All of danohu's Comments + Replies

Cases in Germany have also been spiking the last 2 weeks. I don't see any explanation other than Delta being very, very infectious. Case numbers were halving week-on-week in mid-June (minimal Delta); now (most cases Delta) they are almost doubling each week. You can throw in a few other factors (control system lag, school vacations, a big football tournament), but Delta still must be doing something extreme.

One ray of light is that the increase is largely limited to people in their 20s, who are mostly not yet fully vaccinated. The vaccination rate in that age group should shoot up in the next month, so Delta might get damped more than you would expect from the overall vaccination rate in the population.

The writing offer sounds like exactly what i need. I'm in

[I'll write you directly; also posting here as evidence that The System Works]

What do we know about the nature of infection of vaccinated people, or re-infection of people who have recovered from Covid? It seems to me, this will have a big impact on Covid epidemiology in a mostly-vaccinated country. 

I have three mental models:

  1. Immunocompromise: some people will never build up Covid resistance, no matter how much they get vaccinated or infected
  2. Exposure: a person has resistance, but is exposed to Covid in a way that overwhelms them. High viral load, maybe, or their immune system is somehow having a bad day
  3. Vaccination failure: for
... (read more)

The legal situation around immigration makes the UK much less viable, IMO.

Of course, any location will exclude the majority of the world who aren't allowed to live there. But the US has much of the existing community, and other options (EU, Canada) benefit from either a big population of a liberal migration policy.

Are we living in the timeline where the Bayesian Conspiracy becomes an actual underground movement?

There's another big reason for the jump in vaccinations in Germany (and the rest of the EU): We've just started a new quarter.
Yes, really. 

The discussion on manufacturers vaccine production schedules have largely been on numbers per quarter. Given that the EU is scapegoating them for delays, the manufacturers really really want to avoid missing their targets. And that leads to weirdness at the boundaries of the quarters.

Here's the data on deliveries in Germany

At Biontech, production is going well, and they are comfortably meeting their Q1 commitment (... (read more)

Dogs have supposedly evolved more control over their eye muscles, in comparison to wolves. The suggestion is that this is part of their symbiotic relationship with humans, since we are more likely to look after animals that look cute.

It is interesting to think that dogs may have been selected for hundreds of generation for their ability to influence the emotions of humans.

Absolutely this. Thank you mingyuan!

Also on the retrospective conversation. "Truth and reconciliation" feels like a useful framing, if a bit dramatic (it's more commonly used in the aftermath of war or genocide).

i.e. we want to understand what we did, and how we could have acted better. But we also need to work out how to (re-)build relations with each other and the wider world, when lots of people have behaved selfishly, or suboptimally, or harmfully.

Though becoming less uncomfortable. Regulation/manufacture is belatedly catching up to the idea that you can take a swab from the front of the nose, which is MUCH more bearable.

Much of architecture is a trade-off between price and interestingness. And on that frontier, we largely prioritise cheap over interesting.

I do wonder, though, if we are stuck in a local maximum around rectangular floor-plans. If you start from first principles, you easily find cost- or space-efficient options with irregular room shapes. Think of geodesic domes, or this school layout.

But our world assumes straight walls meeting at right angles. Architects I've talked to say they wouldn't consider anything else, because it would be so risky and expensive to ... (read more)

I think architects are correct to be skeptical of their own ability to do stuff other than right angles. MIT's Stata Center is famously interesting, and also is full of leaks and mold because it doesn't do the basic building thing of keeping the rain out. []
That school layout looks to me like it has a lot of rooms that don't have windows which most people would see as undesireable for a school.

‘people’ doesn’t mean all people, and that it is tractable and common to change who falls into this category or who in it is salient and taken to represent ‘people’.


The default for this, of course, is 'people I spend time with'. Which is why it makes sense for parents to worry so much about their kids hanging out with the wrong crowd.

IMO it is very hard to spend time with people without coming to care what they think. That's why spies go rogue, diplomats go native, and earn-to-give EAs are at high risk of starting to think like their non-altruistic colleagues.

Here are a few of the insights I got from the book, but I'm pretty sure I wouldn't have gotten from visiting Asia frequently:


I'm surprised by this, because those seem like the kind of insights I would expect to get by traveling.

I would, though, expect to make mistakes on what to generalize -- e.g. not knowing whether my hosts sleep with their baby as an individual quirk or a cultural norm. 

The problem is that metaculus points reward some non-obvious combination of making good predictions and being active on the platform. I only care about the first of those, so the current points system doesn't help me much. 

I can't look at a user's points score and figure out how much I should trust their predictions. Or possibly I could, but only by diving into the small print of how scoring works.

I say that as somebody who uses metaculus and believes it has potential. The points system is definitely a weak point

There's no single metric or score that is going to capture everything. Metaculus points as the central platform metric were devised to —as danohu says — reward both participation and accuracy. Both are quite important. It's easy to get a terrific Brier score by cherry-picking questions. (Pick 100 questions that you think have 1% or 99% probability. You'll get a few wrong but your mean Brier score will be ~(few)*0.01. Log score is less susceptible to this). You can also get a fair number of points for just predicting the community prediction — but yo... (read more)

You could also check their track record. It has a calibration curve and much more.

