Posts

Upvote/downvote amounts 2018-01-27T08:00:26.533Z · score: 35 (9 votes)
Principals, agents, negotiation, and precommitments 2012-09-21T03:41:55.923Z · score: 20 (19 votes)

Comments

Comment by gwillen on Making the Monte Hall problem weirder but obvious · 2020-09-18T00:47:39.187Z · score: 3 (2 votes) · LW · GW

It seems like something went wrong with the post mirroring here. It's got a bit of it, and then cuts off without any indication that there's more.

Comment by gwillen on Evaluating Opus · 2020-08-09T22:09:39.173Z · score: 12 (3 votes) · LW · GW

Coauthor here: FWIW I also favor eventually switching to the (more reasonable IMO) streaming approach. But this does require a lot more complexity and state on the server side, so I have not yet attempted to implement it to see how much of an improvement it is. Right now the server is an extremely dumb single-threaded Python program with nginx in front of it, which is performant enough to scale to at least 200 clients. (This is using larger than 200 ms windows.) Switching to a websocket (or even webrtc) approach will add probably an order of magnitude in complexity on the server end. (For webRTC, maybe closer to two orders, from my experiments so far.)

Comment by gwillen on [updated] how does gpt2′s training corpus capture internet discussion?  not well · 2020-08-04T02:50:51.351Z · score: 3 (2 votes) · LW · GW

Do you have examples of that kind of output for comparison? (Is it reproducing formatting from an actual forum of some kind, or the additional "abstraction headroom" over GPT-2 allowing GPT-3 to output a forum-type structure without having matching examples in the training set?)

Comment by gwillen on Betting with Mandatory Post-Mortem · 2020-06-25T00:26:31.177Z · score: 10 (6 votes) · LW · GW

I like this a lot. I would also like to hear a post-mortem from the winner in a lot of cases, although of course it's kind of silly to impose it. But I do sometimes see the winner and the loser agree that the bet turned out to be operationalized wrong -- that they didn't end up betting on the thing they thought they were betting on. I'd like to know whether the winner thinks they won the spirit of the bet, as well as the letter.

Comment by gwillen on Covid-19: My Current Model · 2020-06-01T19:05:29.617Z · score: 15 (5 votes) · LW · GW

I am still skeptical of the strength of "MNM" effects. Control systems with huge lag times are infamously unstable. Are most people really able to judge whether they should be scared or not based on the R value from a week or two ago, which they don't even know but have to eyeball from the trend in cases?

Comment by gwillen on Covid-19: My Current Model · 2020-06-01T18:47:28.770Z · score: 7 (4 votes) · LW · GW

Right, I was thinking the same thing -- not just a person, but medical personnel. So you're going from patient 1, to someone's hands, who is then directly touching patient 2, plausibly even patient 2's mucous membranes. That's much more direct than a typical fomite contact, which is more like face-hands-fomite-hands-face (or if you sneeze on a doorknob, face-fomite-hands-face.)

Comment by gwillen on Coronavirus is Here · 2020-04-15T02:49:24.373Z · score: 4 (2 votes) · LW · GW

No, a bet was never made and accepted. You can see my reply to him as to why I didn't accept his offer. He never replied to my counteroffer.

Comment by gwillen on April Coronavirus Open Thread · 2020-04-05T20:04:32.641Z · score: 11 (3 votes) · LW · GW

For what it's worth, I upvoted your comment.

But since you stated that you had a source already, I don't see how it's asking much for you to post a link to the source you already said you had.

[EDIT: After a couple days, I regret the tone of my comments here. I don't want to discourage anybody from writing posts, or asking for help in composing posts. And I think "there oughta be a rule" was a poor summary of my position and sounded pretty hostile. I think it would be nice if people mentioning the existence of sources would link the sources they mention, and in general I'd like it if people linked source more often. But that wasn't really directed at you personally, it was spillover from elsewhere.]

Comment by gwillen on April Coronavirus Open Thread · 2020-04-05T05:47:54.571Z · score: 2 (1 votes) · LW · GW

Cool, thanks for expanding on that. You might want to link this comment in your other comments about this idea, so people have some details to read. It's a lot more informative than the one I was responding to!

Comment by gwillen on April Coronavirus Open Thread · 2020-04-02T20:01:06.258Z · score: 4 (2 votes) · LW · GW

This is an interesting idea but would benefit from more elaboration.

Why the GI tract in particular -- do you have evidence that this will significantly reduce the risk of respiratory symptoms, or just speculation / "common sense"? Is there evidence that the GI tract as the initial site of exposure will produce an infection / an immune response, but with a reduced chance of the infection spreading to the lungs / respiratory tract? Or with it taking longer to get there, similar to Robin Hanson's thoughts about deliberate exposure with a low dose, like variolation of old?

