AI will get to a point where it is
actively trying to kill us, in our lifetime.
I think many of these doomsayers are misrepresented. They often
claim that AI is the biggest threat to humanity. Presumably, this
means it is the most likely to wipe us out. This says nothing about
our lifetime. Most threats to humanity are on much longer timescales.
In this sense, AI probably is a bigger threat than asteroids or super
volcanoes. Also doesn't mean we shouldn't start developing methods to
control for this threat just like we try to map asteroids and
understand super volcanoes.
That one we might manage to verify by testing samples from a mere half
billion people against each of the other half billion. We leave the
problem of getting that much blood out of each test subject as an
exercise for the reader.
Good idea. If we assume these are independent trials then it's
much more feasible:) We can even do more than two people. An
experiment could be you got the perps DNA and a mix of 5 other
samples. Now can you detect whether or not the perp is in the mix.
Also I'm not worried about the amount of blood. Since we are assuming
the trials are independent, we can tolerate some experimental death.
I'm more worried about the time. Still it's probably doable with some
robotic assistance and is much faster than colonizing the Milky Way.
(In all fairness, colonizing the Milky Way has other benefits.)
If the company wants to claim one in 211 quintillion,
they need to provide a basis for that belief. To apply a mathematical
model to get that number, we'd have to be able to verify the model to
that accuracy, and we'd have to make sure all real-world possibilities
are accounted for. If there's a one in a trillion chance that
accidental contamination of a sample would make it return a false
positive, the probability estimate is off by at least eight orders of
magnitude.
Yes, it seems they have some papers, which as you point out, is still
worthless. Human error is going to completely dominate. My favorite
claim is that will allow the defense to look at their code if they are
paid money at an hourly rate. These guys are some impressive
assholes.
Sorry I couldn't help myself. I figured you didn't read the
article, and the ridiculous claims TrueAllele made. Human error for
DNA testing has been measured to be around 1 in 200, so these tiny
probabilities are just dangerous theatrics. Still it's an interesting challenge
to estimate extreme probability values. I was half hoping you'd shut
me up with some nice technical way around the problem...
As for empirical testing, it makes sense as part of a larger
system of evaluation. Looks like they have some papers to cover the
theory. I don't know if code review would also help, but I see no
reason not to allow the defense access.
This seems to happen every 50 years or so. OMG technology will take
our jobs. Oh wait it took jobs that we didnâ(TM)t want to do and it
created a new market for more jobs.
What's your sample size? If you're arguing it happens every 50
years over the last 200 years then statistically that's not a very
strong argument. Kind of like stock market arguments. Wow it goes up
8% every 20 years which gives maybe a sample size of 5 with the modern
market. Not a strong argument.
Set up a proper benchmark with a relevant number of samples and
confirm whether this (+ any other) program works exactly as expected.
So the article claims a false positive rate of 1 in 211
quintillion for a particular trial. To test that with a 95%
confidence interval we would need at least 600 quintillion samples.
Now we're a bit short on people on this planet. I don't think Earth
could support this many people so we need to colonize other planets.
To make things simple, lets assume the average planet can
support 10 billion people. Therefore, we need to colonize roughly 60
billion planets and test everyone on those planets. I think we can do
that without leaving the Milky Way galaxy, so we should be OK.
Each job should pay what it is worth. Do you think a burger flipper
should make the same as a highly skilled computer programmer?
Do you really think jobs pay what they are worth? Employers will
pay the burgers flippers as little as they can, irrespective of how
much they are worth. They also have little need to compete with each
other on salary since the unskilled workers have been made interchangeable.
Instead the tax payers are in essence subsidizing the owners to help
pay their workers a living wage.
Seriously, not all jobs are meant or worth paying a wage to live off
of....some ARE only for extra money on the side, or starting jobs for
teens.
Then why are people trying to make a living off those jobs. I
don't really care what was intended, I care about what is happening.
This has been the norm for decades, and only recently for
some reason, has everyone started thinking that ANYTHING you could
possibly do for money should pay enough to be your sole source of
income.
I think you've got things backwards. Now, people have to struggle
and do anything just to get enough money to live.
If the individual doesn't like the jobs they are being offered, then
THEY need to figure out what to do or what jobs to seek that do give
enough compensation to live off of solely.....
