A Thousand Chinese Einsteins Every Year Wednesday, Oct 3 2007
intelligence 9:46 am
Accelerating Future reader and professor of economics at Smith College, James D. Miller, recently published an article on embryo screening and intelligence enhancement at Tech Central Station. Here is an excerpt:
“Two British fertility clinics have found a way of safely obtaining thousands of eggs from a woman. Fertility clinics, therefore, will soon be able to give a couple thousands of embryos to pick from. So let’s say that a certain couple’s genes mean that normally they have only a 1% chance of conceiving a child with the genetic potential to reach a genius IQ. With the ability to select among thousands of embryos, however, this couple could now almost guarantee that their offspring has the genetic potential of a genius.
Imagine that in ten years China forces all its college students to get genetic tests. Students with intelligence genes in the top 1% of the top 1% of humankind are then forced to donate sperm or eggs. China then uses the sperm and eggs to create a billion embryos each year. The genetic intellectual potential of all these embryos is checked. Those in the top 10,000 are implanted into women. Each of these embryos has the intellectual potential to be in the top one-billionth of humankind.
Now because of environmental factors many of these embryos won’t turn into intellectual titans. But let’s say that one in ten does. This means that each year 1,000 people with the scientific ability of Einstein will be born. By 2035 they will become adults and start doing scientific research. I imagine these Einsteins will be rather helpful to China’s economy and military.”
I strongly doubt that we will make it all the way to 2035 without radical intelligence enhancement emerging through some other route, such as intelligence enhancement drugs or AI, and I’ll bet James agrees with me that leapfrogging the above scenario is possible, but his presentation is very interesting because few people think about these issues.
Although James’ article may make the process sound more straightforward than it actually would be in reality, it really might be that straightforward, and even a chance of it happening means we should open our eyes to the possibility. Some counterpoint arguments were given by a flustered commenter.
Professor Miller is currently writing a book on intelligence enhancement and the Singularity.
Embryo selection of that type would be greatly complicated by interaction between genes, and the apparently great numbers of genetic variations involved in intelligence. It seems that reproductive cloning will be available well before Miller’s scenario is available, would be much cheaper, and would reduce the risks of surprising negative side effects.
It seems that the wind is blowing strongly in the direction of bans on human reproductive cloning in Western countries.
http://en.wikipedia.org/wiki/Human_cloning#European_Union
We currently need the power of a PFLOPS distributed computing network, Folding@Home, to determine how a single protein will fold. The DNA code for a single protein, on average, will take up about a printed page. My genetic anomalies, at a rough guess, are on the order of seventy thousand printed pages (that’s three hundred kilograms of paper). That’s just the *anomalies*, mind you, that doesn’t take into account how all these proteins will interact with the regular proteins that everyone has.
Not to say that cloning is safe today (with 90+% failure rates and frequent damage from the cloning process), but the advances to get reliable cloning are less daunting than those for mass embryo selection for intelligence.
Intelligence enhancement by way of embryo selection is patently ridiculous. Oh, it would’ve sounded wonderful to Aldous Huxley or some of the early Twentieth-Century eugenicists…but can there possibly be a more awkward process of amping human intelligence? Hell, it would take a century, or at least three or four generations, before a marked general IQ increase in even a nation with the size of China’s population. Plus, intelligence is not strictly determined by genes; there are thousands of pre- and post-natal environmental factors that influence the development of a person’s intelligence.
When dealing with a property that is nothing less than the result of several billion neuronal structural elements, it would stress even a Jupiter brain to predict how even a child with “genius genes” will grow and develop intellectually. Humans are notoriously difficult to programme.
Long before China thinks of breeding a generation of geniuses, the Japanese will have already developed brain augmentations that can be applied to anyone, giving anybody with the proper implants and/or neural tweaking the capacity of a thousand hand-bred baby Einsteins!
I think it would be a matter of significantly shifting the odds for desired outcomes. It would not need to be perfect based on all protein interactions.
DNA chips can screen for 500,000 DNA markers at a time now.
If all of the genes that would hinder intelligence are tossed then it could tilt the odds in favor of supegeniuses a lot.
The thing is if you get pretty good at screening and identifying what to look for then it would not be that big a step to using gene therapy, gene editing, RNA interference, RNA activation etc… to make all of the desired changes. Getting a lot better at the manipulation means that you could tweak as you go.
I view it as redraws in poker to get a better hand. Even super-eugenics is too slow and only allows hand selection at the start.
