What I have been reading: December edition
Apologies for the slight delay. You will notice a decline in X activity given the time constraint, so these posts will get considerably shorter:
Against the narrative of rising markups. Worth noting too that estimates that attempt to measure intangible capital in the cost function do not show an increase.
Lenacapavir is effective as a cure for HIV. We are achieving more and more of these biotech milestones every month! Stem cell transplants also work. Lenacapavir is substantially cheaper, yet requires repeat dosing.
An alternative to transformers that preserves linear scaling?
A simple blood test can predict Alzheimer's long before any substantial cognitive impairment has begun. Another reason why we should not dismiss the amyloid hypothesis altogether. The results replicate too!
Scott Alexander weighs in on the latest of the missing heritability saga. I almost forgot that I wrote briefly on this in the previous post of this series. I'm inclined to agree with the hereditarians here: for the results to be conclusive regarding heritability, a within-family design that pits the two methods directly against each other is necessary. Like Scott, I also made the error of assuming that assortative mating inflates, rather than reduces, heritability estimates. This of course changes the conclusions of my post; the aggregate bias of twin studies are not obvious. Plus the biases are well known so can be accounted for. As Pinker notes, molecular studies are still underpowered as, even with increased measurement of rare variants, they cannot encapsulate all genomic and gene-environment interactions.
Numerous reasons why this could be the case. I suspect one major cause, as readers of my anti-ageing series will know, is that saturated fat intake (highly correlated with obesity) accelerates cognitive decline and other aspects of ageing.
If AI could vote. Amongst the main (and most advanced!) models on the frontier, there seems to be no left/right bias. There exists a liberal bias; likely explained by the fact that intelligence correlates with social liberalism.
Subsidising variance to accelerate natural selection? Hanson is one of the only thinkers who genuinely “thinks outside the box”, unafraid to question social dogma (despite facing numerous cancellation attempts precisely for that reason), which is why his blog is definitely recommended reading.
So perhaps we should not be using the small published effect sizes to downplay the costs of social media?
Joke country, and we wonder why economic growth has been nonexistent here. In fact, I think Britain is a pretty good example of what happens with governance low in elite human capital.