Chaoskey presentation at LCA2017

[bsr]

Very good presentation but I found Keith's comments about testing for randomness a bit disconcerting, specifically how the noisy but non-random image containing Pi passes DieHarder testing.  When I look at the noisy image on the right I see three dots with three vertical traces and also a weak horizontal line, although he says there are two dots.  The point is that the image is clearly not random upon a quick visual inspection and yet he says when he ran the image through DieHarder it said, "Oh yes, that's completely random".  Are there any other (better?) software tools that exist for testing/verifying randomness?  It seems to me that this a catch-22.  Short of building another hardware entropy generator that we accept as truly random (a reference standard to compare against) how can we verify true randomness?  And how would we know our chosen rng standard is truly random to begin with?  It seems hardware measuring decay would be truly random but I'm dubious about a fish tank standard - my black skirt tetras are quite territorial ;)