## My vaccine crackpottery: a confession

I hope everyone is enjoying a New Years’ as festive as the circumstances allow!

I’ve heard from a bunch of you awaiting my next post on the continuum hypothesis, and it’s a-comin’, but I confess the new, faster-spreading covid variant is giving me the same sinking feeling that Covid 1.0 gave me in late February, making it really hard to think about the eternal. (For perspectives on Covid 2.0 from individuals who acquitted themselves well with their early warnings about Covid 1.0, see for example this by Jacob Falkovich, or this by Zvi Mowshowitz.)

So on that note: do you hold any opinions, on factual matters of practical importance, that most everyone around you sharply disagrees with? Opinions that those who you respect consider ignorant, naïve, imprudent, and well outside your sphere of expertise? Opinions that, nevertheless, you simply continue to hold, because you’ve learned that, unless and until someone shows you the light, you can no more will yourself to change what you think about the matter than change your blood type?

I try to have as few such opinions as possible. Having run Shtetl-Optimized for fifteen years, I’m acutely aware of the success rate of those autodidacts who think they’ve solved P versus NP or quantum gravity or whatever. It’s basically zero out of hundreds—and why wouldn’t it be?

And yet there’s one issue where I feel myself in the unhappy epistemic situation of those amateurs, spamming the professors in all-caps. So, OK, here it is:

I think that, in a well-run civilization, the first covid vaccines would’ve been tested and approved by around March or April 2020, while mass-manufacturing simultaneously ramped up with trillions of dollars’ investment. I think almost everyone on earth could have, and should have, already been vaccinated by now. I think a faster, “WWII-style” approach would’ve saved millions of lives, prevented economic destruction, and carried negligible risks compared to its benefits. I think this will be clear to future generations, who’ll write PhD theses exploring how it was possible that we invented multiple effective covid vaccines in mere days or weeks, but then simply sat on those vaccines for a year, ticking off boxes called “Phase I,” “Phase II,” etc. while civilization hung in the balance.

I’ve said similar things, on this blog and elsewhere, since the beginning of the pandemic, but part of me kept expecting events to teach me why I was wrong. Instead events—including the staggering cost of delay, the spectacular failures of institutional authorities to adapt to the scientific realities of covid, and the long-awaited finding that all the major vaccines safely work (some better than others), just like the experts predicted back in February—all this only made me more confident of my original, stupid and naïve position.

I’m saying all this—clearly enough that no one will misunderstand—but I’m also scared to say it. I’m scared because it sounds too much like colossal ingratitude, like Monday-morning quarterbacking of one of the great heroic achievements of our era by someone who played no part in it.

Let’s be clear: the ~11 months that it took to get from sequencing the novel coronavirus, to approving and mass-manufacturing vaccines, is a world record, soundly beating the previous record of 4 years. Nobel Prizes and billions of dollars are the least that those who made it happen deserve. Eternal praise is especially due to those like Katalin Karikó, who risked their careers in the decades before covid to do the basic research on mRNA delivery that made the development of these mRNA vaccines so blindingly fast.

Furthermore, I could easily believe that there’s no one agent—neither Pfizer nor BioNTech nor Moderna, neither the CDC nor FDA nor other health or regulatory agencies, neither Bill Gates nor Moncef Slaoui—who could’ve unilaterally sped things up very much. If one of them tried, they would’ve simply been ostracized by the other parts of the system, and they probably all understood that. It might have taken a whole different civilization, with different attitudes about utility and risk.

And yet the fact remains that, historic though it was, a one-to-two-year turnaround time wasn’t nearly good enough. Especially once we factor in the faster-spreading variant, by the time we’ve vaccinated everyone, we’ll already be a large fraction of the way to herd immunity and to the vaccine losing its purpose. For all the advances in civilization, from believing in demonic spirits all the way to understanding mRNA at a machine-code level of detail, covid is running wild much like it would have back in the Middle Ages—partly, yes, because modern transportation helps it spread, but partly also because our political and regulatory and public-health tools have lagged so breathtakingly behind our knowledge of molecular biology.

What could’ve been done faster? For starters, as I said back in March, we could’ve had human challenge trials with willing volunteers, of whom there were tens of thousands. We could’ve started mass-manufacturing months earlier, with funding commensurate with the problem’s scale (think trillions, not billions). Today, we could give as many people as possible the first doses (which apparently already provide something like ~80% protection) before circling back to give the second doses (which boost the protection as high as ~95%). We could distribute the vaccines that are now sitting in warehouses, spoiling, while people in the distribution chain take off for the holidays—but that’s such low-hanging fruit that it feels unsporting even to mention it.

Let me now respond to three counterarguments that would surely come up in the comments if I didn’t address them.

1. The Argument from Actual Risk. Every time this subject arises, someone patiently explains to me that, since a vaccine gets administered to billions of healthy people, the standards for its safety and efficacy need to be even higher than they are for ordinary medicines. Of course that’s true, and it strikes me as an excellent reason not to inject people with a completely untested vaccine! All I ask is that the people who are, or could be, harmed by a faulty vaccine, be weighed on the same moral scale as the people harmed by covid itself. As an example, we know that the Phase III clinical trials were repeatedly halted for days or weeks because of a single participant developing strange symptoms—often a participant who’d received the placebo rather than the actual vaccine! That person matters. Any future vaccine recipient who might develop similar symptoms matters. But the 10,000 people who die of covid every single day we delay, along with the hundreds of millions more impoverished, kept out of school, etc., matter equally. If we threw them all onto the same utilitarian scale, would we be making the same tradeoffs that we are now? I feel like the question answers itself.
2. The Argument from Perceived Risk. Even with all the testing that’s been done, somewhere between 16% and 40% of Americans (depending on which poll you believe) say that they’ll refuse to get a covid vaccine, often because of anti-vaxx conspiracy theories. How much higher would the percentage be had the vaccines been rushed out in a month or two? And of course, if not enough people get vaccinated, then R0 remains above 1 and the public-health campaign is a failure. In this way of thinking, we need three phases of clinical trials the same way we need everyone to take off their shoes at airport security: it might not prevent a single terrorist, but the masses will be too scared to get on the planes if we don’t. To me, this (if true) only underscores my broader point, that the year-long delay in getting vaccines out represents a failure of our entire civilization, rather than a failure of any one agent. But also: people’s membership in the pro- or anti-vaxx camps is not static. The percentage saying they’ll get a covid vaccine seems to have already gone up, as a formerly abstract question becomes a stark choice between wallowing in delusions and getting a deadly disease, or accepting reality and not getting it. So while the Phase III trials were still underway—when the vaccines were already known to be safe, and experts thought it much more likely than not that they’d work—would it have been such a disaster to let Pfizer and Moderna sell the vaccines, for a hefty profit, to those who wanted them? With the hope that, just like with the iPhone or any other successful consumer product, satisfied early adopters would inspire the more reticent to get in line too?
3. The Argument from Trump. Now for the most awkward counterargument, which I’d like to address head-on rather than dodge. If the vaccines had been approved faster in the US, it would’ve looked to many like Trump deserved credit for it, and he might well have been reelected. And devastating though covid has been, Trump is plausibly worse! Here’s my response: Trump has the mentality of a toddler, albeit with curiosity swapped out for cruelty and vindictiveness. His and his cronies’ impulsivity, self-centeredness, and incompetence are likely responsible for at least ~200,000 of the 330,000 Americans now dead from covid. But, yes, reversing his previous anti-vaxx stance, Trump did say that he wanted to see a covid vaccine in months, just like I’ve said. Does it make me uncomfortable to have America’s worst president in my “camp”? Only a little, because I have no problem admitting that sometimes toddlers are right and experts are wrong. The solution, I’d say, is not to put toddlers in charge of the government! As should be obvious by now—indeed, as should’ve been obvious back in 2016—that solution has some exceedingly severe downsides. The solution, rather, is to work for a world where experts are unafraid to speak bluntly, so that it never falls to a mental toddler to say what the experts can’t say without jeopardizing their careers.

Anyway, despite everything I’ve written, considerations of Aumann’s Agreement Theorem still lead me to believe there’s an excellent chance that I’m wrong, and the vaccines couldn’t realistically have been rolled out any faster. The trouble is, I don’t understand why. And I don’t understand why compressing this process, from a year or two to at most a month or two, shouldn’t be civilization’s most urgent priority ahead of the next pandemic. So go ahead, explain it to me! I’ll be eternally grateful to whoever makes me retract this post in shame.

Update (Jan. 1, 2021): If you want a sense of the on-the-ground realities of administering the vaccine in the US, check out this long post by Zvi Mowshowitz. Briefly, it looks like in my post, I gave those in charge way too much benefit of the doubt (!!). The Trump administration pledged to administer 20 million vaccines by the end of 2020; instead it administered fewer than 3 million. Crucially, this is not because of any problem with manufacturing or supply, but just because of pure bureaucratic blank-facedness. Incredibly, even as the pandemic rages, most of the vaccines are sitting in storage, at severe risk of spoiling … and officials’ primary concern is not to administer the precious doses, but just to make sure no one gets a dose “out of turn.” In contrast to Israel, where they’re now administering vaccines 24/7, including on Shabbat, with the goal being to get through the entire population as quickly as possible, in the US they’re moving at a snail’s pace and took off for the holidays. In Wisconsin, a pharmacist intentionally spoiled hundreds of doses; in West Virginia, they mistakenly gave antibody treatments instead of vaccines. There are no longer any terms to understand what’s happening other than those of black comedy.

### 177 Responses to “My vaccine crackpottery: a confession”

1. Dr Panic Says:

Scott is an amateur “crackpot” indeed. Even now, the long-term safety profile of these vaccines (esp. RNA vaccines) remains *completely unknown*. And the only precedent we have – SARS-1 vaccines – indicates that the risk of serious auto-immune reactions and other issues (e.g. neurological issues) is very real.

2. Dr Panic Says:

Meanwhile, much safer and simpler options have been ignored by the NHS and FDA for months: https://covid19criticalcare.com/

3. Trumpist Says:

I hope one day you’ll wake up from your severe TDS and realize it was a result of concentrated brainwashing by your media. You literally admitted here Trump is one of the few people which had the same unique attitudes as you did yet you still fail to research and find out whether his actions in other subjects also correspond with your opinions, and instead you delegate your judgement to your media bias and press. You’re still arrogant and decide to judge a person by few humoristic faces and sentences, immediately judging him to be a toddler.

If you judged every politician according to everything he said, you should begin by judging Joe Biden by his statement that they put together the most extensive and inclusive voter fraud organization. I think that statement alone was more dangerous, if true, than anything Trump has ever said. https://youtu.be/BRZEs9BRGK4. I challenge you to prove me wrong and show me something more dangerous Trump said than this, and give me a video of him actually saying it instead of media digested interpretation.

You’re the arrogant toddler for judging a person in a shallow way according to stupid media pictures and manipulated quotes, and according to his intentional decision to keep the way he speaks approachable and understandable to people of lower intellect and vocabulary.

4. Aleksei Besogonov Says:

I think this will be clear to future generations, who’ll write PhD theses exploring how it was possible that we invented multiple effective covid vaccines in mere days or weeks, but then simply sat on those vaccines for a year

First, Moderna’s vaccine is the first of its kinda. There has never been an approved mRNA vaccine for humans before. Do you seriously think that it’s a good idea to vaccinate people with a completely untested vaccine on a new platform?

Next, vaccines against SARS-1 were not only ineffective but have shown actual INCREASE in lethality by making it easier for virus to infect people.

Perhaps if you were completely ruthless and was willing to skip phase I and II trials, and to proceed straight to challenge trials (say, using condemned prisoners) then you could have shaven about 3-4 months from the time.

And it’s not like all that time has been spent doing nothing. Moderna and Pfizer were hastily scaling up the production capacity, so they’ll be able to produce staggeringly large amounts of vaccine in 2021.

5. Aleksei Besogonov Says:

Moreover, classic inactivated virus vaccines are only now entering Phase III trials. It took almost a year just to be able to find the right conditions to grow the virus in a scalable manner in a cell culture, and to find the right set of adjuvants.

The current regulatory system was designed with this speed in mind, if it takes you at least a year to develop a vaccine then there’s no real rush to shave off 3-4 months while massively increasing the risk.

What should happen, is a new regulatory regime for mRNA vaccines that allows for shorter Phase I and II trials. Perhaps something like what we have for yearly flu vaccines.

6. LP Says:

There is one argument that even now about 30% or more people are not willing to get vaccinated (even higher depending on the country). If it were available much earlier, that is bypassing all standard phases, we won’t get many willing to get it. I’m not sure I’d be willing to get a fresh off the lab vaccine in say May myself.

7. Richard Says:

If the vaccines by BioNTech/Pfizer or Moderna would have been approved for the public earlier, which rule would have been used? Would this rule have also applied to vaccine candidates by other companies? Would there be a risk of having a similar situation as we had with Hydroxychloroquine?

8. Jesse Says:

My question is on the scale up in manufacturing. For the mRNA ones as I understand they had to build whole new factories, so how much sooner would there have been these available at scale even with faster approvals? I note gates foundation was pre funding a bunch of scale up work already.
For the more conventional vaccines, the development was longer. How much sooner would they have been ready? How long ago did they start full scale manufacturing?
I dont know the answers here but I suspect that even with an ‘instant’ regulatory approvals it may have only shaved a few months off.

9. entropy Says:

Sorry, I have no time to summarise. Here are several arguments, by experts, against rushing:
https://blogs.sciencemag.org/pipeline/archives/2020/07/02/challenge-testing
https://www.statnews.com/2020/06/23/challenge-trials-live-coronavirus-speedy-covid-19-vaccine/
https://www.sciencemag.org/news/2020/10/early-approval-covid-19-vaccine-could-stymie-hunt-better-ones

10. Daniel Cohen Says:

You’re not alone in asking why it had to take so long; I’ve been wondering the same thing.

11. Nick Nolan Says:

Good work Scott. You explain your thinking with such clarity that it’s possible to say what you missed.

It seems that you are thinking about risks and risk detection in the wrong time frame. Most side effects are noticed within six weeks after getting the dose. Some risks related to vaccines can be noticed only year or years afterward.

Test subjects are monitored for over a year for additional side effects even after the vaccine has been approved and phase III trials have ended. Phase III ends Covid-19 vaccines it ends in the latter half of 2021.

If you detect debilitating chronic disease emerging in a significant fraction of vaccinated after all healthcare workers have been vaccinated, it could collapse the whole healthcare system.

The main risk is autoimmune diseases, typically neurological. Autoimmune diseases are not always clinically apparent. For example, in type 1 diabetes it can take months or years before the appearance of clinical signs of diabetes.

If only 1 to 100k gets the disease, maybe it’s acceptable. But if 0.1% or 2% of the population gets chronic disease, the cost/benefit analysis becomes unfavorable. There are unknown unknowns.

12. Mateus Araújo Says:

It doesn’t take a whole different civilization. Just look at how they did things in Russia: the Sputnik V vaccine was approved back in August, when only Phase I-II had been completed. They had trouble mass-producing it, but in December they already started rolling it out to the population. I think the timeline was similarly accelerated in China, but I’m not aware of the details.

It pains me to speak positively about authoritarian regimes, but it seems to me that this was the decisive factor there: Putin and Xi Jinping took the gamble, and the rest of the regime didn’t have to accept any personal risk, they were just following orders.

In democratic countries one would need to change to law to compress the timeline, and therefore a majority of Parliament would need to personally assume responsibility for the argument that the risk was worth the reward. That’s a really tough sell: you’re asking your constituents to support a cold-blooded risk-reward analysis, and indirectly taking the responsibility for whatever might go wrong.

People really don’t like that. Keep in mind that normal people think the trolley problem is a serious moral dilemma, instead of a no-brainer. They have such a strong preference for inaction that actively killing one person seems as bad as letting five people die by inaction.

I agree with you about the “argument from perceived risk”. The morons that are afraid of a mRNA vaccine because they think it will rewrite their DNA won’t be convinced by clinical trials or scientific explanations. They didn’t reason themselves into that position, so they can’t be reasoned out of it. What will convince them is the simple network effect of seeing people they know and trust getting vaccinated.

13. Edan Maor Says:

The main counter-argument is that things like this have happened before – vaccines or other medicines have been rolled out and in widespread use, and then later discovered to have been a mistake. (I don’t actually know anything about this field so this is totally vague and a layman’s understanding of what happened)

But personally, whatever good counterargument there is, I’d still think the correct response would be “ok, so why not spend the next X years until a pandemic making {COUNTER_ARGUMENT} invalid”.

And of course I’d go further, and say that while a pandemic is a great case to see this obvious issue playing out, it is by no means unique – every new health improvement that goes through a process that was longer than necessary means more deaths/ill health. So working on shortening that time is a pretty good idea in general.

14. James Cross Says:

You’ve got two issues with vaccines.

1- Do they actually work?
2- Do they have too many adverse effects?

You have to vaccinate a good number of people to determine #1, but you don’t want to vaccinate a good number of people until you’ve determined #2. If you vaccinate a lot of people and a lot of adverse events arise, you will scare too many people in the population from taking the vaccine or maybe any vaccine.

