00:00:00.000 Like if anything, the bestiality and vegetarian example makes me more pro bestiality than I ever did.
00:00:06.000 But my parents, who rarely listen to the podcast, went on a road trip to Vancouver
00:00:22.000 and decided to listen to our most recent episode on AI. Apparently, our roles have reversed because
00:00:27.920 now I'm swearing every second word. And they were so impressed that you weren't swearing at all.
00:00:34.080 Like the only feedback my mom had for me was, yeah, pretty, you know, pretty nice episode.
00:00:41.200 It's amazing that Vaden doesn't swear. He's so good about not swearing and yet staying so passionate.
00:00:45.920 And then there's kind of fades to silence and I say, okay, what about me? And she says, oh,
00:00:51.360 it's terrible. You're swearing all the time. It's so funny. There is an old clip of us talking
00:00:55.600 about long-termism. So it's kind of apropos that somebody recirculated on Twitter. And I was
00:01:00.240 just swearing like a sailor. And now I'm like a bit more self-conscious about how much I curse.
00:01:05.600 And so I'm trying to try to limit it a little bit, but it's hilarious that your mom picked up on it.
00:01:09.840 And it wasn't all the brilliant things that you and I said that was important, of course.
00:01:18.880 But should we fill up our drink and get into this topic for today? I don't know.
00:01:22.640 How do you feel? Anything you want to say last minute before we dive into it?
00:01:26.560 No, I guess not. Let's just fucking let's rip off the band-aid, get it over with.
00:01:29.920 Okay, let's go. Okay, let me go fill up my drink. And...
00:01:33.520 Cool. Well, let me set up this conversation and then you're going to do an opening burst,
00:01:41.840 because I don't know how to even summarize my thoughts. But so today we are talking about
00:01:47.200 William McCaskell's newest book, What We Know the Future. This is my good friend Benjamin Chug
00:01:53.200 from Carnegie Mellon University. I'm Vaid Masarani from University of British Columbia.
00:01:57.840 And we, in episodes past, have talked a lot about long-termism, kind of stepped away from the
00:02:02.160 conversation for a year or so. But McCaskell's come up with his latest book, Can We Both Read It
00:02:06.880 and Have A Lot of Thoughts About It. And I think today we're just going to go back and forth a
00:02:10.480 little bit, talk about what we liked, what we didn't like. It's kind of an...
00:02:13.600 else between being a tough book to summarize versus being almost trivially obvious to summarize.
00:02:18.640 So I'm struggling to come up with a coherent opening monologue. So Ben volunteered to do so.
00:02:24.880 And I'm going to react to what you think. But that's my setup.
00:02:29.440 It's generous to say I have a lot of thoughts. I'm not here. I feel like every day I wake up and
00:02:34.240 I think about it more. Yeah, I go back and forth as to what's actually going on and how profound
00:02:40.320 I think it is. So I think maybe a helpful backdrop is considering Peter Singer's work on global
00:02:50.960 poverty in the 1970s. And in particular, he brought this influential philosophical article,
00:02:57.760 "Famine Affluence and Morality," sometime in the 1970s. And I think the goal there was to catalyze
00:03:06.240 a moral revolution, so to speak. So he sort of made these arguments that there's these people
00:03:12.880 living halfway around the world. And they're suffering a lot and we're spending money on
00:03:19.840 pretty inconsequential things. And with that money, there's good evidence that we can improve
00:03:25.280 their quality of life drastically and even save their lives in some instances. And basically
00:03:29.680 arguing that we should do that. Okay, so that's a thesis. You can disagree or disagree. But if you
00:03:35.840 adopt that thesis, the consequences are very actionable. So my cast goal, I think,
00:03:41.040 acknowledges that he's also trying to catalyze a moral revolution. But this time, thinking about
00:03:45.120 people in the future, he's got these three phrases sometime at the beginning of the book that basically
00:03:50.320 sums up the thesis. I think future people matter, we can- So future people count, there could be a
00:03:56.640 lot of them. We can make their lives go better. I just noticed your screen name on here is made
00:04:00.400 in the merciful. It's funny to change the name. So he's trying to tell us that like, you know,
00:04:13.840 Singer was saying we should be concerned about people in the world, at my castle hang, we should
00:04:16.880 be concerned about people, thousands, hundreds, thousands, even millions of years in the future.
00:04:22.320 And he wants this to like revolutionize sort of our moral landscape and now we go thinking about
00:04:26.640 the world. But after that, the analogy kind of breaks down because while it's more clear what
00:04:32.080 we can do to help people halfway on the world, obviously you run into the problems that just
00:04:35.760 like predicting the future is quite hard. And so it's much less obvious what we can do to help people
00:04:41.440 in the far future. And my basic read of the book is that he's trying to argue that there are things
00:04:49.920 we can do to positively impact this future. But I don't think he can escape the black hole of
00:04:58.560 uncertainty that surrounds trying to help people so far in the future. And because of that,
00:05:05.920 ends up focusing on problems that are quite present and pressing right now, right? So he ends up
00:05:12.320 focusing on pandemics, whether engineered or natural, ends up focusing on climate chains,
00:05:19.040 ends up focusing on trying to promote economic growth and avoid nuclear war, all things that are
00:05:25.840 very pressing problems that I'm also concerned about. I love that people are talking about. But
00:05:30.800 after he finishes, it's very unclear what the long in long-termism is supposed to do for us, right?
00:05:39.840 These are all problems that you can focus on without adopting his thesis that we should care
00:05:44.880 about people a billion years in the future. So sort of the delta between our current concerns
00:05:50.160 and what the book advocates we think about is pretty small. But because he's trying to be so
00:05:55.840 expansive in his claims, he ends up not really zooming in much on any of the particular problems
00:06:01.040 and just sort of says at a high level, we should think about these problems more carefully and
00:06:07.280 perhaps give more money to them, which perhaps is fine. But if you're trying to catalyze a more
00:06:11.920 revolution, I kind of expected some more actionable things. Like if everyone, I want everyone's going
00:06:16.880 to wake up as a long-termist to borrow. What would change? And I'm not sure that much would
00:06:20.560 change because now we're just being the position about all arguing what we're supposed to do to
00:06:25.520 make the future better. And I think that's what largely we're trying to do anyway, all the time,
00:06:29.920 right now. I think we all have opinions about the direction that we want to see the world go.
00:06:35.200 And that's in part why people disagree about shit all the time.
00:06:40.080 And so these questions are very hard. Like what we do about climate change, what we do about
00:06:43.760 nuclear war, these are intense questions touching on philosophy and geopolitics and technology and
00:06:50.880 economics. And it's fine to write a book saying these are hard problems and we should think about
00:06:56.800 them. But it's sort of inherently in actionable because he can't give actual solutions to all
00:07:03.360 these problems in one book, right? Because they're all like very deep problems. And so I'm just
00:07:08.080 kind of left not sure what I'm supposed to take away from it.
00:07:12.800 I love that I completely agree. And just to give a flavor of what you're talking about with the
00:07:18.800 actionable or lack thereof characteristic of this book. So at the end, he gives three concrete
00:07:24.560 recommendations for what we should do. And they are first, take actions that we can be comparatively
00:07:30.880 confident or good, two, try to increase the number of options open to us, and three, try to learn more.
00:07:37.760 And I get to this point in the book, I'm thinking, well, what is long-termism actually doing to
00:07:41.600 arrive at these recommendations? So if listeners go and check out our episodes from the archive
00:07:49.040 about long-termism, you'll hear us be quite opinionated that it's a bad idea. And we are
00:07:55.760 critiquing a paper that he wrote with Hilary Greaves called The Case for Strong Long-Termism.
00:08:00.240 And it seems like he's just taken out all of the stuff that we were criticizing. And we're left
00:08:05.440 with a book that I just feel like I don't really know what the point of the book is. He's trying
00:08:10.880 to make this case for caring about the future. But when it comes to anything is actionable,
00:08:15.120 you basically don't need it. Most of his recommendations for what we should do are things like
00:08:20.000 find energy sources that are less polluting, try to prevent bioterrorism, try to prevent
00:08:26.240 nuclear weapons. And all this is, from my perspective, just a trivially, obviously good thing to do.
