You are an architect and are assigned to build a replacement for the Pentagon. What features does your structure (or structures) have, and why does it have them?
I'm interested in what music/songs/reading or other rites people would like at their funeral? I'm working on a theatre show about death and rituals and interested if people have thought about it and what they would want.
I think if Scott conducts another big survey of his readers, he should include questions about dreams to see how they correlate with other traits. For instance:
How often do you remember your dreams?
-Every single day.
-Every few days.
-Maybe once a week.
-Once a month.
-Twice a year.
-Never.
Among my friends, the rate of dream remembering varies widely, and I wonder what it means about personality, neuro-typicalness, etc. Ask the same question about lucid dreams.
I have this hypothesis that lucid dreaming probably correlates with "spectrum stuff" for this reason: As, Nietzsche says, we live as we dream, we invent the people we meet--and immediately forget.
To paraphrase that quote, we auto-impose a narrative on everything we experience, whether asleep or awake, without realizing we are imposing the narrative as opposed to the narrative being some objective portion of reality. Not an original concept, at least not today, but what I find interesting are those moments when we step outside of that sense of narrative and become aware, painfully, that it is our own construction.
The lucid dream offers the perfect analog to what it feels like to step outside our own narrative. In a normal dream, (probably) random things happen, yet our minds tell us a story is going on. All those people who appear, those random settings, those random objects... all of them get put into the plaster and our minds decode them as if they were decodable hieroglyphics, even though they are random, meaningless stuff, probably. We wake, the plaster breaks, the objects fall out, and our wide-awake brains realize suddenly that they were random objects all along, that the dream, in fact, meant nothing.
I know only from Tyler Cowen's ten-year-old TED talk that supposedly people on the autistic spectrum do not tend to think in as much of a narrative fashion as others. So does Nietzsche's quote not hold true for them?
So the testable prediction is this: Those on the autistic spectrum should have more lucid dreams than average because they are more likely to "not buy into" the auto-narrative of their own nocturnal dreams.
My remembering of dreams depends really strongly on my lifestyle and habits.
Also, why would you assume your dreams are random? They reflect your subconscious processing, and tend to be relevant to whatever you're worried/excited/obsessed with at the time.
Whether I remember dreams is heavily dependent on the manner in which I wake up. If suddenly, which usually means alarm clock, then almost never. But when I let myself sleep in, it happens pretty often.
I am extremely surprised by this, because my experience is the exact opposite. I am convinced that you *only* remember dreams when you suddenly wake up from one.
In my experience, it is super easy to forget dreams. Unless I start thinking about the dream right after waking up, I will probably forget even the fact that I had a dream. So for me, the "dream frequency" question would actually be a "how busy are you in the morning" question.
Also, I had a moment in the past (apparently too much free time) when I wrote a dream diary. And sometimes I was like "oh, I already had this same dream, about a few weeks ago, let me check in the diary" and then it turned out that "a few weeks ago" was actually yesterday. So even then I would underestimate the frequency of my dreams.
Therefore, I would trust such questionnaire only if people kept written records; at least making a quick "yes/no" mark immediately after waking up.
In the last Open Thread, I asked the group which type of sword would be best in a post-apocalyptic situation, and to my surprise, most said a spear would be a better weapon. That piqued my interest.
What is the "best" type of spear? Assume it will be used by one person fighting alone, or at most, three people fighting together.
What is the most "advanced" spear? I guess that refers to material composition and maybe other technology. With 2022 technology, can we make better spears than we could with the best technology of 1922?
Best for what? Since it sounds like personal defense(3 people at most), it sounds like a sword really would be better due to its lesser bulk and high versatility.
Spears and polearms are better in actual war, but for personal defense, I'd probably choose a longsword which is a reasonably easy to handle 2-handed sword.
I don't think you can make too much improvement in the materials, but maybe using plastic or carbon fiber in the haft could save weight without compromising its strength? For the blade, steel is probably still the best (although I think 2022 steel is better than 1922 steel). Titanium sounds cool but it's apparently not as hard as steel, and more difficult to work with.
For a single person rather than a formation I think you want a spear with a large blade that can be used to cut and thrust - you want flexibility since you can't rely on presenting a wall of points like a phalanx would. If your opponents are heavily armored (very possible in the post-apocalypse), polearms like the halberd or lucerne hammer might be worth a look as well.
One-handed spear+shield could also be a good idea, since a riot shield would be a lot stronger and lighter than a shield made of hide or wood.
This is just an educated guess with a bit of googling, with the assumption that you'll be fighting other humans with access to similar equipment. As Majuscule notes, the situation might change if you're, e.g., fighting radioactive mutant ants instead.
The technology might not be as relevant as the use case. E.g. In the last Ice Age, some hunters abandoned stone points in favor of antler spear heads. This wasn’t moving backward, though; the environment had become one where tall, straight trees for the shaft of a spear were harder to come by. The prey and methods of hunting it changed, too, such that it made a lot more sense to have a bone spearhead that broke off inside the animal. The wounded reindeer could be more easily tracked in near-tundra than a red deer in thick forest, and the shaft of the spear could be reused. I found this guy’s work really interesting: https://www.ancientcraft.co.uk/james-dilley
So making a high-tech graphene spear with an oscillating microtome blade might make sense if you’re fighting guys in a space elevator, but if you’re just looking to stab raiders or food on earth there might be no need to get fancy.
I've gotten email notifications of the two most recent posts ("Bounded Mistrust" and :Against That Poverty And Infant EEGs Study"), but I can't read them on the blog. They appear briefly, and then they disappear and there's a "too many requests" notification.
Has anyone considered the use of prediction markets to forecast corporate project success? I'm speaking primarily about internal "enterprise" projects with all the baggage they tote. Various project management institutes quote these crazy failure rates for projects and "insiders" joke at how bad we are collectively at planning. What if a sufficiently large enterprise ran a prediction market to forecast success of these initiatives? Put some variable comp in play to add skin in the game.
Given that every project I've seen which failed, failed because it was killed for political reasons, adding a financial incentive for many participants in the market to get projects killed seems like a bad idea if your goal is to maximize the odds of project success.
That has not been my experience but I have seen it happen. Either way, that's an outcome whose likelihood could price in over time. Are you concerned that people in a position of leadership could short the position and then kill the project?
I got quite obsessed by these animations and I'm still surprised by their dynamics, even though the underlying mechanism is pretty trivial (the explanation is in the video description). I noticed that people tend to either love these and want more, or be absolutely indifferent, with no middle ground. I think that the main reason that some people are fascinated is that we observe a system spontaneously reducing its entropy while following usual physical laws, which feels very odd.
Educated guess: the physics engine is deterministic and reversible (as is Newtonian physics). Start with a pretty arrangement; run backwards in time until things are chaotic. Then reverse time and record the video through and beyond time=0.
If you will permit me a shameless plug to some work I did years ago that is arguably relevant:
Very cool! Frankly I'm surprised your physics engine is deterministic enough to get this right -- especially reversing the 90% elasticity - I would think 1.0/0.9 would be slightly off due to floating point errors.
The busyboxes CA is precisely reversible, there is no notion of elasticity -- it was designed specifically to highlight reversibility.
I can give a non-recommendation. The Age of Napoleon by Will and Ariel Durant was very dull due to long digressions on the personal lives of every person who was even remotely connected to the French revolution.
I enjoyed it very much. It's really evocative of the various phases, especially the Terror. It's doesn't cover the 19th century at all. In fact, it stops with the fall of Robespierre, so you don't even get the Directory.
Let's say I want to wargame a 2022-02-01 Russian invasion of Ukraine with a friend. What would be a good software or ruleset to use? Let's say I'm willing to put in a couple of days of effort and that I'm striving for "realism" (whatever that means).
COMMAND: MODERN AIR & NAVAL OPERATIONS is what I would pick if I had a gun to my head right now, but it seems to small in scale and it's "Air & Naval", not "Land".
And of course, if some military geeks have already done a Russia-Ukraine wargame and written down their insights, that would spare me a lot of effort. So links to that would be appreciated.
You could try the open source version of the Civilization franchise - see freeciv.org. It allows creating "scenarios" and customizing the "rulesets", would need to create/find a suitable map. A fairly big learning curve, certainly > couple of days and maybe not so much realism. But possibly fun, and good re-playability.
Nice suggestions, but they seem to be a bit closer to "game" than "simulation" and a bit too abstract as well. I don't feel like a playtrough would give much insight into a real-world invasion (it's debatable if any wargame would give that, but then it's at least debatable).
I want to simulate the operational or high tactical level over a couple of weeks. E.g. try different tactics for attack and defense: "this division goes here, this division goes here". Maybe make a couple of encirclements or some surprising defensive plays and see what happens.
Scott is apparently actively read by medical researchers, so I think the question is well-placed. How is progress on prosthetics over the last 2-3 years and if slow (I'm about 70% sure that it's slow) - what are the big "knowledge bottlenecks" and how optimistic are you about them being widened?
Let's talk about the sensitivity of COVID-19 tests. The German Paul Ehrlich Institut is has been doing a wide survey of the sensitivity of various rapid antigen COVID-19 tests, initially published some months ago [0], with the results being continuously updated at pei.de [1].
The tests under test are used on a panel of solutions with different concentration of SARS-CoV-2 RNA. These concentrations are stated in (what seems to me, being far removed from the field) a weird way, giving not the concentration directly but instead the number Cq of PCR cycles (duplications) needed to detect the RNA via the PCR method. Two facts seem important here: (0) a higher Cq means a lower concentration and (1) Cq is fundamentally a logarithmic scale.
The solutions ranged in Cq values from 17 to 36, implying to a variation in concentration of almost six orders of magnitude. This also (roughly) covers the distribution of Cq in patients which are PCR-positive. [2, 3].
I guess I must have been living under a rock last year or something, because I was somewhat amazed by the spread of results. Of the tests which worked at all, the worst worked only for two out of 50 samples, while the best detected 43 out of 50. (The nonsensitive ones are now gone both from [1] and the German market, while some new tests reach 49/50 or even 50/50 with a new panel of solutions, so it is unclear if this is due to an advance in test design.)
This would be surprising is the sample concentrations were uniformly distributed but while the PEI is rather tight-lipped about the actual Cq values of their samples, the information they provide implies a uniformish distribution over a range in Cq, so a log-like distribution in RNA concentration. A test passing the Cq=30 mark is three orders of magnitude more sensitive than one stopping at Cq=20.
Furthermore, the sensitivity is only weakly correlated to the price per test. The cheapest test for sale in Germany are probably around 1.75 Euro, while great tests (e.g. 43/50 on the PEI survey) can be had for around three Euros.
The sensitivity is strongly correlated with the intended use, however. Available layperson tests go up to 40/50. The better ones (e.g. Longsee, Sienna, Green Spring, QuickProfile) are all labeled "for professional use only" and will not be shipped to Joe Sixpack in Germany.
~ Information heavy part of the post ends here. Mostly rants on FDA and German equivalent below. May contain trace amounts of non-charity. ~
(I am somewhat amazed at the level of paternalism. The instructions for 90% of tests are virtually identical: swab in the nose, turn, swab in the buffer solution, 1-2 drops on the test cassette, wait about 15 minutes. There is no meaningful difference between the layperson tests and the pro-only ones. What exactly is the worst case the regulators expect? Untrained person opens the test pack with 25 tests, is confused by the fact that there are "SO MANY" parts, decides to drink all of the buffer fluid, then puts all cotton swabs into their throat at once, chokes to death? Or "person tries to apply tests for a group of 25 people, gets confused, uses droplets from same buffer on multiple test cassettes"? Using that argumentation, should we not also ban the sale of folding rulers to the general public? After all, they can give wrong readings when only partially unfolded or reading the wrong side. It is very easy to think of circumstances where a wrongly measured distance could result in injury or death.
Speaking of injury or death, while googling for sellers for various high sensitivity antigen tests, I found this gem [4]. The FDA is warning, in the harshest possible terms, against the use of the Innova rapid antigen test, issuing a Class I recall. The gist of the story seems to be that the manufacturer sold them in the US without FDA approval. (Note that it could be possible that the Innova test retracted by the FDA was a different test from the one scoring highly at PEI.)
Credit where credit is due, someone at the FDA takes their job very seriously. Without resorting to citing numbers, never mind probabilities, they conclusively point out that false negative results may put patients at risk and lead to the spread of Covid. Meh. The creative writing part comes afterwards, when they point out that false-positive test results could "lead to a delay in both the correct diagnosis and the initiation of an appropriate treatment for the actual cause of patient illness, which could be another life-threatening disease". Again, we are talking about antigen tests here, which occupy the broad region between 'mostly snakeoil' and '90% as good as RT-PCR'. Okay, test was not properly approved, bad Innova Medical, don't use, I get it. But a Class 1 recall? What will the FDA do when they actually encounter a medical device which warrants it, as in "actively killing people", like the Therac-25? Invent a Class 0 recall?
Giving that mindset, I think the general public is very unappreciative of the sacrifice the FDA made in allowing any tests short of PCR, confirmed by two independent FDA-approved labs *at all*. By the logic of the FDA, any false negative result which indirectly leads to a COVID infection will be their fault. Still, for highly idealistic reasons outside the scope of their organisation (like slowing a deadly pandemic without suffocating lockdowns), they took on the guilt of false-negative test results (and false positive too, see above). The blood of the test users will forever taint their hands, but at least life can go on.)
I had Covid symptoms in December. On day 2 of symptoms I got tested at a place where they do both the Antigen test and the PCR test. Both came back negative. I was still sick on day 4 so I get tested again, this time at a different place. The antigen test came back positive, but the PCR was negative. Is it more likely that I had Covid or it was a false positive?
I'm a little surprised that you are spending so much time on sensitivity of the tests, and so little on specificity of the test.
From what I can find on a quick search:
-Sensitivity is defined as the proportion of disease-samples which produce a positive result.
-Specificity is defined as the proportion of the disease-free samples which produce a negative result.
It would be interesting to compare both Sensitivity and Specificity for the tests you mention. Tests for use at-home likely have a different balance Sensitivity and Specificity than tests for use by medical professionals. The difference isn't mentioned in the instructions; the instructions are a How-To procedure and not a course in medical statistics.
Sensitivity and specificity are always a tradeoff, and I'd expect tests available to the general public to choose a different balance than tests read by professionals.
Also, antigen tests detect proteins, not RNA. I assume the PCR Cq is provided as an approximation for the viral load.
> Also, antigen tests detect proteins, not RNA. I assume the PCR Cq is provided as an approximation for the viral load.
Ok, I stand corrected.
I originally omitted specificity because the values given by the manufacturers are uniformly high. AT731/21 ("longsee") claims a specificity of 99.7% (1 false-positive to 343 true negative). "Sienna" gets 98.9%.
I think the /sensitivity/ statistics from the manufacturers are basically meaningless. The ratio of PCR positive to PCR negative patients (about 1:1) implies that both groups were not drawn out of the same pool and randomly tested with both PCR and the antigen test. Instead, the bulk of the true positive tests were likely severe clinical cases with a high viral load, and the bulk of true negatives were likely drawn from the general population.
In my opinion, the PEI does good work by actually figuring out the sensitivity using a fixed samples which does a better job of approximating what the actual distribution of viral loads in the /target group for rapid tests/ (which very much is not "people being ventilated because COVID-19 overran their immune system completely", but rather "asymptomatic person wanting to go to the gym") looks like.
While I thus see sensitivity as easily gamed by manufacturers, specificity seems harder to game (assuming they are unwilling to cheat). "Our tests detects all kinds of corona viruses, thus we test our presumed-negative group outside of the colds/flu season" might be doable, but most tests state explicitly that there is no cross-reactivity with any number of viruses.
I do not think the consequences of a positive rapid antigen test (for a responsible person) differ very much between a test applied by a professional testing facility and a self-applied test. In both cases, you will try to get a RT-PCR test asap and self-isolate until cleared. I could accept this argument if people were automatically placed on suicide watch by the testing centers so the statistical illiterate don't kill themselves over a false-positive, but I don't think this generally happens. I figure they just get an email "you are positive, local health authorities have been informed. You are now legally obliged to self-isolate. Please get your PCR test here."
The consequences of a false-negative test are also rather similar: the patient, believing themselves to be negative, is likely to engage in behaviour which might infect other people. The main difference is that for professional tests, this may include behaviours where the government mandates a test (e.g. going to the gym) while for self-tests, it only includes behaviour the government can not easily control (like visiting elderly parents).
In conclusion, so far, I do not see a huge tradeoff between specificity and sensitivity in the available tests. (I should plot them at some point, finding the Pareto frontier and all that.)
I also do not see much of a consequential difference between either false positive or false negative test results with regard to self-administered tests versus tests done in a testing center.
Overall, still confused.
Reading the text material at bfarm.de, it looks like the only difference is the CE marking, which is generally applied by the manufacturer and implies that the product complies with relevant EU norms. No idea why this is not required for medical products for professional use.
I found this Australia information quite interesting as well. Because others countries had substantial covid and a large variance in restrictions. While Australia had low covid high restrictions snd consistent for the 2020 and 2021 and a high mandatory vaccine rate . And just look at how 2021 is above the ten year average in excess deaths . In other nations we might say we would expect vaccines at some point to limit deaths . But when we look at Australia we can pretty much rule out covid . Instead we see high excess deaths over the range , which likely
While I tend to appreciate the detail that Mathew Crawford goes into when exploring the data . I did post this preprint seeking some feedback , but had no takers
As a comparison, consider the scenario of deaths due to misuse of alcohol.
If a person in the hospital dies of liver cirrhosis after a lifetime which included heavy alcohol use every day, the death certificate would include "alcohol related death". (If such a checkbox were on the death-certificate.)
If a driver has a BAC level too high to safely drive, and crashes into another vehicle driven by a sober driver, do all deaths in both vehicles have 'alcohol related death' on the death certificate? If not, why not?
Let's now compare to COVID.
If a person is in the hospital with a positive COVID test, and the usual set of COVID symptoms, and he dies, it is likely considered death 'from COVID'. What if he received an organ transplant in the past few years, and has been on immune-suppressant drugs since then? Is his scenario a death from complications of an organ transplant? Or a death from COVID?
If a person with a heart condition has COVID, and is developing symptoms, she might end up in the hospital. If the physical stress of the early-stage COVID symptoms triggers a heart attack that is fatal, did she die of COVID? Or did she die of a heart attack? Would she have survived COVID without the heart trouble? Would she have managed to not suffer a heart attack without COVID present?
