Interpreting Data, Sympathizing with AI, Nietzschean Maoism, etc
Links and Best of Twitter, 3/24/23
Meeting slots are still available for half hour meetings. You can book here. I’ve just opened up the week of April 10th.
Here are my recent conversations with Michael Tracey.
Earlier this month I flew out to Columbus, Ohio, to talk to presidential candidate Vivek Ramaswamy about civil rights law for his podcast, not yet released.
As I’ve said before, I’m pretty confident there will be movement on this issue the next time a Republican is in office. I liked the photo here so much I cropped it and made it my profile pic. I now realize how important good lighting is, my previous photos made me look a lot more colorless than I actually am. Always learning.
1. AI doomerism relies on the idea that intelligent systems behave in ways that are coherent. But this argues they’re not, by a clever method of ranking institutions and people by intelligence and coherence, and finding a negative relationship.
2. Razib says don’t romanticize the Vikings. I can’t help it though.
3. On the decline of the Colorado Republicans. Their new chairman tried to legally add “Let’s Go Brandon” to his name.
4. I wholeheartedly endorse Emily Oster’s article on how to interpret data. It echoes what I’ve learned before, which is that not only is correlation not causation, but “correlation that you see even after adding controls still isn’t causation.” The example of breastfeeding:
If you understand this, you’ll realize why the vast majority of social science is bunk. I didn’t learn anything new here as it reflected my own thinking, but if you don’t have well developed thoughts on this topic, I highly recommend reading her piece.
5. Brown spiders will attack and kill black widow spiders that are close by, but are likely to leave other spiders alone. (Popular Science via Apple News)
6. Rebellion from within the Taliban on the ban on girls schooling. Trads appear to be the ones out of touch with human nature.
7. Remember that paper on how having daughters makes members of Congress more liberal on women’s issues? The theory is only supported during a brief window of time. Seems like a classic file-drawer effect, which can afflict natural experiments just as much as it does other kinds of research.
8. My belief in the likelihood of DeSantis being the nominee has gone down in the last few months. Trump has an energy and spirit, and a hold on the base, that just can’t be matched. NYT finds that he increased his lead over the field between January and March. I think the DeSantis position on Ukraine was a huge mistake, as he needs to consolidate the anti-Trump wing of the party, and make sure they don’t vote for Haley or Pence or anyone else. The populists don’t realize that the neocons and warhawks are actually a sizable faction of the base! And Trump’s position on Russia is a large reason that faction has a problem with him. They’ll be much less enthusiastic about DeSantis if he’s Trump-lite on foreign policy.
More fundamentally, it shows weakness by chasing the base on each issue. When you’re dating a girl, sometimes you need to stand your ground on some issue or point of dispute. If you follow her on everything she’ll soon have contempt for you. It’s like this with politicians too. Trump in 2016 almost went out of his way to antagonize the base. This is an extreme form of peacocking, and a strategy that only very rare politicians can employ. More normal politicians, like DeSantis, can’t go as far as Trump did, but they also can’t be pushovers for the base either. Nothing in DeSantis’ past indicates that he should right now take a dovish position on Ukraine. It doesn’t matter if much of the base agrees with him or not. Trump let the base come to him on Russia issues, in part because of his justified disdain for them, which only makes them love him more.
After the 2022 election, my estimate for Trump winning the nomination dropped to about 60%, I would say it’s somewhere around 75% now.
9. Vice video report from inside Andrew Tate’s world shortly before he was arrested. Highly recommended.