We know very well how to increase g: live in a more modern society. That’s what the Flynn effect is all about.

We also know that breastfeeding, having parents who read to children, etc., all help (see the paper I linked last time).

Shortcuts are harder to find, but in a lot of cases even the Stanford-Binet tests where they say “pick the odd one out” include a component that you can only really know reliably if you either happen to feel the same way as the test creator (e.g. number is more important than interior angle), or you have seen enough examples so you know what the “right” answer is in context.

Likewise, what is most important: that only one item is classified as a fruit when it comes to cuisine (“apple”)? That only one item is a non-reproductive body (“carrot”)? That only one item is from a monocot (“corn”)? It’s really hard to write these tests to be utterly unambiguous and to depend on zero knowledge. So I just don’t buy that they’re actually never trainable. Maybe the amount of training to make a measurable difference is impractical for the average person, though.

And the logical tests of the sort “if all foobles are wibbles, and some wibbles are bobbles, then are some bobbles foobles?” really benefit from explicit training in set theory and/or implication. Generally the tests don’t, I think, have all that many logical inference questions, but if you can’t train that part, goodness! What are logic courses for?!