If you have any links/references, please definitely post them. If it's just speculation, it's interesting speculation but tell us what it's based on.

Comment by gwillen on April Coronavirus Open Thread · 2020-04-02T19:57:37.996Z · score: 3 (2 votes) · LW · GW

I have heard, and give some credit to, the theory that silicon valley tech company culture played a role in the bay area's response being relatively early. Tech companies were making contingency plans and sending their employees home, well before there was any kind of government action here. I don't know what fraction of employees / day-to-day interactions that represents. But e.g. all Google employees working from home seems like it could have played a nontrivial role in Mountain View, which was the epicenter of the bay area coronavirus outbreak.

Comment by gwillen on April Coronavirus Open Thread · 2020-04-02T19:50:13.693Z · score: 5 (4 votes) · LW · GW

Can you link whatever you have on this, even before you write it up? Articles, the paper you mentioned, the studies about reducing infections this way, where you get the idea in general?

(Everyone, please do this! It is really helpful, and it's probably easier for you to re-find things than for people to try to find them based on your comment! I wish LW had a rule about doing this!)

Comment by gwillen on March Coronavirus Open Thread · 2020-03-30T07:02:16.491Z · score: 4 (2 votes) · LW · GW

That's extremely interesting. I would love to see someone in our community who I trust to be good at statistics redo the analysis, since all the data is public.

Apparently there are already multiple trials underway, though: https://www.sciencemag.org/news/2020/03/can-century-old-tb-vaccine-steel-immune-system-against-new-coronavirus . The Science article came out before the paper, so I wonder where the idea struck first.

Apparently the broader pro-immune effects of the BCG vaccine for tuberculosis have been known or suspected for a long time; see e.g https://www.ncbi.nlm.nih.gov/pubmed/31055165 "Non-specific effects of BCG vaccine on viral infections", http://sci-hub.tw/10.1016/j.cmi.2019.04.020, which is a fucking wild read and I highly recommend reading the whole paper.

Comment by gwillen on [Update: New URL] Today's Online Meetup: We're Using Mozilla Hubs · 2020-03-29T21:18:03.308Z · score: 2 (1 votes) · LW · GW

I tried taking a video to demonstrate the issues, but I seem to have hung the headset. I think the Quest probably just can't do Mozilla Hubs with adequate performance.

It has no trouble at all with similar environments in native apps, so again I wonder if WebVR is to blame.

EDIT: Honestly I wish I could get a video of the truly comedic amount of problems this has caused the headset. First I got a dialog that "Oculus System" had stopped responding, then the dialog separated into two pieces at a jaunty angle to each other and the display froze. Then I tapped the power button, and got a dialog asking if I wanted to power off the headset, but I was unable to click any buttons. One of the controllers seemed to be tracking inverted from reality somehow (pointed along the opposite vector in VR from how the real controller was pointed.) The other one was sort of gently orbiting.

Honestly, I have found VR to be a pretty buggy experience overall, but this is definitely the worst behavior I have ever seen from this platform. It's pretty funny.

Comment by gwillen on [Update: New URL] Today's Online Meetup: We're Using Mozilla Hubs · 2020-03-29T21:15:35.760Z · score: 2 (1 votes) · LW · GW

I finally made it into one of the rooms. I suspect I'm having performance issues? The tutorial room had 4 avatars and very few objects, and loaded fairly promptly. The other environments are more complex with more people in them. I've finally gotten into one of them, but it's ... bad. Audio is almost unusable. A lot of objects are failing to render. Tracking is hopeless. Movement is impossible.

I suppose I could try rebooting the headset and see if anything improves.

Comment by gwillen on [Update: New URL] Today's Online Meetup: We're Using Mozilla Hubs · 2020-03-29T21:12:55.376Z · score: 3 (2 votes) · LW · GW

I am trying to use Hubs through Oculus Quest. So far I am extremely unimpressed. If I manage to enter one of the non-tutorial rooms without a hang, I might get slightly more impressed, but the audio is also pretty crap for me. Lots of weird static and glitchiness. Sometimes when I turn my head I lose tracking and the world jitters. I think WebVR is not a good substitute for native VR apps.

EDIT: Ok, I rebooted the headset (and switched from the Mozilla browser, "Firefox Reality", to the native browser), and it seems to be working smoothly now. Not sure what the cause was of the issues before.