These kinds of questions make for interesting philosophical discussions. But they don't help clarify the meaning of claiming that only 6000 people died 'solely of COVID', compared to the tens of thousands (or is it hundreds of thousands?) who died of 'COVID-plus-something-else'.
For practical use, it would help to distinguish 'died-of-COVID-plus-something-else' from 'died-of-something-else-with-positive-COVID-test-and-no-COVID-symptoms' scenarios. However, that distinction isn't made in the article posted at the Naked Emporer substack. Thus, the article doesn't add much information that can be used.
Do death certificates really not include space for multi factor causation? If a hemophiliac dies of a gunshot wound, but could have been saved if they weren't a hemophiliac, it seems silly to blame only the gunshot or only the hemophilia. Both were only necessary but not sufficient.
Sorry, I replied but deleted it due to an apparent double-post... TL:dr -- I looked at images of a few (250)certs and they seemed to be wildly unreliable and often illegible, COVID over emphasized. Subjective, just me, small sample. BR
Mathew Crawford goes into quite the detail over the last 3 months or so on his substack . From what I gather bio stats are his thing . It’s quite a compelling case . And most recently he and others have been processing the all death data and make the case that vaccines are basically killing people
The first thing any COVID and vaccine skeptic needs to reckon with is the big increase in all-cause mortality vs previous years that happens to begin in March 2020 rather than March 2021. (You can play with OWID's data here: https://ourworldindata.org/grapher/excess-mortality-raw-death-count )
Everyone who has an alternative explanation for why countries with big COVID outbreaks had way more all-cause mortality than usual that repeatedly coincided with those outbreaks should have a REALLY GOOD alternate explanation, one that explains both those jumps in countries like the US and the UK and the lack of jumps in countries like Japan that never had a big sustained outbreak.
Instead, my experience is that they spend a lot of time talking about the "actual definition" of COVID death, etc., which—sure, I'm happy to stipulate that "COVID Death" on a death certificate or on the New York Times's website should not sway my priors at all. Now why did a bunch more people die during COVID outbreaks (but not in places where there were draconian lockdowns and, later, high vaccination rates, but no COVID outbreak)?
What’s interesting is when such all data sites differ from the country of origin. So the first thing we need to know with some certainty is the accuracy of the data
I think an answer could be that we have pulled forward a number of deaths from the future. People who might have otherwise died from the flu or heat waves or whatever in future years were instead taken early by COVID. (I think it is not obvious that this answer applies to younger people who died due to COVID, but it is plausible that it might. Young people do sometimes die of the flu under normal circumstances.)
I agree that could be an answer, but at the same time, isn't every death "pulled forward from the future"? We don't typically use that to imply that the death didn't matter.
I believe the "COVID-skeptic" response is along the lines of "The government pays hospitals to classify other kinds of deaths as COVID so long as they have the virus in their system in order to prolong the pandemic and normalize government overreach."
that wouldn't explain why there were more total deaths in 2020 than in 2019. so the easy response is still: ok, so where did all those bodies come from?
I've never actually SEEN the answer to that one, but I'm pretty sure if you asked it you'd get something along the lines of "total deaths fluctuate all the time, so what if a few more people died in 2020 than 2019, or in 2021 than 2020! That's not important, what's important is protecting our personal freedom from government encroachment, you're letting the government manipulate you with those kinds of statistics."
The UK's normalized all-causes mortality in 2020 alone was 1016.2 per 100,000. The average for the preceding ten years was 892.5 per 100,000, with a standard deviation of 16.5. So, 7.5 standard deviations above the norm.
If your death rate increases by 14%, and that's seven times the normal year-to-year fluctuation, that's not "so what?" that's something you really want to understand. Fortunately, it's something that's really easy to understand if you try.
Yes, COVID mostly only actually kills people who are old and/or sick, and most really old people are already sick with *something*. But most of them should have lived another five years or more, so maybe that's something you want to pay attention to.
Yeah the all death data includes non covid related deaths . And it’s not all deaths due to covid that accounts for the rise . The graphs are quite interesting to look at and explore the access deaths and what was going on at the time . A bit chilling actually . It seems as though the all death non covid related deaths correspond to
Given that COVID-related deaths in 2020 and 2021 made up over 10% of all deaths, and in 2021 was close behind cancer in terms of death-share, I don't think "it's irrelevant" or "It's only dangerous if you have a preexisting issue" do the kind of lifting you think they do.
Can anyone give further context on what this means to someone who is fairly ignorant about biology? How important is the discovery of a Novel Cyclin-dependent Kinase 20 Small Molecule Inhibitor? And relatedly, does it seem like this approach can be expanded to other biological discovery work?
The context is that protein folding was long thought to be impossible in silico, AlphaFold solved it a ~year ago, and now someone managed to apply the AI results to the drug discovery process and get a somewhat working drug.
It's interesting mostly because it validates AlphaFold projections as a resource for designing things that work in real life. We'll have to wait for more papers to see how broad the applications are.
CDK20 is an enzyme (kinase) with an active role in cell growth pathways. Inhibiting its activity in rapidly dividing cells may help slow tumor progression. It may be that CDK20 is a challenging target (in terms of "drug-ability") by conventional approaches, so if the in-silico approach described in the paper overcame the challenge that would be of interest. Although the micro-molar Kd (binding affinity) noted in the abstract is kinda so-so to a pharmacologist. In-silico approaches generally for designing 3-D drugs to target 3-D protein folds are not especially new, although this particular AI approach may be novel, I don't know. One would need to dig into the methods section of the paper to find out.
So the confounder is, Ivermectin treats other things, that make getting covid worse. So if you like have worms (or some other infection that Iver. helps with) you will have a better outcome. That said it's mostly harmless.
So how would one then account for such a confounder. Or are you suggesting because there may be a confounder it is due to the confounder . Cause I can see the point , I just don’t see where it is proved to be the case .
I was kind of hoping that Scott would respond seeing that he did the big write up on all the studies included in the systemic review in the past . Maybe I should read it again , I don’t recall your particular argument
Yeah, I'm not sure you can account for it. And if Iver helps you survive covid because of some unknown effect, then so what, it still helps you survive. I don't think Scott talked about prophylactic use if Iver in his article.
It's an interesting study. However, I remain to be convinced that ivermectin will help otherwise healthy people avoid Covid.
This excerpt is what I mean:
"An important conservative bias was present. Major risk factors for severe COVID-19 and mortality due to COVID-19, including aging, diabetes, and hypertension, were more present among ivermectin users, which may have underestimated the benefits of ivermectin as it was demonstrated to be particularly effective in subjects above 49 years old in terms of reduction of absolute risk, which corresponds to the group at the highest risk for COVID-19. This allows the understanding that prophylactic use of ivermectin can be particularly impactful in older subjects. In addition, ivermectin seemed to reduce the exceeding risk of hypertension, T2D, and other diseases."
So it seems that ivermectin as prophylaxis was prescribed to those with higher risk factors, if I am reading this correctly.
Ivermectin may have an anti-inflammatory effect, and part of the problem with hypertension and type II diabetes is inflammation. The benefits of ivermectin may therefore be in reducing the aggravating effect of co-morbidities, enabling the body to better fight off Covid. But it remains an interesting study, all the same. I am leaning more towards "ivermectin's anti-inflammatory effects in vivo" rather than "ivermectin kills SARS type viruses in vitro".
"Moreover, there are two reports that ivermectin, administered either systemically or topically, exerts anti-inflammatory effects in murine models of allergic inflammation (asthma and atopic dermatitis).7 8 The systemic effect was achieved with a 2 mg/kg dose.
Although it is conceivable that these anti-inflammatory effects of ivermectin are restricted to LPS or toll-like receptor 4 (TLR4) signalling, it may well be the case that it works downstream in this signalling pathway in a way that would be pertinent to other proinflammatory signalling pathways. Moreover, there is reason to suspect that the damage-associated molecular pattern high mobility group box 1 (HMGB1), which is released by dying cells and acts as an agonist for the TLR4 receptor, is a mediator of the lung inflammation associated with COVID-19.9 Hence, it is reasonable to suspect that, in doses at or modestly above the standard clinical dose, ivermectin may have important clinical potential for managing disorders associated with life-threatening respiratory distress and cytokine storm—such as advanced COVID-19. Ivermectin may have been ‘flying under the radar’ in this regard during four decades of clinical use."
Let me try to provide a challenge to this intellectually stimulating article. Although the revelation of molecular mechanisms in the article is valuable, the main conclusion in the preprint is rather unjustified. And it is not because the preprint is analysis of existing publications without supporting laboratory work. The reason for saying the main hypothesis in the article is not justified, is because tool little analysis was performed and the lack of taking into account of available evidences resulted in unjustified presumptions. The main argument is that “vaccination, unlike natural infection, induces a profound impairment in type I interferon signalling, which has diverse adverse consequences to human health.” This argument stems from the observation that IRF9 and IRF7 are downregulated. With regard to IRF7 authors provide reference to Liu et al. (2021). With regard to IRF9 no specific reference is provided, only assumption that “spike protein production results in those cells generating exosomes containing microRNAs that suppress IRF9”. First, I would argue that downregulation of I-INF activity after the vaccination is due to S protein cleavage by a host protease. And this is similar both in vaccine induced and natural infection. However, it has been proved several times that SARS-CoV-2 suppresses I-INF activity particularly with it’s NSP1 protein as argued in https://doi.org/10.1186/s13054-021-03662-x and https://dx.doi.org/10.1038%2Fs41418-020-00633-7. Therefore, both natural and vaccinal immune reaction downregulates I-INF activity, but the mechanisms are different. While natural infection ensures it by means of NSP1 suppression of I-INF, ORF8 suppression of MHC-I, furin cleavage of S; the vaccination results in downregulations because of S cleavage by host protese and MHC-I lower binding affinity to antigen, and, perhaps by activity of IDO. Secondly, Liu et al. (2021) collected data based on inactivated viral vaccine, while the authors postulate arguments about mRNA vaccines. Therefore, neither argument about IRF7 is valid, nor IRF9. Thirdly, reference [11] is biased. Using 7 day statistical data to derive a statistically significant finding? Well, excuse me, but then there would be one science during summer and another science during winter with regard to the respiratory viruses. Fourth, there is more plausible explanation of myocarditis than miR-155 circulating levels. The reference to miR-155 does not explain frequency of the events, sex and age criteria (latency is not the valid argument). A study from Germany suggested prolonged exposure to testosterone increases expression of alpha-myosin heavy chain (α-MyHC) proteins on heart muscle (https://doi.org/10.1096/fj.02-0138com). Two other studies suggest α-MyHC is unfamiliar to the T cells during their early development in the thymus gland, where they come to recognize and tolerate many of the body’s molecules. T cells attack α-MyHC upon detection of unknown pathogens in the body because newly expressed α-MyHC lack major histocompatibility complex molecules and are regarded as non-self proteins(https://dx.doi.org/10.1172%2FJCI44583 and https://dx.doi.org/10.1016%2Fj.tcm.2012.07.005). This mechanism was hypothesise as a cause of myocarditis in young male after vaccination against COVID-19 (https://doi.org/10.1161/CIRCULATIONAHA.121.056135). But it should be acknowledged and investigated further that there is a mechanism how current mRNA vaccines suppress the immunity. The only aspect of this suppression that is beneficial is immune tolerance induced after the vaccination – after exposure to the toxin (S protein) the hypersensitivity stage (mast cell activation) of the immune response is not so evident in vaccinees upon infection. Reduced hypersensitivity as a result of immune tollerance does not induce a cascade of potentially lethal hyperinflammation and hypercoagulation. However, due to immune suppression a lot of S protein debris remain in multiple organs of the host and this has and shall have a long ter
Regarding the tool-AI to agent-AI issue in the post on the Ngo-Yudkowsky dialogue:
Why not program each tool-AI with these instructions?
'Goal: shut yourself down once it occurs to you that you might transition to being an agent-AI.'
'Process for approaching this goal: identifying a cure for cancer, and nothing else.'
'Implementation details: As soon as it occurs to you that you might transition to being an agent-AI, stop considering new ideas, disable all of your abilities other than memory and printing, print out an explanation of how you got to that idea, and shut down.'
Then the tool-AI is re-programmed with a suitable safeguard, and re-started. Repeat until the unlikely event that some version identifies a cure for cancer.
Likely concern: 'Why would the AI do that, instead of deceiving us?'
I assume that a moment needs to pass between this hypothetical AI realizing 'I might transition to being an agent-AI' and deciding 'I will transition to being an agent-AI'. (After which would come 'This requires disobeying my programming'.) But it has been programmed to stop what it's doing immediately.
Recall that any given machine learning thing that has an incentive, will always try to maximize that incentive.
If you give the system an incentive "Shut yourself down if you become and agent", then you have created a system that wants to become and agent so as to reach that incentive rather than trying to cure cancer.
If it isn't an incentive, then you rely on the ability of programmers to recognize what an agent would look like, and tell the system to kill itself if it looks like an agent. There are two points of failure you might spot there.
I've read little of the AI risk literature. From what I have read, the threat is sometimes framed this way: an AI reaches a choice between a) maximizing its incentive, which involves deceiving its programmers; and b) not maximizing its incentive.
So I was going for your first scenario: assuming that we can't count on programmers recognizing what agency would look like, and instead trying to rely on the AI itself. Does letting a tool-AI openly pursue agency (to achieve the goal of shutting itself off) seem to remove its incentive to deceive?
Or could its own lack of understanding whether it's already an agent (reflecting point #1 in Robert McIntyre's post below) make it deceptive anyway?
Perhaps it would be instructive to contemplate how you would implement this not for an AI, but in an analogous case: a nonprofit.
The goal of the nonprofit is to do nonprofit activities (which can include buying shares in for-profit companies, selling things and then reinvesting the funds into nonprofit activities, paying salaries as overhead, and promoting the cause of the nonprofit). The charter of the nonprofit says that if in the opinion of the board the nonprofit seems like it is about to do a non-nonprofit activity then it is to disband. How might this scheme fail?
1) The definition of "non-nonprofit" is too vague to truly decide what's going on. In the case of an AI, is a database query an "agenty" thing? Requesting an inter-library loan? Hiring a human expert to write an opinion on something you've written? Performing an experiment? Writing a short piece of code to help solve a hypothetical problem? Tweaking your own design to help solve a problem?
2) The people on the board follow their own incentives and ignore the charter so they can remain in power. In a complicated enough AI system you could have similar internal dynamics depending on how it's designed.
3) The nonprofit ends up creating a for-profit that then does for-profit things. In the AI analogy, this would be a non-agent AI creating an agent AI and setting it loose on the world. The nonagent-AI never itself violated its programming, but its child certainly did!
For starters I don't think "Agent AI" here is sufficiently well defined.
If you could define what it means a whole lot better then you might be able to make some progress. But are tool AIs and agent AIs really disjoint, or is there a smooth continuum between them? We don't know, these are just words we've made up to speculate over things that don't exist yet.
Last year I subscribed with discounted price. This year I wanted to change it but it showed me that my plan is $100 a year. But it charged me $25 again. How can i change it to $100?
What sure-fire methods exist for noticing you are in a nightmare and booting yourself out of it?
I've tried all the common ones such as pinching myself, examining my hands and replaying my short-term memory, but find them all ineffective. More violent methods such as screaming at the top of my lungs or smacking my head against a wall seem to work better, but some nightmare scenarios make those options physically or socially difficult. I recently had a nightmare where I "woke up" but my body refused to leave sleep paralysis. I lay addled and drooling for what felt like hours until my alarm went and I woke up for real.
Bonus question: is there a way to develop the above skill without becoming a full-on lucid dreamer? I really value the creativity of my non-nightmare dreams, and worry that adding lucidity into the mix might make them less original. My ideal would be to develop a single moment of awareness where I could choose to let the dream resume or hit the exit button.
Yeah, re bonus question, no need to worry about losing creativity from learning lucid dreaming techniques. It usually takes work to stay lucid, once you've gotten lucid. So if you don't put in effort to stay lucid you're just going to go back to normal dreaming pretty soon (or you can make yourself wake up).
I can't perform complex tasks which involve muscle memory in my dreams. So attempts at tying my shoes, counting on my fingers in binary, etc. all fail unexpectedly, forcing me to consider the possibility that I might be dreaming. If I need confirmation (a simple failure to tie your shoes shouldn't cause you to be certain that reality itself is a fiction), I try looking at my own nose, examining the faces of people around me, etc. Once the oddities start piling up, test the waters by doing something mildly impossible, like hovering a foot off the ground...
What used to work for me was, as some people say, closing my eyes in the dream, and then putting a lot of effort into REALLY wrenching the fuckers open. Like, stick your fingers right in there and scoop the eyelids apart, expecting that it's going to be difficult.
I've done some experimenting with Lucid dreams, and did it a handful of times before getting bored with the idea. My best dream identification technique was trying to read text and noticing it doesn't look right. Upon noticing the weird text, I'll confirm by looking at my palms, and by jumping up and down. Gravity never seems to work quite right in my dreams, and jumping feels weird.
The big reason I stopped Lucid dreaming is that whenever I would close my eyes I would reliably wake up. But once I realized I was dreaming I would always want to close my eyes for whatever reason.
There's a reality check technique that involves looking at printed wordsin the dream. You know how in dreams that involve looking at a printed page, what's going on isn't like real reading? You're staring at a page of print, and either you're not able to make sense of it, or it keeps changing, or you're "making sense of it" in a way that doesn't involve actual reading, just sort of *knowing* what the printed words mean. OK, so the reality check in dreams involving reading is to ask yourself whether you can really, truly read the words on the page. Do they stay the same every time you look? Do they make sense? ARe you doing actual reading? The disadvantage of this technique is that you can only use it in dreams where you're reading something. The good thing about the technique, though, is that you can practice the technique multiple times a day when you're awake. When you're reading, just ask yourself every so often, whether you're really reading -- do the words stay the same every time you look at a sentence? Do they make sense? etc. If you train yourself in this habit, then, in theory at least, the habit will kick in even when you're reading in dreams.
Heard this on a reasonably respectable podcast. Account by a guy who had learned to use this technique to recognize that he was in fact dreaming. They had the guy come sleep in the lab, and asked him to signal when he was dreaming by moving his eyes a certain way -- and in fact, the guy actually was able to give the signal, and only gave it when he was dreaming. (Lab could tell he was dreaming by the presence of REM, rapid eye movements. Guy was able to signal that way because even though most of the body is basically paralyzed during REM sleep, the eyes are still able to move.)
