Wouldn't the best way to do this be to ask questions related to genetic markers? You require 33 yes/no questions that independently divide the population in half, but has near-uniform distribution otherwise (each populace half has no relationship to the other questions).
Are there 33 genetic markers that each has no correlation on the presence of the others?