Spellchecked Supermen


Founded by eight NASA scientists, the Rainbow Mansion is a kind of academic coop, where you have to demonstrate you’re working on something interesting to get a rental agreement. The building itself is true to its name, a mansion spacious within, and surrounded by lush gardens without. Every week they host a group dinner, followed by a speech from an invited guest. This week’s guest was Mike Johnson, a philosopher and transhumanist who is currently working on a treatise that could lay the groundwork for a mathematical model of pain/pleasure.

His talk, however, wasn’t about that, but about another topic and interest of his – genetic typos, the possibility of “correcting” them, and the profound effects that might have on human intelligence and capability if widely implemented.

Genetic editing tools are already coming online that would work for already existing organisms: CRISPR, whole-chromosome DNA synthesis, viral vectors (adenoviri), etc. What can we edit? We could try to maximize for some trait using the GWAS approach (e.g. as BGI is trying to do with IQ). We could go for transgenic bioengineering, you know, the Spiderman/Resident Evil-type stuff. But that’s pretty hard. Just fixing our own broken genes is much easier and could potentially generate tremendous payoffs in increased health, intelligence, and longevity.

We all have varying amounts of “broken genes,” the genetic equivalent of spelling errors. Of those errors that have an effect, the vast majority are bad; as Mike pointed out, if you were to open up a computer program and edit code at random, you are far more likely to ruin or degrade the program than improve it. There are several different definitions and estimates of the numbers of these errors: 100 semi-unique Loss of Function mutations (MacArthur 2012), 1,000 minor IQ-decreasing variants (Hsu 2014), 300 health-decreasing mutations (Leroi 2005).

Broken genes have a broadly linear additive effect on general fitness, which is well approximate by IQ. Stephen Hsu’s research indicates that people have an average of 1000 broken genes, with 30-40 mutations contributing to a stunning -1SD drop in intelligence. In essence, it’s not so much that there are genes for intelligence, as there are genes for stupidity. Fix all of them, and theoretically, you might get IQs never before observed on this planet. As Greg Cochran memorably put it:

What would a spelling-checked person, one with no genetic typos, be like?  Since no such person has ever existed, we have to speculate.  I figure that kind of guy would win the decathlon, steal your shirt and your girl – and  you still couldn’t help liking him.

Here is a list of (optimistic) estimates for other traits that Mike collated from various sources.


While Mike, understandably, did not go into this in his talk, one more important point has to be mentioned: There is also an explicit HBD angle to the theory of genetic load.

Studies show significantly more Loss of Function mutations amongst Africans than Europeans or East Asians, which would tie in not only to well-known psychometric data but Satoshi Kanazawa’s theories on the relatively low atttractiveness of Black women (specifically female beauty, like g, appears to be a good proxy for overall fitness). Cochran ascribes it to heat. I am not so sure. Peak wet bulb temperatures are actually higher today in the Ganges delta and interior China than most of Africa, which has some really cool (temperature-wise) places like the Ethiopians highlands and the Great Rift region. This might not have been quite the case during the Ice Age, of course, but still, 10,000 years is a long time to adjust to a new equilibrium.

Another possible determinant of genetic load is male parental age. Offspring genetic load and paternal (though not maternal) age are positively correlated. Paternal age in traditional societies can differ substantially according to their particular family system. For instance, within the Hajnal Line encompassing most of Western Europe, characterized by nuclear families, average paternal age was considerably higher than amongst say the neighboring Poles and Russians. What specific family system is highly prevalent in the traditional global South, especially in Africa? Polygamy. This implies one dude monopolizing a lot of the chicks. What would he be like? Big, bad, bold – naturally. But he’d also have a reputation, and he’d probably be someone who can spit smooth game. Both the latter require some time to build up. So he would probably be considerably older than fathers elsewhere in the world who entered into monogamous marriages. But this is just a theory, it would be great to actually get concrete anthropological data on average paternal age in traditional Africa.

Though I’m taking steps to remedy this, I am not sufficiently well versed in genetics as to offer a valid judgment on the plausibility of Cochran’s and Hsu’s mutational load theory of IQ. Still, it does appear to have a great deal of face validity to it, though I remain skeptical of whether spellchecking can truly create “superhumans,” as opposed to just some very healthy and athletic 145-175 IQ types with a life expectancy of maybe 105 years. Surely at some point basic biological limits will be hit, and there will be diminishing returns?

Still, the potential for improvement are immense, and eventually, it will be possible to apply them to grown adults as opposed to just embryos. Even raising the global IQ by one SD will basically solve India’s and Africa’s development problems, while making the two odd billions of Europeans, Americans, and Chinese as innovative per capita as the world’s 20 million Ashkenazi Jews. Near instant technological singularity! When asked to give an estimate, Mike Johnson said that this “spellchecking” technology will become available in 5-7 years for billionaires who wish to have a designer baby and are unconstrained by any regulatory restrictions.

