Comments on Kirkegaard's "Smart Fraction Update"
https://astralcodexten.substack.com/p/open-thread-258/comment/11748815
So, three comments,
First, the paper (1) is much better than the blog post and absolute credit to the men, there is a very clear rmd file (2) with all his code and it's even fairly readable. I'm reasonably confident I could replicate these findings in a morning and make simple iterations with new data or models that afternoon. Absolute bravo!
Second, put me down as this being a statistical artifact for two reasons.
-He's feeding the results of one model into another one and that's always much easier to screw up. Specifically, he's not technically using the IQ of the top 5%, he's running a linear regression on what the top 5% "should" be and then feeding that gap, positive or negative, back into the model. (3) I don't want to say this is super bad but it's....much easier to make mistakes when you plug one complex process into another complex process.
-Worse, the residuals he's generating don't make sense, as he himself notes. Some, like South Africa, match our intuitions but Bulgaria and Ghana don't seem like they have a...unusually intelligent elite. Nor can I imagine a reason why Japan, which has virtually no immigration, should have unusually intelligent elites relative to their general population.
So yeah, my gut says he used a technique likely to generate untrustworthy results, the results don't make much sense, but he plugged them in and they worked. Put my money on weird statistical artifact/bug.
Finally, I am somewhat skeptical of "just elites" theories of IQ enhancement. Partly because I think it focuses too much on the benefits of IQ in terms of GDP. One of the things I appreciated about this paper is that it used the Social Progress Index (4), which looks at a wide range of social and economic factors, not just GDP. And a lot of this factors, like "Money stolen" or "freedom of discussion" are much more influenced by the personalities and abilities of common people than a few elites. Frankly, a lot of the happiness in your daily life depends on ordinary interactions with the people around you and there's no substitute for that.
But, and perhaps more importantly, I don't think this properly models the challenges or opportunities of IQ enhancement technology. Specifically, the challenge is credibly proving that it works. If it works, as I guess a "right rationalist", I'm confident that as soon as a bunch of Asian and White nerds start doing this, our enemies will immediately demand that it be provided to everyone, free of charge, by some large and ugly government bureaucracy. Which is awesome, everyone should have this technology, and it's probably inevitable that some large bureaucracy will be in charge of it. Our enemies are our enemies because they, to be overly charitable, they deprioritize the truly important things in life like, family, freedom, and the pursuit of truth to the point where these things are fading away. They are not our enemies because they have bad health care policies or lack opportunism. Frankly, I think most proponents envisage a world where IQ enhancement is unpopular even after it's proven to work, whereas I see a world where those with the ability will hypocritically flip on a dime and try to take it over and build a powerbase out of it, which, again, is super awesome because that's the best way to distribute this to as many people as possible, which is the point.
(3) Specifically this line of the code: "d$IC_reg = lm(x95pct_IQc ~ SAS_IQc, data = d, na.action = na.exclude) %>% resid() %>% standardize()"
(4) https://www.socialprogress.org/global-index-2022-data-definitions#2/0/0