G1(X>z) is so small) that even with a power of large n, (G1(X<=z))^n (remember G1(X<=z)<=1) is still larger than (G2(Y<=z))^m, i.e. we still have smaller probability to generate a genius. He takes the sentence "we still have smaller probability to generate a genius" as empirical fact, and IF n>m, then his conclusion is that G1(X<=z) is much much larger than G2(Y<=z)--i.e. for each individual chinese, the probability to be a genius is even smaller than we thought.
Of course, I am not sure whether "we still have smaller probability to generate a genius" is true, or whether the assumption of iid is true, or wether there is something else.
It's just an argument for fun. Not serious research. Personally I hate a to see such serious racist research--especially when it's so complicated and the truth is so hidden.