My own recollections from the last 50 years: 
 
"astronautics hasn't advanced much since the last moon shot in 1972" 
 
The manned space program is extremely risk adverse and only uses flight proven equipment, i.e., old stuff. NASA's unmanned space program is much more innovative, e.g., the Mars explorers. Also the military is strongly pushing RPV's and robotics. Non-government astronautics has been progressing. 
 
"chemistry seems to have frozen in place except for medical uses" 
 
Computer modeling of chemical structures has significantly improved, e.g., quantum mechanics, transition states, charge distributions, i.e., smart material design. Chemistry plays a critical role in nanotech. Nanotech improves surface characteristics for new catalysts. The materials in modern products are stronger, lighter, and more durable than in prior decades...that is largely due to chemisty. 
 
"Without access to very expensive apparatus, experimental and theoretical physicists have been marking time for much of the last 30 years." 
 
Physics has multi-billion dollar particle accelerators and fusion experiments that employ teams of hundreds of scientists. Some of the most expensive computers are largely devoted to physics. If progress in physics has been slow it is because the problems are extremely difficult and each generation of accelerator or fusion reactor costs far, far more. 
 
There has been significant progress in optics, nanotech, super conductivity, quantum mechanics, and cosmology. 
 
"Nuclear engineering has stagnated" 
 
There are many new, innovative designs. The problem has been a public that fears nuclear power. That is changing. 
 
"unconventional energy schemes such as tidal dams and OTEA have dropped from sight" 
 
The Hawaiian OTEA research station was never able to produce competitively priced electricity. There are many ongoing tidal dam projects around the world. Likewise for geothermal. These projects have major engineering problems which is why windmills and solar power dominate the news. 
 
"huge delays in engineering projects of sorts have become commonplace" 
 
The political/legal environment slows down big projects. 
 
In the 1950's the US didn't even have an interstate highway system. Today cities have massive infrastructure which require maintenance and hampers new development, e.g., digging a ditch means cutting through existing pavement, plumbing, power lines, and communication lines. 
 
"virtually all forms of modern day engineering seem to have atrophied into mere shells of what they once were" 
 
Better materials, better design tools, etc. Buildings must withstand earthquakes and be energy efficient. Compare a modern bullet train to a 1950's train. Compare modern mining equip
More....

JEB: "There are thousands of genes that affect intelligence, and even the smartest people have no more than (say) 52 percent of the "good" alleles,...". 
 
This model assumes random mating with thousands of common additive variants of very small effect determining the genetic component of a person's intelligence. Compare an IQ 150 person to an IQ 50 person where a 4% difference in "good" alleles would account for 50 of the 100 IQ point difference (assuming heritability of 0.5). If the difference between the good and bad variants caused an 0.1 point difference in IQ (genome wide association tests have established an upper limit on the effect size of common variants) there would have to be 12,500 common variants that cause a 0.1 point IQ difference. However it is even worse because the variant effect sizes are likely to follow a power law with only a few common variants producing even 0.1 point difference in IQ. So many more IQ altering variants would be required. 
 
Spread out over the functional areas of the human genome this would imply virtually all regions have some minor impact but no regions have significant impact. This doesn't match the gene expression patterns seen throughout the body. Protein expression in the brain directly affects brain function and some proteins are far more important than others. Brain endophenotypes such as regional differences in gray matter and white matter volumes correlate highly with IQ and are heritable. I.e., the patterns of biological tissue differences don't match this model of genetic intelligence. 
 
So what is happening? 
 
1) There could be many different rare variants of large effect. Or SNP association tests may be missing common causative genetic factors. If so, only a modest number of variants might largely determine intelligence in each person. 
 
2) Assortative mating tends to concentrate good variants and bad variants so the estimate that "the smartest people have no more than (say) 52 percent of the "good" alleles" is wrong. 
 
