There were several debates FOR and AGAINST the "LIKE" button on SSC/ACX. And when it existed on ACX, it was heavily used by at least a portion of the readership. However, since it's been disabled, I've noticed that it's quite rare that somebody simply states (in text) that they like or appreciate a comment. I'd guess that the ratio of "LIKE" button presses, to a written notes of gratitude is 100:1 (maybe 1000:1). If this is close to the truth, then why is it the case?
It's obvious that pressing a button is more convenient than typing a few words. But does this entirely explain the effect?
Is it the anonymity of the Like button?
Is it simply used by many for voting versus liking?
Is it "Like-light"? Kind of like saying "love youuuuu" versus "I love you."
Do many people lack the confidence in using their words, especially in "talking" with strangers?
I'm guessing it's a combination of the above that creates the massive multiplier. Regardless of the answer, I highly recommend spending the few seconds typing appreciative comments. I'd say that those are at least 100x more appreciated by the writer than a clicked "like".
Did I correctly remember that you said you're open to accepting book reviews outside of the Book Review Contest window? If so, I can't find the post where you mentioned this.
I recall Scott saying self-promotion is allowed on these ones so I'm gonna self-promote. I run a music recommendation website, https://theshfl.com, that I've been building on and off for the last couple of years - it's basically a random sampler built from crawling album reviews and best-of lists. What's cool about it though is that it is steerable - as you're flipping through you can change the population sampled from - when the music was made, who recommended it, genre, label, pretty much anything. I built it because I wanted something like it. Anyway try it out if you have trouble finding something new to listen to, or if you want to get some suggestions for good music from a genre you're unfamiliar with.
Since the meetups have started now, can anyone give me some of their experience as to what generally happens at these and what they get out of it? It sounds like an interesting way to meet peoole but if anyone can give me some more info I would be grateful
I talked to some people about a sort of Pascal's Wager theory of medications. Since most medications have stronger main effects than side effects, if you have a serious illness (like COVID) is it worth taking ten untested drugs that each have only a 10% chance of working? After all, probably at least one of them will work, and treating COVID effectively might be worth inflicting ten drugs worth of mild side effects on yourself.
I looked into this more this morning, and the most relevant paper I found was https://www.nature.com/articles/s41467-021-22446-z on HCQ, an untested drug that seemed to have at least a 10% chance of treating COVID earlier in the pandemic. The study finds HCQ increased mortality, by quite a lot. I was originally pretty surprised by this - HCQ gets used as malaria prophylaxis all the time and nobody acts like it's especially deadly. But I think it makes sense in the context where people who are on the verge of dying from COVID and need to be saved from it are actually really weak, and side effects that healthy people would shrug off can become fatal. I think this pretty clearly sinks the Pascal's Wager theory of medications in favor of the usual thing where you don't try anything until you have strong evidence that it's at least safe and probably also effective.
Also, Bram C brought up the idea that pandemics are deadlier than endemic diseases partly because you usually get endemic diseases as a child, and children get less severe versions of many conditions including COVID. Later I found https://www.nature.com/articles/s41597-020-00668-y , which was able to confirm that this pattern holds for many different conditions. Interested in hearing what knowledgeable commenters think about this.
ACX has discussed Hutchinson-Gilford progeria syndrome and aging before. I was involved in some research investigating the underlying cause of disease (other than progerin protein). Biased of course, but I think it goes a long way in explaining the disease's what-we-call 'aging.'
Nuclear membrane ruptures underlie the vascular pathology in a mouse model of Hutchinson-Gilford progeria syndrome
A fun brain teaser. You have an island, centered in a moat. The island is a 10x10 square, and the exterior of the moat forms a 30x30 square. You have two 9x1x1 boards to use to cross the moat. How do you get across? (all units are measured in feet).
Now, there is a “correct” answer, but there are also fun ones that I’ve heard like “accelerate the moat and island to .99 light speed so that relativity causes the distance across the moat to be shorter.” Feasibility of answers is not necessary, this is just a thought exercise I find fun.
Considering how massive education is as an institution, how deeply invested societies and individuals are in education, and the huge body of research and literature on education, it seems shocking to me that there is such a dearth of words/tools to discuss the QUALITY of education.
What tools do we currently have? We can talk about the outcomes -- discussion of which is often limited to graduation rates and standardized tests. We can talk about inputs -- small classes, higher-paid instructors, charter schools, homeschooling, project-based, etc.
But none of these things tell me very much about the objective quality of how Cathy spends 6-10 hours of every school day. And if we drill down to the quality of Cathy’s biology class versus Billy’s Physics class, things get even murkier. We’re stuck with “good teacher” versus “bad teacher”, and “easy” versus “hard”.
I think there may be a single, missing metric that would facilitate a lot of these conversations and analyses: ENGAGEMENT.
How engaged are the students?
And, if they are highly engaged, WHO CARES ABOUT ANYTHING ELSE?!?!
Let’s assume that a bunch of students in a classroom don’t have access to sugar, drugs, or screens. Nor are they fighting or having sex on top of the classroom desks. And let’s imagine that despite all this they are still highly engaged. Whatever magic is happening in that classroom should translate to better learning.
Finding out if a class is good or not should be as simple as seeing the student engagement score for that class. Finding out if a school is good or not should be based on the amount of high engagement hours that the average kid gets in that school. If I was to tell you that School A (an average Public school) provides 1 hour of high engagement per day, School B (a charter school) provides 1.5 hours, and School C (SSC Academy) provides 4 hours of high engagement, would you need any other metric to make a decision?
In honor of Bezos, this is the only metric that focuses on the customer. The customer is the one who is doing the learning. Schools should be obsessed with the quality of the customer experience (quality of engagement), because more engagement equals more learning. We should all be.
If this metric already exists in some form, and I’m missing something, please point it out.
Oh… and how do we measure ENGAGEMENT? Well, we can observe the students (automated tools are getting better at this sort of thing). Or we can ask them. Or we can do some combination of both.
Can somebody point me at any previous intro or basic discussion of prediction markets? I'm struggling to see what the fascination with them is.
To me, they seem to be about as useful predictors of the future as Top Hits or blockbuster lists are as arbiters of taste. Sure, people have to think about some future event and assess its likelihood by putting their money where their mouth is. But since when does having money behind something make it true?
Or maybe they're sources of interesting ideas. Maybe you're curious what questions other folks find interesting but you don't want to be inundated with conspiracy theories and other noise; prediction markets function to add cost, which filters out batshit propositions. Okay, I guess, but how robust is this dynamic? For something to show up as a bet at all, someone has to sufficiently believe that the prevailing opinion is wrong. If the odds are extremely unbalanced, we're dealing with people who think the sun won't come up tomorrow, and it's not very interesting. If the odds are too balanced, then it's a horse race and also not very interesting. This seems a bit... distasteful to me. "Sorry, your idea is neither popular enough nor niche enough for me to care about it".
Regrettably the invite to Winnipeg arrived with too little notice for me to make the 850-mile drive from Chicago. It did at least enhance my morning with a delightful touch of AI weirdness, no harm no foul.
I'm now very curious why, out of the dozens of events listed in that email, Google Calendar decided the Winnipeg event and *only* that event warranted an auto-created entry.
I can't think of any special connections I might have to Winnipeg, real or Google-imagined, and any of the events in Ontario would have been closer by hundreds of miles. Is there something special about the event description? I guess it's the only event that that it doesn't have a "Coordinates" item, and the only one before Europe that mentions a restaurant instead of a park.
There is a lot of covid-related school closure/learning loss literature out there, as well as on broader effects of school closures. I'm posting this one because Hanushek is a big deal in the area of economics of education and because it has a brief section on previous school closures or disruptions which comes to a different conclusion than the ACX piece. https://www.oecd.org/education/The-economic-impacts-of-coronavirus-covid-19-learning-losses.pdf
Oh, so that's why I got a message saying I had a restaurant reservation in Winnipeg. I did wonder about that, but seeing it was for, like, right then, I elected to ignore it rather than try to figure out how to cancel it.
How would you run basic tests for whether a nuclear power plant is leaking radiation, or causing health damage in some other way (assuming you don't trust the government to be honest about it)?
Walking around the area with a Geiger counter seems like the obvious one. Testing the tap water in the nearby town for radiation also seems required (looks like there's a bunch of home water radiation testing kits available online - does anyone know how reliable it scalable they are?) Anything else I'm missing?
On a completely unrelated note, there's rumors that the Israeli nuclear/textile plant in dimona leaks radiation, causing high rates of cancer in dimona. Does anyone know any trustworthy data on this? Is this even a plausible thing to be worried about?
