The genetics of intelligence

I am running into a cultural conflict here, because on one hand the sort of people who don't like meⁱ eagerly agree with me that the "genetic basis of intelligence" is nude and crude pseudoscience, whereas the sort of people who do like meⁱⁱ tend to believe there's something there, if for no other reason than because it bothers so very much the other set ; or else for historical considerations - echoing the "learned" ideas presented by the slavery lobby throughout the 1800s ; or else because it intuitively seems to make sense ; or for any other reason.

So let's proceed by understanding genetics together, then looking at the broader context of our understanding, then making some reasonable predictions and formulating some reasonable expectations.

You probably "know" genetics like any other guy, which is to say like this particle physicist :

The other model Hariezer lays out is that magic lies on a single recessive gene. He reasons squibs have one dominant, non-magical version, and one recessive magical version of the gene. So of kids born to squibs, 1/4 will be wizards. In this version, you either have magic or you don’t, so if wizards married the non-magical, wizards themselves could become more rare, but the power of wizards won’t be diluted. Call this model 2.

There's nothing wrong with this model, outside of the presuppositions it makes. Whether you're a particle physicist or a computer engineer or anything else : do you not find it concerning that the man has a numeric value embedded in his theoretical description ? 1/4, seriously ? What is this magic 1/4 then, a universal constant ? A sort of k in the Van der Waals model ? Is it c, is it e ?

All theoretical textⁱⁱⁱ with numbers in it is suspect, and for good reason : the odds are there that the constant's actually a natural constant, one of the dozen or so magic numbers of the universe. But if the author's inept, the odds are much better that the constant's actually just a creation of his model, which immediately means that his model is fucking stupid. Yes, that strongly. In any case, the confusion of planes does not bode well for the general quality of the thinking process that produced the text. Numeric values in theory are not unlike "and then he was like Oh mai gawd Jennifer and I was totally like ..." interspersed in narrative. In fact, they're exactly the same.

In this case : the model of genetics based on a single alele pair is in the books because fucking peas exhibit it^iv, and because it's easy to teach to biologists, who are people who are bad at basic arithmetics. So you make it look like there's only two genes that can share a locus, and then the whole space becomes trivially enumerable^v : AA, Aa which is the same as aA, and aa. There's a little gotcha^vi, there's numbers and drawings and everything needed to make someone feel good about their high city lernin'. All that satisfies the libertards, form as a fine substitute for function.^vii The only thing absent - some actual understanding, but hey, we don't talk about that sort of stuff in this house.

Luckily we aren't in that house, we're in this house, a place where where you come to think, and to be relieved and protected from the overwhelming stupidity of outside. It's, as the definition goes, a University - something Maryland for instance aims for but does not actually have. And so in this space we can consider some actual learning.

The human body contains 23 chromosomes. These 23 chromosomes aren't all identical, but by and large provide space for something to the tune of one thousand or so genes. These spaces are called loci, plural of locus which just means space in Latin. They are referred to as "genes" by uneducated authors, which is a confusion comparable to counting the capacity of a parking lot in cars. It's true that for all practical purposes you're interested in the car count. Nevertheless, the parking lot is made out of parking spaces, not out of cars, and when you're teaching parking lotness this theoretical fundament is important. Yes, the cars do occupy those spaces. But thinking of a parking lot as a collection of cars rather than a collection of parking spaces empowers a misunderstanding bred by the previous reduction of the genetic model to "single alele pair" above.

So now : let's, for the sake of being idiots, presume that out of the 23 chromosomes weighing in at 1k spaces each, exactly one encodes intelligence (and let's also presume it's not the 23rd, because that'd really be fucking painful politically)^viii. This one magical space will then be occupied by genes. These genes will, through occupying the locus in question, form a genotype, which will manifest visibly as a fenotype.

Let's, for the sake of being idiots, presume that there exists a statistical measure of one characteristic this fenotype, which is called intelligence. Let's presume we actually have definitions for all these words, which we most pointedly do not, and have models and errything. How would that model go ?

For instance : in one point of your head, you have this firm, unexamined conviction that human ability is infinite. There isn't a maximal intelligence to be had, there's no absolute maximum as a matter of the rules of the universe. There is, for instance, a maximal speed : nothing in existence^ix can go faster than c, but nevertheless, human ability is unbounded. You believe this. At the same time, you also believe human stupidity is bounded. There's a 0 somewhere, you can't actually go below that zero.

Meanwhile, in another point of your head, which apparently doesn't ever meet the first, you have this firm, unexamined conviction that human ability is correctly represented by a Gauss distribution. Like this :

That's the right drawing, isn't it ? You've seen it a thousand biliard times and therefore it must be correct, right ? And besides, it makes sense, it's called the "normal distribution", and people are normal. It's the distribution for most natural things, and humans are natural, so... yeah.

Ok. Do you notice that the fucking distribution is bounded on both ends ? It goes from -4 to +4, yo. Yes, you can have unbounded natural distributions, of course you can. On. Both. Ends.