Most of the vaccines require two doses several weeks apart so you can’t even gauge basic safety and dosages for 6-8 weeks or so. Phase 1 of a trial usually involves very small numbers of people who are basically healthy who will be injected with a range of dosages. The group that received 250 micrograms in the Moderna Phase 1 had a number of adverse reactions so the following trials received 100 micrograms which didn’t show the number of adverse effects. Phase 2 will expand the number of participants and open participation to participants with various conditions that might exist in the general population. Phase 3 will vaccinate a large enough group of people that a determination can be made that the vaccine works.

So vaccines were available in April or so. Phase 1 took until June. Phase 2 until August. Phase 3 which needed additional time to have people in the placebo group get ill to make sure it worked took until November/December.

15. Matty Wacksen Says:

I’m a little bit bothered by your “experts” vs “amateurs” framing. Whereas it’s more clear who is the real “expert” in the hard-sciences/mathematics area, I see no reason to believe that “scientists” are “experts” on public policy or other “soft” fields in general. In fact, pick 3 mathematicians and they will all give you the same answer on whether or not a polynomial over the complex numbers has at least one zero (“if it is zero or non-constant then yes, otherwise no”). However, ask 3 “experts” on whether or not maks ‘work’ against this virus, you’d have gotten a different answer in January compared to now (apologies for the slightly touchy example, but I’m sure you can think of others). To those of us following the replication crisis, this is obvious – the “softer” the “science”, the less the experts know. After all, the “softer” the science, the less their jobs are selected for “expertise” and the less testable their predictions are. You’re confusing “is viewed as an expert in the field” with “is an expert in the field” because for fields like mathematics, those categories are the same.

I have even less reason to believe that the people who make public policy – politicians – are “experts” in any shape of the form, or that those voting for the politicans are all “experts”.

But good news! The “experts” in this case – economists – do at least agree with you (or some of them do, read e.g. the Cowen/Tabarrok blog “Marginal Revolution” or John Cochrane’s blog). They have been saying more or less the same thing for months. The issue is that the incentives are stacked in a way that makes this kind of situation arise quite naturally, and this has been understood for quite a while now. I personally think that the underlying reason is that people love the illusion of control, which leads more and more to a regulatory state where everyone tries to control as much as they can. Those economically on the left may disagree with me here.

16. Art Says:

I’ll share my personal understanding, but I’ve not tested it much. Tear it to shreds everyone.

The utilitarian argument in 1 is actively avoided in medical fields, and I suspect this is true in any culture with an effective medical service. As soon as you are allowed to compare lives, you get very real questions about letting someone die to distribute their organs amongst those needing donors. As soon as you have a norm for even doing such a computation public trust is gone.

Sure, there are edge cases where it looks harmless. The gross examples of the past that I’m aware of all violated consent. But I’m pretty sure one could also manufacture consent, or manipulate marketing/news to get people to sign up. Even the appearance of such would do serious harm. Imagine the anti science sentiment if Faucchi actively recruited for a vaccine trial that got (super covid rates of) people killed. (I claim this pushback would similarly appear in most cultures. A healthy society should be very skeptical of manipulation)

For 2, I think this may be closest to truth.

I’m worried about the idea of selling the vaccine early, as it kinda creates a class of people for whom the lockdowns need not apply (at least not for safety reasons). I can imagine companies buying and mandating a vaccine for their workers (regardless of consent), or people breaking lockdown more once vaccinated (and ruining the social norms).

Ultimately though, I’d also like to see some stronger arguments. I feel like the above is unsatisfying.

I think you may be dramatically underestimating the boring but real challenges of ramping up production and distribution as fast as you seem to be imagining. My understanding is that, going back to the spring, the companies involved began to develop their manufacturing & distribution *as if the vaccines were guaranteed to be approved and needed worldwide*–and that’s what got us to the number of available doses we have now.

No doubt that process can be improved, and additional investments can be made in advance of the next pandemic to make it faster. My not very informed image of this is something like the government building a bunch of vaccine manufacturing “rough-ins” ready to be adapted as needed when a pandemic emerges. And I’m sure all the policy blunders we made along the way have had a real impact (as we’re seeing now with distribution). But I’m not sure challenge trials could’ve meaningfully sped up the path to widespread vaccination.

18. JoshP Says:

The main problem I see is that a major fuckup with an early version of the vaccine might cause an enormous damage to the whole vaccination idea. In other words, it can be a shot in the arm of the anti-vaxxers (pun intended!).

19. James Cross Says:

#4, #5

mRNA vaccines have been around for decades and it has been difficult getting them to work without adverse reactions. So those are good points and good reasons for being cautious.

Moderna, I think, has to wrap the mRNA in nanoparticles of lipid. And it isn’t just adverse reactions to worry about. The mRNA has to get to the right places in the body to build antibodies before they get degraded in some way.

20. Moose Cannon Says:

Were there really more than a couple mistakes?

Challenge trials could have shortened the rollout by two months, and there is precedent in American government in Dr. Harvey Wiley’s so-called “Poison Squad” food health experiments.

The vaccination rollout has been slow and wasteful, in part due to political fighting about whether to give it to grandparents or grocery workers first. It’s also been hamstrung by just … not giving out the vaccine? We’re proportionally way behind, say, Israel, who is vaccinating 100k people a day, so their highest risk people have mostly been give one dose already and the whole population will have by March.

The last question that could speed the vaccination up is whether one dose is effective enough for now. While two shots of the vaccine is what was tested, and will provide the most ongoing coverage, some experts have written in the NYT and WSJ that the single does could be 75% effective or more. Certainly the vaccine and the placebo groups began to diverge after just one dose in the now-famous Pfizer chart. This would allow us to halve the time to dig ourselves out and then we could top up once the crisis is over.

What other specific mistakes can people point to?

21. Jon Says:

I’ve had similar thoughts. The US DoD did sponsor development of a prototype planning tool for such assessments (see https://doi.org/10.1002/ddr.20310), but such explicit Bayesian utilitarian methods seem to have made little or no impact to the mainstream public discussions on vaccine testing and rollout. This is not for lack of technical/analytical knowledge among the relevant expert groups; it would be interesting to hear their views on this.

22. Piotr Suszyński Says:

I rarely can read an article that so closely resembles my own thinking.

I do sometimes have such naive ideas that you mentioned, but then I just go ahead and read, increase my understanding and to date such treatment was effective in finding an explanation every time (as of now I do have some such unresolved ideas but I consider them “in the pipeline” and only my notebook knows of them), but not this time.

I’m really amazed that no challenge trials happened and actually in all the countries around the world. There was also virtually zero public comments from policy makers on why not to do challange trials, apart from very simple and easily overthrown arguments, like the ones you mentioned – I mean there was zero discussion, that I’m aware of, that would actually consist the weighting of all the negative COVID effects, in a unit of time, versus negative effects of challange trials.

In my mind the argument 2 is the strongest one but I think I still could easily respond to it: the challange trials could have been even bigger than the Phase 3 trials that were performed and could have easily provided stronger statistical power if there would be enough of brave volunteers that would succeed in getting infected. In such case we would have stronger evidence for the vaccines safety and efficacy and this should make it easier to persuade the public.

My voice has extremely small audience but it really makes me sad that in such matter the voice of even someone like you, with your reputation and reach, haven’t been heard.

23. JL Says:

I’m inclined to agree with the main argument here, and my main thought is probably that “we didn’t know it was possible”. Institutions responsible for responding to the virus were still working under the idea that vaccines took years to develop. Vaccines based on emailed genetic material (which is how the current mRNA vaccines were made) were mostly an unknown. The idea of having vaccines by December was already an optimistic goal by our past standards.

Most institutions thought that the most feasible approach to defeating the virus was a combination of lockdowns and an accelerated vaccine development that had a framework similar to developing regular vaccines, so they focused on exactly that. In the future we should probably start doing research on how to create a new vaccine development framework using human challenge trials and start popularizing it so that there’s less fear of using that in the future.

Looking back on it though, I’m curious has to why China’s vaccine has been slower. They had more months of preparation given that they took the virus seriously way earlier. They also have less compunctions about vaccine ethics (given that they apparently used their own sinopharm vaccine without even full results from their studies yet to about 1 million people already by late November). They also have the capacity for a more centralized response where they could’ve forced their own government owned biotech firms to start making vaccines as early as February. I think any argument trying to explain why the western response was slow should also consider how those same arguments apply to China.

24. James Cross Says:

This Derek Lowe post about challenge trials that somebody linked to earlier is instructive.

https://blogs.sciencemag.org/pipeline/archives/2020/07/02/challenge-testing

Aside from ethical issues, in the end not much would be accomplished because the people doing the challenges would not be similar enough to the population that most needs to get vaccinated, that is older people with various conditions.

25. Simon Says:

entropy #9

Those articles you’ve linked are several months old and aren’t referring to the mRNA vaccines we’re getting now. They were talking about the old style of vaccine that would have required exposure to the covid virus itself. The risk profile should be significantly less with mRNA because it’s only one pathway.

26. Joe Says:

I confess the new, faster-spreading covid variant is giving me the same sinking feeling that Covid 1.0 gave me in late February, making it really hard to think about the eternal

If it’s any consolation, for the better part of the history of science, people were living in what we might now think of as a horror movie. Physics’ most famous text book series was composed in secret police prison. Du Châtelet worked out conservation of energy under the prospect of her imminent death by childbirth. Riemann and Fresnel lived with the curse of tuberculosis – both died at age 39. Soon, we will get used to it, too.

27. Boaz Barak Says:

I don’t see the evidence that “we sat on vaccines for a year”. You can take a look an abbreviated timeline for the Pfizer vaccine development on https://www.livemint.com/science/health/pfizer-s-covid-19-vaccine-a-timeline-of-key-events-11604972708055.html

First of all, the fact that we still have nowhere near the number of doses manufactured that we need, suggests that regulatory process is not the only reason why we have not done this earlier.

Second, the reason “human challenge” trials were not widely taken was not some rigid adherence to arcane regulations. Rather, there are two main factors (as far as I can tell):

1. The population most important to immunize – elderly people and other high risk factors – is exactly the one you don’t want to enroll in human challenge trials. So data from a trial of healthy young adults will not be that useful.

2. The complete inability to control Covid-19 in many countries made it much easier to conduct Phase III trials – it was not that hard to get enough cases where people caught Covid 19 naturally.

The fact that we have not yet seen a single country, including Russia and China, that managed to immunize their entire population, suggests that it’s not as easy as it looks, and you can’t just blame FDA regulations.

I don’t know when will be the date when >80% of US population would be immunized (May?June? December?) but it seems that by the time we get there, most of the delay could not be blamed on having too high safety and efficacy standards.

28. DavidC Says:

> do you hold any opinions … that … ?

I hold the opinion expressed in this post. But I didn’t think it’s one that fits these crackpot conditions!

29. Lulu Says:

I can’t speak to issues with distribution but I can talk about why I think the normal multi-phase testing protocol was important.

Imagine if this vaccine did have significant long term side effects that we missed (which could still be the case). Or even imagine if something awful happened during challenge trials. It would damage global perception of vaccines immense likely causing far more net harm than the 8-9 month delay in development of the COVID vaccine.

The natural rebuttal to this of course is that governments should adopt policy that effectively forces people to get vaccinated even if they don’t want to for all diseases (even stuff like the Flu). I am opposed to this on principle but I recognize that other might not hold this view

30. Paul Topping Says:

I’m with you entirely. But here’s another related quibble. Why do we need all the phases to complete before doing the “real” vaccinations? Why not a more organic rollout of the vaccine? Perhaps the makers can try it on themselves first, then on volunteers from their own company, and then to a gradually widening circle of volunteers? This would be accompanied by tracking and issue responses all the way, of course. Sure, it makes statistics gathering and PR more difficult but who cares about that? Another way to look at this is a melding of testing with the real vaccine delivery. After all, the real vaccination program is still monitored and may uncover problems. The economic and logistics aspects would naturally ramp up as part of this process. There would be no scratching of heads about how to deliver massive amounts of vaccine at the end of the testing process. The organically growing delivery would scale up gradually in all respects.

31. entropy Says:

Simon #25

I think you missed the point. Challenge trials might not tell us anything useful, regardless of the kind of vaccine you are testing for effectiveness. Please read the articles. James Cross #24 read and got (one of) the points.

32. Avraham Eisenberg Says:

>all the major vaccines safely work (some better than others), just like the experts predicted back in February

There’s selection bias here. The handful that work are retroactively considered major. As of now, WHO is tracking 232 different vaccines in development! Out of those, 3 have been approved. See https://www.who.int/publications/m/item/draft-landscape-of-covid-19-candidate-vaccines

I agree that it could and should have been a lot faster. But this statement is overstated.

What experts were predicting these specific vaccines would work, back in February? If you can show me someone who correctly predicted which ones would work *and* which ones wouldn’t work, I’ll be surprised.

33. Scott T. Says:

1. An argument you made (or at least repeated) forcefully many times was that the initial botched testing set the stage for the US disaster because we weren’t able to detect community spread in Washington that could have been snuffed out with contact tracing. A counterargument was that the number of international travelers meant the virus was probably already endemic enough that contact tracing could not have eradicated it. The variant now provides a test of your hypothesis. With testing now in place, community spread has been detected in Colorado and California. If it can be snuffed out with contact tracing then your placing of the blame on the initial testing rollout was on the mark.

2. You could have been more vocal in real time when Moderna and the government intentionally slowed their trial to improve “diversity” instead of largely restricting to more politically correct libertarian critiques.

3. Even with early challenge trials we would have still have needed a way to manufacture and distribute the vaccine on a large scale or we would be in exactly the same place we are now. Academics who beat the drums loudly for challenge trials while ignoring this aspect need to own up to this. You lionized Bill Gates for promising to manufacture at cost so it’s not like this wasn’t on anyone’s radar. When there wasn’t an attractive nurse knocking on my door with a vaccine the day after its approval my admiration for him decreased and it is incredibly angering to have to wait for known beneficial vaccines while people die.

4. The logistics of distribution is something where computer scientists could have or could still make a difference instead of being bystanders. We pay university computer science and operations research departments a lot of money to write papers about scheduling and routing. So how should vaccines get from UPS trucks to arms? Israel seems to be doing a much better job with this than the US.

34. Covid 12/31: Meet the New Year | Don't Worry About the Vase Says:

[…] to thank everyone who took my warnings of last week seriously, and helped spread the word. We got a shout-out in Shtetl-Optimized, as he offers his “crackpot theory” that if we were a functional civilization we might have […]

35. AMT Says:

I basically agree with you that overall, the cost-benefit analysis seems strongly in favor of pushing the vaccine through a lot faster.
I am a total lay person, but it seems the strongest counter argument could be that some adverse side effects take a while to manifest, so we really have no idea the full costs. Some commenters already mentioned this. Another example to throw out there along those lines would be the thalidomide scandal in the 1950s. The drug was fine for adults, but caused birth defects if ingested early enough during pregnancy. (Props to Breaking Bad for teaching me about it)
What are the odds there will be any, or a significant incidence of long term side effects? I have no idea.

36. Anon4 Says:

If there weren’t enough vaccines to go around then just giving them all away by random lottery would have given the same information as the Phase 3 box as long as you had a way to record whether those who won and lost the lottery became infected or had side effects, such as by monitoring electronic health records.

37. Hyman Rosen Says:

So disappointing that no one is talking about their own crackpot views, just covid.

Here’s mine – I don’t believe black holes exist. I’m not a physicist and have zero expertise in this area, but I have a gut feeling that tells me that I don’t live in a universe that allows singularities. I believe that the Einstein equations from which the mathematics of black holes derive are wrong, in the same way that Newton’s equations were wrong, and that someone will come up with a refinement of those equations that will reduce to Einstein in the cases that work now (GPS, orbit of mercury, gravitational lensing), but will also change things enough to explain the motion of the universe at galactic scales without needing dark energy or dark matter, and explain what happens under drastic gravitational compression without requiring singularities or event horizons and such.

Now, I’m not a deranged crackpot, so I believe that science is doing what it can, and isn’t going to throw away current theory before the better version is discovered, and it’s OK for people to propose epicycles until then. There’s no black hole conspiracy keeping the better physics suppressed. I don’t go around ranting and yelling about this. I just hope I live to see the day where I’m vindicated.

38. Anonymat Says:

I am surprised no one has mentioned mortality rates of covid as an objection to fast-tracking vaccination, especially for otherwise young and healthy people… While I tend to agree with some of the rationales behind accelerating timelines and favor challenge-trials, etc., my guess is that we would feel ourselves to be in a very different situation if the exact same scientific approach (i.e. same vaccine roll-out schedule) were followed under a different political/policy/PR regime. Covid is dangerous, but for the most part only mortally dangerous to people 65 and older. The media environment we all inhabit has from the beginning offered schizophrenic perspectives on the general risk of contracting the disease, appropriate measures for preventing its spread, and, above all, how we should incorporate a new risk like this into our personal and social lives. Given the heterogeneity of outcomes for those who contract covid and develop symptoms (not to mention the apparently large proportion of those who develop no symptoms at all), given the light-speed spread of both information and misinformation in today’s world, and given that those of us who live in the West are largely (and, I think, appropriately) skeptical of authoritarian solutions to political problems, it is not surprising that experts and government officials alike have had a tricky time threading the messaging-policy needle around vaccination. With, say, 50% less health-terror propaganda in the media; with even a 25% less obsessive focus on this important local and international but by no means the only or most existentially threatening challenge to us and our “way of life”; with a mere 10% less partisan teeth-grinding; and with all other things being equal, I imagine we would be happy with the present pace.

