WEBVTT - Summer Camp for Poker Players (WSOP preview) 0:00:15.604 --> 0:00:26.484 Pushkin. Welcome back to Risky Business, a show about making 0:00:26.564 --> 0:00:30.284 better decisions. I'm Maria Kannakova and I'm Nate Silver. 0:00:30.524 --> 0:00:32.524 We're going to kick the show off with a lighthearted topic, 0:00:32.724 --> 0:00:35.364 what is the chance that artificial intelligence will destroy every 0:00:35.444 --> 0:00:38.004 last human being and all animals speaking us on Earth? 0:00:38.164 --> 0:00:41.244 Yeah, Nate, that is a very very lighthearted area of discussion. 0:00:41.284 --> 0:00:43.164 I'm looking forward to it. Then we're going to move 0:00:43.204 --> 0:00:47.444 on to the presidential debates and how important they are 0:00:47.484 --> 0:00:48.884 and what they mean for the elections. 0:00:49.204 --> 0:00:51.524 And then for our serious segment, the World Serious Poker 0:00:51.524 --> 0:00:53.644 begins next week in Las Vegas. We're both going to 0:00:53.684 --> 0:00:56.044 be there. I already got my wired transfer to whatever 0:00:56.124 --> 0:00:59.284 new payment provider the world series uses, So we'll do 0:00:59.324 --> 0:01:03.884 a preview of the series on today's show. 0:01:05.364 --> 0:01:08.124 All right, Nate. So you know, I did not think 0:01:08.164 --> 0:01:10.364 that twenty four was going to be the year that 0:01:10.444 --> 0:01:13.364 I would be talking about Scarlett Johansson in reference to 0:01:14.084 --> 0:01:17.484 artificial intelligence and open Ai. But here we are because 0:01:17.924 --> 0:01:20.964 open Ai, which is one of the biggest companies in 0:01:21.004 --> 0:01:26.364 AI research and the fourth behind chat GPT, just unveiled 0:01:26.604 --> 0:01:30.964 the new voice of their newest AI assistant, and that 0:01:31.084 --> 0:01:36.044 voice sounds eerily like Scarlet Johansson, which I guess is 0:01:36.084 --> 0:01:39.244 not surprising since there was this movie Her where Scarlet 0:01:39.524 --> 0:01:44.044 played this AI assistant who becomes very very human like 0:01:44.084 --> 0:01:48.364 and engenders very human like emotions in her co stars. Right, 0:01:48.484 --> 0:01:52.284 and then we see this announcement from Sam Altman, who 0:01:52.364 --> 0:01:56.164 is the founder of Open Ai, about their new voice 0:01:56.404 --> 0:01:59.564 called Sky, and I want to go back to the 0:01:59.644 --> 0:02:01.484 name of the voice in a second when we talk 0:02:01.524 --> 0:02:06.124 about the more substantive issues in open AI, but he 0:02:06.244 --> 0:02:10.404 references her when he tweets about it, and the voice 0:02:10.444 --> 0:02:14.124 sounds early like Scarlett, and she issues a statement that says, hey, guys, 0:02:14.164 --> 0:02:16.644 you know that is not me. I did not agree 0:02:16.644 --> 0:02:20.044 with this, and we're getting some lawyers involved. And then 0:02:20.084 --> 0:02:21.804 the voice is very quickly taken down. 0:02:22.084 --> 0:02:25.124 Yeah, because he had tried to make a proper deal 0:02:25.164 --> 0:02:27.684 with her. She had said nah, cold feet. And then 0:02:27.684 --> 0:02:29.644 they come out anyway with this product that sounds a 0:02:29.644 --> 0:02:32.244 lot like her. So it's I mean, look, this comes 0:02:32.324 --> 0:02:34.444 up in my book a little bit, which I'm always pitching, 0:02:34.444 --> 0:02:38.284 I guess on the show, But I think Silicon Valley 0:02:38.444 --> 0:02:42.764 does not understand politics very well. Right, you have this 0:02:42.884 --> 0:02:46.204 very popular actress people left, right and center. Literally my 0:02:46.244 --> 0:02:50.004 Twitter feed, we're pretty upset for how brazen Sam Altman 0:02:50.044 --> 0:02:52.844 and openay look about this, and that matters. I mean, 0:02:52.844 --> 0:02:56.444 Congress has a lot of power to regulate AI. There 0:02:56.484 --> 0:02:59.604 is actually a broad bipartisan support for more a regulation. 