There’ll inevitably be a lot of hand-wringing about this, lots of soul searching and moral queasiness, and no doubt some attempts at restriction, but it’s hard to stop a moving train. As Mike said, the Chinese and East Asians in general don’t share these concerns; if they can safely have a more intelligent child, well, why on earth not? It is telling that the global focal point for research on the genetics of IQ, which Steve Hsu is incidentally heavily involved with, is the Beijing Genomics Institute. Regardless of their reasons or justifications, those who refuse to get on this train will simply be left behind.


  1. Chinese and East Asians in general don’t share these concerns

    Historically, East Asians care little about ideology/religion. Somehow I start believing higher the IQ, lesser conviction of ideology/religion. Ideology/religion are for stupid people to use as guidance to judge thing they do not understand; for smarter people to use as tools to manipulate the mass.

  2. Nah, I don’t think that stuff about different levels of load per population actually pan outs that way Anatoly –

    http://arxiv.org/abs/1305.2061The deleterious mutation load is insensitive to recent population historyHuman populations have undergone dramatic changes in population size in the past 100,000 years, including a severe bottleneck of non-African populations and recent explosive population growth. There is currently great interest in how these demographic events may have affected the burden of deleterious mutations in individuals and the allele frequency spectrum of disease mutations in populations. Here we use population genetic models to show that–contrary to previous conjectures–recent human demography has likely had very little impact on the average burden of deleterious mutations carried by individuals. This prediction is supported by exome sequence data showing that African American and European American individuals carry very similar burdens of damaging mutations.

    http://arxiv.org/abs/1402.4896“No evidence that natural selection has been less effective at removing deleterious mutations in Europeans than in West Africans”“Non-African populations have experienced major bottlenecks in the time since their split from West Africans, which has led to the hypothesis that natural selection to remove weakly deleterious mutations may have been less effective in non-Africans. To directly test this hypothesis, we measure the per-genome accumulation of deleterious mutations across diverse humans. We fail to detect any significant differences, but find that archaic Denisovans accumulated non-synonymous mutations at a higher rate than modern humans, consistent with the longer separation time of modern and archaic humans.

    Discussion – It is tempting to interpret the indistinguishable accumulations of deleterious mutations across present-day human populations as implying that the overall genetic burden of disease should be similar for diverse populations. To the extent that mutations act additively, this is correct, as it implies that the complex demographic events of the past are not expected to lead to substantial population differences in the prevalence rates of complex diseases that have an additive genetic architecture16,19. However, recessively or epistatically acting mutations work in combination to contribute to disease risk, and, because demography affects allele frequencies, it affects the rate of cooccurrence of alleles. For example, the absolute count of alleles occurring in homozygous form is higher in non-Africans than in Africans for all functional site classes.”.

    (note in this paper, Denisovan is the exception with higher load, not Neanderthal).

    http://arxiv.org/abs/1306.1652On the accumulation of deleterious mutations during range expansions

    References are all 2013-2015; Cochran’s stuff is from 2012, so perhaps the science has moved on.

  3. Anatoly- thank you for the mention, and it’s a nice overview. I do think this is one of the most interesting ‘low-hanging fruit’ areas in biology. I can’t take original credit for the ideas- I stand on the shoulders of many giants in the community, most notably Cochran/Hsu/Leroi.

    A few comments:
    – I’d like to make a graphic distinguishing the different definitions/estimates of genetic load. But in short: MacArthur’s interested in super-rare loss-of-function mutations. We can’t study them with association studies, so we don’t really know what they do– just that they’re probably important, and probably unilaterally bad. Hsu, on the other hand, looks at semi-rare mutations that decrease a specific trait (specifically, IQ). Many of these semi-rare mutations will decrease IQ by virtue of decreasing health; some will decrease IQ and leave everything else fairly untouched; some will decrease IQ but there’ll be some beneficial tradeoff elsewhere. A holistic definition of ‘inherited genetic load’ will combine these two understandings.

    • It’s my understanding Cochran believes paternal age to be a much more significant driver of genetic load than heat. Also, here’s Peter Frost on the general topic: http://www.unz.com/pfrost/on-paternal-age-and-iq/ (I couldn’t find an easy list of paternal ages by culture and era, but it’s probably out there somewhere…)
    • On my “5-7 years” estimate: I should note that this would be wildly optimistic in terms of a ‘full genetic spellcheck’, even if resources, regulations, and morality weren’t factors. However, if we’re just concerned with performing a procedure that gets enough low-hanging fruit to make a significant practical difference, I would stand by the estimate.