00:08:31.920 And then the stuff that goes a little crazy is the AI stuff. And he's significantly downplayed
00:08:36.400 that in this book, which I can't tell if that's to his credit or not. There's a blog post
00:08:40.560 written about this book by an effective altruist author who is much more long-termist than
00:08:47.840 McCaskill seems to be. And he said, and I agreed with this critique a lot because he said that
00:08:53.840 he feels that people are going to get a bit of, they're going to experience a bit of a bait and
00:08:57.520 switch where they read the book. And they think, oh, this is all kind of reasonable and
00:09:01.840 maybe obviously true. And then they go have an 80,000 hours phone call or they go to any
00:09:07.600 effective altruist clubs or they join the forum and they see that the majority of what people
00:09:11.120 are actually talking about is AI safety and AGI super intelligence taking over the world.
00:09:15.200 They're going to be kind of taken aback because the AI stuff, which is a lot of what you and I
00:09:20.720 were critiquing in previous conversations, has basically been reduced to half a chapter.
00:09:24.560 And it's only one of 10 things that he recommends. But there's like then long-termism in the wild,
00:09:31.440 which is a very different beast, which is kind of created by McCaskill and Graves and Ward and
00:09:36.720 stuff. But this book doesn't really represent that, it seems to me. It represents this very
00:09:41.600 watered down version of it that is hard. It's just a unsatisfying read from my perspective.
00:09:48.160 There's something I liked and we'll talk about what some good hearts were. But it just seems to
00:09:52.000 be so watered down that what's actually presented to the public is this weird mix between trying to
00:09:57.840 make rather extreme claims about the billions of people in the future. But then when it
00:10:02.640 bottoms out to what he actually recommends, it's like try to learn more. Well, okay.
00:10:06.720 Obviously I want to learn more. So I just found it this weird super position of being extreme,
00:10:12.960 but then banal at the same time and I couldn't totally grab onto it.
00:10:17.040 Does any of that land with you or resonate? Totally. So the last chapter is about
00:10:23.440 basically summarizing some success stories of people who he thinks are doing good work.
00:10:29.360 But from what I can tell, most of them are not actually, or at least we're not
00:10:33.600 originally motivated by the EA community. Maybe they found some solidarity in the mission in their
00:10:38.320 work. But for example, he cites Isabelle Bomiki doing nuclear power activism, which is fantastic.
00:10:46.160 I love what she's doing. I don't think she did it for long-term as regions. I didn't think she did
00:10:50.480 it because she saw a huge political problem wherein people were pushing only renewables at the
00:10:57.920 expense of nuclear. And this was causing huge energy demand. It was a bad idea. So she started
00:11:03.280 talking about it in public. She didn't need to be motivated by long-termism to do that. Because
00:11:07.600 I almost agree with everything, maybe modulo some of the AI stuff. But even there, most of the actual
00:11:13.520 money, when they talk about AI stuff, it's different from where the actual money is going.
00:11:17.120 Typically, it's just going to run in the middle AI research, which is great. But anyway, so the
00:11:21.920 last chapter is I agree with basically all of these causes that he's talking about. But then I'm
00:11:28.160 like, "Well, it's kind of unfair to credit this to long-termism. I think they would have been doing
00:11:33.520 this anyway." And so maybe that's unfair. Maybe he would say, "Well, what I'm trying to argue is
00:11:41.600 not that there's a whole new set of problems we should focus on. It's instead a reshuffling of
00:11:46.400 our priorities." So it's not that I'm saying no one up to now has been focused on nuclear war or
00:11:53.440 engineered pandemics or something. But I just want people to take that problem much more seriously.
00:11:58.320 Which perhaps is fair. Again, we don't need long-termism to do that. There was like,
00:12:04.480 X-Risk was in the culture well before the phrase long-termism went into the culture.
00:12:08.880 After the pandemic, people are going to obviously be putting much more money in time into pandemic
00:12:14.080 awareness and not awareness pandemic preparation preparedness. And similarly with nuclear weapons
00:12:20.640 and stuff. And so it's like, do I need to make all these tortured philosophical arguments about the
00:12:28.160 importance of people a billion years in the future? Or can I just make down-to-earth concrete
00:12:35.280 arguments about the importance of nuclear energy of cleaning up fossil fuels of preventing bioterrorism,
00:12:43.760 like these, we don't need to go this route. We only need to go this route for the AI stuff,
00:12:48.640 which he's downplayed so significantly that I feel weird critiquing him for that in this book.
00:12:53.920 Because I don't want to critique him for stuff he hasn't written. But it's also, I don't think it's
00:12:58.640 an accurate representation of what long-termism is as practiced by people who call themselves
00:13:04.480 long-termist in the effective altruist community. And again, I will just cite this person who's way
00:13:09.040 more long-termist than McCaskill is. Critiquing McCaskill basically on this point that he didn't
00:13:14.160 adequately or accurately represent how long-termism is being in charge of the community.
00:13:17.840 So that is weird place where it's like, well, do I want to critique long-termism as practiced
00:13:21.600 by the community where I basically think it's like this techno-utopian ideology? Or do I want to
00:13:26.400 critique long-termism as he's presented it here? Which is, it's almost like reading a defensive
00:13:33.920 Islam by a very, very, very moderate Muslim where you don't want to just critique it because it
00:13:40.800 could lead to fanaticism. But you know that there's a fanaticism part of it in existence as well.
00:13:45.120 And it's so it's just this weird place where, and like, and I honestly think that so there's the
00:13:54.720 case for strong long-termism one, and then there's a second case for strong long-termism where they
00:13:58.320 read it in paper. And they took out everything that we quoted as problematic. So in some sense,
00:14:03.200 I'm reading this book thinking that a lot of the stuff that you and I said and wrote
00:14:07.120 actually influenced what is in the book and what's not in the book.
00:14:09.760 Because like, I can't say that for sure, but the evidence for that is that like
00:14:15.040 Shivani and every quote that we raised was stripped for the new version of the case for strong
00:14:20.720 long-termism. And none of that's in this book. And so is this even representing what long-termism
00:14:25.680 is as practiced by the people who call themselves long-termist? I don't think it necessarily is.
00:14:30.960 Yeah, yeah, yeah. I felt that way, especially about, so at the beginning of the book,
00:14:35.680 like I kind of expected him to do, he introduced like the tool of expected value reasoning,
00:14:39.200 right? And he said this is how we're going to reason about things. But then he like,
00:14:42.240 he barely uses it. So he does a bit. So he'll talk about like the probability of certain things
00:14:46.880 happening and you know, the usual conflation happens between his like just guess of what the
00:14:52.960 what the probability is and like some true probability. But he, because if you were to actually take
00:15:00.480 expected value seriously, you'd actually be able to like rank order all these problems, right? And
00:15:05.680 he'd say because the whole point of expected value, you're supposed to be able to tame uncertainty
00:15:11.360 using the numbers, right? That's the whole idea. And so you'd be able to take, you'd be able to bundle
00:15:17.200 all the uncertainty about like which problem to focus on, how much more money to put into that
00:15:21.680 versus some other cause that we're thinking about. Build a bundle that all into the calculation and
00:15:25.600 get very explicit rankings of of how, you know, how much more time and energy should put to
00:15:31.280 problem A versus the problem B. He doesn't do that for good reason, because that would come across as
00:15:35.360 insane to an ordinary reader, right? And be like, well, this is this is this is crazy. But he does
00:15:39.840 cite people who do and he seemed almost sheepish about assigning probability so stuff like he
00:15:44.000 buried some probabilities and footnotes and things. But then he'd cite affirmatively Tobeyour to does
00:15:48.560 this. Yeah. Yeah. Yeah. It's a weird relationship to it. So keep going. Yeah. But so so it's there,
00:15:53.280 but in the background. So so but that's what I mean. Sort of like there's like a partial,
00:15:57.440 only like a partial commitment. And so it's hard to criticize because I do agree with most of
00:16:02.320 these, right? Like I also am worried about technological stagnation. I'm obviously worried about I
00:16:07.440 also don't want to fall into like some totalitarian dictatorship. That also seems bad, right? I'm also,
00:16:11.760 I also want, you know, moral change to like not be locked in forevermore. The part of long terms
00:16:18.320 and long termism that we took issue with earlier was because they were giving like very precise,
00:16:23.760 like you should donate to AI over malaria bed nets because it's better by a factor of 1500 to one
00:16:30.640 or something. And then it was pointing out, okay, well, that's you're not comparing like with like,
00:16:34.640 there's there's some problems here with with your calculations. But because it doesn't make any
00:16:38.400 of those calculations. Yeah, you're just kind of left with this like, there's some problems.
