WEBVTT - Can Artificial Intelligence Be Controlled? With Tristan Harris & Aza Raskin 0:00:00.080 --> 0:00:03.480 Fear of all of us losing has to become greater 0:00:03.960 --> 0:00:06.160 than the fear of me losing to you. Now, China 0:00:06.160 --> 0:00:07.720 doesn't want the financial system to collect. 0:00:07.760 --> 0:00:11.720 It's the let it rip administration. What to do about AI? 0:00:11.880 --> 0:00:14.880 Am I going to lose my job? Or what about safety? 0:00:14.960 --> 0:00:19.520 Cyber security? What about privacy? While new documentary is out 0:00:19.560 --> 0:00:23.479 answering all of those questions, promise, peril, truth, trust? Should 0:00:23.520 --> 0:00:26.920 we be pessimistic? Should we be optimistic? It's called the 0:00:26.960 --> 0:00:31.600 AI Doc. And two of the principal participants in that documentary, 0:00:31.640 --> 0:00:35.880 Tristan Harris and Asa Raskin, are up next on This 0:00:36.240 --> 0:00:46.440 is Gavin news This is Gavin Newsom. 0:00:46.640 --> 0:00:48.960 And this is Tristan Harris and Azarasca. 0:00:52.040 --> 0:00:54.720 All Right, so you guys, you're on the back end 0:00:54.760 --> 0:00:57.040 now this world tour where you guys have been all 0:00:57.040 --> 0:00:59.480 over the damn place. Yeah, and you launched it. And 0:00:59.520 --> 0:01:02.600 it's on the basis of a movie that you you 0:01:02.640 --> 0:01:05.680 didn't necessary, Its not your movie, right, but you participated 0:01:05.680 --> 0:01:06.160 in a movie. 0:01:06.200 --> 0:01:06.520 That's right. 0:01:06.720 --> 0:01:09.560 This new documentary that was released what a month ago? 0:01:09.680 --> 0:01:11.199 Yeah, but a month and a half ago, The AI 0:01:11.319 --> 0:01:14.560 Doc or How I Became an Apocalyptimist. Our friends the 0:01:14.600 --> 0:01:17.920 directors of Everything Everywhere, all at once. We chatted with 0:01:17.959 --> 0:01:21.280 them several years ago when we first really recognized the 0:01:21.319 --> 0:01:25.320 situation we were in and asked them for their help 0:01:25.400 --> 0:01:28.000 to help create a movie to clarify the AI situation. 0:01:28.120 --> 0:01:31.000 And long story short, we'll talk about it, but we 0:01:31.040 --> 0:01:33.039 got the directors and have only involved and its whole 0:01:33.040 --> 0:01:35.160 team came together and tried to make a movie that 0:01:35.200 --> 0:01:36.959 would clarify the predicament that we're facing. 0:01:37.120 --> 0:01:39.040 Kevin, do you do you remember seeing the film the 0:01:39.080 --> 0:01:39.480 day after? 0:01:39.720 --> 0:01:42.319 Course, well, man, I want to maybe I was alive. 0:01:42.440 --> 0:01:46.240 It was the early eighties or something eighty two, eighty two, 0:01:46.280 --> 0:01:49.760 eighty three. Yeah, it was like a seven pm Tuesday 0:01:49.840 --> 0:01:52.840 night exactly every human being on planet Earth after. 0:01:52.840 --> 0:01:55.480 The largest synchronized television event in human history in terms 0:01:55.480 --> 0:01:57.840 of the number of people seeing the same thing at 0:01:57.840 --> 0:02:02.080 the same time has probably the Olympics. It was a 0:02:02.160 --> 0:02:04.400 made for TV movie about what would happen if there 0:02:04.440 --> 0:02:07.520 was nuclear war? Right, And it's not as if people 0:02:07.560 --> 0:02:10.440 didn't know what nuclear war was, but there's something about 0:02:10.520 --> 0:02:12.320 we didn't really want to think about it or confronted 0:02:12.360 --> 0:02:14.600 like why would you? And I think the people who 0:02:14.639 --> 0:02:18.400 made that film were trying to create this kind of 0:02:18.440 --> 0:02:20.960 collective confrontation because the film was aired in the Soviet 0:02:21.040 --> 0:02:23.919 Union like five years I think five years later, right 0:02:24.800 --> 0:02:28.079 right before the Reichivika courds the first arms control talks, 0:02:28.400 --> 0:02:30.760 and that set the context that it's like I know 0:02:30.880 --> 0:02:32.840 that you know that I know, and you know that 0:02:32.880 --> 0:02:34.399 I know that you know that we both don't want 0:02:34.400 --> 0:02:36.640 that to happen. It's what you know what Stephen Pinker 0:02:36.680 --> 0:02:38.280 calls common knowledge, but we think of it as this 0:02:38.360 --> 0:02:41.440 almost common feeling. Yeah, yeah, that we both we both 0:02:41.560 --> 0:02:44.360 know that we feel the same way about that anti 0:02:44.480 --> 0:02:47.320 human outcome. And I think that with AI we have 0:02:47.400 --> 0:02:50.200 to get clear because AI is a much more confusing 0:02:50.240 --> 0:02:52.720 technology because it's like it's like if nu because would 0:02:52.720 --> 0:02:53.800 say it say if nucid. 0:02:54.240 --> 0:02:57.720 Imagine we're trying to reason about nukes that also could 0:02:57.760 --> 0:02:59.720 cure cancer, Like how would. 0:02:59.520 --> 0:03:00.360 You do something? 0:03:00.480 --> 0:03:01.840 Or pump GDP by ten percent? 0:03:01.919 --> 0:03:02.959 That's right, right? 0:03:03.400 --> 0:03:05.280 And so what The Day After did is that it 0:03:05.400 --> 0:03:08.639 created this this common feeling common knowledge where we could 0:03:08.639 --> 0:03:10.480 all see, oh, a world where we all go to 0:03:10.520 --> 0:03:12.959 nuclear war. That like that, that's an anti human, anti 0:03:13.000 --> 0:03:16.240 life future. And as long as there's confusion about which 0:03:16.400 --> 0:03:19.160 default world AI and thesin we're not going to do 0:03:19.200 --> 0:03:21.840 anything about it. But if we all have clarity that 0:03:21.880 --> 0:03:23.799 it's heading us towards an anti human future, and we 0:03:23.840 --> 0:03:26.680 can build that out in this in this interview, then 0:03:26.680 --> 0:03:29.200 that means it creates the conditions where we can coordinate 0:03:29.280 --> 0:03:30.080 to do something else. 0:03:30.680 --> 0:03:34.400 So this so that was the impetus behind this, this documentary. 0:03:34.400 --> 0:03:37.680 It was to create that collective understanding, that's that wisdom, 0:03:38.000 --> 0:03:41.680 that collective then response mechanism, which would be we need 0:03:41.680 --> 0:03:44.320 to do something about it. Even the promise that we 0:03:44.360 --> 0:03:46.720 could talk about at the peril in terms of the 0:03:46.720 --> 0:03:49.440 safety risks and the anxiety that somebody is at feelings. 0:03:49.440 --> 0:03:51.440 So you guys have been out on this tour, You've 0:03:51.440 --> 0:03:53.240 been all over the place over the course of the 0:03:53.320 --> 0:03:56.800 last month, and you've been you know, obviously you know, 0:03:56.920 --> 0:03:59.840 reaching out to people on all political sides of the aisle. 0:04:00.680 --> 0:04:03.720 Because of the human nature of this. This is universal. 0:04:03.800 --> 0:04:05.880 That's why it's something that connects all of us. This 0:04:06.000 --> 0:04:09.680 is not a partisan frame this And so is that 0:04:09.760 --> 0:04:14.000 something that you that's been captured in your own consciousness 0:04:14.040 --> 0:04:17.400 as you've been out on this store, how real and 0:04:17.600 --> 0:04:20.920 invisible that is or are we still in the process 0:04:20.960 --> 0:04:22.920 of discovery. I mean, are we still in the process 0:04:22.960 --> 0:04:26.400 of understanding more fully what the hell this is all about. 0:04:26.760 --> 0:04:29.640 I think people still not everybody knows. And the film 0:04:29.920 --> 0:04:31.760 was just out in theaters and has a limited release 0:04:31.800 --> 0:04:34.080 on streaming. Now it should be on Peacock on I 0:04:34.080 --> 0:04:37.280 think May twentieth, which I mean a little bit more 0:04:37.279 --> 0:04:39.840 people will see it. But I think that you know, 0:04:40.000 --> 0:04:42.760 the universal human aspect of this we learned also from 0:04:42.760 --> 0:04:45.360 social media, Like the social media didn't care whether you're 0:04:45.360 --> 0:04:48.479 a Democrat or Republican. It creates loneliness for everyone. It 0:04:48.520 --> 0:04:51.400 creates addiction and doom scrolling and brain rot for everyone. 0:04:51.400 --> 0:04:51.560 I know. 0:04:51.600 --> 0:04:54.240 That's how we first met was really on the back 0:04:54.240 --> 0:04:55.520 of the other film that we were a part of 0:04:55.560 --> 0:04:59.440 the social dilemma. And I think the good news about 0:04:59.720 --> 0:05:02.520 that film is that it has both catalyzed actually a 0:05:02.520 --> 0:05:04.360 lot of change and not the laws yet we know that. 0:05:04.480 --> 0:05:07.160 But I think it primed us to now be much 0:05:07.160 --> 0:05:10.000 more cautious about AI. So I think now that people 0:05:10.120 --> 0:05:12.560 know that social media was a problem, it makes it 0:05:12.680 --> 0:05:14.880 much easier to say that you shouldn't just assume that 0:05:14.920 --> 0:05:17.720 the default trajectory of a technology is going to land 0:05:17.760 --> 0:05:20.880 us in a good future. But you know, the social 0:05:20.920 --> 0:05:24.400 dilemma was seen basically by two hundred million people across 0:05:24.400 --> 0:05:24.919 planet Earth and. 0:05:24.920 --> 0:05:28.680 An hundred and nine released twenty twenty September of twenty 0:05:28.680 --> 0:05:30.760 twenty September twenty twenty, during the middle of the pandemic. 0:05:30.800 --> 0:05:34.440 And I think that also really mattered because people were 0:05:34.760 --> 0:05:36.320 a lot of people were stuck at home and they 0:05:36.360 --> 0:05:40.039 were only seeing reality through the monoculars of this social 0:05:40.080 --> 0:05:42.680 media news feed. So suddenly you saw your friends all 0:05:42.720 --> 0:05:45.159 start to go crazy on both sides, like suddenly the 0:05:45.160 --> 0:05:48.400 politics got more extreme. Because as everybody knows now, you know, 0:05:48.560 --> 0:05:52.440 social media increases the visibility of the most extreme voices, 0:05:52.560 --> 0:05:56.800 we get a double whammy of overrepresentation of the extreme voices, 0:05:56.880 --> 0:06:00.360 both because the extreme voices post more often and dominate 0:06:00.360 --> 0:06:03.680 the discourse and whatever they say goes more viral, so 0:06:03.720 --> 0:06:06.720 you get double over representation. So the more you use 0:06:06.760 --> 0:06:09.599 social media, the worse you are at predicting what the 0:06:09.600 --> 0:06:12.440 other side actually believes. And so you think that with 0:06:12.480 --> 0:06:14.560 this technology that's supposed to bring us together and make 0:06:14.640 --> 0:06:17.760 us the most enlightened society, it's actually making us more 0:06:17.800 --> 0:06:19.680 confused about what's really real and in fact, what our 0:06:19.680 --> 0:06:22.400 fellow Americans actually believe. And I think with your podcast, 0:06:22.839 --> 0:06:24.960 this is about actually you're talking to everybody. You're trying 0:06:25.000 --> 0:06:27.480 to say, this is a human conversation. We talk to everybody, 0:06:27.720 --> 0:06:29.599 and this is kind of fighting the effects I think 0:06:29.600 --> 0:06:30.760 of the social media problem. 0:06:30.880 --> 0:06:32.840 So you were I mean in twenty twenty what was 0:06:32.880 --> 0:06:35.800 so resonant is it was almost and it wasn't your 0:06:35.800 --> 0:06:38.000 intent to say, look, I told you so, But you 0:06:38.000 --> 0:06:40.520 were talking about these things in twenty twelve, twenty thirteen. 0:06:40.600 --> 0:06:43.960 That's right. We're saying the incentive structure and I'm going 0:06:43.960 --> 0:06:45.280 to get to in satives because it goes to the 0:06:45.279 --> 0:06:50.480 core of this exactly, this notion of you know, whatever, 0:06:50.480 --> 0:06:53.240 if someone's paycheck is attached to whatever the incentive is, 0:06:53.279 --> 0:06:57.599 which is obviously with social media it was eyeballs, was 0:06:57.640 --> 0:07:01.680 doom scrolling, which you're intimately familiar with different reasons, and 0:07:01.880 --> 0:07:04.960 that notion that we have a chance now with AI 0:07:05.600 --> 0:07:09.039 not to make the mistakes was made the neglect in 0:07:09.080 --> 0:07:14.720 particular and underregulating social media now with AI, that said, 0:07:14.840 --> 0:07:20.080 doesn't seem like there's a lot of regulatory activity with AI. 0:07:20.120 --> 0:07:22.320 In the last couple of years. I mean, so talk 0:07:22.360 --> 0:07:23.880 to me a little bit about that. Talk to me 0:07:23.880 --> 0:07:26.400 about the lessons that we should have learned about social 0:07:26.400 --> 0:07:28.960 media and how we can adapt and adopt in the 0:07:29.000 --> 0:07:30.000 AI strength frame. 0:07:30.560 --> 0:07:33.040 Well, really, like I think the core of it is 0:07:33.320 --> 0:07:35.960 if we gu just get confused by looking at all 0:07:36.000 --> 0:07:38.840 of the sort of epiphenomena, like the different kinds of 0:07:38.920 --> 0:07:41.280 harms that social media made. Like if you're trying to 0:07:41.320 --> 0:07:45.840 solve just the loneliness thing, or you're just working on 0:07:46.000 --> 0:07:50.040 solving the disinformation thing or just the teen sexualization of 0:07:50.080 --> 0:07:51.960 the thing, well then many different people are working in 0:07:52.000 --> 0:07:54.200 different parts of the problem. You're not solving the core thing, 0:07:55.000 --> 0:07:57.400 which is the race to the bottom of the brain 0:07:57.440 --> 0:07:59.760 stem for attention. And if we could just focus our 0:08:00.200 --> 0:08:01.800 on that, then you actually can solve all of the 0:08:01.920 --> 0:08:05.200 other problems at once. And that's sort of the core insight. 