0:02:59.684 --> 0:03:03.124 Theydy the AI might be dangerous in different ways, maybe 0:03:03.164 --> 0:03:05.324 not quite in the way that the people will talk 0:03:05.364 --> 0:03:07.324 about in a moment. The doomers think about it, but 0:03:07.404 --> 0:03:09.524 people think it I'll take their jobs, and people think 0:03:09.524 --> 0:03:13.924 it will make misinformation worse, all types of problems algrethmc bias. 0:03:14.004 --> 0:03:16.404 So people have to be people have to be careful. 0:03:16.444 --> 0:03:19.964 I mean, look, I kind of hate politics too, but 0:03:20.084 --> 0:03:22.684 unfortunately they have a lot of regulatory powers. You have 0:03:22.804 --> 0:03:24.324 to you have to put up with some of that 0:03:24.924 --> 0:03:27.924 stupidity and not egging on by doing something twice as 0:03:27.924 --> 0:03:28.684 stupid yourself. 0:03:29.564 --> 0:03:31.964 That's absolutely right, and Nate, I am going to show 0:03:32.004 --> 0:03:34.404 the world how much you're already rubbing off on me. 0:03:34.564 --> 0:03:37.604 In terms of using a sports analogy, this is akin 0:03:37.684 --> 0:03:40.644 to what we would call an unforced error, right, like 0:03:40.684 --> 0:03:44.244 this did not have to happen, And this is in 0:03:44.324 --> 0:03:48.884 a week where open ai desperately, desperately needs some positive publicity, 0:03:49.084 --> 0:03:51.444 because what we really want to be talking about isn't 0:03:51.444 --> 0:03:54.444 Scarlett Johansson, as gorgeous as she is and as amazing 0:03:54.524 --> 0:03:56.764 as her voice is and as much as we love her. 0:03:57.164 --> 0:03:59.924 But the other really really big news from open AI 0:04:00.044 --> 0:04:03.684 this week, which is that one of their co founders 0:04:03.804 --> 0:04:08.684 and chief scientist Ilia Sutskib has resigned. And he is 0:04:08.804 --> 0:04:10.404 not the only one. 0:04:10.604 --> 0:04:14.084 Yeah, I'm not going to pretend I know all these names. 0:04:14.124 --> 0:04:16.364 It's not like their NBA players or how to pronounce them, 0:04:16.364 --> 0:04:21.444 but yeah, unlike a Leopold Aschenbrenner, Pabel Isma Lov, all 0:04:21.484 --> 0:04:23.644 these people that, again I don't know personally, but like 0:04:24.084 --> 0:04:28.444 credible people think are safety oriented open air researchers, and 0:04:29.084 --> 0:04:31.084 a half dozen of them have left in the past 0:04:31.404 --> 0:04:34.204 half year or so. It's a pretty bad sign for 0:04:34.244 --> 0:04:36.724 a company that's marketing itself as once having been a 0:04:36.764 --> 0:04:40.564 nonprofit that was concerned with doing AI the right way. 0:04:40.844 --> 0:04:43.924 Is that people in that coalition are are now leaving 0:04:44.244 --> 0:04:45.404 in some significant number. 0:04:45.724 --> 0:04:48.364 So the reason that people, including the two of us, 0:04:48.604 --> 0:04:52.284 are so incredibly worried about this or thinking about this 0:04:52.324 --> 0:04:54.684 and trying to figure out should we be worried about 0:04:54.684 --> 0:04:58.244 this is because this was the super Alignment team. This 0:04:58.444 --> 0:05:02.644 was the team responsible for making sure that AI doesn't 0:05:02.644 --> 0:05:04.924 destroy the world. And by the way, I'm just going 0:05:05.004 --> 0:05:08.364 to harken back to our Scarlett Johansson discussion for one second. 0:05:08.684 --> 0:05:12.244 They called the sky and like, were they trying to 0:05:12.284 --> 0:05:14.804 refer to Skynet as well? Like, were they just trying 0:05:14.844 --> 0:05:17.484 to put a terminator in the reference right in there, 0:05:17.564 --> 0:05:20.604 saying like, hey guys, we're unleashing something that might destroy 0:05:20.684 --> 0:05:21.164 the world. 0:05:21.684 --> 0:05:25.084 So I think there's a detached irony. Like, look, I 0:05:25.124 --> 0:05:29.244 think Sam Altman's Twitter feed is kind of funny, right, 0:05:29.284 --> 0:05:31.684 And it's kind of funny, and he's combative and he battles. 