And if enough of those jobs don't exist then it's survival of the
fittest.
Damore worded his essay in a way designed to make it look
pro-diversity while cherry picking and exaggerating facts to cause an
opposite impact.
Perhaps, but I'm not a expert, and I've yet to see an expert debunk
his cherry picking. Some would claim he was just trying to present
his evidence.
Some have suggested Damore's honesty and intentions should be
questioned because of that. That's a reasonable thing to do.
That's a bit circular. They question his intent and then use that as
justification to question his intent. I'm not sure why you need the
rationalization. Seems the guy is mildly autistic, so the whole point
is that he's probably hard to read. However, it's fine if you don't
like what he wrote, and you can support that with arguments.
But... I'm not sure they're right. I think it's more a cause of an
expert in one field (computing) thinking that makes him an expert in
others and failing spectacularly trying to prove the real experts
wrong.
He claims many experts support him. Also he has some background
in these areas (PhD student.) Previous slashdot comments linked to a
blogger that did a good job of defending him.
http://slatestarcodex.com/2017...
It's a bit long, but much better than any articles I found that attack
his position. Again, I'm not an expert, and I could be swayed. In
fact, the expert, Adam Grant, he debunks politely rebuts his
comments. By the time I got the far, I was a bit drained, and I found
Grant's arguments somewhat persuasive. However, the blogger did a
sound job dismantling the points. (Though the blogger's reply was a bit
rude which might have stopped a productive back and forth.)
Now you might wonder why few experts vocally supported Damore. (I
think a few did.) However, it's not surprising that supporting
experts did not want to get involved in this firestorm...
You see this on Slashdot all the time. How many people
here disagree with the 97% or so of climate scientists on Climate
Change, for example? It certainly is more than 3%. Are these people
experts in climate change? Do they have all the figures available to
them? Do they have insights that Michael Mann et al don't?
I think the is an interesting issue. For the most part, neither
side of the slashdot debate are experts. How do they determine which
experts to listen to. Why should they even believe that 97% of the
climate scientists believe in climate change? What experts give them
this information? If they are not listening to experts, how do they
form their opinion?
The case for social science is even more complex. I think the
most damning argument against Damore is not that he thinks he can beat
the experts, it's that maybe there are no good experts in this area.
I think this is the real danger; if all science is lumped together,
questionable research can be used to weaken the strong work.
Personally, I assume climate science fits in the strong camp, but I
don't have the time to do extensive research myself, so I have to pick
my experts.
It downplays what caused the initial jump, however. You
don't just start in the middle at "the sudden increase in prices" and
go from there, though. Something in 2006 caused that sudden
increase. The massive increase in subprime lending, and therefore
buyers, is the only reason explanation for that which I've seen.
Well the issue has been sufficiently politicized that it's difficult
to assess expert opinions without a lot of research. However, CRA did
increase the number of mortgages which would increase the prices of
houses. This created a tempting opportunity for real estate
investment. Once the financial industry schemed how to hide bad
mortgages, the flippers and scammers were off to the races. At that
point, a positive feedback loop can cause a rapid rise in prices.
For the most part I think your analysis is accurate and balanced.
Mortgage-backed securities have been around for many
decades. They were quite popular in the 1970s. They were used in the
events of 2006-2008, but they weren't new, so their existence wasn't
the cause.
Credit default
swaps were invented by Blythe Masters from JP Morgan in 1994 and
increased in use in the early 2000s. It's interesting that it took 12
years. I would guess that the financial players are always trying to
cook up new schemes within the current rules. Perhaps they even
considered this a safe technique. It was clearly fine for a while.
However, there are often unintended consequences when trying new
things. Also, it's well know that their models didn't take into account
heavy tail distributions caused by correlations in variables.
The Financial Crisis Inquiry Commission (majority report), Federal
Reserve economists, and several academic researchers have stated that
government affordable housing policies were not the major cause of the
financial crisis.[6][112] They also state that Community Reinvestment
Act loans outperformed other "subprime" mortgages, and GSE mortgages
performed better than private label securitizations.
So some experts disagree. I do think it's probably
noncontroversial to say that without the credit default swap issue, the
crisis would have been much less severe.
But you trust government to "competently" control guns
and healthcare?