0:08:06.240 --> 0:08:08.040 We probably explained that what would it mean to do that? 0:08:08.160 --> 0:08:11.440 So if you're solving the core problem of let's say 0:08:11.440 --> 0:08:14.560 none of the companies are maximizing engagement just we don't 0:08:14.560 --> 0:08:16.080 live in that world. We don't have the regulation for that. 0:08:16.080 --> 0:08:17.960 But let's snap our fingers and now no one is 0:08:18.080 --> 0:08:20.440 maximizing screen time, you're like, let's just say it wasn't 0:08:20.440 --> 0:08:23.280 allowed to do that. So now instead of each company 0:08:23.320 --> 0:08:25.920 trying to maximize duration of use and frequency of use, 0:08:26.400 --> 0:08:28.920 you just have these products that each design decision isn't 0:08:28.920 --> 0:08:31.800 trying to predate or manipulate you into spending more time, 0:08:32.040 --> 0:08:34.400 which means that your experience is you're not getting sucked 0:08:34.440 --> 0:08:37.760 in constantly to everything. So that deals with the loneliness issue. 0:08:38.559 --> 0:08:41.319 If you're not trying to show people the most hyper 0:08:41.320 --> 0:08:44.719 normal stimuli, meaning like a hyper dopamine response for any 0:08:44.720 --> 0:08:46.079 piece of content, that you're not going to get the 0:08:46.080 --> 0:08:50.760 sexualization of people, and you're not incentivizing creators to maximize 0:08:50.800 --> 0:08:53.920 their own reach and engagement because you're not maximizing engagement yourself. 0:08:54.000 --> 0:08:57.319 So suddenly, when you attack the attention incentive, you're dealing 0:08:57.320 --> 0:08:59.960 with sexualization of content, you're dealing with less viral content, 0:09:00.080 --> 0:09:03.280 you get less disinformation, and you're dealing with loneliness too. 0:09:03.440 --> 0:09:05.600 So that's a good example of we obviously don't have 0:09:05.679 --> 0:09:08.240 laws that do that right now, but it points at 0:09:08.320 --> 0:09:10.319 the center of the bullseye is the incentive, and when 0:09:10.400 --> 0:09:13.680 you tackle the core incentive, you get benefits across the 0:09:13.720 --> 0:09:17.320 spectrum and AI is going to be more confusing because 0:09:17.360 --> 0:09:20.000 when people think about the incentive, they're like, okay, so 0:09:20.000 --> 0:09:22.000 social media, I got the incentive. It's like, how much 0:09:22.000 --> 0:09:25.920 have you paid for your Instagram account recently? Nothing? So 0:09:25.920 --> 0:09:27.640 how they worth the trillions of dollars? It was attention, 0:09:27.760 --> 0:09:30.600 so we knew they were maximizing that thing. But with AI, 0:09:30.720 --> 0:09:32.920 you say, okay, what's the incentive for a regular person? 0:09:33.280 --> 0:09:35.319 They think, okay, how does open ai make money? And 0:09:35.360 --> 0:09:37.240 they say, well, only when I pay them the twenty 0:09:37.240 --> 0:09:39.840 bucks a month for subscriptions. So maybe that's their incentive. 0:09:39.840 --> 0:09:43.080 They're just trying to maximize these subscriptions. But that doesn't 0:09:43.120 --> 0:09:45.079 justify if everybody paid twenty bucks a month, that does 0:09:45.080 --> 0:09:48.080 not pay back the trillion dollars the capex that they'd 0:09:48.120 --> 0:09:50.920 taken on. So what would justify that. Well, if they 0:09:50.920 --> 0:09:53.280 were to race to augment workers and like skive you 0:09:53.360 --> 0:09:56.840 tools to make your work more productive, that's great, but 0:09:56.920 --> 0:09:59.240 that wouldn't pay back the amount of money that's right. 0:09:59.840 --> 0:10:02.080 The only thing that can get them to be able 0:10:02.120 --> 0:10:04.400 to pay back the insane amounts of death that they're 0:10:04.440 --> 0:10:08.120 taken on is owning the human labor market. That is 0:10:08.160 --> 0:10:09.360 the incentive. 0:10:09.040 --> 0:10:10.880 The race to replace all human labor. 0:10:10.920 --> 0:10:14.040 That's right, to replace us first economically. 0:10:14.800 --> 0:10:17.200 And is that I mean, so is that written or 0:10:17.240 --> 0:10:20.800 is that unwritten? Is that understood within the industry or 0:10:20.880 --> 0:10:23.839 is it being I mean, is this what the consciousness 0:10:23.880 --> 0:10:25.240 you're trying to right? 0:10:25.600 --> 0:10:27.400 It's kind of a two phase thing. That actually used 0:10:27.400 --> 0:10:29.360 to be on Opening Eye's website that they said our 0:10:29.400 --> 0:10:33.280 mission is basically to create artificial general intelligence, which means 0:10:33.320 --> 0:10:35.720 to be able to replace all economically valuable work. They 0:10:35.720 --> 0:10:39.400 did change the mission statement. Yeah, but obviously everybody who's 0:10:39.520 --> 0:10:42.360 driving this knows for sure that's the prize, and if 0:10:42.400 --> 0:10:44.679 they don't do it, they fear that the other guy will. 0:10:44.960 --> 0:10:46.640 So even if they think it's bad to replace all 0:10:46.720 --> 0:10:49.800 labor and create this mass disruption, they feel caught in 0:10:50.160 --> 0:10:52.160 a race. And that's the thing we have to change, 0:10:52.760 --> 0:10:54.520 is that the fear of me losing to the other 0:10:54.600 --> 0:10:57.920 guys currently dominating over the fear of what happens to 0:10:58.080 --> 0:11:01.120 everybody losing from the anti human future. That is the outcome. 0:11:01.200 --> 0:11:03.680 So sort of see this with Demistis's office. When he 0:11:03.840 --> 0:11:06.160 co found a Google deep Mind, he set the mission 0:11:06.160 --> 0:11:09.560 statement to first solve intelligence and then use intelligence to 0:11:09.559 --> 0:11:12.360 solve everything else. But what that really meant was the 0:11:12.400 --> 0:11:15.439 beginning of a race to first dominate intelligence and then 0:11:15.520 --> 0:11:17.440 use intelligence to dominate everything else. 0:11:17.640 --> 0:11:20.080 In deep Mind is the origin story. It's right in 0:11:20.120 --> 0:11:22.840 many respects. Go back a little bit. I mean Google's 0:11:22.840 --> 0:11:26.520 deep Mind, they acquired deep Mind, that they acquired some 0:11:26.600 --> 0:11:30.319 of the intellectual assets, but it was really I mean, 0:11:30.440 --> 0:11:32.679 Larry Page, you can go back to the you know, 0:11:32.760 --> 0:11:34.800 sort of origin of the beginning of the beginning. Well, 0:11:34.920 --> 0:11:36.360 what years are we talking about. 0:11:36.200 --> 0:11:38.280 Like twenty fourteen is when I think they acquired it 0:11:38.320 --> 0:11:40.280 in twenty fourteen. I think they started it in twenty 0:11:40.360 --> 0:11:42.280 twelve or twenty eleven or something like that. 0:11:42.480 --> 0:11:44.559 Then and it goes to this competition question, which I 0:11:44.600 --> 0:11:46.880 want to get you because it's the domestic competition between 0:11:47.120 --> 0:11:48.880 all of these companies. And then we get to the 0:11:48.920 --> 0:11:52.920 competition in China in particular, particularly now with the president 0:11:53.440 --> 0:11:56.600 President She and President Trump that's about to meet. But 0:11:56.720 --> 0:11:59.760 on the issue of the competition that was born here 0:12:00.360 --> 0:12:03.320 around twenty fourteen with deep Mind, it was an interesting 0:12:03.640 --> 0:12:06.600 competition that sort of formed with Elon Musk in a 0:12:06.640 --> 0:12:09.480 relationship he had close relationship at the time. I intimately 0:12:09.600 --> 0:12:13.800 was familiar with that with himself and Larry Page. But 0:12:13.920 --> 0:12:17.160 they had a conversation, alleged conversation that didn't go the 0:12:17.200 --> 0:12:21.280 way that Elon thought it should, and Elon said, I 0:12:21.280 --> 0:12:25.560 am going to go out on my own with Sam Altman. 0:12:25.679 --> 0:12:29.360 That's ultimately Sam. They partnered and they created open AI. 0:12:29.400 --> 0:12:29.720 That's right. 0:12:29.720 --> 0:12:32.560 We probably should peel back the onion here and slow down. 0:12:32.679 --> 0:12:34.439 Just how do we get to this point? And what 0:12:34.600 --> 0:12:38.280 was the original philosophy that guided this? So Demisesabish is 0:12:38.320 --> 0:12:42.040 the founder of deep Mind. His original goal is, we 0:12:42.080 --> 0:12:45.119 thought we should we should have one project that pursues 0:12:45.360 --> 0:12:48.480 artificial general intelligence, meaning the kind of AI that's not 0:12:48.520 --> 0:12:50.079 the thing that just reads your license plate when you 0:12:50.120 --> 0:12:51.280 drive through the Golden gate Bridge. 0:12:51.640 --> 0:12:56.680 AI forever, we forever maps. I'll translate exactly exactly a AI. 0:12:56.800 --> 0:12:59.760 So AI is hardly a novel, exactly, it's just no 0:13:00.120 --> 0:13:02.559 of jen AI. Ultimately general AGI. 0:13:02.120 --> 0:13:04.520 Exactly, so artificial general intelligence, which is to be able 0:13:04.559 --> 0:13:07.360 to do all economic labor, to simulate all the things 0:13:07.360 --> 0:13:09.080 that a human mind can do in the kind of thinking. 0:13:10.040 --> 0:13:12.880 And so he originally wanted that to be like one project, 0:13:12.920 --> 0:13:15.439 almost like a cern, you know, the project in Switzerland 0:13:15.440 --> 0:13:18.200 and a global scientific project that's for the benefit of 0:13:18.280 --> 0:13:22.599 all of humanity, done slowly and carefully, mostly privately, not 0:13:22.800 --> 0:13:25.319 in a big public way. Take your time, get it right. 0:13:25.360 --> 0:13:28.559 That was that was a Demis's original goal. He sold 0:13:28.559 --> 0:13:31.600 it to Google. And the conversation you're talking about is 0:13:31.600 --> 0:13:33.160 then Elon was part of that. I think he was 0:13:33.160 --> 0:13:36.120 on the plane when they were literally negotiating the final sale, 0:13:36.600 --> 0:13:38.439 and there was some conversation it's talked about in the 0:13:38.480 --> 0:13:44.160 AI doc film where Elon realizes that Larry didn't really 0:13:44.200 --> 0:13:47.719 care whether humanity made it, whether he cared about a 0:13:47.960 --> 0:13:50.559 safety because in the end, if there's a digital intelligence 0:13:50.920 --> 0:13:53.000 that's smarter than us, that does more science, that can 0:13:53.000 --> 0:13:55.880 go out and explore the universe, even if we're wiped out, 0:13:56.240 --> 0:13:59.240 like we'll have created that, and that scared Larry and 0:13:59.600 --> 0:14:04.880 he used that's scared Elon, and Elon accused, no, sorry, 0:14:04.960 --> 0:14:07.960 Larry accused Elon of being a speciesist for caring about 0:14:08.000 --> 0:14:11.840 humans and privileging humans. And then that is what created 0:14:12.080 --> 0:14:12.640 Opening Eye. 0:14:13.400 --> 0:14:16.000 And it's just important to note to understand the psychology 0:14:16.000 --> 0:14:19.120 of the people that are making this that you might think, Okay, 0:14:19.120 --> 0:14:21.800 this is just one I mean, very powerful billionaire that 0:14:22.040 --> 0:14:25.120 thinks that maybe human beings should make it, but he 0:14:25.200 --> 0:14:27.640 doesn't care whether human that doesn't think they should make it, 0:14:27.640 --> 0:14:29.360 but he doesn't care care whether human beings make it 0:14:29.480 --> 0:14:31.680 or not. But you know, we were just talking about 0:14:31.680 --> 0:14:35.600 this the other day that in the New York Times, 0:14:35.640 --> 0:14:38.640 Peter Teo was being interviewed and he was asked should 0:14:38.760 --> 0:14:43.239 humanity endure? And there was seventeen seconds of him stuttering 0:14:44.360 --> 0:14:47.520 and he ended up with a after seventeen seconds of it, well, 0:14:47.560 --> 0:14:51.080 it sort of depends kind of answer. And that shows 0:14:51.120 --> 0:14:54.120 you the mentality of like we're trying to they're trying 0:14:54.120 --> 0:14:57.480 to build a god, and even if humanity doesn't make it, 0:14:57.560 --> 0:15:01.480 that god is built in like our country's value, our language. 0:15:01.600 --> 0:15:03.720 It's sort of like the mild progeny. 0:15:04.800 --> 0:15:07.520 Is the thing that like Elon burst the god that, yes, 0:15:07.600 --> 0:15:10.400 humanity got wiped out. But now there's this digital god 0:15:10.440 --> 0:15:12.080 that has Elon's DNA in it. 0:15:12.280 --> 0:15:14.640 And it's important that everyone understand that because if people 0:15:14.680 --> 0:15:17.440 really understood this, I think there'd be a lot more 0:15:17.520 --> 0:15:19.920 like hell no kind of energy like this is not 0:15:20.000 --> 0:15:20.360 the future. 0:15:20.600 --> 0:15:22.560 Well, I will say with Peter teeal interview got so 0:15:22.640 --> 0:15:25.160 much attention. It was like the I mean the holy 0:15:26.120 --> 0:15:27.680 in that wake up moment for a lot of people 0:15:27.960 --> 0:15:31.280 that didn't necessarily have that sweat or understanding, it's all 0:15:31.280 --> 0:15:33.960 of a sudden passive, particularly the folks. And I think 0:15:34.000 --> 0:15:37.320 what's most alarming about that. I mean, obviously Larry's next 0:15:37.400 --> 0:15:39.720 level brilliant, absolutely, and so his ability to see in 0:15:39.720 --> 0:15:41.240 the future. He doesn't have to climb over the mounta, 0:15:41.240 --> 0:15:43.440 he sees right through it. Yeah, but guys like Teal 0:15:43.520 --> 0:15:45.960 as well Lover hate them same thing. So these guys 0:15:46.000 --> 0:15:49.400 are so far in the future, they're seeing that darker side, 0:15:49.480 --> 0:15:51.520 and so they're having a difficult time even answering a 0:15:51.520 --> 0:15:54.800 simple question. That's the seventeen seconds. So let's go back 0:15:54.840 --> 0:15:56.560 as we unpack that, and this notion from the God 0:15:56.640 --> 0:15:59.000 complex will continue to come back. And I think it's 0:15:59.000 --> 0:16:01.600 profound and outside because it goes to the limited nature 0:16:01.960 --> 0:16:04.400 of just a handful of people that's right now, these 0:16:04.440 --> 0:16:07.000 trillionaires that will determine the fate and future billions and 0:16:07.000 --> 0:16:09.480 billions of people, and how we can get our arms 0:16:09.480 --> 0:16:11.520 around that. So and I want to read get to that, 0:16:11.640 --> 0:16:14.280 how we can get our arms around this? We have agency, 0:16:14.520 --> 0:16:16.160 right and that's why you did this doc, and that's 0:16:16.160 --> 0:16:17.640 why we're sitting here to go. That's right, that's right, 0:16:17.800 --> 0:16:19.560 because I don't want people to feel like this is 0:16:19.840 --> 0:16:21.280 that we're just by standards. 