There is metagenomics which effects what genes are expressed throughout life. It changes based on lifestyle. So you would have to be constantly checking and adjusting the metagenomic expression.
There seems to be idenfication of clear winning genes. Several genes for enhanced muscle growth have been found to be additive in effect. The superstrong mouse can be made twice as strong not just 40%.
SIRT1-SIRT4 genes for longevity seem like clear winners.
Actually meant to say epigenomics not metagenomics
list of notable genes
http://en.wikipedia.org/wiki/List_of_notable_genes
http://en.wikipedia.org/wiki/Epigenetics
http://en.wikipedia.org/wiki/Regulation_of_gene_expression
Interesting stuff. A lot of people have commented above about how impossible this method is, but I think you’ve got a point.
First of all, you don’t have to understand how genes make people clever, you just have to take a large sample of clever people and look at what their genes are, compared to the genes of average people. This deals with Tom McCabe’s objection based on the complexity of analyzing how protiens fold, etcetera.
Derek makes a similar objection: “it would stress even a Jupiter brain to predict how even a child with “genius genes” will grow and develop”
- you don’t need to make predictions for individuals. You know the genes will, on average, make people clever because these are the genes which, on average, clever people have. Think of it like throwing a handful of weighted dice – you can’t predict each individual dice score, but you know that the average score will be high.
Carl Schulman’s objection that “Embryo selection of that type would be greatly complicated by interaction between genes, and the apparently great numbers of genetic variations involved in intelligence”
is moot for the same reasons. Yes, genes interact, and yes, a lot of genes are responsible for intelligence, but the law of averages will work for you. You might sometimes select embryos where the genes interact in an unfavorable way, and fail to give you a genius, but then again you will also get some cases where the genes interact in a favorable way and give you a super-genius. This is actually a good thing, because one super-genius and one retard is worth a lot more than two geniuses. (or should that be genii?)
Roko,
I agree on the general principles, but the numbers matter. The more interactions between genes, the less effective whole-genome-association studies will be in identifying targets. If an effect only appears when you have unusual alleles A, B, C, and D, which are almost never all found with intelligence-boosting (in the studies) alleles X, Y, and Z, then you won’t be able to predict very much about the combination. For the gigagenius extreme of the distribution, where many factors are boosting intelligence together and interactions are overwhelmingly positive, we will have much less data than on combinations to boost IQ to 160. Also, that means that disruption of the interactions in that dataset is more likely to cause harm, so that we would expect a higher ratio of sub-geniuses to super-geniuses than the 1:1 you use in your example.
Costs are also important. The Archon X Prize aims to bring sequencing costs down to $10,000/genome. If additional improvements bring that down to $100 by Miller’s date of 2015, that’s still a hundred billion dollar expenditure just sequencing all these embryos’ DNA, plus other enormous costs. Cloning requires only a small number of nucleus donors, and zygotes could be mass-produced using mass-harvested ova.
Eventually embryo selection will be a powerful tool (before cutting and pasting particular chromosomes, then moving individual genes, and perhaps ultimately applying Venter’s scheme of synthesizing the DNA de novo to humans) but I think it’s not as close to practicality as one using cloning, and that the cost(including reputational costs):(greatly delayed)benefit ratio of the project Miller describes makes such an endeavor unlikely for 2015.
I don’t understand why people think it would be too difficult to use embryo selection to increase average intelligence. Today a woman could increase the average intelligence of her child by choosing a husband that has a high IQ. This woman, clearly, can do this even if she doesn’t understand anything about how genes influence intelligence.
As Roko wrote “you don’t need to make predictions for individuals” to use embryo selection to increase intelligence. All we will need to do for my idea to work is to understand the correlation between genes and IQ. We will easily get these correlations after personal genomics becomes integrated into routine clinical care.
The computer analyzing the data could try to make the embryos have DNA similar to that of a random extremely intelligent person (but a specific one, not an average of many of them) – that would sidestep having to understand what the genes do.
Eric,
In that case why not just clone (possibly inserting a couple of high-powered alleles, like the torsion dystonia mutation)?
James,
“I don’t understand why people think it would be too difficult to use embryo selection to increase average intelligence.” It would be relatively easy to increase average intelligence, and a large increase in the skilled workforce (and the informed electorate) would have major implications, but producing one in a billion intellects even somewhat efficiently consistently is a tougher problem.
@carl:
“I agree on the general principles, but the numbers matter. The more interactions between genes, the less effective whole-genome-association studies will be in identifying targets.”