So here's my problem. My wife has turned anti-vaccine and is refusing to allow our kids to get vaccinated if children are recommended the vaccine (as looks like will happen where we live). What does one do in this situation if (like me) you think the vaccines are the best defence against whatever horrible variant of COVID is currently evolving somewhere?
For background, I always knew that my wife's family were anti-vaccine, but this is the first time it's been imported into our house. I've obviously tried rational argument, persuasion, appeals to authority and the rest––she's not budging on the "vaccines are untested and will make our kids infertile (or whatever)" line. This is unsupported by any evidence beyond anti-vaxx propaganda from wackadoodle world.
This specific problem is about vaccines, but I should also say that I'm finding it very hard not to read my wife's thoughts on the vaccine as a verdict on something bigger.
Is it reasonable to think persons are repeatable? Meaning if you measured every aspect of who I am down to the atoms and then recreated me 10,000 years later, would it be "me" in the sense that I experience life through the lens of my future self? Would I be conscious in that later recreated body?
To think that we could not would indicate something non-material that can't be reconstructed such as a soul. However, until this point we don't have reason to think people can be recreated. We just know that the you of today feels like the you of yesterday. Do we truly need this spacial and temporal continuity for it to be you? If you say yes, is that view consistent with a totally materialist perspective?
In summary, “Four IC elements and the National Intelligence Council assess with low confidence that the initial SARS-CoV-2 infection was most likely caused by natural exposure to an animal infected with it or a close progenitor virus”. However, “one IC element assesses with moderate confidence that the first human infection with SARS-CoV-2 most likely was the result of a laboratory-associated incident”. Three IC elements couldn’t come to a conclusion. Importantly, the IC as a whole agreed that China did not have foreknowledge of the virus before the initial COVID-19 outbreak emerged.
Before the review, only two agencies favored the natural origin theory, according to NYTimes. So whatever secret information the IC has access to that we don’t, it pushed two (three?) of the agencies toward the natural origin theory.
It’s fairly short (9 pages of text) and relatively non-technical by scientific paper standards. To summarize the arguments, (1) lots of zoonotic coronavirus have infected humans; (2) the earliest SARS-CoV-2 cases were clustered around the Huanan wet market, as expected for a zoonotic origin; (3) the WIV was not known to be working on any virus very similar to SARS-CoV-2 (not even as similar as RaTG13), and there would have been no reason to hide such work before the pandemic; (4) more recent common ancestors than RaTG13 have been found in the wild since the pandemic began; (5) past recombinant virus research at WIV has used the WIV1 backbone, which is unrelated to SARS-CoV-2; (6) the furin cleavage site is present in a lot of coronaviruses, and the one in SARS-CoV-2 is suboptimal.
Some interesting facts I learned:
1. “In direct parallel to SARS-CoV-2, HCoV-HKU1, which was first described in a large Chinese city (Shenzhen, Guangdong) in the winter of 2004, has an unknown animal origin, contains a furin cleavage site in its spike protein, and was originally identified in a case of human pneumonia (Woo et al., 2005).”
2. “bat virus RaTG13 from the WIV has reportedly never been isolated nor cultured and only exists as a nucleotide sequence assembled from short sequencing reads”
3. Although RaTG13 has the highest genetic similarity to SARS-CoV-2, a history of recombination means that 3 other viruses have a more recent common ancestor with SARS-CoV-2. These are RmYN02, RpYN06, and PrC31, which were all sequenced after the pandemic had begun. So SARS-CoV-2 was not derived directly from RaTG13 (which we knew already), and scientists are making progress in tracking down the natural reservoir of SARS-CoV-2 (which I didn’t know before).
4. Page 31, showing the geographic distribution of the earliest cases. “Two of the three earliest documented COVID-19 cases were directly linked to this market selling wild animals, as were 28% of all cases reported in December 2019 (WHO, 2021). Overall, 55% of cases during December 2019 had an exposure to either the Huanan or other markets in Wuhan, with these cases more prevalent in the first half of that month (WHO, 2021)”.
This brings up an important philosophical issue: how much weight do you give to geographic coincidences? If you give them a lot of weight, and you believe it’s improbable for a zoonotic pandemic to have begun only 7.4 miles away from the Wuhan Virology Institute, then it should be downright impossible for a non-zoonotic pandemic to have begun 0 miles from a wet market and for a large fraction of the earliest cases to have traceable ties to the wet market. If you think the lack of traceable links between 72% of the December 2019 cases and the wet market is evidence the virus didn’t come from the wet market, then the lack of traceable links between 100% of the December 2019 cases and the WIV should be conclusive evidence that it didn’t come from the WIV. On the other hand, if you don’t put much weight on geographic coincidences, you could argue the wet market link means little because wet markets are crowded places with lots of people where viruses tend to spread easily--but then you’d have to agree that the coincidence that the WIV is in Wuhan means even less, because there’s no traceable link between the earliest cases and the WIV at all.
to people at the meetup: what does Scott's voice sound like? in my head, just because he writes in what I guess I'd term a "thoughtful style", my mental conception of his voice ends up assigning it a pleasant timbre, somewhat quiet and clipped. But as we all know, this is often a terrible assumption to make
Do you ever feign ignorance when debating with someone? I've found myself trying it.
The situation is: you want to say X. You know that saying X is going to make some people you're talking to angry, because X sounds a bit like Y. Y is something that "the other side" say. You think Y is false, and so do your friends. But Y is so toxic, if you say anything associated with it, like X, Y will come to mind. Examples in this blog post: https://www.overcomingbias.com/2008/06/against-disclai.html
I used to clarify, disclaim, and preface a lot: "I'm about to say something that sounds like Y. We all know Y is false and bad. I believe X - that may sound like Y, but it's different because Z".
I've changed my mind; I think the best approach is to say X and let the other person bring up its connection to Y.
If you're the one who brings up the connection, from their point of view, this is a partial admission of a connection.
If they're the one who has to bring up the connection, they're offered the option of not bringing it up, and trying to attack X on its own terms. Offering that option allows them to think about how they might attack X if it wasn't related to Y. Their interest in doing that might entice them away from dwelling on the possible Y connection.
This comes partly from hearing about how to react when you've been cancelled (clarifying is taken as a sign of concession), and partly from Robin Hanson, who wrote the blog post a lot, and appears to live this himself - watch this remarked on amusingly 1m5s into this video https://www.youtube.com/watch?v=kBRwDYNjgog
In American culture, is there a stigma associated with packing-up uneaten food, following an office function? In an office of about 100 staff (we all share the same break room), I am among the few people who will take home uneaten food following birthday celebrations, or retirement events, or whatever. Most people seem content to just leave the remaining food in there, where it will be thrown out within a day. Many of my co-workers earn minimum wage, or just above it, so you'd think that they would have an incentive to grab free food. I find it all a bit odd.
Couldn't Reaper drones be reconfigured to take on light infantry/insurgents, like in Afghanistan? Why haven't they been, to date? I know when most people think of drones, they're thinking of the relatively small ones- and those little guys seem to have be the focus of most warfighting advances recently (as small/cheap bombers). But the America Reaper drone, and the Predator before it, is a 5000 lbs. beast. They're already loaded with missiles- why not the machine gun that goes on an attack helicopter now?
I understand that Reapers now fly at a much higher altitude. But couldn't the US reconfigure one with a mounted machine gun to hover 50ish feet off the ground and take out Taliban insurgents? Seems like a good way to route that sort of light infantry (and strike absolute terror into their hearts), while not risking American lives. If their range is too limited to be effective (hard to believe), a mothership cargo plane could circle over the battlespace and drop them off, then pick them up. Obviously this would be difficult/impossible in a war against Russia or China- but the US does a ton of these counterinsurgency campaigns against very unsophisticated foes all over the world, who lack anti-aircraft weapons.
Why hasn't this been done to date? Is it that the Reaper is vulnerable to small arms fire? I don't believe the Taliban has effective artillery or anti-aircraft weapons- is an AK gonna take the drone down? If so, is it just impossible to armor up a drone enough while still keeping it flyable? Or, is it that that level of firing precision isn't possible with a remote control? (Maybe operators could be in the mothership cargo plane, if latency is an issue).
Seems like a hardened Reaper drone with machine guns, that can't be taken down by small arms fire, would revolutionize remote US wars where we don't want to risk ground troops. Curious if it's technically possible. I am especially hoping to hear from John Schilling, of course
I was thinking about the debate regarding whether the obesity crisis has been caused by psychological factors like food reward / palatability, vs being caused by environmental contamination, which is being researched by Slime Mold Time Mold. One of the questions in my mind is how obesity set points come to be and how they become so stable, which led me to the following idea.