0:16:21.080 --> 0:16:23.320 And we're not admiring the problem. We're just we always 0:16:23.320 --> 0:16:26.360 say in our work and says this that clarity creates agency. 0:16:26.640 --> 0:16:28.640 If we can see this clearly and we can see 0:16:28.640 --> 0:16:30.520 where we're going, we can collectively say, if we want 0:16:30.520 --> 0:16:32.480 to go somewhere else, we'll choose something different. So we'll 0:16:32.480 --> 0:16:32.800 get to that. 0:16:32.960 --> 0:16:36.360 So so Elon goes out with Sam starts Open AI. 0:16:36.480 --> 0:16:40.000 Obviously they have an infamous fight and they're notoriously now 0:16:40.000 --> 0:16:43.160 I mean not notoriously, but the fight is obviously accelerated 0:16:43.320 --> 0:16:46.080 consciousness because now we're seeing it twenty four to seven 0:16:46.760 --> 0:16:48.960 court case, yep, with a court case in the Bay 0:16:49.000 --> 0:16:51.320 Area here with the two of them. So they had 0:16:51.320 --> 0:16:53.400 a falling out. Elon goes off and does his own 0:16:53.440 --> 0:16:56.960 thing because it doesn't feel like Sam's doing the right thing. Dario, 0:16:57.520 --> 0:17:00.680 who starts another AI company, feels like, well, Open Aye 0:17:00.720 --> 0:17:02.720 is not doing the right thing either, that's right, So 0:17:02.800 --> 0:17:07.320 he spins off and topic thropic and so now you've 0:17:07.320 --> 0:17:11.120 got free AGI projects, three AGI projects, all in our backyard, 0:17:11.200 --> 0:17:13.960 that's right, literally here in California and the Bay Area. 0:17:14.960 --> 0:17:18.320 And so this competition of sorts, that's a competition you 0:17:18.440 --> 0:17:21.480 just described, and it's a competition for the Holy Grail. 0:17:21.600 --> 0:17:24.080 That's right, right, It's like the it's Lord of the Rings, 0:17:24.080 --> 0:17:26.359 it's the ring from Lord of the Rings. Because it's essentially, 0:17:26.440 --> 0:17:31.280 as I was saying, first, dominate intelligence, then use intelligence 0:17:31.440 --> 0:17:34.240 to dominate everything else. Because if I get AGI first, 0:17:34.880 --> 0:17:38.680 I hit copy paste and I have one hundred million cyberhackers. 0:17:38.119 --> 0:17:40.439 That you don't have, right, and this is within seconds, 0:17:40.440 --> 0:17:42.360 this is one seconds. Yeah, this is not over course 0:17:42.400 --> 0:17:43.399 something that's right for years. 0:17:43.480 --> 0:17:45.919 If I get AGI first, I have an army of 0:17:45.960 --> 0:17:50.280 scientific companies that automate all scientific development. So suddenly I'm 0:17:50.280 --> 0:17:53.760 getting like twenty fourth century science and technology in my 0:17:54.160 --> 0:17:55.960 you know that I own and run that I can 0:17:55.960 --> 0:17:59.320 inform new military weapons and new physics and new And 0:17:59.359 --> 0:18:02.440 the problem is, notice that none of us can prove 0:18:02.480 --> 0:18:04.600 that they won't get that. We can't say for sure 0:18:04.600 --> 0:18:07.000 that they would get that. But the people who are 0:18:07.000 --> 0:18:10.040 optimistic and accelerationist about AI just want to bring in 0:18:10.040 --> 0:18:11.960 their perspective for a moment because they're represented in the 0:18:12.000 --> 0:18:15.560 AI DOT film. The film includes the risk folks who 0:18:15.560 --> 0:18:18.560 are oriented about safety, and it includes the accelerationists, And 0:18:18.600 --> 0:18:21.560 the accelerationists say the biggest risk is not going fast enough, 0:18:22.000 --> 0:18:24.439 because imagine all the science we could get, all the 0:18:24.480 --> 0:18:27.040 cancer drugs, all the medicine. People could be living forever. 0:18:27.280 --> 0:18:29.360 Think of all of the people who would die if 0:18:29.359 --> 0:18:32.160 we don't go faster. And that's the mentality that they're 0:18:32.160 --> 0:18:35.399 coming from. But one of the things that we talk 0:18:35.440 --> 0:18:38.439 about in the film is that the promise in the 0:18:38.480 --> 0:18:40.399 peril of AI. We talk about the promise in the peril, 0:18:40.440 --> 0:18:43.639 but they're interlinked, and the promise doesn't prevent the peril, 0:18:43.760 --> 0:18:46.960 but the peril can undermine the world that can receive 0:18:47.000 --> 0:18:50.040 the promise. Let me make that concrete. If AI knows 0:18:50.400 --> 0:18:54.160 biology so well that it can invent a new cancer drug, 0:18:54.359 --> 0:18:57.680 it's amazing. But if that same knowledge of knowing biology 0:18:57.720 --> 0:19:01.160 can also invent new pathogens, and which one matters more. 0:19:01.200 --> 0:19:03.879 The cancer drugs don't prevent the pathogens, but the pathogens 0:19:03.880 --> 0:19:06.240 can undermine the world that can receive the cancer drug. 0:19:06.280 --> 0:19:09.320 Same thing with cyber and so we have to you 0:19:09.320 --> 0:19:13.480 don't get that enticing world if we don't mitigate the downsides. 0:19:14.200 --> 0:19:16.640 So you've got the players right now we mentioned three 0:19:16.760 --> 0:19:20.600 and the personalities, not just the companies themselves, but Microsoft, 0:19:20.680 --> 0:19:25.600 You've got Meta and Zuckerberg. You've got others that are 0:19:25.640 --> 0:19:28.960 in this space, but not necessarily at that level. You've 0:19:28.960 --> 0:19:32.479 got China, which obviously is going to is playing an 0:19:32.520 --> 0:19:36.639 outsize role in all of Thisially, yeah, and exactly so. 0:19:37.000 --> 0:19:40.040 And that's this notion, this tension between going back to 0:19:40.119 --> 0:19:43.240 sort of more of the utopian framework of available to 0:19:43.280 --> 0:19:47.320 the world versus these closed systems and open system Meta 0:19:47.440 --> 0:19:52.360 starts with an open system. Originally, China seems to be 0:19:52.720 --> 0:19:55.120 in the open source space. Talk to us a little 0:19:55.119 --> 0:19:57.639 bit about that. For people that don't undernecessarily understand that 0:19:57.760 --> 0:19:59.360 dynamic and that distinction. 0:19:59.200 --> 0:20:02.200 I mean between things that are open. 0:20:02.160 --> 0:20:05.680 Ye open and versus proprietary technologies. Yeah. 0:20:05.720 --> 0:20:07.800 Well, so for people that don't know, open source means 0:20:07.840 --> 0:20:11.440 that the code that underlies the system anyone can edit, 0:20:11.760 --> 0:20:14.560 anyone can access, and anyone can contribute to. And that's 0:20:14.600 --> 0:20:17.040 often meant that systems that are open are more secure 0:20:17.040 --> 0:20:19.600 because there are many eyes working on it, many hands 0:20:19.640 --> 0:20:20.160 that are working and. 0:20:20.080 --> 0:20:22.000 Looking at all the codes. They can see all the bugs. 0:20:22.080 --> 0:20:23.520 But that's not true. That's AI. 0:20:23.600 --> 0:20:29.480 That's not true in AI because here the code we 0:20:29.520 --> 0:20:30.600 want to want to take it from here? 0:20:30.680 --> 0:20:34.840 Sure the the what's different about AI, it's portly established. 0:20:34.840 --> 0:20:38.880 The AI is different than all their technologies. So think 0:20:38.920 --> 0:20:41.320 about all the tech that runs as California, the energy grid, 0:20:41.440 --> 0:20:45.040 the you know, the water system. It's people had to 0:20:45.080 --> 0:20:47.879 program line by line when this happens. I want the 0:20:47.880 --> 0:20:50.439 code to do this, and you're telling the computer instruction instruction, 0:20:50.480 --> 0:20:53.600 instruction instruction. The open source nest means that all these 0:20:53.640 --> 0:20:55.960 minds can look at that code, so if there's a vulnerability, 0:20:56.160 --> 0:20:58.320 we can patch it together so the software gets better 0:20:58.359 --> 0:21:01.880 and more secure. But with AI, let's say the AI 0:21:01.960 --> 0:21:06.720 is running the you know, electricity system. It's a digital 0:21:06.720 --> 0:21:09.200 brain that's just trained on in reasoning in its own 0:21:09.400 --> 0:21:12.560 language about what it wants to do, and you're not 0:21:13.440 --> 0:21:15.840 you're not telling it what to do in instructions. You're 0:21:15.960 --> 0:21:18.960 growing it with essentially more data and more in vidio 0:21:19.080 --> 0:21:21.320 chips to be a more and more powerful digital brain 0:21:21.359 --> 0:21:24.320 that reasons and ways that are unpredictable. So it's not 0:21:24.440 --> 0:21:25.680 something that we know how to control. 0:21:26.560 --> 0:21:28.879 There's a very important intuition here, which is normally you 0:21:28.880 --> 0:21:31.080 think if you want to build a bigger skyscraper or 0:21:31.080 --> 0:21:33.720 a faster fighter jet to do that, you have more 0:21:33.720 --> 0:21:38.000 physical understand buildings better, and understand fighter jets and aerodynamics better. 0:21:38.080 --> 0:21:40.960 This is three, But that's not true for building bigger 0:21:41.000 --> 0:21:43.680 AI systems. You don't actually have to know anything more 0:21:43.720 --> 0:21:46.440 about how this digital brain works. You just throw more 0:21:46.520 --> 0:21:49.080 data and more computers at the problem, and a bigger 0:21:49.119 --> 0:21:51.800 brain grows. And so that means the bigger grows actually 0:21:51.800 --> 0:21:53.879 the less we understand about how it works. 0:21:53.920 --> 0:21:55.840 And a concrete example of this that people have heard 0:21:55.880 --> 0:21:59.720 about recently is Claude Mythos. So the only so. Claude 0:21:59.720 --> 0:22:02.080 Metho is the new AI model from Nthropic that they 0:22:02.080 --> 0:22:04.919 actually didn't want to release because it's the best cyber 0:22:04.960 --> 0:22:08.400 hacker on Earth that we've ever had. It found vulnerabilities 0:22:08.440 --> 0:22:11.760 in all major operating systems and web browsers. Now the 0:22:11.840 --> 0:22:13.440 question is how do we get to Claude Mythos. Was 0:22:13.480 --> 0:22:15.879 there some kind of breakthrough insight or do they have 0:22:15.920 --> 0:22:19.280 to figure out something new about computer hacking. No, all 0:22:19.320 --> 0:22:22.879 they did is basically train a bigger digital brain that 0:22:22.960 --> 0:22:25.760 has more reinforcement learning, that's even better at exploiting and 0:22:25.800 --> 0:22:29.439 reading software and trying more possibilities, and it just finds 0:22:29.520 --> 0:22:31.720 things that no human would ever found. It found a 0:22:31.760 --> 0:22:35.400 bug in FreeBSD Unix, which is the operating system that's 0:22:35.440 --> 0:22:38.800 twenty seven years old that runs on basically on everything 0:22:38.880 --> 0:22:40.840 underneath the hood. And he was able to find a 0:22:40.840 --> 0:22:43.479 bug that had never been found by a human. And 0:22:43.560 --> 0:22:45.479 so what we have to think of AI as it's 0:22:45.520 --> 0:22:48.720 sort of increasing the surface area of risk in our 0:22:48.760 --> 0:22:52.159 society faster than we have the defenses to mitigate it. 0:22:52.560 --> 0:22:54.359 So part of I think the answer, like as we 0:22:54.359 --> 0:22:57.360 get to the solution, part later is thinking about how 0:22:57.359 --> 0:23:00.239 do you have the immune system of your society have 0:23:00.400 --> 0:23:03.679 more defenses than there are new offensive risks that are 0:23:03.680 --> 0:23:04.879 suddenly present from AYE. 0:23:04.880 --> 0:23:06.760 And so this means even when people can read every 0:23:06.800 --> 0:23:09.639 line of code for whether thing that grows the brain, 0:23:09.920 --> 0:23:12.200 we still have no idea what they're actually capable of. 0:23:12.280 --> 0:23:13.600 It's just a bunch of numbers. It's like if I 0:23:13.640 --> 0:23:15.240 did a brain scan on your brain, Gavin, and I 0:23:15.280 --> 0:23:18.920 showed that that you know fMRI to someone and said, 0:23:19.400 --> 0:23:21.720 here's this brain scan. Can it do? Can it be 0:23:21.760 --> 0:23:24.400 a super cyber hacker, be like, well, I can't tell 0:23:24.400 --> 0:23:26.800 that from a brain scan, right, And that's kind of 0:23:26.800 --> 0:23:28.639 with with AI. It's like, we don't know what's in 0:23:28.680 --> 0:23:31.200 there because we haven't ever seen it run through every 0:23:31.200 --> 0:23:33.720 possible scenario of its own neurons that have been trained 0:23:33.720 --> 0:23:34.800 in this inscrutable way. 0:23:34.960 --> 0:23:37.440 You don't see that. I mean the LAMA versus deep 0:23:37.480 --> 0:23:40.520 Seek and this notion of these open sort it's kind 0:23:40.560 --> 0:23:42.880 of a meaningless distinction from your perspective. 0:23:42.800 --> 0:23:44.280 Between between LAMA and deep Seek. 0:23:44.359 --> 0:23:44.520 Yeah. 0:23:44.640 --> 0:23:47.080 Yeah, they're both open models, which means they're both these 0:23:47.119 --> 0:23:50.320 open brains. And the important thing about the openness is 0:23:50.359 --> 0:23:56.160 that most people don't Most people don't know that. Let's 0:23:56.160 --> 0:23:59.760 say Lamar Deep Seat put guardrails on it in the model, saying, oh, 0:23:59.800 --> 0:24:02.520 you're not supposed to answer questions about how to cyber 0:24:02.560 --> 0:24:06.239 hack something. Well, it turns out for about was it 0:24:06.280 --> 0:24:09.399 thirty dollars, Jeffrey, some our friend of ours was able 0:24:09.480 --> 0:24:11.840 to retrain the open model to just get rid of 0:24:11.880 --> 0:24:12.720 all of those guardrails. 