YouSaid · the spoken record

Balaji Srinivasan

lines on the record: 605
first: 2022-10-20
most recent: 2022-10-20
sittings or episodes: 1
sources: podcast

Every line below is reproduced as it was said and linked to the record it came from. Nothing here is summarised or generated. Directory · Search · Corrections

“Are healthy volunteers that could have been soldiers, for example, of varying ages who are there to take a risk their lives for their country potentially. Okay, it could have been just healthy volunteers, not necessarily soldiers, just patriots of whatever kind in any country, not just US. But those healthy volunteers could have gone at the early stages of the pandemic, we didn't know exactly how lethal it was going to be because Li Wen Liang and 30-year-old, 30-somethings in China were dying from this. It seemed like it could be far worse. Halitha.”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“Right, but you can also see the Fauci example shows you why a lot of these folks do not want to be public because they enter a political and media minefield. I'm actually sympathetic to that aspect of it. What I'm not sympathetic to is the concept that in 2022 that the unelected, unfireable, anonymous American regulator should be able to impose regulatory policy for the entire world. Are not the world of 1945? It is not something where these other countries are even consciously consenting to that world. Just as give an example, there's this concept called challenge trials. The Moderna vaccine was available very, very early in the pandemic. You can just synthesize it from the sequence. And challenge trials would have meant that people...”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“Very important point. Like that person who's like named and their face is known as an example, you know Zuckerberg's face and name. Most people don't know Arthur G. Salzberger. They couldn't recognize him yet. He's a guy who inherited the New York Times company from his father's father's father. That is unaccountable power. It's not that they get great coverage, that they get no coverage. You don't even think about them, right? And so it's invisibility, right?”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“And it's not probably neither party would normally think of themselves that way. But if you look at it, when NYT goes and runs stories on a company, it names all the executives. And they get all hit. When it runs stories on a regulator, it just treats the regulator usually as if it was just some abstract entity. It's Zuckerberg's Facebook, but you can't name the people who, the career bureaucrats at FDA. Interesting, right?”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“Was emergency should mean like right now, right? But it was not just taking forever. And so some labs did civil disobedience and they just disobeyed the FDA and just went and tested academic labs with threat of federal penalties because that's what they are. They're like the police. And so we're sort of retroactively granted immunity because NYT went and ran a positive story on them. So NYT's authority is usually greater than that of FDA. If they come into a conflict, NYT runs stories, then FDA kind of gets spanked, right?”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“by the press, and they also aren't accountable to the market because you've got essentially uniform global regulations. Now, the thing about this is it's not just a government thing, it's a regulatory capture thing. Big pharma companies like this as well. Why? Because they can just get their approval in the US and then they can export to the rest of the world. I understand where that comes from as a corporate executive. It's such a pain to get access in one place. So there's a team up, though, between the giant company and the giant government to box out all the small startups and all the small countries and lots of small innovation, right? There are cracks in this now, right? The FDA did not acquit itself well during the pandemic. For example, there's so many issues, but one of the things that even actually New York Times reported, the reason that people thought there were no COVID cases in the US early in the pandemic was because the FDA was denying people the ability to run COVID tests and the emergency use authorization.”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“Small country will outsource their regulation to the USA with all the attendant issues. Because, I mean, you know the names of some politicians. Can you name a single regulator at the FDA? No, right? Yet they will brag on their website that they regulate, I forget the exact numbers, I think it's like 25 cents out of every dollar, something along those lines. It's like double digits. That's a pretty big deal. And the thing about this is people will talk about, quote, our democracy and so on. But many of the positions in, quote, our democracy are actually not subject to democratic accountability. You have tenured professors and you have tax-exempt colleges. You have the Salzburgers that New York Times who have dual class stock. You have a bunch of positions that are out of the reach of the electorate. And that includes regulators who have career tenure after just a few years of not necessarily even continuous service. So not accountable to the electorate.”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“Regulatory harmonization is the mechanism by which U.S. regulators. Impose their regulations on the entire world. So basically, you have a monopoly by U.S. regulators. This is not just the FDA. It is SEC and FAA and so on and so forth. And for the same reason that a small company will use Facebook login that will outsource their login to Facebook.”