Your comparing something not being implemented with the government
implementing it. What you need to do is compare an implementation
using the private sector versus the public sector.
I think the government probably could do a better job than a
private company on national or global key escrow. However, it's
something we don't really need as the costs out weight the benefits
irrespective of who implements it.
We need healthcare. Now one needs to decide who can do the "best"
job. The evidence in the rest of the developed world strongly points
to the government.
When I read the summary, I didn't understand the importance of a 25% improvement. It seemed trivial. Going from impossible to 25 minutes is big. Going from 25 minutes to 18 minutes is minor. (A student in the area could probably optimize the original code to get this kind of improvement.) Maybe I'm missing something, but perhaps the GP has the right explanation...
Thanks for taking the time to give an in depth response. If I can
successfully paraphrase that means I probably understand. Let me give
it a try.
The problem seems to be based on the discrete cycle of payment.
Assume you make money every week and you pay money every week. To
find your profit you sum up what you make with what you pay.
Unfortunately, due to the discrete nature of this process, it bounces
around a lot. You have might have a surplus of money at some point
and then it drops down at the end of the week when you pay things like
salaries. Given that you don't want to always be taking out loans,
this graph must always be on the positive with a reasonable buffer.
It's also exacerbated by other things that are based on a longer
cycles. The problem is with taxes the government just looks at your
"profit" at the end of the year. This will be extra bad for companies
that have Christmas focused profits. For a modest size company this
might generate a large amount of tax.
So is this a problem in the long term? At the start of the next
year you would have a big expense as you pay all your bills.
Shouldn't this balance the profit at the end of the year. I guess if
you are a successful business and things are growing at a reasonable
rate then your profit will always be too high. I agree this doesn't
seem fair. What you need is a way to smooth your money curve. I
assume many companies go to considerable effort to minimize the impact
at the end of the year. You know it's coming so you make sure to
spend what you need at the end of the year. I guess this might be
hard for smaller companies.
I really am curious about this, and it seems plausible that small
business owners are getting screwed, but may naive understanding of a
business/corporation is that these taxes you mention are easy to
avoid. If you really want things to change, you need to educate
people on the problem.
As others have said, we believe a corporation only gets taxed on
profits. Honestly, I don't quite know what happens to these profits.
I guess they just go to the owners, and I'm not sure what taxes owners
have to pay on those already taxed profits. For a big company, some
might go out as dividends (to get taxed again as capital gains) and
some might go into an outside investment (for example Apple's huge
pile of cash). Many companies will just reinvest it all in themselves
which just amounts to an expense (for example Tesla.)
Why can't you set up this type of corporation for your business.
As I said, just pay yourself a big fat salary. This is an expense for
the company, so I assume you will just need to pay the appropriate
income and payroll tax on your salary. In fact, you probably get all
kinds of other tax benefits. With a real business there are all kinds
of things you can start to expense such as a company car. Are there
fixed costs for a corporation that make it too expensive for a small
business owner?
Why do you have to pay all this tax on money you don't bring home.
I find it hard to believe that you own your own business, but don't
know how to hire a good accountant.
This seems like an important issue. Can you explain in more detail? Yes there is wealth tied in the company, but for the most part, you are not taxed on that. Can't you just set yourself up with a salary and reinvest any profit into the company? That way you only pay taxes on the salary.
Why is scaling a big thing. If it works with N parameters, it will
work with N+1.
Scale is big because they didn't have the resources to solve these
problems in the past. Now that they have the resources, new issues
and problems arise.
As for your inductive argument, there are many cases where it is
not true. Many machine learning algorithms suffer from the curse of
dimensionality. Too many parameters and the results are worthless.
One needs to use things like L1 regularization which is related to
compressed sensing to find sparse solutions (if they exist.)
Also, optimal control is a thing and not exclusive,
although optimal control for a data set is not necessarily optimal for
another, or even piecewise continuous. But this comes down to how much
computing power you want to throw at it.
I don't know what you are trying to say.
Anyway, this stuff is 70+ years old and the mathematics
is well understood and common.
I'm sure a lot of the current stuff is a rehash of the old, but some
of it is new. (Necessity is the mother of invention.) As for math,
yes it's old and well understood. The issue is applying it in new
ways to new problems.
