[CuriouslyC]

I don't have the study handy and I don't feel like digging, but I seem to recall longitudinal comparison of IQ scores over long periods of time (something like 30 years) showing a correlation of around 0.7 or so. That's more than a standard deviation, though given the nonlinear scaling of IQ scores I couldn't say exactly how many points it equates to. One study in a shorter period showed swings as high as 21 points on tests performed by the same group. Clearly IQ can change, we just don't know how to control the process.

To the extent that you end up deriving your pleasure and livelihood from abstract symbol manipulation, IQ is important. IQ does matter a lot when you're first learning to do something, but as you master tasks the importance drops. A study on the correlation between chess performance and IQ showed that IQ only predicted chess ability in novices.

I agree that some people have natural proclivities that better match the structure of society. That is an unfortunate side effect of diversity under a fixed system.

I am not arguing that IQ isn't important, but is possible to improve it (regardless of our knowledge as to the process) and beyond a certain basic level it is much less important than creativity - a fact that is only going to become more true with computer automation.