I guess the question is
"does an 'infinite' number of patterns imply 'every' number of patterns?"
For instance, if you could quickly calculate the decimal sequence of π, could you not (in theory of course) come up with an algorithm to search that sequence for some pre-determined sequence?
Then you could do this:
start = findInPi(sequence)
So "sequence" in theory could be a decimal representation of the movie "The Life of Pi". The implication is that all digital knowledge (past, present and future) is bound up in irrational numbers (not just the group of irrational numbers, but each irrational number), and we just need to know the index to pull data out.
Once you know the index and length of data, then you could simply pass this long.
playMovie(piSequence(start, length))
From an encryption standpoint, you could pass the start, length pair around, and the irrational number would be known only by the private key holder.
Am I off base here?
No, this is not the case for every irrational number. For example, the number
$$ 1.01001000100001000001000000100... $$
where each run of zeroes is one longer, is clearly irrational, since the decimal expansion never repeats. But it just as clearly doesn't contain every pattern of digits, because the only digits it contains are 0 and 1.
$\pi$ in particular is suspected (but not proved) to satisfy a stronger property, namely that it is normal, which means that not only does every pattern of digits occur, but every pattern occurs infinitely many times, with the frequency one would assume in a random string of digits.
In a certain technical sense, most numbers are normal, but there are very few expressions that have been proved to produce a normal number. This is a problem for your cryptography idea, because it is hard for the two parties to agree on a particular number that contains all of the messages they want to exchange.