## What we learned from Fast Grants See [[=Tyler Cowen#What we learned from Fast Grants]]. ## Noahpinion interview https://noahpinion.substack.com/p/interview-patrick-collison-co-founder N.S.: So, what are the three things that excite you most about the 2020s? It's hard to restrict to three! But here are the first that jump to mind: First, the explosive expansion in access to opportunity facilitated by the internet. Sounds prosaic but I think still underestimated. Several billion people recently immigrated to the world's most vibrant city and the system hasn't yet equilibrated. When you think about how YouTube is accelerating the dissemination of tacit knowledge, or the number of creative outsiders who can now deploy their talents productively, or the number of brilliant 18 year-olds who can now start companies from their bedrooms, or all the instances of improbable [scenius](https://kk.org/thetechnium/scenius-or-comm/) that are springing up... in the landscape of the global commons, the internet is [nitrogen fertilizer](https://ourworldindata.org/how-many-people-does-synthetic-fertilizer-feed), and we still have a long way to go -- economically, culturally, scientifically, technologically, socially, and everything in between. I challenge anyone to watch [this video](https://www.youtube.com/watch?v=MDG9opjktm8) and not feel optimistic. Second, progress in biology. I think the 2020s are when we'll finally start to understand what's going on with RNA and neurons. Basically, the prevailing idea has been that connections between neurons are how cognition works. (And that’s what neural networks and deep learning are modeled after.) But it looks increasingly likely that stuff that happens inside the neurons -- and inside the connections -- is an important part of the story. One suggestion is that RNA is actually part of how neurons think and not just an incidental intermediate thing between the genome and proteins. Elsewhere, we're starting to spend more time investigating how the microbiome and the immune system interact with things like cancer and neurodegenerative conditions, and I'm optimistic about how that might yield significantly improved treatments. With Alzheimer's, say, we were stuck for a long time on variants of plaque hypotheses (“this bad stuff accumulates and we have to stop it accumulating”)... it's now getting hard to ignore the fact that the immune system clearly plays a major -- and maybe dominant -- role. Elsewhere, we're plausibly on the cusp of effective dengue, AIDS, and malaria vaccines. That's pretty huge. Last, energy technology. There's lots more! New semiconductor technology. Improved ML and everything that that enables. Starlink -- cheap and fast internet everywhere! [Earth-to-earth travel](https://www.nasaspaceflight.com/2020/12/earth-to-earth-supersonic-airliners/) via space plus flying cars. The idea of urbanism that doesn't suck seems to be gaining traction. There's a lot of good stuff on the horizon. Why progress slow? A priori, I think it's completely plausible that many of the changes we've made have not been for the best. Most systems get worse in at least certain ways as they scale. The idea that science could have gotten worse in significant ways sometimes sounds strange to people -- like, we’re doing so much more, how could that be bad? -- but I think that misses the many examples of sensitivity of scientific processes to institutions and culture. Swiss nationals have won more than ten times more science Nobels per capita than Italians have. Ten times! And yet they're neighbors, and Italy certainly isn't lacking in scientific tradition -- Fermi, Galileo, the oldest university in Europe, etc. The "how" of science just really matters. At the micro level, when you look at the careers of individual scientists, the same thing is often striking -- if Born hadn't recommended that Rockefeller give Delbrück a fellowship, he might well have dropped out of science... his track record up to that point wasn't very impressive. But if he had, how far would that have set molecular biology back? So, given that science has changed a lot, we should push ourselves to really understand the effects of those changes, and we shouldn’t assume too casually that they’re all good. (2) Culture. As Ezra Klein recently [described](https://www.nytimes.com/2021/02/11/opinion/california-san-francisco-schools.html) in the New York Times, and Marc Dunkelman has written about in his [great piece](https://www.politico.com/news/magazine/2019/11/29/penn-station-robert-caro-073564) about Penn Station, a particular version of distorted, hypertrophic progressivism that took hold in the 1970s may have had (and still be having!) quite significantly stifling effects. We perhaps shifted from placing emphasis on our collective effectiveness in advancing prosperity and opportunity for people to the perceived fairness that was embodied in whichever particular steps we happened to take. Or, to say that another way, we shifted our focus from sins of omission to sins of commission. Take California, as Ezra does: there is almost endless attention paid to making sure that no single state project has even a tint of impropriety or suboptimality. The result of that cultural shift, however, is that the state as a whole is then often beset with awful results. With this ethos and panoply of strictures, it turns out that California is almost functionally incapable of constructing a high speed rail line connecting its two major metro regions. California has less civilizational capacity than the France of the 70s that built the TGV! (I spoke with Jerry Brown about this a few years ago and he commented that the change over the course of his lifetime on this cultural front was very striking.) California shifted mid century from being the US's fastest-growing state -- 50% population growth between 1950 and 1960 -- to a state that is somehow, improbably, shrinking. This is, obviously, mostly because of the regulations the state's inhabitants put in place that block the housing that's required to support California’s economic success. As a result, California has lost the "technology" of being able to affordably house its inhabitants. In these ways and many others, technology is both advancing rapidly and yet often receding in the state. (3) Institutions and first mover disadvantage. [...] The period of the early twentieth century was an era of building in the broadest sense, from universities to government agencies to cities to highways. The byproduct of this period of building is maintenance and we haven't figured out how to meta-maintain -- that is, how to avoid emergent sclerosis in the stuff we build. I see the exact same thing in financial services, by the way. Nobody in financial services thinks that real-time settlement is a bad idea. Cryptocurrencies show that it is a quite tractable problem. The "enemy", such as it is, is the calcification that follows from an existing install base. And all cultural questions aside, the US simply has a very large existing install base of aged institutions and systems. (4) Talent allocation. Maybe there's something about a substitution effect in where smart people go... perhaps certain sectors are hiring so many of the best people that other sectors have suffered. Most readers will probably have seen versions of the famous [innovator immigration chart](https://twitter.com/patrickc/status/1275288642022289410). [...] Anyway, this is a big topic in its own right, so I won't say too much here except to flag that "allocation of talent across sectors" is a big question and seems plausibly the source of a significant effect. Maybe people are just working on the wrong things. I wish there were more analyses here. [...] And I'll reiterate the point that most cultures through time haven't been, I think, especially conducive to progress or innovation, so I'd also invert -- maybe we should just identify those that were and try to figure out what it was that was special about them. s a meta point, cultural questions are tough for anyone to write about since even true diagnoses will never be wholly convincing (we have no counterfactuals) and because culture is a topic that people are often quite sensitive about. I thought that this was a [pretty neat paper](https://psyarxiv.com/d6qhu/) but lots of estimable onlookers [strongly](https://twitter.com/prof_gabriele/status/1192656942658916352)[disagreed](https://twitter.com/amyfullermorgan/status/1192767594379587585). As a result of these inhibitory forces, I suspect that good scholarship here is underprovisioned. [...] being willing to operate on longer timescales is very important. It’s part of why I’m so drawn to the Long Now (on whose board I serve) and it’s the single biggest thing I’ve taken away from [Derek Parfit’s work](https://en.wikipedia.org/wiki/Reasons_and_Persons). As Jonas Salk said, it’s important to think about whether we’re being good ancestors! verall, though, I worry less about any sector gobbling up all the smart people -- if there really are great opportunities elsewhere, some smart people will see that and things will eventually correct (as happened with finance and tech) -- and worry more about some sectors being structurally inhospitable to very talented people. I don't think that the ambitious upstarts who go into high-speed rail (in America, anyway) are going to have a great time or to have much success in convincing their friends to follow them. And I suspect that, for various reasons, too many domains look somewhat like high-speed rail -- what would a contemporary [William Rainey Harper](https://en.wikipedia.org/wiki/William_Rainey_Harper)'s experience be? There's a view that the internet is a frontier-of-last-resort and I don't think it's totally wrong. ## Ezra Klein and Collison Yep Stripe was mostly "make it much easier and get a lot of the details right". Many big success stories are like this – iPhone not the first smartphone; Slack not first chat, etc. Other major thing: was early to a massive growing market. Thresholds that lead to discontinue growth effects once you past a certain level of ease of use / great execution. Michael Mandell on build up of regulation: "enough stones in a river you have a dam". PC: all the Ubers that don't exist that we deeply can't see and so can't be mad about. We're reducing variance cutting off the tail disasters but also the tail massive wins. EK: everyone says too much regulation. Why do we never have the sense that enough cleaning has been done? PC: great Sunstein paper "a take on his time in government". #todo PC: Sunstein overestimates how well we can cost regulation. So many unintended consequences that are hard to see. PC on learning: heavy reliance on recommendation graphs - link heavy blogs, ppl who share their thoughtstream, pointers, SSC, MR, Twitter. Twitter as sticking a probe into people's heads. Noah Smith, pmarca, Julia galef, Danny Roderick, Stuart brand, Scott Aaronson blog O. Rationality community: one common thread subersive pleasure in realising that they're wrong, calibration. Uniquely truth seeking. Many key figures very free to say what they think. Also very 2nd/3rd/nth order, much less 1st order than e.g. mainstream politics. More attention to tails - positive and negative. Less concern with current limits In 2015 40% of Americans who bought online did so via a Stripe business