There is a lot to say about IQ. I plan to make a video about it. It’s not my field, but I’ve been reading the literature on and off for 17 years. Recently, I have noticed an explosion in what we can (for the purpose of this post) call SecretSauce-ism which is adjacent to a “cult of genius” mindset, i.e. the idea that there is some secret genius juice that lends godlike credibility to a person. This is harmful, so I’ve been rereading the literature, and have over the past week spent about 50-100 hours refamiliarizing myself with the current literature.
It’s essential to know that IQ is primarily used to measure g which is a factor analysis of subdomains: the three primary are perceptual, verbal, and spatial. (Professional gamers would score high on perceptual.) What a lot of people don’t understand is that when people talk about genius or IQ they’re talking very broadly about highly conditional phenomena.
For example, IQ is predictive more downwards than it is upwards. There is a debate in IQ (SPLODR) which presupposes an IQ threshold after which there are diminishing returns or little benefit. This was originally posited at 120 which is trivially easy to refute because e.g. mathematicians have an average of 130. However, it’s much less certain if say 160 IQ will have a benefit over 145 IQ in any meaningful way. (160 is the ceiling on the WAIS IV. If someone says they have e.g. 172, someone used some kind of statistical reaching to get this number — like if you put all of the 4.0 students in a class and used that class’s grade curve to determine who has a “5.0″ GPA.)
To make an analogy, you are scoring the test by the rarity of people who get that score, not some kind of straightforward competence test like the math section of the SAT. If you make 20 free throws in a row, you’re probably pretty good at free throws. If you make 75 in a row, you’re really good. If you make 2000 in a row, you’re one of the few freaks who compete for world records — and the current world record is 5200, which is many SD beyond the 2000 scorer. What is this percentile difference measuring? Likewise with 3.91 GPA vs 4.0 GPA. And, not to get into literal dick measuring, but you would be surprised at how much rarer each .25in of erect penis length is, despite being the same addition of length with each increment.
So, several things are true, which a lot of people don’t know:
general dumbness correlates more than general intelligence, i.e. the subtests are more likely to intercorrelate with lower scores.
at higher scores, there is more likely to be a skew with some subtrait or set of subtraits, e.g. a person is much more likely to have a high spatial relative to verbal at high IQ, whereas this is more likely to be even around the average.
high IQ is less heritable than IQ in general. (going from memory, heritability was .76 for IQ in general and about .54 for high IQ.) This is intuitive if you think of IQ like diet: it’s more about what you avoid than what you add. (Genes that boost IQ and lower IQ are grouped into additive/nonadditive.) It’s also just a lot easier to get the wrong brain configuration than it is to land on the goldilocks zone.
the heritability of IQ is not identical to the midparent IQ correlation, which creates enough variability — especially with higher IQ — that selecting your partner based on what you think their IQ is may be foolish. the intellectual version of this can still happen, in both directions: https://i.imgur.com/oG0crQB.jpg
general intelligence as it applies to humans may have very little to do with general intelligence as it applies to AI. https://i.imgur.com/C87aP24.png (bouchard/johnson 2014).
finally, there is probably a “threshold effect”, also called SPLODR, which shows diminishing returns past a certain point. the original hypothesis was that this occurred at 120 IQ, but this is trivial to refute; it’s less clear that there isn’t a threshold effect for, say, 145IQ. (many reasons for this — one is that it doesn’t pass a sniff test. if there were utility in measuring beyond 160, test companies would do this; further, test companies that are highly invested in predicting professional success, like the creators of the GRE or LSAT, tend to have percentiles that cap at what would be the equivalent of 140-145 as well.)
this is among many other things. This is a toe-dip. But there are many misinformed beliefs about the construct.
There is a lot going on with Nobel prize winners. The most common trait is that they work extremely hard. There have been 40-something g-loaded subtasks that I know of. It’s quite possible that they have an exceptional “recipe” of sub-abilities from these elemental cognitive abilities that won’t show up on a traditional WAIS IV.
But this is to be expected; the primary purpose of IQ testing is (1) to measure cognitive decline or some other kind of medical concern and/or (2) to determine which children go into gifted programs. Adult IQ is rarely tested outside of these settings, yet it is also where people try to draw the most generalizations.
(The reason you can infer that IQ tests aren’t meant to be as much of a measure of ability as they are to do these other two things is because so few safeguards are put in place to prevent cheating. With enough money it is quite possible to retake an IQ test and get 140+; you can even google the answers if you want. They really don’t care. Meanwhile, the SAT is psychotic about cheating and the people who have successfully cheated had to pull off preposterous schemes to do it.)