3) The inheritance models only fit the usual IQ ranges between 70 and 130. I.e., studies of very high IQ populations might show different patterns. E.g., stochastic development factors may play a larger role in the very high IQ. Or non-additive genetic factors may play a larger role. Or the tests used to measure very high IQ might not be very reliable (more noise at the top or the "g" factor begins to separate into non-correlating intelligence subtypes). 
 
4) I don't know. 
 
Note that even within families there can be considerable IQ variance. E.g., my IQ is over 160 while my brother's IQ is around 115 and we share over half our genetic variants. 
 
"...so culling those people out will have very little impact on overall gene frequency, except in the very long run" 
 
Most extreme retardation in white people is or
More....

"Students below roughly 115 IQ won't remember it just due to that." 
 
Does anyone have a good link that discusses the relationship between IQ and episodic or semantic memory. My recollection is that intelligence does not correlate strongly with long term memory. (Intelligence does correlate with working memory.) 
 
When I understand a subject well then I see patterns between the facts. I also see mappings between what I already know and what I am learning. I can then remember specific facts by recalling the pattern or the mapping. So intelligence helps me remember structured information. However, I don't think intelligence helps me remember a phone number. (Intelligence would help me learn a method for memorizing phone numbers.)

"On the other hand, whole swaths of factory work will be eliminated, as robots are more reliable and can work in more hazardous conditions, without benefits." 
 
KIVA warehouse robots: http://www.spectrum.ieee.org/jul08/6380 
MIT greenhouse robots: http://gadgets.softpedia.com/news/MIT-Student-Develop-Robotic-Gardeners-2026-01.html 
 
In the last twenty years processors, sensors, and software have gotten much better and much less expensive. The time is now ripe for automation of many unskilled jobs. We are presently in a transition stage where unskilled humans are still needed for most jobs. However, many of those jobs are being restructured so that robots work together with unskilled laborers. One unskilled laborer + 5 robots can do the work formerly done by several unskilled laborers while also reducing response time and errors. In another decade the robots will be far better and cheaper and more jobs will have been restructured for robots. 
 
Personally I have no idea how a young person should prepare for a modern career. 
 
Plumbers and handymen should still be needed. However, the entry barriers to such jobs are low and my guess is that as low skilled workers lose jobs in manufacturing and agriculture they will move into other low skill jobs and reduce wages. E.g., suppose I'm a skilled plumber with my own business. In the new job market I can hire many unskilled labors. With their help I can complete far more plumbing jobs. These workers will take on more and more of the skilled work. Eventually I will be managing many low paid semi-skilled plumbers. My low cost plumbing business will lower the wages paid to other plumbers. (This already happened in the landscape business.) So low wages should spread from one low skilled occupation to the next. 
 
I'm not sure doctors will fare much better. Suppose a doctor specializes in breast radiology. Medical advances lead to molecular diagnostics that detect early stage cancer from protein markers in the breath or blood. Advanced chemo or immune therapy then cures the cancer. The $250,000 a year breast imaging specialist is no longer needed. 
 
Within a few decades no one, no matter how smart or how well educated, will be able to keep up in a job market where old occupations rapidly disappear or drastically change.

"All you really learned you learned on your own." 
 
If you have an IQ over 130 then you can learn non-science material better on your own than in a classroom. If your IQ is over 140 then you can learn basic science and math better on your own. If your IQ is over 150 you can learn all undergraduate university subjects better on your own. People with very high IQ's pretty much assume that school is worthless because it didn't help them. 
 
The US military might be an excellent model for US education. They track by ability. They only teach topics that are important for career success. They have rigorous standards for success. They have monitoring and feedback to improve training. 
 
Hmmm, here is a modest proposal. Let the grade schools operate as they do now. Let the US military educate students with IQ's less than 130 for grades 7-12. Students with IQ's over 130 would teach themselves under the guidance of experts in various fields. Those students would take online tests to demonstrate competence in the basic subjects. At 18, the average students would have a good basic education and be prepared to enter a trade or begin training for a profession. The very bright students would begin an apprenticeship with a practicing expert, much as graduate students work with their advisor. The expert could be a lawyer, doctor, engineer, scientist, business manager, etc. and the student would learn by doing.