This is true. But you don’t know what those numbers are, because (to my knowledge) no-one has ever taken full DNA samples from 10,000 or so very intelligent people and 10,000 averagely intelligent people and looked for statistical predictors of intelligence. It may be the case that there is absolutely no detectable correlation between someone’s DNA and their intelligence, in which case the whole thing is a fruitless endeavor. From my limited experience with genomes (I worked with the Avida digital life group at CalTech), I very much doubt this. I suspect that there would be some very simple (but weak) predictors – such as the presence or absence of one gene, some moderately complex predictors – such as a combination of 5 different genes and the absence of two others, and some very very complex predictors. Obviously the bigger your sample size, the more of these predictors you would pick up on, and therefore the more clever your “geniuses” would be.
Basically, you don’t know until you try. And if it is some other country with less ethical scruples – like China – that tries, then we may find out the hard way (when these genius children have all grown up and Chinese research and development puts the rest of the world to shame), rather than the easy way.
I’d just like to add that I’m in no position to judge whether screening or cloning would be a more effective strategy, so Carl may be right about that.
The only real technical challenge at this point is finding the genetic markers that correlate to high intelligence. If we can do that, the scenario described here is entirely possible.
The problem is that, for multifactorial phenotypes like intelligence, each gene contributes to only a small percent of the variance, so we need large sample sizes to find them. No studies large enough to handle the task have yet been implemented (that I know of).
BTW, to all the people criticizing the science behind it: you don’t need microarrays or genetic sequencing to get the job done. You just need a QTL analysis of a sample of high and low intelligence individuals to find markers for intelligence. Those markers could then be identified by in situ analysis (such as FISH) of the embryo.
“This is true. But you don’t know what those numbers are, because (to my knowledge) no-one has ever taken full DNA samples from 10,000 or so very intelligent people and 10,000 averagely intelligent people and looked for statistical predictors of intelligence.”
I agree that no one knows those numbers, and it might turn out to be relatively easy to have big effects, e.g. we might find 200 fairly cumulative tweaks, of which the greatest geniuses have about 120, and create kids with 150.
As I said, I think that genome-intelligence correlation studies will definitely yield ways to increase intelligence by quite a lot, but am raising potential difficulties (like the limited sample size of ultra-geniuses available in the entire world) and costs that should affect our predictions and confidence levels.
“Basically, you don’t know until you try.”
But the uncertainty, and the skew of your expectations, affects the expected value of the program.
This was actually covered pretty well by Gregory Cochran as an extension of his work in finding IQ-correlated genes.
http://www.jerrypournelle.com/reports/cochran/overclocking.html
Short version: single gene loci can have huge effects on intelligence (e.g., 10 points in one extreme case), we already know of many to play around with, and the intelligence gene-finding field has progressed in the 5 years since Cochran wrote this. I’ll paste what I find the most relevant, but the whole discussion is worth reading.
There is a good chance that an odd cluster of hereditary neurological diseases among the Ashkenazi Jews is a side-effect of strong selection for increased intelligence. The idea is not really new, but the evidence has gotten stronger with time, and I have recently found some intriguing supporting data.
…
We have found [a gene correlated with both increased IQ and torsion dystonia], which codes for an ATP-binding protein, but as yet I don’t believe that we know exactly how it causes trouble or what it does normally. But I’ll hazard a guess: the change accelerates some brain system tied to cognitive functioning – nearly redlines it, leaves it vulnerable to common insults in a way that can cause spectacular trouble. You might compare to overclocking a chip. Sometimes you get away with it, sometimes you don’t.
More generally, if this is what I think it is, all these Ashkenazi neurological diseases are hints of ways in which one could supercharge intelligence. One, by increasing dendrite growth: two, by fooling with myelin: three, something else, whatever is happening in torsion dystonia. In some cases the difference is probably an aspect of development, not something you can turn on and off. In other cases, the effect might exist when the chemical influence is acting and disappear when the influence does. In either case, it seems likely that we could – if we wanted to – developed pharmaceutical agents that had similar effects. The first kind, those affecting development, would be something that might have to be administered early in life, maybe before birth. while the second kind would be ‘smart pills’ that one could pop as desired or as needed. Of course, we have to hope that we can find ways of improving safety. Would you take a pill that increased your IQ by 10 or 15 points that also had a 10% chance of putting you in a wheel chair?