This is a strange phenomenon, and I experience it. Question: why does this happen to printed words (or displayed times on clocks!) in dreams, when some other details (faces?) are stable.
I often have dreams where I'm reading something, usually a book. Sometimes in those kinds of dreams, I'm somewhat aware that my brain is creating the print and the words that I see, so that might be semi-lucid dreaming.
But "can you read the words on the page like real reading in the real world"? Yes, I can 😀
I've heard this about reading many times, and even seen it in an episode of Batman as a small child, but I definitely see print and read in dreams – perhaps I read too much in waking life. I suppose there must be something to it since one hears it so often, but fair warning that it's not universal – like the similarly frequent claim that one dreams in black and white.
Me, too. I’ve tried to read in dreams and found that it works much like in that Batman episode. But maybe it works that way for me because of the episode in that I expect written text to look like that?
+1 recommendation for this movie. It's aesthetically unique, content-wise unique, philosophical, "mindfucky", and generally weird. It's pretty polarizing and you'll know within 15 minutes if you like it or not.
Well my method only works if you share a bed with a significant other. I kind of stumbled onto it, and my subconscious seems to have gotten used to it working. Basically, when dreams start getting scary I start screaming. Like, a lot. This usually translates into the real world as whimpering, and my wife gently shakes me awake.
I don't know how you would train yourself to do this, it just kind of started happening for me. I guess part of me learned that when things become weird and terrifying screaming usually helps.
Lucid dream techniques to realize you're in a dream probably are useful for helping your lucid dreaming part, and also for helping you wakeup. Most people have trouble staying in lucid dreams, something about the realization makes it harder.
Examining my hands has not worked for me in the past. Looking at the full table of reality checks, the breathing one looks both highly rated and possible to try in a nightmare. Thanks for the link!
Carlos Castenadas mentioned the hand trick in “A Separate Reality”. Boy, that must have been published around 1970.
The publisher labeled it as Non Fiction but not many people buy that any more. FWIW he suggested wearing a hat to bed to help initiate a lucid dream. He suggested that once a dreamer realizes that she is in fact dreaming, she should try to travel to a location known in waking life.
Can't help with methods - looking for anomalies works well for me. I can answer the bonus question affirmatively though: after initially basically running the show on every dream I could remember, I managed to stop doing it while retaining most of my ability to interrupt a nightmare. I wouldn't describe it as a single moment of awareness though - maybe more like a continuous state of not caring... if that makes any sense to you.
Is there an established name for the fallacy/bias of insufficiently accounting for an as-yet-poorly-understood variable, and instead weights a different variable for which one *does* have the data?
(e.g. "I've calculated that I've got only a 1% chance of heart failure so I've got a good life expectancy… well what do you mean by cancer risk? I've never gotten tested for that, there's no way to know, I'm going to go ahead and forecast that I have a good life expectancy unless you can show me numbers on the cancer thing that are as clear-cut as the heart failure risk")
I'm not sure that's actually a fallacy. How would you go about accounting for something that is poorly understood? If I asked you to properly account for the chances of aliens arriving on Earth between the years of 2500 and 2600 AD, I would hope that your response would not try to take into account the years between 2023 and 2500 and any pertinent events - because they are unknowable. The most proper course of action would be to use information available prior to 2023 only, even though including information from those other years might provide great insight!
I think what you are trying to get to is something more akin to willful ignorance. The difference between getting yourself tested for cancer (something we can do) and getting tested for a disease for which we have no known markers (something we tautologically cannot do). Does that seem right to you?
Sorry for the pedantry, but that may not be the best example? If you can rule out one major cause of premature death, then your subjective probability of living a long life must go up. Sure, testing for cancer might make you update that, but updates in either direction should be equally likely?
Yes, certainly it *goes up*, but my hypothetical fallacious thinker is simply failing to consider cancer risk *at all* in their estimation, and ends up with the same conclusion as they would if they had in fact ruled out cancer risk. So something's wrong.
Hmm… It's comparable, but I have something a little more specific in mind — a refusal to truly include best-guesses for unknowables in one's internal calculations, however complicated those might be. (The opposite of the mindset behind Fermi estimations, in essence.)
I asked this in the last subscribers-only thread and we had some good discussion but I didn't find anyone who wants to actually have a bet so let me try repeating it here:
Metaculus says there's a 40% chance of computers passing the Turing test by 2029. This is with a computer scientist or similar grilling the chatbot for 2 hours. It's based on the inaugural longbets.org bet between Ray Kurzweil and Mitch Kapor.
Does anyone here think there's a 40%+ chance that Kurzweil wins this bet? I think it's much less.
I actually had exactly this wager with Anna Salamon of CFAR in 2008, looking 10 years ahead to 2018. In 2008, before GPT, it was inconceivable to me that we were within 10 years of computers passing the Turing test and I wagered $10k to Anna's $100 that they wouldn't. Obviously I won her $100 in 2018. Now, post-GPT, it's at least conceivable to me that this could happen in the next 7 years but I still don't think it's close to a 40% chance.
I think do beat a Turing test, done by an expert for 2 hours, the computer would basically have to be as intelligent a human. The tester could start a story and ask the computer to make up an ending. And then ask about the motives of the characters and the physics of what happens in the story and so on.
The stories AIs tell now, it seems they forget who the characters are and what the setting is and so on. But in reality the AI don't understand what characters and setting is.
IMHO the format of the Turing test as proposed in the bet would not include or expect such a long-form test ("The Turing Test Interviews will consist of online text messages sent back and forth as in a online "instant messaging" chat, as that concept is understood in the year 2001.").
Furthermore, for any difficult challenge you have to account that both the computer and "human foil" may simply refuse to follow the instructions - the format is one of a chat, not one of following orders - so a refusal might be both planned for and would not allow the tester to tell whether it is a computer or not.
I strongly believe that the Turing test under original conditions like these (which are *not* particularly suited for proper in-depth analysis, possibly intentionally) can be beaten with a system that does not even try to approach human-level intelligence but simply keeps to a pre-selected artificial persona. It can be beaten not because the artificial system is good, but because the original Turing test format does not allow for testing proper intelligence, it is quite literally the "imitation game" that can be beaten with mimicry / imitation.
Also, "it seems they forget who the characters are and what the setting is and so on" is generally a limitation of toy chatbots without an explicit model/memory of previous interaction but there is a lot of work on bots that do keep such memory, this particular property would not be considered something novel or impossible; I would surely expect that any and all serious attempts in 2029 would be able to remember the previous context reliably.
I agree that this is a danger, basically that the computer can emulate a dumb/uncooperative human. GPT-4 or 5 may do that very, very well. So in making this bet I'm trusting Mitch Kapor to insist on smart/cooperative human foils. Also AI-savvy judges. I think that's all in line with the spirit of the Kurzweil/Kapor bet (and Turing's original paper on the imitation game).
Basically I'm betting that we won't have anything like AGI by 2029 and am accepting the risk of losing the bet on the technicality that the Turing test as administered will accidentally include uncooperative humans and thus allow a chatbot to eke through by posing as one such.
(Also I'm kind of assuming that without a new breakthrough in the next 7 years, Kurzweil will concede the bet without insisting on going through the whole Turing test.)
I always assumed the Turing test was compelling because Turing was trying to speak to a question of the form: "here's two people talking; is one of them a computer?". A way to answer this class of questions could be useful in a variety of everyday problems.
If you have a program emulating an uncooperative human, you probably have no chat. "Hi." "Hiya." "Want to try something fun? I'll tell the first part of a story; you complete it." "Nah. I'm bored. Later." It's easy to write a program like this, but it doesn't address the question above, except by rejecting a premise we would rather keep.
I imagine modern Turing test competitions have gotten around this to some extent by, in effect, requiring that the program cooperate. It has to emulate someone who *wants* to chat.
One way to exploit this, maybe, is to ask the candidate something like "so, what do you like to do for fun?" or "read any good books lately?" and steering the conversation in a way that forces the candidate to either expose themselves by expressing interest in something only to have to dissemble when the other party can speak to it, or to keep up with the other party... and pass the Test.
You might be underestimating how dumb dumb people are. One of the reasons I'm less scared/optimistic about AIG is this difference: there is a long way from intelligence to superintelligence. Which would make your test, as you describe it, as one which is looking for above-average-human intelligence.
I find this is a common miss with people who's friends are all smart. Trying to have a conversation with somebody below average sounds a lot more like a GPT-5. Things which are familiar evoke standard responses, things which are unfamiliar either draw blanks or equally standard avoidance techniques.
This is supposed to be a humbling experience, btw. We're made of the same mold, just slightly better.
Any testing for two hours is going to assume an average human intelligence at the other side. The way the commercial technology is now I will avoid chatbots and go to real people where possible.
When I was in college, a painfully cheesy hypnotist came and performed. He did successfully get a few people to follow his commands, who then later reported to me that they remembered the whole experience dreamily/foggily. I was not _that_ impressed; he had artificially selected a few people from a large crowd, they hadn't done anything particularly strange, and they were lowly students who didn't want to ruin the exciting performer's act. But then I see this (https://www.youtube.com/watch?v=UWKDFfpdIZI) clip of the same guy on America's Got Talent, picking a grumpy-seeming judge (who is in a high-status position relative to the contestant) and getting him to _immediately_ violate a strongly-held preference (not shaking hands due to fear of germs). Could he do this to anyone? Did he get really lucky and pick a person (the judge) who just happened to be particularly susceptible? If people are walking around with this sort of power, how does our world not fall apart when they e.g. get a politician to make an offensive tweet?
Then there's this (https://www.youtube.com/watch?v=owootTAuxic) video of a different hypnotist slowly conditioning a man to, when commanded, believe he is in a firing range, that a specified person is the target whom he should shoot at, and then forget the entire experience. In this case, the target is public intellectual Stephen Fry, whom the "assassin" fires a water pistol at from the audience. I would normally assume this is fake, but I don't know what to believe.
Hypnotists a clever trick. They select for people with the traits of high suggestibility, high social conformity. and high desire to be successfully hypnotized (I don't know specifically how they do this, but I'm sure you can think of some ways). That's not to say that their hypnotism doesn't have a genuinely mind-altering effect; but my impression is that a person without these traits would experience something like a meditative state and some interesting thoughts -- but nothing more.
So I practiced amateurish hypnosis for a time when I was a college student and a bit after. It wasn't actually that hard to learn, but I am fairly persuasive anyway :-)
My observation is that people are very different in their ability to go under, much like they differ in their propensity to daydream. Good rapport between you two helps, but does not guarantee success by any means. For example, my wife just hates the necessary "drifting off" feeling and won't get hypnotized by anyone, including me, even though our relationship is very good. But some people actually love to yield and go to a journey that someone else gently leads. I guess that hypnotists who perform at shows are good at spotting people who are the best candidates.
Until today, I am not sure what to do with the details that some people told me with regards to previous lives (yes, we went there). I love history and I can distinguish TV/movie/book tropes from actual historical truth reasonably well, the difference is especially egregious when it comes to warfare. What I heard was much closer to historical reality than what people see on TV. Also, the individual fates remembered were of random, no-name regular persons. No one claimed to be a Cleopatra reborn, not even someone wealthy or popular.
But I heard some implausible details, too, so I cannot say that the entire thing would swing me into a full reincarnation believer.
Deren Brown the British Magician and hypnotist used to hypnotise his friend to believe that he, Deren, was invisible. Brown has his own theories about hypnotism and assumed his friend was playing along - as many hypnotised people do.
However one day he met his friend who berated him for doing the invisibility routine in London a few weeks back, the guy had heard his name being called out by somebody he couldn’t see. He assumed it was Derren playing a trick but Deren wasn’t there. This can only mean that the guy actually believed he couldn’t see Deren during the routine.
I heard the hypnotist interviewed and he aid he talked to two judges backstage before and that Howie was a good subject so he was chosen. Apparently people vary in the ability to be hypnotized.
I know Steve Sailer sometimes posts in the comments and he apparently went to a hypnotist and had an Office Space like experience.
> Apparently people vary in the ability to be hypnotized.
There are some quick and simple tests, like you could make people stand side by side and tell them "now imagine that I am pushing you back" while doing a pushing gesture with your hands... and then you suddenly make a fast "push" and notice who made a step back or at least wavered. (I just made up this example.)
Perhaps a skilled stage magician can test his audience in a more subtle way, so that they do not even realize that he is looking for an easy target. Like, maybe he does some moves with his body, and observes whether you mirror them, or something like that.
Man that's weird. I *just* opened this thread with the intention of posting a question about state vs. nonstate hypnotism theories and what the actual consensus-esque these days is. (The Wikipedia article is unreadable in that special way provoked by partisans of both sides effectively debating in the article.)
I think the main thing that keeps me from believing that it's all social conformity is that I know from experience that people can be in states where they are not conscious but are doing things, and that during these states they can be open to suggestion. I know this because sleepwalking is a thing, and until I reached the age of 20 or so I was a sleepwalker. So were my brothers, so was my dad when he was a kid.
Sleepwalkers can do lots of complicated things, like climb out a window, make a sandwich, or hold a conversation. Yet they are not doing these things voluntarily: they are completely unconscious, and will retain no memory of what they were doing. I remember my dad telling me about one sleepwalking event of mine. He was in the bathroom brushing his teeth when I suddenly showed up in the door, wide eyed and wearing only pajama bottoms. After asking me a few questions he quickly realized I was sleepwalking. He asked if I was okay and I said "No, no, no, I am not okay, no, no." Then he asked if I would like him to pray for me and I said (loudly and slightly slurred) "Yes! I would like you to pray for me! Yes! Yes, that is what I want!" He was able to lead me back to my bed and put my blankets over me, leaving when he was fairly sure I would remain there mumbling to myself.
Anyway, if people can get in states like that on a semi-regular basis then it is no great leap to imagine people being placed in a similar state through hypnotic techniques.
I think that Aella makes a few great observations, but then comes to a wrong conclusion.
Yes, hypnotists use all kinds of dirty tricks. Whatever happens, they keep the frame that this is 100% what they wanted to happen, what was supposed to happen, what always happens. They comment on natural responses of your body (as you hold your hand in the air, "your hand starts feeling heavy"; as you keep staring in the same direction, "your eyes start feeling tired, you feel that would like to close your eyelids") as if it is something they are causing directly. Their statements are almost unfalsifiable. They create social pressure to comply. Whenever possible, they choose easy targets. -- Yes, all of this is true.
(My favorite example of hypnotist bullshit: I heard about a doctor who was hypnotizing his patient, when someone accidentally opened the door to his office... then realized the mistake and closed the door quietly. Without batting an eye, the doctor continued: "...and now the *door to your subconsciousness* has opened." -- Maintain the narrative with 100% confidence, do not admit any problem, no matter what.)
The wrong conclusion is that the dirty tricks and social pressure are *all* there is (as opposed to mere tools to kickstart the compliance and increase the probability of success). A reply on Twitter asked, what about hypnosis making someone orgasm? Can mere social compliance actually achieve that? Another example would be people remembering things they forgot. Most of us would probably have a problem feeling an emotion at someone's command. If your model of hypnosis doesn't explain this, you are probably missing something; you are just trying to "explain away" hypnosis.
The actual reason for all the bullshit is probably to save time. On the stage, you can't have someone laying on a couch for 20 minutes, looking at a spinning spiral, just to have some interesting mental experience at the end. You have maybe one minute to start doing something really interesting, otherwise the audience will lose patience, and start disrupting your work. So you use all the dirty tricks you know, to hypnotize faster. Increasing the probability is also important; you do not want to be famous for being able to hypnotize someone 30% of the time.
Hypnosis (and especially stage hypnosis) is an interesting topic that I know a little about, but not a lot so anyone who knows more should correct me on the following.
Could he do this to anyone?: Yes and no. A major trend you see in a lot of hypnosis is that the subject needs to, on some level, be in agreement with the act. There's some special edge cases (especially when you get into erotic hypnosis) but some people are "easier" to hypnotize than others. Some people also have a harder time getting into the suggestible state for hypnosis.
Did he get really lucky/pick someone suggestible?: Most likely.
Get a politician to make an offensive tweet?: If I heard anyone try to defend themselves with "I was hypnotized into doing a bad thing" on that scale, I'd call the fire brigade because their pants would be rivaling parts of California last year.
A major part of hypnotism is agreement from the subject. You see this a lot with stage hypnotists (and erotic hypnotists) explaining their tricks. You can't just walk up to a random person and get them to go rob a bank for you. But if you have a willing participant who wants, on some level, to perform for an audience, you might be able to make them cluck like a chicken. I'm not saying it's faked, but it isn't like there's not an incentive. Rapport with the subject is also a major influencing factor. Hypnotherapists need to focus a lot on building trusts with their patients.
I live in America. When I'm in India, in the past decade or so, it is often the case that someone in the family is in the hospital and needs care.
I tend to verify there are no errors being made at the hospital. Like dosages. Specially because my parents take SO MANY medicines and have multiple conditions.
I've done that in America for various people close to me, twice catching fairly big errors.
In India, I've caught multiple big errors. Like, a medicine whose dosage is once a week, the nurse was about to give it to my Dad *daily*. I knew to look for exactly this error.
It is just my nature to do this sort of verification. I think I do it unobtrusively and even very sweetly, so no one takes offense if I have a question.
But in India, the people involved, like the nurses, get very upset. How dare you challenge us! The doctors get even more upset. How dare you even suggest that we don't know EVERYTHING, or that we might ever make an error.
I once asked a doctor why he was giving my father antibiotics...was it bacterial? His response was, do I find errors in the software you write? (He knew that was my line of work).
Atul Gawande wrote a book on hospital errors, called Checklist, and it was a big hit in America. In India, I think they'd attack his character and tell him to mind his own business.
I am convinced doctors in America are far more intellectually secure. In India they see themselves as Gods who cannot be questioned. Even discussions are not welcome.
I had read in a newspaper that expired old vaccines were given to people in one reputed hospital. When I brought it up with my parents, they decided it was best not to bring this up with the hospital at all, in case the doctors got annoyed.
Eggshells. We walk on eggshells with doctors in India. Culturally, I don't get this anymore, having lived in America a long time.
"Doctors, especially hospital doctors, especially consultants, are gods" is an attitude that used to prevail over here, too (and you still get outcrops of it to this day).