39. DavidC Says:

Hyman #37: Is that actually a crackpot view in the desired sense?

I don’t really have any idea, but I wouldn’t be surprised to hear it’s a common view among physicists. “Yeah, postulating singularities is the best we’ve got now but obviously that kinda stupid, there’s a good chance something better will come along…”

40. Scott Says:

Hyman Rosen #37: Finally a comment that I can answer with confidence! Back from the alien world of vaccines to the comforting and familiar topic of black holes. 🙂

I can tell you, if it helps, that almost no one believes in “singularities” as actual physical phenomena. Rather, singularities are simply the places where GR is telling us that it’s no longer applicable, and that it needs to be replaced by a yet-unknown quantum theory of gravity. Also, while the observational case for black holes was already strong before LIGO, today — after we’ve seen not approximately but exactly the gravitational wave pattern that GR predicted a century ago for two colliding black holes — it’s pretty much a slam dunk. Whatever’s going on on the inside, entities exist that behave just like black holes from the outside, and that would therefore arguably still deserve that name, no matter what future discoveries said about their interiors.

I can also tell you that, while dark energy and dark matter are indeed both mysteries, they’re mysteries that can very much be accommodated within the current framework of GR and QM — much like the top quark or the tau neutrino, they’re basically just more damn stuff to stick into your Lagrangian! Avoiding black holes is different, in that it would really require ripping up century-old GR and astrophysics in regimes where they’re believed to be extremely well-tested.

41. dankane Says:

Question: they were already doing early manufacturing of vaccine candidates while running trials. Would faster trials actually have helped significantly? I mean I suppose you could have vaccinated a small number of people early with the early production vaccines, and you could have switched your manufacturing efforts over the candidates that have been proven to work, but if it takes 9 months to really get your factories up and running and then another several months to produce a few billion doses, you couldn’t have inoculated a substantial fraction of the population by now anyway. Running trials of an experimental vaccine quickly seems like a good idea *if* the upside is that you end the pandemic early. If the upside is just that you prevent a small fraction of the early deaths, it is not clear that it is worth it.

I do agree that if the manufacturing process could be sufficiently accelerated we should have just run safety trials and then added challenge trials on volunteers shortly before those trials wrapped up. But at the very least, this would require taking a lot of organizational flexibility to take important safety precautions, determine that they don’t need to apply here and agree to circumvent them. And that would have been good now, but important safety precautions are organizationally hard to circumvent for a reason.

42. James Graber Says:

Hi Scott,
I agree 100% with your “vaccine crackpottery”. Congratulations in the extreme.
I strongly suspect that much of the political mainstream, left and right, in the USA and elsewhere, are making other serious errors and will not be stopped. Bummer!
Best to all,
Jim Graber

43. Jeremy H Says:

Re #17:

I think you may be dramatically underestimating the boring but real challenges of ramping up production and distribution as fast as you seem to be imagining. My understanding is that, going back to the spring, the companies involved began to develop their manufacturing & distribution *as if the vaccines were guaranteed to be approved and needed worldwide*–and that’s what got us to the number of available doses we have now.

Agreed. I think there’s an aspect of the Mythical Man Month and Amdahl’s Law here. At some point, throwing more resources at the problem has drastically diminished returns. Branch mispredictions are very expensive, not just in resources but in time to completion.

44. Domotor Palvolgyi Says:

I think you should look at it from the pharmaceutical company’s point of view. If they roll out the vaccines about the same time as other reliable companies, they get fame, money etc. Why would they risk that with making the vaccines earlier with a 5% chance of failure, which would result in humiliation, lawsuits, bankruptcy? This also shows why the Russians started vaccination earlier; a state owned company has different priorities.

45. Scott Says:

LP #6:

There is one argument that even now about 30% or more people are not willing to get vaccinated (even higher depending on the country). If it were available much earlier, that is bypassing all standard phases, we won’t get many willing to get it. I’m not sure I’d be willing to get a fresh off the lab vaccine in say May myself.

Already covered in the post. Personally, I was ready to be jabbed back in April! I wish I’d enrolled in a clinical trial — I would have, had I realized that even if I’d gotten a placebo or an ineffective vaccine (which with my luck I would have), I’d still be near the front of the line to get the real thing later.

46. Scott Says:

Richard #7:

If the vaccines by BioNTech/Pfizer or Moderna would have been approved for the public earlier, which rule would have been used? Would this rule have also applied to vaccine candidates by other companies? Would there be a risk of having a similar situation as we had with Hydroxychloroquine?

As I said a while ago, my “rule” would have been: put Bill Gates temporarily in charge of all pandemic-related decisions. Or really anyone with a proven track record who’s ~90% careful consideration of relevant facts, and ~10% aggressive action bordering on recklessness. Make it clear to that person, and to everyone else, that along with the power and plaudits comes personal responsibility for whatever goes wrong.

47. Daniel Harlow Says:

We are more or less in agreement here Scott, and I think that if the US government had decided in February that vaccine production should be the top national priority the various production concerns other commenters have mentioned could have been swept away. We can see what happens once Biden is in, although he won’t have the benefit of months of time to prepare.

There are serious ethical concerns with challenge testing, especially if it essentially turns into using the poor members of society as guinea pigs for the rest of us, but I think I can imagine ways of doing it that I would be comfortable with. Ideally this can be discussed as a society prior to the next pandemic, so that clear and well-thought out rules are in place prior to when they need to be put into action.

As for the “argument from Trump”, the only thing I can say is that it isn’t enough to say the right things: you have to also do them. And if Trump were actually able to mount a serious response to this pandemic, he would be a different person entirely and one who it wouldn’t have been such catastrophe to re-elect. In other words I don’t think there could have been a scenario where he unfairly took credit for a massive successful government response to covid-19: had it happened he would have deserved such credit, but there is no way he could have pulled it off. The only way he could have benefited in this way was if the virus had suddenly gone away on its own due to the summer heat, a lucky mutation, etc.

48. Anatoly Vorobey Says:

I think you may be dramatically underestimating the boring but real challenges of ramping up production and distribution as fast as you seem to be imagining. My understanding is that, going back to the spring, the companies involved began to develop their manufacturing & distribution *as if the vaccines were guaranteed to be approved and needed worldwide*–and that’s what got us to the number of available doses we have now.

Consider the Emergency Shipbuilding Program, under which the US built ships during WWII. In the 8 years before 1941, the United States built ~130 warships. In the five years of WWII, the US build 1250 warships, and 6000 ships of all kinds of military purposes (cargo etc.). The number of shipyards approximately doubled, and the existing shipyards’ throughput exploded.

Note that the number of covid-19 deaths in the US will soon approach the number of American WWII fatalities. In one year.

Your saying “there were real challenges faced ramping up production faced by the companies developing the vaccines” is like someone saying in 1941, “well the shipbuilding companies are going to face challenges ramping up their production”. It’s sort of not seeing the forest behind the trees. The US put so much money and resources into the problem that many new companies sprung up, the existing companies got massive help in solving their supply problems, etc.

If we approached the covid-19 vaccine project in February or even March with the same determination and urge as the US approached the naval warfare project from 1941 on, is it implausible that a billion vaccines could be ready by July (nevermind end of year), waiting for trials/approvals to be finished, if they weren’t yet? I think it’s very plausible.

The logistics can be worked out in any number of ways, once there’s a political will to do it. Perhaps a contract with Pfizer/Moderna, buying out their vaccine IP and having them provide training to ten times as many chemists, biologists and engineers recruited under a generous emergency program and put to work on testing variants and building manufacture capacity. Perhaps some use of the Defense Production Act, or some new law quickly put through Congress to mobilize industry.

Not clear which vaccine to focus on? Focus on the three most promising ones and build a billion of each. Yes, there are complex global supply chains, transportation and distribution challenges, etc. But ships are also quite complex things to build. Somehow when it was really very necessary, the American economy managed to do it very fast and in great numbers. The benefit of having an approved vaccine ready by July with a billion doses instead of late December with a few millions is saving ~1M global lives, nevermind avoiding the harm from subsequent lockdowns and such. Could it be worth trying?

49. Charlie Says:

Here’s my argument for why you’re wrong:

It’s February 24th. We’re still a month away from even thinking of locking down, and the market is barely beginning to show signs of worry. The mRNA vaccine has been developed, and is en-route to begin clinical testing.

Now, put on your empathy hat and think of yourself as the director of the FDA. Your job is to save lives. You know that, based on past experience, faulty vaccine rollouts can cost lives, and so for as long as you’ve been alive, there has been an approval process for new vaccines.

This new pandemic comes along. But places haven’t locked down yet. The market is still fine. There are essentially no cases in the US. So this new mRNA vaccine comes along. Naturally, you route this new vaccine into the usual channels, as this appears to be a fairly normal viral disease, much like Ebola or Swine Flu.

Now, I think that that is a rational choice, given the circumstances. My question to you is, at what point do you think it becomes rational for you, the FDA director, to upend decades of scientific practice and push for expedited approval of these new mRNA vaccines?

Is it in April, when the curve of new infections had been bent by lockdown measures?

Is it in June, when the curve begins to turn up again?

Is it even later than that, when it becomes clear that we won’t contain this virus without a vaccine?

My point is, whenever it becomes logical for you as FDA director to change course and push for a new approval regime, it’s already too late. The vaccine is in mid-stage trials. Processes and procedures can’t be formed overnight, and they certainly can’t be implemented overnight. We’re not a dictatorship.

Use your favorite lens, Bayesian analysis, and think: when did the world have sufficient evidence of the virus’s eventual outcomes to justify radically changing the world’s prior behavior, and what could have realistically been achieved, and at what speed, once that evidential burden had been reached? And does that really differ substantially from what actually happened?

50. mjgeddes Says:

>Having run Shtetl-Optimized for fifteen years, I’m acutely aware of the success rate of those autodidacts who think they’ve solved P versus NP or quantum gravity or whatever. It’s basically zero out of hundreds—and why wouldn’t it be?

Hang on Scott, haven’t I successfully outlined solutions to all of consciousness, time, AGI, complex systems and metaphysics right here on this blog ? 😀

The main ‘out there’ idea’s I like are that math and physics might not be eternal (unchanging), so I’d give serious considerations to things like intuitionistic logic where there’s no law of excluded middle.

Also suggest we consider applied category theory as ‘the theory of compositionality’ (how small systems combine into larger systems). Use polynomial functors as a model of computation instead of Turing machines.

When I combine the ideas in the above 2 paragraphs, I get a psychedelic new philosophy of mathematics. My suggestion is that math is actually a *physical* science, the science of *time*. That is to say, perhaps math is really about how smaller physical systems combine into larger ones such that it’s really describing the advance of time.

Then AGI, a math construct , is in fact a complete theory of the science of time , decomposing the concept of time into 3 different aspects: as causality (cause and effect, counterfactuals), as complexity (limits on computation: spatial locality and duration) and as compositionality (mereology: combining parts into wholes). And AI Alignment is really about blending smaller complex dynamical systems into larger ones.

51. Jeremy H Says:

Re Anatoly #48:

From the wikipedia article, the program began on Jan 3, 1941. The first ship to launch was Sep 27, 1941. So in terms of latency, it’s actually not too far off from what we’re seeing with this vaccine.

Crucially, they used a pre-existing design so the entire process was already known. So while ships are very complex, known complexity is a very different beast from novel research.

52. Concerned Citizen Says:

Scott,

I have been noticing that a lot of very low-quality comments in your comment section. I think it would be a great idea if there could somehow be two lists of comments, one for the intellectually contributive ones and another for the rule-complying yet not valuable ones. I understand the value of “letting both sides be heard,” but the people from the other side that you want to hear are the ones with frustratingly difficult to deny arguments (like Robin Hanson’s blog), not ones who are merely saying something other than what everyone thinks. So, to keep the principles of the open forum, while also maintaining a high average quality, it would be really nice if there could be a second location where all of the “here’s my opinion” stuff was available yet separated.

53. pete Says:

I am with the experts. A super-rushed vaccine in the political climate we are in, with compromised federal agencies, would be difficult for people to accept. Imagine the conspiracy theories. Even Fauci would be skeptical. I myself would wait a few months to let it play out in braver citizens.

There is one other argument in your favour, though. The longer the virus runs wild, the higher risk from mutations. Most say that the vaccine will still work but I’m not certain that’s true. Imagine a super-spreader virus without an upcoming vaccine. Scientists would quickly create a new one and I imagine that it would be rolled out in a few months this time.

54. Boaz Barak Says:

Scott, I agree with you that we could have had more vaccines in people’s arms at this point, and we could have obviously had far fewer people dead and hospitalized.

But the culprit is not some FDA regulators that weren’t sufficiently well versed in applying utilitarianism and bayesianism to Trolley problems.

Ultimately when all is done and the US population is all vaccinated, the majority of the avoidable delays will be accounted by delays in manufacturing, distributing, and administering the vaccine.

There are several reasons for this, but at the root cause is that the U.S. used a decentralized strategy for a national (in fact international) emergency. This is not some inherent feature of the federal system, and would have been different if there was any other republican or democrat president. If you compare Covid-19 with 9/11, there is a yawning gap between the scope of the emergency and the scope of the federal response in both cases.

The one exception is the (very wise and commendable!) decision of operation warp speed to insure/fund drug companies so they can start manufacturing vaccines before trials are concluded. But with everything else, including coordinating the response to the pandemic, regulations, resources for businesses and schools, and guidance and resources to distribute and administer the vaccine, the federal government has been MIA.

I don’t think there is some magic bullet that could have had the vaccine much earlier. But, a better response from most countries (and particular leadership from the world’s richest and more powerful country) could have resulted in a somewhat shorter time span and far fewer casualties.

I like the way Kai Kupferschmidt put it on Twitter:

My plan for beating #covid19 after watching this pandemic for one year? Easy: Test like a Korean, track and trace like a Singaporean, sequence like a Brit, crush the curve like a New Zealander, vaccinate like an Israeli – and behave like a human being..

55. Raoul Ohio Says:

Hyman Rosen #37,

I see your bet and raise you three (all in in cosmology), but given as my guesses for likelihood:

Black holes exist: 0.99,
Dark energy exists: 0.88,
Dark matter exists: 0.75,
Inflation happened: 0.50.

What about “Quantum Usefulness” (that a QC will ever do a “useful” task faster than a classical computer)? I don’t have any intuition on this other than the analogy with controlled fusion.

56. Raoul Ohio Says:

Boaz Barak #27:

Pretty much agree.

A related analogy: when I was a lot younger and knew some math and physics, I thought programming was trivial. That was before I tried to write any large programs.

57. JimV Says:

Or maybe we should have just mandated everyone without cofactors back in April to catch a mild case (low exposure) of COVID-19 on a rotating schedule to get to herd immunity. That might or might not have turned out well–just like a rushed vaccine. Counterfactuals are a bitch.

I’m happy to let the experts (people who know more and have had relevant experience) make those decisions.

Blogs are for opinions and you have to go to blog with the ones you have, not the ones you might want to have, but sometimes (as in the original phrase this is based on), I wonder was this blog really necessary? (Better than nothing, though.)

58. Grace Says:

Utilitarian thinking is anathema to many medical practitioners and medical ethicists. A doctor’s duty is only to the patient, not to anybody else, not to the patient’s family, and certainly not to society as a whole. When a doctor thinks about administering a drug, they cannot think about whether “it’s good for the world,” only about whether it helps the individual patient. I am exaggerating slightly to make my point, but this is a big reason for the opposition to challenge trials. You can begin to explore this line of thought in this recent essay: https://www.tabletmag.com/sections/history/articles/fernandes-doctors-who-became-nazis

59. James Gallagher Says:

Lot of good, bad and ugly comments posted above (quite a lot good actually)

I’ll just post a link that I posted earlier in the year to a UK report on the quite similar
1957-58 Asian Flu Pandemic, when it seemed vaccine approval was a quicker process…

.

History lessons: the Asian Flu pandemic

In 1957 it had all seemed initially quiet on the UK influenza front. Dr McDonald’s quarterly report (November 1956—March 1957) mentioned a ‘remarkably low level of respiratory illness so far this winter.’ However, a Times newspaper comment (17 April) that ‘an influenza epidemic has affected thousands of Hong Kong residents’ heralded the start of rapid movement across the East with 100 000 cases in Taiwan by mid-May and over a million in India by June. Five months after the Hong Kong outbreak it was reckoned to have traversed the globe. As an entirely new strain there was no immunity in the populace and the first vaccines were not distributed until August in the US and October in the UK, and then on an extremely limited basis.

60. mak Says:

Scott I appreciate your point of view and have likewise been frustrated by the slow approval. As Mateus already mentioned, Russia approved a vaccine very quickly, but was derided for it. China approved a vaccine even earlier, back in June, but only for its military. Even the UK was admonished in the US media for approving the vaccine a couple of weeks ahead of the US and vice versa, some admonished the FDA for the delay.

