It's also worth considering that your hypothetical might be anything but! At extremely high levels of intelligence the rate of extreme social disorders skyrockets up to "most" (I recall the cutoff for 50% happening at ~165 IQ, but it continues to rise even after that, limited primarily by our lack of data on IQs much higher). I'm not entirely sure if this analogy is appropriate, but just as organisms with more cells are more likely to produce cancer, it seems minds with more thoughts racing through them are just as damaged by some small percentage of errant thoughts (some are extreme delusions, some are just the product of the isolation of a brilliant mind growing tired of a world, retreating from it, losing touch with how to interact with excellence and subtlety, and seeing even less subtlety in the world, simply growing more tired of it, ending with many extremely intelligent people simply unable to interact with others in any normal way).

>...but also select for a mixture of cancer, OCD, suicidal depression, and plain batshit crazy?

1.3 billion bald, cat cradling, evil geniuses, all saying "mbwah, ha ha" in eerie synchronicity?

What about the assumption that intelligence can be reduced to an IQ number?

This. The generalization of intelligence to just a single number, or even a bunch of numbers, strikes me as a foolish over assessment of our current collective understanding of human intelligence.

So, let's say we have a number. It correlates with ability to solve a number of different puzzles, lines up with our understanding of certain dangers (iodine deficiency, etc.), is consistent, explains some variation of the success in certain tasks or areas that are not explained by experience or upbringing.

I'm not saying it captures every element of our experience, but surely we can point to it and say it represents some subset of our understanding of the idea we point at when we say the word "intelligence". It really isn't competing with anything that's remotely as useful. IQ, although it has faults, seems to be a relatively cheap, standard, well understood metric, that also explains quite a few other phenomenon. If you give me two regression analyses, one of them using IQ and one not, I cannot imagine what understanding you would gain by refusing to acknowledge the former -- and that mistake seems like precisely the type of "foolish over assessment of our current collective understanding" that you deride.

I'm not arguing that IQ doesn't measure anything, my point is simply that it is a one dimensional measure of something for which we do not know how many dimensions there actually are.

I score very highly on IQ tests, but I've worked with a great many people who I would guess do not score more than slightly above average but nevertheless bring a lot to the table that I cannot.

I guess my point is that using IQ as an authoritative measure of "intelligence" is missing the forest for the trees (well I suppose to best fit the metaphor it would be 'tree').