Editing genomes for highly polygenic traits such as IQ is unlikely to be feasible with CRISPR because of off-site damage rates. A 10% rate (i.e. 90% probability of no adverse impact on the zygote) per edit is acceptable for editing low single digit number of sites, but as the number of edits grows, overall success rate drops exponentially. If you try to edit 10 sites, your success rate will be 0.9^10≈1/3. For 20 sites it's 1/8, for 30 sites it's 1/24, for 50 sites - 1/194, and for 100 sites - 1/37,600. If Steve Hsu’s estimates (http://arxiv.org/abs/1408.34210) of the genetic architecture of intelligence are correct and the number of sites that determine g is on the order of 10,000, with about 1,000 sites per standard deviation, editing for g is impracticable unless off-site damage rate per edit is brought down to about 0.25%. This estimate is crude and does not account for possible interactions between edits or for damage to the zygote from the sheer amount of enzymes and guide RNA dumped into it.
As for the kind of welcome that might be extended to hypothetical enhanced populations, a useful comparison is market-dominant minorities.
Good writeup. Of course most of it requires little/no prohibition at least somewhere in the world (which does seem pretty likely).
But I think you're missing what seems to me to be the most interesting consequences:
1 - As people get smarter, the way they think may change.
My personal observation has been that people with highish IQs (~115 to 135) are more leftist than most. It's not clear to me if the same is true of people with extremely high IQs (> 140).
Given that leftist societies tend to collapse, I wonder about the social consequences of rising median population IQs.
(I may be just conflating education effects with intelligence effects, in which case nevermind.)
2 - More important, this is a first step down a path of recursive genetic modification. We've some idea of how we'd change our children, if we could. We have less idea of how those children - different from us - will choose to change their own children - etc. down the generations. The path seems unpredictable, potentially chaotic, and may lead to extinction.
I don't think we have any good reason to think that after 5 or 10 generations of such changes, the result will look anything at all like present humans.
(This is similar to the "AI explosion" recursive improvement argument.)
Thank you for articulating your thoughts on this subject!
Some responses:
A couple points on the stratification of enhanced people into their own subcultures.
1. I expect there will be genetic mixing between enhanced and unenhanced populations. If enhanced males have a similarly broad range of sexual appetites as unenhanced human males do today, then enhanced males will likely continue to have sex with unenhanced women. This provides a point of contact between unenhanced women and enhanced males. I can see unenhanced women being interested in impregnation by enhanced males, even without the prospect of any parental investment. The biggest constraints here are the values and beliefs of the enhanced population. Perhaps it will be considered unethical to create these hybrids. Perhaps enhanced females will consider it low-status if an enhanced male has sired children with unenhanced women. Even if this is the belief in some subcultures, there is also the chance of the opposite beliefs being held. It could be considered an altruistic act for men to sire children with unenhanced women, somewhat similarly to how we treat sperm donation today.
2. Today, there are traditional old money values of respectability that keep rich kids from hanging out with people from different backgrounds. However, among those who have ascended in the network age, we see very different attitudes. It is not unusual for network elites to a) come from unconventional background themselves b) socialize with outcasts, degenerates and fringe people c) travel and engage with a diverse range of cultures from all over the world. While there certainly will be lineages interested in preserving their respectability, I expect that many lineages will propagate characteristics of high openness and xenophilia, creating the will and curiosity to explore beyond the confines of enhanced groups. This point plus point number one definitely creates the possibility of hybrid children, which has all kinds of possible recursions.
I came here from your post on the ACX classified thread, this was an interesting read maybe you should advertise this post over there.
In terms of IQ selection, I can imagine a future where AI is doing most mental work so the economic rewards from high IQ would be a lot lower. I kinda suspect most of the other benefits from high IQ derive from the economic gains or could be selective for independently. I also think high IQ might actually have a negative effect on happiness ( most communities that select for IQ seem quite glum in very specific ways to me, ASX has a pretty pessimistic outlook imo, for example), all else equal. So maybe in that scenario there'd be few incentives to increase it.
Standards for attractiveness seem quite culturally malleable/strongly relative, I'd argue almost everyone in the modern world would be considered attractive by medieval standards, just from good nutrition/hygiene, and would seem positively radiant, in an alien kind of way, to a homo erectus. So maybe if some of what is currently deemed attractive became more common it would start to lessen the appeal in a negative feedback loop, or new standards would keep emerging in a kind of cat and mouse game.
