The problem is that no one knows how to go for geniuses.
I don’t follow. If you have the sorts of genotype/phenotype databases which let you select for a few variants to increase average intelligence a little bit, then you aren’t technologically very far from having the databases to select for a lot of variants to increase average intelligence a lot. I don’t see any reason to expect long-term stagnation where interventions can easily increase by a few points but a lot of points is just impossible.
If you have the sorts of genotype/phenotype databases which let you select for a few variants to increase average intelligence a little bit, then you aren’t technologically very far from having the databases to select for a lot of variants to increase average intelligence a lot.
First, no one has databases which let you select even a few variants. We know a bunch of mutations which reliably decrease intelligence. I don’t think we know what reliably increases it.
Second, the idea that we can just pile all the small improvements together to get a supergenius relies on unlikely assumptions, for example the additivity of these improvements and lack of negative side-effects.
First, no one has databases which let you select even a few variants. We know a bunch of mutations which reliably decrease intelligence. I don’t think we know what reliably increases it.
I am aware of this. But you were the one discussing the hypothetical that the Chinese government would be more likely to do an embryo selection program aimed at modest national-wide increases in averages; clearly you are presupposing that such databases exist, and so I’m not sure why you’re objecting that your hypothetical is currently a hypothetical.
Second, the idea that we can just pile all the small improvements together to get a supergenius relies on unlikely assumptions, for example the additivity of these improvements
My understanding is that, far from being an ‘unlikely assumption’, methods like twin studies & GCTA used to estimate aspects of the genetic contribution to intelligence have long shown that the majority (usually something like ~70%, going off Meng’s citations) is in fact additive.
hypothetical that the Chinese government would be more likely to do an embryo selection program aimed at modest national-wide increases in averages
I probably wasn’t clear. The hypothetical program would not be aimed at modest nation-wide increases. It would be aimed at figuring out how to genetically engineer intelligence. I expect that its first fruits would be modest increases in the averages of the program subjects—not the averages of the whole population of China.
the majority (usually something like ~70%, going off Meng’s citations) is in fact additive.
The studies examined normal ranges of intelligence. The additivity may or may not hold when pushing into genius territory.
High IQ types don’t have that many problems.
That’s not self-evident to me for very high IQ types. Besides, the attempts to genetically engineer high IQ might find different paths in that general direction, some are likely to have serious side effects.
The additivity may or may not hold when pushing into genius territory.
There’s no a priori reason to expect additivity to suddenly fail when going outside. That’s the point of additivity: if they depended on the presence or absence of other variants to have effect, then that would fall into the non-additive parts.
That’s not self-evident to me for very high IQ types.
When we look at regressions for IQ, we almost always see strong positive effects going as high as we can meaningfully measure or get sample sizes. Consider the SMPY studies. I’m not aware of any results from their longitudinal results showing worse problems than your average 100 IQ schmoe. And it should be self-evident: do you associate MIT or Stanford or Harvard or Tsinghua graduates with extremely high flameout rates, shorter lifespans, lower incomes, any of that...?
the attempts to genetically engineer high IQ might find different paths in that general direction, some are likely to have serious side effects.
If there were serious common side effects from the common variants detected by current GWAS, as a statistical necessity, those variants would have been disease hits before they were intelligence hits of small effect.
There’s no a priori reason to expect additivity to suddenly fail when going outside.
We just don’t know at this point. On general grounds I’m suspicious of claims that in highly complex systems stochastic relationships observed for the middle of the distribution necessarily hold far into the tails. In this case I have no strong opinions on whether it will or will not hold.
MIT or Stanford or Harvard or Tsinghua graduates
By “very high IQ types” I mean geniuses. MIT, Stanford, etc. do not graduate geniuses, they graduate merely high-IQ people. Off the top of my head, I would expect geniuses to have a higher rate of mental/emotional issues and a shorter lifespan, though that’s a prior, I haven’t looked at data.
Why do you have your skeptical prior? Where have similar genetic engineering efforts failed? Have we not been able to breed cows and cats and dogs and horses for all sorts of things and traits many SDs beyond their ancestral wild populations?
By “very high IQ types” I mean geniuses. MIT, Stanford, etc. do not graduate geniuses, they graduate merely high-IQ people.
If there are large negative effects then you should be able to show it by looking at the available large samples of high-IQ types, of which MIT/Stanford/SMPY/etc are the best ones. There’s not going to be any magical triggerpoint where IQ 150 people have all the benefits we know high IQ types do and which all the extrapolations predict and we verify up to the limits of our research capability, and then just beyond where we can gather reliable sample sizes, at IQ 151, suddenly they start to lose 20 years of life expectancy and go mad.
