Myers’ Race Car Versus The General Fitness Factor

[Epistemic status: I am not a geneticist, and even the geneticists I know aren’t sure about a lot of this. Take as speculation only.]

I.

PZ Myers argues against Stephen Hsu’s genetic engineering proposal here – a disappointing attitude toward mad science for a guy whose blog header is a crocodile-duck hybrid. The piece has a lot of errors, the worst of which other people have already discussed – but I want to talk about what I think is its strongest point. Myers writes:

Note [Hsu’s] estimate of the number of genes that contribute to IQ: 10,000. That’s half the human genome! Hmmm. I wonder if any of those genes play a role in other processes in human physiology that might be affected by his plan?

Here’s an analogy for you: let’s say a novice car designer has decided that the one quality of an automobile that is most important is speed, raw speed. He doesn’t know much about cars, so he asks more qualified engineers about what elements of the car contribute to acceleration and velocity, and they start off with the obvious…details of the engine, fuel mixes, etc. Then they’re talking tires. Aerodynamics. Weight. Pretty soon they have to admit that just about everything in the car is going to affect the speed at which it travels.

So our blithe designer decides that making a fast car is simple: we just look at each component of the car one by one, and we pick an available option for it entirely on the basis of which option makes the car go faster. We’ll easily be able to make a car that can rocket along at a thousand miles an hour, he thinks.

But we have to ask whether we would want a car where the seats and steering were optimized for speed, where safety options were discarded, where something like visibility or reliability were jettisoned for the sole virtue of going really fast.

This makes a lot of sense. A car in which every component was optimized for speed would probably be uncomfortable, unsafe, ugly, difficult to maintain, and otherwise not the sort of car you want to drive. A human in which every component was optimized for intelligence might well be unhealthy, ugly, physically weak, antisocial, et cetera.

And we can do more than just hand-wave at an analogy to cars. Natural selection constantly weeds out worse alleles and replaces them with better ones. If an allele increases intelligence enough to improve reproductive fitness, but has no negative effects, it should sweep across all human populations in an amount of time proportional to its fitness benefit¹. We see genetic evidence of various alleles sweeping various human populations in both the distant and recent past. These do not include the intelligence-boosting alleles Hsu is talking about.

For example, Hsu cites Rietvald et al‘s finding that rs1487441 is linked to cognitive ability (though it only gives you 0.3 extra IQ points, typical of the generally unimpressive contribution of single genes). About 20% of both Europeans and Japanese have the (A:A) variant, which suggests that however many thousands of years it’s been since Europeans and Japanese diverged from each other isn’t long enough for the gene to undergo much selection. That means neither allele can have any overwhelming advantage, which means there must be some reason to have the opposite allele worth as much as 0.3 IQ points. I think this is the rigorous version of what Myers is saying.

Despite the fact that the race car argument makes perfect sense both analogically and rigorously, it seems to be wrong.

II.

What are the sorts of things we might trade off against intelligence? Perhaps fitness, height, attractiveness, health, longevity, social well-adjustedness?

But in fact none of these trade off against intelligence, many are strongly positively associated with it, and in some the link has been proven genetic!

People with high IQ tend to live longer. For example, a person with IQ 115 (85th percentile) is 20% more likely to survive to age 76 than an average person with IQ 100. One can of course posit many possible connections. Maybe high-IQ people are smart enough to eat healthy and exercise. Maybe rich people can afford both good schools and good doctors. Maybe good health behaviors protect the brain as well as the body and increase both IQ and longevity. But further investigation has cast doubt on all of these theories and strongly supports the hypothesis that no, the same genes that give you high intelligence also make you live longer. See for example the International Journal of Epidemiology: The Link Between Intelligence And Lifespan Is Mostly Genetic, which find genetics explain 95% of the correlation. A few of the genes linking intelligence and longevity may be already known; SSADH seems to be a contributing factor. My favorite study in this area, though, is one that is not yet complete: since all mammals are basically the same [citation needed] some London School of Economics researchers have developed an IQ test for dogs in the hope of checking whether the same correlation applies to them. Since canine intelligence doesn’t affect things like diet, exercise, or tobacco status, a positive correlation in them too would help solidify the finding. We’re still waiting on those results, but even without them the genetic hypothesis is looking pretty strong.

People with high IQ tend to be taller. This is interesting since height is often used as a measure of health and fitness during childhood, and since taller people get a bunch of advantages including being rated as more attractive and earning higher income. Once again we can imagine all sorts of possible confounders; once again studies find that the link is genetic. See for example Common Genetic Variants Explain The Majority Of The Correlation Between Height And Intelligence, The Genetic Correlation Between Height And IQ: Shared Genes Or Assortative Mating, Resolving The Genetic And Environmental Sources Of The Correlation Between Height And Intelligence, On The Height-Intelligence Correlation.

People with high IQ may be more attractive. This is the conclusion of a meta-analysis that finds a positive correlation between intelligence and body symmetry, usually used as a proxy for attractiveness unaffected by things like hairstyle and cosmetics; a second study failed to find this relationship. The jury is out on the positive link, but there certainly isn’t the negative link that Myers’ race car would predict.