This is cool. 
 
Step 1: Look for weak associations between a trait and common variants to get a list of potentially important DNA regions. 
Step 2: Closely examine those DNA regions in people with extremes in that trait looking for rare variants with strong effect. 
Step 3: Use the rare variants of strong effect to identify the important genes, proteins, and molecular pathways underlying a trait. 
Step 4: Tie it all together to get a good DNA-protein-system model of the trait. 
 
This would be a fast track method for predicting phenotype from genotype. 
 
Sequence a person's genome and identify variant DNA. Some of the variants will be common and have a known effect. Some of the variants will be in regions that are known to have little affect on the trait. Some of the novel variants will lie in DNA regions known to be important for that trait. Combined with a good model connecting the genotype through molecular mechanism to phenotype, one then predicts how the novel variant will affect the trait. E.g., a novel variant causes a change in a critical part of a protein. Knowing the protein's function in the biological system the doctor then uses a model to predict the affect of that DNA variant on the trait.

bbartlog: "And why wouldn't it be a viable treatment for HIV?" 
 
For now it is too dangerous, 1/3 die during the transplant. Finding a compatible donor who is also protected against HIV would also be hard. 
 
HIV persists in non-blood cells. However, the protected immune system should hold the disease in check.

I think bone marrow is a primary source for most stem cell types. I.e., stem cells released from the bone marrow can migrate, take up residence in other tissues, and then differentiate and integrate into the tissue. Combined with local growth factors and factors that increase cell turn-over, this could be used to replace most original tissues (replacing too many neurons might cause memory and skill loss). 
 
The bone marrow transplant could be done gradually. Inject new stem cells into the blood each day combined with a drug that increases stem cell migration. Gradually, all of the old stem cells would be replaced. (No radiation or chemotherapy and no loss of immune function during the transition.) (Autoimmune diseases might require wiping out immune "memory cells" before reconstructing the immune system.) 
 
The new stem cells could correct genetic diseases such as sickle cell anemia, fight persistent diseases such as HIV, cure cancer (see cancer resistant "PAR" mouse) or cardiovascular disease, or enhance a person by improving genotype. 
 
The technology needs to improve: 
1) Safety - host/graft rejection. Scientists need better technology for controlling and training the immune system. 
2) Better technology for isolating, growing, and differentiating the stem cells. 
3) Better control of the migration, differentiation, and integration of stem cells into target tissues. 
4) Scaffolding technology to support growth or renewal of critical body structures, e.g., heart valves, joints, and nerves. 
Significant progress is being made in all these areas. Within twenty years I expect to be a chimera. 
 
PS Such technology is one reason I tend to discount concerns about the aging population or dysgenic demographic trends.

Razib: "we really don't know the correlation between # of del. alleles & fitness, do we?" 
 
If you could rank the "fitness penalty" of specific mutations you would likely see a power law distribution. A few mutations are fatal, more are moderately harmful, far more are slightly harmful, and the vast majority are essentially neutral. Evolution through natural selection is strongly weighted toward removing the most harmful mutations. I.e., selection preferentially removes the more harmful mutations lying toward the left side of that distribution. More selection pressure results in less harmful mutations also being removed. Flies occupy the peaks of the fitness landscape while humans wander around the foothills. (E.g., human regulatory sequences tend to be less conserved than mouse regulatory sequences.)

P-ter: "I'm not sure this (a large number of strong selective sweeps) is the same situation considered by Haldane;" 
 
Right. I didn't intend to diss Haldane. 
 
