I think that’s less about being afraid that Amazon will steal their data, and more that they don’t want to give any money to an entity already steamrolling them
Some are legitimately afraid that AWS will deprive them of the ability to scale during peak times, like holiday shopping seasons. I've heard claims of this happening to more than one retailer.
Personally, I wonder if that isn't an emergent property of a lot of people trying to scale at once.
Does anybody know of automated waterfall budgeting solution? I want to be able to forecast the future if I stick to my pay off plans. I want to be able to enter my bills, what I owe on them, then run scenarios on my financial future if I payed certain things off before others in waterfall fashion. Having an idea of the future helps motivate.
Ah the dial-up ISP days. As an young enterprising middle schooler 20+ years ago I started one with the help of my father. I remember stringing together a bunch of modems to a Postmaster and getting a rollover PRI line from the local telco. If my memory serves me it had 24 lines and would roll over if one was busy. We even made some of our own installation floppy disks and placed them at the local video rental store so people could self install. Fun times!
"A Judas goat was a female who would be captured from the wild, tagged with a GPS tracking device, and then released to find other goats, especially lovelorn males.
The sharpshooters would take to the air again, track the Judas goat, find her hidden companions and gun them down, always leaving the Judas goat alive so that the whole process would begin again. Track, slaughter, repeat. The team eventually used 900 Judas goats over the course of a couple of years."
Probably gave those poor females a serious complex. Can you imagine how messed up you'd be if every time you approached someone of the opposite sex, they got shot from a helicopter?
And this is why I don't believe in uprisings and revolutions in modern societies anymore. As an individual, even if you can buy guns, you are not that far away from the goat compared to what tech and training current militaries have.
In the specific, an unarmored helicopter hovering or moving predictably close enough for a marksman to shoot at a target (goat or otherwise) would be easy prey for a handful of people with small arms.
In the more general case, I don't think you quite appreciate how difficult counterinsurgency operations are, or how little tech affects them. Jets can drop some pretty nasty ordinance, but require targeting (the counterinsurgents _really_ don't want to damage infrastructure if they can avoid it) which they have trouble doing. Attack Helicopters are lean mean killing machines, but aren't great in urban environments and also need to PID their targets if the brass gives any fucks about civilian casualties. Tanks are great pretty much everything when it comes to conventional warfare, but _MUST BE SCREENED BY INFANTRY_. Looking at you, giraffe-man.
All this probably makes you think of drones (I'm going to only talk about large UAVs here), which are in fact a pretty useful asset to security forces. They can carry out assassinations, recon, target designation, gather intelligence, and all without risking pilots. However, in a large scale insurgency, they really aren't that useful in a tactical sense.
This is because of the simple truth that the only thing that can hold territory (especially cities) is infantry. A drone can't disperse a mob. A jet can't set up a checkpoint. An attack helicopter can't search a neighborhood for contraband. A tank can't rescue an informant in the middle of the night. Only infantry can do this.
And in spite of any training or technical advantages to date, an insurgent with a rifle can still kill infantry.
A human can carry a shoulder-launched anti-aircraft weapon and destroy the helicopter, while hiding amongst the rubble of a chaotic urban environment.
Asymmetric warfare is very difficult. The US military struggled for many years and spent huge amounts of money fighting insurgencies in Iraq and Afghanistan. Those are small countries compared to the US. Trying to fight an insurgency on home soil, without the support of a tax-paying public? Forget about it!
FWIW insurgent asymmetric warfare was how the US primarily fought its own revolution (there were a few set piece battles as well, e.g. Yorktown, hardly insignificant).
In the end I believe quite a bit of cash spread around Whitehall is what made the difference, though not everyone will agree.
In the US civil war there was much less of this for reasons unclear to me.
One theory would be that CSA's draconian conscription operation (watch or read Cold Mountain, for example) emptied the countryside of its natural defenders. The conscription level may have been necessary given the idiotic location of the CSA capital, but quite a bit about CSA was idiotic.
In Missouri, which didn't secede but had a lot of Southern sympathizers, this conscription wasn't really a thing, and there was lots of asymmetric warfare. It actually started in the 1850s in what would become Kansas. (There are good movies about this too: The Outlaw Josey Wales and Ride with the Devil.)
> The US military struggled for many years and spent huge amounts of money fighting insurgencies in Iraq and Afghanistan
They kill 100 fighters for one death on their side, and have control over the natural resources they want. I don't see the bright side for the opposite party.
> Trying to fight an insurgency on home soil, without the support of a tax-paying public? Forget about it!
That's a better argument. But that would assume the revolution would last. The local population would not have money or food either, but wouldn't have the supply chain or the reserve the military have.
The us lost these engagements though. Truly defeating insurgencies like these is currently beyond the reach of the most technologically advanced and well funded armies in the world.
The local fighters are slaughtered in mass for very little effect toward their objectives of liberation while the US still maintain control of the natural resources the way they want.
I think they must have been doing this on a smaller scale in the 80s as well. When I was a kid my dad was doing some work with the Ecuadorian park service, and I went on a goat hunt on Isabela. The park rangers killed a lot of goats, and we barbecued one of them on the beach.
We use DirectTV Now app on our Apple TV over a Comcast connection. I really like it. The only problem we have is Comcast imposes a data cap of 1TB. With 3 TVs going in the house, kids watching Netflix, and my wife and I watching a separate shows we occasionally hit this limit then get charged an extra $10 per 50GB. Seems ridiculous.
Hi Jermey... Great article. I have a labeled product image dataset (thumbnails, and various sizes) that I'd like to use for the model training. Can you recommend any lightweight methods for commodity hardware?
You can use the approach we used to win the CIFAR 10 training cost section of the competition. If you use fastai/pytorch, then it's ~5 lines of code. Check out lesson 1 of http://course.fast.ai for the basic approach, but when calling `fit()`, add the param `use_clr_beta=(20,20,0.95,0.85)` which will enable 1cycle, and should allow of super convergence. Then train with SGD with a really high learning rate (somewhere from 1-3, generally).