People with high IQ commit much less crime – which is going to be our measure for social well-adjustedness here since it’s well studied. Once again it’s easy to think of possible confounders (I’ll add lead levels to the usual lot). Once again the studies show that at least some of the effect is genetic – here’s one on low-IQ/antisocial-behavior correlation in children, and here’s one cleverly linking fathers’ criminal history to sons’ vs. nephews’ IQ and then throwing enough statistics at it to find that the relationship is genetic. Likewise, the relationship between high IQ and low drug abuse seems to be genetic as well.

People with high IQ tend to be more physically fit. This is the conclusion of a study of 1.2 million Swedes. I don’t have any strong evidence that this relationship is genetically mediated (although Gottfredson on the fitness factor may be relevant here), I just want to note that, once again, there is less than zero evidence for Myers’ race car hypothesis.

People with high IQ have lower rates of heart disease, stroke, circulatory diseases, and diabetes. Intelligence may or may not decrease cancer risk, but again contra the race car hypothesis, it certainly does not increase it. Sibling designs suggest that shared family environment during youth is not responsible for the benefits; differential socioeconomic status as an adult may be, but this status is itself likely caused be the intelligence differences.

So despite its apparent plausibility the race car hypothesis crashes and burns.

III.

In one sense, this is bizarre. It’s as if somebody optimized every part of a race car for speed, and found that by coincidence this also made it the safest, most comfortable, and cheapest car on the road.

Yet if we think about it some more maybe it’s not too surprising. Consider Niels Bohr. He was a Nobel Prize winning physicist, professional football player, activist who helped Jews flee the Nazis, loving father of six children, and so healthy he kept doing science well into his seventies. And his talents show every sign of being at least partly genetic – his brother Harald was a leading mathematician, anti-Nazi activist, and Olympic silver medalist; one of Bohr’s children was also a Nobel Prize winning physicist and another was also an Olympic athlete. So it’s obviously possible to design a human with all-around great genes. Why does evolution restrict such designs to the Bohr family?

I can think of a few possibilities, all of which people who know more than I do are welcome to shoot down.

First, some of these all-around-beneficial genes could be good in heterozygosity but bad in homozygosity. We know something similar is true in the case of sickle-cell anaemia, which is mostly good in heterozygosity (protects vs. malaria) and very bad in homozygosity (causes sickle cell). This is exactly the sort of gene that should exist at a constant low frequency in the population, never getting more or less common. If the frequency got too low, then there’s no risk of two carriers mating, so evolution would encourage it as a free disease cure. If it got too high, evolution would discourage it – any carrier would probably marry another carrier and give their kids sickle cell. Suppose there are ten such genes, each of which grants higher intelligence on heterozygosity and has a frequency of 10% in the population. The average person will on average carry one such gene and have a 10% chance of a horrible genetic disease. Maybe Niels Bohr lucked out and carried all ten such genes without going homozygous on any. Maybe some other poor guy who is lost to history got all ten genes homozygous and died at birth of ten horrible genetic diseases at once.

This would make Hsu’s gene-editing project very promising; all he would need to do is give everybody one copy of the relevant genes (and then never let them mate). But the hypothesis can’t be quite right: I think it would predict that Niels Bohr’s children would have unusually high rates of genetic diseases. In fact, the children of great men regress to the mean a little bit but show no signs at all of being unusually cursed.

Second, we could be talking not about polymorphisms but about mutational load. That means that there’s some genome that works for humans (plus or minus a few hundred thousand polymorphisms that aren’t too important at this level of analysis) and genetic health is determined by how many detrimental mutations you and your parents randomly accreted. If your mother spent too much time near the local nuclear reactor when she was pregnant, maybe you get a few hundred extra mutations and end up with lower IQ, a worse heart, less attractive features, et cetera. And this is obviously true in the case of a literal nuclear reactor, but I’m having trouble figuring out what plays the reactor role in real life. I know Greg Cochran and others have talked about things like paternal age at conception, climate, et cetera, but he applies these only to differences between populations. I’m not sure whether it would work out to expect a big difference in mutational load between Niels Bohr and his underachieving next-door neighbor. Maybe Bohr came from a long line of people who lucked out and got hit by unusually few cosmic rays? I don’t know if this makes sense or not. Part of my problem might be that I still don’t really understand how mutational load ever decreases – I’ve heard “the most heavily-loaded people are weeded out by natural selection”, but it seems like that should only be able to slow the gradual universal dysgenesis.

This would also bode well for Hsu’s project. In fact, it would make it even easier; it would reduce to the modal genome idea (make a baby whose genome, at each location, has the nucleotide which is most common at that location among all humans worldwide) which could be done without even performing the groundwork to see which genes do or don’t influence intelligence.