Comment by gwillen on Will grocery stores thwart social distancing, and when should I eat my food stockpile? · 2020-03-29T08:41:22.054Z · score: 4 (2 votes) · LW · GW

Most or all the 24-hour grocery stores here (bay area) have converted to having closing hours, as far as I know, to help them deal with the logistical problems caused by overwhelming demand. You might expect this to happen in your area too, at some point.

Comment by gwillen on Coronavirus Justified Practical Advice Summary · 2020-03-29T08:03:33.391Z · score: 7 (4 votes) · LW · GW

Beware, some of the very thin bare-looking copper wire you will find is "magnet wire", which is actually coated in a thin layer of clear insulation.

Comment by gwillen on What should we do once infected with COVID-19? · 2020-03-28T21:11:31.212Z · score: 12 (4 votes) · LW · GW

Listing / summarizing some things I've seen elsewhere:

This general summary post by Sarah Constantin: https://srconstantin.github.io/2020/03/27/home-care-mild-COVID19.html

A post by SC specifically on "non-invasive ventilation", meaning CPAP and BiPAP machines (which some people may already have at home), with positive conclusions: https://srconstantin.github.io/2020/03/20/non-invasive-ventilation.html

A document by Matt Bell with information about chloroquine phosphate / hydroxychloroquine: https://docs.google.com/document/d/160RKDODAa-MTORfAqbuc25V8WDkLjqj4itMDyzBTpcc/


One of the most intriguing things I saw was about "proning": https://emcrit.org/pulmcrit/proning-nonintubated/

The author of that post is Josh Farkas, a pulmonologist (i.e. lung specialist) and assistant professor of critical care and pulmonary disease (i.e. lung disease.)

"Prone" here means a face-down lying position, the opposite of "supine" which means face-up. The author says "Typically we prone intubated patients." From context, I am reading "we" to mean his hospital / department, and "prone" to mean "rotate into the prone position for 6-18 hours per day." The commonality of this practice seems to vary among hospitals.

The post, however, is a discussion of proning for awake, non-intubated patients, and concludes that it appears safe and effective. There is a lot of uncertainty around how effective it is, but it looks to me like, if you have pneumonia and hospital treatment is not available to you, there is some evidence that -- perhaps counterintuitively -- you will breathe better lying on your belly, vs. on your back.

(The main counterpoint I have seen to this is that frequently moving around and changing positions is best. I can't tell whether the post is largely about patients who are too out-of-it to do that. I have seen it suggested that, if you're able, sitting up is better than lying down (I have no cite handy for this.) There seems to be overall agreement, at least, on this one point: lying stationary on your back for long periods of time is NOT good when you have lung problems.)

Comment by gwillen on What should we do once infected with COVID-19? · 2020-03-28T20:39:46.874Z · score: 11 (3 votes) · LW · GW

I was under the impression that loss of sense of smell was primarily happening to people who take zinc intranasally. (I don't have numbers handy.)

My impression was that the effect of the zinc was supposed to be on the virus (or the virus's interaction with your cells), not on the body. Which (if true) would seem to imply that prophylactic use shouldn't cause adaptation.

This paper appears to be a discussion of a Cochrane review from 2011, and supports prophylactic use (and also generally supports use, and provides more info):

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3273967/

The 2011 version of the Cochrane review in question: https://www.ncbi.nlm.nih.gov/pubmed/21328251 / http://sci-hub.tw/10.1002/14651858.CD001364.pub3

(Irritatingly, there have been a number of subsequent versions of the Cochrane review, but several of them have been withdrawn, for reasons that are hard for me to interpret, although one at least involved an accusation of plagiarism from another meta-review on the same topic. It feels to me like there may be some kind of political fight over ownership of this Cochrane review.)

ALSO, while looking through Cochrane reviews, I found this one in favor of Vitamin C for the common cold: https://www.cochranelibrary.com/cdsr/doi/10.1002/14651858.CD000980.pub4/full

Comment by gwillen on March Coronavirus Open Thread · 2020-03-25T00:36:17.565Z · score: 3 (2 votes) · LW · GW

Thanks, I believe that article is great advice and I fully endorse it -- I saw it a few days ago but never came back here and updated my comment.

Comment by gwillen on LessWrong Coronavirus Agenda · 2020-03-18T18:14:41.842Z · score: 4 (2 votes) · LW · GW

Please apply the Coronavirus tag to this post.

Comment by gwillen on March Coronavirus Open Thread · 2020-03-17T22:45:48.162Z · score: 2 (1 votes) · LW · GW

Yeah, I think something at the government (or google/facebook) level would be a lot more effective. (Of course, people might have some qualms. China already did it, of course, and it's mandatory -- but that's China.)