00:16:42.560 People should think harder about these problems, which again is like, I'm like, okay, maybe that's
00:16:46.960 good. Like, well, I do want people to be more concerned about like nuclear war or something,
00:16:51.920 right? Like I think on average, if more people concerned about like, just aware of the fact that
00:16:55.760 it was still a huge nuclear arsenal in the world, that would be a good thing. And so I don't want
00:17:00.320 to say that it's not bad or that he shouldn't have written the book or something. It's just the
00:17:04.400 content. It's like the content, the logical consequences of what comes out of the book. If you take the
00:17:09.280 book and be like, what do I do based on this book? It's like very, very small, like logical content
00:17:14.960 that's this book. Yeah, I totally agree. And to the expected value thing, something that I noticed
00:17:20.560 him doing in this book and in podcast conversations is so it's not assigning explicit probability
00:17:28.000 to stuff, although he does do that and he hides some of it in the appendix. And he's kind of like,
00:17:31.360 almost embarrassed to do it, I feel. But he will propose scenarios, which are in direct
00:17:40.080 contradiction with one another and talk about both of them as being plausible. And this is allowed
00:17:46.960 under the expected value reasoning. So for example, he's very worried about AGI super intelligence
00:17:53.520 taking over the world. And he's also worried about technological stagnation. So he's asking us to
00:18:00.240 imagine one future scenario where technology just increases its capabilities and the speed
00:18:05.600 of development of technology grows and grows and grows and grows. And then there's another
00:18:08.800 scenario where technology slows down, slows down, slows down, stops. This is kind of like writing
00:18:13.680 a book where you're worried about overpopulation and you're worried about underpopulation,
00:18:18.000 where you're worried about the climate heating up and the climate cooling down. And it's not a
00:18:23.440 contradiction because under the expected value calculus, these are all equally plausible scenarios
00:18:28.880 and you just assign probabilities to them. But he's not really on the line. He's not advocating
00:18:34.720 a position which then can be falsified by another book. He just can talk about everything happening
00:18:41.280 all at once. So I wrote down a number of these explicit contradictions that I just thought were,
00:18:47.600 I just called it having it both waysism. So he'll talk about how we can make a big
00:18:52.080 difference to future generations. And then later on, he talks about how the future is
00:18:55.680 impossible to predict and anything that we do, the consequences of these are like ripples in a
00:19:00.640 pond. And he'll talk about technological progress speeding up and technological progress slowing
00:19:05.680 down. And he'll talk about how the resiliency of human beings and how we're able to constantly
00:19:12.240 overcome problems and hurdles and difficulties. And then he'll talk about how we're at a significant
00:19:16.640 risk of being taken over by a totalitarian government or a super or super AGI. And so this is this
00:19:23.200 like weird consequence of this expected value reasoning where you don't have to commit yourself
00:19:28.720 to any one position. You can kind of just have them all at the same time. And it's weird because
00:19:33.760 you could just like swap out the arguments and you could use all the arguments for technological
00:19:38.400 stagnation as being reasons why we shouldn't worry about AGI super intelligence. Similarly,
00:19:44.720 you can take all the arguments from AGI super intelligence and use that as reasons why we
00:19:48.000 shouldn't be worried about technological stagnation. And so it's hard to grab onto because he's kind
00:19:53.360 of just he does it all at once. And then if you argue against any one particular thing, he'll just
00:19:58.800 say, well, yeah, there's a lot of uncertainty. Let's reduce our probability for that. And so you
00:20:02.880 can never really pin him to anything. And then in like podcast conversations and stuff, Tyler Cowan
00:20:09.280 asked him about long-termist implications for abortion. And shouldn't you, if you care about
00:20:14.160 future generations, shouldn't you be pro life? And you want there to, and similarly, he tried to
00:20:19.200 pin him down on stuff like expected value reasoning. Would you want to have a, if there's a button
00:20:25.040 that you could press that would have like a 51% chance of like giving everyone infinite life and
00:20:31.600 49% chance of killing everyone? Or even just doubling the population or something, right?
00:20:36.480 Yeah. Yeah. And then he'll just say, well, yeah, so expected value reasoning doesn't get you
00:20:40.960 all the way and we have to have a plurality of views. And then he'll just basically abandon it
00:20:44.560 whenever there's a tough question thrown at him. And it's like, what is this even doing for you
00:20:50.480 besides being a rhetorical device? Like, I just see it now as kind of a rhetorical device that's
00:20:54.560 being used, because he's not committing himself to actually using any of the calculations. He just
00:21:00.480 talks about it. But then when challenged, the band is it the first sense of challenge, I guess.
00:21:06.240 Yeah, interesting. Yeah. Okay. Let me try and defend just a subset of that, which is,
00:21:11.760 if you write a book, I could imagine someone writing a book saying, underpopulation would be bad,
00:21:17.360 because it would slow down progress, people are awesome, etc. So we don't want everyone to stop
00:21:23.760 having kids all of a sudden, because that would be bad. But similarly, we don't want the population
00:21:29.280 to explode within a generation or something. Right. So we should simultaneously be worried,
00:21:34.240 but I'm not saying this is my position. I'm just saying you can imagine someone writing a book
00:21:37.040 saying we should be, we don't want underpopulation and we don't want overpopulation.
00:21:42.000 I don't view that as like having it both ways. I just view that as like pointing out two
00:21:46.160 possible problems. And then you can criticize, you know, like, I don't know if that's a consequence,
00:21:51.520 one of the expected value calculus or just, I think he's just trying to point out a lot of
00:21:55.520 possible problems. And then we can debate about the feasibility of those scenarios.
00:22:00.400 I don't think the analogy is someone writing a book saying underpopulation is bad, overpopulation
00:22:04.160 is bad. I think the analogy would be someone writing a book saying underpopulation is likely,
00:22:08.240 and overpopulation is also likely. And it's not a contradiction, because under this framework,
00:22:16.560 basically, you can just speculate wildly about stuff that might happen. But it does, I think,
00:22:22.320 strain, julity a little bit to talk about two possible futures, which are diametrically opposed
00:22:29.760 to one another. And if you're a reader of this book, and let's say I think that
00:22:37.360 technological stagnation is really likely, and you think that a GI superintelligence is really
00:22:41.040 likely, we're going to donate money to competing organizations, one of which is trying to accelerate
00:22:46.480 technological growth and the other is trying to hinder it. And what's frustrating is it's not
00:22:51.280 a contradiction. I recognize it's not a contradiction. It's just the way that this kind of expected
00:22:56.160 value reasoning allows you to think. And you can just speculate wildly. I want to talk about
00:23:03.120 some of the stuff I liked about the book, because there was some stuff that was genuinely fascinating,
00:23:06.880 but maybe we'll do that in a bit. But there's certain claims that he made that I just thought
00:23:13.040 were flat-out absurd. So one of them is, I'll suggest that there's a significant chance of a third
00:23:18.240 world war in our lifetime. I don't know how anyone can possibly know that. It's just speculation.