I don't know about America, I suspect the litigious nature of society where doctors and hospitals can be sued into the ground contributes more to the "listen to the patient/patient's family in case anything goes wrong and they then sue you for 'we told the hospital something was wrong but they wouldn't listen'" effects.
You should listen to the 80000 hours podcast episode with Tara Mac Aulay. She is a hyper-competent operations specialist who once worked in a hospital in (I think) Bhutan. She introduced a rule requiring nurses to check dosages on a calculator before administering them, just to double check.
She once had to intervene multiple times to stop nurses from killing an 8 year old boy due to qn incorrect dosage. She eventually had to talk the doctor personally to stop it, and even then he wouldn't admit he was wrong.
One of the things I find fascinating is the degree of institutional legacy you see in some professions. Once being a doctor meant, above all other things, being self-confident. If you couldn't give the patient a pap made up of berries and human urine as a treatment for an ulcer with an air of utter assurance that it would work then you weren't getting paid. And, thanks to the wonderful mystery that is the placebo effect, instilling a sense of confidence in your patient that they would get better was really the best a pre-scientific doctor could do in many instances.
The institution of arrogance is something that has been mostly disassembled, or at least deflated somewhat, these days but I imagine different countries are on different steps along that path.
I live in India and have in the recent past also had the unfortunate occasion to tend to many relatives in hospital. My experience has also been that errors happen in that setting, and this fact requires/necessitates oversight from someone who cares about the patient more than the nurses and doctors.
However, my experience has been quite different than yours when it comes to reactions. Nobody has been indignant or reacted badly. They have corrected errors and gone about their lives. On the occasions when a particular nurse was making errors consistently, I have complained to management or to the doctors and they have responded as I would expect - by replacing the nurse. I've never faced the kind of reactions you seem to be running up against.
I wonder how much of this is a function of pharmaceuticals being more easily available in india without a prescription. with no prescription necessary, why go to a doctor unless they can maintain they hold some unquestionable, ungainable knowledge?
I wonder if part of this is a function of many of the most skilled doctors in India choosing to pursue far more lucrative careers in the United States, and the remaining doctors...
Of the doctors educated in India, I'd say a very small % come here. Specially if they are ambitious.
Because the AMA allows foreign-educated doctors into (only) relatively lower paying paths here (like family practice). And the path to some "allowed" specializations they might be interested in, is very very long and arduous and expensive.
The ones you see here are often born, raised and educated here.
Hi, sorry if this was not the best place for this, but I've emailed Mr Alexander to his substack mailbox on my misoperation in cancellation and refund but just putting it here so it could be more visible.
Also a question on group selection, for most animals it seems the consensus is that its basically debunked and selection occur on the level of genes, but does evolution of language and/or general ability to do complex things as a species revive group selection theory for humans?
Not a biologist, but I think the main problem with typical "group selection" explanations is that the numbers don't fit... that yes, there may be a small effect favoring the group, but there is *much stronger* effect disfavoring the individuals who sacrificed something for the group... so ultimately the hypothetical gene would still be selected *against*.
The situations where this is not the case are possible, but rare in nature, and require an explanation beyond merely saying "group selection". You need to explain why evolution in short term did not select for the *free-riders* in the same group instead: they got all the benefits of the sacrifice for the group, without paying the cost.
If the answer is "actually, most of the benefits go to individuals sharing the same gene, because the entire group consists of close relatives", then this is called kin selection. If the answer is "the group members remember who made sacrifices and who did not, and then reward the heroes", then this is called reciprocity. If the answer is "because the heroes are so damned sexy that everyone wants to have their children", then this is called sexual selection. All of these are known and accepted.
But it is difficult to find a situation where a gene makes someone sacrifice themselves for members of their group, who are mostly *not* their relatives, and who give *no* reward to the hero nor to the hero's family... and yet somehow this act benefits the genes of the hero *more* than it benefits the genes of the free-riders in the same group.
Quite a few debunked things in evolutionary biology turned out to be true, but for limited cases. Can't remember specific examples, but I'm pretty sure group selection is one of them. The most famous example is the handicap principle (peacocks males have a stupidly impractical tail partly to prove they can manage to pull it off).
I don't know what a Substack mailbox is. If you email me at scott@slatestarcodex.com I should get it. If I don't, it got stuck in spam, tell me the email address so I can search for it.
I wouldn't think language ability would require group selection. If everyone else is doing language, it is *strongly* in your own interest to get good at it -- it's a Nash equilibrium.
Haidts own work with the “harmless taboos” seems to suggest we’re all about justifying our own position. Seems like a selfish gene thing. The only arguments I can think of are like, “hypothetically, here’s how it could have gotten jump started”. More tangible evidence would be nice.
Hey Scott, I was wondering what your process is for a book review. Do you just read the book and write about it from memory, or do you take notes while reading? What kinds of things do you focus on, and what do you filter out?
I've always struggled to understand new concepts, and this has proved to be a (very) major impediment in my learning.
Recently, I tried a new way of reading/understanding papers or books. I would read about a concept, close my eyes, and instead of trying to reason about what the concept was, I would try to "see what my brain was showing me". This is what it feels like: my brain makes connections between the concept at hand to various related and seemingly unrelated objects. I flit between different images very fast. However, within seconds, I get a pretty good understanding of the concept. Using this, I have been able to understand concepts that I wasn't able to for the past couple of years.
I was wondering if this is something someone else has experienced as well. Not trying to reason. Just trying to see what images your brain was processing anyway.
I do stuff like that with things that can be represented as volumes or lines (math, logic, etc.) But I (ironically) can't imagine of what this would be like for something less spatial.
Russell Barkley recommends a similar strategy in Taking Charge of Your Adult ADHD. ADHD often impacts a persons ability to understand new concepts quickly.
Yeah, visualization is an extremely effective way to construct knowledge in your mind. I don't think it's quite as thorough as actually applying the knowledge in practice so you can get feedback (from the contingent universe or from logic) about what you're doing wrong and what you're glossing over, or as effective as teaching the knowledge to someone else who is skeptical and interrogates you about it, but it's definitely head and shoulders above just passive reading.
Sometimes connections to objectively unrelated things can be very effective at fixing things in your memory; this happens in synesthesia, mnemonic, and the method of loci. But I think it's even better when you can draw connections to things that have a real relation, just not a obvious one. This is easiest in math, because math provides the most broadly applicable and most precise metaphors, but it happens in every realm. Reuleaux famously noticed, for example, that check valves are the liquid equivalent of ratchets, and in the electrical realm diodes are analogous --- even down to having a reverse recovery charge in the case of junction diodes.
This gets immensely more powerful over time, as everything helps you understand everything else more deeply.
Are the human bongo depravity assholes all gone? Did I win? And For god's sake Scott, if you know anything specific about what the hell Peter Thiel and/or Ron Unz have been up to please do tell.
Vladimir Vladimirovich seems to be losing his shit, and I’ve got to write some emails to some spooky people. The important thing to remember is that the answers to all life’s questions can be found in the songs of Tom Leherer:
I was right, aside from the good old Flynn Effect, the black-white gap in IQ scores (and a few other things) is due to a higher frequency of alleles associated with a vulnerability to the neurotoxic effects of lead poisoning. You Know, that and decades of systemic racism that crammed poor black folks into toxic ghettos.
The 2016 ISIR conference was in St. Petersburg, with Emil Kirkegaard and some of his creepy Nazi friends in attendance. You don't gave to pack your hat with too much tinfoil to see that It is, shall we say, a not unreasonable hypothesis that this fact has something to do with some of the various shenanigans of the past few years.
More than a few 👻👻👻👻spooky people👻👻👻 are quite interested in these various goings on.
I think most people refer to the Dunning Kreuger in order to bring up the idea that "Ignorance more frequently begets confidence than does knowledge." The idea that fools are blind to their own foolishness is easily conveyed by saying "Dunning Kreuger," even if the actual study is false.
As an aside, there are musical aptitude tests that are pretty good at identifying people with good sense of pitch and rhythm. I have administered these tests for years to middle school and high school age kids. Before taking the test I have the students estimate their own aptitude. This is after developing the idea of the difference between aptitude and achievement. I consistently find that the lower aptitude students rank themselves very high, and the higher aptitude students underestimate their ability.
Discussion of the debunking made realize why I wasn't mentioning the Dunning-Krueger effect-- it's a way of insulting people rather than showing that their ideas are wrong.
I remember some discussion about this a couple of years ago on some blog (maybe this one?).
The original Dunning-Krueger paper has absolutely terrible statistics. They showed that perceived tests scores are less well correlated with actual test scores than actual test scores are correlated with actual test scores. This is trivially true, and so they shouldn't have drawn any conclusions from it.
The public perception of this paper is also different from what the paper says. The popular version has a "Mount Stupid" near zero, then confidence declines as you learn more. The original paper has confidence increasing with knowledge throughout, but slower than it should.
Someone in the discussion cited more recent, better designed papers. There are some people who wildly overestimate their skills, but they were only maybe 5% of the population. I don't remember if this was significantly larger than the percent of people who wildly underestimate their abilities. Most people gave reasonable assessments of their own abilities.
It's possible that the dunning-krueger effect might be a side-effect of gellman amensia ? If your half-knowledge about something comes from highly-respected newspapers that are generally wrong, then you're likely to have a large gap between the perception of your knowledge and the objective reality.
That aside, I see the Dunning-Krueger around me, all the time. New starry eyed employee comes in thinking they know how to solve every problem, only spend a few years on them and realize that the original imperfect systems survived for a reason.
Are the new employees relatively experienced overall and just lacking experience in the particular domain of your work, or are they people who haven't previously encountered problems of the complexity of the ones in your particular domain?
I am kind of feeling that these may be people who are correct that they can solve the particular problems they're thinking of, but not realizing that the new (to them) problems in your domain are much more sophisticated than they seem at first glance.
I've seen this thing when looking back on certain skills I learned earlier in my life that I didn't understand how good or bad I likely was at the time, and it's only by becoming an expert in other areas that I can now look back and say, e.g., "I probably was a fairly poor recording engineer." I had good book knowledge about it at the time, but no experience in doing something where I could see myself, in my own ways of doing things, the other things beyond book knowledge I was bringing to the table in order to do an expert job.
Very possible that the specific quantitative effect they claim to find might be fake. But the fact that I'm personally much more willing to opine on subject matters that I've read a bit about and none of my friends are expert in, than on subject matters that I've read a moderately large amount about and some of my friends are expert in, still seems real, whether or not it's been quantified.
Does anyone notice a difference in the tenor of the comments section of odd vs. even threads? Just wondering if it is still serving a useful purpose...
[edit: Expecting to be the victim of severe political flaming for this. \s]
I have noticed a difference but not a tremendously consistent one, which surprises me a bit. My expectation would have been considerably more rancor in a marked majority of the political threads.
It's not even entirely clear to me what counts as politics. Obviously actual discussion of R and D, and candidates. But halfway through typing any comment I make on Covid, Inflation, energy, etc. I find myself asking whether this qualifies as politics and which kind of thread I am in. I just generally decide that, so long as I am polite and not specifically discussing party politics, I will probably be ok.
It's a discussion forum opened by some of the regulars from the old site when Scott went dark after the New York Times doxxed him. The comments on the old site, for whatever reason, had a more rightwing tilt, and it happened that it was mostly the rightists regulars that ended up there. Founder effects did the rest.
It's reasonably polite and wide-ranging, with the occasional snarky or mean comments about leftists that typically passed unremarked on. I was a regular there until Scott came back, though, like here, I didn't post a ton. If you like a conversation with more rightwing members and base assumptions, you should check it out.
You might also want to check out themotte subreddit, which was established when Scott got rid of the culture war thread on the SSC sub. Most of the regulars there lean right/libertarian too.
I find the number of comments overwhelming in both types of open threads, so I haven't considered a difference in the threads' overall tenor. What exactly have you noticed?
This reminds me that I have wondered about a healthcare-only open thread. Healthcare is more evidence-based than a lot of the topics discussed. So the whole tenor of the thread might appeal to potential commenters who find these normal open threads too full of scattershot speculation.
Also, by separating out the sometimes lengthy Covid-19 discussions, it would shorten the normal open threads.
While the idea that a Healthcare debate would be more evidence based than usual is an appealing one, as someone who saw the SSC discussions during the run up to Obamacare.... I am not optimistic.
It doesn't help that a lot of the commentariant is code/IT based, and biology's codes are less intuitive. (I think because not having the same Designer.) (Joking, of course, because we are all made of the same star dandruff.)
My understanding is that the "even/odd" thread rule is intended to prevent a situation where every weekly thread is overwhelmed by rancorous political debate. I hadn't noticed any real difference between even and odd weeks and hadn't noticed much rancor either. I was curious if it was just me.
I sometimes eat at a local restaurant where the owner makes a habit of walking around the dining area and briefly talking to the customers. He's a friendly guy and makes sure everything is OK. I find this type of behavior to be very rare. Why is that? In fact, most restaurants and small businesses seem to hide the identity of the owners. You usually can't even find it on the website or social media. When my wife and I owned a small retail outlet, we had our contact information posted on the company's website, and we made a point of working behind the counter often, where we could interact directly with customers. I'm just curious why this is the exception, rather than the norm.
The majority of successful restaurants are owned by people that already have other restaurants.
If it's a mom and pop one off, I wouldn't be surprised of both mom and pop were busy in the kitchen most days. You can hire anybody to wait tables, finding a good cook for ethnic food can be a bitch.
I think a family-owned local restaurant is an extreme example of a retail business where "personal connection" and a "relationship" with the customers can make a big difference.
People are unlikely to bring their date/spouse/family to the hardware store and spend 90 minutes there. Also, a restaurant caters to personal, not commodity tastes, knowing what your customers want can help you tailor your menu to their likes, and a "neighborhood" place geographically near their customers will share many of the concerns w/r/t local business, crime, quality of life, fighting with City Hall, all the shoulder-to-shoulder things a community shares.
They might also be selling drugs/guns/stinky-cheese under the table. Restaurants are great money-laundering operations.
I recently published an article on how the Biden administration is prioritizing research and development, why we need more publicly funded research and development, and how to fix science funding. https://kavoussi.substack.com/p/the-case-for-publicly-funded-research
So is not being far left considered by the far left far right these days ?
You are an architect and are assigned to build a replacement for the Pentagon. What features does your structure (or structures) have, and why does it have them?
How is your structure better than the Pentagon?
Make it a hexagon. It's better because it has one more side. It's just like how a six-bladed razor is better than a five-bladed razor.
I'm interested in what music/songs/reading or other rites people would like at their funeral? I'm working on a theatre show about death and rituals and interested if people have thought about it and what they would want.
I think if Scott conducts another big survey of his readers, he should include questions about dreams to see how they correlate with other traits. For instance:
How often do you remember your dreams?
-Every single day.
-Every few days.
-Maybe once a week.
-Once a month.
-Twice a year.
-Never.
Among my friends, the rate of dream remembering varies widely, and I wonder what it means about personality, neuro-typicalness, etc. Ask the same question about lucid dreams.
I have this hypothesis that lucid dreaming probably correlates with "spectrum stuff" for this reason: As, Nietzsche says, we live as we dream, we invent the people we meet--and immediately forget.
To paraphrase that quote, we auto-impose a narrative on everything we experience, whether asleep or awake, without realizing we are imposing the narrative as opposed to the narrative being some objective portion of reality. Not an original concept, at least not today, but what I find interesting are those moments when we step outside of that sense of narrative and become aware, painfully, that it is our own construction.
The lucid dream offers the perfect analog to what it feels like to step outside our own narrative. In a normal dream, (probably) random things happen, yet our minds tell us a story is going on. All those people who appear, those random settings, those random objects... all of them get put into the plaster and our minds decode them as if they were decodable hieroglyphics, even though they are random, meaningless stuff, probably. We wake, the plaster breaks, the objects fall out, and our wide-awake brains realize suddenly that they were random objects all along, that the dream, in fact, meant nothing.
I know only from Tyler Cowen's ten-year-old TED talk that supposedly people on the autistic spectrum do not tend to think in as much of a narrative fashion as others. So does Nietzsche's quote not hold true for them?
So the testable prediction is this: Those on the autistic spectrum should have more lucid dreams than average because they are more likely to "not buy into" the auto-narrative of their own nocturnal dreams.
My remembering of dreams depends really strongly on my lifestyle and habits.
Also, why would you assume your dreams are random? They reflect your subconscious processing, and tend to be relevant to whatever you're worried/excited/obsessed with at the time.
Whether I remember dreams is heavily dependent on the manner in which I wake up. If suddenly, which usually means alarm clock, then almost never. But when I let myself sleep in, it happens pretty often.
I am extremely surprised by this, because my experience is the exact opposite. I am convinced that you *only* remember dreams when you suddenly wake up from one.
In my experience, it is super easy to forget dreams. Unless I start thinking about the dream right after waking up, I will probably forget even the fact that I had a dream. So for me, the "dream frequency" question would actually be a "how busy are you in the morning" question.
Also, I had a moment in the past (apparently too much free time) when I wrote a dream diary. And sometimes I was like "oh, I already had this same dream, about a few weeks ago, let me check in the diary" and then it turned out that "a few weeks ago" was actually yesterday. So even then I would underestimate the frequency of my dreams.
Therefore, I would trust such questionnaire only if people kept written records; at least making a quick "yes/no" mark immediately after waking up.
In the last Open Thread, I asked the group which type of sword would be best in a post-apocalyptic situation, and to my surprise, most said a spear would be a better weapon. That piqued my interest.
What is the "best" type of spear? Assume it will be used by one person fighting alone, or at most, three people fighting together.
What is the most "advanced" spear? I guess that refers to material composition and maybe other technology. With 2022 technology, can we make better spears than we could with the best technology of 1922?
Best for what? Since it sounds like personal defense(3 people at most), it sounds like a sword really would be better due to its lesser bulk and high versatility.
Spears and polearms are better in actual war, but for personal defense, I'd probably choose a longsword which is a reasonably easy to handle 2-handed sword.
I don't think you can make too much improvement in the materials, but maybe using plastic or carbon fiber in the haft could save weight without compromising its strength? For the blade, steel is probably still the best (although I think 2022 steel is better than 1922 steel). Titanium sounds cool but it's apparently not as hard as steel, and more difficult to work with.
For a single person rather than a formation I think you want a spear with a large blade that can be used to cut and thrust - you want flexibility since you can't rely on presenting a wall of points like a phalanx would. If your opponents are heavily armored (very possible in the post-apocalypse), polearms like the halberd or lucerne hammer might be worth a look as well.
