Billy T said:
Your link did not respond in several tries, will try more later. Do they give any results relating to my question. I.e. given all the Pi sting in base 10 that is now known, what is the probably that it is not normal?
Hmm, funny, it's working for me. Here's an older one (less digits)
http://oldweb.cecm.sfu.ca/personal/jborwein/Kanada_50b.html
Try searching for Kanada and pi if this one doesn't work. Kanada is often at the lead in calculating many digits.
For the 1.2 trillion digits they give
119999636735, 120000035569, 120000620567, 119999716885, 120000114112, 119999710206, 119999941333, 119999740505, 120000830484 ,119999653604
as the number of 0's, 1's, 2's, etc, all pretty evenly distributed, and this gives a chi-squared of 13.13. I have no recollection of what a 'good' value for a chi squared test is, but the data looks pretty darn evenly distributed to me. I believe for a normal number we'd expect in n digits that you'd be no more than sqrt(n) away from the expected values. In this case, sqrt(n) is a little more than 1000000, and the actual distributions are certainly within this distance of 120 billion. There's more to normality than just single digit distributions though, you need to look at 2 digits, 3 digits, etc, all must be uniformly distrubuted in the sense of the asymptotic limit.
The Champernowne constant is 0.12345678910111213141516...., just the naturals in order, this is normal in base ten. see
http://mathworld.wolfram.com/ChampernowneConstant.html I'm quite certain it's not normal in all bases, but I can't recall if there's a proof for an explicit one. Of course you can do the same construction in any base.
Rational numbers are all not normal. In any base they will eventually repeat (possibly 0's).
There's an article somewhere online about using the digits of pi as a random number generator, I don't recall all the details and can't seem to find it right now*. No naturally occuring numbers have been proven to be normal, they are all 'artificial', that is constructed for the purpose of explicitly demonstrating a normal number. I would have my doubts that these artificial ones are terribly useful for this purpose, since there's definite structure in how the digits are laid out. Normal isn't demanding random so much as it's demanding a uniform distribution, which is how random should look in an asymptotic sense.
* press release about an attempt to use pi to generate random numbers
http://www.sciencedaily.com/releases/2005/04/050427094258.htm