WEBVTT - EXCLUSIVE: Peter Thiel on Why AI Can’t Afford to Hit the Pause Button 0:00:10.500 --> 0:00:12.200 Welcome to Tech Stuff. I'm Osvald Ossian. 0:00:12.680 --> 0:00:17.700 Today, a special bonus episode featuring an exclusive conversation with 0:00:17.739 --> 0:00:18.320 Peter Thiel. 0:00:19.239 --> 0:00:21.540 Well, I'm still not in the pause button camp. And 0:00:21.840 --> 0:00:23.320 I'm saying this on your show that I'm not in 0:00:23.340 --> 0:00:26.119 the pause button. We need to hit the accelerator as 0:00:26.180 --> 0:00:27.160 hard as possible. 0:00:27.660 --> 0:00:30.820 That's Peter Thiel talking about the debate over existential AI 0:00:30.880 --> 0:00:34.460 risk that's engulfed the leaders of the frontier labs, the president, 0:00:34.960 --> 0:00:38.350 and set a global conversation alight. The interview with Thiel 0:00:38.390 --> 0:00:42.610 was recorded on Wednesday, September 16th in Washington, D.C. And 0:00:42.630 --> 0:00:46.430 I'm here with Mangesh Hatikula, my co-founder at Kaleidoscope. Mangesh, 0:00:46.750 --> 0:00:48.129 how do we get the scoop from Thiel? 0:00:50.090 --> 0:00:53.449 I mean, this is an excerpt from Kaleidoscope's forthcoming podcast, 0:00:53.470 --> 0:00:57.850 The Builders, with Walter Isaacson. It is coming out in October. And, 0:00:58.020 --> 0:01:00.260 you know, for The Builders, we had Walter sit down 0:01:00.300 --> 0:01:03.820 with everyone from Peter Thiel to Elon Musk, the Amadeus, 0:01:03.860 --> 0:01:08.630 Dario and Daniela. I mean, The list of names is insane. 0:01:09.090 --> 0:01:11.750 And we're kind of in search of this question from 0:01:11.790 --> 0:01:14.550 all of them, right? Whose imagination do we want to 0:01:14.590 --> 0:01:18.250 live in? we're sharing this clip with the tech stuff 0:01:18.290 --> 0:01:22.250 audience early because, you know, Teal's comments were so urgent 0:01:22.350 --> 0:01:25.050 and so timely that we just couldn't hold onto it 0:01:25.110 --> 0:01:27.370 until October. And, you know, we wanted people to get 0:01:27.410 --> 0:01:29.890 a peek and Oz, I mean, I'm curious about this. 0:01:30.050 --> 0:01:32.869 Did you think we'd get Peter Teal? It has taken 0:01:32.950 --> 0:01:35.560 us almost one, like a year to get this interview. 0:01:35.580 --> 0:01:38.660 You flew to London. We did so many calls and 0:01:38.700 --> 0:01:40.839 I'm kind of just pinching myself that it happened. 0:01:41.060 --> 0:01:41.340 We did. 0:01:41.360 --> 0:01:43.649 I mean, we spent, we spent about a year working on, 0:01:43.800 --> 0:01:45.459 on the builders and booking the guests. 0:01:45.700 --> 0:01:49.310 And, uh, you know, Peter Thiel, Elon, others. 0:01:49.390 --> 0:01:51.370 I mean, I think that these types of interviews, like 0:01:51.710 --> 0:01:53.970 you don't quite believe they're going to happen until the 0:01:54.030 --> 0:01:55.690 person is literally sitting in the seat. 0:01:56.230 --> 0:01:57.330 It's insane. 0:01:57.490 --> 0:01:59.470 And yeah, I mean, I didn't, I wasn't sure they 0:01:59.490 --> 0:02:02.660 would happen because we got yeses and then long delays 0:02:02.720 --> 0:02:05.320 and then maybes and then pushed them back to yeses 0:02:05.380 --> 0:02:08.760 and stuff like that. But when we finally got the 0:02:08.780 --> 0:02:12.460 approval from Thiel, it was enormous because it's actually been 0:02:12.500 --> 0:02:14.160 more than a year since he sat down for a 0:02:14.200 --> 0:02:16.850 formal interview. I think the last one was with The 0:02:16.900 --> 0:02:20.880 New York Times and Ross Douthat. And obviously that made 0:02:20.919 --> 0:02:23.640 a ton of waves. But part of the mystique is 0:02:23.700 --> 0:02:25.420 that he doesn't give that many interviews. 