In a large, single, connected, panmictic, stable population in an unchanging environment the specie should be highly adapted to that environment. In this case there would be few beneficial mutations of even moderate selective advantage in the population (harmful mutations are quickly removed by selection). The probability that one animal would have several more good variants than a competing animal would be small (sexual selection and thresholding are thus less efficient at propagating the good variants). There would be few opportunities for beneficial haplotypes to recombine to form even better haplotypes. Instead, a good mutation would slowly sweep the population at a rate determined by the selection advantage of that trait. In that case, Haldane's Limit applies. 
 
P-ter: "If you consider a single selected allele at a selection coefficient of 1%, that 1% is relative to all individuals not carrying that allele (depending on how you model selection). But those individuals are also carrying alleles that, in this model, are highly beneficial. So if this allele had arisen on its own (ie. not in a poppulation with thousands of other selected alleles), I think we have to conclude its selection coefficient would have been orders of magnitude larger?" 
 
P-ter: In my mental model I separate the early stage from the middle and late stages. 
 
In the early stage there are thousands of slightly beneficial variants with selection coefficients of around 0.1% compared to wild type (large population in changing environment generates the variants). These variants are rare so there isn't much interference between variants. In parallel, all the variants slowly increase in frequency. At this stage, the variants neither slow nor accelerate the sweeping of other variants. 
 
In the middle stage, the frequency of the variants has increased so that the variants are beginning to interact. No one variant is common but with thousands of good variants each animal may have several good variants. At this stage, sexual selection and thresholding begin to be more efficient at replacing the wild types with the good variants, i.e., each "selection death" promotes several good variants. Also, good haplotypes recombine to form even better new haplotypes with increased selection advantage. As the new haplotype replace the wild haplotype, several good variants replace the old wild variants. A few of the new haplotypes will have selection coefficients of 1% or more compared to wild type and will begin sweeping rapidly. 
 
In the last stage, the new haplotypes containing the variants are now common. They interfer with each other, they merge and split creating new haplotypes that are optimal for each specific envir
More....

"What is not clear to me is what happens if there are a thousand such genes in the population at once. Is the process additive? Does each advantageous gene increase in frequency independent of the others, as if it was the only advantageous gene in the population? Or does the process get clogged up? If everyone has a random selection of 200 genes each giving them a one percent advantage, then does anyone actually have an advantage over anyone else? How does this work?" 
 
See David's GNXP article: http://www.gnxp.com/blog/2006/04/haldanes-dilemma-should-we-worry.php 
 
Haldane makes assumptions that don't apply to real world populations. 
 
Sexual selection. The most successful stag will likely have many better alleles compared to his rival. That stag will produce far more offspring than would be predicted by considering the relative fitness value of each allele. In one generation, one stag may propagate hundreds of beneficial alleles. 
 
Nonlinear thresholds where many harmful variants combine in the same animal. The animal dies, is infertile, or fails in mate competition and many inferior alleles are removed. 
 
Chromosome recombination during meiosis alters the fitness coefficient. Suppose "A" and "B" are nearby beneficial variants of a chromosome with wild types "a" and "b". At first haplotypes "Ab" and "aB" will have a modest selective advantage over haplotype "ab". Eventually there may be a recombination event creating a new haplotype "AB" which has significant selective advantage over haplotype "ab" and modest selective advantage over haplotypes "Ab" and "aB". The new haplotype will sweep faster and will replace two wild types at the same time. Hence, nearby beneficial variants on the same autosome could accelerate sweeps. 
 
Many thousands of existing variants persist and new variants arise in large populations. The changing environment gives some variants a selective advantage. In parallel, the frequencies of these rare variants increase. Recombination eventually produces superior haplotypes with even greater selective advantage. Those haplotypes sweep the population. Haldane's Limit is left in the dust. 
 
Human populations are even more complicated. Population substructure matters. Migration patterns matter. Inheritance of family wealth, power, or prestige matters. Reality is complex with diverse and changing environmental niches, new haplotypes continually arising, balancing selection, partial sweeps, stochastic historical events, etc. Imagine trying to model the spread of Gengis Khan's Y-chromosome.