If your obesity set point is set by your psychology, would it make sense to use an extreme diet to get down to your desired set point, then take LSD to reset your neural bases, in the same way that Scott has written about before? Is there any possibility that would work?
Since the discussion of aducanumab put me at odds with most of the other commenters, I wanted to make my position more concise and offer predictions/betting opportunities:
1) I did not know that FDA approval may mean forcing Medicare to pay for it. I hope that that didn't happen and if it does, I'm against the approval. I am for allowing doctors to decide based on the available test data whether a drug with nice biomarker action but crappy endpoint evidence has a high enough chance of working on their patient to warrant the cost.
2) I am not so much a fan of aducanumab itself as I am more optimistic than others about the amyloids as a therapeutic target in general. I predict the following, and am willing to bet on these odds:
a) >50% chance of an amyloid targeting drug or a mix containing one showing success on cognitive tests in phase III trials within the next 4 years.
b) >75% within the next 7 years.
c) >30% chance of aducanumab specifically showing success on cognitive tests in some treatment modality within the next 3 years.
d) Conditional on (c) happening, >75% that that treatment modality will involve targeting to early stage or pro-dromic patients, or a combination with a drug against a different target.
Ezra Kline had an interesting conversation with Bessel van der Kolk about his book on trauma, "The Body Keeps Score." During the talk van Der Kolk made a statement that really tickled me. "Tango dancing is probably more effective treating trauma than CBT."
What books would people recommend I read to get a basic grasp on modern science the history of science? I’d like the books to be engaging to a layperson and not a something I need to slog through. Is Steven Weinburg’s “To explain the world” worth buying ?
So I try to stir fry Kung Pao Chicken at home, and it never quite comes out right. I thought maybe it was the heat, so I tried it over a friend's gas grill at max heat to see if that worked - and it still didn't come out right. Is there something to it that most recipes miss? Szechuan peppercorns?
I am getting my first COVID shot tomorrow in anticipation of a plenary shift away from remote work. ACX readers, which vaccine should I get?
I'm leaning towards J&J. I feel ashamed to even raise this point, but I waited so long out fear of unusually high reports of adverse reactions in VAERS. I know this data is not clinically verified and may not be helpful, but I have had a nagging paranoia nonetheless. I trust this community more than I would any other online group, so please feel free to critique my life choices or offer helpful explanations for your answers. At any rate, I will definitely be vaccinated. I am just trying to avoid adverse reactions or long-term effects. I prefer the J&J because I already had 3 viral vector shots for HPV, and to the best of my knowledge there are no long-term effects of adenovirus-based shots.
Is there some kind of good legal/public policy reason why America can't make people save for retirement via a 401k? Like, everyone with an existing 401k registers it with the IRS- those without such a registration, the IRS auto-deducts say 3% from their paycheck (so 2ish% after taxes). Yes, it probably should be more, just trying to be politically realistic. The private 401k is some Vanguard target date fund or mutual fund, with ultra-low fees. If people really feel strongly about it, they can fill out a form and the IRS will stop deducting the retirement monies, and send them their funds back (maybe they can try again in a decade). Yes, it really should be mandatory, but again trying to be politically realistic ('big gubmint can't make me save for no retirement', etc.)
Why hasn't the US done this to date? Social Security (which this is not attempting to replace) is clearly not sending retirees enough money to actually live on. It could be argued this will actually help save SS. The US generally goes for market-oriented solutions way more than any other developed country, and this is not a government-run pension or anything. Tons & tons & tons of people are not saving for retirement, or don't start young enough, and here is a reasonable market-based nudge that's opt-out vs. opt-in, and will probably save the US government money in the very long run. Conservatives should love it, in that there's lots of evidence that having assets makes people more fiscally conservative. Liberals should love that it's reducing inequality, which is mostly about asset prices. I'm sure Vanguard or whomever would love it. Best of all, it doesn't cost the government much of anything that I can see- maybe some administration costs, but not a mondo amount. Why haven't we done this? Seems like virtually free money.
(I'm assuming we can discuss dry public policy like opt-out retirement savings in the 'no politics' thread. If this discussion is considered political, I will delete this comment, apologize, and also commit sepuku)
Weird experience on the Winnipeg reservation: I didn't see that until yesterday, when it popped up on my smartphone. Which I have never used to browse ACX, and I'm pretty sure I haven't used it to read my ACX-registered email account since the Winnpeg post went out.
I attended a casual meet up of modest size recently and exchanged contact info. with a few other attendees. I suggested (individually) that we get lunch, and all seemed amenable. However, none of them have yet responded to my open-ended follow up text.
Am I doing something wrong here? I would have assumed that 1) People are generally friendly and willing to get lunch and 2) Would respond to communications from someone they personally know, that take less than 2 minutes to do so, within a day or so. At least, people in my office at varying levels of rank/seniority have usually been willing to get lunch with me. I wouldn't know if I was, but I don't *think* I was so rude/awkward/annoying that the other attendees actively wouldn't want to interact with me again. (My general approach to conversation is to limit expressing my own opinions and try to ask the other party about their interests and background. I think this is usually pretty effective---although, again, I wouldn't know if it wasn't.)
It's a small sample size, I'm certainly overthinking it, and I suppose it's not impossible that one or two will eventually get back to me. Nonetheless, I find this discouraging. Going to gatherings, exchanging contact info. with a few people you get along with, and following up with lunch seemed like a potentially very effective way to make friends/connections to me. But, without that last step, it seems like the value of going to the gatherings at all is substantially diminished.
Discussion on the topic has been closed for the past 2 years. Starting Wednesday, September 1, discussion will reopen. I hope you all will join me in correcting this most abominable stain on Wikipedia by arguing in favor of giving him his damn infobox.
Is anyone travelling to the ACX meetup in NYC from Westchester? I will be driving down from Yonkers with my wife, and it'd be nice to have someone to split (presumably) eye watering parking fees with.
TL;DR: It's not an information gap: telling group members about an issue affecting their peers changes nothing about their attitude towards it -- if the issue wasn't one they already cared about.
This is not strictly speaking about politics, so I think it's ok for this thread, but if you disagree feel free to flag/delete it.
Disclosure: the author of both the article and the paper is my wife.
Quick question that must have been covered but I can't find it.
The Vitamin D post a while back ( https://astralcodexten.substack.com/p/covidvitamin-d-much-more-than-you ) had the structure of "generally studies says Vitamin D has no effect; take it anyway since its the safe default". Clicking on the studies I found they seemed to be using p-value hypothesis testing.
So in p-value hypothesis testing you have your Alternative and Null statements. The model sees these in terms of "what damage is done if this statement is wrong, but I accept it" -- the Alternative statement is the statement that does more damage if you wrongly accept it; the Null statement is the one that does less damage if you wrongly accept it.
For this reason, the model treats the Null hypothesis as "am I allowed to accept this?" and the Alternative hypothesis is "am I forced to accept this?". The p-value itself is the probability of you wrongly accepting the Alternative hypothesis (hence why lower is better, and at p=0 you just never accept the Alternative hypothesis).
So, the way people /should/ choose their p value, Null hypothesis and Alternative statement is:
1). The Null statement is the safe default; the alternative one is the one you really need to be sure about before accepting (it's like "guilty until proven innocent", because we'd rather let guilty people be free than innocent people be locked up);
2). You should pick your p-value based on the relative danger of incorrectly asserting the Alternative statement; for less dangerous or less costly Alternative statements, higher p-values are sensible; for more dangerous or costly statements, lower p-values are prudent.
However, in every actual paper I've seen, they instead choose as follows:
a). The Alternative statement gives novel information.
b). A set of gold-standard low p-values. As low as is practical is always assumed to be better.
In the Vitamin D article, Scott applied (1) post-hoc by saying it's worth taking vitamin D anyway; correcting their error in those studies that the statements were inverted (they all should have accepted the assumption that Vitamin D is helpful). To be clear: in these studies, taking vitamin D is helpful should have been the Null statement based on (1). Then the hypothesis test setup matches common sense.
And weird concepts like "tending toward significance" I see as post-hoc attempts to apply (2).
But it's very frustrating because we shouldn't need to post-hoc mitigate the substitution of (a) for (1) and (b) for 2: just set up the parameters correctly in the first place.
As an aside, I think it's interesting that (a) is substituted for (1), as if the MOST dangerous thing isn't lives lost, it's us believing something that isn't true. This is really interesting to me and I think is the way a lot of people handle these things.
