[ExI] What if humans were twice as intelligent?

[Anders Sandberg]

(Good overview, Rex!)

Yes, there are multiple factors involved in intelligence. They can be 
complex, but they are not magic nor guaranteed to preclude any 
predictability.

On 2013-10-22 11:20, rex wrote:
> Since humans mate assortatively, I'd like add add it to Anders'
> simulation, but it's not clear to me how to do that with a gene
> simulation. Perhaps Anders can think of a way to add it.

I would do something like this: select one parent with a probability 
proportional to their fitness (BTW, the mysterious selectrandom routine 
is basically just exploiting that if F(x) is a cdf of the distribution 
f(x) and u a uniform [0,1] random number, the variate F^{-1}(u) will 
have the probability distribution f(x)). Then we want to find a mate who 
is selected by fitness *and* has an IQ correlated with the first parent.

One approach would be to randomly generate a set of candidates (select 
random ones by their fitness), get the one with the closest IQ - the 
correlation between parents will be set by the number of candidates. I 
don't know if this actually works.

I suspect the *right* method is selection-rejection or copulas. In the 
first case, generate random pairs of couples (as per above), and reject 
them with a probability proportional to a bivariate Gaussian with the 
right covariance to fit observed assortive mating - continue until you 
get one couple that is probable. The computationally efficient approach 
is likely copula-based, but I do not understand them yet.


-- 
Dr Anders Sandberg
Future of Humanity Institute
Oxford Martin School
Oxford University