Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

The NSA’s machine learning algorithm may be killing thousands of innocent people

arstechnica.co.uk

Post reply on HN
Print view View original

141–150 of 221 posts

Re: The NSA’s machine learning algorithm may be killing thousands of innocent people

#141

Post by IIAOPSW »

Aeolos wrote: post #109
mcv wrote: post #68

Earlier quoted context omitted.

But the end result of all that is just that it generates even more hostility towards the US and feeds terrorism. The US is creating its own enemies this way. Again.

Cynically speaking, this is win-win for everyone involved. Except for the innocents who are killed. The US gets to debug and optimize their assassination program "on brown people that noone cares about". Pakistan and US also get to claim that they are "fighting terrorism". This breeds resentment and new generations of terrorists to be "fought" in the future (divide and conquer 101). Plus, who knows when and where the…

droid program?

Re: The NSA’s machine learning algorithm may be killing thousands of innocent people

#142

Post by merpnderp »

arca_vorago wrote: post #112
hodwik wrote: post #104

Earlier quoted context omitted.

We need a military. All states do -- its part of their duty to their citizenry. The problem in the US is that our hippy-run academia are driving the smart people out of jobs in the military (and police force for that matter), and it's putting us in a terrible position. The only way to fix it is for smart people to start joining the military again.

This is absolute bullshit. The problem is that we need a military that is actually going to man the fuck up when an O-5+ decides to push some unconstitutional program on the people. What has happened isnt that hippy academia is jacking up the military, but rather that the military has been kicking out the kind of people who arent afraid to tell truth to power, and are increasingly making choices and then only hiring…

New world order? The Democrat establishment is doing everything it can to keep the primary from being taken from their anointed disciple. And the Republican primary barely even contains an establishment runner. The GOP power brokers can't decide if they should spend more effort attacking Trump or Cruz, but their man Rubio is trailing a distant third, and their real choice, Bush, is lost in the noise.

If there is a world order, it was never very powerful or orderly.

Re: The NSA’s machine learning algorithm may be killing thousands of innocent people

#143

Post by ctlby »

Houshalter wrote: post #7

They literally named it Skynet. They have an evil sense of humor. Actually using machine learning to detect terrorists isn't a terrible idea. But you are going to get an error rate, and probably a high one in the noisy real world. Maybe only 50% of the people you detect are actually terrorists. Maybe it's even worse than that. We can't even test it because there is no validation set and unreliable labels. The reasona…

> But you are going to get an error rate, and probably a high one in the noisy real world.

They're clearly not implementing a minimum-error rate classifier. Which isn't to defend their choice of loss function (it's not 0-1, but beyond that I have no idea); just to point out the inanity of some of the comments claiming the NSA isn't aware of base rates.

Re: The NSA’s machine learning algorithm may be killing thousands of innocent people

#144

Post by pc86 »

alanwatts wrote: post #139
Aeolos wrote: post #129

Earlier quoted context omitted.

You really want to know? Because of regression towards the mean. Having a single state would mean that the US couldn't afford to have a significantly higher standard of living compared to the rest of the world. In fact, 1st world countries would have to meet 3rd world countries somewhere in the middle. (Assuming you are living in the states) Would you be willing to live on a wage of $5000 a year? Edit: see here https…

You're still thinking in terms of 19th century paper-based processes when you attempt to estimate globally integrated wealth in terms of a nationally fragmented paper-based information system ($). Also, GDP is not indicative of net wealth, as it is a measurement which implicitly asserts the broken window fallacy. Over the long term, breaking down these legacy barriers to trade and freedom of movement would necessaril…

Please try to make your points without quoting random people. Quotes are not sources, they are not facts, they're just something someone said.

Re: The NSA’s machine learning algorithm may be killing thousands of innocent people

#145

Post by jndsn402 »

qeabc wrote: post #62
Houshalter wrote: post #7

They literally named it Skynet. They have an evil sense of humor. Actually using machine learning to detect terrorists isn't a terrible idea. But you are going to get an error rate, and probably a high one in the noisy real world. Maybe only 50% of the people you detect are actually terrorists. Maybe it's even worse than that. We can't even test it because there is no validation set and unreliable labels. The reasona…

Perhaps someone who knows the topic can say whether this mathematical phenomena of "adversarial examples" for neural networks can be translated to random forests: http://www.kdnuggets.com/2015/07/deep-learning-adversarial-e... Barely perceptible changes to an image cause it to be misclassified by neural networks which never saw the image before (this is important, because it rules out simple overfitting). As a non-ex…

With respect to adversarial examples for random forests, that link states:

   Myth: Deep learning is more vulnerable to adversarial examples than other kind of machine learning.