Comment by gwillen on Coronavirus Justified Practical Advice Summary · 2020-03-17T04:01:53.190Z · score: 8 (4 votes) · LW · GW

Yeah, agreed overall. I would not want to discourage literally our only source of direct data on this from doing it again. But ugh, why can't everyone in the entire world please be like, significantly more careful about everything, all the time? (Preceding sentence is rhetorical and is me making fun of myself for making unreasonable demands.)

Comment by gwillen on Credibility of the CDC on SARS-CoV-2 · 2020-03-17T01:04:45.049Z · score: 5 (3 votes) · LW · GW

Note that "survived as an [artificially-generated] aerosol" does not mean that aerosols are generated in substantial numbers in realistic scenarios, nor does it say anything about how infectious the aerosol route is. (Also note that the "3 hour" figure in the preprint's original abstract was grossly misleading; the preprint has been updated to remove it. The real figure implied by their data is longer.)

Comment by gwillen on Coronavirus Justified Practical Advice Summary · 2020-03-16T23:47:37.725Z · score: 5 (3 votes) · LW · GW

Thanks, I am not best pleased about relying on data from a paper that turns out to have been so sloppy. (I guess a rush was understandable under the circumstances, but I think checking all the calculations twice was also arguably imperative under the circumstances! And the misleading abstract was just dumb.)

Comment by gwillen on Coronavirus: Justified Practical Advice Thread · 2020-03-16T23:45:50.838Z · score: 2 (1 votes) · LW · GW

See here: https://www.lesswrong.com/posts/B9qzPZDcPwnX6uEpe/coronavirus-justified-practical-advice-summary?commentId=LuJRfhrNhu4aBanQn

Comment by gwillen on March Coronavirus Open Thread · 2020-03-16T22:31:48.494Z · score: 2 (1 votes) · LW · GW

There are at least two attempts I'm aware of to do almost exactly this, surely more that I'm not aware of, plus some attempts to do other modeling (like, using people's tracks from fitness apps, so they don't have to install a new app.)

I think it's going to be really unlikely to get enough people to use something like this to be useful, but I'd love to be wrong. If you want to help I'm happy to direct you to them.

Comment by gwillen on Coronavirus Justified Practical Advice Summary · 2020-03-16T03:00:54.881Z · score: 4 (2 votes) · LW · GW

Everything I've seen so far seems to suggest that copper oxide is still anti-microbial, but I haven't really attempted to research this, so take that for what it's worth (which is little.) (I've actually become curious whether having it on my hands -- which are visibly turning a bit blue-green where they rub against the copper -- might have further antimicrobial benefits. But this is idle speculation I do not intend to do anything with.)

Comment by gwillen on Coronavirus Justified Practical Advice Summary · 2020-03-16T02:59:17.935Z · score: 6 (3 votes) · LW · GW

FWIW, I started taking Vitamin D without measuring my blood levels after reading https://www.gwern.net/Longevity#vitamin-d . I take 2000 IU/day; the NIH site says the "tolerable upper intake level" is 4,000 IU, so I think I have pretty good margins. https://ods.od.nih.gov/factsheets/VitaminD-HealthProfessional

Comment by gwillen on Coronavirus Justified Practical Advice Summary · 2020-03-16T01:17:48.042Z · score: 18 (8 votes) · LW · GW

Some notes / proposed additions or clarifications:

  • The #1 hazard noted about copper tape turned out to be cutting yourself on the edge of the tape while applying it (4 independent reports I'm aware of, including myself.) Maybe worth mentioning.
  • The 24-hour figure from the abstract of the paper about virus survival is misleading (all those numbers reflected the time of the last datapoint they saw virus in, notwithstanding the time of the first virus-free datapoint.) Those numbers (and the whole abstract) were all removed in v2 of the paper (https://www.medrxiv.org/content/10.1101/2020.03.09.20033217v2.full.pdf -- you may want to update the link in the post.) So "viable virus could be detected up to 24 hours on cardboard" actually means that it dropped below detectability somewhere in the 24-48h window. The half-life estimation graphs were also made much larger and easier to read in v2; eyeballing the graph, the cardboard number looks to have a median of around 24h, but a distribution that extends outward closer to 36h or 48h. (Although, as you say, this is depending on the concentrations they were using; the half-life, which I believe was reported correctly, is the more important figure.)
  • On vitamin-D, if you haven't seen this, someone found a meta-analysis that sort of suggests a benefit even for people with adequate blood levels, if you squint a little bit. (I believe it was a subgroup analysis that showed this, with the subgroup being "taking it daily/weekly, rather than as a bolus". Which does seem to match how anybody self-supplementing would take it.) https://www.bmj.com/content/356/bmj.i6583
Comment by gwillen on March Coronavirus Open Thread · 2020-03-16T00:44:19.499Z · score: 4 (5 votes) · LW · GW