Trans variation should be common due to copy number variation. Multiple copies of the DNA that codes the trans regulatory elements should change the expression levels of the regulatory elements. Extra copies would also be free to mutate and take on new functional roles. Fast track adaptation.

David, this was interesting. 
 
"By far his most common example is that of a quantitative trait controlled by several loci where the selective optimum for the trait is at an intermediate value, i.e. neither the highest nor the lowest that can be produced by the various possible combinations of alleles. In this situation it is likely that the optimum intermediate value of the trait can be produced by different allele combinations. The effect of an allele on fitness (not necessarily on the quantitative trait itself) is epistatic, i.e. dependent on the combination of other genes in the genotype. Which of the relevant alleles are favoured by selection may then depend on the accident of which allele at a locus happens to be most frequent when selection begins, with all other alleles at the locus being driven to extinction." 
 
The above example is particularly interesting. Consider gene expression controlled by trans-acting transcription elements. Being short, such regulatory elements should remain functional long after a duplication event. Thus many functional copies could be spread around the genome. The total number of active copies would be maintainded by selection but no individual copy would be preserved. This mechanism would tend to preserve diversity in gene expression and support fast adaptation to new environments. 
 
If the population consisted of many small tribes with limited gene flow then drift would favor "robust combinations", i.e., trait values would be kept near the optimum even when the genetic background shifted due to drift. If so, long term evolutionary pressure should produce modularization of traits that needed to vary with the environment, e.g., changing tooth shape to match diet without harming other traits.

Punnett Square: "Do you have a link for that? I've never heard that Schiz or any other common mental disorder is that heritable." 
 
Here is one representative link. I don't view such studies as conclusive evidence of high heritability, I only conclude that schizophrenia is a complex disease that is strongly influenced by both genetic and environmental factors. 
 
http://www.ncbi.nlm.nih.gov/pubmed/14662550' 
 
"By using a multigroup twin model, we found evidence for substantial additive genetic effects-the point estimate of heritability in liability to schizophrenia was 81% (95% confidence interval, 73%-90%). Notably, there was consistent evidence across these studies for common or shared environmental influences on liability to schizophrenia-joint estimate, 11% (95% confidence interval, 3%-19%). CONCLUSIONS: Despite evidence of heterogeneity across studies, these meta-analytic results from 12 published twin studies of schizophrenia are consistent with a view of schizophrenia as a complex trait that results from genetic and environmental etiological influences."

Punnett Square: "Scientists know today that most cases of Schiz are the result of environmental damage." 
 
From the DailyScience link you provided: 
 
"Since schizophrenia and autism have a strong (though elusive) genetic component, there is no absolute certainty that infection will cause the disorders in a given case, but it is believed that as many as 21 percent of known cases of schizophrenia may have been triggered in this way. The conclusion is that susceptibility to these disorders is increased by something that occurs to mother or fetus during a bout with the flu. 
 
Now, researchers have isolated a protein that plays a pivotal role in that dire chain of events. A paper containing their results, "Maternal immune activation alters fetal brain development through interleukin-6," will be published in the Oct. 3 issue of the Journal of Neuroscience. 
 
Surprisingly, the finger of blame does not point at the virus itself. Since influenza infection is generally restricted to the mother's respiratory tract, the team speculated that what acts as the mediator is not the mother's infection per se but something in her immune response to it." 
 
I.e., genes and environment together produce disease. 
 
The brain is an exceeding complex organ. Brains can fail from a vast number of distinct causes. Factors may combine to increase the probability of failure. Schizophrenia is a type of brain failure defined by a fuzzy set of symptoms. Specific combinations of genetic factors may increase the probability of schizophrenia...twin studies show heritability over 80%. Specific environmental damage may increase the probability of schizophrenia. Genes and environment interact to produce phenotype. For complex traits such as mental function there will be many genetic and environmental causes.