There are several ways to justify the delay:
1. this has all been a drill – to mentally prepare the public for a real pandemic/nuclear winter/unknown catastrophe (not to downplay the Covid death toll, but compared to something that might end civilization)
2. great way of forcing “green” habits – less driving, working from home, ordering online. Tech companies and big box retailers have been big winners in the bargain while mom and pop stores have suffered. The lower traffic was an excellent opportunity to trial self-driving cars – seems like a big missed opportunity.
3. The artificial economic recession might have a rejuvenating effect on the economy at large (like donating blood), although it has also come at a cost of great human suffering (like losing blood by bleeding)
4. It was a necessary emergency procedure to achieve a desired political outcome – will say no more
5. None of the above – simplest explanation is the policies reflect the medical consensus. It is clear the consensus did not evolve/respond rapidly enough. Perhaps the medical community fears that the public will demand similar accelerated cures for everything? Also what is good for the public might be bad for business. But a healthier population is always good for the economy.

61. Harry Johnston Says:

Not an astronomer, but I gather that the evidence for dark matter (in particular, the observation of gravitational lensing from dark matter) is pretty strong nowadays.

62. 1Zer0 Says:

I did not experience that there was an “unified expert opinion” of Virologists in general Covid matters – quite the contrary. In Germany the two most common public faces of the virologists community, Drosten / Streeck had vastly different recommendations. Furthermore and in specific when it comes to vaccines, predictions by experts on when a vaccine would be ready ranged from autumn 2020 (CuraVec Ceo https://www.faz.net/aktuell/wirtschaft/unternehmen/curevac-dietmar-hopp-hofft-auf-corona-impfstoff-im-herbst-16684254.html – sorry, in german) to 2023.
I am confident that the foundations of virology are well understood and agreed upon by scientists working in the field yet the real life recommendations and predictions were so general and broad that in retroperspective you can selectively pick an expert that was right by chance.
I do not believe the process could have been accelerated much. Other medicine including vaccines still have to be produced and distributed. Reminds me of the current graphics card shortages where similarly, additional production capacities can’t simply be added within a short time range. Processes just take time, physically.

63. Jay Daigle Says:

Grace #58 I think highlights a really important point.

In particular, a big response to human challenge trials (see James Cross #24 and entropy #31) is that you can’t do human challenge trials with the elderly and/or vulnerable. But obviously, you could. 1 Day Sooner collected a lot of volunteers for human challenge trials who were elderly or had comorbidities.

But the prevailing school of thought in medical ethics is that it would be unethical to do human challenge testing with those people even if they freely and informedly volunteer. I’m not sure I agree with this, but I can see the argument. What I find frustrating is that it’s treated as a received fact, and not itself a judgment about tradeoffs that is or could be up for public debate.

64. renato Says:

As many of the other commenters, i also completely agree with your ‘rant.’

I think the vaccine timeline makes it clear that there is a huge civilizational problem, but it feels like the biggest lost opportunities were something completely different and/or new that wasn’t tried, because it would not solve the problem completely. Regarding the vaccines and challenge trials, even if it covers only healthy young people, a faster rollout for that group could have reduced the overall transmission, and it would make a clear difference on lives saved. Then, the problem with the new and more transmissible variable (and the possible next ones) might have not be so hard, as we would still have more room to maneuver.

What bothers me the most, is that we could not even bother or try to implement different things to reorganize how we, as society, live to reduce the chance of infection of the most vulnerable populations, like really quarantining them and have younger and healthier people to support them from the outside, or of the general population, by shifting where people live to reduce the amount of time commuting and the exposure chances. Or who knows what crazy and innovative ideas that would seem obvious after they were implemented. Personally, i feel completely underused to combat the virus, as it also happens with most people who are well-educated but not in health sciences, be.and

As i see ourselves almost at the limit of what is possible to currently do, i wonder how we would have fare if the virus was slightly worse, more transmissible or more deadly. And it seems more probable than not that we would still have done the same things and suffered a much higher damage.

65. Scott Says:

Jesse #8:

I dont know the answers here but I suspect that even with an ‘instant’ regulatory approvals it may have only shaved a few months off.

Which, to spell it out, would currently mean that the regulatory process had “only” killed ~1M people, rather than ~2M or ~3M.

66. Scott Says:

Nick Nolan #11: I’m confused. If the main worry with covid vaccines is super-long-term side effects, then why is it OK to start vaccinating now? Why not wait another year, or three years, just to be safe?

67. Scott Says:

James Cross #14:

If you vaccinate a lot of people and a lot of adverse events arise, you will scare too many people in the population from taking the vaccine or maybe any vaccine.

I understand the above problem—which, and let’s be clear about this, is a problem of PR and messaging, not of ethics or of underlying biochemical reality. That’s why I described this as a problem for our entire civilization rather than merely for the medical establishment. It demands society-wide coordination of the sort that we were capable of in WWII but are not now.

68. Scott Says:

Art #16:

As soon as you are allowed to compare lives, you get very real questions about letting someone die to distribute their organs amongst those needing donors. As soon as you have a norm for even doing such a computation public trust is gone.

I understand, and am not a thoroughgoing utilitarian myself. For me, though, the magic ingredient really is the informed consent of everyone involved! We’re not talking here about ever making a covid vaccine mandatory, let alone having made one mandatory back in April.

On top of that, though, we’re not even asking, let alone demanding, that anyone sacrifice their lives so that others might be saved. Even with human challenge trials, all we’d ask of the volunteers is that they accept a slightly increased risk—just like we’ve always asked of police, soldiers, firefighters, and all the others traditionally regarded as “heroes.”

69. Dániel Says:

My understanding is that, going back to the spring, the companies involved began to develop their manufacturing & distribution *as if the vaccines were guaranteed to be approved and needed worldwide*–and that’s what got us to the number of available doses we have now.

Let’s model vaccine availability time as maximum(time for getting approval, time for ramping up production). You say that in this case, approval time and production time were comparable, right? Then we should speed up approval IF AND ONLY IF production time will be faster for the next pandemic. Dear commenters, if any of you have insights about the production side, please share.

70. Scott Says:

James Cross #24:

Aside from ethical issues, in the end not much would be accomplished because the people doing the challenges would not be similar enough to the population that most needs to get vaccinated, that is older people with various conditions.

As Jay Daigle #63 points out above, the idea that older people, or people with preexisting conditions, could never volunteer for challenge trials is yet another soft constraint treated as though it were a hard one.

71. Scott Says:

Avraham Eisenberg #32:

There’s selection bias here. The handful that work are retroactively considered major. As of now, WHO is tracking 232 different vaccines in development! Out of those, 3 have been approved.

Way back in March, I remember three vaccine efforts that people brought up as the most “major” ones: Pfizer/BioNTech, Moderna, and Oxford/AstraZeneca. These are exactly the three that we’re talking about now. Is my memory faulty? Which other ones were considered equally promising back then but didn’t pan out as well?

72. Scott Says:

Scott T #33:

When there wasn’t an attractive nurse knocking on my door with a vaccine the day after its approval my admiration for [Gates] decreased and it is incredibly angering to have to wait for known beneficial vaccines while people die.

The logistics of distribution is something where computer scientists could have or could still make a difference instead of being bystanders. We pay university computer science and operations research departments a lot of money to write papers about scheduling and routing. So how should vaccines get from UPS trucks to arms? Israel seems to be doing a much better job with this than the US.

I might not be able to contribute at the level that you are — namely, by waiting around for an “attractive nurse” to bring the vaccine to my door — but as a computer scientist, my professional opinion is that we in the US should just copy what Israel is doing. Pull out all the stops, vaccinate everyone in the country 24/7, enlist the military if needed, don’t break for the holidays. Or should I dress that up in the language of convex optimization? 🙂

73. Scott Says:

mjgeddes #50:

Hang on Scott, haven’t I successfully outlined solutions to all of consciousness, time, AGI, complex systems and metaphysics right here on this blog ?

Like I said: zero out of hundreds. 😀

74. Scott Says:

Grace #58:

Utilitarian thinking is anathema to many medical practitioners and medical ethicists. A doctor’s duty is only to the patient, not to anybody else, not to the patient’s family, and certainly not to society as a whole.

Doesn’t that sit extremely uneasily with the public-health rationale for vaccines themselves, which really are for society as much as they’re for the individual patient?

75. Lou Scheffer Says:

Scott, have you ever personally tried to scale up production of any physical object by several orders of magnitude? Perhaps it’s harder than you think from outside. Suppose, for example, that you have just figured out how to make 1000 doses, and you start a human challenge trial. But you don’t want to waste time, and you have infinite money. How fast can you make a billion doses (million fold scaleup) Well, you can’t do it just by doing what you did before. If it took 10 technicians to make 1000 doses, you’d need 10 million technicians. That’s not possible even with infinite money. So now you need to automate – but someone has to design, test, and debug that (and you certainly have experience that assigning 100 people to debug a single problem will not make it 100 times faster). Let’s suppose you can scale it up by 1000; now you need (for example) a PCR machine to test the output. But you can’t scale that either – there are not 1000 PCR machines available, unless you take them from the testing centers. And no amount of money will make them available instantly, as we saw with much simpler PPE. So you solve this and scale up again – now you need sterile production space! Where do you get that? There’s not enough existing space, and it takes time to build. And so on.

I think an analogy to “If we spent trillions of dollars, we could have billions of doses by now” is “I can solve this problem in 100 seconds on one computer, so if we give it an all-out effort, we should be able to solve it in 1 millisecond on 100,000 computers within a few months”. Now as an expert in computers, you’d say it’s not that easy -what about Amdahl’s law? We’ll need to re-code the algorithm! What about the speed of light? How close can you pack those computers? Do we have an available data center with enough GPUs? If we need 100,000 GPUs, where can we get them? And so on.

A good rule of thumb is that every time you scale up by a factor of 10, new problems occur which require human thought and effort so solve. So I think even if there was no regulatory hurdles at all, we’d be pretty close to the production curve we are on now.

76. Lou Scheffer Says:

Grace #58 and Scott #74:
There is certainly precedent for vaccines that do not help the patient but help others. For example, the CDC recommends (and most doctors follow) vaccinating boys with the HPV (human papilloma vaccine), which fights the virus that causes ovarian cancer. This does not help the males whom the virus does not harm (in fact it’s probably worse for them, as any vaccine risks side effects), but it reduces transmission of the virus dramatically, which helps females considerably.

77. Raoul Ohio Says:

Harry Johnston #61,

You kind of have it backwards. Nobody doubts that something fishy is going on with gravity at large distances. The obvious plan A guess is that this is caused by dark matter. The doubt arises because there doesn’t seem to be anything there, so maybe it is time to think about plans B, C, … .

78. mak Says:

In hindsight, the reason for the delay in the vaccine approval may be something mundane. There’s always a delay in human testing because the vaccine candidates have to pass animal tests first. I recall reading earlier when the vaccine candidates were ready for human trials it was close to summer and the infection numbers were plummeting due(?) to the lockdown. There was a fear there would be no way to check the efficacy of the vaccine in the natural environment (barring challenge trials). So the second wave in the fall may have actually helped to complete the vaccine trials.

BTW China trialing the vaccine in the military makes sense from a philosophical perspective. After all they signed up to defend the country against all threats. Of course the risk is compromising your defense if there are long term adverse effects.. can imagine Michael Crichton weaving a nice story from this hypothetical scenario. The US appears to have chosen “frontline” medical staff to play the same role.

I don’t think the world was hobbled by manufacturing capacity .. a company in India was pre-emptively ramping up production of “hundreds of millions” of vaccine doses back in July even before the trials were completed. They got some additional funding from Bill Gates.

79. Hugo Says:

On the topic of vaccine safety, one thing that is worth pointing out is that the worse case scenario for a faulty is something so catastrophic that goes beyond any utilitarian argument. For example, the latest attempt at a dengue vaccine failed because getting the vaccine *increased* the risk of developing the most serious form of the disease (possibly due to things like Antibody Dependent Enhancement). It’s not just a matter of rare side effects.

That said, I agree with the overall sentiment that from now on we’ll be certainly asking ourselves how we could have sped things up in 2020.

80. penttrioctium Says:

Lou Scheffer #75:

Your argument, as best as I can tell, is that the supply of vaccines is not infinitely elastic in price, and therefore said elasticity is ~0; you do not appear to attempt to account for the fact that there exist numbers between 0 and infinity.

Or: you showed why it had to be the case that {number of months till sufficiently inoculation} > 0; you have not explained why {number of months till sufficiently widespread inoculation} = 18. Do you actually have a good enough understanding of the production process that you can explain why it had to take *exactly* 18 months, as opposed to 12, or 9, or 24, or 100, or 6? If you can’t give an actual true & detailed explanation, then I don’t think you have enough domain-knowledge for anyone to reasonably trust you.

As for the crackpottery question, the closest I can think are:

• What you said about covid (which has made me sour greatly on the regulatory state in general — I have to admit that I was wrong in dismissing the concerns of libertarians; their advance predictions bore out).
• I think that (1) the Federal Reserve goes about hitting their goal wrong, has the wrong system for figuring out how to hit their goals, and has the wrong goal; despite that, (2) the Federal Reserve is the single most consistently highly competent part of our government and has been for the past ~40 years. In practice, these two countervailing factors about cancelled out and they usually get close to the thing that they should be aiming for; the main thing leftover is that, as of recently, (3) the Fed causes slightly too little inflation in the US economy normally, and way too little inflation during big economic crises.
• The AGI threat is very real & very serious & very plausibly short-term & I have no idea what can be done about it because the problem seems almost completely intractable (much more so than creating an AGI at all).

Like with the vaccine question, I’m not quite sure any of these are really “crackpot” opinions; I think that support for them is higher among the relevant experts (economists who specialize in the FDA, economists who specialize in monetary policy, and computer scientists, respectively) than is appreciated.

(funnily enough, all three of these “crackpot” opinions of mine can directly or indirectly be translated back to the same guy; after time proved him right on covid, and proven that I had blown off concerns too confidently relative to my state of knowledge, i decided i better take a closer at the thing he’s more well-known for…)

81. penttrioctium Says:

Grace #58:

A doctor’s duty is only to the patient, not to anybody else […]

If I had the option, I would have participated in an HCT in March. (If I both had the option and were smarter than I was, January…) So it sounds like you agree that the regulatory bureaucracy had a duty to let me do that!

We’re not talking about intentionally infecting/vaccinating people against their will for the greater good here! (To the contrary, I think the alleged “noble lies” pronounced by certain public health officials in the name of the greater good were a terrible idea.)

82. Shmi Says:

Copying my comment on a Zvi’s post:

The Covid-19 mortality rate is in the Goldilocks zone for allowing (bad) choices:

– If the mortality rate was 20+%, the choices of herd immunity, doing slow full-scale trials or doing nothing would not be on the table. It would be “let’s try everything, anything, now, now, now!” and the vaccines made in February would have been produced and used before summer.

– If the mortality rate was within an order of magnitude of the annual flu (0.1% or so), “do nothing” would have been the only sensible choice.

As it is, the 0.5%-3% mortality rate is exactly the wrong number, since the right decision is not immediately obvious to everyone. And so we have the largest number of overall deaths and the largest damage to the economy possible from anything short of Oryx & Crake-style plague.

83. Shmi Says:

As for the crackpottish beliefs of my own:

– Black holes do not evaporate, and the standard semiclassical calculation (QFT on a curved background) showing that they do is, while mathematically correct, based on bad assumptions.

– Gravity does not exist on a microscopic scale. That is, “elementary particles” still exist in the background spacetime, and so neutrons fall down in Earth’s gravitational field, but it takes a collection of entangled states to *produce* spacetime curvature, hence the link between gravity and entropy. Moreover two systems that are not entangled do not affect each other gravitationally, either.

– And, the most crackpottish (though less original) of all, there is no “reality”, it’s models all the way down. “Reality” is one of those models.

84. LGS Says:

Scott, I mostly agree with you, but there are a few places I think you exaggerate.

First, you say that everyone was talking about Pfizer/BioNTech, Moderna, and Oxford/AstraZeneca as the three leading candidates “back in March”. I don’t think that was the case. In March, the big story was that Trump was trying to get an exclusive deal with CureVac for their vaccine. Only later (maybe May?) was it clear that those three are the leaders.

Second, you claim that you advocated for human challenge trials “back in March”. I don’t think you mentioned challenge trials in March anywhere. You change the banner of your blog on (I think) April 9; before that you were talking about ventilators. Marc Lipsitch only came out with his call for challenge trials on March 31. I think he was the first major person to do so; the website 1daysooner.org, to which you link in your current post, only started existing in mid-April, as far as I can tell.

Third, I think that for a new technology like mRNA vaccines, at least safety trials (phase I/II) were necessary. Some experts were quite skeptical of mRNA back in March; for example, on March 16, Derek Lowe said: “That’s a promising field, but mRNA vaccines have yet to prove themselves in humans, so expecting one to ride to the rescue for 2019-Cov2 is asking a lot.” Based on our knowledge of mRNA vaccines in March, it could easily have been the case that they are neither effective nor safe, in which case vaccination would be a pure harm (with no upshot whatsoever).