00:23:24.640 And to say that there's a significant chance is, I think he also says one and three at some point.
00:23:32.640 Right? So I didn't see that. Maybe I did. I didn't see that. And he'll say things like,
00:23:36.960 if we could create genetically engineered scientists to have greater research abilities like Einstein,
00:23:44.000 this would compensate for fewer people overall. And it's like, what are you talking about?
00:23:48.640 How could you possibly genetically engineer better research? But you're just allowed to
00:23:54.160 speculate about the future like this. And you're not constrained to anything besides certain chunks
00:23:59.360 of it being plausible when phrased one way. But then you read diametrically opposed arguments
00:24:03.680 like a couple chapters later, and you kind of just forget that he's not contradicting himself.
00:24:08.160 He's just exploring every possible future that could happen all at the same time in a way that
00:24:13.840 you can never pin him down on anything. And it's all probabilistic. And I found it very frustrating.
00:24:19.360 When it leaves all the work to the reader, right? So it's like he's saying, well, this could happen,
00:24:23.120 and this could happen, this could happen. And then the question is, okay, what should we do?
00:24:28.800 But how do we start arguing? How do we debate about which one to do? And just by pointing out
00:24:36.400 all the scenarios, I don't know, maybe there's a role for that. But then when it bottoms out,
00:24:41.920 it bottoms out to stuff that people were doing anyways. It's like, exactly.
00:24:46.480 So what's the extra value of doing this? Sorry, go ahead. I'm just kidding.
00:24:50.960 Yeah, I mean, I'm not sure. I mean, yeah. So it was so, but as an example, like, so yeah,
00:24:56.000 he went on the Tim Ferriss podcast. I mean, dude, they must have pulled out all the stops on
00:25:00.560 their connections. His podcast. Because he was on every podcast in my podcast. But I was just like,
00:25:06.640 this is why I'm like with the exception with the exception of Sean Carroll and Tyler Cowan,
00:25:11.840 everybody else was just like fawning over this book. Like even Sam Harris, and I like have a
00:25:16.880 huge amount of respect for Sam Harris. But I thought he was given such an easy pass on all
00:25:21.360 this whole podcast circuit. And everyone's just like, oh, he's a more of a philosopher. He must be
00:25:25.840 a saint. Sorry. Again, I'm kind of just pitching about things at anointing.
00:25:29.280 Well, because, yeah, I mean, yeah, I, but this is like, sometimes I wake up and I'm like,
00:25:36.400 like today I just, you know, like woke up and like walking to the gym. And I'm like looking up
00:25:40.640 at the stars. And I like, I do sometimes like emotionally, I can like resonate with what he's
00:25:48.240 saying. Like, I think humans are fucking awesome. I want us to like go to the stars. If we are the
00:25:54.240 only like conscious creatures in the university observable universe or whatever. And we did go
00:26:00.960 astink. I do think that would be a huge tragedy. So I it's not like I'm not concerned about like
00:26:08.000 future people, the future of the species. Like I want all the good things to happen to us.
00:26:12.480 In so far as the book makes people think more about that and forget like some of the more
00:26:17.600 parochial problems and, you know, like the classic thing of astronauts going to the space station
00:26:22.640 and be like, Oh, it's just a single blue dot everything so precious and stuff. Those kind of
00:26:27.280 insights are are good. And I do maybe want more people to register that. I think that's my most
00:26:35.040 robust defense is just like in so far as he's talking about that humans are pretty good.
00:26:41.520 And we should run our future to be awesome. And also that we should take our actions and our
00:26:46.800 ideas seriously, right? Like one through line of the book is we can our ideas do matter. And we,
00:26:52.000 you know, if we fuck up, we really we can fuck it up big and nothing's on autopilot, right? Things
00:26:56.960 have gone bad in history before. If we don't stay serious about the power of ideas, then like
00:27:02.320 civilization can collapse and stuff. And all those messages are great because I mean, one,
00:27:07.040 there's one contingent of people in the US, especially who like think civilization is terrible. And
00:27:14.160 or just think like no one has the power to do anything and people are super apathetic and stuff. And so
00:27:19.200 in so far as he's countering messages like that, like fuck yeah, you know, like I. And so
00:27:24.000 sometimes I'm like, can kind of get behind them like yeah, make people think more about this. And
00:27:28.000 you know, like he talks about the value of having kids, which I thought was pretty cool,
00:27:31.840 which no one really wants to talk about anymore. And I was very impressed by right, especially
00:27:35.440 because that's so contra current, current, current environmental, environmentalist messaging,
00:27:40.400 right? So the fact that he took a stand and he said that on multiple podcasts too. So you know,
00:27:43.840 kind of took a stand like yeah, like having kids is a great thing to do totally. And so
00:27:47.200 certain of the messages like that, I'm like, hell yeah, but but then yeah, I mean, once you get
00:27:51.280 into very particular problems, I just thought it was it was quite shallow how we dealt with it.
00:27:55.200 And like one example, you know, he was talking about the green movement in Germany. And how
00:28:01.200 about they sort of catalyzed a bunch of money to go into renewables. And he said, you know, this
00:28:07.600 made renewables a much more respectable thing made other countries take it more seriously,
00:28:11.760 etc. But what he neglected was that that push for renewables also made Germany disband tons of
00:28:17.040 their nuclear plants, which has now landed them in massive problems when it comes to the Russia
00:28:22.480 Ukraine conflict, right? And just environmentally in general. And so but this is a perfect example
00:28:27.200 of how he kind of skates over the complexities when he's dealing with these concrete problems,
00:28:30.960 right? Like imagine being someone who like has spent your whole life thinking about pandemics or
00:28:36.240 nuclear war or something. He's really not adding anything adding anything new to the table. He's
00:28:41.200 just saying this is a problem. And you might be kind of annoyed if you're an expert in this area,
00:28:45.120 like, well, what the fuck I fucking know it's a problem. Like, you know, I'm down here trying to
00:28:48.800 do all the geopolitical heavy lifting to like figure out how to do this stuff. And you're kind of
00:28:54.320 just giving me vacuous statements about how it's important. Anyway, yeah, that kind of sums up my
00:28:58.160 my my my conflicting feelings. Yeah, well, since we're getting to like some of the good parts of
00:29:02.960 the book, I should kind of give a bunch of examples of things that I liked. And I totally agree with
00:29:07.360 you that it was refreshingly like pro human pro progress. One thing that I thought was nice about
00:29:14.960 the book is that there were and you kind of alluded to this, but didn't say it explicitly. But there
00:29:18.640 is like zero traces of volcanism in it. Yeah, which I just thought was nice. Yeah, just quite
00:29:24.560 refreshing. And so he's kind of presenting a different lens, but just in terms of being
00:29:30.640 tainted by seeing oppression everywhere and stuff, there's zero traces of that, which I thought was
00:29:34.160 great. And there's a lot of climate optimism too, which I thought was interesting. Yeah,
00:29:38.400 yeah, which was nice. Like he didn't like he talked about climate change being a problem,
00:29:42.800 but it's not. But he also acknowledged like that we're making a real progress on climate change.
00:29:48.320 And that I think he says somewhere that recent years have given us more cause for hope than any
00:29:53.200 other point in my lifetime. And he pushed back strongly on like the climate movements,
00:29:56.960 encouragement to people for people to not have children. I thought that was great that he was
00:30:03.200 is really a pinker style optimism, I guess. Yeah, totally. And at the end of the book,
00:30:09.440 so just kind of going through a bit of a list of things that I thought was great about it.