I have an idea re: Mandarin tone marks, which are often omitted in normal English transcription of Chinese words. They should be placed as separate characters next to the syllable (example: Xiˊ JinˋPingˊ), similar to Zhuyin, rather than diacritics over the syllable nucleus, as they currently are in Hànyǔ Pīnyīn. This would end both the practice of Gwoyeu Romatzyh type romanization (e.g., "Shaanxi") and end the confusing practice of placing an apostrophe to distinguish two syllables in a single word (e.g., "Xi'an"). I will write a post elaborating on this. Does anyone think this is a good idea? Thoughts?
Hello ACX readers! I recently got a job with the US government for which the most important part of training is learning a new language. I now have to rank my preferences for which languages I'd prefer, but I'm really torn! I know ACX readers tend to be good at making predictions, so I was wondering, of the following languages which do you see as being the most relevant to know in the context of national defense within the next 5 years. Also which, do you predict, will have the most interesting missions to work on?
-Arabic
-Pashto
-Russian
-Persian-Farsi
-Chinese-Mandarin
(I know I'm not providing much information about the job to go off of, I'm just looking for general thoughts)
If people learn to control the DNA of their embryos and we get millions of babies at the level of von Neumann and Tao what do you think life will be like?
Will they get fed up with dumb old farts taking up all the space and do something nasty to the unedited?
I'm hosting a meet up (Buffalo NY) Someone who is not vaccinated wants to come. (He is the only person who has emailed me and said he wants to come. There is one other 'maybe' RSVP.) I'm fine with non-vaccinated people coming, but I don't think this is fair to anyone else who may show up. I'm looking for ideas.
A puzzle adapted from the youtube channel of Michael Penn, one of my favourite maths youtubers. I like it because the solution turns out to depend on something I'd never have guessed was relevant.
n baby chicks sit peacefully in a circle. Suddenly, each chick turns either to left or right (equal chance of each, chosen independently) and viciously pecks one of the chicks sitting next to it.
What is the probability that precisely k chicks are left unpecked?
One of the other posts in this thread had led me to think of an interesting ethical dilemma. Some people have problems with the idea of cloning human beings. One major ethical concern is consent. If von Neumann never provided consent to be cloned, how ethical is it to make a million copies of him. There are other ethical concerns but I want to focus on this particular ones.
There is a concept of implicit consent if you are unable to ask. For example, if I am rushed to a hospital unconscious, they can do what they need to do to help me because it is reasonable to assume that I want to be helped. In a world in which no one wants medical care, this assumption wouldn't be warranted.
If we want to discover whether or not it is ethical to clone von Neumann a million times from a consent perspective, I would argue we should start with 100. If 100a von Neumann's all agree that future cloning is acceptable, then that would be a good reason to think the original von Neumann would think it's acceptable. There would be a number of biases. A major one being that a cloned von Neumann would be biased in favor of cloning because he would not exist otherwise. But if you had a sufficiently large number of von Neumann's, you could probably largely overcome this bias.
A clone of you consenting is not sufficient evidence to think that you would consent but 100 for 100 clones of you consenting is a really good reason to assume implied consent. If the clones didn't agree in large number, then there might be an ethical concern about consent. Obviously this probably has some major complications but it's just an interesting thought I had.
The EU is recommending that member nations block nonessential travel from the USA, per https://www.bbc.co.uk/news/world-us-canada-58386967. I'm afraid that may impact Scott's plan to join ACX meetups across Europe, depending on how the various member states implement the recommendation.
Separately, the UK announced its own changes yesterday, and I don't think they affect US visitors.
I'm figuring out the taxonomy of US healthcare plans, specifically HMO/PPO/EPO/POS. Of course the names are clearly chosen at random and most explanations online are a bit retarded, but looks like at the end of the day it mostly amounts to a 2*2 matrix - out-of-network providers are partially covered or not, specialist appointment requires a physician referral or not.
There are discussions of boosters in the US. Right now, the public discourse is muted because the news cycle has run its course a bit, but I want to think through this a bit.
Most of us in the US have had two shots of an mRNA vaccine (typically Pfizer, atypically Moderna) 6 weeks apart. Some have had the single-shot J&J vaccine (which is an Adenoviral/viral-vector vaccine similar to AstraZeneca popular in Europe and elsewhere).
There seems to be general consensus that the mRNA vaccine efficacy starts waning around the 8 (or 7 or 6) month marker. Are there studies that show this? If so, does the efficacy wane somewhat linearly and if 8 mos represents critical-failure level, does 5 months represent 62.5% emptiness or something like that? How does this work?
There is some evidence that mixing vaccines is beneficial, for e.g., https://science.sciencemag.org/content/372/6549/1392.full. This paper is predicated on previous disease, but I've seen other papers that also suggest that mixing vaccines is beneficial. Overall, I get the sense that boosting a viral-vector vaccine with an mRNA vaccine is a clearly beneficial path. What about when one starts with a 2-dose mRNA vaccine - would they be better off with a 3rd dose of an mRNA vaccine, or would they be better off with a viral-vector dose? Are there studies that are suggestive of one approach or the other?
I visited The Netherlands from North America this August, probably one of a tiny minority to do so. To my complete astonishment, no-one was wearing any masks. In fact, I had trouble finding any kind of covid measure!
It seems unlikely to me that the Dutch population is ignorant and/or brainwashed. The Netherlands in consistently near the top in human development, freedom, quality of life, etc type indexes. They’re not exactly part of the developing world. Also, they have vaxxed a much greater share of their population than the US has.
I’ve heard the same thing is going on in Scandinavia? Perhaps a poster from Holland/Europe could explain the current Covid strategy?
How can I get the most money for some Sterling silver flatware I want to sell? I have several old forks made by "R. Wallace & Sons." A jeweler looked at photos of the pieces that I emailed to him, and he said they had no collectors value, and instead would be melted down. He said the whole collection was worth a few hundred dollars.
Instead of selling the forks to a local jeweler, should I "cut out the middle man" and go directly to a metal smelting plant? Wouldn't I get more money that way? How do such deals work, and where would I go in the Mid-Atlantic to do this?
I am posting this several places because the group of people I'm looking for likely isn't very numerous but...
I'm doing a home remodel and I want to hide some nods to the video game The Witness in a few places. The problem is I'm not very good with visual design. I'm looking for an artist who is familiar with the concepts behind The Witness (at least 2 layers deep, if you've seen that deep then you'll know what I mean) and is open for commissions. I'm looking to pay money for them to brainstorm some design ideas and come up with some rough sketches for 3-8 easter eggs, including a custom glass/iron front door panel (found a company that can do full custom for a good price), and some designs for cabinets.
If you are this person, know someone who might be this person, or have a good idea at all of where to look to find this person, please help me out here.
Looking for housing in the Bay Area, I'm not picky at all on room size/space (lived in Japan) but am hoping for a place that is relatively clean and not too old. I don't mind sharing the space with others and am hoping to meet cool people!
There were several debates FOR and AGAINST the "LIKE" button on SSC/ACX. And when it existed on ACX, it was heavily used by at least a portion of the readership. However, since it's been disabled, I've noticed that it's quite rare that somebody simply states (in text) that they like or appreciate a comment. I'd guess that the ratio of "LIKE" button presses, to a written notes of gratitude is 100:1 (maybe 1000:1). If this is close to the truth, then why is it the case?
It's obvious that pressing a button is more convenient than typing a few words. But does this entirely explain the effect?
Is it the anonymity of the Like button?
Is it simply used by many for voting versus liking?
Is it "Like-light"? Kind of like saying "love youuuuu" versus "I love you."
Do many people lack the confidence in using their words, especially in "talking" with strangers?
I'm guessing it's a combination of the above that creates the massive multiplier. Regardless of the answer, I highly recommend spending the few seconds typing appreciative comments. I'd say that those are at least 100x more appreciated by the writer than a clicked "like".
Did I correctly remember that you said you're open to accepting book reviews outside of the Book Review Contest window? If so, I can't find the post where you mentioned this.
I recall Scott saying self-promotion is allowed on these ones so I'm gonna self-promote. I run a music recommendation website, https://theshfl.com, that I've been building on and off for the last couple of years - it's basically a random sampler built from crawling album reviews and best-of lists. What's cool about it though is that it is steerable - as you're flipping through you can change the population sampled from - when the music was made, who recommended it, genre, label, pretty much anything. I built it because I wanted something like it. Anyway try it out if you have trouble finding something new to listen to, or if you want to get some suggestions for good music from a genre you're unfamiliar with.