I think this doesn't quite change everything, for the following reasons:

  • Even if long-term immunity is unlikely, short-term immunity will push this back towards the flu category, where most people are not getting it acutely at the same time. This will significantly improve the healthcare situation vs what we're seeing in the pandemic phase.
  • Diseases evolve towards increased spread, which usually involves evolving towards reduced lethality / severity. If this becomes endemic it's likely to do the same.
  • If it turns out that this does become a severe endemic disease, there will be a lot of pressure on the development of a vaccine, much more so than has been true for human coronaviruses in the past (when they were much closer to being mostly a nuisance, and included in the general "common cold" category.) Even if long-term immunity is unlikely, we can still improve the situation like we currently do with influenza, giving people periodic boosters based on the current circulating strains.
Comment by gwillen on March Coronavirus Open Thread · 2020-03-16T00:40:24.219Z · score: 6 (3 votes) · LW · GW

This is a fairly late update, but closing the loop on this: I believe the 3-10% number ended up being the secondary attack rate among households where the infected person was isolated after diagnosis. So that's an estimate of the rate of transmission during extended close contact before symptoms/diagnosis, not after, which makes more sense. I assume that extended close contact with a symptomatic infected person will result in very likely transmission.

Comment by gwillen on March Coronavirus Open Thread · 2020-03-16T00:38:38.044Z · score: 2 (1 votes) · LW · GW

Hm, can you say more about information? I believe you should get the most direct information (in the information-theoretic sense) out of running tests where the outcome is most in doubt (i.e. where your prior is approximately 50%, although I think this might budge a bit depending on the FP/FN rates of the test if they are different.) You also get information about their contacts -- if their contacts have a lower-than-50% base rate of exposure, then it seems like you get more of that "secondary information" from a positive than from a negative. (I'm not too confident about that, but certainly at worst it's equal, right?)

Comment by gwillen on March Coronavirus Open Thread · 2020-03-16T00:28:48.195Z · score: 2 (1 votes) · LW · GW

I've already heard that influenza cases are down in countries that enforced social distancing / lockdowns for coronavirus. However, it really only takes one country not doing this for influenza to return to typical incidence -- there's no real reason to believe it will be eradicated. (However, the same seems true for COVID-19, so I'm not sure what to expect there.)

Comment by gwillen on March Coronavirus Open Thread · 2020-03-15T22:04:39.153Z · score: 4 (3 votes) · LW · GW

Right, yes, agreed and good point -- my understanding is that a naive epidemiological model gives a fraction of 1 - (1/R_0) of the population needing to be infected, to drive the effective value of R (new transmissions per infected person) below 1, at which point the population can no longer sustain epidemic spread.

Comment by gwillen on March Coronavirus Open Thread · 2020-03-14T07:23:55.343Z · score: 15 (8 votes) · LW · GW

The first strategy leaves you with a huge population of people with no immunity to the virus, which means you have to keep holding the lid on it indefinitely or you're back to square one.

In the second strategy, everyone ends up either immune or dead, which doesn't mean the virus is gone -- it will remain endemic -- but there will be no giant flood of new cases when people resume their lives.

(Obviously it's not quite as simple as that if the virus doesn't generate durable immunity. Then you end up with something like the flu, where partial immunity keeps it vaguely tamped down with occasional flares.)

Comment by gwillen on March Coronavirus Open Thread · 2020-03-14T03:57:48.115Z · score: 13 (11 votes) · LW · GW

I've spent some time thinking about endgames here. (Not that I feel like I've come to any conclusions. I wish I knew what e.g. the WHO thought the endgame was.) The biggest problem I see with this idea is the lag between input and output -- when you change your quarantine measures, you can't observe the result for at least the 5-7 days it takes the newly infected to get symptoms, and longer if you want to get a lot of confidence in your measurement, over the noise inherent in the system.

Control systems with high lag like this are incredibly difficult to work with. Especially in the presence of exponential growth like this system has -- if you accidentally let R get a bit too high, it will be a week or two before you notice, and in that time you will have seeded a ton of cases that you will have to track down and deal with.

I think the most hopeful endgame here, near-mid-term, is that we find a combination of antivirals with high effectiveness against COVID-19, which reduces the rate of severe pneumonia dramatically. At that point our hardest constraint, ventilators, will get relaxed. Beds are a lot easier to deal with a shortage of.

Mid-long-term, of course, we're all hoping for a vaccine. Who knows whether that's going to happen.

