# [ExI] re end of world

rex rex at nosyntax.net
Fri Sep 11 05:03:21 UTC 2015

```Rafal Smigrodzki <rafal.smigrodzki at gmail.com> [2015-09-10 17:10]:
>    On Thu, Sep 10, 2015 at 6:38 PM, Anders Sandberg <[1]anders at aleph.se>
>    wrote:
>      What is going on is that the intelligence is due to a lot of genes
>      adding about 1% or less each, so removing the <150 IQ population mainly
>      removes some very bad genes and decreases the somewhat bad genes: there
>      is still a lot of genetic diversity. So the new mean goes up, but the
>      variance stays about the same.
>
>    Are you sure the variance would stay the same? Wouldn't
>    removing 99.9570883466% of the population from the reproductive pool lower
>    diversity?
>    The IQ of their offspring would be of course less than 150, with
>    regression to the mean, but there would be still a substantial boost over
>    the current average, to about 130
>    ([2]http://infoproc.blogspot.com/2010/07/assortative-mating-regression-and-all.html)
>    and the residual SD would be about 12.

Good find, Rafał!

I worked on this some years ago and never was able to create a chromosome model
(with crossover, etc) where the variance didn't grow over generations.

A model often used to predict offspring IQ (or other heritable
polygenic trait) is:

oiq = m + h^2*(p-m)   (Crow)

where oiq is the mean IQ of the offspring, h^2 is the narrow
heritability, m = population mean IQ, and m = parental mean IQ.

For IQ, IMO, 0.7 is a reasonable value for h^2 (The heritability of IQ increases
with age and childhood estimates underestimate the adult heritability of IQ.)

I think the value of rho (correlation of mate IQ due to assortative
mating) is fairly "soft," and use 0.25.

R makes it easy to simulate the process.

pop = rnorm(9e6, 100, 16)           #generate 9 million Gaussian randoms, mean 100, SD 16
N1 = 1000                           #size of high-IQ breeder population
hiq = numeric(N1)                   #IQs of breeders
minIQ = 150                         #minimum IQ of breeders
cnt = i = 1                         #pointers
while (cnt <= N1){                  #do until quota filled
if (pop[i] >= minIQ){             #qualified
hiq[cnt] = pop[i]               #add individual to breeder population
cnt = cnt + 1                   #bump number of breeders
}
i = i + 1
}
mean(hiq)
[1] 154.4818
sd(hiq)
[1] 4.12788
plot(density(hiq),main='Minimum IQ 150', xlab='IQ')

As the graph shows, the selected group with minimum IQ has a skewed
distribution. Roughly, the offspring of this group would be expected
to have a mean IQ of:

oiq = 100 + 0.7*(154.5-100) = 138

It would be a very different world, and IQ would drift toward whatever value
was currently selected for.

>    A good reason to make and sequence thousands of zygotes before proceeding
>    to term. For the squeamish you might try to develop a post-meiotic gamete
>    cloning procedure, followed by mass sequencing and zygote generation from
>    vetted but otherwise unmodified gametes.

The mean IQ could be dramatically raised if IQ could be estimated from a cell from
blastocyts. Couples could pick the highest IQ blastocyst from, say, 5. I've a web page
that illustrates mean population IQ over generations if some fraction of parents
follow such a procedure. If there's interest I'll put it on the WWW again.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: high_IQ_selection.png
Type: image/png
Size: 36509 bytes
Desc: not available
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20150910/81079a1c/attachment.png>
```