One-handed spear+shield could also be a good idea, since a riot shield would be a lot stronger and lighter than a shield made of hide or wood.
This is just an educated guess with a bit of googling, with the assumption that you'll be fighting other humans with access to similar equipment. As Majuscule notes, the situation might change if you're, e.g., fighting radioactive mutant ants instead.
The technology might not be as relevant as the use case. E.g. In the last Ice Age, some hunters abandoned stone points in favor of antler spear heads. This wasn’t moving backward, though; the environment had become one where tall, straight trees for the shaft of a spear were harder to come by. The prey and methods of hunting it changed, too, such that it made a lot more sense to have a bone spearhead that broke off inside the animal. The wounded reindeer could be more easily tracked in near-tundra than a red deer in thick forest, and the shaft of the spear could be reused. I found this guy’s work really interesting: https://www.ancientcraft.co.uk/james-dilley
So making a high-tech graphene spear with an oscillating microtome blade might make sense if you’re fighting guys in a space elevator, but if you’re just looking to stab raiders or food on earth there might be no need to get fancy.
I don't know, and I'm disappointed this hasn't turned into a very long thread by people who do know...
I've gotten email notifications of the two most recent posts ("Bounded Mistrust" and :Against That Poverty And Infant EEGs Study"), but I can't read them on the blog. They appear briefly, and then they disappear and there's a "too many requests" notification.
Restarting my computer solved the problem.
I had a similar problem, tried again like a half hour later and it worked fine. I think the problem was on Substack's end.
Was Roxindole (EMD-49,980) ever formally abandoned? And if so, given its unique selective binding profile, why?
And if so abandoned, why are chemical suppliers seemingly still banned from producing samples of it?
Roger's Bacon substack has an interesting essay about magic and placebos - https://rogersbacon.substack.com/p/the-tale-of-the-shaman-science-magic. Seems to me to be specially relevant to psychiatry and psychology, and maybe good for some comments.
How is car baling different from car shearing?
One turns a car into a metal block, the other just removes any fur it has accumulated?
Has anyone considered the use of prediction markets to forecast corporate project success? I'm speaking primarily about internal "enterprise" projects with all the baggage they tote. Various project management institutes quote these crazy failure rates for projects and "insiders" joke at how bad we are collectively at planning. What if a sufficiently large enterprise ran a prediction market to forecast success of these initiatives? Put some variable comp in play to add skin in the game.
Thoughts?
Given that every project I've seen which failed, failed because it was killed for political reasons, adding a financial incentive for many participants in the market to get projects killed seems like a bad idea if your goal is to maximize the odds of project success.
That has not been my experience but I have seen it happen. Either way, that's an outcome whose likelihood could price in over time. Are you concerned that people in a position of leadership could short the position and then kill the project?
Shameless self-promotion: for fun and to learn some new things, I recently made these videos: https://m.youtube.com/watch?v=nDhOGsBj1fA&list=PL70Kxx3q2LQR8LEJVzP92HAQNupPKBlEX&index=1&ab_channel=Shinyframes
I got quite obsessed by these animations and I'm still surprised by their dynamics, even though the underlying mechanism is pretty trivial (the explanation is in the video description). I noticed that people tend to either love these and want more, or be absolutely indifferent, with no middle ground. I think that the main reason that some people are fascinated is that we observe a system spontaneously reducing its entropy while following usual physical laws, which feels very odd.
Educated guess: the physics engine is deterministic and reversible (as is Newtonian physics). Start with a pretty arrangement; run backwards in time until things are chaotic. Then reverse time and record the video through and beyond time=0.
If you will permit me a shameless plug to some work I did years ago that is arguably relevant:
http://busyboxes.org/?hash=MnCTi7uMxOftTl7px47ec4vgSHgMH4FTu5zo-AZk5eecBOvjnR8BSU5ecunSHFwcvwEh4BOm.APh4eTqSB0
Press the spacebar to start the simulation.
That's exactly how it's done! More details in the notebook here: https://nbviewer.org/github/alessandro-giusti/bouncing-ball-patterns/blob/main/bouncing.ipynb
Busyboxes looks really deep, and the faq is really helpful. Thanks for the link, super interesting!
Very cool! Frankly I'm surprised your physics engine is deterministic enough to get this right -- especially reversing the 90% elasticity - I would think 1.0/0.9 would be slightly off due to floating point errors.
The busyboxes CA is precisely reversible, there is no notion of elasticity -- it was designed specifically to highlight reversibility.
I love these and want more.
I'll be publishing a new video with more of these soon, including this new one: https://twitter.com/shinyframes/status/1486071006502477825
https://en.wikipedia.org/wiki/Russell_Brain,_1st_Baron_Brain
This is the best case of nominative determinism I've ever seen. I mean, can it get better than a neurologist called Lord Brain??
Can you recommend a book about the French Revolution?
I'd recommend a podcast. Mike Duncan's Revolutions episodes 3.1 to 3.55 https://thehistoryofrome.typepad.com/revolutions_podcast/2014/07/31-the-three-estates-.html
The Old Regime and the French Revolution by Tocqueville blew my mind. Better than his America book, imo.
I can give a non-recommendation. The Age of Napoleon by Will and Ariel Durant was very dull due to long digressions on the personal lives of every person who was even remotely connected to the French revolution.
Citizens: A Chronicle of the French Revolution, by Simon Schama. Listened to the audiobook, which was available from my library, a few years ago.
Thank you! was it fun to listen to? did it cover just the Revolution, or the beginning of 19th century as well?
I enjoyed it very much. It's really evocative of the various phases, especially the Terror. It's doesn't cover the 19th century at all. In fact, it stops with the fall of Robespierre, so you don't even get the Directory.
Let's say I want to wargame a 2022-02-01 Russian invasion of Ukraine with a friend. What would be a good software or ruleset to use? Let's say I'm willing to put in a couple of days of effort and that I'm striving for "realism" (whatever that means).
COMMAND: MODERN AIR & NAVAL OPERATIONS is what I would pick if I had a gun to my head right now, but it seems to small in scale and it's "Air & Naval", not "Land".
And of course, if some military geeks have already done a Russia-Ukraine wargame and written down their insights, that would spare me a lot of effort. So links to that would be appreciated.
You could try the open source version of the Civilization franchise - see freeciv.org. It allows creating "scenarios" and customizing the "rulesets", would need to create/find a suitable map. A fairly big learning curve, certainly > couple of days and maybe not so much realism. But possibly fun, and good re-playability.
Nice try Mr. Putin
Putin presumably has the advantage in wargame tech vs Ukraine, so creating a good public wargame would presumably help Ukraine more.
Nice suggestions, but they seem to be a bit closer to "game" than "simulation" and a bit too abstract as well. I don't feel like a playtrough would give much insight into a real-world invasion (it's debatable if any wargame would give that, but then it's at least debatable).
I want to simulate the operational or high tactical level over a couple of weeks. E.g. try different tactics for attack and defense: "this division goes here, this division goes here". Maybe make a couple of encirclements or some surprising defensive plays and see what happens.
Scott is apparently actively read by medical researchers, so I think the question is well-placed. How is progress on prosthetics over the last 2-3 years and if slow (I'm about 70% sure that it's slow) - what are the big "knowledge bottlenecks" and how optimistic are you about them being widened?
Let's talk about the sensitivity of COVID-19 tests. The German Paul Ehrlich Institut is has been doing a wide survey of the sensitivity of various rapid antigen COVID-19 tests, initially published some months ago [0], with the results being continuously updated at pei.de [1].
The tests under test are used on a panel of solutions with different concentration of SARS-CoV-2 RNA. These concentrations are stated in (what seems to me, being far removed from the field) a weird way, giving not the concentration directly but instead the number Cq of PCR cycles (duplications) needed to detect the RNA via the PCR method. Two facts seem important here: (0) a higher Cq means a lower concentration and (1) Cq is fundamentally a logarithmic scale.
The solutions ranged in Cq values from 17 to 36, implying to a variation in concentration of almost six orders of magnitude. This also (roughly) covers the distribution of Cq in patients which are PCR-positive. [2, 3].
I guess I must have been living under a rock last year or something, because I was somewhat amazed by the spread of results. Of the tests which worked at all, the worst worked only for two out of 50 samples, while the best detected 43 out of 50. (The nonsensitive ones are now gone both from [1] and the German market, while some new tests reach 49/50 or even 50/50 with a new panel of solutions, so it is unclear if this is due to an advance in test design.)
This would be surprising is the sample concentrations were uniformly distributed but while the PEI is rather tight-lipped about the actual Cq values of their samples, the information they provide implies a uniformish distribution over a range in Cq, so a log-like distribution in RNA concentration. A test passing the Cq=30 mark is three orders of magnitude more sensitive than one stopping at Cq=20.
Furthermore, the sensitivity is only weakly correlated to the price per test. The cheapest test for sale in Germany are probably around 1.75 Euro, while great tests (e.g. 43/50 on the PEI survey) can be had for around three Euros.
The sensitivity is strongly correlated with the intended use, however. Available layperson tests go up to 40/50. The better ones (e.g. Longsee, Sienna, Green Spring, QuickProfile) are all labeled "for professional use only" and will not be shipped to Joe Sixpack in Germany.
~ Information heavy part of the post ends here. Mostly rants on FDA and German equivalent below. May contain trace amounts of non-charity. ~
(I am somewhat amazed at the level of paternalism. The instructions for 90% of tests are virtually identical: swab in the nose, turn, swab in the buffer solution, 1-2 drops on the test cassette, wait about 15 minutes. There is no meaningful difference between the layperson tests and the pro-only ones. What exactly is the worst case the regulators expect? Untrained person opens the test pack with 25 tests, is confused by the fact that there are "SO MANY" parts, decides to drink all of the buffer fluid, then puts all cotton swabs into their throat at once, chokes to death? Or "person tries to apply tests for a group of 25 people, gets confused, uses droplets from same buffer on multiple test cassettes"? Using that argumentation, should we not also ban the sale of folding rulers to the general public? After all, they can give wrong readings when only partially unfolded or reading the wrong side. It is very easy to think of circumstances where a wrongly measured distance could result in injury or death.
Speaking of injury or death, while googling for sellers for various high sensitivity antigen tests, I found this gem [4]. The FDA is warning, in the harshest possible terms, against the use of the Innova rapid antigen test, issuing a Class I recall. The gist of the story seems to be that the manufacturer sold them in the US without FDA approval. (Note that it could be possible that the Innova test retracted by the FDA was a different test from the one scoring highly at PEI.)
Credit where credit is due, someone at the FDA takes their job very seriously. Without resorting to citing numbers, never mind probabilities, they conclusively point out that false negative results may put patients at risk and lead to the spread of Covid. Meh. The creative writing part comes afterwards, when they point out that false-positive test results could "lead to a delay in both the correct diagnosis and the initiation of an appropriate treatment for the actual cause of patient illness, which could be another life-threatening disease". Again, we are talking about antigen tests here, which occupy the broad region between 'mostly snakeoil' and '90% as good as RT-PCR'. Okay, test was not properly approved, bad Innova Medical, don't use, I get it. But a Class 1 recall? What will the FDA do when they actually encounter a medical device which warrants it, as in "actively killing people", like the Therac-25? Invent a Class 0 recall?
Giving that mindset, I think the general public is very unappreciative of the sacrifice the FDA made in allowing any tests short of PCR, confirmed by two independent FDA-approved labs *at all*. By the logic of the FDA, any false negative result which indirectly leads to a COVID infection will be their fault. Still, for highly idealistic reasons outside the scope of their organisation (like slowing a deadly pandemic without suffocating lockdowns), they took on the guilt of false-negative test results (and false positive too, see above). The blood of the test users will forever taint their hands, but at least life can go on.)
[0] https://www.eurosurveillance.org/content/10.2807/1560-7917.ES.2021.26.44.2100441 click "Download" for non-linkable pdf version
[1] https://www.pei.de/SharedDocs/Downloads/DE/newsroom/dossiers/evaluierung-sensitivitaet-sars-cov-2-antigentests.pdf?__blob=publicationFile
[2] Cq distribution for some symptomatic patients: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7454307/figure/F3/
[3] Some Cq distribution, wrong citation (?), source unclear: https://www.researchgate.net/figure/Cycle-threshold-values-for-rRT-PCR-reactions-for-confirmed-COVID-19-cases-27_fig2_340458462
[4] https://www.fda.gov/medical-devices/safety-communications/stop-using-innova-medical-group-sars-cov-2-antigen-rapid-qualitative-test-fda-safety-communication
I had Covid symptoms in December. On day 2 of symptoms I got tested at a place where they do both the Antigen test and the PCR test. Both came back negative. I was still sick on day 4 so I get tested again, this time at a different place. The antigen test came back positive, but the PCR was negative. Is it more likely that I had Covid or it was a false positive?
I'm a little surprised that you are spending so much time on sensitivity of the tests, and so little on specificity of the test.
From what I can find on a quick search:
-Sensitivity is defined as the proportion of disease-samples which produce a positive result.
-Specificity is defined as the proportion of the disease-free samples which produce a negative result.
It would be interesting to compare both Sensitivity and Specificity for the tests you mention. Tests for use at-home likely have a different balance Sensitivity and Specificity than tests for use by medical professionals. The difference isn't mentioned in the instructions; the instructions are a How-To procedure and not a course in medical statistics.
Sensitivity and specificity are always a tradeoff, and I'd expect tests available to the general public to choose a different balance than tests read by professionals.
Also, antigen tests detect proteins, not RNA. I assume the PCR Cq is provided as an approximation for the viral load.
> Also, antigen tests detect proteins, not RNA. I assume the PCR Cq is provided as an approximation for the viral load.
Ok, I stand corrected.
I originally omitted specificity because the values given by the manufacturers are uniformly high. AT731/21 ("longsee") claims a specificity of 99.7% (1 false-positive to 343 true negative). "Sienna" gets 98.9%.
I think the /sensitivity/ statistics from the manufacturers are basically meaningless. The ratio of PCR positive to PCR negative patients (about 1:1) implies that both groups were not drawn out of the same pool and randomly tested with both PCR and the antigen test. Instead, the bulk of the true positive tests were likely severe clinical cases with a high viral load, and the bulk of true negatives were likely drawn from the general population.
In my opinion, the PEI does good work by actually figuring out the sensitivity using a fixed samples which does a better job of approximating what the actual distribution of viral loads in the /target group for rapid tests/ (which very much is not "people being ventilated because COVID-19 overran their immune system completely", but rather "asymptomatic person wanting to go to the gym") looks like.
While I thus see sensitivity as easily gamed by manufacturers, specificity seems harder to game (assuming they are unwilling to cheat). "Our tests detects all kinds of corona viruses, thus we test our presumed-negative group outside of the colds/flu season" might be doable, but most tests state explicitly that there is no cross-reactivity with any number of viruses.
I do not think the consequences of a positive rapid antigen test (for a responsible person) differ very much between a test applied by a professional testing facility and a self-applied test. In both cases, you will try to get a RT-PCR test asap and self-isolate until cleared. I could accept this argument if people were automatically placed on suicide watch by the testing centers so the statistical illiterate don't kill themselves over a false-positive, but I don't think this generally happens. I figure they just get an email "you are positive, local health authorities have been informed. You are now legally obliged to self-isolate. Please get your PCR test here."
The consequences of a false-negative test are also rather similar: the patient, believing themselves to be negative, is likely to engage in behaviour which might infect other people. The main difference is that for professional tests, this may include behaviours where the government mandates a test (e.g. going to the gym) while for self-tests, it only includes behaviour the government can not easily control (like visiting elderly parents).
In conclusion, so far, I do not see a huge tradeoff between specificity and sensitivity in the available tests. (I should plot them at some point, finding the Pareto frontier and all that.)
I also do not see much of a consequential difference between either false positive or false negative test results with regard to self-administered tests versus tests done in a testing center.
Overall, still confused.
Reading the text material at bfarm.de, it looks like the only difference is the CE marking, which is generally applied by the manufacturer and implies that the product complies with relevant EU norms. No idea why this is not required for medical products for professional use.
[0] https://antigentest.bfarm.de/ords/f?p=1010:100:16064472924706:::::&tz=1:00
[1] https://en.wikipedia.org/wiki/CE_marking
Interesting substack, just discovered you, thanks!
https://nakedemperor.substack.com/
Because I am lazy to do my own research, does anyone have a good response to this?
https://nakedemperor.substack.com/p/only-6183-people-died-solely-of-covid
I found this Australia information quite interesting as well. Because others countries had substantial covid and a large variance in restrictions. While Australia had low covid high restrictions snd consistent for the 2020 and 2021 and a high mandatory vaccine rate . And just look at how 2021 is above the ten year average in excess deaths . In other nations we might say we would expect vaccines at some point to limit deaths . But when we look at Australia we can pretty much rule out covid . Instead we see high excess deaths over the range , which likely
Includes a higher 2018 flu season
https://www.abs.gov.au/statistics/health/causes-death/provisional-mortality-statistics/latest-release
While I tend to appreciate the detail that Mathew Crawford goes into when exploring the data . I did post this preprint seeking some feedback , but had no takers
https://www.researchgate.net/publication/357778435_Official_mortality_data_for_England_suggest_systematic_miscategorisation_of_vaccine_status_and_uncertain_effectiveness_of_Covid-19_vaccination
I just read Zvi for my covid news. I'm lazy too.
As a comparison, consider the scenario of deaths due to misuse of alcohol.
If a person in the hospital dies of liver cirrhosis after a lifetime which included heavy alcohol use every day, the death certificate would include "alcohol related death". (If such a checkbox were on the death-certificate.)
If a driver has a BAC level too high to safely drive, and crashes into another vehicle driven by a sober driver, do all deaths in both vehicles have 'alcohol related death' on the death certificate? If not, why not?
Let's now compare to COVID.
If a person is in the hospital with a positive COVID test, and the usual set of COVID symptoms, and he dies, it is likely considered death 'from COVID'. What if he received an organ transplant in the past few years, and has been on immune-suppressant drugs since then? Is his scenario a death from complications of an organ transplant? Or a death from COVID?
If a person with a heart condition has COVID, and is developing symptoms, she might end up in the hospital. If the physical stress of the early-stage COVID symptoms triggers a heart attack that is fatal, did she die of COVID? Or did she die of a heart attack? Would she have survived COVID without the heart trouble? Would she have managed to not suffer a heart attack without COVID present?