Matt: "You'd need >1000 different ways to cause schizophrenia in order to get a 1% mutation-selection equilibrium frequency." 
 
Your estimate depends on the rate of harmful mutations as well as the number of ways that failure may occur. Some types of mutation occur far more frequently than the genome average for SNP's. 
 
E.g., during meiosis recombination can cause duplications/deletions at DNA "hot spots" where similar DNA sequences occur multiple times along the same strand. This type of mutation occurs frequently. I've seen estimates that up to 5% of severe retardation cases are due to recent "hot spot" CNV's in DNA that codes for brain proteins. The children seldom reproduce so the mutations are quickly removed from the gene pool. However the deletions occur so frequently that they account for a significant fraction of mental retardation cases. 
 
Fragile X retardation occurs in 1 in 4000 boys and is caused when a CGG segment exceeds around 200 copies. Normally a person has between 5 and 40 copies. This type of mutation occurs frequently.

Genetic tests are refining disease categorization. E.g., in breast cancer, pathologists using cell morphology or staining can't distinguish between two major classes of ductal carcinoma. One class has only a small chance of remission and doesn't require chemotherapy. The other cancer class is likely to return unless the patient undergoes chemotherapy. A gene expression test, MammaPrint, can reliably determine what type the patient has and whether chemotherapy is appropriate. 
 
Better genetic testing will lead to refined diagnosis which will lead to improved treatment. This will become very important when patients routinely have their genome scanned. That should begin happening in about five years.

"But his main point is that there will be selection in favour of closer linkage between favourable gene combinations on the same chromosomes, and it is therefore a puzzle why recombination is as frequent as it is. I think this remains a problem." 
 
I think it is only a problem if adaptation is slow and that new good alleles are rare. Here is how I see it: 
 
The advantage of recombination combining new good alleles on the same chromosome segment outweighs the disadvantage of recombination breaking old good linkages. I.e., when two good alleles are present at low frequencies on competing chromosome segments then a high recombination rate increases the chance of a new chromosome segment with both good alleles. The new chromosome segment with both good alleles begins sweeping the populace, picking up more good alleles along the way. (I use "chromosome segment" rather than chromosome since the human recombination rate is so high that the unit of inheritance is typically smaller than a chromosome. There are between ten and twenty crossovers for each chromosome pair during meiosis.) 
 
A high rate of recombination favors adaptation when many good alleles are circulating in the populace on competing chromosome segments. This should occur when the environment changes rapidly or when a large population is generating many new good alleles. E.g., the last fifty thousand years of human history. 
 
PS I second the "Bravo!". Excellent post.

Bbartlog, good connection to that earlier paper. Thanks for the link.

re: Neural development of the cortex. 
 
Kaleidoscopik, clearly your concerns are justified. My reply is more of an intuition and a hope than a prediction. 
 
The brain may have repair mechanisms separate from the developmental program. I.e., when a neuron dies then brain stem cells and progenitor cells migrate to the wound site and differentiate into the proper neural type based on the chemical, mechanical, and electrical signals in the local environment. Perhaps many new neurons would be produced and only those that made proper connections would survive. (I suspect that is how the new neurons generated in the dentate gyrus of the hippocampus help in the formation of new memories.) The mouse brain continually replaces neurons lost in the olfactory bulb (where they are exposed to a harsh environment). The new cells must migrate significant distances and properly integrate into functional tissue. So at least some parts of the mouse brain already self-repair. I also believe there is evidence for minor levels of repair in other brain tissues. If so, it might be relatively simple to renew brain tissues by temporarily increasing the neuron production rate and decreasing the new neuron apoptosis rate while targeting senscent cells for destruction. 
 