So how is this different from adaptive control that uses on-line
identification of the process parameters, or modification of
controller gains, thereby obtaining strong robustness
properties. Adaptive controls were applied for the first time in the
aerospace industry in the 1950s, and have found particular success in
that field.
It's not that different. A lot of current machine learning can be
viewed as tractable (often convex) and intractable optimization.
Probably a big difference is the size of the problems; going from
hundreds of parameters to millions/billions of parameters.
Another interesting difference is many machine learning algorithms
don't want to find the "optimal" solution. They either tweak the
problem with regularization or do some type of early stopping. This
is not just done for tractability, but also for the quality of the
solution. It's still got a bit too much magic for my taste.
Slashdot Mirror
I can also weave a story. Now that potential landlords can make twice as much money on apartments, there is a boom in apartment construction. Prices go lower and everyone is happy.
My naive economics seems just as valid as yours. Of course, you or I could make it more complex, but they would both still be crap. What one needs are empirical results which is why people are now doing experiments with UBI.
However, some simple arguments can't be refuted. Guy has no money. UBI gives him money. Now he can buy stuff. Unless you are saying the price of everything goes up by the amount UBI gives (or more if you're a pessimist.)
Thanks for replying with the citations. I hate to complain, since you did go to the effort to send the citations, which were interesting, but do you have any page numbers? You cite a 126 page report and a 336 page book which would take a long time to digest without more information. The short article you did post is interesting, but doesn't support your argument.
My rough calculations show it would take about 10 billion to run the plant for 30 years at peak capacity. A rough cost estimate for a 1GW Tesla battery is about 1 billion which should (hopefully) cover most of the energy and waste used in construction. Of course, this is apple to oranges in many ways...
As you know, it's an unfortunate consequence of capitalism and corporations inherit desire to exploit externalities. An ethical but more expensive way to extract energy will always lose in an unregulated market. I'm not sure, I would attribute it directly to greed. It's more a consequence of a system that allows guilt reducing deniability.
Citation needed.
That's right frogs get ready for the slow boil.
That's good, but I hope everyone else isn't doing that. It would be a great result to improve performance by using outside datasets. In fact many of these neural nets use pretrained nets as building blocks.
And of course the downside of everyone using the same dataset is the statistical bias. The contest have value because the test data is hidden. After the contest is over, people can beat up on that data as much they want. I'm not saying they are cheating by using the test data to train, but enough people testing their algorithms and reporting their positive results will bias things up.
Seems they are exposing/solving a dirty secret of deep learning. Yes, one can claim the network is learning features, but to get the best performance takes a human to design the structure to capture those features. However, I wonder if they are exploiting the massive amounts of data Google can access. One of the hard aspects of ML is the statistical bias given a limited amount of data relative to the set of relevant concepts. If they are coming up with ways of helping these networks with their own data, it's not surprising the AutoML technique works. Seems they are doing something sophisticated since they are using reinforcement learning, but their webpage is light on details. I'd like to see if they are beating humans given a equal set of training data. (Of course, that training data yields more biased results over time...)
I think many of these doomsayers are misrepresented. They often claim that AI is the biggest threat to humanity. Presumably, this means it is the most likely to wipe us out. This says nothing about our lifetime. Most threats to humanity are on much longer timescales. In this sense, AI probably is a bigger threat than asteroids or super volcanoes. Also doesn't mean we shouldn't start developing methods to control for this threat just like we try to map asteroids and understand super volcanoes.
Good idea. If we assume these are independent trials then it's much more feasible :) We can even do more than two people. An
experiment could be you got the perps DNA and a mix of 5 other
samples. Now can you detect whether or not the perp is in the mix.
Also I'm not worried about the amount of blood. Since we are assuming
the trials are independent, we can tolerate some experimental death.
I'm more worried about the time. Still it's probably doable with some
robotic assistance and is much faster than colonizing the Milky Way.
(In all fairness, colonizing the Milky Way has other benefits.)
Yes, it seems they have some papers, which as you point out, is still worthless. Human error is going to completely dominate. My favorite claim is that will allow the defense to look at their code if they are paid money at an hourly rate. These guys are some impressive assholes.