You don't care, but if the goal is to motivate better communal predictions, giving people the incentive to do more predicting seems to make far more sense than having it normed to sum to zero, which would mean that in expectation you only gain points when you outperform the community.

As Zian says: large/liquid/free prediction markets aren't sufficient, they also need to be trusted by enough people, or by powerful enough people. IMO there will be at least a 7 year lag between prediction markets working well, and them being broadly accepted outside of narrow wonk/nerd circles.

In the existing world, covid was already front-page news in Jan/Feb 2020, with speculation about it spreading beyond China*. Few people in the West did anything. The limiting factor wasn't warnings being issued, it was people being able to grok that something Really... (read more)

The stock market is essentially a prediction market so a seperate prediction market on the question whether COVID-19 reaches us wouldn't necessarily produced more information.
I have to question the premises and the strength of the conclusion here. This comparison feels off--I remember things being much more controversial and confusing than "experts expect covid to reach us". But to answer the second part, I guess I would have to ask myself how many public health officials, supply chain influencers, lawmakers, etc. were in the competent minority (and how money-motivated they were). But certainty is never necessary for action. I intuitively imagine a major difference between the sort of uncertainty I felt in Feb 2020 (due to model uncertainty, low trust, competing heuristics, Twitter fog) vs. the uncertainty I would have if prediction markets had said "X% chance of pandemic spread within 3 months". As a human, these different types of uncertainty affect my behavior differently. I was able to replace some of the former uncertainty with the latter uncertainty by watching metaculus, but being able to freely hedge my bets would have been even better. Right, a 14-day shift can make a lot of difference in the early stages of exponential growth. (Or did you mean to say "not enough"?) I think there may be a decent case for prediction markets having only minor overall impact during the notional 7-year lag, but not a strong case. And even if prediction markets wouldn't have prevented any of the largest would be nice to get money for being correct. Or to be able to easily hedge between monetary risk (losing bets), health risks (virus), and professional risks (refusing to go to work in person). ...And to be honest it would have been real nice to ask Vox writers to put their money where their mouth [] was.

Peptide vaccines aren't old and boring, they are new and unproven. From a 2014 review 

The vast majority of candidate peptide vaccines are under Phase I (270 studies) and Phase II (224 studies) stage of development.In a total of 452 studies, only 12 studies have progressed to Phase III level of development.Interestingly, all these 12 studies are on therapeutic candidate peptide vaccines indicated for treatment of multiple types of cancers.

AFAICT no peptide vaccine has ever been approved for human use.

So yes, maybe a peptide vaccine could have worked be... (read more)

I don't know about subtle difference between proteins and peptides, but I would say the relevant category is "recombinant vaccines" and I believe that the first such was the Hepatitis B [] vaccine approved in 1986. This used genetically engineered yeast to produce a protein from the virus that was harvested and injected into people.

They are currently being keelhauled by the EU for having allegedly* diverted vaccine from the EU (which had not approved it) to the UK (which had).

They are committed to not making a profit from the vaccine. So there is no financial incentive for getting more vaccine out into the world, and plenty of reputational risk for doing so.

(*) AFAICT they did not even do this, and still their reputation is being trashed. 

"I hope that they’re giving Europe all the doses we’ve refused to approve here in America"


oh Zvi, you slipped into expecting reasonable behavior.

The EU don't want to inject any vaccine unless it has been made in the EU. At least the published contracts both contain clauses requiring manufacture in the EU/UK. Because clearly that's what we should be worried about right now.

I had wondered why there were separate production chains for the US and the EU. 'Protectionism' is apparently the answer.

I agree! Access to cheap informal testing has made a HUGE difference in my social circles, and I don't understand why it isn't being used all around me. Life is so much easier when you can work around different risk tolerances by giving tests rather than breaking off contact.

Are the antigen tests that much harder to get hold of in other countries?

With antigen tests, I believe the figures are  "P(positive test | positive gold standard test)"

Look at the documentation for another antigen test -- section 14 gives sensitivity compared to PCR. The figure they promote (and is e.g. used for the German government approval) is 97.56% sensitivity. This is both based on PCR, and based on filtering down to specimens with a high viral load (Ct 20-30).

BTW, that German site is a good collection of documentation links for all the antigen tests currently approved.

Ah, gotcha, that makes more sense. And thanks for the awesome antigen test table you linked there!

Depends a lot on where you are. In California, sure. But in the winter in Scandinavia or Canada?

However, there is a high degree of skepticism, since a lot of this evidence is self reported. Some of these reports come from people who believed they had SARS-CoV-2, but never received a clinical confirmation through testing.


That seems like a good opportunity to separate out the effects of covid itself. Collect symptoms from a bunch of these self-reporters, then use antibody tests to determine which actually had corona.

Thanks for the link; added to my reading queue

Are you familiar with the idea of Speech Acts?

Some of the ideas there seem very relevant to your Actor/Scribe distinction. [I only have a very superficial knowledge of this stuff myself, so can't say more than that it might be worth a look]