I suspect that mammals brains are more plastic than insect brains. I.e., brain circuitry in drosophila is genetically controlled down to the low level structures while only the upper level architecture is fixed in mice. If so, then the repair would just have to be in the right ball park and then training would optimize the resulting tissue. If this is the case then we might not need to replicate the developmental path that creates cortical minicolumns. Instead, cells would migrate to a region that was sending signals indicating repair is needed. There they would divide and differentiate into the appropriate neurons. The local tissue environment would provide the signaling needed to organize the cells into a functioning cortical minicolumn. 
 
"another concern is the surgical precision it would take to stick a dot of neural stem cells in the right spot" 
 
Yes, my hope is that other cell migration methods already exist in the brain and that we only need to supply stem cells to the cerebrospinal fluid and increase the rate of cell turnover. If instead we have to directly insert the right cells in the right location and directly controlled the local signaling environment then rejuvenation becomes much harder. In that case we would need robotic micro devices to do the implanting. That would take several more decades of technological advance.

Statins have unexpected effect on pool of powerful brain cells: 
http://www.eurekalert.org/pub_releases/2008-07/uorm-shu070208.php 
 
"Scientists found that both compounds, when used at doses that mimic those that patients take, spur glial progenitor cells to develop into oligodendrocytes. For example, in one experiment, they found about five times as many oligodendrocytes in cultures of human progenitor cells exposed to pravastatin compared to cultures not exposed to the substance. Similarly, they found that the number of progenitor cells was just about one-sixth the level in cultures exposed to simvastatin compared to cultures not exposed to the compound." 
 
"These are the cells ready to respond if you have a region of the brain that is damaged due to trauma, or lack of blood flow like a mini-stroke," said Sim, assistant professor of Neurology. "Researchers need to look very carefully at what happens if these cells have been depleted prematurely." 
 
"Glial progenitor cells are distributed throughout the brain and, according to Sim, make up about 3 percent of our brain cells. While true stem cells that can become any type of cell are very rare in the brain, their progeny, progenitor cells, are much more plentiful. They are slightly more specialized than stem cells but can still develop into different cell types."

bbartlog: "...you're glossing over the fact that a lot of higher-level structures are the result of a developmental process, and the results are not going to be changed by this maintenance or piecemeal replacement therapy" 
 
Yes, I'm glossing over a lot of stuff. There are significant questions regarding how plastic adult bodies really are. That is why I provided the link to "curing" the autistic mouse. There are traits that one would have thought were set during development that can be changed by rather crude genetic engineering on adult cells. I'd love to see mouse experiments that explored the plasticity of the adult animal, especially the adult brain. 
 
I believe that large structural remodeling will soon be achieved. E.g., regrowing a finger or repairing a spinal cord injury. I believe present medical research will lead to rebuilding brain structures that have been destroyed due to injury or disease. It should be possible to enlarge the skull with surgery, allowing the brain to expand. I tend to believe that gradual changes in brain size and structure wouldn't be too disruptive of existing memory and skills. Old people can recover from stroke and young children show significant plasticity in rewiring around damaged areas. 
 
I don't really know what to do about restoring nerve fibers that connect different brain regions. (Researchers are working on restoring the optic nerve connection. http://www.scienceblog.com/cms/node/7070 ) The original connections depended on a carefully orchestrated sequence of cellular events and signaling combined with external environmental feedback and pruning. Duplicating that process would likely destroy memories and skills. For the next few decades we might have to support the existing neurons with their axon connections by rejuvenating oligodendrocytes, astrocytes, and brain stem cells. That would limit potential intelligence improvements. (Brain rejuvenation is difficult.)

Ben G.: ".. taking it would be (wrongly or not) labeled by many group Y members as an admittance of intellectual inferiority[1] (many don't want to contribute to the headline "Group Y seeks out controversial IQ gene in record numbers") 
 
When the technology is commonly available I think such concerns would be marginalized. 
 
The individual would mainly be focused on the difference it would make in their own life. Consider the popularity of cosmetic surgery and then imagine a treatment that made you younger, improved your looks, and made you smarter. Once people saw the results of the treatment on friends and family they would demand access for themselves. 
 