    I’ll send you a better-formatted slide for the super-optimistic estimates. I’ve had problems with the formatting as well.

  4. Erik Sieven says

    Human Biological Diversity could become accepted mainstream as soon as it is gone. When genetic engineering is possible for billionaires, after a few years it will be possible for millionaires. And some years after that some Silicon Valleys Billionaires and international sponsors could finance gigantic genetic engineering projects to enable village people in cameroon and congo to design their babies. Or, of they don´t trust the designing choices those people would make, maybe they would do the designing for themselves.
    Thus every baby born in the world would have the same IQ of 175. Then, and only then people would admit that there had been some differences before

  5. It’s not nearly that simple, I suspect. It’s not as though there were One True Genotype from which we have fallen away in sin – the viability of any gene depends not on its own properties, or of the organism’s environment, but the totality of every gene and environmental influence.

    The overwhelming majority of organisms don’t use reproductive methods which rigorously eliminate suboptimal mutations, even though self-crossing makes it fairly simple to eliminate deleterious recessives. Almost all complex organisms work to outcross instead. Reproduction is generally designed to retain as much diversity as possible even while individual organisms are impaired – and it’s not difficult to see why. At the level of entire species, it’s far more beneficial to keep around a library of variations whose utility will vary depending on external and internal factors than to maintain an optimized monoculture that offers limited potential for selection and will become inefficient as soon as conditions change.

    If it were truly a good idea to optimize the genome, nature would have done it already.

  6. Even raising the global IQ by one SD will basically solve India’s and Africa’s development problems, while making the two odd billions of Europeans, Americans, and Chinese as innovative per capita as the world’s 20 million Ashkenazi Jews. Near instant technological singularity!

    1. East Asian IQ is consistent at 105 or more while european IQ goes as low as 89 for Serbia and Montenegro, 91 for Lithuania and Romania, and 92 for Ireland and Greece. Average eastern european IQ is ~95 while average western european and north american IQ is ~98. The difference is consistently big enough that east asians should be in a category of their own. Source: http://www.ttu.ee/public/m/mart-murdvee/EconPsy/2/Lynn_Meisenberg_2010_National_IQs_calculated_and_validated_for_108_nations.pdf

    2. Where did you find 20 million ashkenazi jews? That’s almost double the actual number. Also, about half the jews in Israel are ashkenazi yet Israel’s IQ is 95.

  7. East Asia already has the IQ advantage over other regions of the world and it also has the advantage of possessing a pragmatic, highly competitive culture that is conducive to taking advantage of genetic manipulations. So, advantage China et al.

    However, maximized human potential will still fall short of Artificial Intelligence and Robotics in the future…

  8. ” Still, it does appear to have a great deal of face validity to it, though I remain skeptical of whether spellchecking can truly create “superhumans,” as opposed to just some very healthy and athletic 145-175 IQ types with a life expectancy of maybe 105 years. Surely at some point basic biological limits will be hit, and there will be diminishing returns?”

    So as always on this site, we wind up talking about Jews.

    Okay what’s the average genetic load of Ashkenazim Jews? (Face it, we really don’t care about any other kind)

    I have read that there are a number of mutations prevalent in that population that are linked to higher IQ.

    So do they, at the same time, have typos that reduce intelligence? And if you removed those, presumably the relative IQ differential remains for our new breed of “corrected” individuals.

    Or do Jews have less typos… for intelligence at least?

    You know it isn’t intelligence, but it would be interesting to test the dna of star athletes and see what their genetic load is like. As indiscriminately as some of them spray sperm around, it shouldn’t be hard to get.

  9. It’s not as though we can just test for ‘genetic load’. We don’t have any general way of evaluating a gene, except by observing lots of phenotypes and trying to correlate, and that’s imprecise at best. Guessing based on gene frequency has some very serious drawbacks.

    A majority of human beings don’t retain lactose tolerance past breastfeeding. Clearly, then, the genes responsible for this trait ought to be eliminated, since lactose intolerance must be metabolically superior.

  10. dave chamberlin says

    We are all guessing at this point to what degree mutational load influences IQ. It still is a fascinating subject worth following but it is all unsubstantiated conjecture at this point. I think it is great that you are now following this subject closely and I would love to read more updates as they become available. I completely agree with your statement that “regardless of their reasons or justifications, those who refuse to get on this train will simply be left behind.” We live in interesting times.

  11. Melendwyr: It’s not as though we can just test for ‘genetic load’. We don’t have any general way of evaluating a gene, except by observing lots of phenotypes and trying to correlate, and that’s imprecise at best. Guessing based on gene frequency has some very serious drawbacks.