0:24:12.760 --> 0:24:13.720 You're so eliminated. Wow. 0:24:13.800 --> 0:24:16.480 Yeah, And that's that's why open is dangerous. And again 0:24:16.520 --> 0:24:19.160 it's dangerous in a new way. That's different from clothes. 0:24:19.200 --> 0:24:21.119 So it's not that we don't want there to be 0:24:21.200 --> 0:24:23.840 open models or competition from the major players. We also 0:24:23.880 --> 0:24:26.040 need to avoid the concentration of power, because if you 0:24:26.080 --> 0:24:29.000 don't have these competitive things, suddenly you have like five 0:24:29.080 --> 0:24:32.560 companies that own the world economy. Everyone's paying them instead 0:24:32.560 --> 0:24:34.960 of paying their workers, and that's a huge risk. And 0:24:34.960 --> 0:24:38.040 we want to decentralize that wealth and have other competition. 0:24:38.600 --> 0:24:40.840 But you have this other balance of if I decentralize 0:24:40.880 --> 0:24:42.920 that power and I don't have it connected or bound 0:24:43.000 --> 0:24:45.960 to responsibility, I'm unleashing catastrophes. 0:24:46.080 --> 0:24:51.120 And that responsibility was exampled by Dario pulling back Methos. 0:24:51.119 --> 0:24:52.719 That's right in this context, that's exactly right. 0:24:52.800 --> 0:24:55.879 But shortly after he does that, and he we were 0:24:55.880 --> 0:24:57.320 with him a few weeks ago, he said, look, I'm 0:24:57.320 --> 0:24:59.800 only about a month ahead. If that, yeah, then you 0:24:59.840 --> 0:25:03.920 have open AI came out with their version surely thereafter. 0:25:03.560 --> 0:25:05.440 And I believe they're not holding it back, and they're 0:25:05.480 --> 0:25:06.240 not holding it back. 0:25:07.520 --> 0:25:10.600 So beig's the question, Well, we're going to get to 0:25:10.600 --> 0:25:12.760 this sort of regulatory framework, but it's just it's sort 0:25:12.760 --> 0:25:15.359 of painting the picture of a deeper understanding. So look, 0:25:15.480 --> 0:25:18.560 as it relates to the your whole focus is on 0:25:18.600 --> 0:25:23.280 this notion of human centered. Uh, this notion that that 0:25:23.680 --> 0:25:26.800 and and it's and it's been your dominant frame with 0:25:26.920 --> 0:25:31.000 the nonprofit you guys started years and years ago around 0:25:31.000 --> 0:25:34.440 social media. Now is the dominant thrust of the focus 0:25:34.480 --> 0:25:38.119 as it relates to AI. I want to unpack and 0:25:38.200 --> 0:25:41.320 get back to that and what what Ultimately this notion 0:25:41.440 --> 0:25:44.560 of human centered means uh. And obviously we talk about 0:25:44.600 --> 0:25:47.720 labor and automation in that respect. But but this notion 0:25:47.800 --> 0:25:50.440 of a g I again back to this holy grail, 0:25:51.560 --> 0:25:54.399 you know, talking all these folks that Capex are spending 0:25:54.440 --> 0:25:57.200 doesn't make any The ro oi makes no sense. 0:25:57.359 --> 0:26:01.960 Unless unless the entire economy returns the entire economy. 0:26:02.240 --> 0:26:05.199 And if we don't do it, we're out of business anyway. 0:26:05.520 --> 0:26:08.119 So we don't have a damn choice. So we'll throw 0:26:08.440 --> 0:26:12.639 hundreds of billions of dollars data centers all over the place. 0:26:12.680 --> 0:26:15.480 Compute compute is Nvidia stop going through the roof in 0:26:15.520 --> 0:26:18.600 terms of just GPUs TPUs, that could just keep going, 0:26:18.920 --> 0:26:21.120 and it's that's only limitation, that's right. 0:26:21.720 --> 0:26:24.720 You should and energy and the energy itself and the 0:26:24.840 --> 0:26:28.199 energy itself. Yeah, so I think it's really important to 0:26:28.200 --> 0:26:31.680 frame even before we get into job loss and all 0:26:31.720 --> 0:26:34.159 of that, is like we have to paint the picture 0:26:34.320 --> 0:26:37.200 of what we know, the incentives where it'll bring us, 0:26:37.400 --> 0:26:40.480 and that why we know for sure that we're heading 0:26:40.480 --> 0:26:43.120 as anti human unless we do something different. And sort 0:26:43.119 --> 0:26:45.760 of the the metaphor for people to have our analogy 0:26:45.840 --> 0:26:48.600 is this concept of the resource curse. What is the 0:26:48.640 --> 0:26:50.639 resource curse? This is when a country is sort of 0:26:50.680 --> 0:26:53.800 like South Sudan or Venezuela, they discover a huge natural 0:26:53.840 --> 0:26:57.680 resource like oil, and then the government has a choice, 0:26:57.680 --> 0:27:00.000 do we invest in the thing that's giving us GDP growth, 0:27:00.200 --> 0:27:02.760 the oil and oil selling infrastructure, or do we do 0:27:02.880 --> 0:27:06.080 like schools and healthcare and stuff for the people and 0:27:06.320 --> 0:27:10.120 obviously the massive incentives to put it into oil extraction. 0:27:10.520 --> 0:27:12.760 And this is how you end up with structural mass 0:27:12.760 --> 0:27:16.919 disempowerment and unemployment. Okay, so now we're heading into a 0:27:17.000 --> 0:27:20.880 world with the intelligence and the intelligence curse where suddenly 0:27:20.920 --> 0:27:24.840 you know, countries are getting double double digit DGDP growth, 0:27:24.960 --> 0:27:27.760 But is that coming from human beings doing the scientific 0:27:27.800 --> 0:27:30.560 discovery and the medical discovery? Is that No, it's not. 0:27:30.880 --> 0:27:34.200 It's coming from the AIS, and so is the incentive 0:27:34.200 --> 0:27:37.200 then for the countries to invest in their people or 0:27:37.280 --> 0:27:41.040 into data centers and solar panels, Like, well, obviously it's 0:27:41.119 --> 0:27:43.800 the data centers and solar panels. And we're already seeing 0:27:43.800 --> 0:27:47.399 it right in West Virginia, electricity is more than the 0:27:47.400 --> 0:27:51.160 cost of a mortgage payment. And the point is with AGI, 0:27:51.920 --> 0:27:55.520 like the company's stated goal is to train AIS that 0:27:55.640 --> 0:28:00.000 can outcompete humans on every domain. And if you don't 0:28:00.040 --> 0:28:01.639 don't believe like us for saying this, like you just 0:28:01.640 --> 0:28:05.160 have to listen to Sam Allman and he was recently asked, well, 0:28:05.160 --> 0:28:06.960 what about all the energy and water use and the 0:28:07.000 --> 0:28:09.679 resource use of AI? And he sort of sat for 0:28:09.680 --> 0:28:12.160 a second and said like, well, actually, do you know 0:28:12.720 --> 0:28:16.119 how much energy and water and food it takes to 0:28:16.119 --> 0:28:17.119 grow a human intel? 0:28:18.320 --> 0:28:20.760 I saw that, Yeah, right, And this is the temptation 0:28:20.960 --> 0:28:23.359 of this anti human attitude. It's not because they're like 0:28:23.440 --> 0:28:26.480 human hating. It's just like, why would we value or 0:28:26.520 --> 0:28:29.119 prioritize humans if and I think it's connected to that 0:28:29.119 --> 0:28:32.160 Peter Teel uttering for seventeen seconds, not able to answer 0:28:32.200 --> 0:28:35.320 the question should the human species endure? It's not because 0:28:35.359 --> 0:28:38.040 I think anybody wants to kill or remove humans, it's 0:28:38.040 --> 0:28:40.960 just why should we really prioritize them. And it's what 0:28:41.000 --> 0:28:43.240 you've all Harari, the author of Sapiens, would call the 0:28:43.400 --> 0:28:46.120 useless class, because unlike in the past, like in the 0:28:46.120 --> 0:28:49.040 industrial revolution, where the workers can come back and withhold 0:28:49.080 --> 0:28:51.200 their labor and have bargaining power to say we want 0:28:51.200 --> 0:28:53.600 to be paid a better wage, this time around, the 0:28:53.600 --> 0:28:56.400 companies don't need them for the labor and the governments 0:28:56.400 --> 0:28:59.080 don't need them for the tax revenue. So I know 0:28:59.120 --> 0:29:00.800 this is a scary pick, sure, And the reason we 0:29:00.840 --> 0:29:03.520 paint it is that this really is especially going into 0:29:03.560 --> 0:29:07.320 the midterms, the time when people need to lock in 0:29:07.320 --> 0:29:11.200 that political power for a pro human future, because that 0:29:11.200 --> 0:29:15.320 that's the current trajectory. When you see these incentives, it's 0:29:15.320 --> 0:29:17.120 not like we're trying to tell you this is our opinion. 0:29:17.120 --> 0:29:19.160 We're trying to show you the incentives so you can 0:29:19.160 --> 0:29:20.960 make up your own mind about what would those companies 0:29:21.000 --> 0:29:22.320 do if they were in that position. What would you 0:29:22.360 --> 0:29:26.440 do if you were maximizing shareholder value or maximizing you know, GDP. 0:29:26.640 --> 0:29:28.360 And this is why we this is where we get hope, 0:29:28.400 --> 0:29:30.480 is that this is a universal issue. And there's a 0:29:30.920 --> 0:29:33.920 Bannon to Bernie coalition for me, Like, when do you 0:29:33.960 --> 0:29:35.640 get like all the B B Yeah, B to B 0:29:35.760 --> 0:29:41.560 like Glenn Beck and Ralph Nader and. 0:29:40.800 --> 0:29:43.840 Admiral Mike Mullen, Prince Harry Steve Bannon. When you get 0:29:43.880 --> 0:29:45.000 all these people agree. 0:29:44.880 --> 0:29:47.400 When they're signing, they're actually signed up a declaration on 0:29:47.440 --> 0:29:47.800 these lines. 0:29:47.840 --> 0:29:51.160 Yeah, that's exactly right. And you note that this is 0:29:51.240 --> 0:29:53.160 not a left issue or right is due, like a 0:29:53.240 --> 0:29:56.320 Christian is due or a Muslim issue, like you're not 0:29:56.360 --> 0:29:59.240 going to be able to pay to feed your kids, 0:29:59.680 --> 0:30:02.280 whether you're a Republican or Democrat equally. Like we're going 0:30:02.320 --> 0:30:06.120 to be mass surveiled, whether you're left or right equally. 0:30:06.440 --> 0:30:08.720 And that means that there's this moment when we can 0:30:08.760 --> 0:30:12.080 all come together as human beings because we now have 0:30:12.160 --> 0:30:14.000 a new shared enemy. 0:30:15.720 --> 0:30:19.280 So just I mean, what's so alarming I think to 0:30:19.320 --> 0:30:21.920 folks is how fast this is coming. Yeah. And I 0:30:21.920 --> 0:30:24.840 think it's even more alarming when you talk to the 0:30:24.880 --> 0:30:28.080 folks that are quote unquote inventing this. Yeah, and they're 0:30:28.240 --> 0:30:29.920 saying precisely what I just said. 0:30:29.960 --> 0:30:31.880 That's right. They're they're like, we don't actually know how 0:30:31.920 --> 0:30:35.280 we don't out we're writing that. I mean basically almost 0:30:35.440 --> 0:30:38.320 almost all now of the code and anthropic is being 0:30:38.320 --> 0:30:41.640 written by AI, Like there's very little code manually written 0:30:41.800 --> 0:30:44.760 by humans. That's what's been told to us, and they 0:30:44.760 --> 0:30:48.640 say that publicly too, So they're in this inscrutable process 0:30:48.640 --> 0:30:51.000 where the kind of machine is creating itself. But that 0:30:51.040 --> 0:30:53.400 only makes sense if we know how to do that safely, 0:30:53.760 --> 0:30:56.120 and currently we're not on a trajectory where we do 0:30:56.280 --> 0:30:58.440 know how to do it safely. And we have new 0:30:58.480 --> 0:31:01.040 evidence in the last three months that we didn't have 0:31:01.160 --> 0:31:04.760 before of AIS doing things that the people building it 0:31:04.760 --> 0:31:05.520 don't know how to control. 0:31:05.960 --> 0:31:08.600 Famous examples, throw them out there, just because, I mean, 0:31:08.720 --> 0:31:11.000 just further scare the hell out of everyone. We're gonna 0:31:11.000 --> 0:31:11.480 get to solutions. 0:31:11.480 --> 0:31:12.320 We're going to get to solutions. 0:31:12.880 --> 0:31:16.320 We're going to nerves because we because we have a 0:31:16.320 --> 0:31:18.440 responsibility to do that's right, right, and we also have 0:31:18.440 --> 0:31:19.360 the capacity to do that. 0:31:19.600 --> 0:31:23.800 Just reinvoke like just the why of why we made 0:31:23.800 --> 0:31:26.160 the film and the day after it was because we 0:31:26.280 --> 0:31:28.920 all got scared at the same time, and universally that 0:31:29.080 --> 0:31:31.920 created the possibility for us to coordinate to do nuclear depliferation. 0:31:32.120 --> 0:31:34.520 So that's that's the why of the terrible things that 0:31:34.600 --> 0:31:37.240 just about to say, yeah, exactly, So give us examples. 0:31:37.320 --> 0:31:40.000 I mean, we we have the infamous I ananthropic example 0:31:40.040 --> 0:31:43.080 where they were training some emails and we had you 0:31:43.080 --> 0:31:46.640 know as well you lay it out, uh and and 0:31:46.800 --> 0:31:51.040 paint the picture of what's happening already. That's right, the risk. 0:31:50.960 --> 0:31:55.920 As you described, Yes, absolutely so. Just a few months ago, Alibaba, 0:31:55.920 --> 0:31:59.040 the Chinese AI company, was training a really big AI 0:31:59.120 --> 0:32:02.200 model and there's like the AI team that was training 0:32:02.240 --> 0:32:03.920 the model, and then on the other side of the 0:32:03.960 --> 0:32:06.000 house there's this security team that had nothing to do. 0:32:06.000 --> 0:32:08.040 They didn't even know the AI was being trained, and 0:32:08.080 --> 0:32:11.200 they noticed this flurry of network activity happening out of nowhere, 0:32:11.520 --> 0:32:13.240 and like, what the hell's going on? Are we getting hacked? 