0:02:25.840 --> 0:02:28.600 He doesn't. And so we're very excited to share this clip. 0:02:29.180 --> 0:02:32.210 We wanted to hit the accelerator as hard as possible, 0:02:32.250 --> 0:02:35.190 you could say, on publishing it. The full episode will 0:02:35.210 --> 0:02:38.370 come out on October 23rd. The series actually launches on 0:02:38.430 --> 0:02:42.269 October 20th with a conversation with Elon Musk. But in 0:02:42.310 --> 0:02:45.260 the meantime, here's an excerpt from The Builders of Peter 0:02:45.280 --> 0:02:47.790 Thiel's in conversation with Walter Isaacson. 0:03:01.980 --> 0:03:06.420 We've seen this enormous backlash in the past few weeks, 0:03:06.900 --> 0:03:11.620 past few months, against AI. What is causing that? 0:03:15.040 --> 0:03:19.810 Man, it's, uh, yeah, I agree with the what that 0:03:19.820 --> 0:03:24.270 it's happening. You know, why questions are always, always much harder, 0:03:25.050 --> 0:03:27.430 much more speculative in nature. I'm tempted to say it's 0:03:27.470 --> 0:03:31.169 somewhat overdetermined. It's very unclear who the winners and losers 0:03:31.210 --> 0:03:34.369 from AI will be. I think it's going to add 0:03:34.389 --> 0:03:35.790 a lot to growth, but it will do so in 0:03:35.810 --> 0:03:39.680 a fairly, fairly disruptive kind of a way. There are, 0:03:39.690 --> 0:03:44.370 of course, these, you know, these dystopian science fiction scenarios that, uh, 0:03:45.030 --> 0:03:48.670 You know, in some ways we believe, and maybe they're 0:03:48.710 --> 0:03:52.810 not entirely unreasonable. So there's some small tale. 0:03:52.830 --> 0:03:54.510 Dystopian mean that we'll all be white. 0:03:54.530 --> 0:03:56.330 The Terminator movie, the Matrix. 0:03:56.350 --> 0:03:59.010 But we've seen that Dario and Elon and Sam are 0:03:59.030 --> 0:03:59.700 now talking about that. 0:03:59.740 --> 0:04:03.360 And to some extent, yeah, to some extent they've, you know, 0:04:03.980 --> 0:04:07.140 leaned into it or partially believed it or too many 0:04:07.180 --> 0:04:11.460 of their employees believed it. And, yeah, there's some prehistory 0:04:11.500 --> 0:04:14.170 where this was, was in the air for a long time. 0:04:14.190 --> 0:04:17.180 I think Elon believes it for a decade. 0:04:17.380 --> 0:04:20.880 Well, you know, it was in some ways when Demis 0:04:20.900 --> 0:04:24.380 has created DeepMind and it gets sold to Google in 0:04:24.400 --> 0:04:28.480 the fall of 2013. Elon and Sam create OpenAI the 0:04:28.560 --> 0:04:33.100 next year or so after that. And it's basically, in 0:04:33.140 --> 0:04:36.820 some ways, it's an anti-Google company, right? What's closed AI? 0:04:37.080 --> 0:04:41.219 Closed AI is Google. And so this technology is too important, 0:04:41.260 --> 0:04:46.599 too dangerous. to be just in a single company. And then, 0:04:46.620 --> 0:04:49.160 of course, in some ways, that's the narrative they have 0:04:49.200 --> 0:04:54.289 internally at OpenAI. And then, of course, there are people 0:04:54.310 --> 0:04:57.150 who take it even more seriously than Altman. I think 0:04:57.230 --> 0:05:01.290 Elon exits around 2018. And then by 2020 or thereabouts, 0:05:01.350 --> 0:05:06.630 it's the seven people who start Anthropic exit from OpenAI. 0:05:06.690 --> 0:05:09.550 In some sense, they want to take these issues even 0:05:09.589 --> 0:05:16.940 more seriously than before. effective altruism, this idea had been 0:05:16.960 --> 0:05:19.500 in the air for a while. So I think there's 0:05:19.520 --> 0:05:20.300 sort of an interesting. 