0:05:31.964 --> 0:05:34.404 He's a poster which if you're like me, who the 0:05:34.404 --> 0:05:38.044 fuck cares? If you're running open AI, then maybe you 0:05:38.164 --> 0:05:41.444 have to be a little bit more buttoned down. And 0:05:41.484 --> 0:05:44.324 they're kind of aloof right. San Francisco and the Silicon 0:05:44.364 --> 0:05:46.404 Valley have always been kind of detached from the rest 0:05:46.804 --> 0:05:49.964 of the country. They're all very smart, all very rich. 0:05:50.004 --> 0:05:52.084 They're also very you know, male for the most part, 0:05:52.124 --> 0:05:55.724 They're probably in a bubble in certain ways. And I 0:05:55.764 --> 0:05:58.964 don't know, like I said before, I might think Congress 0:05:59.004 --> 0:06:02.884 is stupid. I kind of sympathized more with the Silicon 0:06:02.964 --> 0:06:05.844 Valley side and the argument in many ways than the 0:06:05.884 --> 0:06:08.164 DC side. I think they're more innovative. I think technology 0:06:08.204 --> 0:06:10.724 has been good for society. We're all, but like I 0:06:10.724 --> 0:06:13.124 think there is a lot of arrogance there that might 0:06:13.164 --> 0:06:16.244 not end well. You look at what Congress did to Napster, 0:06:16.364 --> 0:06:18.524 for example, or other businesses that we're seen as having 0:06:18.564 --> 0:06:21.324 gotten out of line. And there is a pretty broad 0:06:21.404 --> 0:06:26.644 bipartisan sense of weariness about AI for different reasons. Democrats 0:06:26.724 --> 0:06:29.564 might be concerned about algorithmic bias. Both parties might be 0:06:29.564 --> 0:06:32.764 concerned about jobs being taken over. But the political climate 0:06:32.804 --> 0:06:35.484 on this could shift quite quickly, and that would be 0:06:35.564 --> 0:06:37.124 very bad for Sam Maltman's business. 0:06:37.444 --> 0:06:41.364 So let's talk about this concept known as P doom. 0:06:41.884 --> 0:06:44.324 So do you want to just first to say, like, 0:06:44.404 --> 0:06:45.604 what is P doom? 0:06:45.684 --> 0:06:48.724 I pronounce it more like pay doom because looking myself, 0:06:48.724 --> 0:06:51.804 like you look like a urinary tract problem or something. 0:06:52.364 --> 0:06:53.244 All right, P doom. 0:06:54.644 --> 0:06:57.964 So pay doom is a shorthand for probability of doom, 0:06:58.044 --> 0:07:01.724 which is a shorthand for what's the chance that AI 0:07:02.044 --> 0:07:06.884 destroys civilization or substantially impair civilization, which is actually like 0:07:06.884 --> 0:07:11.884 a pretty big distinction. People have different definitions. Eliezer Yudkowski, 0:07:12.044 --> 0:07:15.324 one of the most prominent doomers, meaning people who are 0:07:15.404 --> 0:07:19.444 very worried about AI risk, defines it literally is all 0:07:19.564 --> 0:07:24.964 human beings and animal species perishing, right, every single living being. 0:07:25.004 --> 0:07:28.204 He has said in a Time magazine article last year. 0:07:28.284 --> 0:07:32.004 I think so he's being very literal. Someone like Ajaia Kotra, 0:07:32.044 --> 0:07:35.284 who is an AI researcher, uses some more inclusive definitions. 0:07:35.284 --> 0:07:39.644 I mean things where humanity loses agency if, for example, 0:07:39.684 --> 0:07:42.804 we are no longer really in charge, maybe our lives 0:07:42.804 --> 0:07:47.364 are okay, or some very rich people in Silicon Valley 0:07:47.404 --> 0:07:50.004 have some agency. But if we kind of hand over 0:07:50.044 --> 0:07:54.644 the keys to machines and it becomes kind of dystopian, 0:07:54.684 --> 0:07:57.444 and that could qualify as p doom as well. That's 0:07:57.484 --> 0:08:00.084 a big distinction. And people also have very wide ranges, 0:08:00.124 --> 0:08:02.484 which when they cite this chance you can get people 0:08:02.484 --> 0:08:05.764 who think it's you know, ninety nine percent in this 0:08:06.564 --> 0:08:11.004 coming from Eliezer Yukowski. The consensus, and again if you 0:08:11.284 --> 0:08:13.284 kind of put aside the fact that the different definitions 0:08:13.324 --> 0:08:16.124 of it is probably in the range of five to 0:08:16.124 --> 0:08:21.364 ten percent. If you survey AI experts, AI researchers give 0:08:21.404 --> 0:08:24.324 them different definitions, they kind of gravitate toward that five 0:08:24.404 --> 0:08:27.364 to ten percent range. Some people in the safety community 0:08:27.444 --> 0:08:30.364 are higher, maybe more the range of ten or twenty percent. 