chain chain chain

This week’s links have some games in them:

Budget Hero: a colorful online game that challenges you to balance the budget. First, you assign yourself various badges to reflect your priorities, like Health and Fitness or National Defense or Fiscal Responsibility. Then, you choose which programs you want to expand or cut, in the form of cards to be played. Then you pull the trigger and see how long you can last before the debt becomes unsustainable. Challenging and deep, despite its simple appearance. I submit that the libertarian fantasy of wantonly slashing federal programs might not even be possible, much less realistic. The infrastructure has become tangled with kudzu and it may be tough to clear it.

I’ve been talking about Bayes a lot lately, and not always as clearly as I’d like. Here’s Cory Doctorow, writing on Bayesian probability and terrorist screening, in the Guardian:

Our innumeracy means that our fight against these super-rarities is likewise ineffective. Statisticians speak of something called the Paradox of the False Positive. Here’s how that works: imagine that you’ve got a disease that strikes one in a million people, and a test for the disease that’s 99% accurate. You administer the test to a million people, and it will be positive for around 10,000 of them – because for every hundred people, it will be wrong once (that’s what 99% accurate means). Yet, statistically, we know that there’s only one infected person in the entire sample. That means that your “99% accurate” test is wrong 9,999 times out of 10,000!

Terrorism is a lot less common than one in a million and automated “tests” for terrorism – data-mined conclusions drawn from transactions, Oyster cards, bank transfers, travel schedules, etc – are a lot less accurate than 99%. That means practically every person who is branded a terrorist by our data-mining efforts is innocent.

In other words, in the effort to find the terrorist needles in our haystacks, we’re just making much bigger haystacks.

You don’t get to understand the statistics of rare events by intuition. It’s something that has to be learned, through formal and informal instruction. If there’s one thing the government and our educational institutions could do to keep us safer, it’s this: teach us how statistics works.

Let’s break up all this depressing political talk with a little alternate universe cosmology:

Among the unnatural aspects of the universe, one stands out: time asymmetry. The microscopic laws of physics that underlie the behavior of the universe do not distinguish between past and future, yet the early universe—hot, dense, homogeneous—is completely different from today’s—cool, dilute, lumpy. The universe started off orderly and has been getting increasingly disorderly ever since. The asymmetry of time, the arrow that points from past to future, plays an unmistakable role in our everyday lives: it accounts for why we cannot turn an omelet into an egg, why ice cubes never spontaneously unmelt in a glass of water, and why we remember the past but not the future. And the origin of the asymmetry we experience can be traced all the way back to the orderliness of the universe near the big bang. Every time you break an egg, you are doing observational cosmology.

The arrow of time is arguably the most blatant feature of the universe that cosmologists are currently at an utter loss to explain. Increasingly, however, this puzzle about the universe we observe hints at the existence of a much larger spacetime we do not observe. It adds support to the notion that we are part of a multiverse whose dynamics help to explain the seemingly unnatural features of our local vicinity.

I hate how the modern news cycle discards stories just as they get interesting: a Texas Appeals court has thrown out the state’s seizure of children from a polygamist compound:

In a ruling that could torpedo the case against the West Texas polygamist sect, a state appeals court Thursday said authorities had no right to seize more than 440 children in a raid on the splinter group’s compound last month.

The Third Court of Appeals in Austin said the state failed to show the youngsters were in any immediate danger, the only grounds in Texas law for taking children from their parents without court action.

It was not clear when the children – now scattered in foster homes across the state – might be returned to their parents. The ruling gave a lower-court judge 10 days to release the youngsters from custody, but the state could appeal to the Texas Supreme Court and block that.

The decision in one of the biggest child-custody cases in U.S. history was a humiliating defeat for the state Child Protective Services agency. It was hailed as vindication by members of the Fundamentalist Church of Jesus Christ of Latter Day Saints, who claimed they were being persecuted for their religious beliefs.

Child Protective Services vs. the Fundamentalist Church of Jesus Christ of Latter Day Saints. I’m honestly not sure who to cheer against here.

I wonder how much worse England would have grown, as a surveillance state, if Orwell hadn’t given us his last name as an adjective. Very few people understand the true dangers of the state described in 1984 – namely, historical revisionism and control of language – but the general thrust, “cameras = bad,” is better than nothing.