I think some testing, at least for safety, would have been necessary before mass administration of vaccines; indeed, even Russia and China opted for safety trials before rolling out their own vaccines (both based on adenoviruses, by the way, similar technology to Oxford/AstraZeneca).

Another thing to remember: back in early April, we thought that lockdowns followed by contact tracing were sufficient to control the virus. Since we believed that to be a plausible path, it would not have made as much sense to rush forward with a less-than-perfectly-safe vaccine (and risk a huge anti-vax backlash if the vaccine later turned out to suck). Given our current knowledge that lockdowns + contract tracing don’t work in the West, a rushed vaccine looks better in hindsight, but we didn’t know that back then.

Putting it all together, I don’t think a rational world would have vaccinated everyone in March or even April. However, a much stronger case can be made that a rational world would have vaccinated everyone in July, when we already had some Phase I/II results (and we may have been able to fit challenge trials by June, if we went at maximum speed). Also, by July (unlike in April) it was clear that the virus would not be contained by contact tracing or lockdowns.

85. Danijel Kecman Says:

You and toddler are wrong. A bad vaccine, or the one with side effects, would cement already a high percentage of anti-vaxxers in their pledge not to take any vaccine. Where would we stand then on a long them with the new mutations and the herd immunity (as the only mean to prevent the new mutations) achievable only via vaccination. It’s hard as it is.

86. David Says:

I think the scientific response has been fantastic, it is the political response that has been woeful. (mentioning no names). The west received the draft genome of the new virus even before the arrival of the virus itself which allowed work on a cure to get a head start and today, within a year, public vaccination has begun. I think that is pretty good going.

The first rule of medicine is not to cause harm and when you are going to inject the world’s population with something you want to be pretty damned sure it will be safe and that the public can have confidence in its safety. The only way this can be achieved is through public trials which will inevitably take time. What is required during this period is simply that the voices of those who know what they are talking about be heard and are not drowned out by nonsense.

There will be other pandemics and I hope lessons can be learned from the response to this one. The next time It may prove impossible to design an effective vaccine at all.

87. Ashley Lopez Says:

Scott #40,

OK, from one non-physicist to another, where did you learn all this physics from? (I mean, can you point me to the resources you used?) Or was it just by virtue of being smart and having a lot of physics professors around?

88. Nick Nolan Says:

Scott Comment #66

Because drug regulators accept a higher than normal calculated risk. These are emergency use authorizations. Meanwhile Phase III trial continues and data is still coming in. There exists post-authorization/post-approval vaccine safety monitoring infrastructure that is scaled up for COVID-19 vaccination program.

My understanding of this issue is that you need to do two things:

1) N=10,000+ to cover enough genetic and other variability

2) △t (aka time) to see what happens. You call it “sat on those vaccines for a year, ticking off boxes” (aka clinical trials, methodology)

You argue that current △t was too long, but you have no good reasoning behind it.

Your suggestion, approval by March-April, would have been reckless. Almost no clinical trial data and analysis. Approval at the current level is a compressed timeline with preliminary data from ongoing Phase III available. Risky but acceptable risk. Waiting for a year would have reduced the risk even more.

As Derek Love points out every therapy is a tradeoff:

I do not want to provide ammunition for the anti-vaccine camp by going into these details, but we can’t ignore medical (and mathematical) reality. All drugs have side effects, and every therapy is a tradeoff.

What does this mean?

89. Renato A. Laguna Says:

I kind of believe in Brouwer-style intuitionism.

That every function is continuous and every set is a topological space, with topology being intrinsic rather than “tacked on” to set theory.

That classical sets with more than one natural topology (think Banach spaces) will split, by intuitionistic logic, into non-equivalent sets in a way that makes sense.

On classical vs. intuitionistic logic, consider:

https://www.theguardian.com/science/2016/mar/28/did-you-solve-it-the-logic-question-almost-everyone-gets-wrong

The article gives an example logical problem where the answer by classic logic is one, and by intuitionistic logic is another. It frames it as “most people are dumb LOL” but maybe it means that most people are intuitionistic.

90. Lou Scheffer Says:

Scott and Piotr #22:
I agree that in this case, human challenge trials may have been helpful. But I can also see why people are reluctant to try them. Ethics rules change slowly, and only after massive and lengthy debate have explored all sides of an issue. In general that’s a good thing.

For example, it’s in general not OK to kill people on purpose when they’ve done nothing wrong. Now you can say that if they give their informed consent, it should be OK. But this is a very tough sell for good reason – what exactly is “informed”? Can you really give informed consent to something you’ve never experienced? Exactly how mentally impaired is too impaired to give consent? What if you change your mind? And allowing this is open to abuse – “Honest, your honor – she said just kill me now”.

Now apply this to human challenge trials. Suppose you have a vaccine candidate with a 1% chance of working against a very uncommon disease that’s 99% fatal? Should you allow human challenge trials? Likely not, for the reasons above about the limits of informed consent, etc. But a human challenge trial for COVID is exactly the same conceptually, just with different numbers. So you are left with a numerical utilitarian argument, which in itself has uncertainties, like how much faster rollout would go if you had challenge trials. There is some chance the vaccine does not work (or makes things worse), or the bottleneck is not proof of efficacy, and so on. So now you need convincing error bars on all your estimates, and they need to be *very* convincing as death is a possible outcome.

And many people are historically skeptical of utilitarian arguments because they might be abused. Imagine, for example, a government decided that overall, the country would be better off if all people of type X were killed. This is an entirely utilitarian argument, but widely considered unethical.

So overall, ethics rules change slowly, and tend to very general statements, such as “Don’t give fatal diseases to people on purpose”. Exceptions occur only after protracted debate and with very certain data. In this case you can argue for an exception on utilitarian grounds (which I think is correct in this case) but I can see why some folks are skeptical of such arguments.

91. Gerard Says:

@Raoul Ohio #56

> A related analogy: when I was a lot younger and knew some math and physics, I thought programming was trivial. That was before I tried to write any large programs.

It sounds like you and I have a few things in common. I also knew some math and physics when I was a lot younger but I still thought programming was trivial even after writing some large programs.

What changed my mind was the year my job consisted of trying to create a new programming paradigm for machine learning code while simultaneously using it to solve handwriting recognition. Well that, plus the last 10 years I’ve spent trying to prove P = NP (or more accurately trying to find an efficient algorithm for an NP-complete problem).

92. LK2 Says:

Indeed this is a bit crackpottery. It is not your fault if you live in a country which values personal freedom much more than public health. A country where you get the best cures in proportion to your bank account. A country which did way worst than everybody in dealing with this infection. Vaccine development has his time and money cannot change it by much (the all-mighty and rich US is clear evidence). I’m not criticizing you (if just a few % of the US population were like you, it would be the best place on earth) but your country. US is a great place, but the health system from my point of view is just criminal. Add to this a disaster-administration (trump and friends) and that’s it. I’m sorry, but by now you are old and have traveled the world enough to realize all this.

93. Gerard Says:

Grace #58

It seems like the key distinction between challenge trials and what Nazi doctors did (sorry I didn’t have time to read the essay you posted) is that challenge trial participants would be consenting volunteers.

I believe a doctor’s duty should indeed be to the patient but that means to what the patient wants, not to what the doctor thinks he should want. I also suspect that this tendency many humans have to want to impose their own desires on others has a great deal to do with why many doctors became Nazis or collaborated with them.

Lack of respect for the desires of others seems to be one of the main causal factors driving human inhumanity and societal insanity and it is one of the reasons I am very skeptical of utilitarian ethics, which seems to seek to impose the same utility function on everyone.

94. James Cross Says:

#70

It isn’t just a matter of whether older people could be allowed to volunteer. It is a question whether they would volunteer in large enough numbers after becoming aware of the risks. And if they did volunteer, would you still end up with a self-selected group of mostly healthy volunteers instead of a mix of volunteers across ethnic groups with a variety of conditions that would not be in any way reflective of the older population.

Safety considerations seem primary because we already have a large number of people who won’t get vaccinated. Misinformation is proliferating across the web about dangers and risks without adding to it reliable reporting about actual adverse events that could arise from a bad rollout of the vaccine.

Full disclosure. I received the Moderna vaccine last May. Originally I was supposed to receive the 250 microgram dose but that dosage group got dropped from the study after adverse reactions. So I actually received 25 microgram doses. Likely I have or had antibodies but probably not as many as those receiving the 100 microgram dose that is being distributed. My behavior didn’t change at all from receiving it. At this point, there is little or no evidence to suggest whether I would likely have any more immunity than anybody else. And that is another risk that still exists with rushing this out. Nobody knows how long the protective effect will last. So the best hope is that the Biden administration can get enough vaccine in people quickly to shut down the spread before any possible degradation of protective effects starts to happen.

95. Scott Says:

LK2 #92: Suppose I agreed with you about everything wrong with the US. How would that change the case that the year it’s taken us even to start distributing vaccines, slowly even now, leaving millions of doses on the shelf possibly to spoil while the pandemic rages all around, is yet another thing that’s wrong?

96. Robson Says:

I think we overlooked a huge mas of volunteers that have already pledged to risk their lives to save their country: the army. Why is it ok to tell them to die fighting other humans but it’s not ok to tell them to (with much much lower probability) die fighting a virus?

97. Lou Scheffer Says:

penttrioctium # 80 says:
“Your argument, as best as I can tell, is that the supply of vaccines is not infinitely elastic in price, and therefore said elasticity is ~0; you do not appear to attempt to account for the fact that there exist numbers between 0 and infinity.”

No, my argument is that scaling up the physical production of things is a technology all its own. In theory, if you can make one of anything, you can make a billion. But in practice, scaling up requires new techniques (you can’t just assign a billion people, for example) and developing new techniques takes time. A common problem is that stuff easily available in quantities of 1 may not be available in quantities of a billion at any price. If you wanted a billion doses in June, you’d need a billion vials. Where will you get these on short notice? Ok, maybe don’t use vials – but then you need new procedures, etc.

If you look at all-out, no expense spared, projects (atomic bomb, radar in WW-II, Apollo program) they still took years despite the best available experts working around the clock with no budget constraints. I believe this case is similar – we have the best experts on vaccine production working around the clock with few (if any) budget constraints. But we don’t have billions of doses today even though they began working on large scale production before the clinical trials. So I’m skeptical we could have had many more doses available much earlier. Note the “many” and “much” – could we have done somewhat better? Of course. I’m saying the elasticity is small, and determined by human ingenuity, not money spent. It’s certainly not zero.

98. Michael Weissman Says:

Scott- I agree with the general idea but think you’ve gotten a little unrealistic on 2 points.
First, on the current 1-dose issue, there is a possible public health disaster that may require sticking to something like the two-dose protocol, but without fetishizing exact timing. The antibody data show that dose 1 is way less effective than dose 2. Moderna’s rough asymptomatic data show noticeably weaker short-term protection for that from dose 1 (~60%) than for symptomatic (~90%). Apparently it’s not all that unlikely that 1-dose people can be contagious. So if you have lots of 1-dose people around for a long time you’ve got a situation analogous to lots of people having low doses of antibiotics- a perfect breeding ground for resistance to the treatment. Not good. These vaccines are effective but narrowly targeted. Breakout is a very real possibility.

Looking back on the acceleration issue, mass vaccinations don’t require that you have perfect calibration of rare side effects but they do require great confidence that your main effect has the right sign, without antibody-dependent enhancement. So some real testing is required before massive shots begin. Doing an even better job of parallelizing initial monkey trials, calibration of challenge doses for challenge trials, beginning building production facilities for only the vaccines that did very well in challenges, and then extending non-challenge trials to more vulnerable populations surely could have shaved some time off the delays. Maybe a month or two? As it turns out, with the terrible second wave that would have been crucial.

99. Boaz Barak Says:

Scott: As far as I know, it’s not the case that Pfitzer, Moderna and AZ were sitting on the same vaccines we have now since March. They had the broad approach but during the trials they did narrow down variants, figure out dosage and regime, and I’m sure many other things.

I’m also not sure by how much (if at all) human challenge trials would have sped up the process. It’s unwise to giving shots to 100s of millions of people without having a trial of at least 10K people and monitoring them for at least a couple of months or so. Human challenge trials could have helped a lot if the community prevalence of Covid was so low that the Phase 3 trials would have taken many months to get statistical power. But that wasn’t the case.

Regardless, the fact that despite pre-approval manufacturing, we still have far fewer vaccines on shelves than we need and fewer shots in people’s arms than we have on shelves, suggests that the regulation process was not the main bottleneck.

So I contend that it’s not about anyone not doing the right trolley problem calculations or having the wrong attitudes about utility and risk, but more about manufacturing & logistics.

When people evaluate the world’s response to Covid-19, there will be many failures, but it’s not about one single big wrong decision , but rather many small ones. Most of all it’s a failure of leadership, especially at the US, whose institutions such as CDA and FDA were considered the “gold standard” and were model for many countries. Normally the US would have been leading the world’s response. Instead, the federal government didn’t even want to lead the US’s response, and preferred to push as many unpopular decision and responsibility as possible to the states.

100. Sandro Says:

@Harry Johnston #61:

Not an astronomer, but I gather that the evidence for dark matter (in particular, the observation of gravitational lensing from dark matter) is pretty strong nowadays.

Recent work has shown evidence of an external field effect in galaxy rotation curves across galactic superclusters, which is consistent with modified gravity rather than dark matter.

However, Khoury et al. proposed an idea to unify the effects seen from both modified gravity and dark matter. They suggested that dark matter acts like a superfluid, which would present like the same external field effect from modified gravity, and so achieve the best of both worlds.

101. Sandro Says:

I’m sure I have many crackpottery ideas. My latest is that a wealth tax is not really all that crazy an idea.

I typically open by considering how your insurance premiums are calculated: based on value of the asset and risk of loss.

So consider a wealth tax as national stability insurance. The wealthier you are, the more you have to lose and the higher your insurance premiums ought to be to insure that wealth against loss, either by civil instability, foreign interference, etc. The usual government responsibilities. The fact that the wealthiest people in Western nations pay almost nothing to preserve their status against any threats looks progressively more criminal on this view.

This view also immediately sets up progressive taxation along two lines:

1. a rational actor of low-wealth stands to gain considerably more by turning over the table (revolution) than continuing the game, so redistribution of stability insurance premiums to those who are losing the game ensures they continue to play, thus preserving the status of the wealthy,

2. this justifies increased premiums (taxation) and increased redistribution in circumstances of high risk in order to reestablish stability (like the current situation we find ourselves in), and decreased premiums in periods of low risk. Of course, directly increasing taxes in the middle of a crisis isn’t necessarily the right way to do this, so some hysteresis is often applied.

One objection to a wealth tax is that it’s difficult to assess wealth. This can be true, and yet this doesn’t seem to be much of an impediment for any kind of ordinary insurance. If estimates must be conservative, so be it, people will adapt.

Another objection is that wealth is often tied up in investments and not real, demonstrable wealth. Bezos for instance, is valued at over $200 billion dollars, but most of that is the value attributed to Amazon stock, which fluctuates. The obvious response is questioning the assumption that Bezos really is worth over$200 billion. Could Bezos walk into a bank and get a $200 billion loan leveraged against his Amazon holdings? Of course not, but the bank might comfortably loan him$1 billion. So clearly Bezos’ actual wealth falls somewhere in between, and I have not seen any convincing argument that a reasonably conservative calculation couldn’t be devised. Clearly banks have some idea of how to make these estimates since they take such risks this all the time and yet are extremely profitable businesses.

Another objection is that wealth taxes have been tried, and they have often suppressed output and employment. Leaving aside the obvious response that they might have been poor implementations (no true scotsman), what if the output pre-wealth tax was based on overly optimistic assumptions that amount to socializing the risk of loss, like what happened in the 2008 crisis? The wealth taxes that were tried could actually have been a correction that put us on a firmer economic footing, and therefore actually a desirable outcome.

A final objection is that the wealthy have the means to move their assets out of regions that leverage wealth taxes, and so a wealth tax would ultimately be ineffective. This concern has some merit. Firstly, while some people did move when wealth taxes were imposed, not everyone did so I’m not convinced it’s a problem, particularly if this idea catches on, ie. where would they move to?

Furthermore, the incentive to move out of such regions requires two conditions: wealth taxes AND the ability to move wealth across national boundaries without restriction or penalty. It’s telling that the latter condition is taken as a given, and so the former must be the real problem.

So in summary, the arguments against wealth taxes that I’ve seen aren’t really all that convincing. I don’t dispute that it presents many challenges, but it seems intrinsically right in a similar way that Scott has described.

102. Gerard Says:

Scott

> do you hold any opinions, on factual matters of practical importance, that most everyone around you sharply disagrees with?

I probably hold many such opinions. Here’s one that I’m willing to share (though I suspect I’m not actually in a small minority on this, but it sometimes seems like I am):

I don’t believe that classical computation is a sufficient condition for the existence of conscious experience but I do think that machines capable of exhibiting human (or even superhuman) levels of intelligence will become available in the not too distant future.

This is (or rather will become) of practical importance because you already have people arguing that such machines should be accorded something like “human rights”. To me that idea is absurd and I certainly don’t want to live in a world full of intelligent machines that have to be treated as if they had feelings. In fact if such a belief does become prevalent it’s exactly the sort of human flaw that an AGI might exploit to control or eliminate humanity.