0:08:30.684 --> 0:08:32.644 And by the way, let's not be too abstract here. 0:08:33.124 --> 0:08:35.204 If it's true that there is like a ten percent 0:08:35.324 --> 0:08:38.524 chance or even frankly a two percent chance of humanity 0:08:38.524 --> 0:08:41.324 going extinct from AI in the next one hundred years, 0:08:41.364 --> 0:08:44.764 this is a tremendously important problem. Yeah, so figuring out 0:08:44.804 --> 0:08:48.004 how reliable those estimates are is an important challenge that 0:08:48.084 --> 0:08:49.884 the show will tackle from time to time. 0:08:50.004 --> 0:08:52.484 Yeah, and I just took round it even more. Let 0:08:52.524 --> 0:08:54.444 me kind of talk you through a problem that has 0:08:54.564 --> 0:08:57.244 nothing to do with AI that I had to deal 0:08:57.284 --> 0:09:01.004 with quite recently, where you kind of start realizing what 0:09:01.044 --> 0:09:03.164 it means, what five percent means, right, what ten percent 0:09:03.164 --> 0:09:06.004 actually means? So I had a leak in the shower, right, 0:09:06.044 --> 0:09:08.484 you know, like that just completely really annoying, like drip 0:09:08.604 --> 0:09:10.804 drip drip from the show our head, and so you know, 0:09:10.844 --> 0:09:12.684 you look it up online, you try to figure out, 0:09:12.724 --> 0:09:15.244 like what's going on? How do we fix this? And 0:09:15.684 --> 0:09:18.084 turns out that you know, this is a gasket problem, 0:09:18.484 --> 0:09:23.684 and according to YouTube, it's super easy to do it yourself, right, 0:09:23.724 --> 0:09:26.684 you just you know, you take it off, you replace it, 0:09:27.084 --> 0:09:30.684 put it back on, and voila, everything's done. So this 0:09:30.764 --> 0:09:34.244 happened in a big apartment building and I had to 0:09:34.284 --> 0:09:35.964 reach out to the management and be like, hey, how 0:09:35.964 --> 0:09:38.044 do I turn off the water in my apartment? Because 0:09:38.084 --> 0:09:40.604 that's step one, right, you turn off the water flow. 0:09:40.964 --> 0:09:44.124 And so they ended up sending someone and she was like, 0:09:44.244 --> 0:09:47.764 are you insane? You do realize that you could end 0:09:47.844 --> 0:09:50.404 up flooding the apartment below and if you do that, 0:09:50.684 --> 0:09:53.324 you're going to be liable for all of those repairs, 0:09:53.324 --> 0:09:57.444 which could go into the tens of thousands of dollars. Now, 0:09:57.484 --> 0:10:00.484 of course, like normal people walking around aren't going to 0:10:00.524 --> 0:10:02.404 give me a percentage and be like, you know, Maria, 0:10:02.444 --> 0:10:07.044 your p flood is approximately fifteen percent here, But she 0:10:07.124 --> 0:10:10.404 made it seem like it was certainly great than five percent, 0:10:10.844 --> 0:10:14.324 and maybe you know, greater than ten percent, especially because 0:10:14.364 --> 0:10:17.844 the building has very hard water. And so there's isn't. 0:10:17.604 --> 0:10:19.244 A real thing that sounds like some bullshit. 0:10:19.804 --> 0:10:22.764 It is a real thing. Date, It's a very real thing. 0:10:22.844 --> 0:10:25.644 As a female who washes my hair all the time. 0:10:25.964 --> 0:10:28.724 Let me tell you, as someone who doesn't have nearly 0:10:28.724 --> 0:10:30.764 as much hair as I do. Sorry, Nate, but that's 0:10:30.844 --> 0:10:34.404 just Them's the facts. God, that hard water is a 0:10:34.564 --> 0:10:37.484 very very real thing and has a very different effect 0:10:37.564 --> 0:10:40.444 on your face, on your hair, on everything, and apparently 0:10:40.684 --> 0:10:43.524 on the level of corrosion when fixing a shower on 0:10:43.564 --> 0:10:46.804 your pea flood. And so here I am trying to 0:10:46.804 --> 0:10:49.684 figure out, okay, well what do I do, like, how 0:10:49.684 --> 0:10:52.244 do I actually decide this? I ended up asking my dad, 0:10:52.484 --> 0:10:56.004 because you know, daddy knows best, and he had a 0:10:56.124 --> 0:10:58.764 very good way of thinking about it. He said, do 0:10:58.924 --> 0:11:01.924 it if there's no visible sign of corrosion and it's 0:11:02.004 --> 0:11:04.324 super easy to take it out, don't do it otherwise. 