Hacker News Reader

Comment by Animats | Hacker News Reader

That's useful when you want random bits from a pulse noise source, such as radioactive decay. Count pulses over a period of time, and take the last bit of the count. Do this twice. 10 => 1, 01 => 0, 00, 11 => retry.

It seems to be a corollary that you cannot get random bits at a guaranteed rate that way.

View on HN

Doesn't that assume the distribution at time 1 is also the distribution at time 2, which isn't entirely true since the source is already sightly less radioactive then?

That's a good point. The effect is small but nonzero.

Still works, with a slight variation: instead of taking two measurements, you can take four successive measurements [A,B,C,D] over equal periods of time, and then do the aforementioned procedure with the values A+D and B+C as input.

If the activity is decreasing at a linear rate, then A+D and B+C should have exactly the same distribution. In reality, they will be ever so slightly different because the decay is exponential rather than linear, but over short periods of time (compared to the half-life) the linear approximation is extremely good.

https://en.wikipedia.org/wiki/Thue%E2%80%93Morse_sequence#Eq...

Coin flips aren't fair either, because gravity will change slightly between successive flips.

OK, but if you toss the coin higher, relativistic time effects come into play.

And if you're flipping near the Earth's surface, there will be aerodynamic effects as well. Which is more streamlined: The Queen's side or the Platypus' side?

Reply on news.ycombinator.com