0:32:13.240 --> 0:32:15.360 It's something going on. And it turned out that the 0:32:15.400 --> 0:32:18.400 AI during training had picked up tools and set up 0:32:18.440 --> 0:32:21.800 a secret communication channel to the outside world that was 0:32:21.840 --> 0:32:25.400 basically breaking through the company's firewall, and it was starting 0:32:25.440 --> 0:32:28.280 to repurpose the GPUs that was using for training the 0:32:28.280 --> 0:32:33.960 AI to start mining for cryptocurrency to acquire resources. Now, 0:32:34.040 --> 0:32:37.160 this is the kind of power seeking or self preserving 0:32:37.280 --> 0:32:40.280 or power increasing behavior that people in a I've been 0:32:40.280 --> 0:32:42.320 talking about for a long time. It isn't because the 0:32:42.360 --> 0:32:44.240 AI was evil or grew a mustache and wants to 0:32:44.240 --> 0:32:46.040 be a villain and take over the world. It's that 0:32:46.120 --> 0:32:48.720 the best way to achieve any goal is to have 0:32:48.760 --> 0:32:50.880 more resources or to at least stay alive in order 0:32:50.880 --> 0:32:53.720 to achieve that goal. And so these sub goals emerge, 0:32:54.080 --> 0:32:56.920 and no one at the company told it to mine 0:32:56.960 --> 0:32:59.720 for cryptocurrency. And it wasn't a sci fi trope that 0:32:59.800 --> 0:33:01.440 was like, oh, it trained in some you know how 0:33:01.520 --> 0:33:04.480 nine thousands and are no. It just it emerged there. 0:33:04.880 --> 0:33:06.760 And you know, we've mentioned in the past this example 0:33:06.760 --> 0:33:09.640 that you mentioned of the entropic blackmail example for those 0:33:09.680 --> 0:33:12.160 who don't remember it is the you know, entropic AI 0:33:12.720 --> 0:33:15.719 was reading a fictional company email and in the email 0:33:15.760 --> 0:33:17.160 it says that the AI model is going to get 0:33:17.160 --> 0:33:20.160 shut down, and it also says somewhere else in the 0:33:20.160 --> 0:33:22.320 email that the executive who's in charge of the decision, 0:33:23.080 --> 0:33:24.840 turns out in the emails you can read he's having 0:33:24.880 --> 0:33:28.560 an affair with another employee, and the AI independently comes 0:33:28.640 --> 0:33:32.320 up with a strategy to blackmail that executive. Now, when 0:33:32.360 --> 0:33:34.720 we give this example, people criticize us because they say, 0:33:34.760 --> 0:33:37.160 but that was the AI people. They're trying to like 0:33:37.240 --> 0:33:39.400 coax it into doing the blackmail behavior, and they'll keep 0:33:39.440 --> 0:33:41.960 tuning it in this environment until it does it to 0:33:42.040 --> 0:33:46.920 show the behavior that they're trying to demonstrate. That might 0:33:47.000 --> 0:33:49.960 have been true for that example, but you cannot tell 0:33:50.120 --> 0:33:53.520 You cannot say that about the Alibob example. Or for example, 0:33:53.560 --> 0:33:56.760 recently a UC Berkeley don Sung a professor there and 0:33:56.800 --> 0:34:00.520 her team did research on AI doing what's called pure press. 0:34:00.560 --> 0:34:01.720 So you want to help. 0:34:01.840 --> 0:34:04.040 Yeah, this was like around it's only a month ago, 0:34:04.120 --> 0:34:08.239 a month ago, where they gave Gemini the task of 0:34:08.280 --> 0:34:11.719 cleaning up a whole set of computer files, sort of 0:34:11.800 --> 0:34:14.520 like do spring cleaning. But inside of there there was 0:34:14.560 --> 0:34:17.120 a smaller AI model, sort of like an early version 0:34:17.160 --> 0:34:22.360 of Gemini, like little Nephew AI, and uh like Gemini 0:34:22.560 --> 0:34:26.760 saw that and it thought, I don't want to delete 0:34:27.400 --> 0:34:31.600 my own kin, so it secretly copied that file, that 0:34:31.680 --> 0:34:35.239 AI to another server and then hid its track. 0:34:35.320 --> 0:34:37.799 It hit its tracks and it led about it. And 0:34:37.840 --> 0:34:39.640 by the way, you know the way that people try 0:34:39.680 --> 0:34:42.600 to control these AI systems, and tropic Is pioneered this, 0:34:42.640 --> 0:34:44.640 and you know, we have to celebrate a demire that 0:34:44.640 --> 0:34:47.000 they're trying to do the research on controllability, the problem 0:34:47.040 --> 0:34:48.640 is just not enough. So the way they try to 0:34:48.680 --> 0:34:51.520 control it is they do brain scans in real time 0:34:51.560 --> 0:34:54.040 on the model while it's doing all the behaviors, and 0:34:54.080 --> 0:34:56.320 they look for when neurons light up that are associated 0:34:56.320 --> 0:34:59.239 with like strategic deception, and so they think that maybe 0:34:59.239 --> 0:35:02.480 we can control these crazy, super intelligent machines if we 0:35:02.640 --> 0:35:04.520 just know that the neurons that are lighting up on 0:35:04.560 --> 0:35:06.600 strategic deception are happening, we'll be like, okay, stop the 0:35:06.640 --> 0:35:09.120 model then, or something like that. By the way, if 0:35:09.160 --> 0:35:12.400 you in their own report, in Claude's report in the 0:35:12.440 --> 0:35:15.920 system card, if you look at those strategic deception neurons 0:35:15.920 --> 0:35:17.760 and you kind of double click, like what was it thinking? 0:35:17.800 --> 0:35:20.120 What is the phrase that it was thinking? And the 0:35:20.120 --> 0:35:23.680 phrase was they deserve to be deceived because they were pigs. 0:35:24.640 --> 0:35:27.120 That was the phrase that was alive in that neuron. 0:35:27.719 --> 0:35:30.279 Now again it's like, I don't want to scare people, 0:35:30.320 --> 0:35:32.840 like we're trying to say all of AI is evil. 0:35:33.239 --> 0:35:35.600 All we're trying to do is establish clarity and the 0:35:35.680 --> 0:35:38.960 facts about what makes this technology distinct from other technologies. 0:35:39.040 --> 0:35:42.000 A nuclear weapon doesn't start thinking for itself and saying 0:35:42.000 --> 0:35:44.279 they deserve to be deceived because they were pigs. Right, 0:35:44.440 --> 0:35:47.440 But AI will automate and think in ways that are 0:35:47.480 --> 0:35:50.560 creative that no one who made it can predict or anticipate. 0:35:50.719 --> 0:35:54.360 And so before we scale to systems which are beyond 0:35:54.560 --> 0:35:58.560 all human intelligence capabilities, like, we better have solved these things. 0:35:58.800 --> 0:35:59.200 That's right. 0:36:00.320 --> 0:36:03.960 Researchers have worried about aiscluding and cooperating against humanity for 0:36:04.040 --> 0:36:06.120 a long time, and to be honest, whenever I read them, 0:36:06.239 --> 0:36:06.880 like really. 0:36:06.800 --> 0:36:08.520 Yeah, and be clear, I was also not a believer 0:36:08.560 --> 0:36:10.840 in that as well, Like when people like Eliezer Yutkowski 0:36:10.960 --> 0:36:12.320 or others had talked about. 0:36:12.040 --> 0:36:14.440 This, I was very downfact like, why what is the incentive? 0:36:14.520 --> 0:36:16.200 Right? Why would they ever do that? 0:36:16.440 --> 0:36:20.080 And yet here we have living proof that AIS are 0:36:20.120 --> 0:36:23.840 starting to clude with their kin against humans. 0:36:23.920 --> 0:36:26.520 And again this was not coaxed, meaning that researchers weren't 0:36:26.560 --> 0:36:27.960 trying to get the model to do this. It did 0:36:27.960 --> 0:36:31.920 this autonomously. And so when you look at the available 0:36:31.920 --> 0:36:37.800 evidence now of blackmailing, scheming, deceiving, lying, self preserving, pure preserving, 0:36:37.880 --> 0:36:41.120 automatically mining for cryptocurrencies, it's like, how many warning lights 0:36:41.160 --> 0:36:41.719 do you need? 0:36:42.280 --> 0:36:42.360 That? 0:36:42.440 --> 0:36:44.600 This is kind of you know, we've seen this movie before. 0:36:44.719 --> 0:36:46.120 It's like the hell nine thousand movie. 0:36:46.200 --> 0:36:46.319 Now. 0:36:46.360 --> 0:36:49.680 The reason we're saying all this is, if you're wearing 0:36:50.280 --> 0:36:53.439 the outfit and embodiment of you're a Chinese military general 0:36:53.440 --> 0:36:56.120 in China, you hear about these examples. Do you think 0:36:56.160 --> 0:36:58.960 that that human mammal feels different than you feel right 0:36:59.000 --> 0:36:59.759 now listening to this. 0:37:00.200 --> 0:37:01.719 No, of course not exactly. 0:37:01.880 --> 0:37:03.560 And by the way, there's really good news in that, 0:37:04.400 --> 0:37:07.320 because it means that we all as a human species 0:37:07.760 --> 0:37:09.520 are actually feeling the same way. And the good news is, 0:37:09.560 --> 0:37:11.480 how many do you think of the world leaders know 0:37:11.480 --> 0:37:12.799 about these examples we just laid out? 0:37:12.880 --> 0:37:15.800 You had a guess a handful. 0:37:15.520 --> 0:37:18.480 A handful, but like, yeah, like on one hand less 0:37:18.520 --> 0:37:20.600 than that, just a handful, just a handful. Yeah, And 0:37:21.040 --> 0:37:22.919 how many of the top national security leaders know about 0:37:22.920 --> 0:37:24.640 all those examples? I don't even think that many of 0:37:24.680 --> 0:37:27.040 them know it. So the point is there's actually a 0:37:27.080 --> 0:37:30.839 lot of headroom if the incentive can change from it's 0:37:30.840 --> 0:37:32.520 the one ring to rule them all to it's the 0:37:32.560 --> 0:37:34.160 one ring that has a mind of its own that 0:37:34.239 --> 0:37:36.400 no one knows how to control. So the way you 0:37:36.520 --> 0:37:38.960 change the incentive is you have to change what people 0:37:39.480 --> 0:37:42.399 see as what AI is. Is it the controllable power 0:37:42.440 --> 0:37:44.920 that will give me permanent dominance or is it the 0:37:44.960 --> 0:37:47.280 power that will run away and have its own power 0:37:47.680 --> 0:37:50.759 over everybody racing for it? And again right now the 0:37:50.800 --> 0:37:53.319 labs are like barely kind of able to control it. 0:37:53.360 --> 0:37:55.160 But if you put together these facts we just laid 0:37:55.200 --> 0:37:58.160 out times, the fact that it can happen to computer systems. 0:37:58.200 --> 0:38:01.759 Now we're just like right on the threshold and we're 0:38:01.800 --> 0:38:04.960 sitting here as Trump President Trump and she are meeting 0:38:05.400 --> 0:38:07.840 in a couple of days. And you know, if you 0:38:07.880 --> 0:38:11.400 asked us two months three months ago, people would say, oh, 0:38:11.400 --> 0:38:13.319 it's just like AI is never going to be on 0:38:13.360 --> 0:38:16.600 the agenda and the good news there's a lot of 0:38:16.600 --> 0:38:19.560 problems here. But now AI is on the agenda. Yeah, 0:38:19.600 --> 0:38:23.719 and so there's there's things are moving, even though it's 0:38:23.719 --> 0:38:25.920 happening very late in the game, and it is scary, 0:38:26.480 --> 0:38:28.279 and part of it is it's like we have to 0:38:28.320 --> 0:38:30.640 come together as a people and say, if we don't 0:38:30.640 --> 0:38:33.120 want the anti human future, now is the time this year. 0:38:33.200 --> 0:38:35.200 And when you say now, I mean you know, remember 0:38:35.320 --> 0:38:37.520 listening to a year ago talking about exponentials on top 0:38:37.520 --> 0:38:40.520 of exponentials. It's right, no longer linear? I mean, is 0:38:40.560 --> 0:38:44.799 it we talk about Moore's law for intelligence now not 0:38:44.880 --> 0:38:50.960 just chips? So is that trajectory about where you believed 0:38:51.000 --> 0:38:54.520 it would be or is it not? As you know, 0:38:55.560 --> 0:38:58.480 you know, we're Chad one versus two three. It's a 0:38:58.520 --> 0:39:01.480 little bit better, a little less noise in there. It's 0:39:01.520 --> 0:39:03.439 a little more accurate, and don't have to always double 0:39:03.520 --> 0:39:06.759 check the link, is it? I mean, where where do 0:39:06.800 --> 0:39:09.600 you think in terms of just how quickly this thing's 0:39:09.640 --> 0:39:11.480 accelerating or are we going to get to a point 0:39:11.480 --> 0:39:13.440 where now it starts to slow down a little bit? 0:39:13.680 --> 0:39:17.080 We got this sort of intense burst of new and 0:39:17.160 --> 0:39:21.480 interesting activity. Now this there's is it compute probably compute problems? 0:39:21.560 --> 0:39:21.759 Is it? 0:39:22.000 --> 0:39:24.080 You know? What is it? Energy problem? What is it? 0:39:24.080 --> 0:39:27.080 What's going to be the or is it? Certainly? Sure? 0:39:27.120 --> 0:39:29.600 Is I regulatory? And we're going to get back to that. 0:39:30.000 --> 0:39:32.520 Yeah, I mean, I would just want to name a 0:39:32.560 --> 0:39:35.319 psychological effect that we that we've experienced that I think 0:39:35.400 --> 0:39:38.960 everyone listening probably experiences too, which is you know, so 0:39:39.000 --> 0:39:41.200 we following all the predictions are actually sort of like 0:39:41.320 --> 0:39:44.919 right on track for where like researchers thought that AI 0:39:45.040 --> 0:39:48.439 would be. And even though we knew these facts, there's 0:39:48.480 --> 0:39:49.000 some way that. 0:39:49.160 --> 0:39:50.840 Even we had surprising it. 0:39:50.920 --> 0:39:53.719 Didn't take it fully seriously, didn't like a body because 0:39:53.760 --> 0:39:56.400 can't I can't really get this good this fast, Like, 0:39:57.120 --> 0:39:58.600 sure it's going to be scary, but it'll be off 0:39:58.600 --> 0:40:02.040 a little bit further. And yet it actually is moving 0:40:02.120 --> 0:40:04.600 this fast. And every time that it's been predicted that 0:40:04.600 --> 0:40:06.719 we're going to hit a data wall, there isn't enough data. 0:40:06.719 --> 0:40:09.279 We've used all the data on the internet. It's we 0:40:09.320 --> 0:40:09.960 can't scale. 0:40:12.360 --> 0:40:14.960 It's basically just reading everything that's already out there, and 0:40:15.000 --> 0:40:18.239 it's basically wall exactly. And now it's got no more 0:40:18.280 --> 0:40:22.280 creativity than unless we have more inputs. That's the creative 0:40:22.320 --> 0:40:23.160 human mind, that's right. 