Links since the last links post, plus a broader discussion of Peter Thiel’s recent public appearences.

In my last links post I said that retracted papers are like zombies that keep going, getting more citations. But it could be worse! It could be that it is cited more approvingly after it gets retracted!

What if you let Word2Vec loose on a corpus of materials science literature? Not only the learned embeddings make sense, but also may help make predictions about what materials will work in the future, thus being able to guide researchers!

Selgin on the interplay between politics and the gold standard

Regular solar panels have a theoretical maximum upper efficiency, known as the Shockley-Queisser limit (around 30%). Apparently it is possible to go a bit above that , as the limit itself relies on a few assumptions that need not be true.

Sabine Hossenfelder to physicists: If you don’t know anything sensible to do with your research funds, why should we pay you?

SEP entry on Russellian Monism which may or may not be a good way to think about consciousness relates to physical stuff.

SEP on retrocausality in QM (wtf)

Markets are efficient, climate warming edition

In science, what correlates with using unconventional methods?

Extremely complex new drug synthesised . Requires 109 steps. I remember when studying organic chemistry and production of chemicals, how Rube-Goldbergian these processes can be, just because we can’t assemble drugs atom by atom… (also, progress in assembling stuff atom by atom)

The interplay of universities, startups, and corporate labs. What is going on with the system of innovation?

If you know anyone from Vietnam, it’s quite likely that their surname is Nguyen. But why?

Free trade does lead to more growth in general

Alice Evan’s podcast with Nathan Lane on industrial policy and the evidence behind it.

Review of “Superior: The return of race science”

Paper proposes a blockchain-based solution to tackle academic misconduct

Remember the Emdrive, that propellentless propulsion system? Coverage of it died among claims that the alleged results finding that it worked due to interference with the surroundings of the device. One of the researchers involved in the original studies built the most accurate measurement device to date, found no thrust.

Google tests cold fusion, finds no effect.

Tabarrok’s study on the Baumol effect

In China, party members are more liberal. This is in line with other research showing that “elites” tend to be more liberal (in the classical sense)

DARPA work on BCIs

Modeling performance incentives in academia

Mass produced, floating, molten-salt nuclear reactors are coming

Patents: Could they be slowing down growth? (Spanish) (Key paper cited here). Also on patents and their use as metrics of innovation

Everything is correlated , and RCTs comparing two groups never have an effect of literally zero. Instead of using NHST as usual to see if an effect should be considered “real”, here’s a proposal to instead quantify the magnitude of the effect itself, if the trial were repeated a number of times.

Via Adam Marblestone: An essay he wrote on making explicit the implicit knowledge hidden away in labs. The closest implemented version seems to be the Fermat Library

Via Michael Nielsen, the Journal of Visualized Experiments. It has videos on experiments across the sciences, including shopping lists for the materials and tooling used!

Does poverty cause poor health? One mechanism that has been proposed is that poverty causes stress, stress raises cortisol levels, and that in turn causes poor health. But there is little evidence that that mechanism is at work.

Neuralink, in context.

Most (75%) new drugs are no better than existing ones, it is argued. The situation is better with cancer and infectious diseases (“Only” 40% add no benefit) and worse with psychiatry (Almost none are beneficial compared to existing drugs)

Lengthy thesis argues against NHST, and explains why it is still in use. It’s the usual critiques, but the added bit is _why those critiques haven’t made people change their minds_.

Does coffee work? There is a sub-literature out there claiming that it doesn’t, that the effects are confounded by the fact that caffeine has a withdrawal effects and that the positive effects that are observed are due to a lower baseline performance caused by caffeine withdrawal.

I don’t generally drink coffee: It tastes bad. And I remember when none of the people my age drank coffee (when I was younger), the transition happened at some point in university. Perhaps due to poor sleeping habits, people are led to think that they coffee will improve their functioning in the morning (After all, that’s what adults do!), and so they start. Then, lacking an objective baseline, come to think that without coffee they would be even more tired and so they continue with the habits.

A review of the current state of gene drives

Why is (medical) education so expensive in the US? Evidence (if true) against Baumol’s cost disease being the only explanation: There is legitimate inefficiency there.

William Eden’s blog, I just discovered it.

Why wasn’t the bicycle invented earlier?

There is huge variation in poverty rates within developing countries.

What’s going on with corporate markup increases?

Students don’t seem to care that much about earnings in the context of choosing what to study at university

After Alex Tabarrok’s work on regulation, it seems like the issues of the US economy are not due to it. But then, new papers comes through arguing that yes, that regulation is stopping the formation of new companies, benefitting larger, established players.

What are recent, and upcoming scientific discoveries that we should know more about?

US infrastructure over time. Something happened in 1970.

Is the galaxy brain real An attempt at explaining why the laws of physics are what they are, involving the universe being conscious.

Podcast with anonymous finance twitterer Jesse Livermore.

How should we think about the value that data has, in the context of large tech platforms? What facebook knows about me is useless in isolation; in context, after pooling with more data, it can be used to drive very specific marketing campaigns.

