Taleb's argumentation relies on fictional scenarios whose conclusion do not match what he would have gotten looking at raw data instead.
Caloric intake in children correlates to height very well below the median(+), and then it stops. You can't grow arbitarily tall by eating more food, but if you underfeed a child enough, it will end up shorter than its more-fed siblings.
In case you need some justificaiton for this intuitive claim, here's some tangential data: https://www.scientificamerican.com/article/how-much-of-human...
" In contrast, in developing countries, nutrition deficits lead to a lower heritability. The fact that the mean height of the U.S. population has almost plateaued in the past decade suggests that the nutrient environment has almost maximized the genetic potential of height, at least in this country. Improved nutrition elsewhere may have similar benefits in terms of stature."
(+) I use median here to mean some unspeicified percentile, as you did.
Here is a good explanation: https://twitter.com/sean_a_mcclure/status/115548736901923225...
But you brought up the same point as the parent. 0 IQ correlates perfectly with death. Really low IQs correlate perfectly with mental disabilities. High IQs correlate (almost) perfectly with socio-economic status.
https://www.ncbi.nlm.nih.gov/pubmed/8162884 https://ehp.niehs.nih.gov/doi/full/10.1289/ehp.7688
Taleb argues precisely that IQ is only a good measure of unintelligence, not intelligence, and that this prevent IQ from being meaningful. He showed that if you IQ test 10K people, 2K are dead (0 IQ) and then test their performance on something, you'll still end up with correlations that'll get you published in prestigious journals, and peddle nonsense conclusions. https://twitter.com/nntaleb/status/1115193783145123840
(the point being that in that scenario, 0 IQ perfectly correlated with 0 performance, can make the population correlation look significant even when "alive" IQ is completely uncorrelated with performance, which, in that example, it is)