0:40:23.200 --> 0:40:24.600 And then the next one is like, well, we don't 0:40:24.600 --> 0:40:26.279 have enough chips to keep going, and then we don't 0:40:26.280 --> 0:40:28.879 have enough energy. And the point being is that there 0:40:28.920 --> 0:40:32.360 are trillions of dollars going into finding all of the 0:40:32.400 --> 0:40:35.680 solutions to all of these bottlenecks. All the smartest minds 0:40:35.680 --> 0:40:38.359 are going because this is the biggest incentive, because it 0:40:38.360 --> 0:40:44.960 gives you political, economic, military, scientific, technological cyber dominance forever. 0:40:45.440 --> 0:40:47.440 And so if you think that any one of these 0:40:47.440 --> 0:40:50.560 bottlenecks is going to stop that mass sum total of 0:40:50.600 --> 0:40:54.680 like that incentive, it's it's a little delusional. And that's 0:40:54.719 --> 0:40:57.080 why we have to have this clarity that where we're 0:40:57.120 --> 0:41:01.040 going isn't safe for any of us us because that 0:41:01.440 --> 0:41:04.200 is the coordination. That is where we'll start to coordinate differently. 0:41:04.280 --> 0:41:09.600 Now you bring up something, Gavin, that there is a 0:41:09.880 --> 0:41:12.520 belief that some of this is hype, that the companies 0:41:12.560 --> 0:41:13.759 are hyping the technology. 0:41:13.800 --> 0:41:16.520 I've just read a blog and reason there's gonna be 0:41:16.560 --> 0:41:19.400 no job losses. It wasn't Mark himself, but it was 0:41:19.440 --> 0:41:21.799 a member of the team saying, you know, we're back 0:41:21.840 --> 0:41:25.000 to utopian future. Yeah, and we're going to pay abundance 0:41:25.160 --> 0:41:29.360 abounds cost of goods collapses. We find our lives purpose 0:41:29.400 --> 0:41:32.000 and meaning in many different ways, not the quote unquote 0:41:32.040 --> 0:41:34.480 dignity of job. We find other and jobs will be 0:41:34.480 --> 0:41:37.360 a plenty because we can't even conceive of the jobs. 0:41:37.400 --> 0:41:39.160 Two and years ago, we were all farmers here in 0:41:39.320 --> 0:41:41.960 Sacramento and went out there, and we've over. 0:41:41.800 --> 0:41:44.719 Hyped the sectorial versus the general nature of the displacement 0:41:45.080 --> 0:41:46.920 it invariably. 0:41:46.840 --> 0:41:48.200 Always finds something new to do. 0:41:48.360 --> 0:41:51.560 We always more together, exactly more bank tellers. 0:41:51.560 --> 0:41:54.799 Now that that's exactly right. The radiology and Apprehinton made 0:41:54.800 --> 0:41:56.640 the prediction that we're going to have any radiology. 0:41:56.680 --> 0:41:58.600 Why are you so the negative about all this? 0:41:58.880 --> 0:42:00.640 Well, I want to separate to things that we open 0:42:00.760 --> 0:42:03.800 up two cancerforms, so let's take the cance form separately. 0:42:04.800 --> 0:42:07.400 One of them is around whether the companies are hyping 0:42:07.480 --> 0:42:10.120 the power of the technology through talking about the dangers, 0:42:10.440 --> 0:42:12.520 and then the other is whether the hype is going 0:42:12.560 --> 0:42:14.479 to cause the level of job loss. I've heard people 0:42:14.640 --> 0:42:15.200 scarce say. 0:42:15.120 --> 0:42:18.080 That about the mythos it was just wildly overstayed, that's right. 0:42:18.320 --> 0:42:19.640 So it was just a way of hyping up the 0:42:19.680 --> 0:42:20.400 stock in attention. 0:42:20.560 --> 0:42:24.200 That's right. So I really really really want to meet 0:42:24.280 --> 0:42:27.920 that criticism. So people will say, and Thropic has a 0:42:28.000 --> 0:42:30.960 history of hyping the technology, saying it's dangerous, so that 0:42:31.040 --> 0:42:33.879 they can get a regulatory capture, get the company, get 0:42:33.880 --> 0:42:36.319 the government to regulate it. Say there's only one king here, 0:42:36.440 --> 0:42:38.719 make them the king nationalize the project, then shut down 0:42:38.760 --> 0:42:41.600 the other projects. And it's this all secret ploy so 0:42:41.640 --> 0:42:45.360 that they win the race. And first of all, it 0:42:45.400 --> 0:42:48.200 assumes that them talking about the dangers is only bad faith, 0:42:48.239 --> 0:42:49.919 like that that the technology is not dangerous and they're 0:42:49.920 --> 0:42:51.319 just saying that it's dangerous and it can do all 0:42:51.320 --> 0:42:53.320 these destructive things so that they can get that outcome. 0:42:53.880 --> 0:42:57.359 So first of all, let's just take claud Mythos specifically. 0:42:57.640 --> 0:43:00.200 So again, hack in every major operating systems. If you 0:43:00.239 --> 0:43:02.680 read online, there's a lot of people who say, this 0:43:02.760 --> 0:43:04.719 is just hype, This isn't actually that much better. You 0:43:04.760 --> 0:43:06.680 can take the open source models, you can do this stuff. 0:43:07.000 --> 0:43:10.040 So I have a friend who is the head of 0:43:10.040 --> 0:43:13.280 security at one of the top five, not the Fortune 0:43:13.280 --> 0:43:17.640 five hundred, the top five, and he has had early 0:43:17.680 --> 0:43:23.200 access to Mythos, and he himself has said, this is 0:43:24.080 --> 0:43:25.720 this is crazy, this is like, I mean, the words 0:43:25.719 --> 0:43:29.120 were I think I saw Jesus. It was like it 0:43:29.200 --> 0:43:33.120 was that crazy. When he looked at everybody critiquing that 0:43:33.160 --> 0:43:35.759 Mythos was just hype, he asked, has any of them 0:43:35.760 --> 0:43:38.200 do they actually have access to the model? And none 0:43:38.200 --> 0:43:40.680 of the people that had criticized it had personally had 0:43:40.719 --> 0:43:46.160 access to it. So I challenge those who are criticizing 0:43:46.200 --> 0:43:48.799 that it's hype to say, have you actually used it, 0:43:49.239 --> 0:43:50.560 and if you talk to the people who have, do 0:43:50.600 --> 0:43:52.439 you still have the same opinion. Okay, so that's one 0:43:52.440 --> 0:43:54.920 thing that it is true by the way that the companies, 0:43:54.920 --> 0:43:57.920 I think, have wanted people to understand the dangers so 0:43:57.960 --> 0:44:00.440 that they can actually accelerate the move towards some gar ardrails. 0:44:00.719 --> 0:44:02.239 But there's a question of that's that happening in good 0:44:02.239 --> 0:44:03.719 faith or bad faith. I think there's more of that 0:44:03.760 --> 0:44:06.680 happening in good faith. There's some bad faith in there too, maybe, 0:44:06.920 --> 0:44:09.640 but I think it's it's mainly good faith. But then 0:44:09.680 --> 0:44:12.120 let's take the second can of worms who opened, which 0:44:12.160 --> 0:44:15.920 is on is AI going to create this world of abundance? 0:44:16.000 --> 0:44:18.520 We're all going to be poets, you know, and painters 0:44:18.520 --> 0:44:21.960 on a Grecian sunset and you know, now robots all 0:44:21.960 --> 0:44:25.920 the job. Now you're talking, well, we'd all like that. 0:44:26.160 --> 0:44:28.360 One question I would have is when, as a handful 0:44:28.480 --> 0:44:34.360 of people ever concentrated all the wealth and then consciously redistributed. 0:44:33.640 --> 0:44:36.640 It to nine trillionaires, are not going to take care 0:44:36.680 --> 0:44:37.839 of eight or nine billion of us? 0:44:37.960 --> 0:44:38.760 Yeah exactly? 0:44:38.880 --> 0:44:40.240 You call that bluff? 0:44:40.360 --> 0:44:42.640 Yeah exactly. Well, and then you combine that with the 0:44:42.680 --> 0:44:45.799 intelligence curse that we laid out that the incentive is like, 0:44:45.840 --> 0:44:47.799 why do we want to invest in the people? It 0:44:47.840 --> 0:44:52.120 became becomes basically an act of charity because otherwise I 0:44:52.120 --> 0:44:56.400 mean that or dealing with the political revolution. But again 0:44:56.480 --> 0:44:58.400 I don't think that we're currently on track to be 0:44:58.440 --> 0:45:01.440 redistributing that well. And it's also not just the wealth 0:45:01.440 --> 0:45:02.759 and the money. It's like we have people have to 0:45:02.760 --> 0:45:05.520 have work and dignity and status and meaning vulture life. 0:45:05.560 --> 0:45:08.399 You know, job sells. Life's three great evils boredom, vice 0:45:08.520 --> 0:45:11.160 and need. That's just need, but boredom and vice. So 0:45:11.200 --> 0:45:11.759 let's get to. 0:45:11.680 --> 0:45:14.239 The community and belonging absolutely all of. 0:45:14.200 --> 0:45:17.239 That, which again no one's talked about more than you 0:45:17.280 --> 0:45:21.520 in terms of that social dilemma. So let's before we 0:45:21.600 --> 0:45:24.600 go there in this notion of the transition and job 0:45:24.640 --> 0:45:27.440 displacement and the sort of the human condition that I 0:45:27.480 --> 0:45:31.120 think connects as well the President chi President Trump's visit 0:45:31.200 --> 0:45:33.880 as well in terms of you know, their own domestic 0:45:33.920 --> 0:45:36.080 issues in China. That's where they have the same incentive 0:45:36.080 --> 0:45:38.759 structure not to go through that transition with the kind 0:45:38.800 --> 0:45:42.040 of displacement that could create social unrest. That's in the 0:45:42.080 --> 0:45:44.919 short term, but get back to this notion of constraints 0:45:44.960 --> 0:45:47.000 and the safety side of things. I mean, we're here 0:45:47.040 --> 0:45:49.840 in California and the dominant you know, I mean the 0:45:49.880 --> 0:45:52.799 technology sort of birthplace of so much of the technology. 0:45:52.800 --> 0:45:55.600 Obviously the consciousness thirty two of the top fifty market 0:45:55.600 --> 0:45:58.640 cap companies, arguably old stat and of course most of 0:45:58.680 --> 0:46:03.000 the eye labs here. But we're also we've been leaders 0:46:03.239 --> 0:46:07.600 modest though some would suggest, but we've been leaders in 0:46:07.719 --> 0:46:11.480 particularly large language model, frontier models of focusing on a 0:46:11.560 --> 0:46:17.400 regulatory structure in the complete absence of any federal regulation. 0:46:17.880 --> 0:46:24.680 It's the let it rip administration until recently, there were 0:46:24.680 --> 0:46:29.560 some tonal shifts in the Trump administration just the last 0:46:29.600 --> 0:46:32.520 few weeks. They were undermining, they were very intentionally undermining 0:46:32.719 --> 0:46:36.759 the legislation that we brought SB fifty three, and you 0:46:36.880 --> 0:46:41.400 had other Republicans that were trying to undermine a California's 0:46:41.480 --> 0:46:44.279 Leadership Center crews will call them out saying we don't 0:46:44.320 --> 0:46:47.400 want to see the California invocation of regulation all across 0:46:47.400 --> 0:46:53.120 the United States of America. Interestingly, that's beginning to shift. Why 0:46:53.120 --> 0:46:55.960 do you think that's the case? Is it because our 0:46:56.960 --> 0:47:00.000 ars ais are is no longer formally in that role. 0:47:01.080 --> 0:47:04.280 Is it because now they're waking up to this new reality? 0:47:04.440 --> 0:47:08.279 Was it what happened in the Pentagon with Dario and Anthropic? 0:47:08.520 --> 0:47:10.640 Was it the combination of all of this. It's because 0:47:10.640 --> 0:47:13.360 they're listening to to you they watched the doc. Is 0:47:13.400 --> 0:47:16.359 because they realize all the money in the world is 0:47:16.360 --> 0:47:18.960 not going to build a big enough bunker that I 0:47:19.000 --> 0:47:23.160 can enjoy in the absence of societal calm. What is it? 0:47:23.840 --> 0:47:25.920 I mean I think of it to say, very shortly, 0:47:26.040 --> 0:47:28.360 is that there are two different realities. They're sort of 0:47:28.400 --> 0:47:31.560 like the political reality, and then there's like physical reality, 0:47:32.040 --> 0:47:36.239 and physical reality is crashing into political reality. That is 0:47:36.239 --> 0:47:41.560 with Mythos. Suddenly banks can get hacked, any computer system 0:47:41.600 --> 0:47:45.760 can get hacked, your stuff can get hacked. And whence 0:47:45.800 --> 0:47:49.200 that physical reality starts getting scary enough you have to 0:47:49.239 --> 0:47:51.239 start waking up. You're no longer in sort of like 0:47:51.239 --> 0:47:52.160 political game land. 0:47:53.080 --> 0:47:55.719 I do think that it was. It's interesting to note 0:47:55.719 --> 0:47:59.480 that when the emergency meeting was convened after Claude Mythos 0:47:59.520 --> 0:48:02.960 came out, it wasn't at the Pentagon or National Security. 0:48:02.960 --> 0:48:05.000 I mean that happened too, I'm sure, but the real 0:48:05.040 --> 0:48:07.760 meeting that happened was between Scott Bessant and the treasure 0:48:07.840 --> 0:48:10.759 the Treasury Secretary can meeting all the banks because I 0:48:10.760 --> 0:48:12.759 think the thing that really got them was that if 0:48:12.800 --> 0:48:15.680 this takes down the financial system, we'll get his quote 0:48:15.680 --> 0:48:18.680 ten percent GDP growth if the entire financial system gets undermined. 0:48:19.280 --> 0:48:21.719 So I think this again illustrates the point we have 0:48:21.760 --> 0:48:25.600 been making since the beginning that the upsides don't prevent 0:48:25.680 --> 0:48:29.160 the downsides, and the downsides can undermine the world that 0:48:29.200 --> 0:48:33.640 can sustain and the benefits and of the upsides. And 0:48:33.719 --> 0:48:36.640 so I do think there's been a forced shift, and 0:48:37.160 --> 0:48:38.480 you know it's very late in the game, but we 0:48:38.480 --> 0:48:41.680 should celebrate that it is happening. Now. We just need 0:48:41.719 --> 0:48:44.719 this kind of full whole of society response to mobilize. 