[Cryptography] How good random number generator is the human brain at conscious level?

Ray Dillinger bear at sonic.net
Mon Dec 18 00:39:36 EST 2017



On 12/16/2017 01:28 AM, Georgi Guninski wrote:
> How good random number generator is the human brain at conscious level?

> Searching the web returns forum speculations that it is bad PRNG 
> and some paywalled medical articles claiming otherwise.

> "Hardware" stuff like neurons firing or muscles moving doesn't count.

Pretty terrible.  When they asked humans for random numbers distributed
between 0 and 16, put one number in each byte, and compressed using
gzip, the output was 2.76 bits per byte input rather than the nominal 4
bits per byte input.  This simple empirical test means that an humble
statistical tool was able to eliminate 2.48 bits of redundant
information from every human-generated "random byte."

That's biases, repetitions, and patterns detectable by an off-the-shelf
compression program that DOES NOT detect all possible (nor even all
likely) biases, patterns, and repetitions.  So the 5.52 bits observed is
a maximum estimate. If you build a more exhaustive test, you can almost
certainly demonstrate an upperbound smaller than that.  Hell, the
current version of gzip likely demonstrates an upperbound smaller than
that; somebody ought to repeat the experiment.

				Bear

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: OpenPGP digital signature
URL: <http://www.metzdowd.com/pipermail/cryptography/attachments/20171217/08134d41/attachment.sig>


More information about the cryptography mailing list