Sorry I couldn't help myself. I figured you didn't read the article, and the ridiculous claims TrueAllele made. Human error for DNA testing has been measured to be around 1 in 200, so these tiny probabilities are just dangerous theatrics. Still it's an interesting challenge to estimate extreme probability values. I was half hoping you'd shut me up with some nice technical way around the problem...
As for empirical testing, it makes sense as part of a larger system of evaluation. Looks like they have some papers to cover the theory. I don't know if code review would also help, but I see no reason not to allow the defense access.
What's your sample size? If you're arguing it happens every 50 years over the last 200 years then statistically that's not a very strong argument. Kind of like stock market arguments. Wow it goes up 8% every 20 years which gives maybe a sample size of 5 with the modern market. Not a strong argument.
So the article claims a false positive rate of 1 in 211 quintillion for a particular trial. To test that with a 95% confidence interval we would need at least 600 quintillion samples. Now we're a bit short on people on this planet. I don't think Earth could support this many people so we need to colonize other planets. To make things simple, lets assume the average planet can support 10 billion people. Therefore, we need to colonize roughly 60 billion planets and test everyone on those planets. I think we can do that without leaving the Milky Way galaxy, so we should be OK.
Do you really think jobs pay what they are worth? Employers will pay the burgers flippers as little as they can, irrespective of how much they are worth. They also have little need to compete with each other on salary since the unskilled workers have been made interchangeable. Instead the tax payers are in essence subsidizing the owners to help pay their workers a living wage.
Then why are people trying to make a living off those jobs. I don't really care what was intended, I care about what is happening.
I think you've got things backwards. Now, people have to struggle and do anything just to get enough money to live.
And if enough of those jobs don't exist then it's survival of the fittest.
Perhaps, but I'm not a expert, and I've yet to see an expert debunk his cherry picking. Some would claim he was just trying to present his evidence.
That's a bit circular. They question his intent and then use that as justification to question his intent. I'm not sure why you need the rationalization. Seems the guy is mildly autistic, so the whole point is that he's probably hard to read. However, it's fine if you don't like what he wrote, and you can support that with arguments.
He claims many experts support him. Also he has some background in these areas (PhD student.) Previous slashdot comments linked to a blogger that did a good job of defending him. http://slatestarcodex.com/2017... It's a bit long, but much better than any articles I found that attack his position. Again, I'm not an expert, and I could be swayed. In fact, the expert, Adam Grant, he debunks politely rebuts his comments. By the time I got the far, I was a bit drained, and I found Grant's arguments somewhat persuasive. However, the blogger did a sound job dismantling the points. (Though the blogger's reply was a bit rude which might have stopped a productive back and forth.)
Now you might wonder why few experts vocally supported Damore. (I think a few did.) However, it's not surprising that supporting experts did not want to get involved in this firestorm...
I think the is an interesting issue. For the most part, neither side of the slashdot debate are experts. How do they determine which experts to listen to. Why should they even believe that 97% of the climate scientists believe in climate change? What experts give them this information? If they are not listening to experts, how do they form their opinion?
The case for social science is even more complex. I think the most damning argument against Damore is not that he thinks he can beat the experts, it's that maybe there are no good experts in this area. I think this is the real danger; if all science is lumped together, questionable research can be used to weaken the strong work. Personally, I assume climate science fits in the strong camp, but I don't have the time to do extensive research myself, so I have to pick my experts.
Well the issue has been sufficiently politicized that it's difficult to assess expert opinions without a lot of research. However, CRA did increase the number of mortgages which would increase the prices of houses. This created a tempting opportunity for real estate investment. Once the financial industry schemed how to hide bad mortgages, the flippers and scammers were off to the races. At that point, a positive feedback loop can cause a rapid rise in prices.
Credit default swaps were invented by Blythe Masters from JP Morgan in 1994 and increased in use in the early 2000s. It's interesting that it took 12 years. I would guess that the financial players are always trying to cook up new schemes within the current rules. Perhaps they even considered this a safe technique. It was clearly fine for a while. However, there are often unintended consequences when trying new things. Also, it's well know that their models didn't take into account heavy tail distributions caused by correlations in variables.
Also from the Subprime Mortgage Wikipedia page/,
So some experts disagree. I do think it's probably noncontroversial to say that without the credit default swap issue, the crisis would have been much less severe.
Your comparing something not being implemented with the government implementing it. What you need to do is compare an implementation using the private sector versus the public sector.
