Beware of geeks making absurd technical claims (and a comment or two on NSA, PRISM etc)

The following recently arrived in my email extolling the merits of what, we shall see momentarily, is a truly extraordinary computational forecasting system:

The critical phrase here is

The simulations produced more than one vigintillion (that’s 10 to the 63rd power) combinations of unique interactions between the groups, Bagherpour said.

Let’s do a little math.[1] Assume—and this is a ridiculously optimistic number—that their software can generate and evaluate those interactions a billion times faster than IBM’s “Deep Thought” chess computer could generate and evaluate chess positions (a half-billion per second). Then evaluating those relations would take a mere 10^45 seconds.[2] By comparison, the age of the universe is roughly 10^19 seconds, give or take a few hundred thousand millennia…so we are missing a mere factor of 10^26 here.

Oh, but there is probably actually very simple explanation [4]: the company doing the analysis simply instantiates 10^26 parallel universes and time-travels to the start and end of each one.[5] Sort of like Map-Reduce. But, sorry, the details are proprietary. Oh, and here’s our invoice: please send the funds to our account in the Cayman Islands. And there’s going to be just a slight delay before we can provide any more results: please be patient.

[Program manager quickly shuffles off to elevator, clutching suitcase stuffed with $100-bills, hoping to restrain an outburst of maniacal laughter until reaching the waiting limo.]

These sorts of things are, of course, the norm rather than the exception in this arena. Your tax-dollars at work, and let’s cut the food-stamps going to all those free-loading children in Appalachia!

Enough said.

Ha!—just kidding. Obviously I need to say at least a few things about the recent NSA surveillance revelations. In no particular order

  • Dianne Feinstein, John McCain et al assuring us that these methods are absolutely vital or shortly after 2001 every Middlesex village and farm [3] would have been forced to implement sharia law in Chinese: “We’ve had classified briefings: these people are very thorough and professional. Why they’ve checked 10 to the 63rd power different interactions…”
  • There was a guy in my dorm at Indiana University in 1970 who belonged to one of those Trotskyite student socialist movements who spent most of their time fighting with other Trotskyite student socialist movements over microscopically subtle nuances of ca. 1930s ideology. But they did really know how to properly organize an anti-war demonstration. He was convinced their group was being infiltrated and monitored by the government. We were all convinced he was paranoid. Years later—this was in the post-Watergate investigations on domestic spying—it emerged that he had been completely correct.
  • Re: PRISM—so let me get this straight: People are getting upset that the government is spying on us using information that Facebook, Google and Apple are already using to spy on us? Oh, but Facebook has Farmville, Google says they aren’t evil, and Steve Jobs always wore black turtlenecks and was cool: that makes it all okay.
  • A thousand Boomer managers read with total disgust Edward Snowden’s weekend revelation that he leaked the memo, having had every intention of coming into the office on Monday and firing at least one irritating twenty-something on suspicion of the crime.

First Bradley Manning, now Snowden, and there’s 800,000 other people out there with security clearances, and every year the number with Cold-War-era unquestioning loyalties declines. Add to this the apparent tendency of younger techies to be disproportionately libertarian and, well, much as Bush/Obama seem intent on creating a high-tech Stasi in the U.S.-of-A, methinks they are going to hit—they already have—some serious limitations. Not like that Stasi thing turned out very well for the DDR either.


1. Which the writer of this piece either thought unnecessary, or lacked the ability to do.

2. They will say, of course “Oh, we don’t have to look at all of those, we filter them.” But even to filter these you have to at least read each one into memory, which takes a decidedly non-zero amount of time, and certainly orders of magnitude more than the 10^-18 sec I’m allowing here. Or, alternatively, extremely broad categories were eliminated without looking at the individual cases. In which case you aren’t, in fact, looking at 10^63 cases. But we knew that.

By the way, there’s a word for that approach: “lying.”

3 Google it.

4. This part I’m making up. I don’t have a wild enough imagination to make up the original quote or web site.

5. Okay, okay, so that 10^26 parallel universes number isn’t correct. Let’s still give them the imaginary 10^18 evaluations/sec computer, and give them about twelve days (10^6 seconds = 11.57 days, plus the remainder for preparing PowerPoint slides. Yellow on green) to do the calculations. Then they actually need to instantiate 10^39 parallel universes. Which may or may not be harder than instantiating 10^26.

But I’m certainly over-estimating the speed of the hypothetical machine. According to Wikipedia, the fastest existing computer runs at about 10^17 floating-point operations per second. I do not know the company’s generate/evaluate algorithm—beyond

  1. Find gullible individual with access to government money
  2. Promise whatever it takes to get the money
  3. Take the money
  4. Disappear
  5. GOTO [1]

but it would certainly require substantially more time, in memory access alone, than 10^-1 FLOP.  From this point we could embark on a Randall Munroe “What If” and demonstrate that the required energy would reduce the machine, and the entire DC metropolitan area surrounding Ft. Meade, to a glowing mass of gamma-ray emitting slag that would quickly sink to the center of the earth…or whatever else you can do with numbers like 10^39, which is pretty much anything…but…isn’t EPSA next week? Didn’t I promise I’d present a paper?

This entry was posted in Methodology, Politics. Bookmark the permalink.

2 Responses to Beware of geeks making absurd technical claims (and a comment or two on NSA, PRISM etc)

  1. womanstats says:

    LOL . . . more seriously, I knew the wind had changed when my high school kid was asked to write a fictional story for his English class, and it turned out Google was the villain . . .

  2. Ralph Hitchens says:

    Thanks to Duck of Minerva for pointing me here. This was amusing & insightful. Re. the hundreds of thousands of us with TS/SCI security clearances, I have long maintained that in the national security business, rogue insiders are an incidental cost of doing business. They simply don’t surface with a statistically significant frequency, if you factor the rogues against the total number holding clearances, and with few exceptions their disclosures of classified information have little long-term impact. (A possible exception would have been the Walker family back in the 1980s, and then only if the Cold War had turned Hot.) Intuitively I think “trusted insiders” are probably a bigger problem in the corporate world, where loyalties may be shallower than the basic patriotism with which most of us are endowed.

    But I am constantly amazed that readily-available network internal security monitoring tools seem not to be getting much use in our classified networks. SIPRNet had download monitoring tools as well as the ability to block i/o ports in workstations, but young Bradley Manning was allowed free rein to do as he pleased — why? There ought to be some network admins breaking rocks with him at Leavenworth, assuming it comes to that. And the private sector certainly uses such tools; only the other day my VISA overlords notified me of an anomaly in usage and briefly suspended my use of the card. Why aren’t similar tools available on classified networks to note anomalies such as a sysadmin accessing or downloading certain very sensitive documents?

    Thanks also for helping with the math — 10 to the 63rd indeed!

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