I wrote a short piece about the possibility of using this kind of technology to reduce peoples' heights' over on the EA forum
Very interesting speculations. I wonder if the culture of a eugenics-obsessed society would place more emphasis on reproduction of the highly enhanced. I would suggest that the current downward trend of fertility among the high IQ might be reversed in such a culture.
Thanks for the primer on genetic enhancement; it's a field of study that I try to keep up on, but I've never done a really deep dive. I'm not sure how I feel about the whole thing - I guess I can see both sides. As a Christian myself, I certainly come at the question of human fertility from the "be fruitful and multiply" angle. But is genetic enhancement "going too far"? I don't know. Part of me wonders if genetic enhancement is only trying to solve problems on the individual level, while overlooking societal problems - how would genetic enhancement solve the problems of systemic racism, environmental degradation, gender inequities, etc.
How do you think computer-human interfaces play into this? Von Neumann could multiply 8 digit numbers in his head, but now everyone can do it just as fast on their phone.
It's pretty easy to understand why people would want to do a few rounds of IVF, do some basic polygenic selection, and increase their kids IQ 15 points and be healthy without changing all that much about them.
When we start talking about this other stuff it just gets really speculative.
What does an IQ of 1,000 even mean?
A very good friend of mine comes from a family where all four kids are very high IQ and got perfect SAT scores.
The father was a weapons engineer. The four sons were a doctor, a Wall Street quant, and robot engineer, and the last was a violent autistic savant that has to be sedated all day long. All were somewhat autistic, and its pretty clear they come from a family that is very selective for IQ.
I'm not sure what would happen if you kept selecting for the traits that made them like they are. You might get more Nobel prizes. You might get more quant trading. Or you might get some super intelligent but mentally damaged individuals.
Imagine if you push a button and it spits out 100 people with an IQ of 1,000, but one of them is pre-disposed to be a "mad scientist" and had the mental tools to do it. The mediocre can't produce Nobel prizes, but at least they are basically harmless.
"Society like ours but a bit better" is easy to grasp, but this sci-fi stuff where people aren't really "human" anymore seems impossible to predict. Though perhaps inevitable I can't really bring myself to have any positive or negative reaction the way that simple and understandable genetic improvement seems like an unabashed positive.
Editing genomes for highly polygenic traits such as IQ is unlikely to be feasible with CRISPR because of off-site damage rates. A 10% rate (i.e. 90% probability of no adverse impact on the zygote) per edit is acceptable for editing low single digit number of sites, but as the number of edits grows, overall success rate drops exponentially. If you try to edit 10 sites, your success rate will be 0.9^10≈1/3. For 20 sites it's 1/8, for 30 sites it's 1/24, for 50 sites - 1/194, and for 100 sites - 1/37,600. If Steve Hsu’s estimates (http://arxiv.org/abs/1408.34210) of the genetic architecture of intelligence are correct and the number of sites that determine g is on the order of 10,000, with about 1,000 sites per standard deviation, editing for g is impracticable unless off-site damage rate per edit is brought down to about 0.25%. This estimate is crude and does not account for possible interactions between edits or for damage to the zygote from the sheer amount of enzymes and guide RNA dumped into it.
As for the kind of welcome that might be extended to hypothetical enhanced populations, a useful comparison is market-dominant minorities.
Good writeup. Of course most of it requires little/no prohibition at least somewhere in the world (which does seem pretty likely).
But I think you're missing what seems to me to be the most interesting consequences:
1 - As people get smarter, the way they think may change.
My personal observation has been that people with highish IQs (~115 to 135) are more leftist than most. It's not clear to me if the same is true of people with extremely high IQs (> 140).
Given that leftist societies tend to collapse, I wonder about the social consequences of rising median population IQs.
(I may be just conflating education effects with intelligence effects, in which case nevermind.)
2 - More important, this is a first step down a path of recursive genetic modification. We've some idea of how we'd change our children, if we could. We have less idea of how those children - different from us - will choose to change their own children - etc. down the generations. The path seems unpredictable, potentially chaotic, and may lead to extinction.
I don't think we have any good reason to think that after 5 or 10 generations of such changes, the result will look anything at all like present humans.
(This is similar to the "AI explosion" recursive improvement argument.)
How are you so informed on this topic?
Do you do all of this research yourself in your spare time, or do you work in this space?
Thank you for articulating your thoughts on this subject!
Some responses:
A couple points on the stratification of enhanced people into their own subcultures.