Since the meetups have started now, can anyone give me some of their experience as to what generally happens at these and what they get out of it? It sounds like an interesting way to meet peoole but if anyone can give me some more info I would be grateful
Updates to some conversations yesterday:
I talked to some people about a sort of Pascal's Wager theory of medications. Since most medications have stronger main effects than side effects, if you have a serious illness (like COVID) is it worth taking ten untested drugs that each have only a 10% chance of working? After all, probably at least one of them will work, and treating COVID effectively might be worth inflicting ten drugs worth of mild side effects on yourself.
I looked into this more this morning, and the most relevant paper I found was https://www.nature.com/articles/s41467-021-22446-z on HCQ, an untested drug that seemed to have at least a 10% chance of treating COVID earlier in the pandemic. The study finds HCQ increased mortality, by quite a lot. I was originally pretty surprised by this - HCQ gets used as malaria prophylaxis all the time and nobody acts like it's especially deadly. But I think it makes sense in the context where people who are on the verge of dying from COVID and need to be saved from it are actually really weak, and side effects that healthy people would shrug off can become fatal. I think this pretty clearly sinks the Pascal's Wager theory of medications in favor of the usual thing where you don't try anything until you have strong evidence that it's at least safe and probably also effective.
Also, Bram C brought up the idea that pandemics are deadlier than endemic diseases partly because you usually get endemic diseases as a child, and children get less severe versions of many conditions including COVID. Later I found https://www.nature.com/articles/s41597-020-00668-y , which was able to confirm that this pattern holds for many different conditions. Interested in hearing what knowledgeable commenters think about this.
ACX has discussed Hutchinson-Gilford progeria syndrome and aging before. I was involved in some research investigating the underlying cause of disease (other than progerin protein). Biased of course, but I think it goes a long way in explaining the disease's what-we-call 'aging.'
Nuclear membrane ruptures underlie the vascular pathology in a mouse model of Hutchinson-Gilford progeria syndrome
https://insight.jci.org/articles/view/151515
A fun brain teaser. You have an island, centered in a moat. The island is a 10x10 square, and the exterior of the moat forms a 30x30 square. You have two 9x1x1 boards to use to cross the moat. How do you get across? (all units are measured in feet).
Now, there is a “correct” answer, but there are also fun ones that I’ve heard like “accelerate the moat and island to .99 light speed so that relativity causes the distance across the moat to be shorter.” Feasibility of answers is not necessary, this is just a thought exercise I find fun.
I'm glad I now know where that weird even in my calendar came from.
I would love to see someone first steelman bitcoin maximalism, and then explain why it’s wrong.
I’ve looked at every argument I can find against bitcoin, and very few of them see to try anything like a steel-man approach.
Considering how massive education is as an institution, how deeply invested societies and individuals are in education, and the huge body of research and literature on education, it seems shocking to me that there is such a dearth of words/tools to discuss the QUALITY of education.
What tools do we currently have? We can talk about the outcomes -- discussion of which is often limited to graduation rates and standardized tests. We can talk about inputs -- small classes, higher-paid instructors, charter schools, homeschooling, project-based, etc.
But none of these things tell me very much about the objective quality of how Cathy spends 6-10 hours of every school day. And if we drill down to the quality of Cathy’s biology class versus Billy’s Physics class, things get even murkier. We’re stuck with “good teacher” versus “bad teacher”, and “easy” versus “hard”.
I think there may be a single, missing metric that would facilitate a lot of these conversations and analyses: ENGAGEMENT.
How engaged are the students?
And, if they are highly engaged, WHO CARES ABOUT ANYTHING ELSE?!?!
Let’s assume that a bunch of students in a classroom don’t have access to sugar, drugs, or screens. Nor are they fighting or having sex on top of the classroom desks. And let’s imagine that despite all this they are still highly engaged. Whatever magic is happening in that classroom should translate to better learning.
Finding out if a class is good or not should be as simple as seeing the student engagement score for that class. Finding out if a school is good or not should be based on the amount of high engagement hours that the average kid gets in that school. If I was to tell you that School A (an average Public school) provides 1 hour of high engagement per day, School B (a charter school) provides 1.5 hours, and School C (SSC Academy) provides 4 hours of high engagement, would you need any other metric to make a decision?
In honor of Bezos, this is the only metric that focuses on the customer. The customer is the one who is doing the learning. Schools should be obsessed with the quality of the customer experience (quality of engagement), because more engagement equals more learning. We should all be.
If this metric already exists in some form, and I’m missing something, please point it out.
Oh… and how do we measure ENGAGEMENT? Well, we can observe the students (automated tools are getting better at this sort of thing). Or we can ask them. Or we can do some combination of both.
I just hope that if anyone accidentally went to Winnipeg, it was in the summer. Lovely in the summer ...
I’ve written 100 tweets about Richard Rorty https://mobile.twitter.com/ZoharAtkins/status/1430897432699629572
Can somebody point me at any previous intro or basic discussion of prediction markets? I'm struggling to see what the fascination with them is.
To me, they seem to be about as useful predictors of the future as Top Hits or blockbuster lists are as arbiters of taste. Sure, people have to think about some future event and assess its likelihood by putting their money where their mouth is. But since when does having money behind something make it true?
Or maybe they're sources of interesting ideas. Maybe you're curious what questions other folks find interesting but you don't want to be inundated with conspiracy theories and other noise; prediction markets function to add cost, which filters out batshit propositions. Okay, I guess, but how robust is this dynamic? For something to show up as a bet at all, someone has to sufficiently believe that the prevailing opinion is wrong. If the odds are extremely unbalanced, we're dealing with people who think the sun won't come up tomorrow, and it's not very interesting. If the odds are too balanced, then it's a horse race and also not very interesting. This seems a bit... distasteful to me. "Sorry, your idea is neither popular enough nor niche enough for me to care about it".
Regrettably the invite to Winnipeg arrived with too little notice for me to make the 850-mile drive from Chicago. It did at least enhance my morning with a delightful touch of AI weirdness, no harm no foul.
I'm now very curious why, out of the dozens of events listed in that email, Google Calendar decided the Winnipeg event and *only* that event warranted an auto-created entry.
I can't think of any special connections I might have to Winnipeg, real or Google-imagined, and any of the events in Ontario would have been closer by hundreds of miles. Is there something special about the event description? I guess it's the only event that that it doesn't have a "Coordinates" item, and the only one before Europe that mentions a restaurant instead of a park.
There is a lot of covid-related school closure/learning loss literature out there, as well as on broader effects of school closures. I'm posting this one because Hanushek is a big deal in the area of economics of education and because it has a brief section on previous school closures or disruptions which comes to a different conclusion than the ACX piece. https://www.oecd.org/education/The-economic-impacts-of-coronavirus-covid-19-learning-losses.pdf
Oh, so that's why I got a message saying I had a restaurant reservation in Winnipeg. I did wonder about that, but seeing it was for, like, right then, I elected to ignore it rather than try to figure out how to cancel it.
I intermittently get likes on ACX comments. How?
How would you run basic tests for whether a nuclear power plant is leaking radiation, or causing health damage in some other way (assuming you don't trust the government to be honest about it)?
Walking around the area with a Geiger counter seems like the obvious one. Testing the tap water in the nearby town for radiation also seems required (looks like there's a bunch of home water radiation testing kits available online - does anyone know how reliable it scalable they are?) Anything else I'm missing?
On a completely unrelated note, there's rumors that the Israeli nuclear/textile plant in dimona leaks radiation, causing high rates of cancer in dimona. Does anyone know any trustworthy data on this? Is this even a plausible thing to be worried about?
So here's my problem. My wife has turned anti-vaccine and is refusing to allow our kids to get vaccinated if children are recommended the vaccine (as looks like will happen where we live). What does one do in this situation if (like me) you think the vaccines are the best defence against whatever horrible variant of COVID is currently evolving somewhere?
For background, I always knew that my wife's family were anti-vaccine, but this is the first time it's been imported into our house. I've obviously tried rational argument, persuasion, appeals to authority and the rest––she's not budging on the "vaccines are untested and will make our kids infertile (or whatever)" line. This is unsupported by any evidence beyond anti-vaxx propaganda from wackadoodle world.
This specific problem is about vaccines, but I should also say that I'm finding it very hard not to read my wife's thoughts on the vaccine as a verdict on something bigger.
Is it reasonable to think persons are repeatable? Meaning if you measured every aspect of who I am down to the atoms and then recreated me 10,000 years later, would it be "me" in the sense that I experience life through the lens of my future self? Would I be conscious in that later recreated body?
To think that we could not would indicate something non-material that can't be reconstructed such as a soul. However, until this point we don't have reason to think people can be recreated. We just know that the you of today feels like the you of yesterday. Do we truly need this spacial and temporal continuity for it to be you? If you say yes, is that view consistent with a totally materialist perspective?
