# Brave New World

Read Steve Hsu’s Nautilus article on Super-Intelligence. If so-called IQ-related genetic variants are truly additive then his estimates are probably correct. His postulated being could possibly understand the fine details of any topic in less than a day or shorter. Instead of taking several years to learn enough differential geometry to develop Einstein’s General Relativity (which is what it took for Einstein), a super-intelligence could perhaps do it in an afternoon or during a coffee break. Personally, I believe that nothing is free and that there will always be tradeoffs. I’m not sure what the cost of super-intelligence will be but there will likely be something. Variability in a population is always good for the population although not so great for each individual. An effective way to make a species go extinct is to remove variability. If pests had no genetic variability then it would be a simple matter to eliminate them with some toxin. Perhaps, humans will be able to innovate fast enough to buffer them against environmental changes. Maybe cognitive variability can compensate for genetic variability. I really don’t know.

## 3 thoughts on “Brave New World”

1. Rick G says:

I had always thought that the small contribution of each gene to intelligence was bad news, but if this is correct then it is potentially very good news (from the perspective of cognitive enhancement).

I wondered if this model depended on the distribution of each variant, for example whether these intelligence-associated genes (A) each had two equally common variants, or (B) had one very common plus one unusual variant. So I generated random numbers from (A) a binomial distribution with N=10,000, p=0.5, or from (B) a distribution given by the sum of two binomials, one with N=5,000, p=0.95, and the other with N=5,000, p=0.05.

For (A), as Steven says, you need to play with about sqrt(10,000)=100 genes to increase an individual by one standard deviation. One standard deviation = 50, but of 100 random genes, 50 will already be the right variant, so you only need to flip 50 others.

For (B), the mean is the same as (A), but the standard deviation is only about ~ 22. So you’ve reduced the number you need to flip by more than half. And now your ceiling is about 5000/22 ~ 225 standard deviations above the mean, instead of the 100 standard deviations in (A).

So accounting for the fact that intelligence-related genes probably do not have a maximum entropy distribution in the population actually increases an estimate of the limits of cognitive enhancement (under these generous total additivity assumptions).

Like

2. Rick G says:

It’s amazing to imagine that if the average person is a 5000 on a 10000 point scale, that the smartest people who have ever lived may not have been above 5500, with 99.7% of all people between 4850-5150.

Like

3. I am so far out of these issues I shouldn’t be commenting (but i have a big mouth which gets me in a fair amount of trouble—so often i pretend i’m to be autistic which may be partly true.). A physicist (phD caltech) who is a socialist type said Hsu scares him (me too a bit but that’s acutally universally true for me).

I dislike addi(c)tive models (the ones without the c). There seem to be others which emerge from say Turing models, or Rene Thom’s catstrophe theory. . I’m not sure thinking about cognition this way makes sense or is very useful. Also, how does one define intelligence. Steve Hsu, Steve Colbert, Dr Dre, Kenneth Rogoff? I personally prefer models more like ‘collective intelligence’, ‘PDP’, or a la superogranisms—Keller Segal model of slime molds (which seem to be an apt term for describing this world).

Also, to go out of my league, my view is any distribution can be viewed as one of maximum entropy—all you do is add constraints. This was pointed out not long ago by a mathematical biologist in online comments on an artilcle in Science on economic inequality; i also made basically the same comment a different way since i hadnt noticed his.

Superintelligence which seems popular among some people (who sometimes come off as eugenics proponents) i am not sure of—it almost seemed like a ‘scam’ used to sell pop books by people who decide physics research was too hard, low paying, and no suiperstar status usually (not everyone is a feynman). However one such proponent did make some points i agreed with (ie precautionary principle). To have an extreme profusion of means in an age with huge confusions about ends (einstein) may not be a priority—does isis need faster computers?

This may be happening already—some people live in the brave new world, others in middle ages or prehistroy. In Brazil some will be going to Olympics events and others selling themselves in favellas to tourists. Also, statistician I J Good (wikipedia) apparently had a similar idea long ago—50’s or 60’s.
. .. .

Like