0:11:04.604 --> 0:11:06.524 And so I, you know, opened it and there was 0:11:06.604 --> 0:11:09.164 tons of corrosion. It was very clear that I could 0:11:09.164 --> 0:11:12.084 not take it out easily. Hired a plumber, got it done. 0:11:12.324 --> 0:11:16.324 But it made me realize that those tiny chances, when 0:11:16.324 --> 0:11:19.524 there's a huge risk involved that can actually translate to 0:11:19.604 --> 0:11:22.044 a lot of money, you have to take that seriously. 0:11:22.244 --> 0:11:25.884 And obviously this is so ridiculously silly, right me destroying 0:11:26.204 --> 0:11:31.244 someone's bathroom below mind versus humanity going extinct. But it 0:11:31.364 --> 0:11:34.244 just illustrates that even if you said, you know, the 0:11:34.244 --> 0:11:36.404 p doom is five percent, and that's not just five 0:11:36.444 --> 0:11:39.644 to ten percent, but five percent or two percent, that 0:11:39.804 --> 0:11:44.364 is absolutely astronomical when the actual cost is the end 0:11:44.364 --> 0:11:44.924 of the world. 0:11:45.884 --> 0:11:48.404 Yeah. In fact, I'm not even sure that it makes 0:11:48.444 --> 0:11:50.684 sense to think of it an expected value terms, because 0:11:50.684 --> 0:11:54.084 expected value is the average or some number of outcomes. 0:11:54.164 --> 0:11:54.404 Right. 0:11:54.524 --> 0:11:56.724 I made a bet on a soccer match the other 0:11:56.804 --> 0:11:59.604 day that you know, I had to win one out 0:11:59.604 --> 0:12:02.324 of six times to be profitable. It did not win, 0:12:02.724 --> 0:12:05.404 but in the long run you would make money from 0:12:05.404 --> 0:12:09.884 that ben Right. If you destroy civilization, however, you don't 0:12:09.884 --> 0:12:11.964 have at no longer, I don't have a view over. 0:12:12.164 --> 0:12:14.684 This is a very very important point. Yes, if I 0:12:14.724 --> 0:12:17.284 destroyed the downstairs apartment, I will be able to pay 0:12:17.284 --> 0:12:19.084 for it. So this is a risk that I can 0:12:19.124 --> 0:12:21.804 think about. If I destroy the world, I'm not going 0:12:21.844 --> 0:12:23.124 to be able to make it all better with a 0:12:23.164 --> 0:12:23.604 band aid. 0:12:23.924 --> 0:12:26.764 And what's very weird about the politics of all of 0:12:26.804 --> 0:12:29.004 this is that the people who are most worried about 0:12:29.044 --> 0:12:32.124 p doom tend to be the people who are building 0:12:32.524 --> 0:12:35.244 the AI systems. Which to me people try to like 0:12:35.524 --> 0:12:37.524 over complicate this and say, well, maybe they have some 0:12:37.564 --> 0:12:40.044 marketing and go like, that's fucking stupid. They're worried about 0:12:40.044 --> 0:12:45.964 it because a it's made progress much faster than people anticipated, 0:12:45.964 --> 0:12:48.324 at least recently. I mean, I think very few researchers 0:12:48.324 --> 0:12:52.524 would have thought that large language models like CHATCHYPT can 0:12:52.924 --> 0:12:55.444 accomplish as much as they do right now. Or they can, 0:12:55.724 --> 0:13:00.124 you know, answer any question with a degree ranging from 0:13:00.444 --> 0:13:04.564 somewhat competent bullshitter to actually very good on many questions. 0:13:04.564 --> 0:13:06.964 They can like solve some mathematical equations, they could do 0:13:07.044 --> 0:13:09.844 some programming, they can draw shitty art, they can make 0:13:09.884 --> 0:13:12.124 mediocre music, right. But like, these are all things that 0:13:12.164 --> 0:13:16.164 would have surprised most experts if you'd ask them five 0:13:16.244 --> 0:13:20.044 years ago. The other thing that worries experts is that 0:13:20.084 --> 0:13:23.244 we don't really know how all of us works. You 0:13:23.324 --> 0:13:26.564 kind of feed a giant, you populate a giant matrix 0:13:26.604 --> 0:13:29.644 of what are called