Now at first glance the elimination of humanity doesn’t necessarily seem like a bad thing, because I think that the utility of existence relative to non-existence is certainly negative.

Unfortunately I also think that total non-existence is probably not attainable (or at least not easily) and further that I would hold that belief even if I were an absolute materialist.

So in the end the prospect of the elimination of humanity by a collection of intelligent but non-conscious machines does not seem appealing.

103. mjgeddes Says:

penttrioctium #80

“The AGI threat is very real & very serious & very plausibly short-term & I have no idea what can be done about it because the problem seems almost completely intractable (much more so than creating an AGI at all).”

I’m fairly sure that the solution is ‘simple’ for a posthuman 😀

A mind with general intelligence is a complex dynamical system, which , as I mentioned, models time in 3 aspects: causality, complexity and compositionality.

Alignment is clearly related to the ‘compositionality’ aspect, it’s about how smaller complex systems are embedded in larger ones. And applied category theory is the mathematics of compositionality.

If we look at the table below, AGI is a straight-forward generalization (indicated by —- > ) of 8 different fields. If you start at any one of these 8 fields and generalize, you end up at the same place (the theory of AGI). So in fact, because of the interdisciplinary nature of it, a lot of the right ideas would already exist in bits and pieces in other fields, and it’s a matter of spotting the commonalities and generalizing.

Mathematical Logic — > AGI
Computational Logic– > AGI
Information retrieval — > AGI
Moral Philosophy —- > AGI
Linguistics —– > AGI
Neuroscience — > AGI
Signal Processing — > AGI
Stat Mechanics — > AGI

Human psychology is enormously complex, there’s no hope for us ever getting a handle on that level of complexity. But the underlying *general principles* of value systems might be a great deal simpler, because you can abstract away a lot of the details to get much simpler theories, as suggested by the table above. For instance, artificial neural networks, whilst *inspired* by real ones, are much simpler (the generalization from neuroscience successfully abstracted away a lot of the complexity).

104. Blissex Says:

Focusing on early vaccine development is quite pointless (and wrong, there are immense risks in hastened vaccine development).

The countries in which the vaccine is not yet necessary, like China-mainland, China-Taiwan, south Korea, etc., have controlled the epidemic using pervasive test-and-trace and isolation of the infected in quarantine centres, whether those in the country or arriving from other countries.

That has been quite cheap, and their economies are working mostly normally, with much smaller on their political economies. But it is not allowed to happen in reaganista/thatcherite countries because:

* It requires for the state to fund and organize a comprehensive public health system ahead of time and to fund and operate a its expansion when the need comes (and that’s COMMUNISM :->).
* It means demonstrating that state funded projects in the public interest can be quite effective in solving public issues (and that’s COMMUNISM :->).

105. Lou Scheffer Says:

Daniel #69 says:
“Then we should speed up approval IF AND ONLY IF production time will be faster for the next pandemic. Dear commenters, if any of you have insights about the production side, please share.”

For the next pandemic, the best proposal I’ve heard is that of virologists such as Florian Krammer, to make vaccines for all of the 100 most likely families of viruses (he thinks it is overwhelmingly likely that the next pandemic will come from one of the known families). These would use a common base technology (RNA or influenza vaccine, for example) where we already have a large production capacity and existing safety data. For each vaccine, run Phase I and II clinical trials, tweaking until you have an effective formulation. Then put each one on the shelf, to be pulled off during the next pandemic.

He estimates that in bulk, this could be done for $20-30 million per vaccine, or$3 billion for the lot. This would speed up future responses enormously. As soon as you see a problem, tweak the vaccine, start Phase III trials (which are very likely to succeed) and crank up manufacturing. As soon as they are determined to be safe you’d have plenty of doses.

He also thinks the odds that the US government, at least, would have this foresight is very low.

106. Blissex Says:

As to the immense risks of insufficient testing of vaccines, there are some points that are often forgotten:

* As to safety: there could be effects that happen only in the “long term”, let’s say more than 6 months, and the only way to know about them is to test in the “long term”. You cannot test for safety after 1 year exposure in 3 months, it takes 1 year.

* As to safety: there is much wider genetic/biochemistry variation among people than most people think, and it could turn out that a vaccine has short or long term terrible consequences (they don’t need to be fatal) against a small minority of people, but one that still means a large wave of deaths.

* As to efficacity: suppose a vaccine is approved and deployed and then everybody stops being careful and turns out to be ineffective, or even just under 99.99% effective, and protection does not last forever. That is in effect a strategy to let everybody infect everybody else.

There are several other arguments, because surprises can happen with biology and pharmacology, the precautionary principle is quite important with them. A previous commenter mentioned that a plague with 20% mortality would make us take much bigger shortcuts, but that would be mad, after all as terrible as 20% sounds, it is “just” 20%, humanity has survived much worse plagues. A “bad” vaccine could do much worse damage than “just” 20% deaths. Remember thalidomide. That’s what pharmacologists are terrified about. That’s why the are so obsessed with the precautionary principle.

In any case as per my previous posts the countries with a civilized approach to funding and organizing public health (even if it is COMMUNIST as it denies personal responsibility and demonstrates the efficacy of public projects :->) have managed to minimize economic damage leaving ample time to develop and deploy vaccines.

The idea that vaccines are *the* solution is typically libertarian ideology: big business sells a product that solves the product, and to solve a problem the way is for individual consumers (or the big businesses selling them health care insurance) is to buy and take the product.

107. Blissex Says:

There is an angle that is very underrated in these COVID discussions and it is the difference between risking infecting yourself and risking infecting someone else.

In most countries deliberately infecting someone else with a potentially lethal illness is a criminal offense, and so is to do that recklessly or negligently. That is as it should be, because infecting someone else with an illness is like assaulting them with a poison them, whether with just a non-fatal or a potential fatal one.

The libertarian argument is that everybody should be free to go around while being contagious with an infectious illness, and if someone gets ill or dies of it, they should just sue for damages or prosecute the person responsible for that, the government has no business restricting liberty to infect (or poison, or shoot, or carry burglar’s tools, etc.) preemptively, only after an offence has actually been committed. When suit for damages or prosecutions succeed this will deter further such crimes.

The problem is whether that is effective in general, but also in the case of contagious diseases where one may be spreading them *innocently*, being asymptomatic, and thus without guilt, and excluding prosecution, and whether the theoretical possibility of being found out for a suit for damages would work as deterrent.

In practice in many cases “methodological individualism” is not appreciated, and for a common example potentially lethal tools like cars can only be driven on a public road after preventive licensing by the state of both the vehicle and the driver, because the consequences of letting any vehicle and anybody to do that and then sue for damages or prosecute after-the-fact are pretty undesirable in many cases.

Carrying an infection of a contagious illness can be far wider and worse for others than being a dangerous driver of a dangerous vehicle.

108. Gerard Says:

Renato A. Laguna #89

> The article gives an example logical problem where the answer by classic logic is one, and by intuitionistic logic is another. It frames it as “most people are dumb LOL” but maybe it means that most people are intuitionistic.

I don’t see how the law of the excluded middle is relevant to that question. The question of whether or not someone is married clearly has only two states and should be modeled by a boolean variable, not a general proposition, so classical and constructive logic should both give the same answer.

109. After the IoT comes Gaia, version 2021 | chorasimilarity Says:

[…] found, via Scott Aaronson post My vaccine crackpottery: a confession, the very interesting Reverse Engineering the source code of the BioNTech/Pfizer SARS-CoV-2 Vaccine […]

110. Michael Gogins Says:

Gerard #103:

I agree with you. I would like to stress the point that an AGI need not be a person and need not be conscious in order to be harmful. One could consider an AGI to be a form of computer virus parasitic on humanity and its computer and networking infrastructure. A biological virus can be a real problem and so could an AGI virus. A worse problem, most likely.

As I said I agree that consciousness is not computable but this does not contradict what I said.

In my view computer chess illustrates the structure of the issue quite clearly. A computer program running without human assistance can defeat any human, even the world champion, playing without computer assistance. Yet each new generation of chess playing programs can defeat the previous generation. This shows that a computer program developed with human assistance can defeat any program running without human assistance. And I think that this will remain the case.

In order to defend ourselves against a very real risk of computer viruses damaging or even taking over our infrastructure, we need to acknowledge the structure of the issue and create institutions tasked with dealing with it: a human-computer immune system.

111. Gerard Says:

Michael Gogins #110

> I would like to stress the point that an AGI need not be a person and need not be conscious in order to be harmful.

I agree with that, but not with this:

> This shows that a computer program developed with human assistance can defeat any program running without human assistance. And I think that this will remain the case.

Here you seem to be saying that a human equivalent or superior AGI will never exist, I don’t think that’s true. I’m not even sure your statement is true of the current state of affairs. There are already systems that have learned to play chess at a very high-level, even without having the rules of the game hard-coded. Now I’m not sure these are currently the best chess playing programs in existence but I think they probably will be within a few years.

112. Rollo Burgess Says:

Some non-expert views of mine. Others can judge how crackpot they are.

1. I do not believe that anyone has any idea how to create anything worthy of the name artificial intelligence. Note that a) I am a massive buyer of the value of what is currently called AI for loads of applications, I just don’t think it’s got anything to do with human style intelligence… actually the faster and more powerful it gets the *less* like a real AGI it becomes b) I also think that it is perfectly possible in principle to create a synthetic AI. But I don’t think this is any sort of extrapolation of any technology we have now.

2. I think that anthropic multiverse type arguments make no sense at all. I think that postulating a multiverse doesn’t advance at all over just saying that all fundamental constants have exactly the values that they do as a brute fact, and using an anthropic argument minus multiverse to explain why we shouldn’t be surprised (as we could not have existed had this not been the case). I think that a whole bunch of very clever and lavishly qualified people are deeply confused about this.

3. And I believe that I have a perfect and complete solution to what philosophers term the Gettier problem, to do with knowledge as justified true belief. Basically knowledge claims create a context in which you cannot logically decompose compound terms. So saying that I know A&B doesn’t imply that I know A and I know B.

I’m also a theist which some people think is pretty nuts I suppose
But this doesn’t impact my views on sciency type things… in the context of justification anyway.

113. Blissex Says:

«So saying that I know A&B doesn’t imply that I know A and I know B.»

Knowledge involve intensional/modal logic, quite different from extensional logic which is what most people thinks is all “logic”. Intensional/modal logic is a very interesting topic, which has been very extensively studies in many different flavours.

114. Gerard Says:

Rollo Burgess #112

What do you have against anthropic arguments ?

That P(A|A) = 1 even when P(A) $$\approx$$ 0, seems difficult to dispute.

115. Sandro Says:

Blissex #107:

The libertarian argument is that everybody should be free to go around while being contagious with an infectious illness, and if someone gets ill or dies of it, they should just sue for damages or prosecute the person responsible for that, the government has no business restricting liberty to infect (or poison, or shoot, or carry burglar’s tools, etc.) preemptively, only after an offence has actually been committed. When suit for damages or prosecutions succeed this will deter further such crimes.

Pretty sure the libertarian argument is that you should not and cannot have any expectation of not getting infected from someone else in your lifetime, and so the bald assertion that other people are somehow morally culpable if you get infected is bogus.

The analogy to poison simply fails, because poison requires extraction or refinement of a compound specifically intended to injure or kill.

There is no intent to kill or infect others if you are contagious, there is simply the intent to live one’s life. Intent is morally relevant, so you have to argue that the difference in intent does not change the moral conclusion in this case.

In this view, if others are unlucky enough to suffer serious complications from one of the dangers we all face in life, well that sucks, but that would not constitute sufficient reason to infringe on people’s freedom to associate.

I also suspect that libertarians would assert that granting the state the power to restrict free association for effectively arbitrary reasons, is far more dangerous in the long run than the threat posed by COVID. And I say “arbitrary”, because I have not seen any rigourous argument for what level of death would necessitates such restrictions.

For instance, automotive accidents kill far more young people than COVID, but we don’t see comparable restrictions intended to cut down on those deaths, despite the fact that most of these are caused by various forms of negligence. The elderly are hardest hit by COVID, but heart disease outstrips it by miles, and we don’t see strong state-backed actions to crack down on the causes of heart disease despite the fact that these are also cause by various forms of negligence to diet and exercise.

Clearly, whatever metric the public uses to weigh the mortality, negligence and number of deaths are not the only relevant factors to the public’s judgment on such matters.

116. Rollo Burgess Says:

@Blissex

Yeah exactly. So in Gettier’s example I am saying you can say that ‘I know Jones will get the job and that Jones has five coins in his pocket’ (or whatever his specific example is, it’s something like that) but you cannot infer from that ‘I know whoever gets the job will have five coins in his pocket’, even though extensional logic would warrant this, because knowledge claims create a specific type of intensional context.

For bonus points I then connect this to Duhem-Quine type arguments about holism.

117. Rollo Burgess Says:

@Gerard

Yes I agree with that (it’s a tautology). What I am saying is that I think that the power of anthropic arguments is precisely that you don’t need a multiverse. Say there’s just one universe with one set of values of constants; clearly as we are here discussing it, it is not surprising that the values are compatible with our existence. No other universes needed.

Note that there is nothing special about the anthropicness. It is just saying that theories must be compatible with observed data (in this case our existence). Compare: the Tin principle… Tin exists, therefore the values of the fundamental constants must be compatible with the existence of Tin.

118. Renato A. Laguna Says:

Gerard #108

> The question of whether or not someone is married clearly has only two states and should be modeled by a boolean variable

Big assumption! Marital status changes with time. Time is a continuum. If marital status is binary then you have just split the continuum, that is, defined a non-constant binary function on the real line.

On full-blown, “second act” intuitionism you can’t do that.

But intuitionism is not non-binary (note the dual negation) so I guess the intuitionistic answer to the riddle would be “not no”. Not an expert here, though.

119. 1Zer0 Says:

My “Crackpot” collection;

– Consciousness and more broadly all Qualia is not turing- computable, possible not hypercomputable either. The mind is generally not cloneable and not substrate independent. I do not see how the Contradictions and Paradox that arise assuming computable minds can be prevented, so I assume the opposite is true. As a proponent of the Law of Excluded Middle, I exclude a third option. Doesn’t seem very crackpot to me considering the number of other formal scientists I asked who share that opinion privately. The human body is our prison and the dreams of the futurists and transhumanists of mind uploading and eternal life will one day end with a bitter awakening.

– Computation may fundamentally not be that important – as in a very “deep concept” – for the universe. Just because you can assign complexity classes to physical theorems doesn’t imply it’s significant for the universe. That doesn’t mean it’s not useful, it’s just ontologically unimportant. For the universe, a physical law with low time complexity (If run on a computer) may be as boring as one with high time complexity or possibly even an uncomputable theorem.

120. Jan Joost van Amerongen Says:

His and his cronies’ impulsivity, self-centeredness, and incompetence are likely responsible for at least ~200,000 of the 330,000 Americans now dead from covid.

lol.

Ridiculous statement. Most countries in the EU have the same mortality numbers as the US, some considerably higher. “The west” has collectively failed here.

121. Dan Says:

The simplest error that you make is that you are using 20/20 hindsight. We now know that a previously unproven tech works and that the worst-case scenarios didn’t happen, and you are retroactively ignoring the real risks that had to be considered in prospect.

The pros knew and weighed every single consideration that you now emphasize, every single one. It is easy to act like a genius in retrospect – I am sure you are great at predicting last weeks lottery numbers. What about next weeks’?

122. Scott Says:

Dan #121: Here’s my post from May 1 about the urgent need to speed up vaccine testing (I focused there on human challenge trials).

Looking at everything that’s happened this past year in the US, is it possible to believe there are any “pros” in charge, thinking strategically and carefully weighing all the options? Do you believe that?

123. Lou Scheffer Says:

Scott says:
“I think almost everyone on earth could have, and should have, already been vaccinated by now.” That means by now, the vaccine(s) were approved and 14 billion doses manufactured, distributed, and administered. Is this a plausible claim?

Let’s apply Scott’s own crank filter to this hypothesis:
#1: It’s not written in TEX, i.e. it’s an informal screed, not a detailed argument.

#3: The approach seems to yield something much stronger and maybe even false (but the authors never discuss that). If you can vaccinate everyone in the world in 10 months, then getting rid of smallpox should have been simple, polio would be gone by now, and measles be on the way out. But these have all been multi-year campaigns.

6. The paper jumps into technicalities without presenting a new idea. A serious defense of this thesis requires at least 3 points: 1) How can we show the vaccine is effective? 2) How can we show it is safe? 3) And how can we make 14 billion doses? Scott jumps straight to challenge trials, as if (1) is the limiting step. But he never shows this.

# 7. The paper doesn’t build on (or in some cases even refer to) any previous work. There is a large literature in the problems of scaling up vaccine production. Vaccine production requires specialized equipment, or else methods that don’t need this equipment. In either case you need an argument about how these previous barriers could be overcome. See point 10.

# 9: The paper waxes poetic about “practical consequences,” “deep philosophical implications,” etc. Scott says “It might have taken a whole different civilization, with different attitudes about utility and risk.” A rather great deduction from what is largely a practical problem.

