Does NYT have policies? 2020-06-24T04:06:57.556Z · score: 17 (3 votes)
Douglas_Knight's Shortform 2020-03-28T17:02:42.450Z · score: 8 (1 votes)
Perception of the Concrete vs Statistical: Corruption 2016-03-23T01:19:33.856Z · score: 5 (4 votes)
Would you notice if science died? 2016-03-08T04:04:49.587Z · score: 5 (10 votes)
Actually existing prediction markets? 2015-09-02T22:24:45.470Z · score: 11 (10 votes)
The Cold War divided Science 2014-04-05T23:10:38.181Z · score: 21 (22 votes)
Games People Play 2010-11-20T04:41:39.635Z · score: 11 (10 votes)


Comment by douglas_knight on SlateStarCodex deleted because NYT wants to dox Scott · 2020-06-29T18:32:38.213Z · score: 4 (3 votes) · LW · GW

I've seen a lot of complaints about Metz's history, but they all seem backwards to me. They seem like a satire of virtue ethics.

Who do you think he's "working for"? If he is working for outside forces (eg, keeping a source happy), then drawing attention to it is exactly the best way to take it out of his hands and force him to work for his editor; and force his editor to work for the paper.

Writing puff pieces sounds more lazy than malicious to me.

Comment by douglas_knight on Douglas_Knight's Shortform · 2020-06-27T02:09:00.723Z · score: 3 (2 votes) · LW · GW

Are there compendiums or classifications of trolley problems?

What is the most extreme real-world trolley problem? By "real-world" I mean something that really happens, emphasis on the plural. I don't want one-off examples where one person has the moral luck of having to face it and everyone else can breathe easy that they didn't have to think about it. I want examples where there is a definite, known policy. By "extreme," I mean something that really pushes people's buttons. By a classification, I mean a classification of which features make it more like a visceral trolley problem and which more like a blurry statistical haze that allows trading lives.

I propose a candidate: the dengue vaccine. In any event, I think people will find it interesting.

Dengue fever is an often-fatal mosquito-born tropical viral disease. People develop immunity, so we could make a vaccine. Obvious candidate, except ... Since we are all now experts in antibodies, we all know about the crazy phenomenon of antibody-dependent enhancement, mainly observed in dengue. It is not one virus, but four closely related strains with different envelope proteins and different immunity. If you get one, it's a non-lethal disease and you become immune to that strain. But you're still vulnerable to the other strains and, for not entirely clear reasons, infection with a new strain is much worse.

If you've already had some variant of dengue, any vaccine is better than none. But if you've never been exposed, it might be worse than not vaccinating. So of course the vaccine is a combination of all four variants. What if each of the four vaccines had a 95% chance of working, independent? Then someone receiving the vaccine would have about a 20% chance of not being vaccinated for all four. Let's say that's worse than nothing. Vaccinating everyone is a trolley problem benefiting people who have been exposed at the expense of those who have not been exposed. Both the benefit and harm is statistical (you don't know that you'll ever get dengue in the future), but the two groups of people can be identified ahead of time, not in a God's eye view of who will be bitten, but in a really potentially testable way. You could just test people for antibodies. If you're first-world-rich, perhaps a tourist from the first world, you can get repeated testing for antibodies and if you ever test positive, then you should get the vaccine. But the testing is more expensive than the vaccine (and logistically complicated) and Filipinos are poor, so we're not going to pay to test them. Should we choose some simple criterion like an age threshold and living in a badly hit area and just vaccinate everyone?

This was a hypothetical and I'm not sure if people were ever faced with this decision. If so, they decided not to pull the switch and instead kept working on the vaccine until it was much better than 95% effective. It was so effective (at least as measured by producing antibodies) enough that they rounded it off to 100% declared the problem solved and vaccinated a bunch of Filipinos who were old enough that they'd probably had it once.

And then the data trickled in and it saved lots of (net) lives, but it wasn't quite as good as hoped. People who had been vaccinated still got dengue, just not as often. But surely that meant that people who hadn't been exposed before were promoting mild to severe dengue? This seems pretty obvious, but they put their fingers in their ears and waited for the data to pin that down. That waiting, or maybe something else, burned their credibility and now the WHO policy is that you shouldn't give anyone the vaccine without an antibody test. Practically speaking, that means no vaccines.

This is a trolley problem that happened in the real world and the fact that the groups of people are potentially knowable seems to really important to reluctance to switching tracks. But the rejection of the vaccine is not purely the result of the trolley problem, but also about burnt credibility.

Comment by douglas_knight on Douglas_Knight's Shortform · 2020-06-26T04:48:48.977Z · score: 7 (4 votes) · LW · GW

A common pedagogical example of the perils of correlation analysis that ice cream consumption is correlated with homicide. The common cause is seasonal variation. This is usually presented as an absurd example, a mistake no one would make, but there is an extremely similar example that was nationally prominent. Polio was blamed on ice cream consumption because they had the same seasonal pattern. I wonder if the standard example was engineered from the real example. Perhaps it is better (eg, more absurd), but one doesn't have to choose just one example; surely it is better to also include the historical example.

Comment by douglas_knight on SlateStarCodex deleted because NYT wants to dox Scott · 2020-06-25T16:30:06.129Z · score: 2 (1 votes) · LW · GW

What do you mean by "mistake theory" and "conflict theory"?

I'm really confused by this comment and I think you are using the terms backwards. Telling someone that they've made a mistake is a violent act, a form of conflict, but it is an example of mistake theory.

Some people theorize that there is an irreducible conflict. They generally recommend that their side not talk to NYT. Until the doxing came up, they were the dominant voices on the topic of this article in preparation, or at least the ones causing discussion. But after the topic moved on to doxing, they have nothing more to say and have been overwhelmed by

This LW thread is almost entirely about mistake theory. Maybe you see different things on twitter, but if so, you should say that, because the one thing all your readers have in common is that they're on LW.

Comment by douglas_knight on The point of a memory palace · 2020-06-23T17:30:36.601Z · score: 2 (1 votes) · LW · GW

Is this one of those exercises in which you write out your argument and then reverse the valence of every claim in order to see if there was an argument? That is, was this originally a list of the form: "Memory palaces are a bad idea because they produce memorization at the expense of ___"?

Does anyone know what exercise I'm talking about? I think it was in the Sequences.

Comment by douglas_knight on SlateStarCodex deleted because NYT wants to dox Scott · 2020-06-23T17:21:15.758Z · score: 8 (6 votes) · LW · GW

I don't think that their principal goal is to doxx him. But there is a big difference between a habit and a rule. It's not that they used the name without thinking about it, but they specifically rejected his complaint and said that they were just following orders.

On many other places I see people discussing this, they point out that the reporter's claim that there is an NYT policy is a bald-faced lie. You are the first person I have seen that took it at face value. This LW discussion is striking because no one else acknowledges the claim at all. I think that they believe that it is a lie, but don't want to rudely point that out, so they pretend it was not uttered.

Added, next day: I estimate that 99% of the time that NYT writes about someone with a professional pseudonym, they treat it as a real name. 1% of the time, they note that it is a pseudonym and 1/10 of those times, 1/1000 of all times they print the real name.

Seriously, 99% of the time. I am not being hyperbolic. The main source of uncertainty is how often they write about someone with a professional pseudonym. I estimate that NYT writes about someone with a professional surname every day.

Comment by douglas_knight on SlateStarCodex deleted because NYT wants to dox Scott · 2020-06-23T15:48:19.571Z · score: 3 (2 votes) · LW · GW

If you want to clarify, edit your original post.

Comment by douglas_knight on Estimating COVID-19 Mortality Rates · 2020-06-10T21:37:31.371Z · score: 2 (1 votes) · LW · GW
except trusting that he isn't picking and choosing his arguments

Well, don't do that. I told you this before.

What's his confidence interval?

What's CBG's confidence interval? When he says 0.5-1%, does he mean something? Does he mean a confidence interval, or a distribution of "normal" situations or a distribution of more general situations? Or does he not mean anything?

Later on in that thread CBG also acknowledges it may be higher in than 1% in some places and conditions.

It's nice that he says that, but that's exactly the situation that you cited him in the other thread, claiming <=1%. I'm guessing that the pseudo-detail is exactly what caused you to not understand his claims. If you don't know what he claims, how can you assess his work? At least with GC you're not fooling yourself about what you've done.

And I still don't know what he claims. He seems to claim that NYC had IFR <=1%. Was NYC normal or not? In any event he's wrong. If NYC defines the upper range, then this affects his conclusion. If NYC doesn't count, I dunno, but I'm pretty sure that people are equivocating on whether it counts.

Comment by douglas_knight on Covid-19: My Current Model · 2020-06-10T04:07:51.949Z · score: 5 (3 votes) · LW · GW
The Default Infection Fatality Rate (IFR) Is 0.5%-1%

Why do you believe that? We can only measure IFR in the worst outbreaks, such as NYC and Lombardy, where it was 1-2%. Maybe hospitals that aren't overrun have half the morality rate, but how do you know?

Amer­ica in gen­eral could be as high as 1.2% IFR with­out mak­ing the data stop mak­ing sense.

What about the data wouldn't make sense if the IFR were 2% in America ex NYC? Outside of a massive outbreak, we can measure neither deaths nor infections. Sure, if you assume that only 33% of deaths are missed, then we can measure deaths. But why assume that? Isn't that only true in NYC because there was pressure to record untested pneumonia deaths? Elsewhere there is less attention and less pressure.

Comment by douglas_knight on Estimating COVID-19 Mortality Rates · 2020-06-10T03:58:22.987Z · score: 2 (1 votes) · LW · GW

You say that like detail is a pure good. "Greg Cochran says 1.2%" is better than any number of words from CBG. Anyhow, you repudiated this. When I pushed you on it, you came up with the number 1.4%.

Comment by douglas_knight on Estimating COVID-19 Mortality Rates · 2020-06-10T03:58:07.348Z · score: 2 (1 votes) · LW · GW
start with seroprevalence data

Because of false positives, seroprevalence is massively overestimated everywhere that there hasn't been a massive outbreak. In those places the IFR is 1-2%. But can we extrapolate to normal outbreaks? If, as widely believed, an overrun medical system has worse mortality, then maybe the normal IFR really is only 0.5-1%. But if your meta-analysis directly measures that, it is not well-done.

Comment by douglas_knight on How to evaluate (50%) predictions · 2020-05-04T18:54:54.994Z · score: 2 (1 votes) · LW · GW

Yes, exactly: this post conflates accuracy and calibration. Thus it is a poor antidote to people who make that mistake.

Comment by douglas_knight on 2020 predictions · 2020-05-04T15:49:37.216Z · score: 2 (1 votes) · LW · GW

It is striking how errors in discussions of this topic are systematically in the direction of downplaying the severity. Probably 95% of errors.

assuming a runaway infection we'd have R=3 so ~220M infected

This is a math error. Herd immunity is achieved once 1-1/R is infected. The goal of "flattening the curve," is to just barely reach this number. But in a "runaway" scenario, it is much higher. The epidemic final size of the SIR model is 94%.

Since Lombardy had a population fatality rate of 0.2%, I'm not going to look at your citations. I assume the problem is that they ignore most of the deaths.

Comment by douglas_knight on How to evaluate (50%) predictions · 2020-05-04T02:43:47.741Z · score: 2 (1 votes) · LW · GW

Well, that's something, but I don't see how it's relevant to this thread.

Comment by douglas_knight on Douglas_Knight's Shortform · 2020-05-04T02:36:51.260Z · score: 6 (3 votes) · LW · GW

Nothing in academic biology makes sense except in the light of feudalism.

Comment by douglas_knight on 2020 predictions · 2020-05-04T02:23:49.697Z · score: 5 (3 votes) · LW · GW
Given Coronavirus IFR <1% then with a US population of 330 million this seems almost certain. I would have put this probability higher if there was a higher option.

If a lot of people get infected, the hospital systems will collapse and the IFR will be higher than 1%, as it was in Wuhan, Lombardy, and NYC. If the whole population gets infected, it will be much higher. Also, the IFC is probably >1% even without collapse.

Comment by douglas_knight on How to evaluate (50%) predictions · 2020-05-04T02:17:12.837Z · score: 2 (1 votes) · LW · GW

So you probably won't convince me that these people know what the claim is, but you haven't even attempted to convince me that you know what the claim is. Do you see that I asked multiple questions?

Comment by douglas_knight on How to evaluate (50%) predictions · 2020-05-03T17:46:53.657Z · score: 2 (1 votes) · LW · GW

Could you give an example?

Could you give an example where the claim is that 50% predictions are less meaningful than 10% predictions?

How do you know that it is about accuracy?

Comment by douglas_knight on How four guys helped redirect Japan's coronavirus policy · 2020-04-24T16:49:22.470Z · score: 2 (1 votes) · LW · GW

Note that you have rewritten and cherry-picked his predictions.

His precise predictions were all wrong. Maybe Japan was undercounting cases by 5x, but so was everyone else. Cases were rising at 8% per day and deaths were rising at 8% per day. Cases and deaths have both continued to rise at 8% day. For factual purposes, the best prediction was to trust the data and simply extrapolate.

The consensus that Japan was OK was wrong, but it was directly contradicted by the official data. Exponential growth is bad. 8% per day is unacceptable. But would it be possible to simply point that out? I don't know. Maybe the only way to get attention was to claim that the data was wrong.

Here is another story of people refusing to acknowledge exponentials.

Comment by douglas_knight on How likely is the COVID-19 apocalyptic scenario? · 2020-04-23T16:42:13.685Z · score: 2 (1 votes) · LW · GW
Covid seems to leave permanent lung damage even in young people.

It is way too soon to say that. Maybe reason by analogy with SARS. It is widely claimed that SARS caused permanent lung damage. But "permanent" seems to mean 3 or 6 months! Here is a paper showing substantial improvement from 6 months to 9 months.

Comment by douglas_knight on Coronavirus: Justified Key Insights Thread · 2020-04-15T03:29:21.049Z · score: 1 (3 votes) · LW · GW

Here are some quotes from the paper. What is the simplest model you can make from them? Forget the word "model"; what conclusions can you draw?

Most of the infections on board the Diamond Princess cruise ship appear to have occurred before or around the start of the 2-week quarantine
the delay, D, between infection and onset of symptomatic infection (i.e. the incubation period) ... at 6.4 ... days
Comment by douglas_knight on Coronavirus: Justified Key Insights Thread · 2020-04-15T03:15:27.396Z · score: 2 (1 votes) · LW · GW

I think Michael's position doesn't have much to do with covid, but a lot to do with 6b. The crisis is a coordination point for bursting a bubble. 6b says that there is very little new information, but the stock market is acknowledging a large backlog of negative news. The whole market was a bubble, but Boeing was the worst of the bubble (plus covid is news specifically about transport). In some sense this is a very strong rejection of EMH, but, "the market can stay irrational," "noise traders," etc.

Comment by douglas_knight on Coronavirus: Justified Key Insights Thread · 2020-04-14T19:49:39.980Z · score: 2 (1 votes) · LW · GW

First of all, it is very important to distinguish data from inferences.

Second, the inference is idiotic. It's probably a calculation error, but it's just not worth reading to determine what went wrong.

Comment by douglas_knight on Coronavirus: Justified Key Insights Thread · 2020-04-14T19:44:46.090Z · score: 3 (2 votes) · LW · GW

So, yes, it is simply misquoting the source that I cited.

Comment by douglas_knight on Coronavirus: Justified Key Insights Thread · 2020-04-14T18:50:29.135Z · score: 4 (3 votes) · LW · GW

Where do you get your data on the Diamond Princess? As far as I know, there are no updates on symptoms. Perhaps you get it from this, which is not data, but an inference?

Comment by douglas_knight on Coronavirus: Justified Key Insights Thread · 2020-04-14T16:38:25.044Z · score: 3 (2 votes) · LW · GW

The Diamond Princess was 50% asymptomatic after 2 weeks of [in cabin] quarantine. It would be nice to have more follow-up on it, but that's already much better than most measures of asymptomatic cases.

Comment by douglas_knight on How to evaluate (50%) predictions · 2020-04-11T16:57:04.815Z · score: 6 (3 votes) · LW · GW

The title and first sentence are about calibration. You never hear very smart people saying that 50% predictions are meaningless in the context of accuracy.

There's nothing magical about 50%. The closer the predictions are to 50%, the harder it is to judge calibration.

Comment by douglas_knight on Coronavirus: the four levels of social distancing, and when and why we might transition between them · 2020-03-30T04:59:04.804Z · score: 3 (2 votes) · LW · GW

There is another measure, which you might think of as level 5, but you should probably think of as an independent axis that could be applied at earlier levels, which is quarantining infected people. That is, taking them out of their home and putting them in makeshift hospitals.

Comment by douglas_knight on Douglas_Knight's Shortform · 2020-03-28T17:11:16.562Z · score: 2 (1 votes) · LW · GW

two tweets:

[Carl Schmitt is a good philosopher but] One nightmarish way to understand The Discourse is that somehow Carl Schmitt became the obvious, agreed-upon, common-sense interpretation of politics. All sides nod sagely, but each fetishizes a different book.
So the left gets Political Theology, the right gets Nomos of the Earth. Quilette-style centrists are 100% indebted to the Concept of the Political.
Comment by douglas_knight on Douglas_Knight's Shortform · 2020-03-28T17:02:42.773Z · score: 4 (2 votes) · LW · GW

The Wealth of Nations by Adam Smith is a great book. My favorite part is Book III, Chapter 4 on the end of feudalism. In particular, I like these two paragraphs:

In a country which has neither foreign commerce, nor any of the finer manufactures, a great proprietor, having nothing for which he can exchange the greater part of the produce of his lands which is over and above the maintenance of the cultivators, consumes the whole in rustic hospitality at home. If this surplus produce is sufficient to maintain a hundred or a thousand men, he can make use of it in no other way than by maintaining a hundred or a thousand men. He is at all times, therefore, surrounded with a multitude of retainers and dependants, who, having no equivalent to give in return for their maintenance, but being fed entirely by his bounty, must obey him, for the same reason that soldiers must obey the prince who pays them.

In a country where there is no foreign commerce, nor any of the finer manufactures, a man of ten thousand a year cannot well employ his revenue in any other way than in maintaining, perhaps, a thousand families, who are all of them necessarily at his command. In the present state of Europe, a man of ten thousand a year can spend his whole revenue, and he generally does so, without directly maintaining twenty people, or being able to command more than ten footmen not worth the commanding. Indirectly, perhaps, he maintains as great or even a greater number of people than he could have done by the ancient method of expense. For though the quantity of precious productions for which he exchanges his whole revenue be very small, the number of workmen employed in collecting and preparing it must necessarily have been very great. Its great price generally arises from the wages of their labour, and the profits of all their immediate employers. By paying that price he indirectly pays all those wages and profits and thus indirectly contributes to the maintenance of all the workmen and their employers. He generally contributes, however, but a very small proportion to that of each, to very few perhaps a tenth, to many not a hundredth, and to some not a thousandth, nor even a ten-thousandth part of their whole annual maintenance. Though he contributes, therefore, to the maintenance of them all, they are all more or less independent of him, because generally they can all be maintained without him.
Comment by douglas_knight on mind viruses about body viruses · 2020-03-28T16:18:09.137Z · score: 14 (8 votes) · LW · GW
Okay, words aside, does the right strat­egy look like the fa­mous GIF taken liter­ally, or like a feed­back sys­tem where we keep turn­ing so­cial dis­tanc­ing on and off so the graph looks like a heart rate mon­i­tor, or like a “ham­mer” re­set fol­lowed by a suc­cess­ful em­u­la­tion of South Korea, or
I don’t know and you don’t know and To­mas doesn’t know and Carl doesn’t know. It’s hard!

No, it is really easy. If the society is capable of executing the feedback system, then it has a method of driving R0 below 1. If it can do that, it should execute the hammer. This requires less aggregate time under extreme measures than the feedback system. If the society fails to execute the dance, then it can try the feedback system, but that sounds a lot harder to me than the dance.

Like me, these people make both substantive and semantic objections.  In fact, theirs are a strict superset of mine (see that last Bergstrom thread re: Gaussians!).
I am not saying “look, I was right, the experts agree with me, please recognize this.”  I mean, I am saying that.

But you were wrong about Gaussians. So much the worse for the experts.

There are only two choices: herd immunity, or the dance. Herd immunity can be accomplished by vaccine, by permanent change to society to reduce R0, or by having 1-1/R0 proportion of people survive the disease. Without a vaccine and without permanent change to society, the area 1-1/R0 is fixed, although a full-speed epidemic could infect even more.

Comment by douglas_knight on Authorities and Amateurs · 2020-03-26T01:13:32.841Z · score: 3 (4 votes) · LW · GW

You've lost track of the object level here.

What did his post originally mean? I'm not allowed to read people's minds. He admits that no one took from it what he wanted them to take from it. Lanrian said that it was "a reasonable critique...that it doesn't make sense to assume a normal distribution." That was a qualitative complaint and he admitted that it was qualitatively wrong.

Comment by douglas_knight on Authorities and Amateurs · 2020-03-26T00:49:11.518Z · score: 3 (2 votes) · LW · GW


Comment by douglas_knight on History's Biggest Natural Experiment · 2020-03-25T18:14:27.695Z · score: 4 (2 votes) · LW · GW

Yes, I meant exactly a 1 year effect and a random 5 year window and I should have spelled that out. Cancer and many autoimmune diseases take years to notice.

And Jim is right that my example is poor because a prenatal effect pinpoints when the exposure must have occurred. But, actually, I believe that such simple studies don't find an effect for the normal variation in flu (3% some years, 15% others), except for the 1918 cohort. The studies that find an effect rely on asking the mother if she had flu during pregnancy (which has some post hoc problems).

Comment by douglas_knight on Authorities and Amateurs · 2020-03-25T14:34:34.191Z · score: 9 (13 votes) · LW · GW

The gaussian assumption makes no difference. Nostalgebrist's post is a math error. He later admitted that at SSC, but barely updated his post.

Comment by douglas_knight on Against Dog Ownership · 2020-03-24T20:36:19.638Z · score: 3 (2 votes) · LW · GW

Where can I read about this? It seems easier to verify than the claims in the OP.


I looked at google images for "dog body language." I see some people saying that showing the belly is happy, while others said it was submission (and one saying both!). This seems like a good example of what you are saying, but are there more? Added: nose lick as stressed vs peace, although maybe that isn't even a confusion, but just opening the box of what it means to sue for peace.

Comment by douglas_knight on History's Biggest Natural Experiment · 2020-03-24T16:29:05.812Z · score: 5 (3 votes) · LW · GW

I am skeptical that this will have measurable effects because most these diseases indirectly caused by infection have long lag times, often years, if only time to diagnosis. For an extreme example, prenatal flu increases the odds of schizophrenia, 20 years later.

Comment by douglas_knight on Using smart thermometer data to estimate the number of coronavirus cases · 2020-03-23T22:56:49.721Z · score: 2 (1 votes) · LW · GW

OK, maybe it does show a hot spot in Seattle. (though I'm not convinced)

But what about Florida? It seems to me that when you predict a million coronavirus cases, your method is predicting that half of them are in Florida. So you really have to look closely at the state. Surely the decline starting on March 19 does not represent the very minimal social distancing imposed on March 17.

Comment by douglas_knight on Using smart thermometer data to estimate the number of coronavirus cases · 2020-03-23T18:27:07.468Z · score: 2 (3 votes) · LW · GW

Is this plausible? Look at the fine-grained results.

Hot spot in New York.

No hot spot in Seattle.

The biggest hot spot is Florida. The State is handling this very badly, so it is plausible that it the worst but not yet recognized; it would be useful to have an alternate source of information. But look at the time series. It peaked in Florida (all counties show the same) on March 19th and declined afterwards. Given how badly Florida has handled this and the delay from infection to fever, it doesn't seem plausible to me that this measures covid-19.

Comment by douglas_knight on Forecasting an 80% chance of an effective anti COVID-19 drug (probably Remdesivir) · 2020-03-17T16:25:51.983Z · score: 4 (3 votes) · LW · GW

I don't see him saying that. I didn't listen to the podcast, but I used speech-to-text to search. Every youtube video uploaded these days has a transcript. On a computer (not phone) click on the three dots under the video to the right of share/save and choose "open transcript." I looked at the 16 times he mentioned antivirals. Speech-to-text isn't good enough to search for "remdesivir," but it was mentioned at 44 minutes. (I also searched for side effects. This is more a concern for vaccines given to healthy people than antivirals given to people who are already sick.)

(Also, castbox is a podcast app that has transcripts of podcasts not on youtube, but I haven't had a lot of luck with it.)

Comment by douglas_knight on Has a technological dependency graph been made? · 2020-02-28T16:05:53.215Z · score: 6 (4 votes) · LW · GW

That's a good idea, but a word of warning. History is based on such little data that it must be the product of circular reasoning. Most historical claims about what technologies existed are based on implicit beliefs about tech trees, while most beliefs about tech trees are based on explicit beliefs about what technologies existed. Yes, it is better to make the beliefs about tech trees explicit, although there is risk of ossifying these beliefs. Anyhow, I recommend that you widen your priors by reading Lucio Russo's book Forgotten Revolution about how much science and technology the ancient Greeks had.

Comment by douglas_knight on What is Life in an Immoral Maze? · 2020-01-06T04:59:24.409Z · score: 19 (6 votes) · LW · GW
in Moral Mazes there are at least 25 (!) levels of management

Exponential growth makes that implausible.

In the US military there are 25 ranks, but a hierarchy of half that depth with a branching factor of 3. Commissioned officer ranks correspond pretty well to the hierarchy, but there are only 3 levels of enlisted hierarchy below them.

You seem to be referring to this passage:

A weeding-out process takes place among the lower ranks of managers during the first several years of their experience. The early careers of promising young managers are highly variegated; the more promise managers show, the more probations they must undergo. Take, for example, the case of a young man newly graduated in 1965 from one of the South’s leading universities. He joined Weft Corporation and spent the next two years in the company’s production management training program. Then he became a first-line supervisor on the third shift at a small mill. Shortly thereafter, he was promoted to the night superintendent’s job of that mill and given overall responsibility for the night shift. After six months, he became a department head for weaving operations in another mill. After another six months, he was assigned to head a larger weaving department in yet another plant. After still another six months, he became assistant plant manager at a medium-sized mill and kept that job for four years. Then he moved to a still larger mill in the same capacity for another two years. Then he became plant manager of a medium-sized mill for two years. Finally, he was named one of two group managers with six plant managers reporting to him. At the age of 36, he has reached grade 20, the “breaking point” on a scale of 29, placing him in the top 12.17 percent of management in Weft Corporation with, he hopes, a clear shot at becoming vice-president of manufacturing. Similarly variegated careers are evident for young marketing and sales managers in Weft’s northern offices. In Alchemy Inc., whether in sales, marketing, manufacturing, or finance, the “breaking point” in the hierarchy is generally thought to be grade 13 out of 25 or the top 8.5 percent of management.

The low levels of these ranks probably provide for recognition of non-management employees, like the enlisted and warrant ranks in the US military. With a branching factor of 2, top 12% would mean 3 levels up from the bottom, not 20. With 9 levels above above 20, a total of 13 levels of managers. The other company, perhaps 15. But probably the top of the hierarchy is not actually 9 or 12 ranks, but sparser than they suggest, not as sparse as the lower ranks, but not completely full like the US military.

Comment by douglas_knight on Could someone please start a bright home lighting company? · 2019-12-05T18:43:46.902Z · score: 4 (3 votes) · LW · GW

What do you want?

Do you want to buy something for yourself, or do you want a company to change the world?

Yes, there is room for a better product, but I think that off-the-shelf products are pretty good and you should just get them. If you want to change the world, maybe you should just promote these existing products. In particular, for your short term needs, just do it.

I think that the right answer for most people and most purposes is Raemon's instructions, $300 for 300 watts, same total wattage as coelux. Why did you write this post already knowing about Raemon's instructions? What are they lacking? That they require installation? If you have 24 separate bulbs spread around the room, installation is unavoidable. Light strips may be a better solution, but they require even more installation.

Some people want different things. David Chapman seems to want to illuminate his desk, not his room, so he might not like Raemon's setup. If you want to minimize installation, you might want a single light. This leads to Ben and Ashen's suggestions. They probably aren't as nice as coelux, so, yes, it would be nice if someone made nicer versions (which should be possible). Ashen's outdoor floodlights probably have lousy CRI. Ben's corncob isn't the standard residential fixture, and thus required some assembly. Both products probably shine outwards to illuminate an area, rather than the coelux which is intended to mimic the sun through a window pushing light in a sharp line. This illusion is probably luxurious, but I'm skeptical that it is actually good for the goal.

I was going to follow up by saying that if you like the form factor of coelux, there are similar products on the market for maybe $2/watt, only twice as expensive as Raemon's setup. They aren't as bright as coelux, but you could get 5 or 10. There is the second product Ashen linked or light therapy boxes (apparently 72W) are probably a good option with full spectrum and good lenses. But then I read more I heard a lot of accusations of poor quality and fraud around light boxes, so I dunno.

Comment by douglas_knight on CO2 Stripper Postmortem Thoughts · 2019-12-05T03:53:28.616Z · score: 2 (1 votes) · LW · GW

You asked for an expert consensus and I gave it to you. Naval researchers are the experts.

No, "experiments yield results in different directions" is not an accurate summary. Experiments with large interventions trump experiments with small interventions.

But, it's true, I left out the most convincing evidence, which is back of the envelope calculations with gross anatomy.

Comment by douglas_knight on CO2 Stripper Postmortem Thoughts · 2019-12-04T22:53:09.954Z · score: 5 (2 votes) · LW · GW

When people say that ventilation helps them, I believe them. They might even be far on an axis of response to pollution. But how would they know that the particular pollutant they respond to is CO2? They should be cautious in assigning blame and trying specific interventions. Gwern points out that one of the studies that most impressed Paul about CO2 actually found larger effects from mold, which is a big problem in the foggy slums of Berkeley. In theory there are ways to isolate human pollution from house pollution, such as varying the number of roommates, but I doubt people are careful enough to disentangle that and CO2 isn't even the only human pollutant. [Added: but submarines are equally subject to all human pollutants, so that should limit the possibilities to the short list of what they scrub.]

Are submariners selected on that axis? I'm skeptical. In any event, the naval studies don't restrict to submariners.

Comment by douglas_knight on Long-lasting Effects of Suspensions? · 2019-12-04T17:07:49.110Z · score: 2 (1 votes) · LW · GW

It would be nicer if there were more randomization, but it would also be nicer if more information were extracted from the few people who are randomized. For example, I know someone who participated in an RCT of breastfeeding/formula. It was aimed at a specific (acute, adverse) infant outcome. I'm not sure it even looked at other infant metrics, but it certainly did not have long-term follow-up, not even at 5 years. Not only did the study make a big investment in persuading the subjects for such little measurement, but it is now impossible to do a better experiment, because RCTs of breastfeeding are now considered unethical because of the damage their null results do to the authors' careers. (Similarly the Swedish and Australian twin registries are the right way to do twin studies.)

On the other hand, sometimes you can't randomize and you'd like to know how well you can do correlational studies. If your employer is so enthusiastic about experiments, maybe it apply that enthusiasm to itself and do an experiment to see how well its employees can do observational analysis?

Comment by douglas_knight on CO2 Stripper Postmortem Thoughts · 2019-12-02T19:15:50.033Z · score: 3 (2 votes) · LW · GW

Well, who are the experts? Submarines routinely have CO2 levels much higher than even Berkeley group homes. Naval researchers do experiments with higher levels still, showing little effect. There seems to be an illegible LW consensus to the opposite, probably from people pretending to read this post. People praise Gwern for his quantity, but they don't actually read him.

Again, most research is about ventilation and is thus confounded by other pollutants. I don't usually speak up about this because most discussion of this doesn't depend on the CO2 hypothesis.

Comment by douglas_knight on CO2 Stripper Postmortem Thoughts · 2019-12-02T05:09:15.650Z · score: 7 (2 votes) · LW · GW

Ventilation has the advantage that it dumps all pollutants, not just CO2. In fact, the premise that CO2 affects cognition is false.

Comment by douglas_knight on The unexpected difficulty of comparing AlphaStar to humans · 2019-11-30T02:33:57.203Z · score: 4 (2 votes) · LW · GW

This is super tangential, but I think you're making a technical error here. It's true that poker is imperfect information and it's true that this makes it require more computational resources, which matches the main text, but not this comment. But does imperfect information suggest mixed strategies? Does optimal play in poker require mixed strategies? I see this slogan repeated a lot and I'm curious where you learned it. Was it in a technical context? Did you encounter technical justification for it?

Games where players move simultaneously, like rock-paper-scissors require mixed strategies, and that applies to SC. But I'm not sure that requires extra computational resources. Whether they count as "imperfect information" is subject of conflicting conventions. Whereas play alternates in poker. I suspect that this meme propagates because of a specific error. Imperfect information demands bluffing and people widely believe that bluffing is a mixed strategy. But it isn't. The simplest version of poker to induce bluffing is von Neumann poker, which has a unique (pure) Nash equilibrium in which one bets on a good hand or a bad hand and checks on a medium hand. I suspect that for poker based on a discrete deck that the optimal strategy is mixed, but close to being deterministic and mixed only because of discretization error.

Comment by douglas_knight on Market Rate Food Is Luxury Food · 2019-11-29T02:19:08.583Z · score: 2 (1 votes) · LW · GW

Maybe this argument is a straw man. That is, maybe it's not accurately describing the arguments that people use. But that is a very different problem than saying this argument might be OK.