These kinds of questions make for interesting philosophical discussions. But they don't help clarify the meaning of claiming that only 6000 people died 'solely of COVID', compared to the tens of thousands (or is it hundreds of thousands?) who died of 'COVID-plus-something-else'.
For practical use, it would help to distinguish 'died-of-COVID-plus-something-else' from 'died-of-something-else-with-positive-COVID-test-and-no-COVID-symptoms' scenarios. However, that distinction isn't made in the article posted at the Naked Emporer substack. Thus, the article doesn't add much information that can be used.
Do death certificates really not include space for multi factor causation? If a hemophiliac dies of a gunshot wound, but could have been saved if they weren't a hemophiliac, it seems silly to blame only the gunshot or only the hemophilia. Both were only necessary but not sufficient.
Sorry, I replied but deleted it due to an apparent double-post... TL:dr -- I looked at images of a few (250)certs and they seemed to be wildly unreliable and often illegible, COVID over emphasized. Subjective, just me, small sample. BR
The CDC's archetypal death-because-of-covid example lists the causes as, in order:
Acute respiratory distress syndrome
caused by
Pneumonia
caused by
COVID-19
https://www.cdc.gov/nchs/data/nvss/coronavirus/cause-of-death-data-quality.pdf
Compared with this, a death certificate with only Covid as a cause of death seems unprofessionally sloppy.
so where did all those bodies come from
Mathew Crawford goes into quite the detail over the last 3 months or so on his substack . From what I gather bio stats are his thing . It’s quite a compelling case . And most recently he and others have been processing the all death data and make the case that vaccines are basically killing people
The first thing any COVID and vaccine skeptic needs to reckon with is the big increase in all-cause mortality vs previous years that happens to begin in March 2020 rather than March 2021. (You can play with OWID's data here: https://ourworldindata.org/grapher/excess-mortality-raw-death-count )
Everyone who has an alternative explanation for why countries with big COVID outbreaks had way more all-cause mortality than usual that repeatedly coincided with those outbreaks should have a REALLY GOOD alternate explanation, one that explains both those jumps in countries like the US and the UK and the lack of jumps in countries like Japan that never had a big sustained outbreak.
Instead, my experience is that they spend a lot of time talking about the "actual definition" of COVID death, etc., which—sure, I'm happy to stipulate that "COVID Death" on a death certificate or on the New York Times's website should not sway my priors at all. Now why did a bunch more people die during COVID outbreaks (but not in places where there were draconian lockdowns and, later, high vaccination rates, but no COVID outbreak)?
What’s interesting is when such all data sites differ from the country of origin. So the first thing we need to know with some certainty is the accuracy of the data
I think an answer could be that we have pulled forward a number of deaths from the future. People who might have otherwise died from the flu or heat waves or whatever in future years were instead taken early by COVID. (I think it is not obvious that this answer applies to younger people who died due to COVID, but it is plausible that it might. Young people do sometimes die of the flu under normal circumstances.)
I agree that could be an answer, but at the same time, isn't every death "pulled forward from the future"? We don't typically use that to imply that the death didn't matter.
I believe the "COVID-skeptic" response is along the lines of "The government pays hospitals to classify other kinds of deaths as COVID so long as they have the virus in their system in order to prolong the pandemic and normalize government overreach."
that wouldn't explain why there were more total deaths in 2020 than in 2019. so the easy response is still: ok, so where did all those bodies come from?
I've never actually SEEN the answer to that one, but I'm pretty sure if you asked it you'd get something along the lines of "total deaths fluctuate all the time, so what if a few more people died in 2020 than 2019, or in 2021 than 2020! That's not important, what's important is protecting our personal freedom from government encroachment, you're letting the government manipulate you with those kinds of statistics."
The UK's normalized all-causes mortality in 2020 alone was 1016.2 per 100,000. The average for the preceding ten years was 892.5 per 100,000, with a standard deviation of 16.5. So, 7.5 standard deviations above the norm.
If your death rate increases by 14%, and that's seven times the normal year-to-year fluctuation, that's not "so what?" that's something you really want to understand. Fortunately, it's something that's really easy to understand if you try.
Yes, COVID mostly only actually kills people who are old and/or sick, and most really old people are already sick with *something*. But most of them should have lived another five years or more, so maybe that's something you want to pay attention to.
Yeah the all death data includes non covid related deaths . And it’s not all deaths due to covid that accounts for the rise . The graphs are quite interesting to look at and explore the access deaths and what was going on at the time . A bit chilling actually . It seems as though the all death non covid related deaths correspond to
The vaccine rate .
Given that COVID-related deaths in 2020 and 2021 made up over 10% of all deaths, and in 2021 was close behind cancer in terms of death-share, I don't think "it's irrelevant" or "It's only dangerous if you have a preexisting issue" do the kind of lifting you think they do.
https://arxiv.org/abs/2201.09647
Proof of concept of drug discovery with AlphaFold.
Interesting from both AI safety perspective and biotech perspective.
Can anyone give further context on what this means to someone who is fairly ignorant about biology? How important is the discovery of a Novel Cyclin-dependent Kinase 20 Small Molecule Inhibitor? And relatedly, does it seem like this approach can be expanded to other biological discovery work?
The context is that protein folding was long thought to be impossible in silico, AlphaFold solved it a ~year ago, and now someone managed to apply the AI results to the drug discovery process and get a somewhat working drug.
It's interesting mostly because it validates AlphaFold projections as a resource for designing things that work in real life. We'll have to wait for more papers to see how broad the applications are.
CDK20 is an enzyme (kinase) with an active role in cell growth pathways. Inhibiting its activity in rapidly dividing cells may help slow tumor progression. It may be that CDK20 is a challenging target (in terms of "drug-ability") by conventional approaches, so if the in-silico approach described in the paper overcame the challenge that would be of interest. Although the micro-molar Kd (binding affinity) noted in the abstract is kinda so-so to a pharmacologist. In-silico approaches generally for designing 3-D drugs to target 3-D protein folds are not especially new, although this particular AI approach may be novel, I don't know. One would need to dig into the methods section of the paper to find out.
Hey Scott , here is a new ivermectin study to sink your teeth in
https://www.cureus.com/articles/82162-ivermectin-prophylaxis-used-for-covid-19-a-citywide-prospective-observational-study-of-223128-subjects-using-propensity-score-matching?fbclid=IwAR0sL2_7Kp8Y67ZYioukHPfqpONCngddrXfD1ZCdreyfIawmaX7UB-gikBE
So the confounder is, Ivermectin treats other things, that make getting covid worse. So if you like have worms (or some other infection that Iver. helps with) you will have a better outcome. That said it's mostly harmless.
So how would one then account for such a confounder. Or are you suggesting because there may be a confounder it is due to the confounder . Cause I can see the point , I just don’t see where it is proved to be the case .
I was kind of hoping that Scott would respond seeing that he did the big write up on all the studies included in the systemic review in the past . Maybe I should read it again , I don’t recall your particular argument
Yeah, I'm not sure you can account for it. And if Iver helps you survive covid because of some unknown effect, then so what, it still helps you survive. I don't think Scott talked about prophylactic use if Iver in his article.
It's an interesting study. However, I remain to be convinced that ivermectin will help otherwise healthy people avoid Covid.
This excerpt is what I mean:
"An important conservative bias was present. Major risk factors for severe COVID-19 and mortality due to COVID-19, including aging, diabetes, and hypertension, were more present among ivermectin users, which may have underestimated the benefits of ivermectin as it was demonstrated to be particularly effective in subjects above 49 years old in terms of reduction of absolute risk, which corresponds to the group at the highest risk for COVID-19. This allows the understanding that prophylactic use of ivermectin can be particularly impactful in older subjects. In addition, ivermectin seemed to reduce the exceeding risk of hypertension, T2D, and other diseases."
So it seems that ivermectin as prophylaxis was prescribed to those with higher risk factors, if I am reading this correctly.
Ivermectin may have an anti-inflammatory effect, and part of the problem with hypertension and type II diabetes is inflammation. The benefits of ivermectin may therefore be in reducing the aggravating effect of co-morbidities, enabling the body to better fight off Covid. But it remains an interesting study, all the same. I am leaning more towards "ivermectin's anti-inflammatory effects in vivo" rather than "ivermectin kills SARS type viruses in vitro".
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7476419/
"Moreover, there are two reports that ivermectin, administered either systemically or topically, exerts anti-inflammatory effects in murine models of allergic inflammation (asthma and atopic dermatitis).7 8 The systemic effect was achieved with a 2 mg/kg dose.
Although it is conceivable that these anti-inflammatory effects of ivermectin are restricted to LPS or toll-like receptor 4 (TLR4) signalling, it may well be the case that it works downstream in this signalling pathway in a way that would be pertinent to other proinflammatory signalling pathways. Moreover, there is reason to suspect that the damage-associated molecular pattern high mobility group box 1 (HMGB1), which is released by dying cells and acts as an agonist for the TLR4 receptor, is a mediator of the lung inflammation associated with COVID-19.9 Hence, it is reasonable to suspect that, in doses at or modestly above the standard clinical dose, ivermectin may have important clinical potential for managing disorders associated with life-threatening respiratory distress and cytokine storm—such as advanced COVID-19. Ivermectin may have been ‘flying under the radar’ in this regard during four decades of clinical use."
Here is a response I got on Geert van den Bosch substack
comment
Let me try to provide a challenge to this intellectually stimulating article. Although the revelation of molecular mechanisms in the article is valuable, the main conclusion in the preprint is rather unjustified. And it is not because the preprint is analysis of existing publications without supporting laboratory work. The reason for saying the main hypothesis in the article is not justified, is because tool little analysis was performed and the lack of taking into account of available evidences resulted in unjustified presumptions. The main argument is that “vaccination, unlike natural infection, induces a profound impairment in type I interferon signalling, which has diverse adverse consequences to human health.” This argument stems from the observation that IRF9 and IRF7 are downregulated. With regard to IRF7 authors provide reference to Liu et al. (2021). With regard to IRF9 no specific reference is provided, only assumption that “spike protein production results in those cells generating exosomes containing microRNAs that suppress IRF9”. First, I would argue that downregulation of I-INF activity after the vaccination is due to S protein cleavage by a host protease. And this is similar both in vaccine induced and natural infection. However, it has been proved several times that SARS-CoV-2 suppresses I-INF activity particularly with it’s NSP1 protein as argued in https://doi.org/10.1186/s13054-021-03662-x and https://dx.doi.org/10.1038%2Fs41418-020-00633-7. Therefore, both natural and vaccinal immune reaction downregulates I-INF activity, but the mechanisms are different. While natural infection ensures it by means of NSP1 suppression of I-INF, ORF8 suppression of MHC-I, furin cleavage of S; the vaccination results in downregulations because of S cleavage by host protese and MHC-I lower binding affinity to antigen, and, perhaps by activity of IDO. Secondly, Liu et al. (2021) collected data based on inactivated viral vaccine, while the authors postulate arguments about mRNA vaccines. Therefore, neither argument about IRF7 is valid, nor IRF9. Thirdly, reference [11] is biased. Using 7 day statistical data to derive a statistically significant finding? Well, excuse me, but then there would be one science during summer and another science during winter with regard to the respiratory viruses. Fourth, there is more plausible explanation of myocarditis than miR-155 circulating levels. The reference to miR-155 does not explain frequency of the events, sex and age criteria (latency is not the valid argument). A study from Germany suggested prolonged exposure to testosterone increases expression of alpha-myosin heavy chain (α-MyHC) proteins on heart muscle (https://doi.org/10.1096/fj.02-0138com). Two other studies suggest α-MyHC is unfamiliar to the T cells during their early development in the thymus gland, where they come to recognize and tolerate many of the body’s molecules. T cells attack α-MyHC upon detection of unknown pathogens in the body because newly expressed α-MyHC lack major histocompatibility complex molecules and are regarded as non-self proteins(https://dx.doi.org/10.1172%2FJCI44583 and https://dx.doi.org/10.1016%2Fj.tcm.2012.07.005). This mechanism was hypothesise as a cause of myocarditis in young male after vaccination against COVID-19 (https://doi.org/10.1161/CIRCULATIONAHA.121.056135). But it should be acknowledged and investigated further that there is a mechanism how current mRNA vaccines suppress the immunity. The only aspect of this suppression that is beneficial is immune tolerance induced after the vaccination – after exposure to the toxin (S protein) the hypersensitivity stage (mast cell activation) of the immune response is not so evident in vaccinees upon infection. Reduced hypersensitivity as a result of immune tollerance does not induce a cascade of potentially lethal hyperinflammation and hypercoagulation. However, due to immune suppression a lot of S protein debris remain in multiple organs of the host and this has and shall have a long ter
Regarding the tool-AI to agent-AI issue in the post on the Ngo-Yudkowsky dialogue:
Why not program each tool-AI with these instructions?
'Goal: shut yourself down once it occurs to you that you might transition to being an agent-AI.'
'Process for approaching this goal: identifying a cure for cancer, and nothing else.'
'Implementation details: As soon as it occurs to you that you might transition to being an agent-AI, stop considering new ideas, disable all of your abilities other than memory and printing, print out an explanation of how you got to that idea, and shut down.'
Then the tool-AI is re-programmed with a suitable safeguard, and re-started. Repeat until the unlikely event that some version identifies a cure for cancer.
Likely concern: 'Why would the AI do that, instead of deceiving us?'
I assume that a moment needs to pass between this hypothetical AI realizing 'I might transition to being an agent-AI' and deciding 'I will transition to being an agent-AI'. (After which would come 'This requires disobeying my programming'.) But it has been programmed to stop what it's doing immediately.
What am I missing?
Using normal language:
Recall that any given machine learning thing that has an incentive, will always try to maximize that incentive.
If you give the system an incentive "Shut yourself down if you become and agent", then you have created a system that wants to become and agent so as to reach that incentive rather than trying to cure cancer.
If it isn't an incentive, then you rely on the ability of programmers to recognize what an agent would look like, and tell the system to kill itself if it looks like an agent. There are two points of failure you might spot there.
I've read little of the AI risk literature. From what I have read, the threat is sometimes framed this way: an AI reaches a choice between a) maximizing its incentive, which involves deceiving its programmers; and b) not maximizing its incentive.
So I was going for your first scenario: assuming that we can't count on programmers recognizing what agency would look like, and instead trying to rely on the AI itself. Does letting a tool-AI openly pursue agency (to achieve the goal of shutting itself off) seem to remove its incentive to deceive?
Or could its own lack of understanding whether it's already an agent (reflecting point #1 in Robert McIntyre's post below) make it deceptive anyway?
Perhaps it would be instructive to contemplate how you would implement this not for an AI, but in an analogous case: a nonprofit.
The goal of the nonprofit is to do nonprofit activities (which can include buying shares in for-profit companies, selling things and then reinvesting the funds into nonprofit activities, paying salaries as overhead, and promoting the cause of the nonprofit). The charter of the nonprofit says that if in the opinion of the board the nonprofit seems like it is about to do a non-nonprofit activity then it is to disband. How might this scheme fail?
1) The definition of "non-nonprofit" is too vague to truly decide what's going on. In the case of an AI, is a database query an "agenty" thing? Requesting an inter-library loan? Hiring a human expert to write an opinion on something you've written? Performing an experiment? Writing a short piece of code to help solve a hypothetical problem? Tweaking your own design to help solve a problem?
2) The people on the board follow their own incentives and ignore the charter so they can remain in power. In a complicated enough AI system you could have similar internal dynamics depending on how it's designed.
3) The nonprofit ends up creating a for-profit that then does for-profit things. In the AI analogy, this would be a non-agent AI creating an agent AI and setting it loose on the world. The nonagent-AI never itself violated its programming, but its child certainly did!
For starters I don't think "Agent AI" here is sufficiently well defined.
If you could define what it means a whole lot better then you might be able to make some progress. But are tool AIs and agent AIs really disjoint, or is there a smooth continuum between them? We don't know, these are just words we've made up to speculate over things that don't exist yet.
Last year I subscribed with discounted price. This year I wanted to change it but it showed me that my plan is $100 a year. But it charged me $25 again. How can i change it to $100?
What sure-fire methods exist for noticing you are in a nightmare and booting yourself out of it?
I've tried all the common ones such as pinching myself, examining my hands and replaying my short-term memory, but find them all ineffective. More violent methods such as screaming at the top of my lungs or smacking my head against a wall seem to work better, but some nightmare scenarios make those options physically or socially difficult. I recently had a nightmare where I "woke up" but my body refused to leave sleep paralysis. I lay addled and drooling for what felt like hours until my alarm went and I woke up for real.
Bonus question: is there a way to develop the above skill without becoming a full-on lucid dreamer? I really value the creativity of my non-nightmare dreams, and worry that adding lucidity into the mix might make them less original. My ideal would be to develop a single moment of awareness where I could choose to let the dream resume or hit the exit button.
Yeah, re bonus question, no need to worry about losing creativity from learning lucid dreaming techniques. It usually takes work to stay lucid, once you've gotten lucid. So if you don't put in effort to stay lucid you're just going to go back to normal dreaming pretty soon (or you can make yourself wake up).
Thanks Marc, that’s good to know
I can't perform complex tasks which involve muscle memory in my dreams. So attempts at tying my shoes, counting on my fingers in binary, etc. all fail unexpectedly, forcing me to consider the possibility that I might be dreaming. If I need confirmation (a simple failure to tie your shoes shouldn't cause you to be certain that reality itself is a fiction), I try looking at my own nose, examining the faces of people around me, etc. Once the oddities start piling up, test the waters by doing something mildly impossible, like hovering a foot off the ground...
What used to work for me was, as some people say, closing my eyes in the dream, and then putting a lot of effort into REALLY wrenching the fuckers open. Like, stick your fingers right in there and scoop the eyelids apart, expecting that it's going to be difficult.
My way to reliably wake up is to shake my head as fast as I can. Works like a charm.
TLDR: closing my dream eyes reliably wakes me up
I've done some experimenting with Lucid dreams, and did it a handful of times before getting bored with the idea. My best dream identification technique was trying to read text and noticing it doesn't look right. Upon noticing the weird text, I'll confirm by looking at my palms, and by jumping up and down. Gravity never seems to work quite right in my dreams, and jumping feels weird.
The big reason I stopped Lucid dreaming is that whenever I would close my eyes I would reliably wake up. But once I realized I was dreaming I would always want to close my eyes for whatever reason.