0:05:20.900 --> 0:05:23.880 Wait, explain that, because people are now saying it's sort 0:05:23.920 --> 0:05:27.280 of a psyops, or David Sachs says, conspiracy theory of 0:05:27.360 --> 0:05:31.220 effective altruism. And I think even people like me vaguely 0:05:31.279 --> 0:05:33.320 know what Sam Bankman-Fried was on about, but we don't 0:05:33.480 --> 0:05:34.700 get that. Tell me about that. 0:05:35.360 --> 0:05:38.940 Well, there's a lot of history, but even if you 0:05:39.140 --> 0:05:42.680 go back to the 2000s, there was a in the 0:05:42.820 --> 0:05:46.320 East Bay of Silicon Valley, Berkeley, Oakland, some of those places. 0:05:46.700 --> 0:05:48.680 It wasn't where the big tech companies were being built, 0:05:49.080 --> 0:05:52.520 but there was sort of a creativity with ideas. It 0:05:52.580 --> 0:05:55.710 was a little bit adjacent to the, somewhat adjacent to 0:05:55.750 --> 0:05:59.790 the new atheists, like Richard Dawkins and people like that, 0:05:59.810 --> 0:06:02.830 but then it had sort of a specific tech angle and, 0:06:03.089 --> 0:06:06.050 you know, how could you be more reasonable, rational, where 0:06:06.089 --> 0:06:09.790 was the world heading? And one big idea that these 0:06:09.830 --> 0:06:14.940 sort of futurists kept coming back to was that perhaps 0:06:15.050 --> 0:06:19.049 something like AI could happen, that perhaps it was around. 0:06:20.050 --> 0:06:23.230 It was closer than people thought. There had been an 0:06:23.490 --> 0:06:26.930 AI desert for a long time. It had been overpromised 0:06:27.170 --> 0:06:30.170 for many decades. But on the other hand, Moore's law 0:06:30.190 --> 0:06:34.610 was still progressing. The computers were relentlessly getting more powerful, 0:06:34.650 --> 0:06:38.430 and perhaps it was not a question of necessarily a 0:06:38.490 --> 0:06:40.630 much better algorithm, even though there were some of those 0:06:40.790 --> 0:06:43.010 things that happened on the algorithm level. But perhaps it 0:06:43.050 --> 0:06:46.510 was just almost a hardware problem that once you got 0:06:46.550 --> 0:06:50.860 the chips and the computers that were powerful enough, it 0:06:50.890 --> 0:06:53.480 would be more likely that something like this would happen. 0:06:55.020 --> 0:06:56.980 And then that this would be a very, very big 0:06:57.060 --> 0:07:02.060 thing when it happened. In certain dimensions, computers were already 0:07:02.100 --> 0:07:05.940 smarter than people. It was the late 1940s where computers 0:07:05.980 --> 0:07:09.310 had beat mathematicians and who could do square roots faster. 0:07:09.320 --> 0:07:12.920 I was sort of a pretty good chess player. And 0:07:12.960 --> 0:07:19.610 so 1997, when Deep Blue beats Kasparov, the IBM just 0:07:19.630 --> 0:07:24.130 computer beats Kasparov, wow, you've solved chess. And isn't chess 0:07:24.150 --> 0:07:27.770 the quintessentially human activity? But there was sort of this 0:07:27.870 --> 0:07:30.120 idea by the mid-2000s that this was going to start 0:07:30.150 --> 0:07:33.960 happening in a number of other domains at some point 0:07:34.020 --> 0:07:38.540 the next decade or two, and that this would be 0:07:38.580 --> 0:07:44.290 an enormously socially transformative thing. It was maybe potentially very dangerous. 0:07:45.070 --> 0:07:47.680 Maybe it could be weaponized. Maybe there were these science 0:07:47.700 --> 0:07:50.900 fiction scenarios where the AI would almost become autonomous. So 0:07:50.920 --> 0:07:54.260 there were very many different kinds of scenarios. And then 0:07:54.280 --> 0:07:55.720 there were, of course, parts of it that were, I 0:07:56.340 --> 0:08:02.690 don't know, crazy and almost cult-like. But it had maybe 0:08:02.710 --> 0:08:05.230 the good dimension of a cult where there was one 0:08:05.550 --> 0:08:08.630 big thing they knew that the other people did not. 0:08:08.650 --> 0:08:11.040 And I think, even though there are ways I'd like 0:08:11.060 --> 0:08:13.250 to be very critical of them today, There were ways 0:08:13.260 --> 0:08:15.890 they got that one, you know, that one big thing, right. 