    That’s why the optimum strategy would be more (at least first) to just nix anything under say 0.1%. Won’t hit the lactase persistance allele (lot higher frequency than 0.1% in practically any population where its present at all) or any weird pigment variants frequent in your population; will hit much more random low frequency mutational noise than it hits low frequency incredible genes.

    And take note of population structure when you do it – compare a Chinese against the consensus genome for 10,000 Chinese, not 10,000 French, for instance, even though that’s still kind of a low sample size in both cases. Yes, you’ll lose some good mutations in the process, but many, many more bad mutations will go.

  12. Yes, you’ll lose some good mutations in the process, but many, many more bad mutations will go.,

    But that’s just it – outside of specific contexts, there is no such thing as a ‘good’ or ‘bad’ gene. And the context includes every other gene an organism possesses.

    Screening out all the rare genes, or all the genes that are deleterious to some aspect of our traits in the most common contexts, would massively impair our ability to evolve in response to selection pressures. We were probably never rigorously optimized – in the long-term, a strictly optimized organism will rapidly go extinct as conditions change and the target moves out from under it. But to the degree that our genes on average produced a near-optimum result, it was in our hunter-gatherer days. We’ve been changing our environment rapidly since then – and our genome has changed rapidly, too. It would be foolish to think that our present genes are optimal for our present conditions. And as for future conditions… who knows? Sabotaging our genetic library of options just as our environment is being revolutionized is a recipe for disaster.

    We have recessive traits for a reason. We’re designed to conserve recessive traits, when it would be easy to eliminate them, for a reason. That reason has nothing to do with individual human survival or happiness, and everything to do with the long-term health of our species.

  13. Melendwyr: Screening out all the rare genes, or all the genes that are deleterious to some aspect of our traits in the most common contexts, would massively impair our ability to evolve in response to selection pressures.

    The impression I get is there’s enough standing variation at levels beyond the 0.1% deleterious mutation frequency cutoff this isn’t an issue at all, compared to the gains from reducing mutational load.

    However, if you really wanted, they could always develop a smarter cross species variant – e.g. if certain deleterious mutations are always lower than 0.1% frequency in every mammal, and humans don’t show a pattern of being the outgroup for the gene, and they’re also lower than 0.1% in humans, maybe we don’t want them? There are plenty of options for spellchecking (although the more species and subpopulations you check and the more thresholds you get the more complex the process becomes).

    In the long run evolvability to become a vastly different organism doesn’t seem like what humanity would even want compared to quality within its strategy – I’m not sure humans would even want to preserve variations which would allow us to evolve to become as different a species as say pigs are from mankind.

    This is one of those discussions that could probably run and run between geneticists who know about evolvability vs load.

  14. Personality and cognitive functions are profoundly influenced by genetics. And as many have speculated, major technological shifts have probably been both the cause and consequence of changes in certain mental traits.

    Optimizing performance at the moment, by eliminating rare variants, no matter what metric is used to define ‘optimal’, will cripple development. The currently-maladaptive genes are the building blocks of future adaptation.

    There are likely genetic causes behind the ascension of ancient China to the world’s most advanced civilization… and behind its centuries-long stagnation. Remove dissidents from the gene pool over a long enough period of time, and the traits of the population change. That’s just with indirect selection by proxy – imagine what damage we’ll do to ourselves once we start picking and choosing directly.

  15. What about spell-checking for aggression, violence, criminality, jihad,..? Or can this be grafted in?

    I’d take a kind-hearted, homely, stupid person over a brilliant, gorgeous psychopathic killer (e.g., Rosamund Pike’s chararcter in Gone Girl) But of course, millions of people rely on this worst of human flaws. What we do with all the law enforcement-homeland security-military-security industrial complex workers?? Have them pick daisies?

  16. Honorary Thief says

    Also, about half the jews in Israel are ashkenazi yet Israel’s IQ is 95.

    Sephardic Jews and Arabs.

  17. Forgive me for what is likely a stupid question, but if we gain the capability to edit deleterious versions of genes, would we not also gain the ability to replace “good” genes with better ones?

    Is there a reason why we would only delete bad genes and not good ones?

    And if we are able to do both, would the people that result be close to the sort of superhumans Cochran envisions? Or would we still run into the sort of diminishing returns you mention?

  18. Melendwyr-

    There’s natural genetic variation, and there are errors. This idea of ‘spellcheck’ leaves natural genetic variation (and most balanced polymorphisms) intact, while fixing the obvious, low-hanging-fruit errors.

    It’s reasonable to say context is important, but it’s unreasonable to say “outside of specific contexts, there is no such thing as a ‘good’ or ‘bad’ gene”– many errors are simply bad. Negative. They break the gene, and this hurts us. Full stop.