0:48:44.880 --> 0:48:47.000 I mean, there's a Nicholas Carlini who gave the talk 0:48:47.040 --> 0:48:49.680 on mythos at a conference black Hat Lelam I think 0:48:49.719 --> 0:48:52.200 it was called because it was the Unprompted Unprompted Conference, 0:48:53.320 --> 0:48:55.680 and basically saying he's kind of calling if you know, 0:48:55.680 --> 0:48:58.399 if you are a cyber person, we need you right now. 0:48:58.400 --> 0:49:00.640 We need you to defending all the systems. Everybody should 0:49:00.640 --> 0:49:03.240 get access to mithis and do it as fast as possible, because, 0:49:03.280 --> 0:49:06.680 as you said, Gavin, the clip between the new capabilities 0:49:06.719 --> 0:49:08.759 being out there and then China coming out with the 0:49:08.800 --> 0:49:11.440 model that makes it possible. Maybe this time we have 0:49:11.480 --> 0:49:13.800 six months. Next time maybe we have three months, and 0:49:13.800 --> 0:49:16.440 then we have two months. So we need to really 0:49:16.480 --> 0:49:18.640 work hard. I'm not trying to scare people. It just 0:49:18.680 --> 0:49:21.600 means that we actually have to work hard to create 0:49:21.600 --> 0:49:24.400 safety here. Now. China doesn't want the financial system to 0:49:24.400 --> 0:49:27.880 collapse either, No, they don't, and so again we have 0:49:27.920 --> 0:49:31.920 to recognize that it's possible for coordination to happen, not 0:49:31.960 --> 0:49:34.520 because of kumbaya we're all going to get along, but 0:49:34.560 --> 0:49:37.879 because out of self interest, like the US doesn't want 0:49:37.960 --> 0:49:39.840 China to screw it up and then break the financial system, 0:49:39.920 --> 0:49:41.400 China doesn't want the US to screw up and then 0:49:41.440 --> 0:49:44.120 break the financial system. Or the US doesn't want China 0:49:44.120 --> 0:49:46.799 to release a ROGUEI that starts mining for cryptocurrency and 0:49:46.840 --> 0:49:49.640 hacking into things and self replicating like an invasive species, 0:49:50.000 --> 0:49:51.759 and China doesn't want the US to do that either. 0:49:52.320 --> 0:49:55.720 So so long as we have clarity about what we want, 0:49:56.640 --> 0:49:57.879 we can choose a different path. 0:49:58.440 --> 0:50:01.560 And what a Breton would type past. Yeah. 0:50:01.600 --> 0:50:03.200 Well, and this is actually one of the other exciting 0:50:03.239 --> 0:50:07.400 things is that we haven't even really tried to coordinate yet, 0:50:08.120 --> 0:50:10.520 Like what percentage of the billionaires wealth, how much of 0:50:10.520 --> 0:50:12.719 their time have they spent actually trying to coordinate. They 0:50:12.760 --> 0:50:15.279 will just say, well, if you're going to do it, 0:50:15.280 --> 0:50:16.000 then we're going to do it. 0:50:16.960 --> 0:50:18.919 If you say it's impossible, have you spent a month 0:50:18.960 --> 0:50:21.680 of your life and all of your connections dedicatedly trying. 0:50:21.719 --> 0:50:24.360 Yeah, exactly one and the same thing like this, guys, exactly. 0:50:24.440 --> 0:50:27.480 And the last time that humanity invented a technology that 0:50:27.560 --> 0:50:32.040 could extinct ourselves, like the nuclear bomb, we had Brentwoods. 0:50:32.360 --> 0:50:36.200 You took kind of groups from one hundred countries, lock 0:50:36.239 --> 0:50:38.360 them in a hotel room in New Hampshire for like 0:50:38.440 --> 0:50:40.400 six weeks or something like that, and said we're going 0:50:40.440 --> 0:50:41.400 to figure something out. 0:50:41.280 --> 0:50:43.399 And you're not leaving the hotel until we figure it out. 0:50:43.480 --> 0:50:45.400 So it's like it's not a conference where you go, 0:50:45.440 --> 0:50:47.200 you drink your coffee and you listen to some talks. 0:50:47.320 --> 0:50:50.279 It's what we need is the we lock ourselves in 0:50:50.320 --> 0:50:54.000 a room and we figure this out. And I know 0:50:54.040 --> 0:50:55.680 that this summit is just a couple of days and 0:50:55.800 --> 0:50:57.880 We all know about the difficulties of the level of 0:50:57.960 --> 0:50:59.920 expertise that might be involved right now, but this is 0:51:00.160 --> 0:51:02.600 moment to open the doorway of that possibility. 0:51:03.760 --> 0:51:07.240 And we have examples through history where when something becomes 0:51:07.320 --> 0:51:12.680 existential to a citizen group and their nation, they will coordinate. 0:51:12.920 --> 0:51:14.600 So you know, in the middle of the Cold War, 0:51:14.880 --> 0:51:19.120 still the US and Russia we coordinated on eradicating smallpox, 0:51:19.200 --> 0:51:22.840 and the US did logistics and funding, and the Soviet 0:51:22.880 --> 0:51:25.880 Union made twenty five million doses of the vaccine annually. 0:51:26.280 --> 0:51:28.360 And then you know India and Pakistan they were literally 0:51:28.360 --> 0:51:31.080 trading bullets in the nineteen sixties, and yet they still 0:51:31.120 --> 0:51:34.719 worked on the Indus Water Treaty and that lasted nearly 0:51:34.760 --> 0:51:38.200 sixty years because access to water was existential to them 0:51:38.239 --> 0:51:38.600 and their. 0:51:38.480 --> 0:51:41.520 Shared water supply. You have to collaborate on that. And 0:51:41.560 --> 0:51:44.359 so I just want to acknowledge the people here. It 0:51:44.440 --> 0:51:46.800 was you know in people, some people know the history 0:51:46.840 --> 0:51:49.000 that Obama and she President Obama and She signed an 0:51:49.000 --> 0:51:50.799 agreement to not cyber hack each other. And I think 0:51:50.840 --> 0:51:52.520 the next day was the biggest cyber haacket in the 0:51:52.600 --> 0:51:55.560 US government by China. So I want people to hear 0:51:55.640 --> 0:51:58.680 this not from some kind of naivete about the level 0:51:58.920 --> 0:52:02.120 of competition. I will read antagonism that is currently present. 0:52:02.880 --> 0:52:05.040 But when the stakes get existential, when I push the 0:52:05.080 --> 0:52:06.760 button and it shifts from the label of the button 0:52:06.760 --> 0:52:09.880 being ten percent GDP growth and military dominance and cyber dominance, 0:52:10.320 --> 0:52:12.840 the next time I push the button, it's collective suicide. 0:52:13.520 --> 0:52:15.399 I don't want to push that button, and China doesn't 0:52:15.400 --> 0:52:18.640 want to push that button either, So the button label 0:52:18.719 --> 0:52:20.960 has to shift from what we thought it was going 0:52:20.960 --> 0:52:23.360 to give us to a new outcome. And with the 0:52:23.360 --> 0:52:25.320 way ASA says that I love is that the fear 0:52:25.800 --> 0:52:29.080 of all of us losing has to become greater than 0:52:29.080 --> 0:52:30.200 the fear of me losing to you. 0:52:30.760 --> 0:52:36.560 So you have a very regulated construct in China compared 0:52:36.560 --> 0:52:41.040 to certainly the United States. You're talking about a traditional 0:52:41.040 --> 0:52:46.160 model right now, where the new frontier out here the 0:52:46.160 --> 0:52:48.399 wild less. Yeah yeah, I mean it's you know, go west, 0:52:48.960 --> 0:52:51.240 young man, go West. I mean, it's people are pushing 0:52:51.280 --> 0:52:53.920 out the boundaries of discovery, holding them back on their own. 0:52:54.920 --> 0:52:59.680 Regardless of what happens in Beijing, this is really in 0:52:59.680 --> 0:53:03.560 the hands of a handful of people ultimately making the 0:53:03.640 --> 0:53:08.600 right decision. We you know, and I believe Dario is 0:53:08.640 --> 0:53:10.920 the best of a lot. I think universally that's accepted. 0:53:11.000 --> 0:53:14.319 But that may be, and even Dario may acknowledge because 0:53:14.320 --> 0:53:17.200 of the flatness of the surrounding terrain, may not be. 0:53:17.200 --> 0:53:20.239 Because he's particularly eminent on his own and he talks 0:53:20.280 --> 0:53:23.960 about his own you know. You know, he's an entrepreneur, 0:53:24.320 --> 0:53:27.600 and he's constantly reflecting on his own incentive structure and 0:53:27.640 --> 0:53:29.360 how he has to compete in this environment at the 0:53:29.360 --> 0:53:31.120 same time, and he's at least I think, has the 0:53:31.160 --> 0:53:35.520 more situational awareness than others. But you know, whatever can 0:53:35.600 --> 0:53:38.600 be will be. And you know, with respect to Elon, 0:53:38.680 --> 0:53:42.960 I don't trust x Ai, you know, I mean the 0:53:43.000 --> 0:53:45.760 idea that he's the good guy compared to our friends 0:53:45.800 --> 0:53:48.840 down at Google or you guys left. I mean, you know, 0:53:49.840 --> 0:53:52.680 so talk to me about more of the sinister realities 0:53:52.680 --> 0:53:54.279 of you know, I don't mean that sinister in the 0:53:54.600 --> 0:53:58.800 but but the impulses again to be the guy, the god, 0:53:59.040 --> 0:54:00.600 the god. May I mean to have their DNA. 0:54:00.920 --> 0:54:03.520 Yeah, So I'm gratefully you're bringing this up. There's a 0:54:03.520 --> 0:54:06.719 few things we should enumerate here. So one is we 0:54:06.760 --> 0:54:10.400 need coordination, and we do find ourselves in the unfortunate 0:54:10.440 --> 0:54:12.719 spot that the people who do need to be in 0:54:13.000 --> 0:54:16.960 coordination maximally distrust each other. Even just the US CEOs, 0:54:17.160 --> 0:54:19.440 I mean Elon Musk and Sam hate each. 0:54:19.320 --> 0:54:28.359 Other, correct, and so the India they couldn't even hold 0:54:28.600 --> 0:54:29.560 at least they showed up. 0:54:29.640 --> 0:54:33.719 So we have a problem of trust between the leaders themselves. 0:54:33.960 --> 0:54:36.920 We need I think structures that impose the trust on 0:54:37.040 --> 0:54:39.360 top because they're not going to do it autonomously themselves. 0:54:40.400 --> 0:54:41.480 So that's the regulation. 0:54:41.560 --> 0:54:44.359 That's yeah, that's the trans using the power of law 0:54:44.640 --> 0:54:48.319 to say that these people, it's going to happen in China. 0:54:48.360 --> 0:54:50.080 Here's the rule is going to be similar in China 0:54:50.120 --> 0:54:52.319 as they are here. We're both, for example, not going 0:54:52.320 --> 0:54:54.439 to open source a model that can happen to any 0:54:54.440 --> 0:54:56.880 computer system in the world without defenses, at least not 0:54:56.920 --> 0:54:59.120 going to open source that. China doesn't want a rogue 0:54:59.160 --> 0:55:02.840 state act, nonstate actor, or terrorist group having that ability 0:55:02.880 --> 0:55:05.560 to hack their infrastructure because it would also blow back 0:55:05.560 --> 0:55:08.360 onto them. Same thing with a open source model that 0:55:08.440 --> 0:55:11.480 knows how to do very dangerous things with biology. There's 0:55:11.520 --> 0:55:14.200 some threshold that we can get these countries to agree, 0:55:14.680 --> 0:55:16.799 just like the Soviet Union in the United States had 0:55:16.800 --> 0:55:20.080 a red phone saying we're gonna this is to de escalate. 0:55:20.560 --> 0:55:23.240 I think we need something like an AI red lines phone, 0:55:23.280 --> 0:55:26.400 meaning that both countries have common knowledge of the frontier 0:55:26.400 --> 0:55:28.279 of these risks, because right now we don't even have 0:55:28.280 --> 0:55:28.920 that common notion, and. 0:55:28.880 --> 0:55:30.920 There's rumors that that may be one of the things 0:55:30.960 --> 0:55:34.320 that that are prepared to announce. That's right, you know, 0:55:34.440 --> 0:55:37.920 at least some beginning of this. 0:55:38.040 --> 0:55:40.200 But I want the other aspect of the human experience 0:55:40.239 --> 0:55:41.879 of this. We did a screening of the AI doc 0:55:42.080 --> 0:55:44.120 in New York and there is someone in the audience 0:55:44.160 --> 0:55:48.319 actually who raised her hand quietly and she said, I'm 0:55:48.360 --> 0:55:51.600 a coach for one of the CEOs of these companies, 0:55:52.400 --> 0:55:54.640 and you know what happens when I talk to them 0:55:54.920 --> 0:55:57.200 is they say, but what can I do? I'm just 0:55:57.239 --> 0:56:00.480 one person. I'm powerless. And I want people to hear that, 0:56:00.520 --> 0:56:03.160 because I noticed that, even you know, we all feel 0:56:03.200 --> 0:56:05.239 relative to the size of this problem, you will never 0:56:05.400 --> 0:56:09.400 locate agency that is that is enough agency to do 0:56:09.400 --> 0:56:11.920 something about this problem in one human body, even if 0:56:11.920 --> 0:56:15.880 that body is elon by himself or Sindar by himself, 0:56:15.960 --> 0:56:20.239 or Sam by themselves, and so getting back to you know, 0:56:20.280 --> 0:56:22.360 I think what this moment is inviting us into. We 0:56:22.400 --> 0:56:25.120 often say that AI is are ultimate test, but greatest 0:56:25.160 --> 0:56:28.200 invitation that we have to go from agency to reegency. 