…
Let us assume that we really could make drugs that increased intelligence. I’m pretty sure it’s actually possible, and the approach that makes it easy is looking to see what the results of natural selection for increased intelligence are, rather than trying to understand everything about human bioneurochemistry from the ground up. We didn’t need to know everything about the biochemistry of lactation to get Jersey cows – indeed we didn’t know anything.
“I think it would be a matter of significantly shifting the odds for desired outcomes. It would not need to be perfect based on all protein interactions.”
Simply shifting around pre-existing alleles isn’t going to give us anything new; we’ve been doing that naturally for millions of years.
“First of all, you don’t have to understand how genes make people clever, you just have to take a large sample of clever people and look at what their genes are, compared to the genes of average people.”
If it were that simple, natural selection would have done it already. Almost all genes depend on other genes to function. If you simply took the top one hundred alleles in smart people worldwide and deposited them into embryos, you’d probably get above average intelligence, but you’d also get a lot of… er… interesting side effects. To get Einstein-level intelligence, you need either a great deal of luck and the right environment, or entirely new alleles. I do think that there are some simple hacks which can be made genetically to make the brain work faster/better, but even those are currently beyond us. Probably by 2020 or so.
“but then again you will also get some cases where the genes interact in a favorable way and give you a super-genius.”
If these genes do exist, in any significant numbers, they would *already have interacted* because of cross-breeding. If you dump a lot of genes in, a few will improve intelligence, and most of the rest will just sit there gumming up the works. Down’s Syndrome is caused by an extra copy of chromosome 21- simply taking genes that are *already being expressed* and making another copy of them can have major physiological effects.
“Would you take a pill that increased your IQ by 10 or 15 points that also had a 10% chance of putting you in a wheel chair?”
This is an excellent summary of the current state of research. Soon, we’ll have enough computing power to figure out directly what these loci are doing, and hopefully block them when they’re dangerous. We’re getting there.
“The computer analyzing the data could try to make the embryos have DNA similar to that of a random extremely intelligent person”
If you dumped Einstein’s DNA into a random embryo, you’d get a smart person, but you wouldn’t get *Einstein*. Keep in mind that neither of Einstein’s parents was particularly intelligent, so there’s no obvious, heritable genetic “switch” to throw.
Shifting the odds not just alleles.
We are identifying the intelligence impairment genes.
If we can clearly identify the genes that when active are quite harmful, then it would make sense to turn them off or deselect them.
Say an 60-95% reduction in those with IQ 100 or less.
At the same time we identify the IQ boosters 10-15% or better.
Increasing the percent of higher intelligence people.
Thus the odds for top 1% could go up by 4-10 times.
Even if 10% would need a wheelchair – we would have exoskeletons, plus Stephen Hawking still makes great contributions.
Note: I think that a lot of being a genius or super-expert can be improved with better training.
Deliberate practice entails more than simply repeating a task — playing a C-minor scale 100 times, for instance, or hitting tennis serves until your shoulder pops out of its socket. Rather, it involves setting specific goals, obtaining immediate feedback and concentrating as much on technique as on outcome.
Their work, compiled in the “Cambridge Handbook of Expertise and Expert Performance,” a 900-page academic book makes a rather startling assertion: the trait we commonly call talent is highly overrated. Or, put another way, expert performers — whether in memory or surgery, ballet or computer programming — are nearly always made, not born.
Tom,
I absolutely agree that the complexities here are staggering, but initial results with gene therapy have shown that, at least when mucking around with alleles which are normally found in the human gene pool, the body is remarkably accommodating. Usually. There’s so much genetic junk, copying errors, copy number variations, structural variations, inversions, etc variation between peoples’ genomes that adding a gene here and there that adding a gene here or there isn’t as big of a physiological shock as it would seem. Usually.
I also agree that environmental effects are important, but I don’t agree that this sort of approach couldn’t create some spectacularly intelligent people– as Cochran notes,
- if we’re talking the Ashkenazi mutations, hardly anyone has more than one, just about nobody more than two. One in two thousand Ashkenazi, at most, carry a Tay-Sachs mutation and a Gaucher mutation, the two most common. But using drugs, we could in principle give you the torsion dystonia effect _and_ the Gaucher-carrier effect _and_ the Tay-Sachs – carrier effect _and the Canavan-carrier effect. _ and the familial dysautonomia carrier effect. As a rough guess, might give you considerably more than 20 pts – torsion dystonia gives about ten all by itself.