A few weeks ago I examined the lab leak hypothesis for COVID-19 (https://astralcodexten.substack.com/p/open-thread-178/comments#comment-2270264). Alex Power suggested that I put it on a web page, so I did:
https://medium.com/@pseudodionysus/did-covid-19-come-from-a-lab-a-critical-examination-208f0eff7c3
There are also some new developments on this front. President Joe Biden asked the intelligence community (IC) to prepare a report for him on the origins of COVID-19 within 90 days. They did so, and the declassified summary is here (https://www.dni.gov/files/ODNI/documents/assessments/Unclassified-Summary-of-Assessment-on-COVID-19-Origins.pdf).
In summary, “Four IC elements and the National Intelligence Council assess with low confidence that the initial SARS-CoV-2 infection was most likely caused by natural exposure to an animal infected with it or a close progenitor virus”. However, “one IC element assesses with moderate confidence that the first human infection with SARS-CoV-2 most likely was the result of a laboratory-associated incident”. Three IC elements couldn’t come to a conclusion. Importantly, the IC as a whole agreed that China did not have foreknowledge of the virus before the initial COVID-19 outbreak emerged.
Before the review, only two agencies favored the natural origin theory, according to NYTimes. So whatever secret information the IC has access to that we don’t, it pushed two (three?) of the agencies toward the natural origin theory.
I also stumbled across this paper in Cell, critically reviewing the lab leak hypothesis: https://www.cell.com/cell/pdf/S0092-8674(21)00991-0.pdf
It’s fairly short (9 pages of text) and relatively non-technical by scientific paper standards. To summarize the arguments, (1) lots of zoonotic coronavirus have infected humans; (2) the earliest SARS-CoV-2 cases were clustered around the Huanan wet market, as expected for a zoonotic origin; (3) the WIV was not known to be working on any virus very similar to SARS-CoV-2 (not even as similar as RaTG13), and there would have been no reason to hide such work before the pandemic; (4) more recent common ancestors than RaTG13 have been found in the wild since the pandemic began; (5) past recombinant virus research at WIV has used the WIV1 backbone, which is unrelated to SARS-CoV-2; (6) the furin cleavage site is present in a lot of coronaviruses, and the one in SARS-CoV-2 is suboptimal.
Some interesting facts I learned:
1. “In direct parallel to SARS-CoV-2, HCoV-HKU1, which was first described in a large Chinese city (Shenzhen, Guangdong) in the winter of 2004, has an unknown animal origin, contains a furin cleavage site in its spike protein, and was originally identified in a case of human pneumonia (Woo et al., 2005).”
2. “bat virus RaTG13 from the WIV has reportedly never been isolated nor cultured and only exists as a nucleotide sequence assembled from short sequencing reads”
3. Although RaTG13 has the highest genetic similarity to SARS-CoV-2, a history of recombination means that 3 other viruses have a more recent common ancestor with SARS-CoV-2. These are RmYN02, RpYN06, and PrC31, which were all sequenced after the pandemic had begun. So SARS-CoV-2 was not derived directly from RaTG13 (which we knew already), and scientists are making progress in tracking down the natural reservoir of SARS-CoV-2 (which I didn’t know before).
4. Page 31, showing the geographic distribution of the earliest cases. “Two of the three earliest documented COVID-19 cases were directly linked to this market selling wild animals, as were 28% of all cases reported in December 2019 (WHO, 2021). Overall, 55% of cases during December 2019 had an exposure to either the Huanan or other markets in Wuhan, with these cases more prevalent in the first half of that month (WHO, 2021)”.
This brings up an important philosophical issue: how much weight do you give to geographic coincidences? If you give them a lot of weight, and you believe it’s improbable for a zoonotic pandemic to have begun only 7.4 miles away from the Wuhan Virology Institute, then it should be downright impossible for a non-zoonotic pandemic to have begun 0 miles from a wet market and for a large fraction of the earliest cases to have traceable ties to the wet market. If you think the lack of traceable links between 72% of the December 2019 cases and the wet market is evidence the virus didn’t come from the wet market, then the lack of traceable links between 100% of the December 2019 cases and the WIV should be conclusive evidence that it didn’t come from the WIV. On the other hand, if you don’t put much weight on geographic coincidences, you could argue the wet market link means little because wet markets are crowded places with lots of people where viruses tend to spread easily--but then you’d have to agree that the coincidence that the WIV is in Wuhan means even less, because there’s no traceable link between the earliest cases and the WIV at all.
to people at the meetup: what does Scott's voice sound like? in my head, just because he writes in what I guess I'd term a "thoughtful style", my mental conception of his voice ends up assigning it a pleasant timbre, somewhat quiet and clipped. But as we all know, this is often a terrible assumption to make
Do you ever feign ignorance when debating with someone? I've found myself trying it.
The situation is: you want to say X. You know that saying X is going to make some people you're talking to angry, because X sounds a bit like Y. Y is something that "the other side" say. You think Y is false, and so do your friends. But Y is so toxic, if you say anything associated with it, like X, Y will come to mind. Examples in this blog post: https://www.overcomingbias.com/2008/06/against-disclai.html
I used to clarify, disclaim, and preface a lot: "I'm about to say something that sounds like Y. We all know Y is false and bad. I believe X - that may sound like Y, but it's different because Z".
I've changed my mind; I think the best approach is to say X and let the other person bring up its connection to Y.
If you're the one who brings up the connection, from their point of view, this is a partial admission of a connection.
If they're the one who has to bring up the connection, they're offered the option of not bringing it up, and trying to attack X on its own terms. Offering that option allows them to think about how they might attack X if it wasn't related to Y. Their interest in doing that might entice them away from dwelling on the possible Y connection.
This comes partly from hearing about how to react when you've been cancelled (clarifying is taken as a sign of concession), and partly from Robin Hanson, who wrote the blog post a lot, and appears to live this himself - watch this remarked on amusingly 1m5s into this video https://www.youtube.com/watch?v=kBRwDYNjgog
In American culture, is there a stigma associated with packing-up uneaten food, following an office function? In an office of about 100 staff (we all share the same break room), I am among the few people who will take home uneaten food following birthday celebrations, or retirement events, or whatever. Most people seem content to just leave the remaining food in there, where it will be thrown out within a day. Many of my co-workers earn minimum wage, or just above it, so you'd think that they would have an incentive to grab free food. I find it all a bit odd.
Couldn't Reaper drones be reconfigured to take on light infantry/insurgents, like in Afghanistan? Why haven't they been, to date? I know when most people think of drones, they're thinking of the relatively small ones- and those little guys seem to have be the focus of most warfighting advances recently (as small/cheap bombers). But the America Reaper drone, and the Predator before it, is a 5000 lbs. beast. They're already loaded with missiles- why not the machine gun that goes on an attack helicopter now?
I understand that Reapers now fly at a much higher altitude. But couldn't the US reconfigure one with a mounted machine gun to hover 50ish feet off the ground and take out Taliban insurgents? Seems like a good way to route that sort of light infantry (and strike absolute terror into their hearts), while not risking American lives. If their range is too limited to be effective (hard to believe), a mothership cargo plane could circle over the battlespace and drop them off, then pick them up. Obviously this would be difficult/impossible in a war against Russia or China- but the US does a ton of these counterinsurgency campaigns against very unsophisticated foes all over the world, who lack anti-aircraft weapons.
Why hasn't this been done to date? Is it that the Reaper is vulnerable to small arms fire? I don't believe the Taliban has effective artillery or anti-aircraft weapons- is an AK gonna take the drone down? If so, is it just impossible to armor up a drone enough while still keeping it flyable? Or, is it that that level of firing precision isn't possible with a remote control? (Maybe operators could be in the mothership cargo plane, if latency is an issue).
Seems like a hardened Reaper drone with machine guns, that can't be taken down by small arms fire, would revolutionize remote US wars where we don't want to risk ground troops. Curious if it's technically possible. I am especially hoping to hear from John Schilling, of course
I was thinking about the debate regarding whether the obesity crisis has been caused by psychological factors like food reward / palatability, vs being caused by environmental contamination, which is being researched by Slime Mold Time Mold. One of the questions in my mind is how obesity set points come to be and how they become so stable, which led me to the following idea.
If your obesity set point is set by your psychology, would it make sense to use an extreme diet to get down to your desired set point, then take LSD to reset your neural bases, in the same way that Scott has written about before? Is there any possibility that would work?