Third, maybe all of these other good things are trading off against things that were important in the environment of evolutionary adaptedness but not today. Greg Cochran brings up infectious disease resistance; some commenters bring up calorie requirements. This latter seems especially plausible; the brain uses a lot of energy and energy was a scarce resource through much of evolutionary history. Either of these would explain why evolution kept the seemingly detrimental version around for so long, and why right now in our low-infection high-calorie modern civilization one allele or another seems to be an unalloyed good.

Again, this would bode well for Hsu’s project, although the supergeniuses so produced would probably want to stay well away from any malarial swamps.

Or maybe it is some mixture of all four possibilities – Myers’ trade-offs, heterozygosity, mutational load, and disease burden. The latter three could provide a sufficiently positive effect for intelligence to hide the negative effect of the first. I would be really surprised if something like this wasn’t true – the theoretical argument for the first seems compelling, and it ought to happen at least a little even if it isn’t the main source driving intelligence differences.

I was going to write that in this case we’d have to sort through every intelligence gene one-by-one to make sure we weren’t getting one of the trade-off ones, but maybe this isn’t true – a genius designed by Hsu’s method should have on average the same number of trade-offs versus unalloyed-goods as a genius born normally (right? or am I missing something?). Since most of us would prefer a natural-born baby with IQ 150 to a natural-born baby with IQ 100, it seems whatever trade-offs are necessary to reach that point are widely considered worth it. So unless there’s a difference I’m missing between normal recombination and Hsu’s method, we should be okay with designing an IQ 150 baby as well (from a purely health-related perspective, at least).

Whether this generalizes to creating an IQ IQ 200 or 300 baby depends not just on ethics, but on whether for some reason the costs and trade-offs of intelligence compound more than linearly. It’s possible, for example, that there are ten different genes coding for something that protects heart health, all of which can be traded off against intelligence. If you switch one gene from heart health to intelligence, whatever, you still have nine genes protecting your heart. But switching all of them at once would be a bad idea. I don’t know any reason to think this is true, but it’s a possibility that might give us pause between the IQ 150 and the IQ 300 level.

Overall though, I think the race car idea, despite its plausibility, is likely to be less of an impediment to genetic engineering than it might seem.

Only One Footnote, But It’s Really Long

1. This is an assumption I’m granting for the sake of argument, but possibly not true at the margin.

Consider that evolution doesn’t care about intelligence nearly as much as we do. The most recent common ancestor of Europeans and Japanese wasn’t going to use her intelligence to design a mammoth-seeking rocket. In fact, it’s not totally clear why humans did evolve intelligence before the modern age; sure, tools are nice, but early hominids stuck to the same tools for a million years at a stretch; that doesn’t exactly give a tight feedback loop to work with. The most convincing argument I’ve heard is the Machiavellian intelligence hypothesis which says that our ancestors used intelligence to navigate tribal politics and gain status within a social group.

But this theory would naively predict that the smartest person in high school would be the most popular. If intelligence is for gaining status, it seems to have diminishing returns beyond a certain point, which would explain why evolution didn’t generally make us more intelligent even though greater-than-average intelligence is clearly possible (eg geniuses).

In the rare cases where evolution did have an incentive to evolve higher intelligence, it did so quickly and effectively. Several highly mercantile societies independently evolved the same set of genes producing higher IQ. The most notable were the Ashkenazi Jews, who have an average IQ 12-15 points higher than their European neighbors and whose genes show strong signatures of recent selection for intelligence; this most likely occurred during the Middle Ages when they were the mercantile class of Europe, since non-Ashkenazi Jews show no such effect. The genes involved tend to produce sphingolipidoses when homozygous, which shows a pretty good reason why evolution didn’t do this kind of thing more often. Myers has previously dismissed this research, but I think wrongly – the paper itself considers and rejects the all the criticisms he raises (see pages 15 – 31). The very short summary is that Myers dismisses the genetic pattern as “variations amplified by chance,” but the expected level of chance variations can be calculated and this isn’t it. Ashkenazim have similar heterozygosity to other Europeans in neutral markers – ruling out a simple bottleneck – and the mutations involved are too potentially deleterious in homozygosity to persist for many generations by chance alone. Further, the mutations are all clustered in a few key pathways, many of which are clearly linked to intelligence. For example, Ashkenazim are at high risk for torsion dystonia, which is associated with higher IQ in sufferers.

So in response to the argument that evolution must trade off against something else, I would argue that evolution doesn’t share our exchange rate. Suppose that we could gain 20 IQ points at the cost of having larger heads that are harder to fit through a birth canal (remember, some of the known genes for intelligence are associated with head size, and the obstetrical dilemma used to be a big deal!). For hunter-gatherers, who had little use for IQ but lots of use for getting through birth canals, this was a bad deal and evolution didn’t take it. For moderns, who can use IQ points to cure cancer and explore space, and who have modern obstetric techniques, it’s a lot more attractive. So yes, let’s be cautious, but I think we’d all feel pretty stupid if we avoided bootstrapping our way to superintelligence out of fears of “things man was not meant to meddle with”, only to learn later that the whole problem could have been solved with c-sections.