In Singapore, and China-outside-Hubei, my impression is that very aggressive high-bandwidth contact tracing is working effectively. Unfortunately, at least Seattle has already given up on that, as far as I can tell. But if we can simultaneously raise our ability to do contact tracing effectively, and lower the value of R below 1 until we get the number of cases under some kind of control, we ought to be able to use a combination of contact tracing and more moderate measures to keep it there. I hope.

Of course, the organization primarily responsible for contact tracing in the US is currently rather indisposed. But in theory, the states should be just as able to do it, although some scaling up may be in order.

Comment by gwillen on March Coronavirus Open Thread · 2020-03-13T01:07:58.939Z · score: 4 (2 votes) · LW · GW

I hope that there is some actual epidemiology going on behind the scenes here that is being oversimplified for the press, but there's nothing in the article to really indicate that the estimate has anything meaningful behind it...

Comment by gwillen on March Coronavirus Open Thread · 2020-03-12T08:08:10.734Z · score: 5 (4 votes) · LW · GW

I haven't checked your models quantitatively, but qualitatively I absolutely believe you that the options here are "bad" and "really really bad", and that neither one of them gets us down to where we need to be.

The difference between 4% and 10% could still save a lot of lives; at that level it may be close to 1:1 (every bed freed up is a life saved), since only the most critical cases will be getting beds at that point.

But you're right that this is clearly not adequate, and the graphic showing the flatter curve as peaking under the capacity line is pretty misleading. (There are versions of the graphic which don't, but they appear to have been memetically outcompeted by those that do.)

I think it's still true that "flattening the curve" will save lives, potentially a lot of lives, so even if the graphic might be a bit misleading as to the possibility of flattening it below the critical threshold, I think it's still a reasonable meme to promote.

But really the ultimate goal has to be reducing R below 1, which will arguably flatten the curve, just not quite in the way the meme seems to be trying to get at. I don't want to steer too close to dark side epistemology here, but if the meme gets people to stay inside, cancel their parties, and wash their fucking hands... it's hard for me to be too against it, and I think it's probably true enough?

Comment by gwillen on March Coronavirus Open Thread · 2020-03-12T01:13:56.163Z · score: 8 (2 votes) · LW · GW

A friend of mine (who lives in the SF Bay Area, currently somewhat of a coronavirus hotspot) posted to Facebook that he hasn't been feeling well recently and he thinks he might be sick (and was having trouble focusing at work yesterday). I posted the following; I don't know him well enough to know how he'll take it, but we'll see. I feel like we're still at a point on the curve where this kind of individual outreach can potentially have substantial value, so I'm offering it as perhaps a template for other people to use.

Yikes, good luck -- I hope you feel better!

If you can, please consider staying home from work and other events. Santa Clara County is a coronavirus hotspot right now, and we need everybody who is able to help reduce the spread. (That doesn't mean you have it, but you very much could.)

If it is the coronavirus: most cases (~80% average, but more than that in younger people, and people without major preexisting conditions) will resolve on their own without treatment; but if you do need medical attention, make sure you call ahead first. I believe self-quarantine is recommend until end of symptoms plus 24 hours, but there is evidence of viral spread until up to 10 days after start of symptoms, so I would suggest whichever is later. "Secondary attack rate" -- that is, the rate of spread to close contacts -- has been estimated at up to 10% for family/household members, so keep that in mind.

Please let me know if there's any more information I can provide or anything I can do to help.

Comment by gwillen on March Coronavirus Open Thread · 2020-03-11T23:47:59.288Z · score: 11 (4 votes) · LW · GW

I think the problem with this is the burden of wiping down surfaces often enough to be effective. Copper appears to take a few hours to effectively render the virus undetectable (according to https://www.statnews.com/2020/03/09/people-shed-high-levels-of-coronavirus-study-finds-but-most-are-likely-not-infectious-after-recovery-begins/, just out today.) So to get the same effect from wipes, I would presumably need to fully wipe down all those surfaces every few hours at least, and I would need to not run out of wipes.

I think for high-touch surfaces in widely-shared environments (e.g. offices), probably disinfecting every few hours or more often is a good idea, and better than the passive benefits of copper. At home I think copper is likely to win out, unless your house is really on the ball about disinfecting things.

Comment by gwillen on March Coronavirus Open Thread · 2020-03-11T23:43:36.136Z · score: 4 (2 votes) · LW · GW

Thanks, will maybe do that today, it's on the queue.

Comment by gwillen on When to Reverse Quarantine and Other COVID-19 Considerations · 2020-03-11T23:43:07.793Z · score: 2 (1 votes) · LW · GW

Wow, heh, we're at N=4 at least then, counting Ray above. Probably future advice about copper tape should clearly document this hazard (I'm not even joking...)