0:08:16.430 --> 0:08:20.230 And then, um, has a cult ever started a company 0:08:20.270 --> 0:08:23.150 as successful as Anthropic. You know, I don't know. I 0:08:23.170 --> 0:08:24.850 don't know what the most success, I mean, like even a, 0:08:25.370 --> 0:08:27.650 even a billion dollar company seems like bigger than a 0:08:27.850 --> 0:08:29.710 crazed cult could build. So you have a cult that's 0:08:29.750 --> 0:08:33.330 built a company that's worth probably close to 2 trillion 0:08:33.370 --> 0:08:36.310 at the IPO. It would probably be the, I don't know, sixth, 0:08:36.370 --> 0:08:39.890 seventh largest company by market cap in the world built 0:08:39.910 --> 0:08:43.329 in like six, seven years. Has a cult ever done 0:08:43.390 --> 0:08:47.010 something like that? So, yes. So I want to be 0:08:47.050 --> 0:08:50.750 critical of them, and they're things I'm uncomfortable and critical about. 0:08:51.240 --> 0:08:53.660 But the first thing I would try to do is always, 0:08:53.679 --> 0:08:55.380 you always want to steel man your opponents. You want 0:08:55.420 --> 0:08:59.060 to try to understand them as well as they understand 0:08:59.100 --> 0:09:02.040 themselves or even make their arguments better. And the way 0:09:02.080 --> 0:09:04.709 I would steel man the effective altruists is, first off, 0:09:05.160 --> 0:09:06.990 they got one enormously big thing right. 0:09:07.610 --> 0:09:10.150 And that one enormously big thing was? 0:09:10.450 --> 0:09:11.530 AI is around the corner. 0:09:11.820 --> 0:09:15.980 But also that it could become misaligned with human values. 0:09:18.660 --> 0:09:22.280 And then, of course, the way AI was classically defined, 0:09:22.340 --> 0:09:24.800 let's say. There were all these different definitions of AI, 0:09:24.820 --> 0:09:28.679 of course. It could be just sort of machine learning 0:09:29.200 --> 0:09:32.140 and sort of like maybe cameras everywhere or something like that. 0:09:32.160 --> 0:09:34.459 It's a sort of totalitarian AI that you have in China. 0:09:34.980 --> 0:09:39.540 Or it could be this sort of super duper AI 0:09:39.620 --> 0:09:42.329 with a godlike intelligence or something. which was sort of 0:09:42.350 --> 0:09:46.290 the Nick Bostrom kind of fantasy. But the in-between version 0:09:46.330 --> 0:09:49.569 of AI was not that it would be this dumb 0:09:50.170 --> 0:09:52.989 surveillance cameras, not that it would be godlike, but it 0:09:53.010 --> 0:09:55.610 would be like a human being. That had been the 0:09:55.690 --> 0:10:00.230 AI test. The Turing test from the beginning was it's 0:10:00.260 --> 0:10:03.030 not that it is a human being, but that can 0:10:03.130 --> 0:10:06.770 fool you into thinking it's a person. And from the outputs, 0:10:07.110 --> 0:10:12.010 you can't tell the difference. And yeah, that's going to 0:10:12.030 --> 0:10:14.670 be a very, very big deal. And it raises even 0:10:14.720 --> 0:10:19.210 basic economics questions. Is it economic compliment to people? Is 0:10:19.230 --> 0:10:20.470 it a substitute for people? 0:10:21.120 --> 0:10:23.720 Well, let's do both of those. There's two big things 0:10:23.780 --> 0:10:28.000 it raises. One is, what will it do to human labor? 