tokens, which are words which are 0:13:29.764 --> 0:13:33.404 kind of compressed into vector space. I'm deliberately using some terminology, 0:13:33.444 --> 0:13:37.244 but like the inner workings of these transformer models are 0:13:37.804 --> 0:13:41.284 not entirely understood. To be fair the inner workings of 0:13:41.284 --> 0:13:43.964 the human brain are not very well understood either. But 0:13:44.084 --> 0:13:47.004 to have a very powerful technology that purports to do 0:13:47.124 --> 0:13:49.844 superhuman things within a span of five or ten or 0:13:49.884 --> 0:13:53.724 twenty years and which we don't understand very well seems 0:13:53.764 --> 0:13:56.484 like kind of self evidently a risk we ought to 0:13:56.524 --> 0:13:58.964 be concerned about. That's kind of like the short Steelman 0:13:59.084 --> 0:14:02.044 version of why it's not ridiculous to be worried about 0:14:02.084 --> 0:14:03.284 about ped doom. 0:14:03.684 --> 0:14:06.164 Well, and I think that an even shorter version is 0:14:06.204 --> 0:14:08.884 something that you said right at the beginning, which is 0:14:08.964 --> 0:14:12.124 that the people who are designing it, that's the same 0:14:12.244 --> 0:14:14.604 subset as the people who are most worried about it, 0:14:14.644 --> 0:14:17.444 and whose p dooms tend to be very high. That 0:14:17.564 --> 0:14:20.724 tells us something, Right, if everyone who was actually working 0:14:20.924 --> 0:14:24.484 in AI said nah, this is fine, like it's all good, 0:14:24.844 --> 0:14:27.924 and we were hearing this from pundits and from politicians 0:14:28.044 --> 0:14:30.364 and from you know, people on the other end of it, 0:14:30.564 --> 0:14:33.124 then we'd say, oh, you know, maybe we should discredit 0:14:33.164 --> 0:14:35.084 some of this. But it's coming from the people who 0:14:35.124 --> 0:14:38.284 are actually doing the designing, who are actually doing the building, 0:14:38.684 --> 0:14:41.284 And so that makes me, as someone who doesn't know 0:14:41.324 --> 0:14:45.444 anything about these black box workings worried, especially because we 0:14:45.524 --> 0:14:48.844 know that other processes that have had black box components 0:14:48.844 --> 0:14:50.764 have gone off the rails and have gone wrong and 0:14:50.764 --> 0:14:53.764 have glitched in the past. I mean, we've had this 0:14:53.844 --> 0:14:57.364 with trading algorithms, right, Like this shit happens all the time. 0:14:57.724 --> 0:15:00.364 Yeah, Look, there are an abundance of scenarios, including that 0:15:00.404 --> 0:15:05.244 people are over confident about AI capabilities and therefore hand 0:15:05.284 --> 0:15:08.484 control of key systems over to AIS and they fuck up. Right, 0:15:08.684 --> 0:15:11.324 I don't see the concerns that kind of DC has 0:15:11.364 --> 0:15:15.084 about AI and the one Silicon Valley does that's necessarily 0:15:16.044 --> 0:15:18.604 mutually exclusive. They could both. You know, there are various 0:15:18.604 --> 0:15:22.444 categories of mild, medium, and terrible things, but there's also 0:15:22.524 --> 0:15:24.804 like a lot of upside in general. I don't know 0:15:24.804 --> 0:15:26.564 if there's some way to quantify this, but the best 0:15:27.124 --> 0:15:31.964 majority of technologies move humanity forward in different ways. Like, 0:15:32.524 --> 0:15:35.364 you know, the human condition is much much, much much 0:15:35.404 --> 0:15:38.444 better than it was centuries ago because of modern technology. 0:15:38.724 --> 0:15:42.124 People who don't know that are basically totally misinformed. And 0:15:42.204 --> 0:15:43.924 like I we'd bet that half of people, even I 0:15:43.924 --> 0:15:46.844 mean our listeners might know, but like the average voter 0:15:46.964 --> 0:15:49.204 probably doesn't know very much about kind of economic history 0:15:49.204 --> 0:15:51.324 and things like that. Technology has been a force for 0:15:51.404 --> 0:15:54.884 good mostly, but we haven't dealt with that many technologies 0:15:54.884 --> 0:15:58.044 that were predicted to have the power to like destroy civilization. 0:15:58.164 --> 0:16:00.684 And by the way, nuclear weapons, you can say, well, 0:16:00.724 --> 0:16:03.284 you know, nuclear weapons have gone okay so far, well 0:16:03.284 --> 0:16:05.964 only since nineteen forty five. If there's like a one 0:16:06.004 --> 0:16:09.164 percent chance risk every year of nuclear weapons been used 0:16:09.164 --> 0:16:11.524 in common, which, by the way, happened in nineteen forty 0:16:11.524 --> 0:16:14.924 five Nagasaki and Hiroshima. You know, seventy eighty years of 0:16:15.004 --> 0:16:17.924 data are not enough to nullify a one in one 0:16:18.004 --> 0:16:18.684 hundred chance. 0:16:18.804 --> 0:16:21.644 So, Nate, I remember someone when I was kind of 0:16:22.004 --> 0:16:24.564 reading about all of the open AI stuff, something that 0:16:24.644 --> 0:16:28.444 really jumped out at me was a CEO who described 0:16:28.924 --> 0:16:32.604 Sam Altman's approach to open AI regulation and open AI 0:16:32.644 --> 0:16:37.884 in general as ready fire aim And that to me 0:16:38.044 --> 0:16:40.724 is a little bit backwards. It doesn't mean we can't fire, 0:16:40.764 --> 0:16:43.284 but we should probably be aiming. And the people who left, 0:16:43.364 --> 0:16:47.244 the people who've resigned the most high profile departures, those 0:16:47.284 --> 0:16:50.724 were our aim people. So Nate, I think we should 0:16:51.204 --> 0:16:53.924 be revisiting our p DO estimates in the coming weeks 0:16:54.044 --> 0:16:56.364 and the coming months to see what happens. I think 0:16:56.404 --> 0:16:58.644 I'm a little bit more pessimistic than you are, and 0:16:58.724 --> 0:17:01.284 I think probably these departures have made both of us 0:17:01.284 --> 0:17:04.884 slightly more pessimistic than we were before. And I would 0:17:04.924 --> 0:17:07.964 just absolutely love to be proven wrong here. This is 0:17:08.004 --> 0:17:10.844 one area where I hope that ourp dooms are all 0:17:10.844 --> 0:17:13.564 way too high. And you know, we can revisit this 0:17:14.044 --> 0:17:17.044 in fifty years and say, ha haha, remember when we 0:17:17.044 --> 0:17:18.244 were all so scared. 0:17:34.164 --> 0:17:34.684 So nate. 0:17:34.804 --> 0:17:37.404 As you know, we are in an election year and 0:17:37.444 --> 0:17:40.564 there were actually some big news for this early on, 0:17:40.924 --> 0:17:44.644 and they had to do with the debates, the presidential debates. 0:17:44.964 --> 0:17:47.844 This has been something on everyone's minds, and it seems 0:17:47.884 --> 0:17:52.524 like Biden and Trump have mostly agreed for their debate 0:17:52.604 --> 0:17:55.564 terms for their first debate, which has now been pushed 0:17:55.684 --> 0:17:59.084 up from the fall to June twenty seventh on CNN. 0:17:59.404 --> 0:18:01.844 So that is like that's in a month, right, that 0:18:01.964 --> 0:18:04.884 is coming up right around the corner, and then there's 0:18:04.924 --> 0:18:09.084 going to be another debate in September. And this is 0:18:09.124 --> 0:18:12.124 big news for a few reasons. So I'd love you 0:18:12.164 --> 0:18:15.204 to talk us through why this is important and why 0:18:15.244 --> 0:18:18.604 we should you know, actually care about this news, about 0:18:18.764 --> 0:18:19.844 the presidential debates. 0:18:20.404 --> 0:18:22.124 So the reason you should care is that we are 0:18:23.524 --> 0:18:25.044 how to put this nicely, You. 0:18:25.004 --> 0:18:26.964 Don't have to put it nicely, Nate. I know we're 0:18:26.964 --> 0:18:28.044 among friends here. 0:18:28.724 --> 0:18:32.124 I think this benefits. This maneuver by Biden, which I'll 0:18:32.124 --> 0:18:35.964 explain in a moment, benefits from people who understand incentives 0:18:35.964 --> 0:18:39.964 and strategy and kind of revealed preference versus stated preference, 0:18:40.004 --> 0:18:44.764 all those fancy terminologies, fancy terms I should say. So 0:18:44.964 --> 0:18:48.404 last week, Biden does a tape segment on Morning Joe. 0:18:48.444 --> 0:18:50.604 By the way, I can't even like actually go live 0:18:51.164 --> 0:18:53.124 on the show. You know, if you think you could 0:18:53.124 --> 0:18:54.844 debate anywhere anytime, you could do a live interview, but 0:18:54.924 --> 0:18:56.924 leave that alone for a second. Does a tape little 0:18:56.964 --> 0:19:00.644 segment saying I challenge you Donald to not one, but 0:19:00.764 --> 0:19:04.404 two debates. Half an hour later, his campaign puts out 0:19:04.404 --> 0:19:06.884 a memo which does confirm that he has challenged Trump 0:19:06.924 --> 0:19:08.724 to two debates, and also says in the third paragraph, 0:19:08.764 --> 0:19:11.604 by the way, we are actually pulling out of the 0:19:11.604 --> 0:19:14.124 Commission on Presidential Debates the three debates that they had 0:19:14.164 --> 0:19:17.764 planned for later this year. And look, the CPD has problems. 0:19:17.764 --> 0:19:20.524 I mean Trump ragged them all the time. In twenty 0:19:20.884 --> 0:19:23.684 and twenty. He also didn't do any GOP primary debates. 0:19:23.884 --> 0:19:28.724 But basically Biden traded three debates after Labor Day for 0:19:29.084 --> 0:19:31.444 one debate after Labor Day and then one on CNN, 0:19:31.524 --> 0:19:33.684 a cable network that will happen in the middle of 0:19:33.764 --> 0:19:35.884 June that probably everyone will forget about by the time 0:19:36.284 --> 0:19:38.444 the conventions happened in July and August, and certainly by 0:19:38.444 --> 0:19:39.964 the time we get to the stretch round of the campaign. 0:19:40.124 --> 0:19:43.844 So what Biden is actually doing here is reducing his 0:19:43.924 --> 0:19:48.284 exposure to debates, which is interesting for various And by 0:19:48.284 --> 0:19:51.484 the way, if you have doubt about that, Biden says too, 0:19:51.564 --> 0:19:53.884 Trump agrees to two, they say how about two more, 0:19:54.644 --> 0:19:57.844 and then Biden's campaign chairwoman says no, no, no, no, no. 0:19:57.844 --> 0:20:00.564 That would create chaos. When they just create a chaos 0:20:00.844 --> 0:20:03.644 that morning by blowing up the schedule. It's an interesting 0:20:04.124 --> 0:20:07.564 kind of game theory problem because on the one hand, 0:20:07.564 --> 0:20:11.884 the debates are zero, one candidate gains more than the 0:20:11.924 --> 0:20:14.124 other almost by definition. On the other hand, if you 0:20:15.044 --> 0:20:17.324 duck the debates, that might be seen as worse than 0:20:17.884 --> 0:20:20.204 even a bad outcome of the debate. So it's kind 0:20:20.204 --> 0:20:21.844 of like the opposite of prisoners subletment, where you have 0:20:21.844 --> 0:20:23.684 an incentive not to cooperate. Here, you have an intetive 0:20:24.044 --> 0:20:26.364 to reach an agreement. You want to look too difficult 0:20:26.484 --> 0:20:31.124 or diffident. But usually when you're the candidate who's behind 0:20:31.164 --> 0:20:33.884 in the race, as Biden is in the swing states 0:20:33.924 --> 0:20:37.764 in the large majority of polls, you want more uncertainty 0:20:38.084 --> 0:20:41.564 and more variance, meaning that you want to take more risks. 0:20:41.844 --> 0:20:44.444 You should want more debates because they add volatility the election, 0:20:44.564 --> 0:20:48.684 and instead Biden wants fewer. That to me is a 0:20:48.724 --> 0:20:52.124 really really, really really bad sign for his campaign. It 0:20:52.164 --> 0:20:56.164 indicates his campaign either is delusional and doesn't believe the polls, 0:20:56.164 --> 0:20:58.404 and there's a lot of reporting that they don't or 0:20:58.444 --> 0:21:01.964 that they think their candidate is unlikely to commit himself 0:21:02.004 --> 0:21:06.324 well in spontaneous public interactions because he's eighty one years old, 0:21:06.404 --> 0:21:09.244 he has always stammered a lot, but you know, seems 0:21:09.284 --> 0:21:13.084 like audibly and visibly worse now. That is a bearish 0:21:13.084 --> 0:21:15.484 sign that you're constraining your stategic options, your chances to 0:21:15.524 --> 0:21:18.444 come back because you're worried about how well your guy 0:21:18.484 --> 0:21:22.004