Klein: think rationality crowd is more of a herd than you do. Company Culture when growing: it's not that we knew who we were at the start and gradually lose that. We're constantly trying to figure out who we should be, how to be more like ourselves. Metaphor: chiseling away. EK: You often start out with less of an idea who you are than you think. Respect for Bezos consistently using long time horizons, patience, to his advantage. Amazing how he managed to combined access to capital of being public with patience of being private. "He earned the right shareholders." Strong Recs David Deutsch beginning Ibfinity Really takes ideas, explanations seriuoy. Dream machine biography of licklieder. Really a book about ideas, how non accidental, right support, vision, communication setup, culture, produced enormous effects. Art of science and engineering. I think ppl don't think ## Crawford and Collison On Fast Grants: [existing funding bodies] are not good at responding when speed is of the essence and so we actually reached out to at the NIH and the NSF to try to get a sense for a sort of you know what it was that they were doing to respond to this  [thing] that is setting back global civilization [...] and we were struck and this is actually kind of public you know back in April if you just search the NIH is public database for how many COVID or coronavirus grants have been given at the time I think the number as of April was twelve or something like that and so we realized that it it may actually be possible and because so few organizations are suited to moving quickly to do something that unlocked scientists to to get started more quickly to run trials more quickly to test new potential drugs more quickly. So a week after this idea we launched Fast Grants, initially I think at launch with twelve million dollars. And within I don't know the exact details but within about two or three weeks we've made on the order of a hundred grants and so I think we were - I don't want to definitively claim this and in fact I hope I'm wrong - but we became as far as I know at the leading funder at least by a number of grants in the world. and these were really top scientists - when we published our open call we received 4,000 applications within the first week and as maybe you know if you ever put up an application on the internet you know or anybody can apply you know 50% plus but you get won't be that good we were astonished the vast majority of the applications like let's say at least 3,000 of them were very sensible very plausible perhaps even possibly worth funding and so Fast Grants had made the promise to respond with a definitive funding decision 48 hours subsequent to the application so we had some you know pretty intense reviewing to do [...] anyway so we made the grants they're now off the races and we actually are just in the process of receiving they're kind of second month of updates and it's proving really encouraging there are really promising drug candidates already developed there are you know things that are allowed even with the trials or vaccine candidates and so you know if one posits or accepts that COVID is one of the sort of greatest threats to progress in this calendar year you know we were having to be able to you know do something you know small but hopefully helpful on margins [...] and I don't want to be at all dismissive of other efforts that are taking place or you know kind of self aggrandizing for Fast Grants like you know others did and would eventually have acted and we'll see over time and how much Fast Grants appears to have actually mattered but at least as of April it sure seemed that the set of people who were committed to moving really quickly to actually acting on supporting and funding the top scientists in the world to solve our new tough medical challenge was actually exceptionally small. In fact Berkeley published an article maybe a month ago [May 2020] saying that Berkeley scientists had applied for all kinds of repurposing of existing NIH grants they already had, that they at the time had not heard back to any of those applications but again at the time I think four or five Berkeley scientists maybe even more had been funded by Fast Grants and so I guess this was a very potent example of this broader curtain phenomenon but by no means the only one. Q: Compare and contrast the effective altruism movement with progress studies. On EA… with anything that's a totalising framework (and I don't mean that pejoratively) I just mean it's like in principle everyone could have an EA mindset and I'm sure at least some EA members think that everyone should have an EA mindset. And then you can ask the question: would it be good for everyone to have an EA mindset? And then you can ask well is EA on the current margin a good new way for people to be thinking [...] would it be good for 5% of people or 10% of people. In that sense like was it a good shift on the margin I think EA has been great and I'm so delighted that they have had the progress that they've had. If the question is: should everyone be an EA or even in an individual sense am I or do I think I should be… um obviously there's a kind of heterogeneity within the field but my general sense is that the EA movement is always very focussed on a kind of rigid… not rigid that's unfair perhaps… on a sort of estimation analytical you know quantification and utilitarian calculation and I think that as a practical matter means you end up too focussed on things that you can measure which again means as a practical matter you're too focussed on things that are short term, you know bed nets or deworming or whatever. And like are those good causes I would say almost definitely yes although now we've seen some new data that they're maybe not as good as they initially seemed. But it's hard to see how writing you know a Treatise of Human Nature would score really highly in an EA framework and yet you know ex post that looks like a really valuable thing for Hume to do. And similarly, as we look in hindsight at things that seem like very good things to have happened in the world, it's often unclear to me how an EA oriented intuition might have caused somebody to do so. And so I guess I think of EA as like a metal detector in the backyard of good things to do, and they've invented a new kind of metal detector that's really good at detecting some metals that other detectors are not very good at detecting but actually I think we need some diversity in the different metallic substances to which our detectors are tuned and for me it would not be the only one. Q: What about the question of global catastrophic and existential risk? This is where I get the most concern from people about progress studies. I think it's probably true that the optimal rate of technological change is not monotonically better the more that there is. There probably are shear force with society at a certain point. I think the question is as a practical matter, should our concern be having too much or too little. And actually by the way I don't want to conflate progress studies with purely technological advancement although it's clearly a significant part of it. Looking historically I think that too little progress has been the problem rather than too much. And still today, too many lives are not as a good as they could be in a very tangible sense. For every 1 unit of concern that there's too much I give 4 or 5 that there's too little. Now there is an asymmetry there where the existential risk are you know, existential, and I thought Toby Ord's book was a great contribution and broadly I think the existential risk folks have introduced a good line of thinking that we really should be taking seriously. I suspect that it's possible to mitigate most of those risks relatively effectively without redirecting vast swathes of society, and I think the more difficult problem will actually be how do we generate enough progress. Later, audience question: What do we need to do to mitigate existential risk? [Pause] I don't know. I've wondered about it quite a bit. This is, as you know, something that top scientists and others wrote about, mostly starting in the aftermath of the second world war because it was obviously a salient question after the atom bomb… and… I don't know yet. Q: What have you what have you learned from Fast Grants what's been most surprising or unexpected for you? um well a big lesson in general is like you you know I started out in rural Ireland and I you know I so I kind of felt far from sort of where things were happening and I guess I sort of assumed that in the places where things are happening you know really enlightened people are making really enlightened decisions and the adults are in charge and again and again through my adulthood I've learned the lesson that there's nobody behind the curtain and I think that's simultaneously good and bad news in the sense that it's bad news if you have to rely on upon smart people taking care of things that you might think are a really big deal that you should not assume that they exist but it's good news in the sense that I think you can upgrade your estimation of your own potential impact too if you choose to pursue one of these things. -- Q: Is Bell Labs, Xerox Parc a thing of the past? One thing that is underestimated in stories about places like these: they competed on compensation. They just paid more than any other potential source of employment. Their strategy to aggregating the best computer scientists in the world was to pay them more than they would have been earning in academia. -- On Vanever Bush: I've huge respect for Bush yet I also think that The Endless Frontier sowed the seeds for the destruction of science. That's kind of hyperbolic but proposing a centralised federal agency actually creates some very pernicious incentive structures. The main thing I would do if I were writing such a report today is to argue for heterogeneity, decentralisation and competing approaches. I don't know what the best approaches are in a kind of singular sense, I don't know that it's knowable, I suspect that it does not even exist. [...] Many scientists at the time opposed [the Bush proposal], accurately foreseeing the hegemonic effects. -- I think GDP per capita is pretty good like I redid this analysis I think the correlation between GDP per capita and self-reported happiness is 0.76 and GDP against life expectancy is even higher it's about 0.8 A thing I believe about progress is that many of the relevant dynamics are less about key individuals and more about the dynamics of small groups and a really productive cluster of individuals and certainly that's what you tend to see in the past. [...] Clearly the degree to which the culture supports heterodoxy, that worked well for the Dutch, the Scottish, the English. [...] Basic liberal values, can anyones ideas be taken seriously. Entrepreneurship I see as the deployment mechanism for many scientific discoveries, though I don't think those are fully disjoint, ent clearly feeds back in. --- As Romer and others have pointed out what's interesting about ideas is that they can be nearly instantly deployed to produce a gain across an arbitrarily large economy. [...] So there is a scale invariance to knowledge. --- -- On fallibilism: I like the weight Deustch puts on explanations as a focal point