0:10:28.740 --> 0:10:32.220 Will it create more jobs and more productivity? Or will 0:10:32.280 --> 0:10:34.450 it pretty much put us out of work and we'll 0:10:34.490 --> 0:10:38.290 have to figure out what we do? That's part of 0:10:38.309 --> 0:10:42.189 the backlash, especially my students at Tulane are like, wait 0:10:42.230 --> 0:10:46.650 a minute. So tell me your feeling about that backlash. 0:10:48.260 --> 0:10:50.460 Okay, so let me start by answering what I think 0:10:50.480 --> 0:10:53.400 will happen with a qualifier that nobody really knows and 0:10:53.620 --> 0:10:58.579 nobody can be certain. I think in aggregate, on a 0:10:58.640 --> 0:11:03.220 macroeconomic level, AI will increase productivity a lot. It will 0:11:03.240 --> 0:11:09.240 be a fairly disinflationary force in our economy. And then 0:11:09.700 --> 0:11:14.550 unless it's incredibly narrowly captured by a few companies, you'd 0:11:14.590 --> 0:11:17.730 expect the median person to be better off or things like. 0:11:18.030 --> 0:11:20.750 This was, you know, the Luddites in the 19th century 0:11:20.770 --> 0:11:23.590 were always worried that the machines would replace the people. 0:11:24.130 --> 0:11:26.450 And then, of course, what happened in practice was, yeah, 0:11:26.929 --> 0:11:30.270 there were definitely various industries, sometimes even big industries, where 0:11:30.290 --> 0:11:34.150 there was a lot of disruption. But then it would 0:11:34.210 --> 0:11:36.689 free people up to do more productive things, to do 0:11:37.250 --> 0:11:42.390 other things. And that would still be my intuition with AI. 0:11:42.510 --> 0:11:45.270 But, Wilk, before you go on, you said unless it's 0:11:45.330 --> 0:11:48.410 captured by a few companies, are you worried? that indeed 0:11:48.470 --> 0:11:52.570 we're about to have four or five companies capture it? 0:11:52.670 --> 0:11:55.570 Yeah, most of the history of tech innovation, I would 0:11:55.610 --> 0:12:01.290 say for 200, 250 years, relatively small parts of the 0:12:01.309 --> 0:12:04.710 innovation were captured by business. They mostly went to consumers, 0:12:04.770 --> 0:12:08.170 to workers, to the rest of the economy. There are 0:12:08.230 --> 0:12:13.170 ways that software and the internet were somewhat of an exception. 0:12:13.450 --> 0:12:19.520 And so we had these enormously successful businesses. Maybe they 0:12:19.559 --> 0:12:22.860 were natural monopolies of sorts. They were in some sense 0:12:23.600 --> 0:12:28.679 good for consumers, but also extremely profitable. And there was 0:12:28.720 --> 0:12:32.719 something about if you have a single e-commerce company like 0:12:32.820 --> 0:12:37.000 Amazon that dominates e-commerce, it can be cheaper for consumers 0:12:37.059 --> 0:12:39.830 because you have sort of one super efficient system of warehouses, 0:12:40.300 --> 0:12:43.550 but then you also can have very large monopoly profits. 0:12:43.570 --> 0:12:46.050 So there is a way that a lot of the 0:12:46.090 --> 0:12:49.660 value gets captured by by one or a few big 0:12:49.700 --> 0:12:53.839 internet businesses. And so there's a pattern matching to what 0:12:53.880 --> 0:12:57.959 happened with the internet that I think people quite reasonably 0:12:58.020 --> 0:13:00.500 are worried about will happen with AI, that it's an 0:13:00.570 --> 0:13:04.090 even bigger market, it's a bigger technology. And so perhaps- 0:13:04.110 --> 0:13:05.830 So you would try to stop that. In other words, 0:13:05.929 --> 0:13:09.690 our antitrust laws- Well, the question is, is that how 0:13:09.730 --> 0:13:13.330 likely do you think that's going to happen? The intuition 0:13:13.350 --> 0:13:17.320 that I have on the other side is that- You know, 0:13:17.340 --> 0:13:18.679 one of the, one of the questions I always like 0:13:18.780 --> 0:13:21.740 asking companies is tell me why you have a monopoly because, 0:13:21.800 --> 0:13:24.340 you know, we can sort of debate whether they're good 0:13:24.360 --> 0:13:26.460 or bad for society. Obviously we have IP, we have 0:13:26.480 --> 0:13:30.540 all kinds of good monopoly forms of monopoly. Um, but on, 0:13:30.700 --> 0:13:33.220 on the inside, yeah, you don't want to have just 0:13:33.280 --> 0:13:37.260 cutthroat competition where all the profits get competed away. And, um, 0:13:37.740 --> 0:13:39.780 and the kind of narratives people tell me when they 0:13:39.820 --> 0:13:44.260 pitch me on AI companies, they're narratives of growth of 0:13:45.230 --> 0:13:49.630 terrific technology. They are incredibly weak on the monopoly story. 0:13:50.010 --> 0:13:52.630 And then the felt sense of it is that it's 0:13:52.670 --> 0:13:56.920 like this incredibly fast, dynamic race where it's been changing 0:13:56.980 --> 0:13:59.900 a lot. DeepMind, in a sense, was in the lead 0:13:59.940 --> 0:14:04.599 in the early 2010s. Then OpenAI with ChatGPT and the 0:14:04.640 --> 0:14:08.219 transformer models was able to replace that. And then in 0:14:08.340 --> 0:14:11.970 some ways, Altman locked up a lot of the chips 0:14:12.110 --> 0:14:14.210 and built the data centers preemptively. So there were sort 0:14:14.230 --> 0:14:17.679 of all sorts of ways he correctly leaned into this 0:14:17.880 --> 0:14:22.320 incredibly CapEx intensive strategy. And then there still is a 0:14:22.340 --> 0:14:27.060 sense where, you know, maybe Anthropic is a little bit 0:14:27.080 --> 0:14:30.610 more in the pole position now than OpenAI was. And, 0:14:30.620 --> 0:14:33.610 you know, it's always hard to exactly score these things, 0:14:33.630 --> 0:14:36.230 but I would say that it also still feels that 0:14:36.490 --> 0:14:41.290 Google and Meta and Grok are not that far behind. 0:14:41.310 --> 0:14:44.320 You know, I'm not sure China is that far behind. 0:14:45.050 --> 0:14:48.040 So it feels, if we describe it on the level 0:14:48.060 --> 0:14:53.020 of the Shakespearean drama among these people, it's like really crazy. 0:14:53.320 --> 0:14:56.640 So they're acting like, man, it is a fight to 0:14:56.660 --> 0:14:59.880 the death. It is the furthest thing from monopoly imaginable. 0:15:00.300 --> 0:15:03.070 And if it's not a monopoly, that's where I think 0:15:03.590 --> 0:15:07.750 the benefits will be likely distributed pretty widely in society. Now, 0:15:07.930 --> 0:15:09.790 if I had to steel man the anxiety, which was 0:15:10.350 --> 0:15:12.750 your starting question, I think you could have a lot 0:15:12.770 --> 0:15:15.460 of growth. And the median person could be better off. 0:15:15.880 --> 0:15:18.740 But it's with a lot of volatility. You expect to 0:15:18.760 --> 0:15:24.250 get a 3% wage hike, but it's plus or minus 10% 0:15:24.250 --> 0:15:26.930 or something like that. And we're kind of in this 0:15:27.010 --> 0:15:30.989 society that doesn't really like risk, doesn't like that sort 0:15:31.010 --> 0:15:37.070 of volatility. It doesn't like disruption. The universities are in 0:15:37.130 --> 0:15:41.350 some ways all still very geared towards these 20th century 0:15:42.870 --> 0:15:45.930 industries these 20th century tracks and then maybe these tracks 0:15:45.990 --> 0:15:48.650 still work maybe some of them get disrupted a lot 0:15:49.170 --> 0:15:52.770 so um maybe good for the median person the good 0:15:52.810 --> 0:15:55.460 may be sort of unknown the bad things are more 0:15:55.540 --> 0:15:59.540 predictable and the disruptions disruptions clear so i think even 0:15:59.580 --> 0:16:02.740 on the economic level i can understand the anxiety even 0:16:02.760 --> 0:16:05.910 though i think it is misguided well. 0:16:05.760 --> 0:16:08.810 You talk about there being a lot of competition among 0:16:08.830 --> 0:16:12.630 these and that seems like generally a good thing Although, 0:16:12.910 --> 0:16:15.350 in some ways, if it was just one or two companies, 0:16:15.850 --> 0:16:19.370 you could control it better, regulate it for safety better. 