I mention this as a prelude to the Home Office’s plan to monitor every single phone call and e-mail sent in the UK:

A Home Office spokesman said: “The Communications Data Bill will help ensure that crucial capabilities in the use of communications data for counter-terrorism and investigation of crime continue to be available.

“These powers will continue to be subject to strict safeguards to ensure the right balance between privacy and protecting the public.”

The spokesman said changes need to be made to the Regulation of Investigatory Powers Act 2000 “to ensure that public authorities can continue to obtain and have access to communications data essential for counter-terrorism and investigation of crime purposes”.

But the Information Commission, an independent authority set up to protect personal information, said the database “may well be a step too far” and highlighted the risk of data being lost, traded or stolen.

Assistant information commissioner Jonathan Bamford said: “We are not aware of any justification for the state to hold every UK citizen’s phone and internet records. We have real doubts that such a measure can be justified, or is proportionate or desirable.

“Defeating crime and terrorism is of the utmost importance, but we are not aware of any pressing need to justify the government itself holding this sort of data.”

Let’s break up the depressing news with some photographs.

Here’s some wedding photography, taken during last week’s 7.9 quake in China:
Chinese earthquake wedding

Finally, the Economist does some math:

Data centres consumed 0.6% of the world’s electricity in 2000, and 1% in 2005. Globally, they are already responsible for more carbon-dioxide emissions per year than Argentina or the Netherlands, according to a recent study by McKinsey, a consultancy, and the Uptime Institute, a think-tank. If today’s trends hold, these emissions will have grown four-fold by 2020, reaching 670m tonnes. By some estimates, the carbon footprint of cloud computing will then be larger than that of aviation.

I’m innately leery of any figure that arises from predicting 12 years worth of trends. But a “peak server” crash could be just as bad as a peak oil crash. Having a 1980 volume of oil in 2020 would be terrible; having a 1980 level of computing power in 2020 would fuck us. Hard.


8 Responses

  1. I often wonder about how much my field is contributing to the warming of the Earth. We’re trying like mad to reduce the amount of power used by our machines like mad here, not because of environmental concerns, but because the power grid in my old building can’t handle it. But, people keep insisting on more and more in computing, so even with machine virtualization, I don’t see this trend going down any time soon.

    I do wonder more, actually, about how much the dilapidated state of the American power grid is contributing to global warming. I remember reading somewhere that there is a significant (I don’t recall the percent, but it was high) amount of power lost because the wiring is degrading without replacement. I also wonder how much the inefficient traffic system in the Boston area is contributing to global warming since cars are required to stop WAY more frequently than they should and how much gas that wastes for people that would drive regardless of how many stops they would make.

    You really don’t hear as much about these causes, do you?

  2. The logical conclusion from this is not that we should not screen for terrorists (since the consequences of letting terrorists slip through undetected could be very, VERY bad) but rather that we should be aware that most of those who test positive for being terrorists probably aren’t, and hence we should be careful not to abuse them in the process of investigating them.

  3. Mark: interesting. Bastiat wrote a lot about seen and unseen costs.

    Jordan: regarding your sentence, “we should be aware that most of those who test positive for being terrorists probably aren’t, and hence we should be careful not to abuse them in the process of investigating them,” I have three questions:

    (1) Who are “we”?

    (2) How “careful”?

    (3) “Should” implies an obligation. Who’s doing the obligating, with what force and with what penalty for noncompliance?

  4. That Cory Doctorow example is ridiculous. Saying that the 99% accurate test is wrong “9,999 out of 10,000 times” because you are choosing to limit the group being talked about to the cases where it was wrong is cherry-picking to an absurd degree. Is he also saying that it’s right 990,000 out of 990,000 times and therefore it is the greatest 99% accurate test ever?

  5. Mike – I’m not sure what exactly you’re saying. A 99% accurate test, if applied to a population where the testable variable only appears 1 time in 1,000,000 must, by definition, generate 9,999 false positives. If you accept the math, then what’s your issue with Doctorow’s statement?

  6. It does not need to generate 9,999 false positives. Some of the “non-accurate” results could be false negatives. Mike is pointing out that Doctorow was assuming they’d all be false positives for no good reason.

  7. Drederick – Presuming the 1:1,000,000 incidence, then … true, it could generate 9,998 false positives and 1 false negative. Is that really worth splitting hairs, over?

  8. Yep, you’re right, it’s not worth worrying about.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: