Posts

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

Comments

Comment by Douglas_Knight on Covid 3/12: New CDC Guidelines Available · 2021-03-16T03:01:02.627Z · LW · GW

That's a group selection argument.

Comment by Douglas_Knight on Covid 3/12: New CDC Guidelines Available · 2021-03-15T03:52:34.890Z · LW · GW

Although it is widely held by biologists that there is a tradeoff between infectiousness and virulence, people (eg, Paul Ewald) who actually study the evolution of virulence say the opposite, both in theory and data. In the case of sars2, it is overdetermined: death is due to immune overreaction, after the window of selection is over.

Comment by Douglas_Knight on Douglas_Knight's Shortform · 2021-03-02T21:32:12.897Z · LW · GW

Recently I've run across several people offhandedly offering disclaimers that they've never personally checked that the earth is round. I've never thought to check either, but a moment's reflection reveals that I've traveled far enough, both north-south and east-west that the curvature is obvious

Time zones are a measure of the curvature of the earth. When I travel from New York to California, I know that noon has changed, trusting only my wristwatch. In fact, it was pretty clear to my circadian rhythm. Or just make a phone call to someone you trust on the other coast. Most of the time I don't discuss sunlight on such calls, but it has come up.

I've traveled shorter distances north-south. I've been to Glasgow, which is 15 degrees north of NYC. If I went in the spring, there's probably little to notice, but I went near the solstice, when it was obvious that the nights were much shorter. Similarly, if you go south 15 degrees to Miami, I'm told that the winter and summer day lengths are obviously moderated.

Accurate clocks and instant communication give us a big advantage over the ancients, but the north-south method is largely unchanged.

Added: East-West travel produces a linear effect. North-South travel produces non-linear effects, which can be easy to notice. If I were measuring the height of the sun at noon, that would be linear in the latitude. 15 degrees might be enough to measure without instruments, if I chose to think about it. But the length of the night is not linear. Summer solstice night heads to zero not at the north pole, but just at the arctic circle. So summer solstice in Glasgow was obviously shorter than any night I had previously experienced, maybe cut in half. Whereas summer solstice in Miami is shorter than summer solstice in New York, but just an ordinary length day from other times of the year. Maybe if I had thought to ask the question I could have told the difference without a clock, but I didn't think about it, whereas the night in Scotland was striking and a topic of conversation. If someone from Miami comes to New York for the solstice, he will experience the shortest night of his life, which might be obvious, but it won't as dramatic as half the length he's used to. I knew a guy who moved from Miami to New York and he noticed it, but I think it was about the experience of life, not a single night. There is another non-linear effect as you head to the tropics, which is the solstice shadows at noon shorten to zero. That might be obvious to some people, but it's not the kind of thing I pay attention to.

Comment by Douglas_Knight on Covid 2/4: Safe and Effective Vaccines Aplenty · 2021-02-08T21:04:14.537Z · LW · GW

All the modeling efforts talked about in the write-up are doomed because they don’t understand the role of the control system

Is that a forward-looking prediction? What consequences was UIUC doomed to? The article ends its coverage of UIUC in early September, declaring it a failure. But, in fact, it achieved its goal of keeping infections below 5k. You could credit that to "the control system," the panic caused by the early spike that got all the news coverage, but it's still a success.

Comment by Douglas_Knight on Making Vaccine · 2021-02-06T03:41:43.023Z · LW · GW

If it's so cheap and easy to make vaccines, why aren't commercial ones made this way? In particular, the Novavax vaccine sounds similar, so why wasn't that the first vaccine to market?

Added: Specifically, the ultimate purpose of a vaccine is to get protein into the body. Traditional vaccines grow the virus using its own reproductive apparatus. Fancy new vaccines, like the adenovirus and mRNA vaccines inject instructions and induce the subject to manufacture proteins. But if it's so easy to just print proteins, why don't we do that? That's what Novavax does, unlike the ~7 vaccines that beat it to market.

Added: one difference is that all the vaccines that made it to market, including, I think, Novavax, used the whole spike protein, whereas this proposal uses short peptides. Identifying the right short snippets takes time, while using the whole protein is simpler and more likely to work. The cost of peptides is probably super-linear in length. Still, I remain confused about Novavax.

Comment by Douglas_Knight on Covid 1/28: Muddling Through · 2021-01-28T19:20:15.814Z · LW · GW

By default, we should expect viruses to become less deadly over time rather than more severe, but more severe is always a risk. We also believe the new strain carries generally higher viral loads, which could plausibly be a cause of higher severity.

That is the consensus among biologists, or maybe even evolutionary biologists, but the consensus among people who study the evolution of parasites is exactly the opposite. Theory predicts and observation agrees that parasites evolve to become more virulent over time, especially ones that have just jumped from one host to another. See Paul Ewald, for example.

Comment by Douglas_Knight on Covid: Bill Gates and Vaccine Production · 2021-01-28T17:12:29.900Z · LW · GW

I suspect that Gates had a long-standing specific plan for manufacturing old-fashioned vaccines, but was unable to pivot to funding new vaccines. It's a lot harder to spend money to speed up deployment of new technologies, especially at arm's length.

AZ claims this week that the EU negotiations being delayed for a couple months delayed their factories. Why couldn't they just start earlier? This is a clear claim that money would have mattered. But maybe there is a lot more to this than physical construction. The EU is currently threatening to confiscate AZ vaccine, so maybe AZ didn't see any point in building factories in countries that hadn't pre-paid.

Comment by Douglas_Knight on Covid: Bill Gates and Vaccine Production · 2021-01-28T16:58:06.979Z · LW · GW

Why is the Gates foundation a charity, as opposed to just a non-profit? If he wants to take Buffett's money and give him a tax benefit, then it has to be charity, but for spending his own money, he doesn't need this status.

Comment by Douglas_Knight on Covid 1/21: Turning the Corner · 2021-01-22T05:21:47.287Z · LW · GW

Given the high dimension of the search space, I think (b) is negligible and the linear model (a) of your first comment is better. In low dimension the boundary of the unit sphere is small and you can have a lot of copies on the inside, having to pass through the sphere to reach new terrain. Whereas, in high dimensions, the population will quickly thin out and all be unique, so what matters is the total volume of space explored, not how long it takes to get anywhere.

Comment by Douglas_Knight on Covid 1/21: Turning the Corner · 2021-01-22T04:37:45.824Z · LW · GW

Dynamically, that predicts that the advantage would rise over time, as a substantial proportion of the population got infected by the original strain. I think we've been monitoring the UK variant enough to see that this is not a large portion of its advantage.

Even statically, I think it's difficult to make the numbers work out. The idea that the population has heterogeneous risk makes a lot of sense, but it doesn't seem born out by the basic prediction that HIT is much lower than predicted by initial R. I suppose the control system might make it hard to observe how close we are getting to HIT, but I'm skeptical. And you need zero cross-immunity. But then wouldn't the observed reinfections skyrocket and be obvious? Maybe if the reinfections are asymptomatic, but just as contagious. I guess that this very specific scenario predicts that the new variant has passed through the riskiest portion of the London population and no longer has an advantage there.

Comment by Douglas_Knight on Covid 1/21: Turning the Corner · 2021-01-21T21:02:33.755Z · LW · GW

My (highly speculative!) hypothesis is that the emergence of these variant viruses arises in cases of chronic infection during which the immune system places great pressure on the virus to escape immunity and the virus does so by getting really good at getting into cells. 11/19

That’s plausible, but doesn’t explain why the chronic infections hadn’t done this earlier, and the English strain doesn’t escape immunity in this way (and we don’t know about the others) so I notice it doesn’t feel like it explains things.

That isn't plausible to me. Not only does it not explain everything, I don't think it explains anything. There are two traits, immune escape and infectiousness, which is pretty much the same as fitness. It makes sense that chronic infection causes immune escape, but we aren't looking at immune escape. There might be tons of strains with immune escape out there, but we won't notice until we achieve substantial herd immunity, at which point they will have increased fitness. If that happens, and we need to explain immune escape in future strains, then this is a possibility. But I see no reason to believe it about these strains. We are filtering by increased fitness, so we need to explain increased fitness. Immune escape is probably a side effect of increased fitness, because it's a potential side effect of any change.

Fitness is fitness. There is no need for a weird environment to explain selection for fitness, because it's the main story. In fact, this is exactly backwards. If there is any trade-off between reproduction between hosts and reproduction in hosts, then selection by chronic infection will favor the latter at the expense of the former and probably move away from optimal fitness. Whereas it appears, as Bedford claims, that these mutations are a free lunch. This is quite plausible for a young parasite that hasn't finished adapting to its new host. But then what does a weird selection pressure explain? Every infection should be an opportunity to develop this. What matters is the number of roughly the number of virus-host-days. A virus trapped in one individual for 100 days is under roughly the same selection pressure as a virus passing through 20 individuals under the same time. I've seen a lot of people call that "fast evolution" because 20x as much selection happens in the host, but it's not any faster in time.

Bedford seems to allude to this view, but he also uses similar points to make a different argument, which is that the UK variant appeared out of nowhere with a lot of mutations, without the intermediate forms being observed. Arising from a chronic infection would explain this. But we also know that all the mutations have arisen before, without seeming to do much on their own. So we know that it only took off when it reached the magic combination, which explains why we didn't see the intermediate forms. I don't think chronic infection adds much to this, although I could imagine a scenario in which the individual changes happen to be selected for in chronic infection at the slight expense of general fitness. So it needs the different landscape of a chronic infection to cross the valley and reach the final form. But this seems like excess detail to me.

Comment by Douglas_Knight on Covid 1/7: The Fire of a Thousand Suns · 2021-01-09T14:49:45.940Z · LW · GW

I don't believe that every infected cell is killed.

Comment by Douglas_Knight on Covid 1/7: The Fire of a Thousand Suns · 2021-01-08T19:59:14.589Z · LW · GW

This is about the change that is shared between UK and SA, not about the change in the SA variant that is uniquely worrisome.

Comment by Douglas_Knight on Covid 1/7: The Fire of a Thousand Suns · 2021-01-08T18:24:33.321Z · LW · GW

That doesn't distinguish the new vaccines from traditional attenuated vaccines, which also inject nucleic acid into the cell, although inactivated vaccines don't. Or, for that matter, getting infected. Every year I get a couple of colds, where way more cells get way more nucleic acid than in vaccine.

Comment by Douglas_Knight on Covid 1/7: The Fire of a Thousand Suns · 2021-01-08T04:39:07.297Z · LW · GW

10% chance of survival is "unlikely" but it's lives, not warm fuzzies. It may not have been cost-effective, but it was buying something real. (And it may well have been cost-effective if it was marginal use of facilities that already existed.)

10% is the number I have heard for the specific category I have heard triaged against in LA, namely patients whose heart the EMT can't restart. Instead they can perform CPR to manually pump the heart on the trip to the hospital.

Comment by Douglas_Knight on Covid 1/7: The Fire of a Thousand Suns · 2021-01-08T04:26:48.464Z · LW · GW

this is 3,809 deaths per day and the average had not at the time been over 2,700 for any 7-day period so far, that that projection was made on December 31, so that seems like a rather bold prediction

If cases are in unprecedented territory, you should expect deaths to go into unprecedented territory. Forecasting deaths 3 weeks in the future should be easy, since it's forecasting the progression of the disease among people already infected. You just take the number infected of cases and multiply by the CFR. At 220k cases/day and a CFR of 1.75%, that's 3850 deaths/day, so that sounds totally reasonable.

(This is all from eyeballing 7-day rolling averages. I got the CFR from comparing the plateau of cases in September to the plateau of deaths in October. I haven't checked if it has held up or what the lag should be. The 220k cases/day is the 7-day average peak at Christmas. So this requires assuming that the dip after Christmas was an illusion, which has been vindicated by the week since the prediction was made.)

Added: over at Zvi's blog, Alexander Gordon-Brown claims that CDC does not predict 80k deaths, but that its point estimate is only 64k. Also, that the prediction is dated 12/21.

Comment by Douglas_Knight on New SARS-CoV-2 variant · 2020-12-28T15:52:08.795Z · LW · GW

If you make any of these hypotheses precise enough to calculate, then I don't think that they are likely enough to be worth calculating. The point was just to give suggest how big the space of unknown unknowns is. I think you need an outside view to estimate it. You might hope to get that from the virologists, but they are dismissing it as a "founder effect" which is even more specific, rather than accepting the ignorance of an outside view.

I think I got them all from Francois Balloux, though I'm not sure what he was saying and I may have interpolated a lot of detail. I got 2a and maybe 1 from here. 2b is from here, a response to the first thread. Added: actually, I think I got 2a from the "Does it matter" video, which was generally hostile to reason and knowledge epidemiology, but did suggest something like this at the end.

Comment by Douglas_Knight on New SARS-CoV-2 variant · 2020-12-28T03:35:22.357Z · LW · GW

How likely is it that the spread of this new strain was caused by a few superspreaders, and that most of the above is blown out of proportion?

The probability is basically zero. But the question is whether the obvious hypothesis is false, not whether a particular alternate hypothesis is true. There are many alternate hypotheses more likely than that, and most important the final bucket of "other," hypotheses that I have not thought of. Here are a few suggestions I have heard: (1) It is biologically different and spreads easily in children, but if they closed the schools, it would be negligibly different. (2) Measurement error. 2a: All sars2 produces RNA after infectiousness is over, effectively false positives and maybe this produces lots of RNA for much longer, effectively higher false positives and more lagging signal. 2b: It is mainly detected by applying the usual 3 tests for sars2 and failing a particular one. So it is inherently a noisier test, although it is not at all clear how this could produce the observed pattern.

Comment by Douglas_Knight on How long does it take to become Gaussian? · 2020-12-15T02:52:12.016Z · LW · GW

If the center goes to a unique limit and the tails go to a unique limit, then the two unique limits must be the same. CLT applies to anything with finite variance. So if something has gaussian tails, it just is the gaussian. But there are infinite variance examples with fat tails that have limits. What would it mean to ask if the center is gaussian? Where does the center end and the tails begin? So I don't think it makes sense to ask if the center is gaussian. But there are coarser statements you can consider. Is the limit continuous? How fast does a discrete distribution converge to a continuous one? Then you can look at the first and second derivative of the pdf. This is like an infinitesimal version of variance.

The central limit theorem is a beautiful theorem that captures something relating multiple phenomena. It is valuable to study that relationship, even if you should ultimately disentangle them. In contrast, skew is an arbitrary formula that mixes things together as a kludge with rough edges. It is adequate for dealing with one tail, but a serious mistake for two. It is easy to give examples of a distribution with zero skew that isn't gaussian: any symmetric distribution.

Comment by Douglas_Knight on How long does it take to become Gaussian? · 2020-12-14T16:09:58.777Z · LW · GW

The central limit theorem is a piece of pure math that mixes together two things that should be pretty separate. One is what the distribution looks like in the center and the other is what the distribution looks like at the tails. Both aspects of the distribution converge to the Gaussian, but if you want to measure how fast they converge, you should probably choose one or the other and define your metric (and your graphs) based on that target.

Skewness is a red herring. You should care about both tails separately and they don't cancel out.

Comment by Douglas_Knight on Are the social sciences challenging because of fundamental difficulties or because of imposed ones? · 2020-11-16T19:57:06.429Z · LW · GW

There's an inherent difficulty you don't list. You might file it under "political agendas," but the big problem isn't the external constraint of conscious agendas, but of people fooling themselves.

Comment by Douglas_Knight on Covid 11/12: The Winds of Winter · 2020-11-13T03:59:53.229Z · LW · GW

Press releases about defined endpoints of phase 3 trials are the ones that move the stock price the most, next to mergers. Probably across all companies, not just pharma. The SEC would come calling if they were lies.

Comment by Douglas_Knight on Douglas_Knight's Shortform · 2020-11-12T23:11:10.639Z · LW · GW

It's pretty common for there to be coverups with no crime protected. People just close ranks and reflexively lie. So coverups are rarely good evidence of the primary crime. But they are evidence of a sick culture.

Comment by Douglas_Knight on Covid 11/12: The Winds of Winter · 2020-11-12T20:13:31.392Z · LW · GW

But 10% of the entire population developing a new mental illness every three months is much worse! 

That's not what the paper says. It says that 10% of people with the flu or a broken bone in 2020 are getting diagnosed. It doesn't say how many people with the flu or a broken bone were diagnosed in 2019, nor how many people without any other reason to go to the doctor in 2020.

Comment by Douglas_Knight on Covid Covid Covid Covid Covid 10/29: All We Ever Talk About · 2020-11-05T17:03:21.734Z · LW · GW

So what? There are individuals that have reached 100% infection. I'm talking about America, which is currently undergoing an epidemic. I'm predicting that the past 2 weeks of behavior will continue for the next two weeks. To predict otherwise on the grounds of herd immunity is to claim that it has been achieved just this week. Whereas, Qatar has steady levels, suggesting herd immunity in the current environment.

Comment by Douglas_Knight on Covid Covid Covid Covid Covid 10/29: All We Ever Talk About · 2020-10-31T20:28:08.785Z · LW · GW

I'm talking about people infected in the next 2 weeks. I don't see how that is an answer unless we have already achieved herd immunity. And I'm skeptical that you can call the peak that precisely.

Comment by Douglas_Knight on Covid Covid Covid Covid Covid 10/29: All We Ever Talk About · 2020-10-31T12:30:56.926Z · LW · GW

But on the question of whether the worst is behind us, the answer is probably yes.

This is underspecified. There are lots of ways that things could be bad.

In March and April, the hospital system broke down. Supply lines were hanging by a thread and we were having trouble finding ways to put literal food on our literal table, especially meat. Thousands were dying each day. Supply chains and the whole economy and the market on the verge of collapse. It’s easy to forget how bad things were at first.

OK, the disruption probably isn't coming back. But the deaths? You're predicting deaths shooting up; from current cases, I predict 1500/day. And how much higher will cases go? It's easy to imagine that they go up another 50% and deaths exceed previous rates. That's only about 2 weeks of growth at current rates. Why would they stop here?

Comment by Douglas_Knight on Douglas_Knight's Shortform · 2020-10-23T21:58:33.773Z · LW · GW

A common historical paradox is that centralizing forces can break apart large organizations. Usually what happens is that the large organization was fake, nominally claiming wide domain while actually being weak. When real power centralizes enough to defy the fake power, it secedes, producing the appearance of decentralization.

At least, my cached thought is that it's common. I can't remember what examples lead me to it. The only example I can think of right now is the Holy Roman Empire. A less paradoxical situation is that rapidly changing power produces uncertainty and civil war. Maybe my previous examples were things like the English Bill of Rights, where the King makes an explicit concession, but this is only necessary because the centralizing forces made the king powerful enough to need to clarify how powerful. (And Parliament is almost as centralized as the King, so this hardly even has the appearance of decentralization.)

I was reminded of this by people pushing back on Samo Burja's claim on the centralizing effects of the printing press. I think that this is a logical error. Just because the press broke the Western Church, doesn't mean that it did so for decentralizing reasons. On the other hand, a false argument doesn't mean a false conclusion. You have to look at the details to decide whether the mechanism was centralizing or decentralizing, which is a lot to ask for a tweet. FWIW, Burja only claimed a net centralizing effect, classifying the effect on the Church as decentralizing:

The printing press reduced the Catholic Church’s control over intellectual institutions. But it also paved the way for the standardization of language and for more direct control by state bureaucracies. Society was vastly more centralized in 1750 than it was in 1400.

Added: So, of course, I wrote this because my first thought on seeing the tweets was that Reformation was an example of this, but then I became uncertain about the example. Now I'm wondering if it was actually the motivating example when I first cached this thought. Anyhow, I do think that the nominal power and organization of the Church are misleading. Added: Yes, I think the Western Schism was my original example. I still think that's right, that it was caused by centralizing forces. I'm just not sure how the printing press fits in.

Comment by Douglas_Knight on What Does "Signalling" Mean? · 2020-09-21T18:37:39.794Z · LW · GW

But we already have a term for signalling desirable properties about yourself: virtue signalling!

That's not what "virtue signalling" means.

Comment by Douglas_Knight on Covid 9/17: It’s Worse · 2020-09-17T23:14:51.667Z · LW · GW

For those worried, yes, the halted vaccine trial from last week has resumed

I don't believe that this is true. All the articles from last week say that it's only the British trial resuming, not the American. I believe that British trial is almost full, so it's irrelevant that it resume, whereas the American trial, the biggest, has barely begun. I'm not sure how far Brazil and India have gotten. I guess India resumed on Tuesday. Here is an article about FDA not resuming.

Comment by Douglas_Knight on Were vaccines relevant to 20th century US mortality improvements? · 2020-09-11T02:55:59.048Z · LW · GW

I believe that a much stronger statement is true. For almost every viral disease with a vaccine, there was a 90% reduction in mortality before the advent of the vaccine. The only graph I have on hand is measles:

measles mortality time series

Of course, if A causes a 90% reduction in mortality and B causes a 90% reduction in mortality, and they are independent, in a causal sense they are equal and you shouldn't judge their effects based on which one is deployed first. But once one is deployed, the marginal value of adding the other is only 10% as much. Even if B causes a 100% reduction, its marginal value beyond A is only 10% of the initial value of A.

(There is also a theory that measles resets your immune system and wipes out acquired immunity, so avoiding measles saves even more lives. So then a vaccine would be much more valuable than surviving measles.)

Comment by Douglas_Knight on Covid 8/20: A Little Progress · 2020-09-02T15:54:55.475Z · LW · GW

It is funded by Emergent Ventures, which was founded with the purpose of being a one-man shop. I'm pretty sure he still makes the final decisions, but is that the same thing? As you say, contra Zvi's claim, Emergent Ventures has grown into a bureaucracy. What's wrong with bureaucracies? Cowen appears to have solid control over EV. If the bureaucracy fails at the legible goal of being fast, he will simply fire it. But it is doing some filtering that he used to do, which is harder to police.

Comment by Douglas_Knight on Covid 8/27: The Fall of the CDC · 2020-09-02T15:25:10.107Z · LW · GW
a decrease of 3.2%

Don't say that. Say a decrease of 3.2 points. Or "percentage points," for clarity. Worst case scenario is that people are confused, which is better than wrong.

Comment by Douglas_Knight on Douglas_Knight's Shortform · 2020-08-24T22:05:52.551Z · LW · GW

Sprinkling in the words "supply" and "demand" doesn't make it an argument.

I would be very happy if it were just a model, but it is not even a model. That's exactly the problem.

If you want to make an argument, you have to actually say something about supply and demand. You have to connect slavery (or any other aspect of any particular time and place) to supply or demand, or, better, both.

It is much more popular to argue that cheap labor caused the industrial revolution than that expensive labor caused the industrial revolution. Maybe expensive labor caused the agricultural revolution, which overshot and produced cheap labor, which in turn caused the industrial revolution. But if you can't tell the difference between that claim and the claim that expensive labor causing the industrial revolution, then you don't actually mean anything when you claim to have a model.

Comment by Douglas_Knight on Douglas_Knight's Shortform · 2020-08-22T01:48:32.216Z · LW · GW

There's an extremely common argument that the reason that ancient Greek science didn't lead to Greek engineering is that Athens was a slave society and slave societies are brimming with labor and have no demand for labor-saving devices.

I have never been able to make head or tail of this argument. Also, the conclusion and premise of the argument are false. Conclusion: Greek engineering was better than Roman engineering. It was awesome and we're really not sure how far it went. Premise: Greek scientists weren't in Athens, but in many places in the Hellenistic world, especially Alexandria. Was Alexandria a slave society? I don't think anyone really knows. Some sources claim that it was full of slaves; some empty. Some that it had slaves everywhere, but others only in the fields, not in the workshops.

An alternate theory is that in (some) slave societies, the master is not supposed to think about the kind of work performed by slaves.

Here is a third argument that slave societies are not inventive. I just ran across Carroll Quigley:

a society whose productive system was based on slavery would probably be uninventive, because the slaves, who knew the productive process most intimately, would have little incentive to devise new methods since these would be unlikely to benefit themselves, while the slaveowners would have only a distant acquaintance with the productive processes

This argument seems too narrow to me. Is this about slavery, or about big organizations? I thought Adam Smith wrote something broader, but he actually wrote exactly the same:

A great part of the machines made use of in those manufactures in which labour is most subdivided, were originally the invention of common workmen, who, being each of them employed in some very simple operation, naturally turned their thoughts towards finding out easier and readier methods of performing it. — Book I, Chapter 1
Slaves, however, are very seldom inventive; and all the most important improvements, either in machinery, or in the arrangement and distribution of work, which facilitate and abridge labour have been the discoveries of freemen. Should a slave propose any improvement of this kind, his master would be very apt to consider the proposal as the suggestion of laziness, and of a desire to save his own labour at the master's expense. The poor slave, instead of reward would probably meet with much abuse, perhaps with some punishment. In the manufactures carried on by slaves, therefore, more labour must generally have been employed to execute the same quantity of work, than in those carried on by freemen. — Book IV, Chapter 9

Again, what's special about slaves? Why would the master think any differently of the slave than of the apprentice? Doesn't "at his master's expense" apply both times? Is the master's reaction supposed to be correct or prejudiced? Perhaps the master should be willing to let the underling try out cheap experiments, but should be cautious about investing in building the machine. The master tradesman is different from the slave driver in having more experience in the task, but that's mainly a question of the depth of the hierarchy, not the legal status. My memory was that Smith went on to say that the master was right that the apprentice was crying wolf and that real progress required the apprentice to set up his own shop and try out his inventions at his own expense, with skin in the game. I wouldn't be surprised if he does say something along these lines elsewhere, but not connected to this passage.

Bonus: here are the sentences before and after the second quote. They sound rather odd to me, perhaps like the first theory I rejected above:

the great body of the people were in effect excluded from all the trades which are, now commonly exercised by the lower sort of the inhabitants of towns. Such trades were, at Athens and Rome, all occupied by the slaves of the rich, who exercised them for the benefit of their masters, whose wealth, power, and protection made it almost impossible for a poor freeman to find a market for his work, when it came into competition with that of the slaves of the rich. [Slavery and invention.] The Hungarian mines, it is remarked by Mr. Montesquieu, though not richer, have always been wrought with less expense, and therefore with more profit, than the Turkish mines in their neighbourhood. The Turkish mines are wrought by slaves; and the arms of those slaves are the only machines which the Turks have ever thought of employing. The Hungarian mines are wrought by freemen, who employ a great deal of machinery, by which they facilitate and abridge their own labour.

First, there is the mystery of why the Turks don't copy the Hungarians next door. It's one thing to ignore innovations from slaves, but why do they ignore proven innovations? Second, if the Hungarians can compete with the Turks, why can't free tradesmen of Athens and Rome compete with the slave tradesmen? One possibility is vertical integration ("for the benefit of their masters"). Another is that something outside of economics has been smuggled in with the words "power and protection." It seems like most of the sentences exists to refute the first few, so what's going on?

Comment by Douglas_Knight on Douglas_Knight's Shortform · 2020-08-18T20:11:03.369Z · LW · GW

How much of the Copenhagen Interpretation of Ethics is just feudalism?

That is, that is, taking on a patron-client relationship is a discrete change with abrupt ethical consequences.

I think that feudalism explains a lot about the ethics of employment. In particular, the employer is responsible for a living wage, not supplemented by charity, which is for wards of the state; and the employer is responsible for the sins of the ward. I'm not sure it explains the examples Jai's original post. He did a good job of producing diverse examples that aren't explained by a common factor like employment. But I wonder whether it is an echo of feudalism.

Comment by Douglas_Knight on Matt Botvinick on the spontaneous emergence of learning algorithms · 2020-08-18T18:26:19.968Z · LW · GW

I asked you if you were talking about starving to death and you didn't answer. Does your abstract claim correspond to a concrete claim, or do you just observe that anorexics seem to have a goal and assume that everything must flow from that and the details don't matter? That's a perfectly reasonable claim, but it's a weak claim so I'd like to know if that's what you mean.

Abrupt suicides by anorexics are just as mysterious as suicides by schizophrenics and don't seem to flow from the apparent goal of thinness. Suicide is a good example of something, but I don't think it's useful to attach it to anorexia rather than schizophrenia or bipolar.

Long-term health damage would be a reasonable claim, which I tried to concede in my original comment. I'm not sure I agree with it. I could pose a lot of complaints about it, but I wouldn't. If it's clear that it is the claim, then I think it's clearly a weak claim and that's OK. (As for the objection you propose, I would rather say: lots of people take badly calibrated risks without being labeled insane.)

Comment by Douglas_Knight on Matt Botvinick on the spontaneous emergence of learning algorithms · 2020-08-17T14:59:11.256Z · LW · GW

I'm objecting to the claim that it fits your criterion of "catastrophic." Maybe it's such a clear example, with such a clear goal, that we should sacrifice the criterion of catastrophic, but you keep using that word.

Comment by Douglas_Knight on Douglas_Knight's Shortform · 2020-08-17T02:18:27.923Z · LW · GW
The essence of systems thinking: Every persistent biological or cultural structure exists because of a positive feedback loop. Sometimes it’s hard to see. But to understand the structure, you must understand the loop.
Kevin Simler

Elaboration:

Positive loops by themselves are unstable, yes, and both are needed for stability, but positive loops are primary. Without a positive loop, there would be nothing for a negative loop to stabilize.
Comment by Douglas_Knight on Matt Botvinick on the spontaneous emergence of learning algorithms · 2020-08-17T01:49:25.148Z · LW · GW

Why do you single out anorexia? Do you mean people starving themselves to death? My understanding is that is very rare. Anorexics have a high death rate and some of that is long-term damage from starvation. They also (abruptly) kill themselves at a high rate, comparable to schizophrenics, but why single that out? There's a theory that they have practice with internal conflict, which does seem relevant, but I think that's just a theory, not clear cut at all.

Comment by Douglas_Knight on What a 20-year-lead in military tech might look like · 2020-08-06T02:03:10.564Z · LW · GW

OK, if you change the topic from occupation to conquest, then technology matters. Cortes and Pizarro didn't equip and train their allies. Maybe in the beginning, say, 1680-1750, the East India Company acted like Cortes and Pizarro, allying with armies that they didn't train or equip and maybe their advantages were a reason to side with them. But starting in 1750 they trained and equipped native armies and lost all their advantages.

I explicitly addressed this: the British had no advantage of guns. The Sepoy mutiny was sparked by the beef fat in the cutting-edge rifles that the British gave to the Indians.

For that matter, your wikipedia link says that the Sultan of Zanzibar did have a Maxim gun. Technology travels fast! It was more a battle of artillery. The British Raj did respond to the Sepoy Mutiny by moving artillery out of the hands of natives, but the Company had managed 50-100 years trusting artillery to Indians.

Comment by Douglas_Knight on What a 20-year-lead in military tech might look like · 2020-07-31T19:08:44.870Z · LW · GW

I think phrasing this in terms of "power" is not so helpful. Cortes and Pizarro brought only a few hundred men to destroy empires. But they didn't nuke them; rather, they allied with existing armies. Similarly, most of the Indian Army was native. Most of the administrators were Indians, too. The British had better weapons and social technology of training, but in creating the Indian Army, they gave up that advantage. The key was some tacit social technology of hierarchy, to keep command of army and country. At the beginning, the Empire had 80k British soldiers, but the East India company tried to get away with using only 40k. It failed in the Sepoy Mutiny, which is why it was replaced by the Empire, but the units with more British were more likely to rebel, so it's not a matter of pure numbers, but of how they were deployed. (Specifically, units with more British had segregation and racism, while the units with fewer British had camaraderie. And probably responsiveness to specific concerns, like the tallow bullets.)

Comment by Douglas_Knight on What a 20-year-lead in military tech might look like · 2020-07-31T02:53:54.143Z · LW · GW

The British Raj had 125k colonists at the beginning in 1861 and 166k near the end in 1921, so about one per 2,000 subjects.

Comment by Douglas_Knight on New Paper on Herd Immunity Thresholds · 2020-07-30T21:18:01.887Z · LW · GW
1. People are identical, and have identical susceptibility to the virus.
2. People are identical, and have identical ability to spread the virus.
3. People are identical, and have identical exposure to the virus.
4. People are identical, and have contacts completely at random.
5. The only intervention considered is immunity. No help from behavior adjustments.
All five of these mis­takes are large, and all point in the same di­rec­tion.

I think you are making an error about 5. There are several questions you could ask the SIR model. If you mix them up, you get the wrong answer, but that's not the fault of the model. The SIR model allows non-immunity changes by just changing R. The question of what would herd immunity be without behavior adjustments is a perfectly reasonable question. It is the question of what level of immunity would allow us to go back to normal without risking an outbreak.

Maybe I don't understand what you mean by 2 and 3, but I don't see how they predict systematic deviation from the SIR model, unless the effects in 2 and 3 are correlated. Probably I would just subsume 2 and 3 into 1 and 4.

I see three main deviations from the SIR model. One is natural immunity. Like Owain, I think that this is overplayed, at least in Europe. The second is the network difference you talk about between the connected and the isolated. But the third is the obvious network structure of cities. Talking about whether Italy has achieved herd immunity is an error: Milan can achieve it without protecting Naples. Talking about a national immunity threshold is a category error and using national PCR and antibody numbers is not so useful. (I'm not sure how badly this paper makes this mistake. It does talk about Madrid and Catalonia, but in other countries I think it uses the only data it can.)

Comment by Douglas_Knight on New Paper on Herd Immunity Thresholds · 2020-07-30T19:10:20.115Z · LW · GW

The Imperial model is worse than the SIR model.

It accreted detail for a decade just to prove that they were doing something. It is a good demonstration of the typical failure modes of an agent-based model. A useful model has very few parameters abstract parameters, so that they can be measured from reality. Agent-based models are useful to explore the space of relevant parameters, not to simulate a country. If simulating a country is "sophisticated," then I don't want to be a sophist.

Comment by Douglas_Knight on Douglas_Knight's Shortform · 2020-07-21T20:46:18.157Z · LW · GW

I could have said "The Aumann Agreement Theorem, also known as Bulverism," which is more broadly true. But the converse is a more valuable statement.

Comment by Douglas_Knight on Douglas_Knight's Shortform · 2020-07-20T15:56:47.194Z · LW · GW

Bulverism, also known as the Aumann Agreement Theorem.

Comment by Douglas_Knight on SlateStarCodex deleted because NYT wants to dox Scott · 2020-06-29T18:32:38.213Z · 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 · 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 · 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.