1. I expect there will be genetic mixing between enhanced and unenhanced populations. If enhanced males have a similarly broad range of sexual appetites as unenhanced human males do today, then enhanced males will likely continue to have sex with unenhanced women. This provides a point of contact between unenhanced women and enhanced males. I can see unenhanced women being interested in impregnation by enhanced males, even without the prospect of any parental investment. The biggest constraints here are the values and beliefs of the enhanced population. Perhaps it will be considered unethical to create these hybrids. Perhaps enhanced females will consider it low-status if an enhanced male has sired children with unenhanced women. Even if this is the belief in some subcultures, there is also the chance of the opposite beliefs being held. It could be considered an altruistic act for men to sire children with unenhanced women, somewhat similarly to how we treat sperm donation today.
2. Today, there are traditional old money values of respectability that keep rich kids from hanging out with people from different backgrounds. However, among those who have ascended in the network age, we see very different attitudes. It is not unusual for network elites to a) come from unconventional background themselves b) socialize with outcasts, degenerates and fringe people c) travel and engage with a diverse range of cultures from all over the world. While there certainly will be lineages interested in preserving their respectability, I expect that many lineages will propagate characteristics of high openness and xenophilia, creating the will and curiosity to explore beyond the confines of enhanced groups. This point plus point number one definitely creates the possibility of hybrid children, which has all kinds of possible recursions.
I came here from your post on the ACX classified thread, this was an interesting read maybe you should advertise this post over there.
In terms of IQ selection, I can imagine a future where AI is doing most mental work so the economic rewards from high IQ would be a lot lower. I kinda suspect most of the other benefits from high IQ derive from the economic gains or could be selective for independently. I also think high IQ might actually have a negative effect on happiness ( most communities that select for IQ seem quite glum in very specific ways to me, ASX has a pretty pessimistic outlook imo, for example), all else equal. So maybe in that scenario there'd be few incentives to increase it.
Standards for attractiveness seem quite culturally malleable/strongly relative, I'd argue almost everyone in the modern world would be considered attractive by medieval standards, just from good nutrition/hygiene, and would seem positively radiant, in an alien kind of way, to a homo erectus. So maybe if some of what is currently deemed attractive became more common it would start to lessen the appeal in a negative feedback loop, or new standards would keep emerging in a kind of cat and mouse game.
I wrote a short piece about the possibility of using this kind of technology to reduce peoples' heights' over on the EA forum
https://forum.effectivealtruism.org/posts/3caZ7LhMsvsS7kRrz/hobbit-manifesto
If you're interested.
Very interesting speculations. I wonder if the culture of a eugenics-obsessed society would place more emphasis on reproduction of the highly enhanced. I would suggest that the current downward trend of fertility among the high IQ might be reversed in such a culture.
Thanks for the primer on genetic enhancement; it's a field of study that I try to keep up on, but I've never done a really deep dive. I'm not sure how I feel about the whole thing - I guess I can see both sides. As a Christian myself, I certainly come at the question of human fertility from the "be fruitful and multiply" angle. But is genetic enhancement "going too far"? I don't know. Part of me wonders if genetic enhancement is only trying to solve problems on the individual level, while overlooking societal problems - how would genetic enhancement solve the problems of systemic racism, environmental degradation, gender inequities, etc.
How do you think computer-human interfaces play into this? Von Neumann could multiply 8 digit numbers in his head, but now everyone can do it just as fast on their phone.
Excellent essay.
Do you have a Metaculus profile of your own with public bets, Parrhesia?
It's pretty easy to understand why people would want to do a few rounds of IVF, do some basic polygenic selection, and increase their kids IQ 15 points and be healthy without changing all that much about them.
When we start talking about this other stuff it just gets really speculative.
What does an IQ of 1,000 even mean?
A very good friend of mine comes from a family where all four kids are very high IQ and got perfect SAT scores.
The father was a weapons engineer. The four sons were a doctor, a Wall Street quant, and robot engineer, and the last was a violent autistic savant that has to be sedated all day long. All were somewhat autistic, and its pretty clear they come from a family that is very selective for IQ.
I'm not sure what would happen if you kept selecting for the traits that made them like they are. You might get more Nobel prizes. You might get more quant trading. Or you might get some super intelligent but mentally damaged individuals.
Imagine if you push a button and it spits out 100 people with an IQ of 1,000, but one of them is pre-disposed to be a "mad scientist" and had the mental tools to do it. The mediocre can't produce Nobel prizes, but at least they are basically harmless.
"Society like ours but a bit better" is easy to grasp, but this sci-fi stuff where people aren't really "human" anymore seems impossible to predict. Though perhaps inevitable I can't really bring myself to have any positive or negative reaction the way that simple and understandable genetic improvement seems like an unabashed positive.