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“Sure. So, first, let me talk about the FDA thing. So, I was considered for a senior role at FDA, but I do believe that, and this is a whole topic, we can talk about the FDA, I do believe that just as it was easier to create Bitcoin than to form the Fed, right, reforming the Fed basically still hasn't happened, right? So just as it was easier to create Bitcoin than to reform the Fed, it will literally be easier to start a new country than to reform the FDA. It may take 10 or 20 years. I mean, think about Bitcoin, it's only about 13 years old. It may take 10 or 20 years to start a new network state with a different biomedical policy. But that is how we get out from perhaps the single worst thing in the world, which is harmonization, regulatory harmonization.”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“So now you go from a log file, which is just the individual record of one servers action, to A decade worth of data on literally billions of people. Of their online life, like arguably that's why I say that's actually what the written history was of the 2010s was this giant digital history. As you go to the 2020s and the 2030s, more of that is going to move from merely online to on chain and then cryptographically verifiable. So that soft subject of history becomes something that you can calculate things like Google Trends and Ngrams and stuff like that.”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“10 to 20 years, he was Arguably, that's already there, right? And here's the thing. You were saying, well, you might drop off when you hit psychology or history. Actually, I think it's the softer sciences that are going to harden up. Why? One of the things I talk about a lot in the book is, for example, with history, the concept of crypto history makes history computable. One way of thinking about it is remember my Britney Spears example, right, where Queen Brittany, right? Okay, so at first it's kind of a funny thing to say a computer scientist term for history is the log files. Until we realized that what would a future historian, how would they write about the history of the 2010s? Well, a huge part of that history occurred on the servers of Twitter and Facebook.”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“Here's something which I totally agree with him on. And he actually got trolled or attacked for this, but I completely agree. Within 10, 20 years, nearly every branch of science will be, for all intents and purposes, a branch of computer science, computational physics, computational chemistry, computational biology, computational medicine, even computational archaeology, realistic simulations, big data analysis, and ML everywhere. That to me is incredibly obvious why. First of all, all we're doing every day is PDFs and data analysis on a computer, right? And so every single one of those areas can be reduced to the analog digital step, and then it's all digital. Then you're flying, you're in the cloud, right?”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“Okay, so Let me give the strong form version, right? So there's a guy who I think is a great machine learning guy, creator of actually cheros, who he disagrees with me on Franco Schiller.”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“Well, yeah, yeah. So here's why you know, like Python has this concept of batteries included for the standard library, right? Because it lets you just import, import, import, and just get to work, right? That means you can fly. Whereas if you couldn't do all these things and you had to rewrite string handling, you would only be able to do incremental things. Libraries actually allow for greater innovation. That's my counter.”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“Okay, so I understand your point And I also agree that some papers like Francis Crick on the Klaustrum, others are theoretical. They're more about where to dig than the data itself and so on and so forth, right? So I agree with that. Still, the counter argument is. Rather than a thousand people reading this paper to try to rebuild the whole thing and do it with errors, when they can just import, they can more easily build upon what others have done, right?”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“Though, of course, it's got political overtones. Point being that the pseudonymous online publication at Project.gitub.io happened when it was taboo to do so. So we're back to the age of pseudonymous publication where only the arguments can be argued with the person can't be attacked. This is actually something that used to happen in the past, like there's a famous story where Newton solved a problem, someone said, I know the lion by his claw or something like that, right? People used to do pseudonymous publication in the past so that they would be judged on part by their scientific ideas and not the person themselves, right? And so I do disagree that this is the incremental stuff. This is actually the innovative stuff. The incremental stuff is going to be the institutional gerontocracy that's academia where it's like, you know, do you know who I am? I'm a Harvard professor. Yeah, science.”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“That's this revolutionary thing that came outside of it. Early in the pandemic, there was something called projectence.github.io, which accumulated all of the evidence for the coronavirus possibly having been a lab leak, when that was a very controversial thing to discuss, right? Alina Chand, to her credit, you know, Matt Ridley and Alina Chan have written this book whether the coronavirus was a lab leak or not. I think it's plausible that it was. I can't say I'm 100% sure, but I think at least it is a hypothesis wherey of discussion.”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“And so this entire centralized scientific establishment was set up at the peak of the centralized century. And it might have been the right thing to do at that time, but is now showing its age. And it's no longer actually geared up for what we have. Where are the huge innovations coming out? Well, Satoshi Nakamoto was not to our knowledge a professor, right?”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“1969, you have Hoover Dam, you have the Manhattan Project, you have Apollo. That generation was acclimatized to a centralized U.S. government that could accomplish great things, partly because technology favored centralization going into 1950 and then started favoring decentralization going out of it. I've talked about this in the book. Sovereign digital has talked about this, but very roughly you go up into 1950 and you have mass media and mass production and just centralization of all kinds giant nation states slugging it out in the world stage. And you go out in 1950 and you get cable news and personal computers and the internet and mobile phones and cryptocurrency.”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“Necessary condition for the presence of science. Neither is even the United States. Obviously, a lot of these discoveries Newton was before the, I believe it was before the American, let me find the exact, it's actually less old than people think. Okay, so Newton died 1727, right? So I knew that, you know, it was like in the 1700s. So Newton was before the American Revolution, right? Obviously, that meant huge innovations could happen before the US government, before NIH, before NSF, right? Which means they are not a necessary condition, number one. That itself is crucial because a lot of people say the government is necessary for basic science. It is not necessary for basic science. It is one possible catalyst. And I would argue that mid-century, it was okay because mid-century was the time when the middle of the centralized century. 1933, 1945,”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“And the answer to this is actually to say, well, what if we go further back than 1950? Did science happen before 1950? Well, I think it did. Bernoulli and Maxwell and Newton, were they funded by NSF? No, they weren't, right? Were aviation railroads, automobiles, gigantic industries that arose and about were stimulated by and stimulated development of pure science. Were they funded by NSF? No, they were not, right? Therefore, NSF is not.”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“The government hath granted us the Internet and self-driving cars and spaceflight and so on. How can you possibly be against the US government, kneel and repent for its bounty? And really, the reason they kind of, they don't say it quite in that way, but that's really the underpinning kind of thing because they've replaced GOD with GOV. They really think of the U.S. Garbman as God. You know, the conservative will think of the U.S. government as like the all-powerful military abroad and the progressive will think of it as the benign, all-powerful, you know, like nurturing parent at home. Okay. But in this context, they're like, how come you as, you know, some tech bro could possibly think you could ever do basic science without the funding of the U.S. government? Has it not developed all basic science, right?”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“And notice how, first of all, the average age is moving way upward before you become NIHPI. Second is it's a court of guys, people, who are just awarding grants to each other. Clearly, what's happening? That's the underlying dynamic. They're not wording grants to folks who are much younger because those folks haven't proven themselves yet, right? So this is what happens when you get prestigious citation rather than independent replication. The age just keeps creeping up. And this was 10 years ago, and it's gotten even worse. It's become even more gerratic, even more hidebound, right? And so the thing is the structures that Vannevar Bush and others set up the entire post-war science establishment. One thing I'll often find is people will say,”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“Innovator, for example, NIH has this graph, which is like, I think it's age of recipients of R01 grants. And what it shows is basically it's like a hump that moves over time, roughly plus one year forward for the average age as the year moves on. Okay, I'll see if I can find the GIF. Why is this? Let me see if I can find it actually. Look at this movie just for a second. It's a ridiculously powerful movie and it's 30 seconds. Sent in”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“Well, I mean, there's different, each of the steps that I'm talking about, right? There's like data being public and everything. Just having the paper being public, that's like V1, right? Then you have the thing being regenerated from code and data, like the PDF being regenerated from code and data. Then you have the citations as import statements. Then you have the full citation graph as an import statement. So you just follow it all the way back, right? And now you have, that gives you auditability. Then you have the off-chain, the analog digital crypto custody, right? Like where you're hashing things and streaming things. So you have the chain of custody. Each of those is kind of like a level up and adds to complexity, but it also adds to the audibility and the verifiability and the reproducibility. But one thing I'd say I wanted to respond to that you said was that you think this would be good for incremental but not innovative. I actually think it's quite the opposite. I think academia is institutional and it's not”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“My point is if you have a crypto instrument, you can have various hashes and stuff of the data as a chain of custody for the data itself that are streamed and written on chain, that the manufacturer can program into this. For anything that's really, and you might say, well, boy, boy, that's overkill, right? I'm saying actually not. You know why? If you're doing a study whose results are going to be used to influence a policy that's going to control the lives of millions of people, every single step has to be totally audible. You need the glass box model. You need to be able to go back to the raw data. You need to be able to interrogate that. And again, anybody who's a good scientist will embrace this.”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“Various sequencing runs. It'll tell you what instrument and what time is run and who ran it, and so on and so forth. That does is allows you to correct for things like batch effects. Sometimes you will sequence on this day and the next day, and maybe the humidity or something like that makes it look like there's a ciscally significant difference between your two results, but it was just actually batch effects.”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“Neither is it just trust the centralized academic establishment. Instead of saying trust because you can verify, because we can download things and run them. The crucial thing that I'm assuming here is the billions of supercomputers around the world that we have, all the MacBooks and iPhones that can crank through lots and lots of computation. So everything digital, we can verify it locally. Now, there's one last step, which is I mentioned the instruments, whether it's your sequencing machine or your accelerometer or something like that is generating the data that you are reporting in your paper when you put it on chain. Basically, you think that's the analog digital interface. We can cryptoify that too. Why? For example, an Illumina sequencing machine has an experiment manifest. And that's written to, there's a website called NCBI National Center for Biotechnology Information. You can see the experiment metadata on”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“You can internally check the consistency of these tables and graphs, and often you'll find fraud or things that don't add up that way because all the code and the data is there, right? And now you've made it so that anybody in Brazil, in India, Nigeria, they may not have an academic library access, but they can get into this. All right. Now, how do you fund all of this? Well, good thing is crypto actually allows tools for that as well. Andrew Hume and others have started doing things like with NFTs to funder lab. I can talk about the funding aspect. There's things like researchhub.com, which are trying to issue tokens for labs. But a lab isn't that expensive to fund. Maybe it's a few hundred thousand, a few million a year, depending on where you are. Crypto does generate money. And so you can probably imagine various tools, whether it's tokens or NFTs or something like that to fund. Finally, what this does is it is not QAnon, right? It is not saying don't trust anybody.”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“And putting them on chain. It's about the scale of, let's say, Wikipedia. Okay. So it's like, I don't know, a few hundred thousand, a few million papers. I don't know the exact number, but it'll be of that level. Okay. So now you've got number one. These things that are on chain. Okay. Number two, you've turned citations into import statements. Number three, anybody can now at a minimum download that code. And while they may not have the instruments, and I'll come back to that point, while they may not have the instruments, they can do internal checks. The Benford's law stuff we were just talking about.”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“Okay, so if it's citing something, you can now look it up and look it up and look it up. And a surprising number of papers actually, you know, their antecedents don't terminate or the original source says something different or it just kind of got garbled like a telephone game. And there's this famous thing on the spinach does actually have iron in it or something like that. I forget the details on this story, but it was something where you track back the citations and people are contradicting each other. But it's just something that just gets copy-pasted and it's a fact that's not actually a fact because it's not audited properly. This allows you to cheaply audit, in theory, all the way back to Maxwell or Newton or something like that. Okay. Now, what I'm describing is a big problem, but it's a finite problem. It's essentially taking all the important papers.”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“This itself changes how we think about papers. And actually, by the way, the inspiration for PageRank was actually citations. It was like the impact factor out of Academia. That's where Larry Page and Sergey Brin got the concept out of, right? So now you've got a web of citations that are import statements on chain. In theory, you could track back a paper all the way back to its antecedents.”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“That is facilitated if their entire paper is reproducible research that is generated. You can just literally import that Python and then you can generate your figure off of it, right? Moreover, think about how that aids reproducibility because you don't have to reproduce in the literal sense every single snippet of code that they did. You can literally use their code import it. People start compounding on each other. It's better science. Now I talked about this, but actually there's a few folks who have been actually building this. So there's usecholar.org, which actually has a demo of this, like just a v1 kind of prototype where it shows two stats papers on chain. And one of them is citing the other with an import statement. There's also a thing like called, I think, decide.com, which is trying to do this, right? Decentralized science.”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“That's cool, right? So now you're not just getting composable finance Like DeFi, where you have one interest rate calculator calling another, you have composable science. And now you can say this paper on this, especially in ML, right, you'll often cite a previous paper in its benchmark or its method, right? You're going to want to scatterplot sometimes your paper, your algorithm versus theirs on the same data set.”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“The open access movement has been fighting this for the last 27 years. There's various levels of this like green and gold open access. The first step is the code, the data, and the PDF go on chain. Step number one The second thing is once you've got, so you can anybody who is, and that could be the Ethereum chain, it could be its own dedicated chain, whatever, okay? It could be something where there's just the URLs are on the Ethereum chain and stored on Filecoin, many different implementations. But let's call that on-chain broadly, okay? Not just online on-chain. When it's on-chain, it's public and anybody can get it. So that's first. Second is once you've got something where you can regenerate the code or the PDF from the code and the data on chain, guess what? You can have citations between two papers turn into import statements.”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“Reproducible research is the idea that the PDF should be regenerated from the data and code. So you should be able to hit enter and regenerate it. Why is this really important as a concept? John Clairboo and David Donahue at Stanford 20 years ago pioneered this in stats because the text alone often doesn't describe every parameter that goes into a figure or something, right? You kind of sometimes just need to look at the code and then it's easy. And without that, it's hard. So reproducible research means you regenerate the PDF from the code and the data. You hit enter. Now, one issue is that many papers out there, science, nature, et cetera, are not reproducible research. Moreover, the data isn't even public. Moreover, sometimes a paper isn't even public.”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“You can sort of divide it into the analog to digital and the purely digital. Okay, so the analog to digital is you're running some instruments, you're getting some data, and then once you've got the data, you're generating figures and tables and text and a PDF from that data, right? Leave aside the data collection set for now. I'll come back to that, right? Just the purely digital part. What does the ideal quote academic paper look like in 2022, 2023? First, There's this concept called reproducible research, okay?”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“Such levels of uncertainty were not acknowledged. Instead, people were basically lying in the name of science and public policy wasn't public health, it was political health. So something like that, you're just spending on all the credibility of an institution for basically nothing. And so in such a circumstance, what do you do? Break glass, decentralize. What does that look like? Okay. Let me describe what I call crypto science by analogy to just like there's fiat science crypto science, fiat economics. So in any experiment, any paper when it comes out”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“But in part, they were also about power and whether the centralized entity would write all the rules or the decentralized one would. And so what happened was obviously Catholicism still exists, but Protestantism also exists. And similarly, here you've got this ossified central institution where forget about, I mean, there's complicated studies that are difficult to summarize. But when you have the science saying masks don't work and then they do, which everybody saw. And this is not like, you know, everybody knew that there was not like some massive study that came out that changed our perspective on mask wearing. It was something that was just insistently asserted as this is what the science says. And then without any acknowledgement, the science said something different, you know, the next day, right? I remember because I was in the middle of this debate. And I think you could justify masks early in the pandemic as a useful precaution. And then later, you know, post-vaccination, perhaps not necessary. I think that's like the rational way of thinking about it. But the point was that.”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“That's right. And so one way of thinking about this is I've said this in, I think I tweeted this or something, but Western civilization actually has a break classification emergency button. It's called decentralization, right? Martin Luther hit it. When the Catholic Church was too ossified and centralized, decentralized with the Protestant Reformation. He said, you know, at the time people were able to pay for indulgences, like that is to say they could sin. They could say, okay, I sinned five times yesterday. Here's the equivalent of 50 bucks. Okay, I'm done with my sin. I can go and sin some more. They really buy their way out of sin. Now, people debate as to how frequent those indulgences were. These are one of the things he invade against in the 95 theses. So decentralization, boom, break away from this ossified church, start something new, right? And in theory, the quote religious wars of the 1600s that ensued were about things like where the wafer was a body of Christ or what have you.”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“So many games devolve into the metagame of who writes the rules of the game, become essentially games of Nomic. Proof of work is so amazing because it didn't devolve in such a way, right? It became very hard to rewrite the rules once they got set up, very financially and technically expensive. That's not to say it will always be like that, but it's very hard to change.”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“Nomic is a game where the rule of the game is to change the rules of the game. At first, it seems insane. Then you realize that's Congress. Literally, what it is so meta because there are laws for elections that elect the editors of those laws who then change the laws that get them elected with gerrymandering and other stuff, right? That's a bad version. The other way of thinking about it is this is what every software engineer is doing. You are constantly changing the rules by editing software and pushing code updates and so on, right?”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“Exactly. It's very hard to come up with something that's stable in this way. There's actually on the meta point, gosh, there's a game where the rule of the game is to change the rules.”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“A fascinating thing. We can talk about that. Basically, once people know that you're using this as a measure, they will start to game it. And so then you have this cycle where sometimes you have a fixed point like Satoshi with proof of work was miraculously able to come up with a game where the gaming of it was difficult without just buying more compute. So it's actually, it's a rare kind of game where knowledge of the game's rules didn't allow people to game the game.”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“Whereas some paper that came out last week in Science or Nature may have zero independent replications, yet it is being cited publicly as prestigious scientists from Stanford and, you know, Harvard and MIT all came up with X, right? And so the prestige is a substitute for the actual replication. So there's a concept called Goodhart's Law. I'm just going to quote it. When a measure becomes a target, it sees it to be a good measure. For example, backlinks on the web were a good signal for Google to use when people didn't know they were being used as a signal.”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“A good one, right? Quote unquote science is a paper that came out last week. And the key thing is that capitalist science, real science, is about independent replication, not prestigious citation. That's the definition like all the journal stuff, the professors, all that stuff is just a superstructure that was set on top to make experiments more reproducible. That superstructure is now like dominating the underlying thing because people are just fixating on the prestige and the citation and not the replication, right? So how does that apply here? Once you start thinking about how many replications does this thing have, Maxwell's equation Every single time you pick up a cell phone or use a computer, you're putting our knowledge to the test, right?”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“And so academia right now, one of the big problems, this is kind of related to the oh boy, okay, the current institutions, we don't have trust in them, but the answer is that the answer to trust no one, right? And I think the alternative is decentralized cryptographic trust or verification. How does that apply to academia? First observation is we are seeing Science Being abused in the name of quote unquote science. Okay. Capital Science is Maxwell's equations.”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“Well, so here's the thing is. What you can do is let's say you did have a unit tested page for everything that's on Wikipedia. First of all, it makes a page more useful. You can download it, you can run it, you can import it, and so on. Second, it leads into one of the things that we can talk about. I've sort of like a roadmap for building alternatives to not just existing companies, but to many existing US institutions, from media and tech companies to courts and government and academia and nonprofits. The Wikipedia discussion actually relates to how you improve on academia, right?”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“Right. And the thing about this is, as I said, because the truth is a global constant, but like incorrectness, go ahead. Every happy family.”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source
“Yeah, exactly. At a minimum, for example, it's not WYSIWYG, right? So medium or something like that, or Ghost, you can just go in and type and it looks exactly like it looks on the page. Here you have to go to a market language where there can be editor conflicts and you hit enter and someone is overwritten your edit or something like that and you don't know how it looks on the page. You might have to do a few previews or what have you. So number one. So editing you talk about bearish editing. That's the thing. Number two is Given that it might be read a thousand times for every one time it's written, it is important to actually have the mathematical things unit tested if they can be given that we've got modern technology. And that's something that's hard to like retrofit into this because it's so kind of ossified, right?”
2022-10-20 · Lex Fridman Podcast · #331 – Balaji Srinivasan: How to Fix Government, Twitter, Science, and the FDA · IDENTIFIED FROM THE TRANSCRIPT · source