00:30:12.640 At the end, I thought he gave some really great career advice. And his career advice was basically
00:30:17.760 to treat your career like an experiment. And so he says, in practical terms, you might follow
00:30:24.480 these steps, research your options, make your best guess about the long term path for you,
00:30:29.200 try it for a couple of years, update your best guess and repeat. And this is just like perfectly
00:30:33.920 poparian advice about how to make a good career. And I thought that was great. And there's just a
00:30:40.080 couple of random points that he made that I just wanted to highlight because I didn't hear these
00:30:43.520 said on podcasts before, but he made the very excellent point that I thought was just such
00:30:48.960 obviously a good idea that I can't believe I hadn't heard of it or thought of it, which is that
00:30:53.040 not a single country allowed the vaccine to be bought on the free market prior to testing.
00:30:57.840 By those who understood the risks, even on the condition that they report,
00:31:00.640 whether they were subsequently infected, just this idea that there could have been and should
00:31:05.040 have been a free market for COVID vaccines amongst volunteer. I thought was a really important insight.
00:31:11.680 And one that I hadn't heard before. And so just props to my Casco for saying that.
00:31:16.720 Yeah, regarding the vaccine, he made some fascinating points there. One, he pointed out how homogenous
00:31:23.280 the global, more or less, how homogenous the global response to COVID was. Obviously,
00:31:28.320 there was some regional variation, right? South Korea comes to mind, right? But if you consider
00:31:34.240 the massive space of possible responses to something like COVID and then how little of that space was
00:31:39.360 inhabited by most countries, even when other stuff seems like actually quite good ideas,
00:31:44.000 like challenge trials, etc. That was quite interesting. And then he follows that up by saying,
00:31:48.480 he proposed the idea of charter cities. So did you have a thought there? That was quite interesting.
00:31:55.600 At first, I thought that was super interesting. And then I started thinking,
00:31:58.720 well, what's the difference between charter cities and just different cities?
00:32:01.520 Like, how do you put into it? So like, maybe you should say for the audience,
00:32:05.920 what charter cities are first before I opine on it a little bit.
00:32:08.560 Yeah. So my understanding is they're sort of like independent regions, so they wouldn't be bound
00:32:14.880 to like, they wouldn't be bound to like a specific country, say. And there would be cities that
00:32:19.200 are trying out a specific, perhaps like economic plan. So he kind of hedges and says, it could be
00:32:25.520 like different moral views or it could be different economic views. But maybe you have a charter cities
00:32:29.520 for like, like democratic socialists who want to try something more along that vein, then you have a
00:32:35.680 charter city for like, libertarians or like anarcho capitalists or something who want to live in a
00:32:40.960 much different way, right? And everything's like voluntary and stuff. But you have these like,
00:32:44.960 cities that are trying out very different experiments on how to live. And to so that civil,
00:32:50.160 you know, we can kind of see it's like running little experiments, like, what's going to work,
00:32:52.960 what's not going to work. Obviously, there's some practical issues, right? Like how, you know,
00:32:56.480 how are these systems economic, they have to relate to each other economically somehow. We need
00:33:00.560 like global free trade and whatnot to some extent. And so they have to be bound in some sort of
00:33:05.440 social, international norms and stuff. So the details seem quite fuzzy, but
00:33:09.040 I can get behind a general sentiment that like, we should be running more experiments.
00:33:14.240 So at first, I liked it, but then I don't see how you could implement this without
00:33:18.320 being an authoritarian totalitarian. See, like you could do it in China, I guess. But
00:33:25.440 you have to give people free movement. Yeah, for sure. And then you have to decide how a city's
00:33:31.360 going to like, which economic policy city is going to implement and to do that, again,
00:33:35.840 in a non totalitarian way, it has to be democratically elected policies. And then you just have people
00:33:41.200 voting for economic policies in different geographical regions, which I think just gets us back to
00:33:46.640 regular cities. Well, like, I don't know how you would go like, how would we do this in Canada or
00:33:50.880 the United States, for example, just you could get a bunch of libertarians and they would go off to
00:33:55.840 it. Like, would we just take existing cities and replace libertarians or replace the people with
00:34:00.880 a bunch of libertarians and stuff? Like, it just doesn't seem like it's at all feasible to do
00:34:07.040 without violating civil liberties. Maybe I mean, I can kind of see natural experiments arising
00:34:15.120 as long as you let cities be a little more autonomous than they're lawmaking. So I'm considering like,
00:34:20.240 something like San Francisco versus Austin, Texas, right? So San Francisco tends to have people who
00:34:24.960 vote for like much more state intervention, right? They might vote on much more left leaning policies.
00:34:29.840 And that's going to also attract people who think that's a better idea. Okay. Yeah, San Francisco is
00:34:33.600 not having a great time at the moment. And so that could almost be considered one type of experiment.
00:34:38.400 Obviously, San Francisco is still bound by like certain state laws and stuff. So what you might
00:34:42.000 want to say is like, okay, now San Francisco's can just like vote on everything. They don't have to
00:34:46.320 abide by the state regulations. They just vote on everything. And then people don't like that,
00:34:50.480 which seems to be a lot of people are like moving out of San Francisco, moving into places like
00:34:54.080 Austin, Texas, and you know, Texas tends to be more on the libertarian side of things. So these are
00:34:59.520 not like mandated charter cities, but they're getting closer to that sort of idea, right? Like
00:35:05.280 San Francisco right now is definitely an experiment in living. One that arguably is not
00:35:10.160 but there's still this big difference between collecting data and analyzing natural experiments.
00:35:15.200 So experiments which have just naturally taken place based on people's free choice and movement
00:35:19.440 and voting policies. First, implementing this as a strategy and saying we are going to
00:35:25.680 decide that San Francisco is going to be more blue and Austin, Texas is going to be more
00:35:30.560 libertarian or what have you. So yeah, because as charter cities were proposed in the book, it
00:35:34.640 seemed like it was like experiments that would be run by who exactly like it. I just don't know
00:35:42.000 how you could do this except yeah, without being like a China and just forcing people to live in
00:35:46.240 places. Yeah, yeah, yeah. So it was an interesting idea for sure, but then when I thought a bit more
00:35:50.640 about it, I don't know how you could do it with it. Yeah, possibly important. Yeah. I just think
00:35:54.640 also there's an issue. I was talking about this with Mira and we're talking about some of the
00:35:58.880 difficulties. She was like, well, okay, but what if you want to have like, what if like a city of
00:36:04.320 people come together, right? So you could think of like certain religious extremists who are like,
00:36:09.360 we want to live in a state where women have no say in how they how they live, right? And so
00:36:18.320 you're just going to allow that to happen because it's like quote unquote an experiment living.
00:36:22.560 Like probably not. So it still has to be even a charter city idea. There's no it can't be anything
00:36:28.800 goes right. It still has to be implemented in some sort of like liberal order or something. So
00:36:33.520 you can only push it so far. So here, what did you think about the whole value lock in? Because I
00:36:38.080 might I'm slightly skeptical, but I'm curious what your thoughts are and that maybe we should
00:36:42.720 summarize for the audience what what value lock in is. Sure. I mean, I guess yeah, even more broadly,
00:36:48.320 maybe we should just give a slight outline of the book because we've been talking free for him so
00:36:52.240 far, but we should say like, right? So he's concerned about the long term future, but he breaks the book
00:36:57.760 into several parts and parts two and three talk about two sort of distinct ways that you might go
00:37:04.000 about influence in the future or ways in which the future could be good or bad. Part two is about
00:37:08.560 trajectory changes. And so this is about values, basically over time. So the two sub parts of that
00:37:16.800 are about moral change and value lock in value lock in being like if we got locked into some
00:37:22.000 total totalitarian dictatorships or something like say the Germans had won second world war and we
00:37:28.080 were all living in a thousand year Reich at the moment that would be like a form of a value lock
00:37:32.480 in assuming like it couldn't be changed or something. And that's where he gets into some of the AI
00:37:36.800 stuff like AI could could help with the cementing totalitarian values. Yeah, it was the AGI. It was
00:37:42.400 like the the driver of the value lock in. Although he didn't say that explicitly, the main arguments
00:37:48.240 for value lock in were about AGI. So that's where the AGI system came in. Yeah, yeah. And then the
00:37:54.400 next part is about safeguarding civilization. And this is I would say most of what we've been talking
00:37:58.320 about so far. So like the extinction risk, technological stagnation, civilizational collapse,
00:38:04.000 things that you most commonly associate under like this sort of a broad umbrella of like
00:38:08.320 more existential risk type type stuff. So yeah, what did I think of the value lock in? I thought
00:38:13.280 I thought there were the parts that were of great merit like it is there are some fascinating
00:38:19.280 historical case studies of how certain ideas emerged over time and we're able to like win out
00:38:24.720 against other ideas and became entrenched. And so like he gives a whole background on
00:38:30.560 Confucianism and the Han Dynasty and the Maoists, right? And like how they were sort of battling
00:38:37.280 for prestige and then how Confucianism like ended up ended up winning out over time and basically
00:38:42.800 cemented itself as a certain ideology over the span of like 2000 years. And so parts of it are
00:38:49.280 just testament to like how powerful ideas are and how much we should care about spreading norms of
00:38:56.320 like open inquiry and debate and making sure that I mean basically like the classic liberal values
00:39:03.520 of like free speech and people need to be allowed to point out what they see as problems and talk
00:39:08.400 about things right because we don't want to get ourselves in a situation where we get we don't
00:39:13.920 tolerate dissenting views in case we're sliding into some sort of bad moral system or something.