    Any hypothetical ‘spellcheck’ process wouldn’t have much impact on ongoing accumulation of de novo mutations, so evolution would still happen if it was a one-time thing. If it was an ongoing process, sure, evolution would slow down (but you still have a gazillion possible combinatorial arrangements of existing genes, so it’d by no means stop). But something tells me worrying about genetic engineering having the capacity to stop evolution is kind of missing the point: I guarantee genetic engineering will lead to a lot more genetic change, not less.

  19. Mike- yes in theory, if we can edit the genome, we can replace ‘good’ genes with better ones.

    But it’ll be really complicated to do this. Generally, if there’s a ‘good’ version of a gene, it won’t be easy to improve upon it in a way that doesn’t involve subtle (or obvious) tradeoffs. Eventually this sort of ‘from good to amazing’ enhancement will happen, I’m sure, but it’ll take deep knowledge of how genes contribute to our phenotype, and a lot of trial-and-error, and maybe there’s not a ton of room for improvement without completely remaking our genomes.

    Meanwhile, we have a bunch of broken genes laying around, and we already know the ‘good’ versions work (since almost everybody carries the good variant of any given gene). We don’t have to know nearly as much about the nuts-and-bolts of how things work to fix these errors. The neat thing is the potential benefits from doing this are surprisingly large.

  20. AK: OT to this post, but a friendly heads-up.

    I caught your map of the three nations of Ukraine on your twitter page; by lumping together the former “blue” Yanukovich-voting Ukraine into one “nation” your article may be outdated. Numerous polls, and election results,, have seen a real split within the formerly Yanukovich-voting part of the country.* Arguably Dnipropetrovsk, Kherson and Mykolaiv oblasts may be placed together in the same “nation” as Kiev. Or, they may be their own one.

    *If you recall, Yanukovich stood for an independent but Russia-oriented, pro-Russian speaking Ukraine (though towards the end he swung for the EU then suddenly changed course, then was overthrown). The war has split his former electorate along geographic and ethnic lines. The East (Donbas, Kharkiv, largely Zaporozhia) continues its pro-Russian orientation – but the war has swung the ethnic Ukrainian South (Dnipropetrovsk plus the rural Ukrainian-inhabited Kherson and Mykolaiv oblasts) away from Russia. These regions are not as hardcore anti-Russian as Kiev now is, but they are about where central Ukrainian regions were a couple of years ago. Odessa is split.

    While it may be pleasant for pro-Russians to imagine a “New Russia” linking Crimea and Transnistria to Russia proper, this idea seems to be based on a combination of wishful thinking misreading Yanukovich’s voters, and setting aside the results of the war on public opinion. Just as Donbas had become much more anti-Kiev, other parts of the country have become much more anti-Russian. Kiev and the Center have become about as nationalistic as Galicia used to be, and the non-Odessa South has become rather like the Center used to be.

  21. There’s natural genetic variation, and there are errors.

    Wrong. It’s all ‘errors’. Errors which turned out to be useful become the default. Errors which aren’t useful tend to be retained in case of future utility.

  22. It seems to me you want to say something about mutation, adaptation, and ecological niches, but aren’t quite clear on how to phrase the statement.

    Are you suggesting that no gene variant can ever be more across-the-board adaptive than any other gene variant? Or that if a random mutation happens, our probability estimate shouldn’t skew toward it likely having a negative effect? Or that there’s no such thing as a “loss-of-function” mutation, since function is contextual? Or that random mutation is the core engine behind adaptive evolution, so it has a net positive effect on our genomes? Or that accumulating mutations in not-so-important-or-inactivated-genes is useful as a pool of genetic variance, that can be drawn upon during times of increased selection pressures (basic ‘punctuated equilibrium’ theory)? Or that evolution, for all its blind chance, is probably smarter than humans would be if we tried to ‘fix’ our genomes, so we should leave well enough alone?

    The above statements are all very different. A couple are reasonable, with some caveats, whereas others go against every foundational equation in the field of population genetics.

  23. The last question is the only one that has any relationship to what I’ve been trying to tell you. As for “not being clear how to phrase the statement”, let me make this as simple and explicit as possible. It won’t be short, but that’s what happens when things are spelled out.

    1) Genes code for biochemistry, not for the traits we ordinarily observe. You cannot make a mouse grow a trunk by splicing the genes for trunks from elephants, because there are no “genes for trunks”.

    2) Very few traits are ultimately the result of one or a few genes. Most traits arise from the interacting effects of many genes – dozens, hundreds, even tens of thousands or more.