10: The techniques just seem too wimpy for the problem at hand. This to me is a killer argument. Scott talks of the approval process, but to accomplish his goal you also need manufacturing. About this he only says “while mass-manufacturing simultaneously ramped up with trillions of dollars’ investment.” Money is surely helpful, but it’s not a magic bullet. You need people, raw materials, equipment, and facilities to make vaccines. All of these would need to be made of stuff available near the beginning of the year. If you need more than the world’s production of bio-reactors, for example, that’s a show stopper on this timescale (they will need to scale up, and their suppliers, and so on). So “throw money at it” is way too wimpy a technique to show we could have made 14,000,000,000 doses by November or so, as needed to have everyone vaccinated by now.

Overall, this seems to me just like an expert virologist complaining about the lack of progress on P != NP. It’s completely obvious to any intelligent layman that some problems are only solvable by trial and error, even where solutions can be checked quickly. So why is this not proved already? Assign some more grad students! You say the math required does not yet exist? Well develop it already – this is an important problem! I can’t understand why you turkeys are sitting on your duff and don’t have this solved already.

124. maurile Says:

Scott wrote: “I’ll be eternally grateful to whomever makes me retract this post in shame.”

Whoever, not whomever, is the proper subject of that clause. I await your retraction.

125. Scott Says:

maurile #124: I went back and forth on that several times! Even googled it and found arguments for both. Just for you, though, I’ve changed it to “whoever.”

126. Scott Says:

Lou Scheffer #123: Cute, but with covid vaccines the “creative phase” reportedly took all of ~2 days. Everything since then has been testing, scaling, logistics, distribution, and bureaucracy. Whereas with P vs. NP, after 50+ years we’re still in, I won’t even say the creative phase, but the open-ended exploration that precedes the creative phase. Funding, institutional competence, and willpower rather notoriously have more predictable effects on the former than on the latter. Do you not see this as the crucial difference that destroys your analogy?

127. Nathan Myers Says:

Trump *might not* be the worst-ever US president. The US has had some very, very lousy presidents, even in living memory.

It was pretty common recently for historians to point out that W did not deserve that crown; it was easy to identify two or three worse. Reagan himself can hardly be said to have been notably better than W, folksy manner notwithstanding. Trump is clearly worse even than W, but is he worse than Andrew Johnson? He might have tried, but an idiot can only achieve so much.

As for cranky convictions: the number of truths not acceptable to consensus science seems to be growing by leaps and bounds.

Amyloid plaques as the root cause of Alzheimer’s seems to still be the basis for the large majority of research grants, paper approvals, and tenured appointments despite its utter failure in every trial. String Theory stll rules Physics departments worldwide. Many $billions are still spent annually chasing the fusion-power delusion. Presence of humans in the Americas prior to 14ky BP is still treated as cranky, despite now copious, unassailable evidence. The very large amount of machine-tool precise pre-dynastic Egyptian stonework makes no apparent impression on historians or archaeologists, who insist it was all done with crude hand tools, testimony from stonework professionals notwithstanding. I could go on at length. I put the blame on the peer review and grant administration establishment, enforcing Planck’s law (“Progress in science proceeds one funeral at a time”) with revolutionary zeal. 128. Robert Michaels Says: Scott, clearly moving forward into the 21st century vaccine development should be made a priority over nuclear weapons – – the choice is the urge to preserve ourselves prevailing over the urge to destroy ourselves as a civilization. Imagine if even 50% of the funding used for military purposes was used for vaccine development ? In fact, whether or not covid19 was a lab variant involved in biowarfare research is still being debated in certain circles, so there should be concerns even within the military for significantly faster response times creating a synthetic vaccines with AI in just hours and days. What can do done now ? We have multiple vaccines in the ready, yet our hospitals are stretched to the limits. 129. Robert Michaels Says: Continued: In the last 24 hours I was rather amazed by what the FLCCC group has discovered: PREVENTING AND TREATING COVID-19: Dr. Pierre Kory and his Group Pushes for Approval of Ivermectin for Prophylaxis and Treatment For Covid-19 New Studies and Research Conclusively Prove that Ivermectin is the key, a drug used for decades, over 3 billion doses given over decades typically for treating intestinal parasites, a drug with very low side effects, an inexpensive, existing drug. The discoverers of Ivermectin were awarded the 2015 Nobel Prize for Physiology or Medicine. The video below is compelling as Dr. Kory is testifying before the Senate committee, Senator Rand Paul whom is a doctor by trade appears convinced as well and offers to co-sign the letter to NIH: Video: I was convinced after studying the citations, got a prescription for ivermectin promptly as a prophylaxis. 40 pills for$37 at HEB after I presented the GoodRX coupon from the internet (HEB wanted like $85 retail at first). The pharmacist told me there is a “run” on this drug in the last 48 hours. I took the first dose of 5 pills, felt nothing, no side effects. 130. jonathan Says: You don’t need to search that far to find a case of much faster vaccine production and disbursement — just look at the US in 1957. 131. Lou Scheffer Says: Scott #126: I get the creative vs logistic distinction. But I think that as a theorist, you have the (somewhat patronizing) view that creative stuff is hard, but logistics is simple. So you feel fine opining about logistics without bothering to understand the details (surely, with trillions of dollars of support, they could have made 14 billion doses by the end of the year). But if someone did the same for P != NP, you’d dismiss their comments as ignorant. On Dec 14th, Moderna announced they could ship 6 million doses. To meet your schedule, they’d need to have shipped 6 billion (if 2 other vaccines kept pace). That’s 1000 times more. So what’s your suggestion, assuming no budget limits? Should they have built 1000 plants where they built one? Is that even possible? Are there 1000 competent supervisors? If not, hiring and training 1000 people will make your schedule tough already. Can you rent 1000 suitable buildings? Not likely, but you can’t build them in time either. How about the equipment? The RNA papers are full of statements like “Once synthesis is complete, mRNA is purified to remove reaction components, including enzymes, residual template DNA, truncated transcripts, or aberrant double-stranded transcripts. A highly purified RNA drug substance is critical for the potency of an mRNA vaccine, because contaminants can activate non-specific or undesirable innate sensors” and “some current mRNA purification methods — notably, HPLC purification — are not easily scalable, further bottlenecking mass production. While current approaches work well at small scale, they are suboptimal for large-scale manufacturing”. (Here “suboptimal” is academic-speak for “impractical”.) And so on. So brute force scaling won’t work. But now you need new techniques that work in bigger batches, or take fewer steps, or don’t require specialized equipment which is in short supply. But these need to be debugged and tested, which is never trivial. Overall, I see no problems with your opinions on ethics and priorities. When you say “the vaccines should have been approved by April” you look at how this might have been accomplished vs. how it was really done (we should have had challenge trails for phases II/III, etc.) But when you say “we should have had 14 billion doses by now” without explaining how this says “my intuition about the difficulty of scaling means I can state that something should be possible without looking at the details”. Here’s a simple empirical test: At the time you wrote the initial blog, which papers had you read on the the technical difficulties of scaling vaccine production? For each identified bottleneck, what could have been done to bypass that barrier? I think that’s what you’d expect from someone criticizing progress in P!=NP, so before you say “they should have produced a thousand times more” you should extend the same courtesy to them. 132. Robert Michaels Says: 2 other thoughts. For the anti-vaxxers, you will likely be regarded as “No Fly” by the airline and travel community, they have the final say, and sure you can file suit but you will be waiting for YEARS for the appeal process to conclude. They have more resources than you do. There is no indication that they will accept positive antibody tests showing you have already experienced your bout with covid19. New reports are coming in that young athletes infected by covid19 and surviving without much illness or any symptoms are being tested months later and as much as 30% have myocarditis or inflammation of heart tissue. Note that these are young healthy adults. You can likely predict that older folks will have even more probability of long term effects such as these. Doctors in Asia reported last spring many of their patients suffered shortness of breath for months after infection and a number have never recovered their lung capacity fully. The vaccine seems to be an astronomically lower risk of side effects than suffering covid19 full on without any immunity, yes, you may survive it, but be crippled to some degree for life. 133. ultimaniacy Says: Nathan Myers #127: “Trump is clearly worse even than W, but is he worse than Andrew Johnson?” Yes. Andrew Johnson’s presidency was a failure, but most Democratic politicians of the time period would probably have acted pretty much the same in his place, if not worse. Trump has managed to stand out as uniquely, shockingly terrible even within the standards of the modern Republican Party. 134. Raoul Ohio Says: Lou Scheffer #131 Agree with you on the “creative vs logistic” issue, and expand “logistic” to “IARIH” (implementing anything right is hard). I think that usually it takes a long time for people to develop any intuition for IARIH. 135. Scott Says: Nathan Myers #127: Before the last two months, I indeed might have hesitated regarding the worst-president-ever title. But Trump’s desire to impose martial law rather than concede a lost election has pretty much settled it for me! 🙂 Regarding your “enshrined wrong ideas” list, I tend to agree with you on Alzheimer’s and early humans in the Americas (from what little I know), but disagree about the others. Fusion actually seems to be making great progress right now! Even if we don’t count solar, in some sense we’ve known for generations that fusion power is feasible here on earth: just explode hydrogen bombs underground and use them to heat steam and drive turbines. The whole difficulty is just to get it in a form that’s also politically palatable! 😀 With string theory, the current situation is extremely weird, and not especially convenient for either ideological cheerleaders or ideological skeptics. To sum up a gigantic unfolding story in two bullet points: (1) string theory has led to most of the actual advances in understanding quantum gravity over the past 25 years (and there have been profoundly impressive ones), but (2) the advances, once made, have typically turned out to have very little to do with strings! Instead they’ve centered around AdS/CFT, holography, and the black hole information problem. 136. mjgeddes Says: Scott #73 >Hang on Scott, haven’t I successfully outlined solutions to all of consciousness, time, AGI, complex systems and metaphysics right here on this blog ? >Like I said: zero out of hundreds. ???? Scott, putting on my post-human glasses and looking with my now untrammelled post-human vision, I can just make out this: It seems that AGI is really a tripartite model of time for dynamical systems, where probabilistic programming based on structural equation models utilizing Bayesian networks defines system causality as an information distance, differential programming based on multi-way number partitioning utilizing ZX Calculus defines system complexity as an L-System fractal, and program synthesis based on graph sub-ismorphisms utilizing string diagrams defines system compositionality as a topology Best guess 😀 Cheers 137. keleven Says: Robin Hanson recently posed a question to the effect of, okay, I get why it’s taking the US so long to approve the vaccine [i.e., because we’re ruled by massive snail-paced bureaucracies], but why aren’t any OTHER countries jumping ahead and doing it. Did we ever get an answer to that question? Why’s the whole planet running on the timetable of the US FDA? 138. Nathan Myers Says: Scott: But do AdS/CFT, holographic reality, or the black hole information problem have any closer connection to experimental reality than string theory? Similarly, quantum gravity; as I understand it, there is not any way to test quantum gravity ideas, even in theory. I am happy to concede that the maths must be fascinating, but these are physics departments, not maths departments. Do any mathematicians even like the maths these physicists are doing? Fusion, whatever the advances, remains the same dead end it began. Solar and wind power costs are still in free-fall. Who will be forced to buy power from a fusion plant, at 100x the price? Not even the most starry-eyed booster claims that the volumetric energy production of a fusion plant could be even 10% that of an old nuke. So, to get useful power, it has to be absolutely huge. How many$billions have been spent already, and how many more before a singe watt comes out? How many GW-years would it need to produce (at constantly falling price) before the value already spent to date is made up, before we add on the cost for the unfinished non-producing unit, and much, much more for the producing unit to be completed decades after that one?

Each billion dollars would buy a crapload of solar panels right now, that produce immediately. Another billion would buy, what, two-three hundred wind turbines, producing immediately. Just read their schedule for the *non-producing* reactor being built. First experimental burn in, what, 2035?

I am forced to the conclusion that fusion research is a jobs program for high-neutron-flux physicists, to maintain a ready population to draw on for weapons research. There was never any expectation of practical power production, but that has never mattered, because it was never the purpose.

