No, that's not quite right. Randomly picking people from a distribution should eventually end up giving the original distribution. Taking n people from a pool

No, that's not quite right. Randomly picking people from a distribution should eventually end up giving the original distribution. Taking n people from a pool m times, then the means of those m samplings will be normally distributed however (given the constraints of the CLT of course).