Since the discussion of aducanumab put me at odds with most of the other commenters, I wanted to make my position more concise and offer predictions/betting opportunities:
1) I did not know that FDA approval may mean forcing Medicare to pay for it. I hope that that didn't happen and if it does, I'm against the approval. I am for allowing doctors to decide based on the available test data whether a drug with nice biomarker action but crappy endpoint evidence has a high enough chance of working on their patient to warrant the cost.
2) I am not so much a fan of aducanumab itself as I am more optimistic than others about the amyloids as a therapeutic target in general. I predict the following, and am willing to bet on these odds:
a) >50% chance of an amyloid targeting drug or a mix containing one showing success on cognitive tests in phase III trials within the next 4 years.
b) >75% within the next 7 years.
c) >30% chance of aducanumab specifically showing success on cognitive tests in some treatment modality within the next 3 years.
d) Conditional on (c) happening, >75% that that treatment modality will involve targeting to early stage or pro-dromic patients, or a combination with a drug against a different target.
The idea of being tricked into going to Winnipeg makes me think of https://what-if.xkcd.com/155/
Ezra Kline had an interesting conversation with Bessel van der Kolk about his book on trauma, "The Body Keeps Score." During the talk van Der Kolk made a statement that really tickled me. "Tango dancing is probably more effective treating trauma than CBT."
Podcast link:
https://www.nytimes.com/2021/08/24/opinion/ezra-klein-podcast-van-der-kolk.html
Just let Winnipeg happen, Scott!
I live in Salt Lake City now, but I'm from Winnipeg originally, so I found the calendar thing doubly confusing.
What books would people recommend I read to get a basic grasp on modern science the history of science? I’d like the books to be engaging to a layperson and not a something I need to slog through. Is Steven Weinburg’s “To explain the world” worth buying ?
> 4
What a relief, I thought I was being targeted by someone nefarious. Anyway.
Next up, in the quixotic ululations of AI Defense in Depth:
Could you have stopped Chernobyl?
https://aidid.substack.com/p/could-you-have-stopped-chernobyl
So I try to stir fry Kung Pao Chicken at home, and it never quite comes out right. I thought maybe it was the heat, so I tried it over a friend's gas grill at max heat to see if that worked - and it still didn't come out right. Is there something to it that most recipes miss? Szechuan peppercorns?
I am getting my first COVID shot tomorrow in anticipation of a plenary shift away from remote work. ACX readers, which vaccine should I get?
I'm leaning towards J&J. I feel ashamed to even raise this point, but I waited so long out fear of unusually high reports of adverse reactions in VAERS. I know this data is not clinically verified and may not be helpful, but I have had a nagging paranoia nonetheless. I trust this community more than I would any other online group, so please feel free to critique my life choices or offer helpful explanations for your answers. At any rate, I will definitely be vaccinated. I am just trying to avoid adverse reactions or long-term effects. I prefer the J&J because I already had 3 viral vector shots for HPV, and to the best of my knowledge there are no long-term effects of adenovirus-based shots.
Is there some kind of good legal/public policy reason why America can't make people save for retirement via a 401k? Like, everyone with an existing 401k registers it with the IRS- those without such a registration, the IRS auto-deducts say 3% from their paycheck (so 2ish% after taxes). Yes, it probably should be more, just trying to be politically realistic. The private 401k is some Vanguard target date fund or mutual fund, with ultra-low fees. If people really feel strongly about it, they can fill out a form and the IRS will stop deducting the retirement monies, and send them their funds back (maybe they can try again in a decade). Yes, it really should be mandatory, but again trying to be politically realistic ('big gubmint can't make me save for no retirement', etc.)
Why hasn't the US done this to date? Social Security (which this is not attempting to replace) is clearly not sending retirees enough money to actually live on. It could be argued this will actually help save SS. The US generally goes for market-oriented solutions way more than any other developed country, and this is not a government-run pension or anything. Tons & tons & tons of people are not saving for retirement, or don't start young enough, and here is a reasonable market-based nudge that's opt-out vs. opt-in, and will probably save the US government money in the very long run. Conservatives should love it, in that there's lots of evidence that having assets makes people more fiscally conservative. Liberals should love that it's reducing inequality, which is mostly about asset prices. I'm sure Vanguard or whomever would love it. Best of all, it doesn't cost the government much of anything that I can see- maybe some administration costs, but not a mondo amount. Why haven't we done this? Seems like virtually free money.
(I'm assuming we can discuss dry public policy like opt-out retirement savings in the 'no politics' thread. If this discussion is considered political, I will delete this comment, apologize, and also commit sepuku)
I read this study on rationality tests: https://link.springer.com/article/10.1007/s11109-019-09579-0
Does anyone have access to the tests they use in these studies? They reference Tversky and Kahneman. Is that a test anyone is familiar with?
I was wondering where that calendar appointment came from!
Weird experience on the Winnipeg reservation: I didn't see that until yesterday, when it popped up on my smartphone. Which I have never used to browse ACX, and I'm pretty sure I haven't used it to read my ACX-registered email account since the Winnpeg post went out.
Networking: Am I missing something?
I attended a casual meet up of modest size recently and exchanged contact info. with a few other attendees. I suggested (individually) that we get lunch, and all seemed amenable. However, none of them have yet responded to my open-ended follow up text.
Am I doing something wrong here? I would have assumed that 1) People are generally friendly and willing to get lunch and 2) Would respond to communications from someone they personally know, that take less than 2 minutes to do so, within a day or so. At least, people in my office at varying levels of rank/seniority have usually been willing to get lunch with me. I wouldn't know if I was, but I don't *think* I was so rude/awkward/annoying that the other attendees actively wouldn't want to interact with me again. (My general approach to conversation is to limit expressing my own opinions and try to ask the other party about their interests and background. I think this is usually pretty effective---although, again, I wouldn't know if it wasn't.)
It's a small sample size, I'm certainly overthinking it, and I suppose it's not impossible that one or two will eventually get back to me. Nonetheless, I find this discouraging. Going to gatherings, exchanging contact info. with a few people you get along with, and following up with lunch seemed like a potentially very effective way to make friends/connections to me. But, without that last step, it seems like the value of going to the gatherings at all is substantially diminished.
I commented in an open thread back in the spring about the wretched thing that is Stanley Kubrick's current Wikipedia article and the controversy surrounding adding an infobox to his article. https://astralcodexten.substack.com/p/open-thread-163/comments#comment-1499620
Discussion on the topic has been closed for the past 2 years. Starting Wednesday, September 1, discussion will reopen. I hope you all will join me in correcting this most abominable stain on Wikipedia by arguing in favor of giving him his damn infobox.
Is anyone travelling to the ACX meetup in NYC from Westchester? I will be driving down from Yonkers with my wife, and it'd be nice to have someone to split (presumably) eye watering parking fees with.
This is an article about a paper studying the reasons why certain groups care about some social issues but not about others that affect their group disproportionately (e.g. there wasn't an LGBT+ #MeToo, even though LGBT+ people get sexually harassed much more than any other group): https://www.scientificamerican.com/article/why-we-rally-around-some-social-issues-and-not-others/
TL;DR: It's not an information gap: telling group members about an issue affecting their peers changes nothing about their attitude towards it -- if the issue wasn't one they already cared about.
This is not strictly speaking about politics, so I think it's ok for this thread, but if you disagree feel free to flag/delete it.
Disclosure: the author of both the article and the paper is my wife.
Quick question that must have been covered but I can't find it.
The Vitamin D post a while back ( https://astralcodexten.substack.com/p/covidvitamin-d-much-more-than-you ) had the structure of "generally studies says Vitamin D has no effect; take it anyway since its the safe default". Clicking on the studies I found they seemed to be using p-value hypothesis testing.
So in p-value hypothesis testing you have your Alternative and Null statements. The model sees these in terms of "what damage is done if this statement is wrong, but I accept it" -- the Alternative statement is the statement that does more damage if you wrongly accept it; the Null statement is the one that does less damage if you wrongly accept it.
For this reason, the model treats the Null hypothesis as "am I allowed to accept this?" and the Alternative hypothesis is "am I forced to accept this?". The p-value itself is the probability of you wrongly accepting the Alternative hypothesis (hence why lower is better, and at p=0 you just never accept the Alternative hypothesis).
So, the way people /should/ choose their p value, Null hypothesis and Alternative statement is:
1). The Null statement is the safe default; the alternative one is the one you really need to be sure about before accepting (it's like "guilty until proven innocent", because we'd rather let guilty people be free than innocent people be locked up);
2). You should pick your p-value based on the relative danger of incorrectly asserting the Alternative statement; for less dangerous or less costly Alternative statements, higher p-values are sensible; for more dangerous or costly statements, lower p-values are prudent.
