On Monday, February 23, 2004, at 01:10 PM, Eugene Salamin wrote:
Measure each output beam with a detector that can detect individual photons. For each horizontally polarized photon, output a zero; for each vertically polarized photon, output a one. Now you have a random bit stream.
These exist as devices to plug into your USB port. Um, here's one: http://www.gapoptic.unige.ch/Prototypes/QRNG/ I wonder whether that's easier than using radioactive decay, the other obvious way that the laws of quantum mechanics give us genuine randomness?
Because of alignment and material imperfections, the probability of 0 and 1 cannot be made precisely 1/2.
There are lots of ways to make "almost random" streams better. If you really have independent, identially distributed Bernoulli trials and the only worry is that p=1/2+epsilon, the easiest remedy is Von Neumann's: look at generated bits two at a time, and apply "10" -> 0 "01" -> 1 throw away "11" and "00". This cuts your output rate in half (-2epsilon^2), but if there's really no short-term correlation, you get true randomness. --Michael Kleber kleber@brandeis.edu