I think the government probably could do a better job than a private company on national or global key escrow. However, it's something we don't really need as the costs out weight the benefits irrespective of who implements it.
We need healthcare. Now one needs to decide who can do the "best" job. The evidence in the rest of the developed world strongly points to the government.
When I read the summary, I didn't understand the importance of a 25% improvement. It seemed trivial. Going from impossible to 25 minutes is big. Going from 25 minutes to 18 minutes is minor. (A student in the area could probably optimize the original code to get this kind of improvement.) Maybe I'm missing something, but perhaps the GP has the right explanation...
Thanks for taking the time to give an in depth response. If I can successfully paraphrase that means I probably understand. Let me give it a try.
The problem seems to be based on the discrete cycle of payment. Assume you make money every week and you pay money every week. To find your profit you sum up what you make with what you pay. Unfortunately, due to the discrete nature of this process, it bounces around a lot. You have might have a surplus of money at some point and then it drops down at the end of the week when you pay things like salaries. Given that you don't want to always be taking out loans, this graph must always be on the positive with a reasonable buffer. It's also exacerbated by other things that are based on a longer cycles. The problem is with taxes the government just looks at your "profit" at the end of the year. This will be extra bad for companies that have Christmas focused profits. For a modest size company this might generate a large amount of tax.
So is this a problem in the long term? At the start of the next year you would have a big expense as you pay all your bills. Shouldn't this balance the profit at the end of the year. I guess if you are a successful business and things are growing at a reasonable rate then your profit will always be too high. I agree this doesn't seem fair. What you need is a way to smooth your money curve. I assume many companies go to considerable effort to minimize the impact at the end of the year. You know it's coming so you make sure to spend what you need at the end of the year. I guess this might be hard for smaller companies.
I really am curious about this, and it seems plausible that small business owners are getting screwed, but may naive understanding of a business/corporation is that these taxes you mention are easy to avoid. If you really want things to change, you need to educate people on the problem.
As others have said, we believe a corporation only gets taxed on profits. Honestly, I don't quite know what happens to these profits. I guess they just go to the owners, and I'm not sure what taxes owners have to pay on those already taxed profits. For a big company, some might go out as dividends (to get taxed again as capital gains) and some might go into an outside investment (for example Apple's huge pile of cash). Many companies will just reinvest it all in themselves which just amounts to an expense (for example Tesla.)
Why can't you set up this type of corporation for your business. As I said, just pay yourself a big fat salary. This is an expense for the company, so I assume you will just need to pay the appropriate income and payroll tax on your salary. In fact, you probably get all kinds of other tax benefits. With a real business there are all kinds of things you can start to expense such as a company car. Are there fixed costs for a corporation that make it too expensive for a small business owner?
Why do you have to pay all this tax on money you don't bring home. I find it hard to believe that you own your own business, but don't know how to hire a good accountant.
This seems like an important issue. Can you explain in more detail? Yes there is wealth tied in the company, but for the most part, you are not taxed on that. Can't you just set yourself up with a salary and reinvest any profit into the company? That way you only pay taxes on the salary.
Scale is big because they didn't have the resources to solve these problems in the past. Now that they have the resources, new issues and problems arise.
As for your inductive argument, there are many cases where it is not true. Many machine learning algorithms suffer from the curse of dimensionality. Too many parameters and the results are worthless. One needs to use things like L1 regularization which is related to compressed sensing to find sparse solutions (if they exist.)
I don't know what you are trying to say.
I'm sure a lot of the current stuff is a rehash of the old, but some of it is new. (Necessity is the mother of invention.) As for math, yes it's old and well understood. The issue is applying it in new ways to new problems.
It's not that different. A lot of current machine learning can be viewed as tractable (often convex) and intractable optimization. Probably a big difference is the size of the problems; going from hundreds of parameters to millions/billions of parameters.
Another interesting difference is many machine learning algorithms don't want to find the "optimal" solution. They either tweak the problem with regularization or do some type of early stopping. This is not just done for tractability, but also for the quality of the solution. It's still got a bit too much magic for my taste.
Yes the cooling plan got some recent attention. http://www.iflscience.com/envi...
My nexus 6 started skipping when I upgraded to nougat. Maybe it's the OS?