Unless you're capable of coming up with some room for infinite stupidity in your model, you can't at the same time have the gaussian and the unbounded human potential, you must pick exactly one. Never thought about that, have you ? Why not ?

But let's presume you've somehow solved this difficulty, and now you're looking at an actual distribution of an actual measure of actual intelligence as characteristic of a bunch of fenotypes, god help you. What you wish to do with this is to experimentally pinpoint which gene combinations on the locus do what exactly to the fenotype as understood through this single measure you came up with.

Obviously being too precise won't work on the first pass, you can't conceivably make a full description, so remembering your analysis classes (which you somehow took in spite of not managing to do arithmetics in your head, or understanding the basics of statistics^x ) you decide to look at the limit. What exactly causes the topmost performance ?

Let's, for the sake of argument, propose that there's a total of 13 genes that could fit on that locus we're looking at that are also dominant^xi, note them A, B, C, D, E, F, G, H, I, J, K, L. Let's also propose there's a total of 23 genes that could fit on that locus that are recessive, note them a, b, c, d, e, f, g, h, i, j, k, l, m, n, o, p, q, r, s, t, u, v, w.

What happens when these mix ? Holy hell, don't ask. KC for instance makes the kid grow eyes on the back of his head. Jg makes this hand conformation^xii :

But whatever, you know this is the locus of intelligence, in spite all conceivable evidence, and so onward we trudge.

So, there's a grand total of (16 + 23) * (16 + 23 - 1) = 1`482 unique combinations of genes you could be confronted with. Question : if a sack has 39 differently marked balls in it, and you get to pick two at a time, how many picks do you need to make before you've seen all ball combinations at least once ?

And the answer to that question lies before door #∞. If you must be absolutely certain you've seen all combinations, you must make an infinity of picks. So what's your tolerance to risk ? Suppose we say 99% is good enough. That means you have a 1% chance to miss a combo, and that combo could be exactly the genius one. You don't know. Which is the problem here : when you're trying to model exceptional things, such as high performance or other catastrophes, you can't really rely on tolerances. Taleb has written extensively about this, go read him.

But let's, for the sake of still being idiots^xiii come up with a numeric solution to a theoretical problem. We'll tolerate a billion to one odds of missing out a group. So how many picks do we need then ?

That's not so hard to figure out. The odds of a specified pair not coming out in one try would be 1`481/1`482, or 99.93%. In two tries, it goes down to a shocking 99.86%. In a thousand tries it's about even odds, which means that if you look at 1`000 different people, you have about 50% chance to have seen any particular genotype. In 10`000 tries it fades to 0.117%, so you can almost claim 99.9% certainty you've seen each genotype at least once - except that wouldn't actually be correct. Each individual combination has a 99.88% chance of having been seen by the time you've done 10`000 assays, but since there's 1`482 of them, there's still a 17.6% chance you've not seen one. If you do another 10`000 assays that finally drops to 0.21%, so what you can now say is this : if you look at 20`000 people, you can be 99% confident you've seen each individual genotype at least once.

What if you need more than one instance to do anything useful with it ? Multiply the assays.^xiv What if 99% confidence isn't really close enough to "one in a billion" for your comfort ? Multiply the assays.^xv What if the balls-in-a-sack model isn't really useful for genetics, seeing how they are intertwined with society and human choice, and so you end up with selected pockets everywhere ? Multiply the assays.^xvi

By now your model (single locus, few genes) is so demanding you actually need to study more people than are alive. Let's not remind you that it's unlikely this is how it works - as best we can tell multiple loci on multiple chromosomes are involved, not to mention an entire dazzling array of indirectly genetic complexity.^xvii Let's not even get into a discussion about how it's unlikely but still perfectly possible that peak performance is actually the result of a trisomy^xviii (god knows the bottom performance often is), and you're lost.

The human genome was sequenced a decade ago, the science junklets posing as "peer reviewed journals" regularly publish "studies" done on three digit populations and here you sit needing a few trilion samples just to get started. It's not clear there's ever been a few trillion humans alive to date in total, which brings the problem home : you're worse than the pseudoscientists with the "global warming" and "carbon"^xix charade. What they've done is, they've taken data from the past coupla decades, extrapolated it into the past, then pretended like their extrapolations are actual data, and put the dataset thus obtained through a process that produces the same results no matter what data you put into it. This they call "science", and they're getting away with it because it's politically useful, like any pseudoscience the quacks ever got away with.^xx

But their position is better than yours : the temperature at every spin since the Earth's spinning could in fact have been measured. There wasn't anyone there to do it, of course, but the temperature was present, and ready to be measured. Meanwhile you... you have to measure people that didn't even exist yet! If we invent time travel, the Earth pseudo-scientists could methodically go back day by day and become actual scientists. You couldn't, even with a time machine, even if you go back to every moment and measure every soul that ever lived^xxi you still wouldn't have enough, and probably by a very fat margin. You likely wouldn't have enough to get within 1% of having had enough.

That's what "the genetics of intelligence" would be like. It's perhaps possible, eventually. If it is or if it isn't - what you're doing isn't related. It's not even science. It's just nonsense, except on the other side of the political isle from the "global warming" quacks.