I would expect geniuses to have a higher rate of mental/emotional issues and a shorter lifespan,
It sounds like your beliefs on this topic are molded by some outdated Romantic myths about genius.
I’m talking about different paths.
I have no idea what you mean. All proposals are for using GWAS results based on existing variation (since no one knows what other genetic changes one would make!), and my argument for safety works there. What different paths?
Have we not been able to breed cows and cats and dogs and horses for all sorts of things and traits many SDs beyond their ancestral wild populations?
Not for intelligence, as far as I know. Though dog breeds are widely considered to vary in intelligence—have there been any attempts to quantify it?
As to “failing”, traditional genetic engineering certainly ran into some limits. To continue with dogs, large breeds have shorter lifespans. Many breeds have well-known pervasive genetic problems (hip dysplasia in German shepherds, etc.).
It sounds like your beliefs on this topic are molded by some outdated Romantic myths about genius.
I doubt it’s Romantic myths since people that come to my mind mostly lived in the XX century, but yes, I’ve said that it’s a prior and I’m open to evidence other than handwaving.
since no one knows what other genetic changes one would make
I doubt it’s Romantic myths since people that come to my mind mostly lived in the XX century
The Romantics invented the myth of insane geniuses touched by divinity, but that doesn’t mean people holding that belief suffer from amnesia and are unable to list any examples from after the Romantics… Given the lifetime prevalence of any mental illness in the general population, it would be surprising if one couldn’t list some anecdotes like Godel.
That’s the point of experimenting :-)
There’s a practically infinite number of genetic changes one could make. Understanding of genetic networks influencing cognition will have to be extraordinarily good before any researchers can write down a completely novel gene or variant which has no natural examples and experiment with it. For better or worse, for the next several decades, we’re stuck exploiting natural variants—all interventions are going to look something like ‘people with X seem to be smarter, let’s try adding X to others or select for it’ or ‘Y is a rare or de novo variant, maybe it’s harmful, let’s remove it or select against’.
That’s not my mental model at all. I haven’t thought deeply about it, but I probably imagine geniuses as an overclocked, supercharged, often highly specialized piece of wetware running on the same-reliability components, possibly crowding out some other capabilities, and frequently having social problems just due to the fact that 99.9%+ of people around you are quite different from yourself.
Though dog breeds are widely considered to vary in intelligence—have there been any attempts to quantify it?
Yes; take a look at this, and the generic wikipedia page. One of the more visible tests is the number of times a new command must be repeated to be learned. Overall, there’s not too much agreement because there are a number of different interpretations of what it means for a dog to be intelligent, and no one has (to my knowledge) done the factor analysis to look for g in dogs.
We do have such a test battery for primates, though, the Primate Cognitive Test Battery (came up 2014 in showing chimp intelligence is, of course, heritable). Cross-species comparisons have been done and don’t show much difference aside from humans: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3317657/ (and from a different avenue, http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3049098/ ). This is a bit surprising to me but I suppose it’s not like we’ve deliberately bred any of those species for intelligence or anything.
I don’t follow. If you have the sorts of genotype/phenotype databases which let you select for a few variants to increase average intelligence a little bit, then you aren’t technologically very far from having the databases to select for a lot of variants to increase average intelligence a lot. I don’t see any reason to expect long-term stagnation where interventions can easily increase by a few points but a lot of points is just impossible.
First, no one has databases which let you select even a few variants. We know a bunch of mutations which reliably decrease intelligence. I don’t think we know what reliably increases it.
Second, the idea that we can just pile all the small improvements together to get a supergenius relies on unlikely assumptions, for example the additivity of these improvements and lack of negative side-effects.
I am aware of this. But you were the one discussing the hypothetical that the Chinese government would be more likely to do an embryo selection program aimed at modest national-wide increases in averages; clearly you are presupposing that such databases exist, and so I’m not sure why you’re objecting that your hypothetical is currently a hypothetical.
My understanding is that, far from being an ‘unlikely assumption’, methods like twin studies & GCTA used to estimate aspects of the genetic contribution to intelligence have long shown that the majority (usually something like ~70%, going off Meng’s citations) is in fact additive.
High IQ types don’t have that many problems.
I probably wasn’t clear. The hypothetical program would not be aimed at modest nation-wide increases. It would be aimed at figuring out how to genetically engineer intelligence. I expect that its first fruits would be modest increases in the averages of the program subjects—not the averages of the whole population of China.
The studies examined normal ranges of intelligence. The additivity may or may not hold when pushing into genius territory.