The early adapters would be old, rich people. There would be strong public pressure to make the treatments available to everyone as a matter of social justice. 
 
Those people that are committed to identity politics would want to make certain that "their people" didn't get left behind. There might be racial pride initiatives to mobilize the group members to rapidly use the treatments. 
 
I do see potential problems. The first treatments might go awry in numerous ways. People might demand access to the treatment before all the bugs are worked out. Also, what would the parent of an autistic child do? 
 
Finally, I may be naive but I believe most people do want a better world for everyone. I suspect many liberals completely reject biological determinism because it provides no solution to social injustice. Likewise many conservatives reject "socially progressive" programs because they believe the programs don't work and often make the problems worse. An effective technology would win converts from both groups.

Keith: "Wouldn't implanted stem cells get rejected by the immune system?" 
 
If you just injected adult stem cells from an immune incompatible donor without first destroying the the patient's immune system then the transplanted cells would be rejected. As long as the donor cells are immune compatible then there is no rejection. There are multiple solutions to the rejection problem. 
 
Choose a donor cell line that is immune compatible. 
 
Do minor genetic engineering on the donor cell line so that it is compatible with all humans. 
 
As part of the rejuvenation program, rebuild the immune system. This has the advantage that new immune system might attack the senescent cells of the old type. Immune suppressant drugs could be used to control the rate the old cells were attacked. 
 
Advances in biotech to manipulate the immune system will have advanced to the stage where rejection can be controlled or stopped. I believe this level of immune system technology will be available before we have the technology for tissue rejuvenation. (Advanced immune system technology should also reduce the threat of cancer.)

"wouldn't it become a point of group pride not to take the "other guy's gene's" pill?" 
 
I believe there are enough exceptional members in all large groups that a person could choose a donor from their own group. It would be up to the individual whether to change groups or not. 
 
An exceptional individual with an IQ over 200 and demonstrated accomplishment might prefer to keep his own genes. Such an individual might convince others to use him as a donor. I imagine becoming a donor would confer high status. 
 
I don't really know how most people would react to this choice. For myself, it would be an easy. Grow old and deteriorate or renew myself in a better body. If the renewal process caused me to lose many important memories, the decision would be harder. I'd probably defer it until very old. 
 
Also, I don't really think of this as gene engineering, i.e., selecting and combining specific alleles. The stem cells would be from individuals known to have superior phenotype. Stem cell transplantation should be easier than genetic engineering of superior phenotypes. It would be more like cloning a superstar.

gc: "1) SNPs are a priori going to be of small effect as they change only one bp. CNVs, by contrast, are changing kilobases or megabases at a time (i.e. entire genes or exons)." 
 
Relevant links: 
http://microarraybulletin.com/community/article.php?p=284&page=1 
http://www.bio-medicine.org/medicine-news/Genetics-Of-Mental-Retardation-13248-1/ 
 
5% of the retardation cases are likely due to recent CNV's in brain genes. As far as I know all races are equally likely to experience such hotspot CNV's. These retardation CNV's are so harmful that they are rapidly eliminated by selection. I doubt they are the source of genetic group differences. 
 
Other CNV's with less drastic effects are likely common. It would be interesting to compare the hotspot distribution differences between racial groups. Looking at CNV hotspots might make identifying CNV variation more efficient. 
 
SNP's in regulatory DNA could significantly affect gene expression. Since there is twice as much conserved regulatory DNA as coding DNA in the human genome there are a lot of potential variants in the human population.

Gradually replace existing stem cells in all body niches with stem cells cultured from the best genotypes of each race. Use drugs to increase senescent cell turn-over. Use growth factors and drugs to increase the rate of tissue replacement. Within a few years this would remodel the body and brain. Some differences fixed during development would remain, but much should improve. "Cosmetic surgery" for the masses. 
 
DNA would no longer be destiny.