Steve Hsu writes about his recent research predicting height from DNA with stunning accuracy:
Had we taken a poll on the eve of releasing our biorxiv article, I suspect 90+ percent of genomics researchers would have said that ~1 inch accuracy in predicted human height from genotype alone was impossible.

Source: http://infoproc.blogspot.com/2018/08/scientists-of-stature.html
In the comment section I wrote the following:
This is absolutely brilliant research and in some ways it’s even more important than predicting IQ because height is a person’s most salient property and the most important datum police use in finding suspects.
Now that you have proven you can strongly predict height within a specific time and place (21st century Western countries), the next step is predicting the CAUSAL component of height, by building an algorithm that can predict height in ANY time and place. The scientific standard for proving A causes B seems to be that A precedes B in time and A correlates with B in all times and places. We know for example that the Y chromosome causes height because it precedes the development of height in time and because people with Y chromosomes are taller than those without it in virtually all countries and all centuries, so we know it’s not just some local interaction causing a spurious correlation.
So it would be even more fascinating and useful if your team had the computer learn how to predict heights on a global sample (i.e. a mix of people living in Europe, Asia, and Africa) since you have access to data from all of these continents. Now there would be a much larger margin of error because of the enormous differences in health and nutrition, but to help with this, I would enter the Human Development Index (HDI) of the country each individual was from.
The result would be an algorithm that could predict the height of anyone anywhere using just SNPs and HDI and because the predictor would work everywhere and everywhen, it would not just be correlative, but causal. It would prove the genome and HDI each have an INDEPENDENT (i.e. causal) effect on height. The advantage of causal predictors is not just that it would debunk the skeptics, but it might even allow you to predict the height someone living 40,000 years ago would have had he been born and raised in 21st century America (by entering modern America’s HDI into the formula)
The same type of study could also be done for IQ if you had a highly g loaded IQ test that was sufficiently culture reduced to compare people in First and Third World countries. Such a universal algorithm might even be used to estimate the IQ Neanderthals, Cromagnons or Denisovans would have if they were born and raised in 21st century America (assuming comparable genomic architecture)