Lighthouse provision in Edo period Japan. Like the post-Coase literature, it reveals an interesting interplay of public and private.

Quantifying how much worse sleep deprivation makes you at writing code

The next big thing in battery technology?

A brief history of financial regulation in the US, its aims and its consequences.

Medical nihilism: dentistry edition.

Consider the maxim that everyone should visit the dentist twice a year for cleanings. We hear it so often, and from such a young age, that we’ve internalized it as truth. But this supposed commandment of oral health has no scientific grounding. Scholars have traced its origins to a few potential sources, including a toothpaste advertisement from the 1930s and an illustrated pamphlet from 1849 that follows the travails of a man with a severe toothache. Today, an increasing number of dentists acknowledge that adults with good oral hygiene need to see a dentist only once every 12 to 16 months. Many standard dental treatments—to say nothing of all the recent innovations and cosmetic extravagances—are likewise not well substantiated by research. Many have never been tested in meticulous clinical trials. And the data that are available are not always reassuring.

This norm, however, seems to vary by country, in Spain it's once a year, not twice.; similarly dental sealants are in use (from personal experience, maybe the place I usually go to is an outlier)

No plane-based rocket launch stagnation

People don't really care about privacy, in the context of, say, Google or FB having data about you, as I imagined. This is also why I am pessimistic re blockchain, one of the genuinely new things that it facilitates is private transactions, but if people don't value that...

Science nihilism: When studies are retracted, they continue to be cited. 91% of those citations are positive. One of the basic rules of research at Nintil is that papers have to be checked for these sort of things before citing them.  Academics still have room to improve!

Candidate gene studies have been known for a while to be fishy, now it becomes very obvious that at least even entire candidate gene literatures that had been meta-analysed are wrong. As originally seen in SSC.

Surprisingly, a relatively simple educational intervention shows a fairly large (d=0.83) effect size in an RCT. This is just comparing assigning math problems that are similar i blocks vs assigning alternating kinds of problems. This shows that at least there are some things that can be done to improve education. But will these techniques show up in actual education?

Self-driving cars may take longer than it seems, decades rather than 1--2 years.

Latest review paper on the minimum wage wars: Econ 101 is still right, minimum wages are bad, especially for low skilled workers. A survey of economists (in the thread) from this year shows that consensus still seems to be that effects are mostly harmful.

US state Idaho legislators, possessed by the ghost of Murray Rothbard just repealed all regulations. Now they have to decide what to keep on a case by case basis.

The benefits of electric cars are usually discussed in terms of CO2 savings, which vary with the electricity generation mix. What is often forgotten, as I point out in a reply to the thread is that EVs more importantly move away pollutants from cities to powerplants, where they can be safely addressed.

One of the best reviews of Radical Markets so far, by David Levine. On data as labour, he joins the consensus that it is the worst part of the book.

Open-sourcing civil society, essay by Vlad Tarko

Jason Collins reviews The Mind is Flat. There are some issues similar to what I found in The Elephant in the Brain: Extrapolating too much from relatively niche studies like split-brain patients, studies that don't replicate.

Critique of "The Mating Mind" by @Evolving_Moloch. I too side with the view that a sexual-selection explanation for art and music is all not that obvious.

What is Ethereum actually used for? Gambling and cryptokitties.

Good review on graphical models, clearer than Judea Pearl's own exposition in The Book of Why

Why Popper (falsificationism) is wrong, essay from Philippe Lemoine. This also seems to me the consensus view in philosophy in general. A pro-Popperian paper posted in the previous thread as a counterpoint.

Twin study on religion and crime. (Spoiler: Religion is not causally associated with less crime here).

As I've been saying on twitter for years, the rich are nice people over broad measures of "nice". A study that found otherwise fails to replicate in a large transnational sample. The study finds either a null effect or a small positive effect of income on generosity.

Also as I said earlier here, the "expanding circle" narrative of moral progresshas issues, it greatly depends on what one counts as a moral patient. If one holds a pro-life position, then the millions of worldwide abortion negate all moral progress. If one is pro-life, one has to accept that in at least one instance, there was a narrowing of the circle of moral concern.

ChinAI, a newsletter on developments in AI in China

Comments from WordPress

  • Rational Feed – deluks917 2019-05-20T19:56:27Z

    […] Links 27 by Artir – Link post. Sampler: AI progress in China, expect delays in self driving cars, people don’t care about privacy, a history of financial regulation, Etherium’s actual uses. […]