That’s not self-evident to me for very high IQ types. Besides, the attempts to genetically engineer high IQ might find different paths in that general direction, some are likely to have serious side effects.
There’s no a priori reason to expect additivity to suddenly fail when going outside. That’s the point of additivity: if they depended on the presence or absence of other variants to have effect, then that would fall into the non-additive parts.
When we look at regressions for IQ, we almost always see strong positive effects going as high as we can meaningfully measure or get sample sizes. Consider the SMPY studies. I’m not aware of any results from their longitudinal results showing worse problems than your average 100 IQ schmoe. And it should be self-evident: do you associate MIT or Stanford or Harvard or Tsinghua graduates with extremely high flameout rates, shorter lifespans, lower incomes, any of that...?
If there were serious common side effects from the common variants detected by current GWAS, as a statistical necessity, those variants would have been disease hits before they were intelligence hits of small effect.
We just don’t know at this point. On general grounds I’m suspicious of claims that in highly complex systems stochastic relationships observed for the middle of the distribution necessarily hold far into the tails. In this case I have no strong opinions on whether it will or will not hold.
By “very high IQ types” I mean geniuses. MIT, Stanford, etc. do not graduate geniuses, they graduate merely high-IQ people. Off the top of my head, I would expect geniuses to have a higher rate of mental/emotional issues and a shorter lifespan, though that’s a prior, I haven’t looked at data.
I’m talking about different paths.
Why do you have your skeptical prior? Where have similar genetic engineering efforts failed? Have we not been able to breed cows and cats and dogs and horses for all sorts of things and traits many SDs beyond their ancestral wild populations?
If there are large negative effects then you should be able to show it by looking at the available large samples of high-IQ types, of which MIT/Stanford/SMPY/etc are the best ones. There’s not going to be any magical triggerpoint where IQ 150 people have all the benefits we know high IQ types do and which all the extrapolations predict and we verify up to the limits of our research capability, and then just beyond where we can gather reliable sample sizes, at IQ 151, suddenly they start to lose 20 years of life expectancy and go mad.
It sounds like your beliefs on this topic are molded by some outdated Romantic myths about genius.
I have no idea what you mean. All proposals are for using GWAS results based on existing variation (since no one knows what other genetic changes one would make!), and my argument for safety works there. What different paths?
Not for intelligence, as far as I know. Though dog breeds are widely considered to vary in intelligence—have there been any attempts to quantify it?
As to “failing”, traditional genetic engineering certainly ran into some limits. To continue with dogs, large breeds have shorter lifespans. Many breeds have well-known pervasive genetic problems (hip dysplasia in German shepherds, etc.).
I doubt it’s Romantic myths since people that come to my mind mostly lived in the XX century, but yes, I’ve said that it’s a prior and I’m open to evidence other than handwaving.
That’s the point of experimenting :-)
The Romantics invented the myth of insane geniuses touched by divinity, but that doesn’t mean people holding that belief suffer from amnesia and are unable to list any examples from after the Romantics… Given the lifetime prevalence of any mental illness in the general population, it would be surprising if one couldn’t list some anecdotes like Godel.
There’s a practically infinite number of genetic changes one could make. Understanding of genetic networks influencing cognition will have to be extraordinarily good before any researchers can write down a completely novel gene or variant which has no natural examples and experiment with it. For better or worse, for the next several decades, we’re stuck exploiting natural variants—all interventions are going to look something like ‘people with X seem to be smarter, let’s try adding X to others or select for it’ or ‘Y is a rare or de novo variant, maybe it’s harmful, let’s remove it or select against’.
That’s not my mental model at all. I haven’t thought deeply about it, but I probably imagine geniuses as an overclocked, supercharged, often highly specialized piece of wetware running on the same-reliability components, possibly crowding out some other capabilities, and frequently having social problems just due to the fact that 99.9%+ of people around you are quite different from yourself.
Yes; take a look at this, and the generic wikipedia page. One of the more visible tests is the number of times a new command must be repeated to be learned. Overall, there’s not too much agreement because there are a number of different interpretations of what it means for a dog to be intelligent, and no one has (to my knowledge) done the factor analysis to look for g in dogs.
We do have such a test battery for primates, though, the Primate Cognitive Test Battery (came up 2014 in showing chimp intelligence is, of course, heritable). Cross-species comparisons have been done and don’t show much difference aside from humans: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3317657/ (and from a different avenue, http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3049098/ ). This is a bit surprising to me but I suppose it’s not like we’ve deliberately bred any of those species for intelligence or anything.