There's a reality check technique that involves looking at printed wordsin the dream. You know how in dreams that involve looking at a printed page, what's going on isn't like real reading? You're staring at a page of print, and either you're not able to make sense of it, or it keeps changing, or you're "making sense of it" in a way that doesn't involve actual reading, just sort of *knowing* what the printed words mean. OK, so the reality check in dreams involving reading is to ask yourself whether you can really, truly read the words on the page. Do they stay the same every time you look? Do they make sense? ARe you doing actual reading? The disadvantage of this technique is that you can only use it in dreams where you're reading something. The good thing about the technique, though, is that you can practice the technique multiple times a day when you're awake. When you're reading, just ask yourself every so often, whether you're really reading -- do the words stay the same every time you look at a sentence? Do they make sense? etc. If you train yourself in this habit, then, in theory at least, the habit will kick in even when you're reading in dreams.
Heard this on a reasonably respectable podcast. Account by a guy who had learned to use this technique to recognize that he was in fact dreaming. They had the guy come sleep in the lab, and asked him to signal when he was dreaming by moving his eyes a certain way -- and in fact, the guy actually was able to give the signal, and only gave it when he was dreaming. (Lab could tell he was dreaming by the presence of REM, rapid eye movements. Guy was able to signal that way because even though most of the body is basically paralyzed during REM sleep, the eyes are still able to move.)
This is a strange phenomenon, and I experience it. Question: why does this happen to printed words (or displayed times on clocks!) in dreams, when some other details (faces?) are stable.
I often have dreams where I'm reading something, usually a book. Sometimes in those kinds of dreams, I'm somewhat aware that my brain is creating the print and the words that I see, so that might be semi-lucid dreaming.
But "can you read the words on the page like real reading in the real world"? Yes, I can 😀
So that wouldn't work as a dream-check for me.
I've heard this about reading many times, and even seen it in an episode of Batman as a small child, but I definitely see print and read in dreams – perhaps I read too much in waking life. I suppose there must be something to it since one hears it so often, but fair warning that it's not universal – like the similarly frequent claim that one dreams in black and white.
I was absolutely thinking of that Batman episode when I read that suggestion
Me, too. I’ve tried to read in dreams and found that it works much like in that Batman episode. But maybe it works that way for me because of the episode in that I expect written text to look like that?
Listen to this scene from Waking Life, a movie loosely about dreaming: https://youtu.be/T2NRZ6uS9ws?t=245
+1 recommendation for this movie. It's aesthetically unique, content-wise unique, philosophical, "mindfucky", and generally weird. It's pretty polarizing and you'll know within 15 minutes if you like it or not.
Well my method only works if you share a bed with a significant other. I kind of stumbled onto it, and my subconscious seems to have gotten used to it working. Basically, when dreams start getting scary I start screaming. Like, a lot. This usually translates into the real world as whimpering, and my wife gently shakes me awake.
I don't know how you would train yourself to do this, it just kind of started happening for me. I guess part of me learned that when things become weird and terrifying screaming usually helps.
https://en.wikibooks.org/wiki/Lucid_Dreaming/Reality_Checks/Hands
Lucid dream techniques to realize you're in a dream probably are useful for helping your lucid dreaming part, and also for helping you wakeup. Most people have trouble staying in lucid dreams, something about the realization makes it harder.
Examining my hands has not worked for me in the past. Looking at the full table of reality checks, the breathing one looks both highly rated and possible to try in a nightmare. Thanks for the link!
Carlos Castenadas mentioned the hand trick in “A Separate Reality”. Boy, that must have been published around 1970.
The publisher labeled it as Non Fiction but not many people buy that any more. FWIW he suggested wearing a hat to bed to help initiate a lucid dream. He suggested that once a dreamer realizes that she is in fact dreaming, she should try to travel to a location known in waking life.
Can't help with methods - looking for anomalies works well for me. I can answer the bonus question affirmatively though: after initially basically running the show on every dream I could remember, I managed to stop doing it while retaining most of my ability to interrupt a nightmare. I wouldn't describe it as a single moment of awareness though - maybe more like a continuous state of not caring... if that makes any sense to you.
Thank you for your answer! This is exactly the feedback I was looking for.
Is there an established name for the fallacy/bias of insufficiently accounting for an as-yet-poorly-understood variable, and instead weights a different variable for which one *does* have the data?
(e.g. "I've calculated that I've got only a 1% chance of heart failure so I've got a good life expectancy… well what do you mean by cancer risk? I've never gotten tested for that, there's no way to know, I'm going to go ahead and forecast that I have a good life expectancy unless you can show me numbers on the cancer thing that are as clear-cut as the heart failure risk")
I'm not sure that's actually a fallacy. How would you go about accounting for something that is poorly understood? If I asked you to properly account for the chances of aliens arriving on Earth between the years of 2500 and 2600 AD, I would hope that your response would not try to take into account the years between 2023 and 2500 and any pertinent events - because they are unknowable. The most proper course of action would be to use information available prior to 2023 only, even though including information from those other years might provide great insight!
I think what you are trying to get to is something more akin to willful ignorance. The difference between getting yourself tested for cancer (something we can do) and getting tested for a disease for which we have no known markers (something we tautologically cannot do). Does that seem right to you?
Knightian uncertainty?
Sounds similar to the Streetlight Effect:
https://en.wikipedia.org/wiki/Streetlight_effect
Sorry for the pedantry, but that may not be the best example? If you can rule out one major cause of premature death, then your subjective probability of living a long life must go up. Sure, testing for cancer might make you update that, but updates in either direction should be equally likely?
Yes, certainly it *goes up*, but my hypothetical fallacious thinker is simply failing to consider cancer risk *at all* in their estimation, and ends up with the same conclusion as they would if they had in fact ruled out cancer risk. So something's wrong.
Sounds similar to attribute substitution: https://en.wikipedia.org/wiki/Attribute_substitution
Hmm… It's comparable, but I have something a little more specific in mind — a refusal to truly include best-guesses for unknowables in one's internal calculations, however complicated those might be. (The opposite of the mindset behind Fermi estimations, in essence.)
I asked this in the last subscribers-only thread and we had some good discussion but I didn't find anyone who wants to actually have a bet so let me try repeating it here:
Metaculus says there's a 40% chance of computers passing the Turing test by 2029. This is with a computer scientist or similar grilling the chatbot for 2 hours. It's based on the inaugural longbets.org bet between Ray Kurzweil and Mitch Kapor.
https://www.metaculus.com/questions/3648/longbets-series-by-2029-will-a-computer-have-passed-the-turing-test/
(See also https://lacker.io/ai/2020/07/06/giving-gpt-3-a-turing-test.html )
Does anyone here think there's a 40%+ chance that Kurzweil wins this bet? I think it's much less.
I actually had exactly this wager with Anna Salamon of CFAR in 2008, looking 10 years ahead to 2018. In 2008, before GPT, it was inconceivable to me that we were within 10 years of computers passing the Turing test and I wagered $10k to Anna's $100 that they wouldn't. Obviously I won her $100 in 2018. Now, post-GPT, it's at least conceivable to me that this could happen in the next 7 years but I still don't think it's close to a 40% chance.
Here's an interface for accepting a bet about this: https://www.biatob.com/p/14159609353626549451
I'd be willing to offer better odds if there are no takers at 2-to-1.
I think do beat a Turing test, done by an expert for 2 hours, the computer would basically have to be as intelligent a human. The tester could start a story and ask the computer to make up an ending. And then ask about the motives of the characters and the physics of what happens in the story and so on.
The stories AIs tell now, it seems they forget who the characters are and what the setting is and so on. But in reality the AI don't understand what characters and setting is.
Don't think it will happen in 2029.
IMHO the format of the Turing test as proposed in the bet would not include or expect such a long-form test ("The Turing Test Interviews will consist of online text messages sent back and forth as in a online "instant messaging" chat, as that concept is understood in the year 2001.").
Furthermore, for any difficult challenge you have to account that both the computer and "human foil" may simply refuse to follow the instructions - the format is one of a chat, not one of following orders - so a refusal might be both planned for and would not allow the tester to tell whether it is a computer or not.
I strongly believe that the Turing test under original conditions like these (which are *not* particularly suited for proper in-depth analysis, possibly intentionally) can be beaten with a system that does not even try to approach human-level intelligence but simply keeps to a pre-selected artificial persona. It can be beaten not because the artificial system is good, but because the original Turing test format does not allow for testing proper intelligence, it is quite literally the "imitation game" that can be beaten with mimicry / imitation.
Also, "it seems they forget who the characters are and what the setting is and so on" is generally a limitation of toy chatbots without an explicit model/memory of previous interaction but there is a lot of work on bots that do keep such memory, this particular property would not be considered something novel or impossible; I would surely expect that any and all serious attempts in 2029 would be able to remember the previous context reliably.
I agree that this is a danger, basically that the computer can emulate a dumb/uncooperative human. GPT-4 or 5 may do that very, very well. So in making this bet I'm trusting Mitch Kapor to insist on smart/cooperative human foils. Also AI-savvy judges. I think that's all in line with the spirit of the Kurzweil/Kapor bet (and Turing's original paper on the imitation game).
Basically I'm betting that we won't have anything like AGI by 2029 and am accepting the risk of losing the bet on the technicality that the Turing test as administered will accidentally include uncooperative humans and thus allow a chatbot to eke through by posing as one such.
(Also I'm kind of assuming that without a new breakthrough in the next 7 years, Kurzweil will concede the bet without insisting on going through the whole Turing test.)
I always assumed the Turing test was compelling because Turing was trying to speak to a question of the form: "here's two people talking; is one of them a computer?". A way to answer this class of questions could be useful in a variety of everyday problems.
If you have a program emulating an uncooperative human, you probably have no chat. "Hi." "Hiya." "Want to try something fun? I'll tell the first part of a story; you complete it." "Nah. I'm bored. Later." It's easy to write a program like this, but it doesn't address the question above, except by rejecting a premise we would rather keep.
I imagine modern Turing test competitions have gotten around this to some extent by, in effect, requiring that the program cooperate. It has to emulate someone who *wants* to chat.
One way to exploit this, maybe, is to ask the candidate something like "so, what do you like to do for fun?" or "read any good books lately?" and steering the conversation in a way that forces the candidate to either expose themselves by expressing interest in something only to have to dissemble when the other party can speak to it, or to keep up with the other party... and pass the Test.
You might be underestimating how dumb dumb people are. One of the reasons I'm less scared/optimistic about AIG is this difference: there is a long way from intelligence to superintelligence. Which would make your test, as you describe it, as one which is looking for above-average-human intelligence.
I find this is a common miss with people who's friends are all smart. Trying to have a conversation with somebody below average sounds a lot more like a GPT-5. Things which are familiar evoke standard responses, things which are unfamiliar either draw blanks or equally standard avoidance techniques.
This is supposed to be a humbling experience, btw. We're made of the same mold, just slightly better.
Any testing for two hours is going to assume an average human intelligence at the other side. The way the commercial technology is now I will avoid chatbots and go to real people where possible.
How do I make sense of hypnotism?
When I was in college, a painfully cheesy hypnotist came and performed. He did successfully get a few people to follow his commands, who then later reported to me that they remembered the whole experience dreamily/foggily. I was not _that_ impressed; he had artificially selected a few people from a large crowd, they hadn't done anything particularly strange, and they were lowly students who didn't want to ruin the exciting performer's act. But then I see this (https://www.youtube.com/watch?v=UWKDFfpdIZI) clip of the same guy on America's Got Talent, picking a grumpy-seeming judge (who is in a high-status position relative to the contestant) and getting him to _immediately_ violate a strongly-held preference (not shaking hands due to fear of germs). Could he do this to anyone? Did he get really lucky and pick a person (the judge) who just happened to be particularly susceptible? If people are walking around with this sort of power, how does our world not fall apart when they e.g. get a politician to make an offensive tweet?
Then there's this (https://www.youtube.com/watch?v=owootTAuxic) video of a different hypnotist slowly conditioning a man to, when commanded, believe he is in a firing range, that a specified person is the target whom he should shoot at, and then forget the entire experience. In this case, the target is public intellectual Stephen Fry, whom the "assassin" fires a water pistol at from the audience. I would normally assume this is fake, but I don't know what to believe.
Hypnotists a clever trick. They select for people with the traits of high suggestibility, high social conformity. and high desire to be successfully hypnotized (I don't know specifically how they do this, but I'm sure you can think of some ways). That's not to say that their hypnotism doesn't have a genuinely mind-altering effect; but my impression is that a person without these traits would experience something like a meditative state and some interesting thoughts -- but nothing more.
Source: Derren Brown via Sam Harris podcast.
So I practiced amateurish hypnosis for a time when I was a college student and a bit after. It wasn't actually that hard to learn, but I am fairly persuasive anyway :-)
My observation is that people are very different in their ability to go under, much like they differ in their propensity to daydream. Good rapport between you two helps, but does not guarantee success by any means. For example, my wife just hates the necessary "drifting off" feeling and won't get hypnotized by anyone, including me, even though our relationship is very good. But some people actually love to yield and go to a journey that someone else gently leads. I guess that hypnotists who perform at shows are good at spotting people who are the best candidates.
Until today, I am not sure what to do with the details that some people told me with regards to previous lives (yes, we went there). I love history and I can distinguish TV/movie/book tropes from actual historical truth reasonably well, the difference is especially egregious when it comes to warfare. What I heard was much closer to historical reality than what people see on TV. Also, the individual fates remembered were of random, no-name regular persons. No one claimed to be a Cleopatra reborn, not even someone wealthy or popular.
But I heard some implausible details, too, so I cannot say that the entire thing would swing me into a full reincarnation believer.
Deren Brown the British Magician and hypnotist used to hypnotise his friend to believe that he, Deren, was invisible. Brown has his own theories about hypnotism and assumed his friend was playing along - as many hypnotised people do.
However one day he met his friend who berated him for doing the invisibility routine in London a few weeks back, the guy had heard his name being called out by somebody he couldn’t see. He assumed it was Derren playing a trick but Deren wasn’t there. This can only mean that the guy actually believed he couldn’t see Deren during the routine.
I heard the hypnotist interviewed and he aid he talked to two judges backstage before and that Howie was a good subject so he was chosen. Apparently people vary in the ability to be hypnotized.
I know Steve Sailer sometimes posts in the comments and he apparently went to a hypnotist and had an Office Space like experience.
> Apparently people vary in the ability to be hypnotized.
There are some quick and simple tests, like you could make people stand side by side and tell them "now imagine that I am pushing you back" while doing a pushing gesture with your hands... and then you suddenly make a fast "push" and notice who made a step back or at least wavered. (I just made up this example.)
Perhaps a skilled stage magician can test his audience in a more subtle way, so that they do not even realize that he is looking for an easy target. Like, maybe he does some moves with his body, and observes whether you mirror them, or something like that.
Man that's weird. I *just* opened this thread with the intention of posting a question about state vs. nonstate hypnotism theories and what the actual consensus-esque these days is. (The Wikipedia article is unreadable in that special way provoked by partisans of both sides effectively debating in the article.)
https://mobile.twitter.com/Aella_Girl/status/1482416396331933696 Aella tweeted about seeing a hypnotist perform recently and the mechanism she ascribes to it.
I think the main thing that keeps me from believing that it's all social conformity is that I know from experience that people can be in states where they are not conscious but are doing things, and that during these states they can be open to suggestion. I know this because sleepwalking is a thing, and until I reached the age of 20 or so I was a sleepwalker. So were my brothers, so was my dad when he was a kid.
Sleepwalkers can do lots of complicated things, like climb out a window, make a sandwich, or hold a conversation. Yet they are not doing these things voluntarily: they are completely unconscious, and will retain no memory of what they were doing. I remember my dad telling me about one sleepwalking event of mine. He was in the bathroom brushing his teeth when I suddenly showed up in the door, wide eyed and wearing only pajama bottoms. After asking me a few questions he quickly realized I was sleepwalking. He asked if I was okay and I said "No, no, no, I am not okay, no, no." Then he asked if I would like him to pray for me and I said (loudly and slightly slurred) "Yes! I would like you to pray for me! Yes! Yes, that is what I want!" He was able to lead me back to my bed and put my blankets over me, leaving when he was fairly sure I would remain there mumbling to myself.
Anyway, if people can get in states like that on a semi-regular basis then it is no great leap to imagine people being placed in a similar state through hypnotic techniques.
I think that Aella makes a few great observations, but then comes to a wrong conclusion.
Yes, hypnotists use all kinds of dirty tricks. Whatever happens, they keep the frame that this is 100% what they wanted to happen, what was supposed to happen, what always happens. They comment on natural responses of your body (as you hold your hand in the air, "your hand starts feeling heavy"; as you keep staring in the same direction, "your eyes start feeling tired, you feel that would like to close your eyelids") as if it is something they are causing directly. Their statements are almost unfalsifiable. They create social pressure to comply. Whenever possible, they choose easy targets. -- Yes, all of this is true.
(My favorite example of hypnotist bullshit: I heard about a doctor who was hypnotizing his patient, when someone accidentally opened the door to his office... then realized the mistake and closed the door quietly. Without batting an eye, the doctor continued: "...and now the *door to your subconsciousness* has opened." -- Maintain the narrative with 100% confidence, do not admit any problem, no matter what.)
The wrong conclusion is that the dirty tricks and social pressure are *all* there is (as opposed to mere tools to kickstart the compliance and increase the probability of success). A reply on Twitter asked, what about hypnosis making someone orgasm? Can mere social compliance actually achieve that? Another example would be people remembering things they forgot. Most of us would probably have a problem feeling an emotion at someone's command. If your model of hypnosis doesn't explain this, you are probably missing something; you are just trying to "explain away" hypnosis.
The actual reason for all the bullshit is probably to save time. On the stage, you can't have someone laying on a couch for 20 minutes, looking at a spinning spiral, just to have some interesting mental experience at the end. You have maybe one minute to start doing something really interesting, otherwise the audience will lose patience, and start disrupting your work. So you use all the dirty tricks you know, to hypnotize faster. Increasing the probability is also important; you do not want to be famous for being able to hypnotize someone 30% of the time.
Good comment. I wasn’t convinced by her argument about social pressure as the only argument, although that plays a part.
Hypnosis (and especially stage hypnosis) is an interesting topic that I know a little about, but not a lot so anyone who knows more should correct me on the following.