However, in every actual paper I've seen, they instead choose as follows:
a). The Alternative statement gives novel information.
b). A set of gold-standard low p-values. As low as is practical is always assumed to be better.
In the Vitamin D article, Scott applied (1) post-hoc by saying it's worth taking vitamin D anyway; correcting their error in those studies that the statements were inverted (they all should have accepted the assumption that Vitamin D is helpful). To be clear: in these studies, taking vitamin D is helpful should have been the Null statement based on (1). Then the hypothesis test setup matches common sense.
And weird concepts like "tending toward significance" I see as post-hoc attempts to apply (2).
But it's very frustrating because we shouldn't need to post-hoc mitigate the substitution of (a) for (1) and (b) for 2: just set up the parameters correctly in the first place.
As an aside, I think it's interesting that (a) is substituted for (1), as if the MOST dangerous thing isn't lives lost, it's us believing something that isn't true. This is really interesting to me and I think is the way a lot of people handle these things.
Sorry for length.
I have an idea re: Mandarin tone marks, which are often omitted in normal English transcription of Chinese words. They should be placed as separate characters next to the syllable (example: Xiˊ JinˋPingˊ), similar to Zhuyin, rather than diacritics over the syllable nucleus, as they currently are in Hànyǔ Pīnyīn. This would end both the practice of Gwoyeu Romatzyh type romanization (e.g., "Shaanxi") and end the confusing practice of placing an apostrophe to distinguish two syllables in a single word (e.g., "Xi'an"). I will write a post elaborating on this. Does anyone think this is a good idea? Thoughts?
Hello ACX readers! I recently got a job with the US government for which the most important part of training is learning a new language. I now have to rank my preferences for which languages I'd prefer, but I'm really torn! I know ACX readers tend to be good at making predictions, so I was wondering, of the following languages which do you see as being the most relevant to know in the context of national defense within the next 5 years. Also which, do you predict, will have the most interesting missions to work on?
-Arabic
-Pashto
-Russian
-Persian-Farsi
-Chinese-Mandarin
(I know I'm not providing much information about the job to go off of, I'm just looking for general thoughts)
If people learn to control the DNA of their embryos and we get millions of babies at the level of von Neumann and Tao what do you think life will be like?
Will they get fed up with dumb old farts taking up all the space and do something nasty to the unedited?
I'm hosting a meet up (Buffalo NY) Someone who is not vaccinated wants to come. (He is the only person who has emailed me and said he wants to come. There is one other 'maybe' RSVP.) I'm fine with non-vaccinated people coming, but I don't think this is fair to anyone else who may show up. I'm looking for ideas.
What is the wisdom in the words: "make haste slowly?"
This was one of the Roman emperor Augustus' favorite sayings. I'm trying to figure out what it means.
A puzzle adapted from the youtube channel of Michael Penn, one of my favourite maths youtubers. I like it because the solution turns out to depend on something I'd never have guessed was relevant.
n baby chicks sit peacefully in a circle. Suddenly, each chick turns either to left or right (equal chance of each, chosen independently) and viciously pecks one of the chicks sitting next to it.
What is the probability that precisely k chicks are left unpecked?
One of the other posts in this thread had led me to think of an interesting ethical dilemma. Some people have problems with the idea of cloning human beings. One major ethical concern is consent. If von Neumann never provided consent to be cloned, how ethical is it to make a million copies of him. There are other ethical concerns but I want to focus on this particular ones.
There is a concept of implicit consent if you are unable to ask. For example, if I am rushed to a hospital unconscious, they can do what they need to do to help me because it is reasonable to assume that I want to be helped. In a world in which no one wants medical care, this assumption wouldn't be warranted.
If we want to discover whether or not it is ethical to clone von Neumann a million times from a consent perspective, I would argue we should start with 100. If 100a von Neumann's all agree that future cloning is acceptable, then that would be a good reason to think the original von Neumann would think it's acceptable. There would be a number of biases. A major one being that a cloned von Neumann would be biased in favor of cloning because he would not exist otherwise. But if you had a sufficiently large number of von Neumann's, you could probably largely overcome this bias.
A clone of you consenting is not sufficient evidence to think that you would consent but 100 for 100 clones of you consenting is a really good reason to assume implied consent. If the clones didn't agree in large number, then there might be an ethical concern about consent. Obviously this probably has some major complications but it's just an interesting thought I had.
The EU is recommending that member nations block nonessential travel from the USA, per https://www.bbc.co.uk/news/world-us-canada-58386967. I'm afraid that may impact Scott's plan to join ACX meetups across Europe, depending on how the various member states implement the recommendation.
Separately, the UK announced its own changes yesterday, and I don't think they affect US visitors.
I'm figuring out the taxonomy of US healthcare plans, specifically HMO/PPO/EPO/POS. Of course the names are clearly chosen at random and most explanations online are a bit retarded, but looks like at the end of the day it mostly amounts to a 2*2 matrix - out-of-network providers are partially covered or not, specialist appointment requires a physician referral or not.
Namely, my understanding is as follows:
POS: Out-of-network allowed, physician referral required
PPO: Out-of-network allowed, physician referral not required
HMO: Out-of-network not allowed, physician referral required
EPO: Out-of-network not allowed, physician referral not required
Can someone confirm this is generally correct?
There are discussions of boosters in the US. Right now, the public discourse is muted because the news cycle has run its course a bit, but I want to think through this a bit.
Most of us in the US have had two shots of an mRNA vaccine (typically Pfizer, atypically Moderna) 6 weeks apart. Some have had the single-shot J&J vaccine (which is an Adenoviral/viral-vector vaccine similar to AstraZeneca popular in Europe and elsewhere).
There seems to be general consensus that the mRNA vaccine efficacy starts waning around the 8 (or 7 or 6) month marker. Are there studies that show this? If so, does the efficacy wane somewhat linearly and if 8 mos represents critical-failure level, does 5 months represent 62.5% emptiness or something like that? How does this work?
There is some evidence that mixing vaccines is beneficial, for e.g., https://science.sciencemag.org/content/372/6549/1392.full. This paper is predicated on previous disease, but I've seen other papers that also suggest that mixing vaccines is beneficial. Overall, I get the sense that boosting a viral-vector vaccine with an mRNA vaccine is a clearly beneficial path. What about when one starts with a 2-dose mRNA vaccine - would they be better off with a 3rd dose of an mRNA vaccine, or would they be better off with a viral-vector dose? Are there studies that are suggestive of one approach or the other?
I visited The Netherlands from North America this August, probably one of a tiny minority to do so. To my complete astonishment, no-one was wearing any masks. In fact, I had trouble finding any kind of covid measure!
It seems unlikely to me that the Dutch population is ignorant and/or brainwashed. The Netherlands in consistently near the top in human development, freedom, quality of life, etc type indexes. They’re not exactly part of the developing world. Also, they have vaxxed a much greater share of their population than the US has.
I’ve heard the same thing is going on in Scandinavia? Perhaps a poster from Holland/Europe could explain the current Covid strategy?
How can I get the most money for some Sterling silver flatware I want to sell? I have several old forks made by "R. Wallace & Sons." A jeweler looked at photos of the pieces that I emailed to him, and he said they had no collectors value, and instead would be melted down. He said the whole collection was worth a few hundred dollars.
Instead of selling the forks to a local jeweler, should I "cut out the middle man" and go directly to a metal smelting plant? Wouldn't I get more money that way? How do such deals work, and where would I go in the Mid-Atlantic to do this?
Why are there way more MtF than FtM trans people?
I am posting this several places because the group of people I'm looking for likely isn't very numerous but...
I'm doing a home remodel and I want to hide some nods to the video game The Witness in a few places. The problem is I'm not very good with visual design. I'm looking for an artist who is familiar with the concepts behind The Witness (at least 2 layers deep, if you've seen that deep then you'll know what I mean) and is open for commissions. I'm looking to pay money for them to brainstorm some design ideas and come up with some rough sketches for 3-8 easter eggs, including a custom glass/iron front door panel (found a company that can do full custom for a good price), and some designs for cabinets.
If you are this person, know someone who might be this person, or have a good idea at all of where to look to find this person, please help me out here.
Looking for housing in the Bay Area, I'm not picky at all on room size/space (lived in Japan) but am hoping for a place that is relatively clean and not too old. I don't mind sharing the space with others and am hoping to meet cool people!
A math puzzle - http://www.acritch.com/media/math/Self-assigned_student_numbers.pdf from Scott aaronson’s comments.
Anyone have any suggestions for good Crypto trading groups?