00:39:19.120 I found the value lock in stuff to be a little difficult to conceptualize because I think values
00:39:25.520 tend to get locked in when they are maximally beneficial to the most amount of people. The
00:39:30.720 historical examples all run into the problem that you just didn't have global communication back then.
00:39:36.320 So you didn't have ideas being able to be spread quickly. Also when you go back that far in time,
00:39:44.320 we just weren't as morally or scientifically violent to his pre-enlightenment. And so I think
00:39:50.240 post-enlightenment, I have a difficult time imagining a future where we all get locked in
00:39:54.880 to a set of values which are making everybody miserable but nobody thinks to question them.
00:39:59.120 And this is one of the things he says somewhere which is that he's worried about us all getting
00:40:03.440 locked into a state of nature where we're all being made absolutely miserable. It's in the future.
00:40:10.160 So there's obviously communication and stuff but no one challenges these values.
00:40:15.520 Like once you have a dystopia like that, this is a problem to be solved. And I think the
00:40:22.080 history of the 20th century in the Cold War in particular highlights that totalitarian regimes
00:40:27.520 just aren't fundamentally stable because you don't have flourishing and people want to
00:40:33.200 have a better life for themselves and their children and they're naturally going to start
00:40:36.160 challenging certain values like slavery, anti-Semitism, these kind of values which are creating misery.
00:40:42.800 And so I didn't find it very persuasive and also because most of what he argued was like an
00:40:49.280 AGI getting into power and locking in these values. And we don't need to talk too much
00:40:53.520 about the AGI stuff because we talked about that ad nauseam. But he used historical examples in
00:40:58.640 AGI and I don't think both, either of these things work for different reasons but historical
00:41:02.720 examples, particular, it's pre-enlightenment and so the world is just a very different place,
00:41:07.040 post-enlightenment, scientific revolution. So although I didn't find that too persuasive but
00:41:12.320 I guess- You're making me realize that I guess there's two different kinds of value lock-in
00:41:17.600 that maybe he distinguished but I can't remember. So one is sort of what he talked about with just
00:41:23.440 ideas winning out over one another and there I agree with you it's hard to imagine a world
00:41:29.440 right now where everyone just starts to think a certain way and it's making us miserable and no
00:41:34.400 one thinks to question it, especially with things like the proliferation of the internet and stuff.
00:41:38.160 But then there's the question of like, okay, but if the values were enforced from top down
00:41:42.480 and he goes like the AGI route but you can even just imagine like a North Korea type situation,
00:41:47.280 right? Like North Korea takes over the world. There you have certain ideas in force and it's not
00:41:51.920 that no one thinks to question them, it's just that no one can question them because you're living
00:41:55.360 under some form of dictatorship. And so to that I say yeah that would be shitty. I think we're all,
00:42:01.840 no one wants to live in a dictatorship. And so that was kind of one of the things where I was like,
00:42:05.360 yeah, I mean that's a problem. If you are, if you can point to something right now and say this is
00:42:12.240 a sign of a dictatorship and I'm getting really worried about it, you know, like, I don't know.
00:42:17.280 Some country is like sliding into some sort of dictatorship then yeah that'd be a problem we
00:42:24.080 should worry about and then we'd start dealing with it as best we can. But without that sort of
00:42:29.840 concrete problem to point at aside from the actual dictatorships that are surviving right now,
00:42:36.640 something I'm not sure what to do with that information, right? It's kind of again one of
00:42:39.760 those vacuous things where everyone can agree that's a problem that would be shitty. Obviously we
00:42:43.440 should kind of be on guard for that but it's quite unclear what to do with that. It's not really
00:42:49.120 going to change my day to day life, right? Yeah, yeah, yeah. So you were going through the
00:42:52.960 structure of the book and I thought that was super useful so I derailed it slightly by asking
00:42:56.080 okay, but perhaps you want to finish the structure of the book. Right, so part two trajectory changes,
00:43:00.640 part three safeguarding civilization and then okay, we should go at this a bit. So part four is
00:43:09.040 an introduction to population ethics. Fucking population ethics, I can't stand population ethics.
00:43:16.400 Population ethics is the study of how to compare different populations of abstract people basically.
00:43:24.720 So these are the classic questions of like, is it good to bring like a life of positive utility
00:43:32.880 into existence? How do I compare a population of people with a population of 100, 1000 people
00:43:41.200 with utility 75 against a population of a billion people with utility 45 and stuff, right?
00:43:48.000 And this is Derek Parfett's like the one who created this field. Yeah, he created this field.
00:43:53.040 And would you say it's a field or is it more just like an area? I don't know. I've been meaning to
00:43:59.520 look at textbook on population ethics and see one like how well founded, how many people are
00:44:07.920 studying population ethics? Like, I don't know, like because the part that really bugs me here is
00:44:13.120 is in the book and in interviews, McCaskill will say like any view you have on these questions
00:44:19.920 leads you to paradoxes, right? So he'll say like, but then what that means, I think we actually
00:44:25.680 talked about this in like one of the earliest episodes, but what that means in technical terms
00:44:30.000 is if you are willing to assign one number that captures a person's utility and you're willing to
00:44:36.640 aggregate that utility over everyone living, right? So you're a lot, basically to sum up the utility
00:44:43.120 of a bunch of people living in a world and say that this world has utility A, if you allow sort
00:44:48.640 of like moral mathematics like that, then you run into problems of like basic arithmetic,
00:44:53.280 which is which is quite clear. Basic, leading leading to conclusions that people find
00:44:57.840 repugnant. That's the yeah, exactly. Exactly. To end up preferring like, you end up preferring
00:45:03.360 a world, a huge world with tons of people where everyone's like just above misery to a world
00:45:09.440 full of a much smaller world full of like very, very happy people. In some sense, it's true to say
00:45:14.560 that is like an impossibility result, he'll call it, but it leaves out the axiom that you have to
00:45:22.000 be willing to like assign a number. Exactly. Like a number that captures the utility of an entire
00:45:28.880 world. And so I'm not sure where to go with this exactly. Like we could actually talk about
00:45:33.840 population ethics, but even in service of the whole book, I was quite unclear about what this
00:45:37.200 chapter was doing. Like to my eye, like, I don't think you really have to argue that the world
00:45:44.080 ending tomorrow is a bad thing. Maybe there's a few people like David Benatar, who would think
00:45:48.560 it's a great thing, but I don't think they are reading this book anyway. And so I was kind of
00:45:52.800 unclear what that chapter was about, except for like maybe signaling some philosophical
00:45:57.360 bona fides or something. Yeah, like, I totally agree. And it's like, for me, when I read about
00:46:03.360 the repugnant conclusion is just so obvious to me that it's like we can't reduce the well-being
00:46:09.280 of a human to a single number and compare them. And so I just, that's the axiom that I'm
00:46:15.360 saying I want to get rid of this one. Let's not try to compare my well-being against your well-being
00:46:20.240 because it can't be done. And I view the Republican conclusion is just showing that it can't be done.