0:16:19.850 --> 0:16:24.910 And secondly, the competition, either amongst these companies or with China, 0:16:25.510 --> 0:16:29.530 means you can't pause the way people are saying now, 0:16:29.850 --> 0:16:31.229 shouldn't we be pausing this? 0:16:32.210 --> 0:16:35.330 That seems correct to me. And, you know, there's sort 0:16:35.350 --> 0:16:39.220 of all these very complicated questions one can ask about China, 0:16:39.250 --> 0:16:43.979 but I'm, you know, I worry about totalitarianism. I don't 0:16:44.040 --> 0:16:46.760 like it. And I think there is a certain type of, 0:16:46.770 --> 0:16:52.470 you know, high tech totalitarianism that's being tried out in China. Um, 0:16:52.650 --> 0:16:55.970 it's not as violent as, you know, Stalin or Mao or, 0:16:56.370 --> 0:16:59.480 or Hitler. You don't need to kill as many people, but, uh, 0:16:59.880 --> 0:17:03.340 it can go to the inner recesses of your thoughts 0:17:03.400 --> 0:17:06.000 and it knows more about you than you know about yourself. 0:17:06.520 --> 0:17:09.830 And there's something about that, that, uh, We really don't 0:17:09.850 --> 0:17:11.190 want that to be the model that wins. 0:17:11.230 --> 0:17:14.030 And your old friend, David Sachs, and of course, President 0:17:14.070 --> 0:17:18.040 Trump are now just very recently saying, because of the 0:17:18.130 --> 0:17:22.859 competition with China, stop talking about pausing AI, even if 0:17:22.900 --> 0:17:25.500 you think it could be an existential threat. Do you 0:17:25.520 --> 0:17:26.020 agree with that? 0:17:29.119 --> 0:17:29.320 Yes. 0:17:29.500 --> 0:17:29.660 Yes. 0:17:30.160 --> 0:17:33.240 I broadly agree with that. And I've been friends with 0:17:33.320 --> 0:17:39.280 David Sachs for many decades. The thing that makes... me 0:17:39.340 --> 0:17:44.140 less comfortable making those kinds of arguments publicly is that 0:17:44.570 --> 0:17:47.630 on some level, it's hard when the big companies themselves 0:17:47.670 --> 0:17:49.590 are telling you to stop and to slow down. 0:17:49.890 --> 0:17:52.370 So that's caused you to have some pause because you 0:17:52.470 --> 0:17:54.850 were not in the pause button camp. 0:17:54.950 --> 0:17:57.250 Well, I'm still not in the pause button camp. And 0:17:57.550 --> 0:17:59.030 I'm saying this on your show that I'm not in 0:17:59.050 --> 0:18:01.810 the pause button. We need to hit the accelerator as 0:18:01.869 --> 0:18:02.850 hard as possible. 0:18:03.190 --> 0:18:05.119 Well, wait a minute. Let me stop that. Why haven't 0:18:05.340 --> 0:18:09.550 the past few weeks caused you to be more amenable 0:18:09.590 --> 0:18:10.250 to oppose by. 0:18:19.650 --> 0:18:23.609 I mean, anthropic as a cult, hitting the accelerator, monopolies 0:18:23.650 --> 0:18:27.290 are maybe good. The peek into his mind is so fascinating. 0:18:27.770 --> 0:18:30.830 And obviously that is just a snippet and Thiel gets 0:18:30.869 --> 0:18:31.970 into so much more. 0:18:32.670 --> 0:18:35.850 He certainly does. We'll be publishing The Builder's Feet very soon. 0:18:36.350 --> 0:18:39.149 We'll be promoting it on Tech Stuff, of course. But 0:18:39.190 --> 0:18:41.800 come back on October 20th for the conversation with Elon 0:18:41.820 --> 0:18:45.990 Musk and October 23rd for the full conversation with Peter Thiel. 0:18:46.290 --> 0:18:47.110 We can't wait for you to listen.