Lion of the Blogosphere writes:

There’s a study publixhed in Molecular Psychiatry.

The only major newspaper to write an article about it is The Telegraph and their paywall prevents me from reading what they have to say about it.

Currently, genetic testing can only reveals 7% of intelligence differences between people but I’m sure as research and methods in genetic sequencing and computer analysis improve, that will eventually become a much higher percent. And then we will also finally have proof that blacks are less intelligent than whites because of genes, and not discrimination or poverty.

I took a look at the study he’s citing and found a few relevant quotes:

Using our meta-analytic dataset on intelligence we carried out polygenic prediction into UK Biobank subsamples following their removal from the meta-analysis. Between 3.64 and 6.84% of phenotypic intelligence (as measured by the VNR Test in UK Biobank) could be predicted (Supplementary Table 10); the upper limit is an improvement of ~43% on the largest reported estimate to date, of 4.8%

So if I understand correctly, it sounds like 7% is the upper limit of some kind of margin of error, but seeing as the lower limit is still around 4%, the single best estimate remains around 5%.  However taking the square root of 5% tells us that genomic predictions correlate 0.22 with IQ, which is a weak (though not terrible) correlation.

However the good news for behavioral genetics is that the IQ test used in this study (the verbal-numerical reasoning test, abbreviated VNR) sounds shockingly bad:

The VNR test consists of 13 items, 6 verbal and 7 numerical questions, all of which are multiple choice. An individual’s verbal numerical reasoning score was measured by summing the number of correct responses given within a 2 minute time period.

Tests with only 13 items (scored right or wrong) almost never have high loadings on g (the general intelligence factor) because the reliability is too low.  My educated guess is that the VNR has a g loading of only 0.65.  Dividing the polygenetic predictive power (0.22) by the estimated g loading of the VNR (0.65), gives 0.34, which is a reasonable estimate of the genomic correlation with a hypothetically perfect measure of g.

A 0.34 correlation is still only moderate, but even modest correlations add up, because by the logic of regression, for every 1 standard deviation increase in the genomic score, general intelligence should increase 0.34 standard deviations on average (5 IQ points).  This is not trivial.  And I agree with Lion that predictive power will increase dramatically as the technology advances.

Of course none of this tells us anything about black-white IQ differences unless the races have been found to differ significantly on these genomic scores.

But of course as commenter Mug of Pee points out, all these predictions are in Western countries so the genotype-phenotype correlation could just be a local phenomenon and not reflect a truly independent genetic effect.  We have no idea whether these genomic scores would predict IQ in societies with radically different environments.