When we talk about creating more “Einsteins” I would say we’re nudging into the realm of speaking about achievement, not intelligence… but on the intelligence angle, adding 1.3 standard deviations of IQ onto an already brilliant person (and who is motivated to make use of their intelligence) would be substantial.
Clearly safety issues come up when we speak about loading many intelligence genes already correlated with disorders into a person, but as Cochran notes,
- these mutations are very recent and have not been refined by natural selection. And most people with that torsion dystonia mutation never get sick. Since that is the case, it is probably easy to improve them, reduce side effects, etc. They are non-optimized and so can be optimized. If nothing else, take a break now and then from the drugs. Carriers can’t do that.
Interesting stuff.
“adding a gene here and there that adding a gene here or there isn’t as big of a physiological shock as it would seem. Usually.”
Agreed. The problem isn’t adding one new gene, it’s adding dozens or hundreds of new genes, which will be necessary if you plan to enhance intelligence through gene statistics.
“but on the intelligence angle, adding 1.3 standard deviations of IQ onto an already brilliant person (and who is motivated to make use of their intelligence) would be substantial.”
A gene does not move everyone up exactly one point three standard deviations on the human-designed IQ test, no matter how intelligent they already are. It doesn’t work that way.
[...] Accelerating Future (do check out some of the interesting comments made, regarding both implications as well as [...]
I do not agree about the need for hundreds of genes. On the link I provided, Cochran found that a gene connected with risk of torsion dystonia provided an average of 10 IQ points. 2/3 of a standard deviation, all by itself. With genes like that out there should be no need to dump in hundreds of genes to get significant results.
And, yes, we are talking “on average” here– splicing in said genes might effect people differently (if TD is an average of 10, that means sometimes it’s less, but sometimes it’s more), and one might need to be born with it to get the full effect. Still, we’re not talking hundreds by any measure.
As to your point about diminishing returns on already-smart people: I don’t know. Hard to say.
[...] bunk and hysteria? Read the article and you tell [...]
“On the link I provided, Cochran found that a gene connected with risk of torsion dystonia provided an average of 10 IQ points. 2/3 of a standard deviation, all by itself.”
I’m sure genes like this exist, although probably with major side effects. The difficulty is finding them, and you’re not going to find them with statistical methods. Suppose gene A is, on average, found more in intelligent people. Gene A could help to make you more intelligent. Or it could simply be next to Gene B, which helps make you more intelligent, as the two get copied together during meiosis. Or it could be a non-coding intron sequence that falls close to an intelligence-enhancing gene. Or it really could make you intelligent, but only when working in concert with Genes C, D, and E, all of which are buried somewhere in the morass of data because they don’t stand out individually. Or it could make you more intelligent in concert with Gene F, but will kill you when acting by itself. Poof, there goes your research funding.
The neat thing about Cochran’s approach is that he has data, so we don’t need to deal in hypotheticals (or, rather, it allows us to deal in more grounded hypotheticals). I would recommend reading
http://www.jerrypournelle.com/reports/cochran/overclocking.html
Cochran has released at least one followup paper: see
http://homepage.mac.com/harpend/.Public/AshkenaziIQ.jbiosocsci.pdf
It’s aimed at proving an evolutionary hypothesis, rather than identifying intelligence genes, but it does so as part of the argument– see e.g. Table 6.
He’s analyzing a fairly endogamous group, which makes the statistical analysis and teasing apart causality much easier.
I fell myself clever, but clever is not in life, the interaction with your parents or your family, to love be clever and don’t wish to be something else. So what for to built a million-superclever factory if the will be frustrated or hate the situation they are living. Better is to forbidden fornication and adultery, rapes and many things that affect the pure conception and impact directly the life.
I’ve got an Amazon gift certificate burning holes in my pocket,
and I want to get the most bang for my buck.
Enter the Secret Amazon Web Pages:
http://bargains-hunter.blogspot.com/2007/10/how-to-find-bargains-at-amazons-secret.html
Amazon Secret Web Pages
This is where you’re going to find the “latest sales, rebates, and limited-time offers” from
Amazon, and you can score some pretty deep discounts if you’re a savvy shopper.
Next, there’s the special Sale link. This is open every Friday, and ONLY on Fridays.
You can find the same good discounts here as you would in hidden Deals, although some
Fridays you can really get lucky and make off like an Amazon bandit – I’ve seen discounts
there as low as 75% off sticker price.
I found this article bookmarked and I really liked what I read. will surely bookmark it as well and also go through your other posts later.