Comment by gwillen on March Coronavirus Open Thread · 2020-03-11T23:39:59.977Z · score: 12 (4 votes) · LW · GW

That's an interesting question that seems like it ought to be able to be checked numerically.

I made an attempt using this simulator of the fairly-naive "SIR" model of disease transmission:

http://www.public.asu.edu/~hnesse/classes/sir.html?Alpha=0.3&Beta=0.07&initialS=1000&initialI=100&initialR=0&iters=50

Note that this simulator appears to be someone's class project. However, its behavior seems to track more or less with what I'd expect. But I'd love for someone with more experience to reproduce this relatively simple model and check it.

You can read about the model at https://en.wikipedia.org/wiki/Compartmental_models_in_epidemiology#The_SIR_model .

I have limited confidence that I've understood it correctly, so take this for what it's worth. It looks to me the time step used in this simulator is one day. So the gamma parameter (rate of recovery per unit time) should be (Wikpedia says) 1/D where D is the duration of the disease. (For transmission modeling purposes, this should be the infectious duration, not the duration of symptoms.) I chose gamma=0.7, meaning D ~= 14 days, semi-arbitrarily, based on https://www.medrxiv.org/content/10.1101/2020.03.05.20030502v1 (which says 10 days after start of symptoms) and the general figure of 14-day quarantines.

The beta parameter is the transition rate from "susceptible" to "infected" per person infected per unit time. (That is, betaI is the transition rate overall.) I think therefore R = Dbeta (the total number of new infections per person should equal the duration times the number of infections per unit time), so beta = R/D = R*gamma.

All that being said, given those assumptions, here are what I think the plots look like for various R values. (Note that the names of the parameters given in the URL do not appear to match the names in the UI. I think the URL parameter names are just wrong; the model behaves as I would expect it to. It's a very simple model and I'd love for someone to independently check this.)

R=4.82 (beta=0.34) (upper cited estimate from Wikipedia): http://www.public.asu.edu/~hnesse/classes/sir.html?Alpha=0.344&Beta=0.07&initialS=1000&initialI=100&initialR=0&iters=50

R=3.5 (beta=.25): http://www.public.asu.edu/~hnesse/classes/sir.html?Alpha=0.25&Beta=0.07&initialS=1000&initialI=100&initialR=0&iters=50

R=2.28 (beta=.16) (estimate based on the Diamond Princess data, https://www.ncbi.nlm.nih.gov/pubmed/32097725): http://www.public.asu.edu/~hnesse/classes/sir.html?Alpha=0.16&Beta=0.07&initialS=1000&initialI=100&initialR=0&iters=50

R=2 (beta=.14): http://www.public.asu.edu/~hnesse/classes/sir.html?Alpha=0.14&Beta=0.07&initialS=1000&initialI=100&initialR=0&iters=50

So it looks to me like very substantial curve-flattening ought to be possible, based on this simplified model, at quite realistic R values. Whether it's possible to flatten it enough to prevent health system overload is anybody's guess -- likely not everywhere -- but it looks like there are substantial benefits possible.

Comment by gwillen on When to Reverse Quarantine and Other COVID-19 Considerations · 2020-03-11T08:13:33.410Z · score: 4 (2 votes) · LW · GW

"Be careful when applying, since the edge of a copper sheet can cut you."

Did you discover this independently, or was this from my, uh, case report on Facebook?

Comment by gwillen on Coronavirus is Here · 2020-03-11T07:18:57.672Z · score: 7 (4 votes) · LW · GW

Well, my level of annoyance at you for making misleading and poorly-sourced claims is definitely coming to a middle.

Mildly annoying is your vague specification of "end of spring" as the endpoint. More annoyingly, after spending some time investigating your newest vaguely-sourced claim, I found that Wikipedia tells me -- on the page https://en.wikipedia.org/wiki/2009_flu_pandemic_in_the_United_States -- that the US had 593 deaths, as of September 3, 2009. Someone else helped me find https://www.cdc.gov/media/transcripts/2009/t091112.htm, which gives an estimate of 3,900 as of October 17, which is still under 4,000 and implausibly late for "end of summer". (I don't think this is really material to the outcome of the bet, but it is a demonstration of exactly the same sorts of issues I've had with your previous comments.)

However, if you're still willing to make the bet, still 1:1, using the actual figure of 593, and let's specify May 31 as what seems to be the consensus last day of spring, I am happy to do that. $50 is fine, but it has been pointed out to me that betting on the coronavirus is considered to be in poor taste, which seems fair enough. So I ask that, if I win, my winnings be in the form of a donation to the Against Malaria Foundation, or another GiveWell top charity of your choice, with receipt provided. For your side, I would suggest something similar, but I will pay cash if you so desire.