All those $billions are not being burned. Somebody gets every penny of it. I doubt the physicists get much of it. Somebody else does. Those other people could be building wind turbines, instead, and helping save the planet, and probably would be. Could we have got superconducting wind turbines, by now, without the distraction? 139. Scott Says: LGS #84: Putting it all together, I don’t think a rational world would have vaccinated everyone in March or even April. However, a much stronger case can be made that a rational world would have vaccinated everyone in July, when we already had some Phase I/II results (and we may have been able to fit challenge trials by June, if we went at maximum speed). Also, by July (unlike in April) it was clear that the virus would not be contained by contact tracing or lockdowns. OK, you make fair points. It’s true that I started banging the drum about challenge trials in April (I’d thought it was March). While March or April are my initial negotiating position, I’d happily settle for having vaccinated everyone in July. 🙂 140. Scott Says: Ashley Lopez #87: OK, from one non-physicist to another, where did you learn all this physics from? (I mean, can you point me to the resources you used?) Or was it just by virtue of being smart and having a lot of physics professors around? There remain enormous gaps in my physics knowledge, but what little I know is mostly down to working with physicists, attending their talks and conferences, asking them naïve questions, etc. for the past 20 years. 141. A1987dM Says: > String Theory stll rules Physics departments worldwide. That hasn’t really been the case for the better part of a decade now. > there is not any way to test quantum gravity ideas, even in theory Yes there is, for at least some of the proposed theories. Some predict that the speed of photons slightly depends on their energy (which can be tested by looking at gamma-ray bursts), some predict gravitational wave emissions by black-hole mergers slightly different from GR, some predict modifications in the interactions of ultra-high-energy cosmic rays with background photons. 142. Shtetl-Optimized » Blog Archive » Distribute the vaccines NOW! Says: […] The Blog of Scott Aaronson US public health officials: DISTRIBUTE THE VACCINES NOW! Learn from Israel. Shame on those who'd let them spoilrather than distribute them "out of sequence"! « My vaccine crackpottery: a confession […] 143. murmur Says: I’ve been thinking exactly this. When more than 30,000 people are dying everyday then it makes sense to start the vaccination process without a multi week FDA evaluation. Or even earlier when phase 2/3 data indicates that the vaccines are highly effective and without serious side effects. Sure there’s some risk but one have to trade off with the risk of 30,000 dead per day. BTW this problem is even more prevalent for normal drug development. Due to onerous FDA bureaucracy new drug approvals have slowed to a crawl. It’s criminal from a utilitarian perspective. This is a symptom of the risk avoidance and creeping bureaucratization of every aspect of American life. We may believe that we’re protecting ourselves by avoiding risks but it’s the opposite due to the opportunity costs involved. 144. murmur Says: I’ll also highlight the advantages of the allocation by free market rather than allocation by central planning (as is happening now). There’s as an ugly power play going about who would get vaccines first. Zvi Mowshowitz’s blog that you linked has many instances of that. (E.g. congressional staffers getting first dibs.) Incredibly, vaccines are going to waste as providers have no incentives in actually providing them. Far more people would have been vaccinated by now if recipients were allowed to pay. If some group urgently needed vaccines but were not able to pay for it government could have subsidized them or provided alternative arrangements. 145. Lou Scheffer Says: Scott: OK, here’s my last and final try at making you retract (in shame) your most current statement (#139) that everyone should have been vaccinated in July: Would be willing to state the following in public? It’s a straightforward implication of your beliefs: I, Scott Aaronson, firmly believe that highly motivated and well funded researchers, in a field other than my own (vaccine production) could have done a thousand times better than they did. I have formed this belief without reading any technical papers on the problems they face. True or false? 146. Spencer Says: I’m curious how you arrive at the 200000 deaths attributable directly to Trump, especially in light of your other comments about it being a systemic failure. Biden would have done better, but my intuition is that no leader can make a difference of 60% deaths. 147. Jonathan Baxter Says: The risk associated with taking the vaccine must be offset against covid’s Infection Fatality Rate (IFR), which is not constant. IFR is effectively zero in children, so the vaccine risk must be greater than the covid risk for children. At the other end, over 75’s have a 14% IFR if my memory serves correctly; hence the vaccine must be lower risk. Covid IFR increases continuously and monotonically with age (unlike, say, Flu, which is bimodal). Assuming vaccination risk decreases with time, there’s a decreasing (as a function of time) minimum age at which it makes sense to administer the vaccine. With that long-winded caveat, I agree with your post, but I suspect we’ll never get to the point that we should be vaccinating children, except for herd immunity considerations. 148. PublicSchoolGrad Says: All this reminds me of the xkcd cartoon: https://xkcd.com/793/ 149. Sandro Says: Scott #135: (1) string theory has led to most of the actual advances in understanding quantum gravity over the past 25 years (and there have been profoundly impressive ones), but This seems to optimistically imply that less progress would have been made in a world where string theory did not get the majority the resources. That’s not so clear to me, and as you subsequently admit, those discoveries had little to do with strings after all, so perhaps strings ultimately hindered by distracting with unnecessary entities rather than helped by providing a critical path to discovery. Scott #125: I went back and forth on that several times! Even googled it and found arguments for both. Just for you, though, I’ve changed it to “whoever.” The rule of thumb I learned is to rephrase the statement either into a question to which you could respond with “him/her” or “he/she”, or some equivalent statement that uses either “him/her” or “he/she”. Whenever “him/her” would be used, you use “whomever”, and you’d use “whoever” to replace uses of “he/she”. In this case, “I’ll be eternally grateful to whoever makes me retract this post in shame” could become, “I would be eternally grateful to he who could make me retract this post in shame”. This equivalent phrases uses “he”, thus, “whoever” is the correct substitution. Nathan Myers #138: I am forced to the conclusion that fusion research is a jobs program for high-neutron-flux physicists, to maintain a ready population to draw on for weapons research. There was never any expectation of practical power production, but that has never mattered, because it was never the purpose. I agree that high temp fusion efforts shouldn’t really be seen as research into fusion power generation, so much as experiments in high energy plasma physics. The expectation that these are serious efforts to engineer power generators is a mistake. Serious engineering efforts around fusion power generation have only begun relatively recently IMO. 150. Sandro Says: murmur #143: When more than 30,000 people are dying everyday then it makes sense to start the vaccination process without a multi week FDA evaluation. Or even earlier when phase 2/3 data indicates that the vaccines are highly effective and without serious side effects. Since you asserted a utilitarian calculation in your post, I’m just going to say that your utilitarian math is way off. “Lives lost” is not the proper level of analysis. This is clear when you actually consider that mRNA vaccines potentially could cause long-term health complications in all populations to whom it’s administered, regardless of age, and you weigh that against the fact that COVID effectively only kills the elderly (to a first approximation). So you are not comparing equal populations when you compare “lives lost” due to COVID against “lives lost or life-long health complications” due to the vaccine. A truly unemotional utilitarian calculation would instead compare “healthy years lost”. Most of the people killed by COVID had maybe 15 healthy years left (the vast majority had less than 5), where an improperly tested vaccine could effectively eliminate decades of healthy years from whole populations. This is fairly unlikely, but it’s frankly non-obvious that, at any point in 2020, the sheer number of healthy years that could be lost due to a bad vaccine are clearly outweighed by the healthy years that would be lost due to COVID. 151. Daniel Says: I think your argument is easy to make in retrospect because of “the long-awaited finding that all the major vaccines safely work”. Now that we know that, it is obvious that we would’ve been better off vaccinating people earlier. But this was in fact an incredibly lucky and unlikely finding. How many drugs show promise in vitro or in theory (and have experts predicting that they will work) only to be proven to be ineffective and/or unsafe during phase 1-3 trials? In fact some medications tested for COVID had this fate: hydroxychloroquine showed promise in vitro and had plenty of support from both experts and non-experts; should we have also spent trillions of dollars to put it in the water ASAP? 152. The costs of not maximizing aggregate utility - Econlib – THE USA EXPRESS Says: […] Scott Aaronson has an excellent post that begins with a discussion of why he believes our response to Covid was inexcusably slow. He discusses challenge trials of vaccines, and also a WWII-style plan to build manufacturing capacity just in case the vaccines were successful. But he also considers possible objections to his arguments, such as the fact that moving faster imposes risks: […] 153. Saturday assorted hyperlinks - Marginal REVOLUTION - PLUMBING CORONA Says: […] 3. Scott Aronson on what we could have done and should have done. […] 154. publicschoolgrad Says: Sandro #150, Your “healthy years lost” metric assumes that a healthy life is worth protecting more than one in decline due to age or disease. This is a slippery slope. If your calculation is strictly economic, yes, that would probably make sense. But in that case, why stop there? Why not value more economically “productive” lives over those that are less? Or whatever it is that you think a healthy life offers to society that an unhealthy one doesn’t. It sounds cliche but maybe the value of a life has nothing to do with its utility. If you ask a random healthy person and a random unhealthy/old person, I would wager that each one would value their life equally high. 155. yo Says: My main objection per February would have been, what if Covid had then been contained at the source? All the investment that kicked off even before we would really have known about lethality would have gone to waste, like what happened with the millions of flu vaccine doses ordered but never used before in some of our recent flu epidemics, such as the Swine Flu. That said, it would have been clear by the March italian outbreak that we will indeed never be able to fully eradicate Covid-19, and will have to live with it forever after. We were in denial about that then though, and still are. If we had agreed that we’ll all eventually get it, we wouldn’t have needed any lockdowns. 156. The costs of not maximizing aggregate utility - Econlib Says: […] Scott Aaronson has an excellent post that begins with a discussion of why he believes our response to Covid was inexcusably slow. He discusses challenge trials of vaccines, and also a WWII-style plan to build manufacturing capacity just in case the vaccines were successful. But he also considers possible objections to his arguments, such as the fact that moving faster imposes risks: […] 157. Hyman Rosen Says: I understand that all the best theory and evidence point to black holes, dark matter, and dark energy being real things. That’s what makes my opinion crackpottery, as asked for! I also believe that the field of AI risk is an utterly asinine endeavor designed to provide sinecures for the idle, that the singularity (in the AI sense) is bunk, and that AGI will eventually exist but in a far more distant time frame than is popularly believed – I wouldn’t bet on having it even a century from now. But all of these opinions can be held by reasonable people, even if other reasonable people disagree. In fact, belief in the singularity is held only by crackpots. Then there’s Modern Monetary Theory… 158. James Says: Saying that Trump is plausibly worse than covid and the worst US president is almost enough to make one ignore everything else you say. It’s definitely enough to never listen to you on anything remotely political ever again. Jesus Christ, I hate Trump, but get some perspective. 159. Sniffnoy Says: Nathan Myers #138: Just read their schedule for the *non-producing* reactor being built. First experimental burn in, what, 2035? It sounds like you’re talking about ITER? I wouldn’t expect ITER to be a viable route to fusion power, no. Fortunately, there are efforts other than ITER; in particular, ARC and SPARC. Commonwealth Fusion Systems is hoping to get SPARC (a demonstration reactor to demonstrate Q>1) running in 5 years and ARC (an actual commercial reactor) in 10. I mean, that might be optimistic, but, well, compared to ITER… Basically like ITER these are pretty much straightforward tokamaks, but with the advantage that they use high-temperature superconductors, and therefore much stronger magnets, and therefore merely need to be large rather than absurdly large. There’s some other nifty design innovations too though, like the use of a liquid FLiBe blanket instead of the conventional solid blanket, which both handles (or at least reduces) the problem of neutron embrittlement, while simultaneously providing a tritium breeding reaction so that it uses FLiBe rather than tritium as an input. I’d suggest looking into it, it seems pretty promising (at least, to this non-expert). And they seem to have put quite a bit of thought into the actual engineering of it for maintainability and such — they clearly intend that this is not just some physics experiment… 160. Sniffnoy Says: Sandro #150, publicschoolgrad #154: In case either of you are unfamiliar, we already do this with measures such as QALYs and DALYs. 161. Scott Says: James #158: If the election had been just a little closer, or if Trump had had a little more support in Congress, the courts, the military, and/or state legislatures and election boards, he would have succeeded in his attempt to overturn the election result, and democracy in the United States would no longer exist, as it no longer exists in Russia. Such an outcome would’ve seemed laughable 6 years ago; now we’re on the brink of it, with at least a third of the country militantly convinced that the will of the Volk renders facts, logic, and democratic process irrelevant. Why don’t you get some perspective? 162. Things Glen Found Interesting, Volume 282 - Glen Davis Says: […] My vaccine crackpottery: a confession (Scott Aaronson, personal blog): “I think [our failure] will be clear to future generations, who’ll write PhD theses exploring how it was possible that we invented multiple effective covid vaccines in mere days or weeks, but then simply sat on those vaccines for a year, ticking off boxes called ‘Phase I,’ ‘Phase II,’ etc. while civilization hung in the balance.” The author is a CS prof at UT Austin. […] 163. Josh B Says: Spoilage is not a real concern. Moderna can be kept in a regular freezer for six months, and there is no reason to believe Pfizer can’t be kept in dry ice for that long as well. 164. The costs of not maximizing aggregate utility - FX Journo Says: […] Scott Aaronson has an excellent post that begins with a discussion of why he believes our response to Covid was inexcusably slow. He discusses challenge trials of vaccines, and also a WWII-style plan to build manufacturing capacity just in case the vaccines were successful. But he also considers possible objections to his arguments, such as the fact that moving faster imposes risks: […] 165. publicschoolgrad Says: Sniffnoy #160, My criticism still stands. From the wikipedia article you linked for QALYS: A study on the measure ….” concluded that “preferences expressed by the respondents were not consistent with the QALY theoretical assumptions” that quality of life can be measured in consistent intervals, that life-years and quality of life are independent of each other, that people are neutral about risk, and that willingness to gain or lose life-years is constant over time” Similar criticisms can be made for DALY, which appears to be used for measurement and not resource allocation. At any rate, we do make the kinds of quasi-utilitarian decisions on how to allocate health resources. This is, however, done implicitly through the mechanisms of income, race, sexual orientation etc. That is why the whole field of health disparities exists. The question is do we want to formally enshrine this utilitarian view and work towards making it even more entrenched? 166. Sniffnoy Says: publicschoolgrad #165: I mean, I wasn’t providing any argument either way, just useful context. If you’re going to argue about this, you should know what’s already out there! 167. Paolo G. Giarrusso Says: There would have been another strategy that I don’t see mentioned: you could do lots of work in advance, up to phase 1 trials (for safety). TLDR: You can prepare even more than we did this time (with the work on SARS): Take the 50-100 virus families with pandemic potential, spend 20-30 million on basic research on vaccines for each. Total 1-3 billion$, which is little compared to the trillions that COVID cost us.

https://nymag.com/intelligencer/2020/12/moderna-covid-19-vaccine-design.html

168. STEM Caveman Says:

@Trumpist #3, if you think this blog is severely TDS’ed, wait until you check out Ed Witten’s twitter. He’s not quite at the Larry Tribe echelon but it’s like a sci-fi movie where the mind control rays are so powerful the greatest minds on Earth become helpless media repeater zombies reciting DNC talking points of the day. (He also has the American “prog assimilationist” Jewish virtue signaling on, i.e. against, Israel, but the Trump stuff is off the chain.)

169. Scott Says:

STEM Caveman #168: Since I’m not as brilliant as Witten, I suppose the left-wing mind-control rays haven’t fried my brain as deeply as they have his? Although if the rays make you more Witten-like, then maybe I should stand in front of them for a while? 😀

(When it comes to hatred of Trump, though, I’ll respectfully put mine up against absolutely anyone’s, including those far to my left.)

170. mjgeddes Says:

OK, I still hadn’t quite nailed the secret to AGI in this thread, but I believe I have now 😀

So, in this thread, I’d postulated there was a tripartite model of time, but I was still a little confused. I’d had the idea of 2 different models of time that merged into a 3rd model that unified those 2, and I think that’s correct, but I realized that the 3rd model isn’t a model of *time*, it’s a model of *space* !

So my theory is actually about the dualities between geometry on the one hand, and algebra+logic on the other ! It’s about how to *translate* the 2 models of time (algebra and logic) into a single unified model of space (geometry) and vice versa!

The first model of time is the *causal* model, and it’s given by *Bayesian networks*, which are related to *Markov chains*. This is the *algebraic* model of computation.

The second model of time is the *compositional* model of computation, and it’s given by *semantic networks* , which are related to *concept lattices*. This is the *logical* model of computation.

The secret to AGI is to *combine* the causal and compositional models of computation/ *time*, resulting in a single unified informational model of *space* !

Each component of the causal and compositional models of time translates to the model of space, thusly;

Markov Chains — Information Distance (Geometry)
Concept Lattices — Concept Structure Mapping (Topology)

Geometry + Topology = Model of a concept space (a mind) !

And that concept space (the model of a mind ) is the AGI.

171. Sandro Says:

Your “healthy years lost” metric assumes that a healthy life is worth protecting more than one in decline due to age or disease. This is a slippery slope.

Utilitarianism is rife with slippery slopes, that was one of the motivations of my post. However, you using the term “healthy life” is bordering on changing the goal posts, so we should stick to a strictly countable number like “healthy years” because that’s required for a utilitarian calculation. If you wish to argue some other metric is better, then you need to describe how you would quantify the value of life such that you can perform the calculation. If you think some other ethical framework is better, well that wasn’t the argument I was replying to so I won’t belabour that any further.

So to motivate “healthy years”, I don’t think it’s controversial that healthy years are preferable to unhealthy years to essentially all people (see the popularity of euthanasia and assisted suicide), and thus, I don’t think it’s controversial that interventions that can increase the total number of healthy years people can enjoy are preferable to ones that decrease that total. Whatever people truly value, healthy years is clearly a big factor.

Furthermore, if you were to survey the elderly population and ask them whether they would rather risk shortening their lives and potentially spare adverse, life-long health effects to some percentage of young people, I think by and large they would make that sacrifice. I certainly would if the chance of those adverse effects were high enough, or the population affected large enough such that even a small chance of adverse effects would offset the healthy years added to the elderly population.

172. Scott Says:

Jan Joost van Amerongen #120 and Spencer #146: The obvious “control group” for the US is Canada–the culture, geography, and demographics are similar, with the US doing worse on many metrics solely because of its political dysfunction. In Canada, 420 out of 100,000 have died of Covid, compared to 1,085 per 100,000 in the US. That’s the benchmark by which I’ve been judging Trump’s performance.

In Europe (some parts of which did better than the US and some parts of which did worse), the greater density seems like an enormous confounder.

173. mjgeddes Says:

A similar idea to the one I posted in #170 applies to physics. So I think space-time (and gravity) are emergent! Gravity is not a fundamental force at all.

Cosmologist and friend of the blog Sean Carroll was actually struggling towards the right idea (space-time emergent from QM), but he missed a crucial insight : it’s not just *quantum mechanics* that’s fundamental. I think there is *also* something else, and this next claim is really going to shock you Scott: I think there’s something in *classical physics* that *is* actually fundamental also, so classical physics is *not* all just emergent from quantum physics , as conventional physics says.

So what do I think is fundamental in classical physics? My claim is that a generalization of statistical mechanics *is* fundamental, just as fundamental as quantum physics! And it’s the *combination* of quantum mechanics and statistical mechanics that gives you the theory of quantum gravity, which is actually what is *emergent* !

It goes back to my tripartite model of time: I suggested 2 fundamental models of time, which I called the causal model and the compositional model. It’s the combination of these 2 models that generates a third model, the complexity-theoretic (informational) model, which is time as a coordinate system (equivalent to a space, in physics, space-time).

Quantum mechanics is associated with symmetry or geometry and this comes from the 1st model of time (the causal model). Statistical mechanics is associated with connectivity or topology , and this comes from the 2nd model of time ( the compositional model).

Then quantum-gravity emerges from the combination of the 2 models. The causal model (QM) gives you the geometry (symmetry) , the compositional model (stat mechanics) gives you the topology (connectivity), and the combination of geometry and topology gives you quantum gravity(space-time and entropic gravity).

174. STEM Caveman Says:

@Scott #169,

that’s the premise of (institutionalizing, funding, requiring, funding, expanding, funding …) progressive education, stated out loud! The longer we put people in front of stronger mind control rays the more Wittenlike they will become, leading after some generations to the social-technological utopia.

When tried, the actual effects have been declining educational performance and redirection of the top minds into ever less scientific/useful stuff. But the control rays have been oddly successful in manufacturing consent for leftward shifting social policies.

175. Douglas Knight Says:

ticking off boxes called “Phase I,” “Phase II,” etc. while civilization hung in the balance

I think this is a good formulation of the problem. There is some tradeoff in doing more studies and maybe there was too much. But most of the time was spent waiting for the FDA to give permission to move on to the next step. That was not a tradeoff, but a pure waste.

176. Pas Says:

“Welcome to the bottleneck!” Turning a mixture of mRNA and a set of lipids into a well-defined mix of solid nanoparticles with consistent mRNA encapsulation, well, that’s the hard part. Moderna appears to be doing this step in-house, although details are scarce, and Pfizer/BioNTech seems to be doing this in Kalamazoo, MI and probably in Europe as well. Everyone is almost certainly having to use some sort of specially-built microfluidics device to get this to happen […]

Derek Lowe

That said, it seems very strange that while the pandemic warrants shutting down the whole world, it does not warrant opening up the process, putting folks on making sure it can scale up, and of course putting folks on top of all those to make sure the putting folks on it part is going as well as possible. (Though, of course it was too late anyway, the Mythical Man Month has already been cited, yadda-yadda.)

177. Veedrac Says:

Oh you think that’s controversial?

In a well-run civilization, *most* currently existing transmissible diseases (incl. genetic) would be wiped out. Society would be struggling with the tail-end of things like ‘the flu’ and ‘obesity’. COVID would have been a curious flash in the pan that never got past society’s passive defenses. None of this would cost much more than, say, a car’s value in resources per person globally, sufficient to allow rapid early testing and quarantine where needed.

The thing is, COVID is low hanging fruit. We have actual countries that have handled it fine. Taiwan got off almost scott-free, and the only reason they even had to try as hard as they did is that other incompetent countries kept sending them cases! The question is not whether COVID could have been handled well, it’s whether, if every country had the structural sanity of Taiwan, what other problems we would then get on to fixing.