   Fact: So far we have been able to generate adversarial examples for every model we have tested, including simple traditional machine learning models like nearest neighbor. Deep learning with adversarial training is the most resistant technique we have studied so far.
But the larger point is that adversarial examples are just one demonstration that algorithms are still quite primitive. Nobody should be killed on the basis of a statistical algorithm alone (and to be fair, the article does not show that anyone has been).

Re: The NSA’s machine learning algorithm may be killing thousands of innocent people

#146

Post by mratzloff »

flashman wrote: post #27
Houshalter wrote: post #7

They literally named it Skynet. They have an evil sense of humor. Actually using machine learning to detect terrorists isn't a terrible idea. But you are going to get an error rate, and probably a high one in the noisy real world. Maybe only 50% of the people you detect are actually terrorists. Maybe it's even worse than that. We can't even test it because there is no validation set and unreliable labels. The reasona…

If you are going to put surveillance on potential terrorists, you should also be putting surveillance on people not suspected of being terrorists, so you can judge the rate at which your surveillance personnel falsely accuse people being terrorists based on having been selected for surveillance. Single or double blind that process.

That would be true if you were genuinely interested in avoiding innocent deaths. If you placed very little or no value on non-Western lives, then there would be no incentive to take that kind of extra care.

Re: The NSA’s machine learning algorithm may be killing thousands of innocent people

#147

Post by anovikov »

If that system skill kills enough terrorists to make them scared and limit inflow if new recruits into that system than why not? Obviously rest of the population of Pakistan pays their price, but what they expect if they can't (or really, don't want to) shut down terrorist activity in their own country themselves?

Re: The NSA’s machine learning algorithm may be killing thousands of innocent people

#148

Post by jndsn402 »

pc86 wrote: post #130
eli_gottlieb wrote: post #100

Earlier quoted context omitted.

Wait wait wait. So they're killing people based on a model that fails Statistics 101 !? No accounting for base rates? No cross-validation!? God fucking damnit, why the fuck does our government think killing should be cheaper and easier than healing!? We put new medical treatments through decades of difficult scientific tests, but have none at all for strictly invalid ways to kill people?

I do not think you completely understood the article.

The models as described in the article do seem to have glaring flaws. However, the article did not show that anyone has been killed based on the models alone.

Re: The NSA’s machine learning algorithm may be killing thousands of innocent people

#149

Post by alanwatts »

pc86 wrote: post #144
alanwatts wrote: post #139

Earlier quoted context omitted.

You're still thinking in terms of 19th century paper-based processes when you attempt to estimate globally integrated wealth in terms of a nationally fragmented paper-based information system ($). Also, GDP is not indicative of net wealth, as it is a measurement which implicitly asserts the broken window fallacy. Over the long term, breaking down these legacy barriers to trade and freedom of movement would necessaril…

Please try to make your points without quoting random people. Quotes are not sources, they are not facts, they're just something someone said.

If you think the thoughts of great scientists and engineers like Buckminster Fuller (Former president of the Mensa Society) and Carl Sagan are unwarranted in a place like hackernews then you should find a new forum. Do you have anything to contribute to this discussion?

Edit:

"The argument from authority (Latin: argumentum ad verecundiam) also appeal to authority, is a common argument form which can be fallacious, such as when an authority is cited on a topic outside their area of expertise, when the authority cited is not a true expert."

Please do tell who might be a true expert on science and technology as it relates to societal organization, if not Fuller or Sagan.

Re: The NSA’s machine learning algorithm may be killing thousands of innocent people

#150

Post by pc86 »

alanwatts wrote: post #149
pc86 wrote: post #144

Earlier quoted context omitted.

Please try to make your points without quoting random people. Quotes are not sources, they are not facts, they're just something someone said.

If you think the thoughts of great scientists and engineers like Buckminster Fuller (Former president of the Mensa Society) and Carl Sagan are unwarranted in a place like hackernews then you should find a new forum. Do you have anything to contribute to this discussion? Edit: "The argument from authority (Latin: argumentum ad verecundiam) also appeal to authority, is a common argument form which can be fallacious, su…

Since you're a fan of linking to Wikipedia articles about fallacious arguments, I'll just leave this:

https://en.wikipedia.org/wiki/Argument_from_authority

Post reply on HN