for what we are or sort of should be engaged in. It's easy to get into a justificationist mindset where you're always seeking justifications for your beliefs. [...] Every explanation you ever have is going to be superseded by one that sort of fits reality that much better. [...] So I really like that mindset of accepting that we're always in error and engaging in the process of iterative refinement of our understanding. -- Advice for teenagers: I worry a bit about realising or thinking of oneself too soon as precocious or you know super talented. If you are super talented you start to realise that at 13, but I sometimes worry that it's easy to encourage people to kind of skip over the learning that you know should be a factor, that needs to happen. And this may be a flipside to some of these gifted youth programs. I guess it depends on the individual but do they make you think "man I'm capable of more than I thought" or do they make you think you're even more capable than you in fact are. [...] And so… if you're here, if you're a teenager listening to this course, you should probably be wary of overcorrection and that probably there are five or ten more years of knowledge acquisition of understanding and deepening that you should engage in before potentially committing to definitive courses of action. Later, responding to audience question: I actually went back to college to respond to this concern, because I thought I had not learned enough. [...] I think software is more forgiving for various reasons for people who have not acquired a lot of formal knowledge. -- We are still just so bad at building software, it still feels to me that we're building MS-DOS programs and there are so many layers beyond that which we're currently realising. Still day 1. --- On building innovative organisations: I think correlation between size and innovativeness is present but imperfect in that most small organisations are not highly innovative, some parts of large organisations are very innovative. A simple question is: are you attracting innovative people, and how enabled are they? We do a lot of customer interviews with our staff: how difficult or straightforward is it for them to get their ideas realised? I suspect that just pursuing those two questions gets you a long way. Q: What do you think about studying liberal arts if you're in college for technology? I think culture is hugely important if you view liberal arts as a study of and to some extend a determinant of culture I think it's a hugely important topic. But I think you really want to make sure that you're studying the right works and the right philosophies. [...] If you have some theory about how you can identify the right works, ideally a unique perspective on that, then I think it could be super high return. Joel Mokyr A Culture of Growth strong rec https://rootsofprogress.org/teasing-apart-the-s-curves ## 2021-03-13: Tyler and Patrick, Works in Progress PC: Japan and Switzerland -- how are they managing such a high quality of services? PC: LISP is a language for individuals, all about enabling a single individual to do as much as possible. TC" Stripe is a writing company. PC John is funnier than I am PC would watch more movies if he could watch them at 3x. Charlie Munger philanthropy strategy: support the best places, wants to double down on places that are already good. Fast Grants: a very large fraction of most promising work that was underfunded was in fact at the top institutions. There is in fact very strong self selection among scientists. A surprisingly large fraction of best work is happening at the best places. Even within top institutions, scientists are nowhere close to optimally supported. Even the very best can't get funded easily when a pandemic comes along. Why excited about carbon capture? -- how little has been tried. Supposedly sophisticated "well actually if you run the numbers it won't work". Underrates possibility of tech improvement, innovation. Could look like batteries and solar 20 years ago. --- Given the success of Stripe Press, when will you start Stripe Hollywood? --- What would it take to make sins of omission similarly or more salient than sins of commission? Do we need a marketing expert to reframe this somehow? Rory Sutherland? --- David Roodman of the Open Philanthropy project has a paper called "Modelling the Human Trajectory" where he points out that world GDP has grown at a superlinear rate over the past few centuries. He points out that our best models of world GDP predict extraordinary growth, collapse, or something that mandates a major changes to the models. What do you make of this? --- You both recognise that progress must be sustainable. What do you think about Nick Bostrom's claim that we'll need radically more effective global surveillance and policing capabilities to maintain a reasonable level of stability in the face of growing technological capabilities? --- How are you guys feeling about the results of Fast Grants, one year on? What did you learn from working on it? --- In the Noah Smith interview, Patrick said that dysfunctional talent allocation between sectors might be a big part of what is slowing progress. What kind of things might improve this? --- Tyler has a blog post about the high returns that can come from raising the aspirations of particular individuals. This seems good, but as a 1-1 intervention, it's doesn't seem very scalable. What are some underrated 1-to-many interventions along these lines? --- Peter Singer's drowning child thought experiment seems to have inspired many complacent people to care more about people far away from them. Are there similarly good thought experiments for progress?