    3) Since the biochemical consequences of a given gene depend not only on the environment external to the organism but its internal chemistry, which is affected by the activity of all other genes it possesses, what a gene does can ultimately be evaluated only by looking at the sum total of all the genes in a given genome. This has been repeatedly demonstrated in simple organisms and is thought to hold true in complex ones – the only reason we can’t reliably demonstrate it in animals is that they’re too complex for us to understand their genetics fully. Plants are easier. For example, it’s known that if you take what seems to be a perfectly healthy strain of mitochrondria from one variety of plant and place it into a perfectly healthy alternative variety of that plant (or a closely-related species capable of interbreeding) the result is sometimes male-sterile, completely sterile, or even totally inviable. The mitochrondrial DNA in itself cannot be said to be viable or inviable, nor can the nuclear DNA be either – we can only speak of their relative compatibility or incompatibility.

    Similar effects at high levels of analysis are known to occur in animals, whose genetic mechanisms are much subtler. In embryonic neurological development, for example, we know that different neural systems mature at different times and in response to different types of stimuli. Alter the order in which their development occurs or cause multiple systems to try to develop at once, and development ‘crashes’. Chickens’ visual cortices don’t start to wire themselves vigorously until the chick hatches and is flooded with light. If you remove part of the eggshell, and form a sort of window, the developing chick receives the stimuli that induce visual cortex development early – and it screws up not only visual processing but whatever was maturing when the window was made, because the resources that their physiology was ‘expecting’ to be devoted to a single system are now being competed over by different parts of the growing brain.

    4) We know that the overwhelming majority of ‘higher’ organisms have mechanisms designed to prevent self-crossing or breeding with close relatives. This can vary from having separate sexes, to timing male- and female-fertile periods differently, to complex biochemical signals which permit or prevent fertilization. There are some examples of organisms which freely permit self-crossing. Many of them are plant and animal varieties that have been extensively modified by human selective breeding, and the wild original strains or species that are their ancestors are often found to retain the preventative mechanisms.

    5) A little playing around with Punnett squares quickly demonstrates that obligate in- or self-breeding rapidly produces ‘pure strains’ that are homozygous for any given gene variant, while obligate out-breeders and -crossers tend to retain heterozygosity with variations that are imperfectly dominant or recessive, even if those genes significantly negatively impact the organism’s survival when two or more copies are present. Even when breeders want to eliminate such traits, it’s virtually impossible to do so through selection when dealing with outbreeding species. The best they can usually do is reduce the frequency of the recessives below a certain value through generations of back-crossing and screening.

    6) Because inbreeding ‘fixes’ traits by eliminating heterozygosity over time, it’s generally easier to select among displayed traits with inbreeders. This is an important reason for why so many of our crop plants self-cross even when very few wild plants will tolerate it – we selected against the plants’ own selectivity.

    7) However, there are serious limits to what can be accomplished with inbreeders. Take peas, the cultivated forms of which rarely cross-pollinate although the wild ancestor species does so most of the time. A mutation that alters the conditions under which the host plant will thrive – say, by altering the activity of a key enzyme at different temperatures – will probably have significant negative effects on fitness under normal conditions when expressed. Because peas are inbreeders, they tend to become homozygous – and so the allele will tend not to be compensated for by a ‘standard’ allele, and its frequency will diminish over time.

    But that means that there’s little available diversity to select among – alleles that render peas less fit under standard conditions – but potentially more fit in different conditions – aren’t retained for long. So if you wanted a strain of pea that could be planted in the beginning of summer, instead of very early spring or early fall, you’d have to plant out cold-adapted peas and coddle them while waiting for new, mutated variants to arise. Because the peas don’t tolerate genetic variants that are poorly adaptive, therefore pretty much all plants can be expected to be fine-tuned to their normal conditions – but there’s little scope for change.

    Corn, on the other hand, is an obligate outbreeder. Self-crossing is quickly lethal, and a relatively large population is needed to maintain the health of a gene pool. A mutation that renders a plant less fit – even if partially compensated for by the normal allele – can’t be easily removed, and a gene that’s fully recessive is very hard to select out. An allele that rendered a key enzyme inefficient at hot summer temperatures – when corn normally thrives – would tend to be retained at a low level in the population even though it would be grossly maladaptive normally. Some plants in every generation would be unlucky and inherit maladaptive copies of that gene, and probably die. But despite that, there would be carriers that would transmit the bad gene to the next generation.