Pushing back against space colonization.

Once again, against corporate irrational short-termism: leave the share buybacks alone.

Economics as a highly incestuous field.

Women tend to get more autoimmune diseases, I recently learned. A theory about why that is the case.

Is the Belt and Road Initiative a big mistake for China?

More often than not, when I make my point that physics is over, I get a physicist in the audience saying “But topological matter!”. So I guess we should read more about topological matter.

Peter Thiel

It’s Thiel week! Thiel’s speech at the National Conservatism conference (Thanks Bonnie Kavoussi for the notes) , and Thiel’s podcast with Eric Weinstein.

  • The podcast is way, way better, Weinstein did a great job. Even for those who have been following Thiel and Weinstein, there are a few new ideas in there.
  • On the NC speech, I have more mixed thoughts. If one takes his “Why is Chinese money flowing out of China” point, one can think of obvious explanations: China has a large savings rate and a large population (Remember: GDP growth and stock market performance, with no adjustments, don’t have much of a correlation). I didn’t have to search much for proposed explanations. In the speech, Thiel hints at this being bad somehow, but doesn’t go deeper on whether it’s a normal thing, and why it would be bad.
  • He has some good points: Indeed, it is naive to assume that free trade will necessarily be beneficial to a given country, especially in the short term. If one values citizen welfare more than foreigner’s welfare (A common moral view I disagree with) and if American workers are made worse off -again, in the short term- due to the China shock, if I were an American worker it would be in my self-interest to ask for protectionism. Even if you knew than in aggregate, in 20 years time, global GDP will be higher, the choice you are making might be a world where you are employed and your chinese counterpart makes less money or a world where you are unemployed and your chinese counterpart makes way more money, understandably voters, if faced with that tradeoff would vote for protection.
  • But is this for real? Is Chinese competition hurting US workers? That seems to remain a contended question. I don’t know the exact evidence that’s behind Peter Thiel’s claim, but it would be helpful if he blogged them for everyone to assess.
  • He is right on education, one can point to Caplan’s Case Against Education for the empirical backing.
  • On political correctness, it’s hard to pin down the underlying thinking with full clarity. The claim is twofold: One that there is more political correctness now, and two that that has negative effects throughout the economy, affecting even innovation.
    • The usual interpretation of political correctness these day is around topics of gender and race, and indeed the examples that Thiel gives point in that direction. That kind of political correctness seems to be on the rise. But does that lead to a decrease in innovation? I can see it affecting the social sciences, but not the STEM fields that Thiel and I think are the key to the innovations we all think about when we think of innovations.
    • For example, there are some diseases that affect more some races than others. If one does not acknowledge that race plays a role in medicine, then, to put it in hyperbolic terms, political correctness kills people. But does it though? What’s the effect, on the margin?
    • Then we can think of PC as not questioning consensus truths, and having sacred cows. This I don’t know if it’s on the rise, but it is indeed an issue.
      • Consider Hanson’s call for cutting healthcare
      • Ditto for Caplan and education
      • And Hossenfelder and not randomly throwing money at high particle physics
    • These three proposals, and even milder versions thereof are being stopped by what seems to be a wall of hurt feelings. But if successful, and if they are right, it would greatly increase productivity in those areas (and society overall, given the large amounts of money poured into healthcare and education), freeing up resources to be used in other areas.
    • Now, this is definitely worth investigating, and maybe this is what Thiel means all along, but it’d be nice to hear in a clearer, blog-like way!
  • On technological innovation, the podcast is more nuanced (e.g. they grant that maybe differential progress is just due to the relative difficulty of each field of inquiry);
    • On the specific atoms vs bits distinction, on the one hand it is true that making things in software is easier than in hardware. Software has almost zero barriers to entry and no fixed costs, you can start small and scale. Starting an airplane company from scratch is expensive. Also, hardware is constrained by the laws of physics, software is not, insofar as computation is concerned, or at least to a lesser degree. e.g. compare doing mergesort vs doing mergesort in real life, physically rearranging objects, how fast can that be done? Or, in a videogame one can have things that violate physics, like magic. Of course, those things are simulated and not real, software deals with relations of abstractions, not with real atoms. In sum, software is less constrained, thus having more room to improve.
    • But it doesn’t seem like it is software that is going faster: It has been argued that software is getting more bloated and slower because we now have more computing power, enabling us to be lazier. Sure, there is a plethora of new apps, but then we can also point at progress in the world of bits: semiconductors (This is atoms not bits!), photovoltaics, gene sequencing, assembly line manufacturing of cars or ships. Even the poster child for economic slowdown in the Founder Fund’s manifesto, airplanes -whose speed, has indeed, stagnated-, has on the other hand experienced a 50% cost reduction since the 80s! And this is not merely by flying the same old jets in with an underpayed crew, there has been legitimate improvements in fuel efficiency enabling those cost reductions.
    • Tl;dr Bits vs Atoms does make sense as a key distinction to explain the differential behaviour of software and hardware, but it does not seem to me to be the key thing to understand progress. We also have to look at other things.