00:46:26.640 But philosophers view this as being the focus of PhD dissertations and stuff because
00:46:31.760 for some reason they don't want to stop this game of assigning you till points to human beings.
00:46:37.760 Well, I think what they would argue is like, well, from, you know, they would say,
00:46:41.280 this is just an abstraction of what we have to do in like public policy anyway, for example,
00:46:46.080 right? Like there we have to compare the effect of certain policy decisions on groups of people.
00:46:51.680 One could benefit and like a larger group of people to a lesser extent, one could benefit
00:46:56.160 a smaller number of people to a greater extent. How do we think about those kind of trade-offs
00:47:00.000 and then population ethics just emerges from there. But it can't be divorced from the concrete
00:47:06.080 problem. Exactly. I agree that you have to make these kind of trade-offs. And for certain
00:47:12.080 cases, it's a good heuristic to simplify people to single well-being points. But when you try to
00:47:21.040 apply this to every problem and divorce it from real world situations, then you need to
00:47:26.960 absurdize. And so, okay, we can't do that. And also in public policy, you don't have the option
00:47:33.920 of adding a population of 150 billion, barely happy people. But you can't just make up people
00:47:40.480 out of the void like this. When you're talking about real people's, realize sometimes you do
00:47:45.360 have to make simplifying assumptions. And that's only to make progress on real concrete, tangible,
00:47:50.480 specific problems. In terms of what it was doing in the service of the book, I guess it's because
00:47:57.680 we're talking about the well-being of a potentially infinite future population. And so, the philosophical
00:48:05.040 area that deals with this is population ethics. But I honestly, I just read part of it and skipped
00:48:10.240 it because it just seemed to be completely unnecessary for what he was trying to say. So,
00:48:14.880 I couldn't give you a concrete answer for why that chapter was there.
00:48:18.720 I think maybe the distinction between someone like you and someone who finds that chapter very
00:48:25.040 necessary is an interesting focus on consistency. So, I think someone would want to say,
00:48:33.280 I don't want to have beliefs in the public policy realm unless they can be extrapolated to every
00:48:40.480 possible abstract world, lest I realize my beliefs are being inconsistent in some way.
00:48:46.240 And this is also interesting related to Dutch books, arguments and stuff, arguments in
00:48:50.720 favor of Bayesianism. There, it's all about at every moment in time you'll have consistent
00:48:55.120 beliefs. You want to bet 51% on something and then 52% on its opposite and then people are going to
00:49:02.160 be able to make bets against you. Yeah, I don't know. There's some focus on... Obviously,
00:49:07.120 having consistent beliefs is a good thing. But there's something with you take this to the limit.
00:49:15.280 And yeah, I'm not sure exactly what's going on there.
00:49:18.240 I have some thoughts there. So, the focus on consistency, I think it's moral philosophers
00:49:23.360 are trying to do what mathematicians do, which is come up with a perfect framework. They have
00:49:28.800 one framework, call it utilitarianism and they have another framework, call it consequentialism.
00:49:33.040 And they're trying to find a framework which can answer all moral questions no matter how
00:49:39.520 difficult or complicated. And they'll never be able to do this. And so, what happens is they
00:49:44.560 come up with some perfect framework and then everyone finds counter examples to show that
00:49:48.880 this leads to moral scenarios which violate our moral intuitions and everyone throws up their
00:49:54.160 hands. The person who's made the most amount of sense on moral philosophy for me is Noah Smith,
00:50:00.640 who in one podcast interview, he said, "I don't have a framework, a moral framework,
00:50:05.520 what have our basic moral intuitions that sometimes can map onto some frameworks occasionally." But
00:50:12.000 the main thing that I use are my intuitions and then I use these frameworks to kind of help me
00:50:16.560 think through corner cases a little bit. But I think that is consistent because it's not
00:50:21.280 attempting to come up with a perfect framework. This is recognizing that so much of our moral
00:50:26.480 decision making is based on these evolutionary pre-programmed intuitions we have. And that's
00:50:32.080 kind of what I want to say too. I have certain intuitions about the morality of slavery and I
00:50:37.200 have other intuitions about the morality of eating meat. And for whatever reason, I don't wake up
00:50:43.360 every morning thinking too much about the suffering of chickens and cows. Other people do and power
00:50:49.280 to them. But I'm riding on my moral intuitions here. And so, I think that that's not an inconsistent
00:50:56.240 worldview. It's more just an acknowledgement that this is what most people are doing most of the
00:50:59.680 time. And frameworks can sometimes help to think through the corner cases occasionally. But
00:51:06.960 we shouldn't seek to find a perfect framework. And this is foundationalism to use a Deutsche
00:51:12.800 Interm. But it's just seeking this perfect thing that can answer all future questions is a futile
00:51:18.720 quest. And instead, we should try to reason through our moral intuitions and use these frameworks
00:51:23.120 occasionally when they're helpful. But in all instances, we have to just go case by case.
00:51:27.360 In some sense, I don't know. Did any of that land?
00:51:31.040 Yeah, I think so. I mean, I'm trying to think what some more sympathetic to population ethics
00:51:36.160 would say. And I think they would say, like, well, formalizing things lets you spot
00:51:40.080 errors in your reasoning or gaps in your intuition or like conflicts in your intuition, right? Like
00:51:46.160 you can't simultaneously think, I don't know, that like, bestiality is wrong, but like,
00:51:51.920 for the reason because it like violates in animals, what sort of sovereignty or something?
00:51:58.720 Yeah, yeah, but yeah, exactly. Like, by the way, autonomy, by the way, there we go,
00:52:02.240 violates their autonomy. And then simultaneous to be okay with like killing them for food.
00:52:06.560 Yeah. And so I guess that's not a situation in which you're like putting numbers on things.
00:52:13.920 You're just like making an argument that these two are inconsistent. But I think what
00:52:17.280 people are trying to do when they're putting numbers on things is abstract away the details
00:52:22.800 that you can see if you hold to completely incompatible views. Then there's a question of like, well,
00:52:28.720 once you've done the abstraction, if you find any compatibility, is that actually telling you
00:52:33.520 anything useful about the world? I think you would say no, and they would say yes. And maybe
00:52:37.200 that's where that's what the difference is. Like, what's the, what's the actual utility?
00:52:40.640 I wouldn't necessarily say no. It wouldn't necessarily say no. Like, if anything, the
00:52:45.200 bestiality and vegetarian example makes me more pro bestiality than I would have been before.
00:52:51.760 So I think you're pointing out, I think you're like pointing out contradictions. Excuse me,
00:52:58.160 I think pointing out contradictions is very important. And there are many times when our
00:53:02.240 moral intuitions do need to be changed. So I'm not saying moral intuitions are always right by no
00:53:06.480 means. I'm more saying moral intuitions are my starting point, and then they need to be kind of
00:53:13.120 buffeted and constrained and directed by kind of moral argument. But that's not the same as saying.
00:53:20.800 I seek a perfect moral framework that can answer all problems between now and the end of time.
00:53:27.920 And then to throw out my hands in despair when I find counter examples to these frameworks.
00:53:34.880 Because like it's so often a Cascale will talk about how like expected value theory is
00:53:40.160 lots of problems with it, but it's the best tool we have. And it's like the seeking a perfect
00:53:47.280 framework. And so it's like, it was not expected value reasoning that showed me a different
00:53:50.640 framework that has fewer problems rather than just trying to overcome the need to have a perfect
00:53:56.400 question answering framework. It's like they want like a fountain of wisdom that they can just ask
00:54:00.640 questions to and have it tell them what to do in all these moral circumstances when really we
00:54:07.360 should just recognize you have to go circumstance by circumstance and not try to find the perfect
00:54:13.120 framework. Yeah, yeah. I mean, yeah, okay, I guess we've talked about expected value now.
