Re:Computing 10,000X more efficiently (2010) [pdf]
by samstave on Monday February 24, 2014 @08:40PM
(#7294649)
And these are precisely the things, which we know little, that the NSA may be exploiting with extreme effectiveness that scares the shit out of me.
I recall an article from 2005 that described a system that could track BULLETS IN MID FLIGHT over a vast swath, in-the DARK, which provided detailed view in all ballistics in a fighting theater....
Re:Computing 10,000X more efficiently (2010) [pdf]
by m_darkTemplar on Monday February 24, 2014 @10:23PM
(#7295193)
It's probably not of much use to the NSA where breaking encryption requires precise math that will fail quickly if operations are off by 1%.
Re:Computing 10,000X more efficiently (2010) [pdf]
by sophacles on Monday February 24, 2014 @11:15PM
(#7295486)
I can however see lots of application in various pattern analysis techniques based on machine learning. There's lots of places in those that a bit of fuzz won't really matter, at least for first pass filtering.
Similarly voice recognition would probably have to deal with far larger errors from the transmission and capture of sound anyway.
Just because it's not good for decryption, doesn't mean it isn't good for the overall set of NSA operations.
Re:Computing 10,000X more efficiently (2010) [pdf]
by klipt on Wednesday February 26, 2014 @02:48AM
(#7303581)
Yeah, even human brains can do voice recognition, and they're noisy as fuck.
Re:Computing 10,000X more efficiently (2010) [pdf]
by Houshalter on Tuesday February 25, 2014 @04:54AM
(#7296619)
For floating point operations sure, but if I understand it correctly it could also provide a speed advantage to other operations. It means 1% of the operations will be wrong, but what does that matter when it allows you to search 10,000 more keys. Perhaps you could get even more speed up with specialized hardware for decryption, and then buy rooms full of them.
http://www.defensetechbriefs.com/component/content/article/1...
http://proceedings.spiedigitallibrary.org/proceeding.aspx?ar...
From the paper: "The hardware is designed to perform high dynamic range arithmetic with roughly 1% error per operation. Singular has developed and studied varied algorithms for producing high quality results despite the approximate hardware. These studies use a perfect simulation of the accelerator’s arithmetic. Tasks that have been explored include summing thousands of numbers without accumulating error, k-nearest neighbor classification (KNN), foreground/background separation using Gaussian mixture models, iterative reconstruction tomography, deblurring using the Richardson-Lucy deconvolution algorithm, FFTs, radar processing, neural net learning, and other tasks. Most of these algorithms need slight adaptations to prevent cumulative effects of the 1% error, but with those adaptations all perform as desired."