So no, it's not that I don't "believe in a genetic explanation of intelligence". For one thing, I don't believe period. For the other thing, I actually know what explanations must look like. Here's a hint : it doesn't matter if they're convenient. It matters if they're sound.

Go do something useful with your time.

1. Shorthand : libertards.

   Longhand : the sort of idiot that believes that

   1. the problems of the individual are to be solved by the group ;
   2. form is an adequate substitute for function ;
   3. everything's one thing ;

   [↩]

2. Make your own definition. [↩]
3. Software is very much this, incidentally. [↩]
4. So it's experimentally verifiable, go grow some peas at home. Takes a coupla weeks and the results are usually striking. [↩]
5. Ever seen the inumerate do division ? It goes something like this : "divide 150 by 9". "Well... 141... 132... 123... 114... 105... 96... 87... 78... 69... 60... 51... 42... 33... 24... 15... 6... uhh 9.5. I mean 15.5. Or something."

   Can't fucking count, these people, and instead of beating them every day until they can count, you're teaching them genetics. Shame on you, seriously. [↩]

6. Hey, aA has 50% of the space, as opposed to the other two with 25% each, but that's because of the genetic commutativity so ha-ha how smart are we that we know things! notwithstanding that genetic commutativity is not quite like the hamiltonian-unitary relationship or anything [↩]
7. You've almost taught someone something, right ? Now they know about "genetics". So what if your shorthand approach to teaching them stuff that's actually correct has in fact produced in their heads more erroneous or just plain silly preconceived notions than anything useful or usable - that's not your business, is it ? You go to the hospital, get treated by someone who doesn't care about anything but "doing his job", suffer from it and then blame those asinine doctors who won't consider the broader context.

   Then you go back to school, and continue to teach the kids in exactly the manner that down the road has produced the doctors that crippled you. Why should you care, the bit you actually did was actually correct, nobody can say that this isn't how genetics work, at least some of the time. So what if kid decides that's how they work all the time ? Not your problem, good for him, he has good self-esteem, all the better.

   Idiot. [↩]

8. You see, I make presumptions too. Unlike you, I delineate mine, and plainly. This is the difference. [↩]
9. Mark my crafty qualification. [↩]
10. By the way : the "computer security is hard" folk aren't statisticians. It's pretty much the same thing. [↩]
11. This doesn't exactly jive with what you imagine dominant means, I'm sure, because you probably formed the mistaken notion out of the peas 1/4 nonsense that "there can only be one". This is false. [↩]
12. I'm cheating, Patau is a trisomy, but anyway. [↩]
13. "Oh George, you are so sweet. Don’t ever change." [↩]
14. Say you need what, like 1k samples of each type to meaningfully measure their intelligence ? Let's be generous and say 1k suffices. So you want... 20 million now. [↩]
15. Let's pretend like you go from one in a hundred to one in a billion by simple multiplication, your 20 mn are now 200 trillion. This isn't necessarily true - for all you know rare genes show up less often than that - note that at issue here aren't "all the genes that are", but "all the genes that could be". The simple fact of the matter is that even the question of "how many samples do we need to look at to be confident we've seen the whole diversity" is not ready to receive an answer just yet. [↩]
16. And I'll forever be thankful if you won't look at me to tell you by how much. [↩]
17. The entire "it's genetic!" thing completely misunderstands how genes yield fenotypes. In simple words : your genes don't encode "for blue eyes". They encode "for blue eyes if X Y Z nutrients are available", which is to say they're well written code, that fails gracefully (most of the time - which is to say that if genetics worked at the speed code writing works, genetics would not yet have failed ungracefully once). People genetically predisposed to grow tall don't magically overcome a lack of protein in their diet by that mystical power of "genegod hath so ordained", because that mystical power does not exist. [↩]
18. A pathological situation when a body has three instead of two copies of a gene. These are usually copies of either one or the other, so commonly it'd be aaF or Dbb or something - but they don't have to be! Trisomy can also work with a mangled copy of both genes, something like qCX, where X is made out of a fragment of q and a fragment of C. This doesn't usually result in a functional gene, but then again it very well might! How would you know ? (Multiply the assays!)

    The odds of a trisomy surviving in the first place are dismal, and adding to that the odds of a mixing being successful, which are yet again dismal, and all this happening on the previously heavy basis makes your numbers lose count of their zeros. Sorry. [↩]

19. Carbon dioxide. Carbon motherfucking dioxide!

    It's not just "carbon", you objectionable louts, and your confusion on this topic, and your inclination to herd following and comfort seeking is indicative of just how intellectually fucking useless you are. [↩]

20. In truth, these idiots have over the alchemists and the Richters of past the advantage of having to deal with a bureaucracy rather than an actual leader. The sultan may well decide to hang you on the third pass of failing to produce gold, but in decapitated America absolutely nobody ever hung for buying IBM. [↩]
21. Not just a cheek swab, mind you. You must account for chimeras. You must account for their actual intelligence, in languages and cultural contexts you don't even know! [↩]