math

With the right large algorithm the entire works of Shakespeare can be extracted from any random data. 

With the right small algorithm, e.g. understanding English, the entire works of Shakespeare can be extracted from the entire works of Shakespeare. 

In each case the entropy of the system is the same. Either you need an algorithm that can extract the works of Shakespeare from randomness, in which case the algorithm will have entropy equal to the works of Shakespeare itself, or the algorithm will be small (understand English) and the data will be patterned (Shakespeare).

So 

entropy(massive_algorithm(randomness)) = entropy(understand_english(works_of_shakespeare))

This is why the factoid that the entire works of Shakespeare can be found within the digits of pi is not as awesome as it initially seems. This is true for literally any data set in the universe, given the right algorithm. 

Either the information is in the data set and the algorithm is small, or the information is stored in the algorithm.