    Suddenly change the environment – by planting corn a month or two earlier than normal – and heterozygotes (or even previously-defective homozygotes) might have a serious advantage over the plants whose enzyme can’t function in colder temperatures. Because the gene pool retained alleles that weren’t adaptive in one context, they’re available in other contexts where they might prove valuable. Some plants in every generation will lose the genetic lottery because of retained maladaptive stuff, but that’s just the cost of maintaining a library of variations for the possible benefit of the entire species. The species is less finely adapted as a whole, but it’s much more flexible in its ability to cope with change.

    8) There is also some interesting evidence that heterozygosity is itself sometimes responsible for hybrid vigor, rather than hypothetical defective recessives being generally cancelled by healthy dominants in the cross. We don’t know the details of how or why, but in some cases the traditional explanation for why crossing two strains produces a healthier and more-fit hybrid seems to be wrong.

    9) Eliminating relatively rare alleles from the human gene pool would make us unlike corn but like peas: the variations we retained would likely work more efficiently on the biochemical level with each other, and probably would result in a greater fitness in the environment those variations were selected in and by. But despite rapid changes which we now know took place fairly recently, we’re still adapted primarily for the ancestral environment – hunter-gather food sources, social structures, and so on. Screening out uncommon alleles would probably have a minor negative effect on our ability to cope with our modern environment – and would severely cripple our ability to continue to adapt to our relatively recent conditions, not to mention conditions yet to arise.

    10) We currently believe that IQ variation within normal limits (for example) doesn’t make people in the few hunter-gatherer societies we’ve observed any more successful or fit. The cognitive features that make us well-adapted to live in modern societies are either neutral or maladaptive in their societies – if only because it’s impossible to prioritize everything at once. For example, it’s been suggested that what we now consider ADD – being highly distractible by modern standards – might be highly adaptive for hunters who have to keep a lookout for prey and predators, and the ability to ‘focus’ we now consider ‘normal and healthy’ would probably get us killed or leave us hungry under ancestral conditions.

    If the screening-out of rare, ‘maladaptive’ variations you’re proposing had been carried out in the ancestral environment, you would have likely eliminated the occasional case of maladaption and rendered the population more efficient and more fit. But you would probably have scuttled the emergence of animal husbandry, agriculture, technological development that depends upon a settled lifestyle, even the instincts and emotions that cause us to be settled rather than nomadic wanderers.

    Carry out your plan in the modern world, and at best you’ll abort our continued genetic adaptation to the world we have created and are creating for ourselves.

  24. tl; dr: here’s a shorter way of looking at it.

    I’m told that novices trying to play chess find the vast number of options too difficult to think about, so they concentrate on eliminating pieces until the board is relatively clear. With only a few pieces, the game has been simplified enough that they can think more than a few moves ahead. Experienced players, however, don’t try to make things simpler. They’ve grown able to deal with the complexities, and so they’ll trounce a novice every time.

    Humans are good at making simplified models of reality and applying them. As a result, we can find clean, simple, rational solutions. The blind idiot god Evolution doesn’t make models, and it doesn’t think ahead. It doesn’t think at all. So there are many cases where it seems we can out-engineer it.

    But while evolution lacks our bias towards simplicity, it also lacks our bias against complexity. And so by randomly searching through possibilities, it can produce solutions we can’t comprehend and would never design. The complexities of our biology – particularly our neurology – likely represent just such a case.

    There are already biological systems that, by their nature, do to their own gene pool pretty much what you’re suggesting we artificially do to our own. Those systems are very rare. If the strategy were a good one, nature would probably have implemented it – the fact that it has not, and has gone with what at first seems like a needlessly baroque and elaborate strategy instead, is a powerful indication that there are fewer advantages and more problems with it than we might at first think.

    A lot of genetic properties are directly pleiotropic, and most of them are in subtle ways – if only because biochemical pathways can interfere and compete with each other. There’s no such thing as a good or bad gene, only genes which have certain effects in certain contexts. Eliminating rare variations is probably not going to be as beneficial as you’re hoping – and it’s going to have serious drawbacks that you seem to be completely overlooking. Genetics is messy and complex, like a game of chess from Hell, and the models you’re proposing are probably simplified to the point of major error. Our Promethean foresight likely cannot win the chess game against the blind idiot god – evolution has us outnumbered. Trying to simplify our models to the point that we understand them will only make them unsuitable to grasp the complexities evolution has vomited up by accident – and applying those models will lead to disaster.

  25. As far as I can tell, you’re making a valid point that gene function depends on context, a valid point that balanced polymorphisms exist, and a valid point that mutations are an important factor in long-term adaptation.

    But then you go for what seems to me to be two ridiculous conclusions, that (1) any genetic engineering whatsoever is necessarily doomed to failure, because any optimization we can possibly think of involves unforseeable tradeoffs, even if it’s just fixing a genome’s most obvious Loss-of-Function mutations, and (2) that it’s always problematic to call any given mutation “negative”, because who knows, it could do something good in some context.