Deal?

Comment by gwillen on March Coronavirus Open Thread · 2020-03-11T02:04:30.238Z · score: 16 (7 votes) · LW · GW

What are the issues involved in receiving delivery food during this pandemic?

Can one safely receive and eat delivery food as follows: Avoid contact with the deliveryperson (have them leave it outside), carefully dispose of the packaging in the same way you would for a package delivery, then take the delivered food and reheat it in the oven for a time/cooking temp that will kill the virus?

The respiratory viruses as a family do not appear very resistant to heat (as compared to e.g. some of the foodborne illnesses.) From https://www.quora.com/At-what-temperature-does-the-cold-virus-die/answer/Thomas-Basterfield (I didn't check the citations yet), it seems like 70C for 25 minutes will kill most respiratory viruses thoroughly. This is such a low temperature that I wonder if hot food is inherently inhospitable to them even without the reheating step. My oven dial doesn't even go that low. (Getting the center of the food to this temperature could be challenging without using a higher oven temperature, but you really only need to do the surface; the center has already been cooked, and any relevant contamination will be on the surface from post-cooking handling.)

(There is also a mention of autoclaves, which apparently require <300F and <20 minutes at 15 PSI over atmospheric to sterilize surgical instruments. (Eyeballed numbers, do not trust.) In fact, I was able to find a published paper which concludes that a pressure cooker can be used in place of an autoclave with good results, https://www.ncbi.nlm.nih.gov/pubmed/12267939. So a pressure cooker (e.g. an instant pot) seems like another route to ensure delivery food is heated to a temperature that will kill viruses, although it's probably overkill and is likely to damage the texture of some foods, but is maybe a better option for foods that would dry out objectionably in the oven.)

(It also seems very likely to me that a microwave could achieve sufficient conditions for this, although I think it would be harder to be confident about it without at least having a meat thermometer.)

Does this seem like a reasonable approach?

[One thing I still have to figure out: I have been receiving "ready-to-heat" meals from https://www.freshly.com/. They are precooked but refrigerated, and require reheating. I'm not sure what my risk here is, and how much I can reduce it by just heating them longer in the microwave than the instructions call for.]

Comment by gwillen on March Coronavirus Open Thread · 2020-03-11T00:51:13.894Z · score: 2 (1 votes) · LW · GW

I would also like to know the answer to this.

One thing I'm not sure about: how hard is it to get your hands on HCoV-OC43? With high confidence and in quantities suitable for pretty much guaranteeing to give someone a cold / some immunity? (Do excessive quantities lead to a more severe cold?)

This does really seem like something someone should be working on. Probably someone is, somewhere...

EDIT: Here is one paper on the consequences of HCoV-OC43 infection:

https://www.ncbi.nlm.nih.gov/pubmed/23337903

Among other things: "Recent studies have suggested [that human coronaviruses] can cause severe lower respiratory tract illnesses in children." and "In our population, HCoV-OC43 infections generally caused upper respiratory tract infection, but can be associated with lower respiratory tract infection especially in those coinfected with other respiratory viruses."

So safety might be in question.

EDIT 2: Scihub link: https://sci-hub.tw/10.1097/INF.0b013e3182812787

EDIT 3: I would really love for someone who knows things to take a look at this paper actually, and help interpret it. It is only studying children, and notes that "HCoV-OC43 infections tend to occur before 2 years of age" (does that mean adults can't get it? or they aren't exposed to it much? Does exposing them to it generate a useful immune response?), and also that, among the children selected for the study, children with HCoV-OC43 had better outcomes than controls (but I have no idea how to normalize this for statistical issues; the subjects were children who tested positive for HCoV-OC43, whereas the controls were children who were tested for respiratory viruses but were negative for HCoV-OC43.)

Comment by gwillen on How useful are masks during an epidemic? · 2020-03-10T23:38:09.051Z · score: 4 (2 votes) · LW · GW

Wrong link, should be https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3306645/.

I'm curious whether "normal" is meant to simulate a proper OSHA-approved fit, or "how normal people wear masks". My impression was that the OSHA fit test standard for N95 masks was meant to achieve an extremely high-quality seal -- they test with volatile substances, and you fail if you can smell them. (https://www.osha.gov/video/respiratory_protection/fittesting_transcript.html) It seems like the "normal" setting here is probably a lower standard than what an OSHA fit is supposed to achieve, whereas "fully sealed" (with tape) is possibly higher.