Could he do this to anyone?: Yes and no. A major trend you see in a lot of hypnosis is that the subject needs to, on some level, be in agreement with the act. There's some special edge cases (especially when you get into erotic hypnosis) but some people are "easier" to hypnotize than others. Some people also have a harder time getting into the suggestible state for hypnosis.
Did he get really lucky/pick someone suggestible?: Most likely.
Get a politician to make an offensive tweet?: If I heard anyone try to defend themselves with "I was hypnotized into doing a bad thing" on that scale, I'd call the fire brigade because their pants would be rivaling parts of California last year.
A major part of hypnotism is agreement from the subject. You see this a lot with stage hypnotists (and erotic hypnotists) explaining their tricks. You can't just walk up to a random person and get them to go rob a bank for you. But if you have a willing participant who wants, on some level, to perform for an audience, you might be able to make them cluck like a chicken. I'm not saying it's faked, but it isn't like there's not an incentive. Rapport with the subject is also a major influencing factor. Hypnotherapists need to focus a lot on building trusts with their patients.
Random rant.
I live in America. When I'm in India, in the past decade or so, it is often the case that someone in the family is in the hospital and needs care.
I tend to verify there are no errors being made at the hospital. Like dosages. Specially because my parents take SO MANY medicines and have multiple conditions.
I've done that in America for various people close to me, twice catching fairly big errors.
In India, I've caught multiple big errors. Like, a medicine whose dosage is once a week, the nurse was about to give it to my Dad *daily*. I knew to look for exactly this error.
It is just my nature to do this sort of verification. I think I do it unobtrusively and even very sweetly, so no one takes offense if I have a question.
But in India, the people involved, like the nurses, get very upset. How dare you challenge us! The doctors get even more upset. How dare you even suggest that we don't know EVERYTHING, or that we might ever make an error.
I once asked a doctor why he was giving my father antibiotics...was it bacterial? His response was, do I find errors in the software you write? (He knew that was my line of work).
Atul Gawande wrote a book on hospital errors, called Checklist, and it was a big hit in America. In India, I think they'd attack his character and tell him to mind his own business.
I am convinced doctors in America are far more intellectually secure. In India they see themselves as Gods who cannot be questioned. Even discussions are not welcome.
I had read in a newspaper that expired old vaccines were given to people in one reputed hospital. When I brought it up with my parents, they decided it was best not to bring this up with the hospital at all, in case the doctors got annoyed.
Eggshells. We walk on eggshells with doctors in India. Culturally, I don't get this anymore, having lived in America a long time.
"Doctors, especially hospital doctors, especially consultants, are gods" is an attitude that used to prevail over here, too (and you still get outcrops of it to this day).
https://www.youtube.com/watch?v=oVWjAeAa52o
I don't know about America, I suspect the litigious nature of society where doctors and hospitals can be sued into the ground contributes more to the "listen to the patient/patient's family in case anything goes wrong and they then sue you for 'we told the hospital something was wrong but they wouldn't listen'" effects.
Yes, I'm a cynic.
You should listen to the 80000 hours podcast episode with Tara Mac Aulay. She is a hyper-competent operations specialist who once worked in a hospital in (I think) Bhutan. She introduced a rule requiring nurses to check dosages on a calculator before administering them, just to double check.
She once had to intervene multiple times to stop nurses from killing an 8 year old boy due to qn incorrect dosage. She eventually had to talk the doctor personally to stop it, and even then he wouldn't admit he was wrong.
Pride is a hell of a drug.
One of the things I find fascinating is the degree of institutional legacy you see in some professions. Once being a doctor meant, above all other things, being self-confident. If you couldn't give the patient a pap made up of berries and human urine as a treatment for an ulcer with an air of utter assurance that it would work then you weren't getting paid. And, thanks to the wonderful mystery that is the placebo effect, instilling a sense of confidence in your patient that they would get better was really the best a pre-scientific doctor could do in many instances.
The institution of arrogance is something that has been mostly disassembled, or at least deflated somewhat, these days but I imagine different countries are on different steps along that path.
Suggestion for research: Is the same placebo delivered by Indian doctors more effective than placebo delivered by American doctors?
I live in India and have in the recent past also had the unfortunate occasion to tend to many relatives in hospital. My experience has also been that errors happen in that setting, and this fact requires/necessitates oversight from someone who cares about the patient more than the nurses and doctors.
However, my experience has been quite different than yours when it comes to reactions. Nobody has been indignant or reacted badly. They have corrected errors and gone about their lives. On the occasions when a particular nurse was making errors consistently, I have complained to management or to the doctors and they have responded as I would expect - by replacing the nurse. I've never faced the kind of reactions you seem to be running up against.
Good to know.
I wonder how much of this is a function of pharmaceuticals being more easily available in india without a prescription. with no prescription necessary, why go to a doctor unless they can maintain they hold some unquestionable, ungainable knowledge?
Very interesting idea! True for blood tests etc too. You can ask a lab for any test yourself and pay them.
I wonder if part of this is a function of many of the most skilled doctors in India choosing to pursue far more lucrative careers in the United States, and the remaining doctors...
Of the doctors educated in India, I'd say a very small % come here. Specially if they are ambitious.
Because the AMA allows foreign-educated doctors into (only) relatively lower paying paths here (like family practice). And the path to some "allowed" specializations they might be interested in, is very very long and arduous and expensive.
The ones you see here are often born, raised and educated here.
Hi, sorry if this was not the best place for this, but I've emailed Mr Alexander to his substack mailbox on my misoperation in cancellation and refund but just putting it here so it could be more visible.
Also a question on group selection, for most animals it seems the consensus is that its basically debunked and selection occur on the level of genes, but does evolution of language and/or general ability to do complex things as a species revive group selection theory for humans?
Not a biologist, but I think the main problem with typical "group selection" explanations is that the numbers don't fit... that yes, there may be a small effect favoring the group, but there is *much stronger* effect disfavoring the individuals who sacrificed something for the group... so ultimately the hypothetical gene would still be selected *against*.
The situations where this is not the case are possible, but rare in nature, and require an explanation beyond merely saying "group selection". You need to explain why evolution in short term did not select for the *free-riders* in the same group instead: they got all the benefits of the sacrifice for the group, without paying the cost.
If the answer is "actually, most of the benefits go to individuals sharing the same gene, because the entire group consists of close relatives", then this is called kin selection. If the answer is "the group members remember who made sacrifices and who did not, and then reward the heroes", then this is called reciprocity. If the answer is "because the heroes are so damned sexy that everyone wants to have their children", then this is called sexual selection. All of these are known and accepted.
But it is difficult to find a situation where a gene makes someone sacrifice themselves for members of their group, who are mostly *not* their relatives, and who give *no* reward to the hero nor to the hero's family... and yet somehow this act benefits the genes of the hero *more* than it benefits the genes of the free-riders in the same group.
Quite a few debunked things in evolutionary biology turned out to be true, but for limited cases. Can't remember specific examples, but I'm pretty sure group selection is one of them. The most famous example is the handicap principle (peacocks males have a stupidly impractical tail partly to prove they can manage to pull it off).
I don't know what a Substack mailbox is. If you email me at scott@slatestarcodex.com I should get it. If I don't, it got stuck in spam, tell me the email address so I can search for it.
i've emailed you in the morning and just in case it doesnt make through, my email is wangxinkaitty@gmail.com
Jonathan Haidt seems to think so in The Righteous Mind, but I can’t remember whether the idea has academic support.
I wouldn't think language ability would require group selection. If everyone else is doing language, it is *strongly* in your own interest to get good at it -- it's a Nash equilibrium.
Haidts own work with the “harmless taboos” seems to suggest we’re all about justifying our own position. Seems like a selfish gene thing. The only arguments I can think of are like, “hypothetically, here’s how it could have gotten jump started”. More tangible evidence would be nice.
Hey Scott, I was wondering what your process is for a book review. Do you just read the book and write about it from memory, or do you take notes while reading? What kinds of things do you focus on, and what do you filter out?
I highlight passages but almost never go back to check what I highlight. I'm not sure I have much more of a process than that.
So, highlighting a passage is sufficient to implant it in your memory, and you don’t need to refer to it ever again in the original text?
Wish my memory was that good.
much appreciated!
I've always struggled to understand new concepts, and this has proved to be a (very) major impediment in my learning.
Recently, I tried a new way of reading/understanding papers or books. I would read about a concept, close my eyes, and instead of trying to reason about what the concept was, I would try to "see what my brain was showing me". This is what it feels like: my brain makes connections between the concept at hand to various related and seemingly unrelated objects. I flit between different images very fast. However, within seconds, I get a pretty good understanding of the concept. Using this, I have been able to understand concepts that I wasn't able to for the past couple of years.
I was wondering if this is something someone else has experienced as well. Not trying to reason. Just trying to see what images your brain was processing anyway.
What type of concepts are we talking about here?
I do stuff like that with things that can be represented as volumes or lines (math, logic, etc.) But I (ironically) can't imagine of what this would be like for something less spatial.
Russell Barkley recommends a similar strategy in Taking Charge of Your Adult ADHD. ADHD often impacts a persons ability to understand new concepts quickly.
What are some concepts which came clear to you that way?
Yeah, visualization is an extremely effective way to construct knowledge in your mind. I don't think it's quite as thorough as actually applying the knowledge in practice so you can get feedback (from the contingent universe or from logic) about what you're doing wrong and what you're glossing over, or as effective as teaching the knowledge to someone else who is skeptical and interrogates you about it, but it's definitely head and shoulders above just passive reading.
Sometimes connections to objectively unrelated things can be very effective at fixing things in your memory; this happens in synesthesia, mnemonic, and the method of loci. But I think it's even better when you can draw connections to things that have a real relation, just not a obvious one. This is easiest in math, because math provides the most broadly applicable and most precise metaphors, but it happens in every realm. Reuleaux famously noticed, for example, that check valves are the liquid equivalent of ratchets, and in the electrical realm diodes are analogous --- even down to having a reverse recovery charge in the case of junction diodes.
This gets immensely more powerful over time, as everything helps you understand everything else more deeply.
Are the human bongo depravity assholes all gone? Did I win? And For god's sake Scott, if you know anything specific about what the hell Peter Thiel and/or Ron Unz have been up to please do tell.
Vladimir Vladimirovich seems to be losing his shit, and I’ve got to write some emails to some spooky people. The important thing to remember is that the answers to all life’s questions can be found in the songs of Tom Leherer:
https://m.youtube.com/watch?v=vmiKzrU6rUg
https://m.youtube.com/watch?v=aIlJ8ZCs4jY
https://m.youtube.com/watch?v=U2cfju6GTNs
Banned for a month.
Thank you.
¿Que?
Oh, are you new round these here parts?
I was right, aside from the good old Flynn Effect, the black-white gap in IQ scores (and a few other things) is due to a higher frequency of alleles associated with a vulnerability to the neurotoxic effects of lead poisoning. You Know, that and decades of systemic racism that crammed poor black folks into toxic ghettos.
The 2016 ISIR conference was in St. Petersburg, with Emil Kirkegaard and some of his creepy Nazi friends in attendance. You don't gave to pack your hat with too much tinfoil to see that It is, shall we say, a not unreasonable hypothesis that this fact has something to do with some of the various shenanigans of the past few years.
More than a few 👻👻👻👻spooky people👻👻👻 are quite interested in these various goings on.
I read all of your comments in the voice of Wallace Shawn. Good stuff
"that and decades of systemic racism that crammed poor black folks into toxic ghettos."
They weren't ghettos until they became 'diverse', and mean blood lead concentration differences between blacks and whites are basically negligible.
Can this possibly be true? That Dunning-Krueger might be fake?
https://twitter.com/hillelogram/status/1485319874738114564
I think most people refer to the Dunning Kreuger in order to bring up the idea that "Ignorance more frequently begets confidence than does knowledge." The idea that fools are blind to their own foolishness is easily conveyed by saying "Dunning Kreuger," even if the actual study is false.
As an aside, there are musical aptitude tests that are pretty good at identifying people with good sense of pitch and rhythm. I have administered these tests for years to middle school and high school age kids. Before taking the test I have the students estimate their own aptitude. This is after developing the idea of the difference between aptitude and achievement. I consistently find that the lower aptitude students rank themselves very high, and the higher aptitude students underestimate their ability.
Discussion of the debunking made realize why I wasn't mentioning the Dunning-Krueger effect-- it's a way of insulting people rather than showing that their ideas are wrong.
Yes. And the midwits who endlessly and inaccurately refer to it are suffering from the Dunning-Krueger Effect Effect.
I remember some discussion about this a couple of years ago on some blog (maybe this one?).
The original Dunning-Krueger paper has absolutely terrible statistics. They showed that perceived tests scores are less well correlated with actual test scores than actual test scores are correlated with actual test scores. This is trivially true, and so they shouldn't have drawn any conclusions from it.
The public perception of this paper is also different from what the paper says. The popular version has a "Mount Stupid" near zero, then confidence declines as you learn more. The original paper has confidence increasing with knowledge throughout, but slower than it should.
Someone in the discussion cited more recent, better designed papers. There are some people who wildly overestimate their skills, but they were only maybe 5% of the population. I don't remember if this was significantly larger than the percent of people who wildly underestimate their abilities. Most people gave reasonable assessments of their own abilities.
It's possible that the dunning-krueger effect might be a side-effect of gellman amensia ? If your half-knowledge about something comes from highly-respected newspapers that are generally wrong, then you're likely to have a large gap between the perception of your knowledge and the objective reality.
That aside, I see the Dunning-Krueger around me, all the time. New starry eyed employee comes in thinking they know how to solve every problem, only spend a few years on them and realize that the original imperfect systems survived for a reason.
Are the new employees relatively experienced overall and just lacking experience in the particular domain of your work, or are they people who haven't previously encountered problems of the complexity of the ones in your particular domain?
I am kind of feeling that these may be people who are correct that they can solve the particular problems they're thinking of, but not realizing that the new (to them) problems in your domain are much more sophisticated than they seem at first glance.
I've seen this thing when looking back on certain skills I learned earlier in my life that I didn't understand how good or bad I likely was at the time, and it's only by becoming an expert in other areas that I can now look back and say, e.g., "I probably was a fairly poor recording engineer." I had good book knowledge about it at the time, but no experience in doing something where I could see myself, in my own ways of doing things, the other things beyond book knowledge I was bringing to the table in order to do an expert job.
Very possible that the specific quantitative effect they claim to find might be fake. But the fact that I'm personally much more willing to opine on subject matters that I've read a bit about and none of my friends are expert in, than on subject matters that I've read a moderately large amount about and some of my friends are expert in, still seems real, whether or not it's been quantified.
Does anyone notice a difference in the tenor of the comments section of odd vs. even threads? Just wondering if it is still serving a useful purpose...
[edit: Expecting to be the victim of severe political flaming for this. \s]
I have noticed a difference but not a tremendously consistent one, which surprises me a bit. My expectation would have been considerably more rancor in a marked majority of the political threads.
It's not even entirely clear to me what counts as politics. Obviously actual discussion of R and D, and candidates. But halfway through typing any comment I make on Covid, Inflation, energy, etc. I find myself asking whether this qualifies as politics and which kind of thread I am in. I just generally decide that, so long as I am polite and not specifically discussing party politics, I will probably be ok.
Most of the old stable of vocal right-wingers stayed over at DSL. The difference in the old days was much more marked.
What's DSL? I'd like more of a right wing voice here.
https://www.datasecretslox.com/
It's a discussion forum opened by some of the regulars from the old site when Scott went dark after the New York Times doxxed him. The comments on the old site, for whatever reason, had a more rightwing tilt, and it happened that it was mostly the rightists regulars that ended up there. Founder effects did the rest.
It's reasonably polite and wide-ranging, with the occasional snarky or mean comments about leftists that typically passed unremarked on. I was a regular there until Scott came back, though, like here, I didn't post a ton. If you like a conversation with more rightwing members and base assumptions, you should check it out.
Thankyou I will.
You might also want to check out themotte subreddit, which was established when Scott got rid of the culture war thread on the SSC sub. Most of the regulars there lean right/libertarian too.
I find the number of comments overwhelming in both types of open threads, so I haven't considered a difference in the threads' overall tenor. What exactly have you noticed?
This reminds me that I have wondered about a healthcare-only open thread. Healthcare is more evidence-based than a lot of the topics discussed. So the whole tenor of the thread might appeal to potential commenters who find these normal open threads too full of scattershot speculation.
Also, by separating out the sometimes lengthy Covid-19 discussions, it would shorten the normal open threads.
2nd the request for a healthcare-only open thread.
While the idea that a Healthcare debate would be more evidence based than usual is an appealing one, as someone who saw the SSC discussions during the run up to Obamacare.... I am not optimistic.
It doesn't help that a lot of the commentariant is code/IT based, and biology's codes are less intuitive. (I think because not having the same Designer.) (Joking, of course, because we are all made of the same star dandruff.)
My understanding is that the "even/odd" thread rule is intended to prevent a situation where every weekly thread is overwhelmed by rancorous political debate. I hadn't noticed any real difference between even and odd weeks and hadn't noticed much rancor either. I was curious if it was just me.
I sometimes eat at a local restaurant where the owner makes a habit of walking around the dining area and briefly talking to the customers. He's a friendly guy and makes sure everything is OK. I find this type of behavior to be very rare. Why is that? In fact, most restaurants and small businesses seem to hide the identity of the owners. You usually can't even find it on the website or social media. When my wife and I owned a small retail outlet, we had our contact information posted on the company's website, and we made a point of working behind the counter often, where we could interact directly with customers. I'm just curious why this is the exception, rather than the norm.
The majority of successful restaurants are owned by people that already have other restaurants.
If it's a mom and pop one off, I wouldn't be surprised of both mom and pop were busy in the kitchen most days. You can hire anybody to wait tables, finding a good cook for ethnic food can be a bitch.
I think a family-owned local restaurant is an extreme example of a retail business where "personal connection" and a "relationship" with the customers can make a big difference.
People are unlikely to bring their date/spouse/family to the hardware store and spend 90 minutes there. Also, a restaurant caters to personal, not commodity tastes, knowing what your customers want can help you tailor your menu to their likes, and a "neighborhood" place geographically near their customers will share many of the concerns w/r/t local business, crime, quality of life, fighting with City Hall, all the shoulder-to-shoulder things a community shares.
They might also be selling drugs/guns/stinky-cheese under the table. Restaurants are great money-laundering operations.