00:54:16.640 Yeah, but yeah, I totally agree. Like the difference between public policy and population ethics is
00:54:24.800 you're constrained by the actual policies that you're considering. And so you're not in this
00:54:28.720 abstracted world with utilities. And it's like very specific to the prominent hand. And also,
00:54:34.160 in public policy, you can also always come up with new options. Whereas like the population
00:54:38.480 ethics is very sort of game theoretic in nature where like you have a set of options in front of
00:54:42.880 you. And that's all you're allowed to do. Because you've abstracted yourself into this mathematical
00:54:47.120 world, obviously you can't just like create new options. You can't say, Oh, maybe this third policy
00:54:51.600 is better. And so which of course is how real life works. And then I well, that's the, that's
00:54:58.240 almost the last chapter of the lat the last chapter is is taking action, which is tiny
00:55:02.800 clarification. So the last part, so the book is split into five parts of the source and chapters.
00:55:07.280 Yeah. So the last part part five consists of chapter 10. Yeah, part five is, is yeah,
00:55:11.600 solely chapter 10, which is just about taking action, which we talked a little bit about earlier,
00:55:18.080 like he talks about what people can do, which in the abstract is quite vacuous. But then gives
00:55:23.520 example of people who are working on problems. But like we said earlier to work on this problem,
00:55:28.000 you don't need to adopt a dog long termism. And so I found that chapter to be quite underwhelming.
00:55:33.200 And even he himself, like I was talking with a Tim Ferriss podcast earlier on, and on there Tim
00:55:37.760 Ferriss kind of presses him and says like, like, you know, where should people donate? Like what,
00:55:44.080 what should we do? Like what are some, what are some wins, some concrete wins from long termism?
00:55:48.080 And McCaskell gives a few examples. He gives an example of, of far UVC. So this is like a type of
00:55:54.720 lighting, I guess, that treats rooms for certain bacteria, right? So this,
00:56:02.160 I thought again, was a great example of like, totally, totally,
00:56:05.680 the craziest. And then he says, early detection of new pathogens, which is, which is related.
00:56:11.520 Also, like obviously, fuck yeah, better PPE equipment for people working with dangerous
00:56:17.840 chemicals, and then some technical AI safety stuff, which again, is mostly cashed out in terms of
00:56:23.360 just giving money to PhD students who are doing classic ML research, just focused on on robustness
00:56:29.280 and whatnot. No, I think that's all I got. I mean, this so that yeah, taking action is,
00:56:32.800 is the last part of the book, it's good, but doesn't seem to require long termism. And so,
00:56:39.680 like one thing I couldn't help while reading, help think while reading the book was like,
00:56:43.280 presumably if you introduce like a new view, right? So I was to introduce like a new scientific theory,
00:56:49.120 and it agreed with general relativity 99.5% of the time. That's not a benefit of the new theory,
00:57:01.920 right? You wouldn't take that as evidence that the theory was true. What you look at is the 0.5%
00:57:06.560 of times where deviates from general relativity and say it doesn't make better predictions in that
00:57:12.080 area, right? So sort of like the difference, the delta between these two theories. And I was trying
00:57:17.200 to do that a bit with long termism, like in so far as it doesn't line up with what I think most people
00:57:24.800 would consider to be totally valid problems. What more is it recommending beyond that? And the answer
00:57:31.680 seems to be relatively vague things and then some AI stuff. And so considering that delta,
00:57:39.440 it's not the most impressive thing ever. Yeah, but I also think that the consequence of this kind of
00:57:46.400 book, and I include Toby Ward, is that in five years, and this is a falsifiable prediction,
00:57:52.400 that in five years long term, this research is going to be dominating the EA giving space,
00:57:57.760 then say global poverty research, which is what McCaskill's first book was focused on. And that's
00:58:03.520 going to hurt real people where there's a finite amount of malarial dead nets, etc. And if funding
00:58:11.280 for these kind of concrete problems is being siphoned away and instead we're getting AI
00:58:16.720 research, AI safety research, and long termism research, etc. That's going to hurt real people.
00:58:21.120 And I think that this book definitely, I think it's contributing to that while being able to deny it
00:58:28.000 and that's maybe an unfair critique, but that's my honest impression that the book is so
00:58:33.280 anodyne, but it's like a gateway drug to the more intense stuff, which is on the internet and which
00:58:39.440 McCaskill's written himself. And so it's just a strange kind of superposition between dangerous
00:58:47.440 and banal, I guess. So I read the book once, or I listened to it, and then prepping for it,
00:58:53.120 I read a few more chapters. I've never had kind of less interesting things to say
00:58:57.600 about a book I've actually dedicated time to thinking about. I almost feel like embarrassed
00:59:03.120 that I don't have that much to say, but I truly was just kind of at a loss for what to think.
00:59:08.960 Like I would, yeah, sometimes I'd be like, yeah, I mean, taking the future seriously,
00:59:12.720 that's a good idea. Phemons are cool, I want to see us flourish. That's great. Other times,
00:59:17.520 I'm like, you're talking about AI replicating, taking over the world, like, you know, this is
00:59:22.800 crazy. And so it's just such a bizarre thing to reach. And I can tell, like, am I just too deep in
00:59:32.560 the historicals? It's like, there's nothing terribly new, but it was just like a boring version of
00:59:40.480 stuff we've already talked about before. So it wasn't like, like, there's a couple things like the
00:59:47.440 vaccines on the free market. I thought that was a great idea. Hadn't heard that, hadn't thought
00:59:51.840 about that before. Obviously, like, duh. But most of it is like, okay, here's the expected value stuff.
00:59:57.200 Here's the AI stuff. Like the value logic. Check this on that personally. Yeah, there we go.
01:00:01.680 There we go. But yeah, in general, I agree. Time to move on. It's a weird space, man. It's
01:00:07.600 weird. I mean, even like Sam Harris, like, yeah, like, have him on, though, like, you'll look,
01:00:10.880 they'll, like, get excited about the ideas. But I have to wonder, like, is this influencing Sam's
01:00:17.040 actual gift? Like, I know he donates like a lot of money, you know, with this foundation and stuff.
01:00:20.800 The last I checked, he donated a bunch of stuff to like Afghan refugees or something. Yeah,
01:00:25.200 I don't think that's a long term is caused. Like, yeah, I'm like, is this act like,
01:00:29.280 there's some like, is this just intellectual cotton candy or something that we're just like,
01:00:33.760 eating up? Because it's like, it's like, yeah, fun. Yeah. Like, there was a huge splash when
01:00:40.320 McCaskill went on the podcast circuit. But like, then it was gone radio silent. Totally.
01:00:45.200 And maybe it's just going to be radio silent forever. Like, what is he actually advocated
01:00:50.960 that we do differently? Exactly. But donate more to like fossil fuel replacements and like, okay,
01:00:56.320 great, man. I agree we should do we should do that. But yeah. Yeah. I also don't want
01:01:02.480 civilization to collapse. Like, I agree. Now what's one concrete thing that I can do to avoid that?
01:01:08.640 Like, yeah, no, exactly. Yeah. You learn more and donate. It's like, yeah, learn and donate.
01:01:13.920 Goodness. I'm not trying much more to say, to be honest.
01:01:18.880 Me too. I think we've been going for an hour and a half. So we definitely got to got a bunch
01:01:24.320 out there. So I don't know what we're going to do next time. I think we're having some guests on
01:01:27.520 at this point. But we should be careful not to announce them because you never know.
01:01:33.280 Yeah, never know. But hopefully that'll happen soon. And this will probably be the last thing
01:01:37.360 we say about long term is this like, my cast garage is another book that.
01:01:40.240 Yeah. I hope so at least. Yeah. I want to move on. Yeah. Yeah. But cool then. Well, this is fun.