    This misunderstanding seems rather fundamental, and resting on very different understandings of these ‘typos’ I suggest fixing: it feels like you’re arguing off of your definition of genetic variation, and I’m arguing off population genetics’ definition of genetic load, and ne’er shall the twain meet. Which is fine, but I think we may be getting into diminishing returns from discussion. If you want to see genetic load more as population genetics does, I can recommend the following fairly accessible sources:

    You can also just google ‘genetic load population genetics’ or ‘genetic load Haldane’ (co-founder of population genetics, who also coined the concept of genetic load) and you’ll find a lot to read. Some of the literature uses the term ‘mutational load’ interchangeably with ‘genetic load’, which can be confusing. But I assure you, this isn’t exactly an unstudied problem…

  26. Not “any genetic engineering whatsoever”, no. But most attempts will have serious unexpected side effects – and most have. Pleiotropy is nothing to play around with.

    The traditional understanding of ‘genetic load’ is wrong-headed. ‘Load’ isn’t a problem that nature can’t eliminate – it’s perfectly capable of getting rid of rare variants. It’s not a bug, it’s a feature. Getting rid of it will have only mild benefits to individual organisms in the short-term and will greatly harm their species in the long-term.

    I don’t think I have much else to say on this matter.

  27. dave chamberlin says

    Thanks for your lengthy posts Melendwyr, I learned from them. Here is a link to a fascinating Steven Hsu paper which I expect some people will appreciate.http://arxiv.org/abs/1408.3421

    If Steven Hsu is right that we will know much of the genetic architecture of intelligence in ten years or so that is incredible. I have earlier stated that nobody has any idea how much mutational load influences IQ but we might begin to find out if we understand the very complex nature of what Steve Hsu expects to be the additive components of genetic intelligence. We live in interesting times and I recommend following Steve Hsu’s blog here http://infoproc.blogspot.com/

  28. Heterosis (hybrid vigor) may be able to produce even superior humans without engineering but we need to find suitable populations.

  29. Kevin O'Keeffe says

    “Historically, East Asians care little about ideology/religion. Somehow I start believing higher the IQ, lesser conviction of ideology/religion.”

    Perhaps, but as a Christian who is not stupid, I see no Christian-derived theological, philosophical, or moral basis for raising any objection to this possibly quite wonderful innovation in human genetic health.

    So many of the ideas I come across in contemporary science articles, strike me as trite and innane, or occasionally horrifying, but this idea just seems exciting and brilliant.

  30. Interestingly, at least with phenotypes that are caused by genetic load, there is evidence that is rather pointed, on several points.

    This study find no relationship between IQ and facial attractiveness in a large twin sample (Whites). The raters for looks did not know the IQs of the twins.

    From this sample, it was also found that the heritability of facial attractiveness was entirely due to additive genetic variance. There wasn’t non-additive variance as you might expect from various combinations of facial features. This indicates genetic load is involved. But, contra previous results, the lack of correlation between IQ and attractiveness means whatever load that exists is specific to both and IQ and to attractiveness. There was no overlap as you might expect from pleiotropic mutations.

  31. If it were truly a good idea to optimize the genome, nature would have done it already.

    You misunderstand the concept behind genetic load. Nature has been trying to…

    That said, I agree, I don’t think there’s One True Optimal Phenotype.

  32. Optimizing performance at the moment, by eliminating rare variants, no matter what metric is used to define ‘optimal’, will cripple development. The currently-maladaptive genes are the building blocks of future adaptation.

    This is true to some extent. But there are plenty of mutant alleles that are just bad all around.

    There are immensely more ways of screwing something up than improving it. Even with variable conditions considered.

  33. Historically, East Asians care little about ideology/religion

    You must have a really constrained understanding of those terms.

  34. You don’t understand. The “lots of ways to screw something up, few ways to get it right” has nothing to do with biochemistry and everything to do with how you evaluate things.

    Nature already has methods of purging mutations which reduce fitness. It doesn’t use them. Instead, it goes out of its way to arrange for reproductive patterns which conserve rare mutations. Because in reality, if not in human minds, what “getting it right” means isn’t well-defined, and tends to change rapidly. It even changes every time something comes along which works better.

  35. Could you posts the links for the references for the ‘super-optimistic estimations’?

    I haven’t seen any published estimates for most of those.

  36. The high genetic load of Africans may actually be caused by young, and not old, male parental age. Africans are known to breed early and often due to the diseases and wildlife hazards found in the jungle. and a recent Cambridge study found that young fathers (under 20) pass on more mutations to their children. http://news.investors.com/022415-740732-young-dads-mutation-links.htm