I take issue with pretty much all tests of potential. In my humble view, all they really do is measure how your experience, up to the point when the test is taken, has prepared you for whatever it is you're testing for. A sufficiently intelligent person should be able, given enough training, to become proficient in any area of expertise.
I have no hard data to back this up, save for my own observations.
I think the two are not as mutually exclusive as you present it to be.
Consider something like language acquisition for example; if you had a 30 year old that already knew 3 languages they would almost certainly perform better on an exam to test potential to learn new languages than a 30 year old that knew one language.
If I am reading you correctly you are saying "that's not actually a legitimate potential predictor; it just indicates what experience you have, not a true measure of your capability going forward", but in reality if you have only learned one language by the time you are 30, you are at a point where your gene pool is not a great predictor. By the time you are 30, your lack of previous experience has actually had a physical effect on your brain that prevents you from being as good at language acquisition going forward.
It's not merely the fact that you have experience in a particular area that improves your direct ability to answer questions; lacking experience actually makes it more difficult to acquire the same levels of experience.
In other words, having experience makes you more able to gain experience; a test that measures your experience is a legitimate indicator of future learning potential. It's exactly along the line of colleges teaching you "how to learn" and any practical skills that you gain along the way are largely incidental.
But according to the article (and it's also pasted [here](http://news.ycombinator.com/item?id=2903166)), they've identified something independent of experience and preparation.
A sufficiently intelligent person should be able, given enough training, to become proficient in any area of expertise.
Indeed! Combine several tests of mental potential into an omnibus test, and run a large group of people through the omnibus. If you run the results through common factor analysis, you will find that getting a good score on one subtest is highly correlated to getting a good score on each of the other subtests. Psychologists have named the common underlying cause g, the general intelligence factor. (Scaled for age and population variation, g becomes an IQ score.)
I have no hard data to back this up, save for my own observations.