A child learns much more and more deeply about language from just a fraction of the amount of unsupervised data. The point is that the mechanisms are entirely different, it's not very useful to compare.