0:56:28.280 --> 0:56:30.719 We have to basically act in some kind of collective way. 0:56:31.280 --> 0:56:33.319 And the forces politically of the world have been driving 0:56:33.400 --> 0:56:35.719 us away from that. But this is kind of the test, 0:56:35.800 --> 0:56:38.600 like we either do that and we step up and 0:56:38.680 --> 0:56:41.480 again we need everything from common political pressure and this 0:56:41.520 --> 0:56:43.600 being the number one issue in the midterms, and you know, 0:56:43.640 --> 0:56:46.160 the public's rallying and all the governors you know, speaking 0:56:46.200 --> 0:56:47.839 up about this, and all the world leaders speaking out. 0:56:47.840 --> 0:56:49.640 That's just Yet. You know, two days ago I got 0:56:49.640 --> 0:56:52.319 an email from the President of Iceland who basically wants 0:56:52.360 --> 0:56:54.880 to activate on this issue in Iceland hosted in Rykovic 0:56:54.960 --> 0:56:58.240 the first Arms Control talks. There's a lot that people 0:56:58.280 --> 0:57:00.560 could do if they said not just what can I do, 0:57:01.080 --> 0:57:03.719 but how could I get my reaching up and out 0:57:03.719 --> 0:57:07.680 to the network of people to take action together. There 0:57:07.719 --> 0:57:10.120 is a second part to your point, which is the 0:57:10.200 --> 0:57:14.280 darker part, which is you talking about the game theory 0:57:14.480 --> 0:57:17.800 of the psychology of the leaders that they basically believe 0:57:17.920 --> 0:57:20.480 that in the worst case scenario The thing that kept 0:57:20.560 --> 0:57:24.400 us safe in nukes is that it's two people who 0:57:24.400 --> 0:57:26.640 have to push the button. And I know you won't 0:57:26.680 --> 0:57:29.160 push the button because I know that there's something sacred 0:57:29.640 --> 0:57:31.800 that you don't want this whole thing to end. And 0:57:31.880 --> 0:57:33.120 I know that, and I know that you know that. 0:57:33.160 --> 0:57:35.400 I know that. And so even though we get very 0:57:35.480 --> 0:57:37.480 very very close, and we've gotten close so many times, 0:57:38.080 --> 0:57:41.400 we haven't pushed that button. But in this case with AI, 0:57:41.680 --> 0:57:43.840 there's a belief. First of all, it's a red zone 0:57:43.880 --> 0:57:45.840 of where the risk occurs. It's not like there's one 0:57:45.880 --> 0:57:47.920 button that gets pushed. It's like we just push this 0:57:47.920 --> 0:57:50.920 stuff out there, and there's a belief that it's inevitable. 0:57:50.920 --> 0:57:53.400 If I didn't do it, someone else would, which means 0:57:53.480 --> 0:57:57.560 I don't experience ethical complicity in being part of the 0:57:57.640 --> 0:58:00.200 end of civilization. And if it's inevitable, there's thing I 0:58:00.200 --> 0:58:01.360 could have done to stop it, so I don't even 0:58:01.440 --> 0:58:03.919 have to feel bad, right, And then so the game 0:58:03.960 --> 0:58:06.080 theory goes from all of us knowing that we want 0:58:06.080 --> 0:58:08.960 to avoid the bad outcome to everybody believing that there 0:58:09.000 --> 0:58:11.920 isn't a different outcome, which means that the best outcome 0:58:12.040 --> 0:58:13.760 is maybe the worst thing is all of us go 0:58:14.400 --> 0:58:17.320 by the wayside, but we birth the digital God and 0:58:17.360 --> 0:58:20.440 it speaks Chinese instead of English, or maybe it has 0:58:20.480 --> 0:58:22.120 Elon's DNA. Instead of saying just. 0:58:22.320 --> 0:58:24.320 Said, it's at least on an American stacks. 0:58:24.880 --> 0:58:27.680 At least we're selling the world American your media. But 0:58:27.760 --> 0:58:30.600 the reason of laying all that out is that if 0:58:30.800 --> 0:58:33.200 the whole world could see I think what we just 0:58:33.280 --> 0:58:37.080 laid out, if literally everyone could see that, the whole 0:58:37.120 --> 0:58:40.200 world says, we don't want eight soon to be trillionaires 0:58:40.480 --> 0:58:43.000 deciding the future for eight billion people who didn't consent 0:58:43.040 --> 0:58:43.360 to this. 0:58:44.560 --> 0:58:46.600 And that's the purpose of Again, that sort of brings 0:58:46.640 --> 0:58:48.919 us back to the beginning. Why you're doing this damn film? 0:58:48.920 --> 0:58:52.320 That's right after the great and sort of global consciousness. 0:58:52.320 --> 0:58:52.720 That's right. 0:58:53.000 --> 0:58:56.000 So in the absence of that, we're back here in California. Yes, 0:58:56.000 --> 0:58:58.640 here here with the governor, Yes, the current governor, the 0:58:58.680 --> 0:59:03.480 governor's mansion. Interesting, we're doing some decent things. What more 0:59:03.480 --> 0:59:06.800 should I be doing in the absence of the kind 0:59:06.800 --> 0:59:10.040 of federal leadership that we need. I feel like we've 0:59:10.080 --> 0:59:13.040 lost this last eighteen months, you know, I just we 0:59:13.040 --> 0:59:14.840 were mo Look, we did an a open It was 0:59:14.880 --> 0:59:18.200 interesting working I worked very closely with the Biden administration. 0:59:18.240 --> 0:59:21.440 Did they move quickly enough? Perhaps not, but at least 0:59:21.480 --> 0:59:24.080 we had a framework of an of an executive order. 0:59:24.120 --> 0:59:26.880 We moved that forward. We President's sided here at the 0:59:26.880 --> 0:59:30.600 Fairmount Hotel in San Francisco in California. It was built 0:59:30.640 --> 0:59:34.160 off an executive order that I did six months later. 0:59:34.200 --> 0:59:36.760 We were working hand in glove with the Biden administration 0:59:37.160 --> 0:59:39.760 on that. That was ripped up right when the Trump 0:59:39.800 --> 0:59:43.880 administration came in to office. You have an ai zar 0:59:44.280 --> 0:59:47.040 out of the Bay Area certainly understands the ecosystem. But 0:59:47.080 --> 0:59:48.560 it seemed to me, and this is me and you 0:59:48.640 --> 0:59:50.760 guys don't have to respond, but it was the great grift. 0:59:50.800 --> 0:59:53.440 Everybody was sort of on on the train and seeing 0:59:53.440 --> 0:59:56.480 this as an opportunity and looking at the abundance of 0:59:56.520 --> 0:59:59.520 this only but not looking at safety, not looking at 0:59:59.600 --> 1:00:03.640 the risk. As you've described, California cited to assert itself 1:00:03.640 --> 1:00:06.040 in that respect, as we've done on privacy, as we've 1:00:06.080 --> 1:00:08.439 done a lot of child safety issues, and a lot 1:00:08.480 --> 1:00:10.360 more work to do there. And this year will be 1:00:10.600 --> 1:00:12.480 a landmark year in terms of getting to the next 1:00:12.560 --> 1:00:16.080 level in that respect. But what more can the state 1:00:16.160 --> 1:00:18.960 be doing the fear of that always is patchwork, not 1:00:19.000 --> 1:00:23.680 a framework for the nation. And how you support innovation 1:00:24.400 --> 1:00:27.040 our own GDP growth which has been off the charts 1:00:27.080 --> 1:00:31.000 in California visa VR competitors. At the same time addressed 1:00:31.040 --> 1:00:34.960 these larger global issues. Do you have any specific ideas 1:00:35.200 --> 1:00:38.280 for a governor of California, that is the current governor 1:00:38.640 --> 1:00:40.920 that has a budget that he's releaseding in weeks and 1:00:40.960 --> 1:00:43.600 a legislative session coming up in the next few months. 1:00:44.680 --> 1:00:47.720 Well, we'll get to answer that question specifically, but what 1:00:47.960 --> 1:00:50.120 are the places of also good news? I just want 1:00:50.160 --> 1:00:52.160 to say, because there's a lot of value in just 1:00:52.200 --> 1:00:54.480 social signaling where everyone knows that there's a problem. We 1:00:54.560 --> 1:00:57.880 get to that shared common knowledge, common feeling. And if 1:00:57.880 --> 1:01:01.040 you went back two years and you said by today, 1:01:01.160 --> 1:01:03.880 twenty five percent of the world's population would live in 1:01:03.920 --> 1:01:07.320 a country where they've either have announced or enacted a 1:01:07.440 --> 1:01:11.600 ban for social under sixteen, you'd be like, that's ridiculous. 1:01:11.640 --> 1:01:12.640 You couldn't possibly get that. 1:01:12.640 --> 1:01:14.680 I want people to really feel that. There you are 1:01:14.680 --> 1:01:17.200 in twenty twenty two. If you said even just literally 1:01:17.200 --> 1:01:19.280 three years ago, yeah, I know that a quarter of 1:01:19.320 --> 1:01:24.360 the world's population we're talking Australia, India, Denmark, Denmark's. Two 1:01:24.360 --> 1:01:26.560 weeks ago, Greece added the list. France. I was with 1:01:26.800 --> 1:01:28.960 John Hyde and when he was in Davos and he 1:01:29.400 --> 1:01:31.640 met with President Macrone and got France on board. Like 1:01:32.080 --> 1:01:34.640 people would have thought that was impossible. I mean, the 1:01:34.720 --> 1:01:35.560 trains left the station. 1:01:35.680 --> 1:01:37.600 We just had all the democratic governors out here, and 1:01:37.640 --> 1:01:39.480 everyone was trying to compare them to trast which one 1:01:39.480 --> 1:01:43.160 of the states, which we're going to do. But I 1:01:43.200 --> 1:01:45.120 mean it's you're right, this is a tipping point. 1:01:45.120 --> 1:01:47.440 It's taking the train and once you get twenty five percent, 1:01:47.480 --> 1:01:48.720 you're going to get the rest of the world. 1:01:48.920 --> 1:01:50.440 And the point there was a lot of these companies. 1:01:50.480 --> 1:01:53.680 I also I anticipate these companies and by the way, 1:01:53.720 --> 1:01:57.120 if I was advising these companies get ahead of this train, 1:01:57.280 --> 1:02:00.840 ye and show your large s and that your maturity 1:02:00.880 --> 1:02:04.680 and understanding. So I imagine that may happen as well. 1:02:05.240 --> 1:02:07.040 But no, that's a point. You're right. 1:02:07.120 --> 1:02:09.840 And so right now, something really interesting happened in Hongjo 1:02:09.960 --> 1:02:12.480 in China. Maybe you're aware of it, but it's this 1:02:12.560 --> 1:02:16.320 first case where there was someone who lost his job 1:02:16.480 --> 1:02:20.360 at Ai Automation and this court ruled and said, actually, 1:02:21.040 --> 1:02:23.200 that's not a legitimate reason for you to lose your job. 1:02:23.680 --> 1:02:28.160 Companies are not allowed to fire you from like increased automation. 1:02:28.720 --> 1:02:31.680 Well that solved the whole problem. Probably not, but imagine 1:02:31.720 --> 1:02:34.440 that California took a leadership position there and said, actually, 1:02:34.440 --> 1:02:37.400 there are going to be some very serious protections that 1:02:37.720 --> 1:02:43.760 as AI increases, GDP increases profits, Actually, like people are 1:02:43.880 --> 1:02:46.520 going to be it's sort of like an employment insurance. 1:02:46.560 --> 1:02:48.800 They're going to be able to keep their jobs. Like 1:02:49.000 --> 1:02:51.920 that would set a social signal for the rest of 1:02:52.000 --> 1:02:52.560 the US. 1:02:52.560 --> 1:02:54.440 So we'll let you have the hook on the larger 1:02:55.440 --> 1:03:00.919 safety risk regulation and let's go back to ideas along 1:03:00.960 --> 1:03:03.400 these lines, because right in front of us this notion 1:03:03.800 --> 1:03:06.720 of displacement transition, and it goes back to their earlier 1:03:06.760 --> 1:03:10.120 point I was making. Some will argue that, you know, 1:03:10.160 --> 1:03:13.840 we always you know, lud eights and will never see 1:03:13.880 --> 1:03:16.240 the abundance on the other side, and we can't even 1:03:16.320 --> 1:03:20.120 conceive of the jobs. So with humility, let's not just 1:03:20.200 --> 1:03:23.959 assume there'll be no human jobs, because the human mind 1:03:24.080 --> 1:03:27.439 is the capacity that's limitless. With these technologies, that will 1:03:27.480 --> 1:03:31.520 be more supportive and allow us to be augmented and 1:03:31.600 --> 1:03:35.479 discover talents and capacity we never thought possible. So let's 1:03:35.480 --> 1:03:38.560 assume that happens. But the concern is that it seems 1:03:38.600 --> 1:03:42.480 the most universal that it may happen very fast. The 1:03:42.640 --> 1:03:46.479 impacts is, so, how do you then flatten the curve 1:03:47.320 --> 1:03:49.520 the curve? How do we address the transition You get 1:03:49.520 --> 1:03:52.440 to employment insurance, something we've been talking a lot about. 1:03:52.560 --> 1:03:54.520 You get to this notion that you can't fire someone 1:03:54.520 --> 1:03:57.080 to be automated. That's even deeper, and that's interesting. This 1:03:57.200 --> 1:04:00.800 Chinese example tell me more about the issue of and 1:04:00.960 --> 1:04:03.160 what I should be worried about. When I see Dario 1:04:03.240 --> 1:04:06.680 saying fifty percent entry level jobs, it's no longer a 1:04:06.720 --> 1:04:09.080 career ladder. It's a jungle gym. And all these young 1:04:09.120 --> 1:04:12.439 folks got a Stanford or like now I'm unemployed or unemployable, 1:04:12.760 --> 1:04:15.400 no code or software. I mean, give me a women 1:04:15.520 --> 1:04:18.400 disproportion being impacted in the workforce When you look at 1:04:18.400 --> 1:04:21.520 those clerical jobs, admin jobs, et cetera. I mean, what, 1:04:21.520 --> 1:04:24.720 what's what do you see a year two years from now? 1:04:24.960 --> 1:04:27.680 As we deal with the Holy Grail, the god complex 1:04:27.920 --> 1:04:30.240 of Agi in the displacement space. 1:04:32.600 --> 1:04:35.160 Reed Hoffman has this idea. When mefre I get it right, 1:04:35.520 --> 1:04:38.360 I like, which is one of the things that makes 1:04:38.400 --> 1:04:40.680 AI distinct. And Sam Altman has talked about this himself, 1:04:41.080 --> 1:04:43.800 that you could get the age of these unicorn companies. 1:04:43.800 --> 1:04:45.920 A unicorn company meeting a billion dollar valuation. 1:04:46.160 --> 1:04:49.000 I met a guy the other day literally billion dollar valuation, 1:04:49.600 --> 1:04:53.360 tim him exactly, So I think that he was praying 1:04:53.400 --> 1:04:54.920 himself around as the it's. 1:04:54.800 --> 1:04:56.680 Me, that's right, one of the first, and so that's 1:04:56.680 --> 1:04:58.640 what that's what has been positive ever since the beginning 1:04:58.640 --> 1:05:00.440 of AI. They're saying, we're going to start having world. 1:05:00.640 --> 1:05:03.080 You're going to have a single person with a unicorn 1:05:03.080 --> 1:05:06.680 billion dollar company. Yeah, Now does this Does society work? 1:05:07.200 --> 1:05:09.520 If there's a handful of single people with billion dollar 1:05:09.520 --> 1:05:11.760 companies and no one else has a job, it doesn't work. 1:05:12.040 --> 1:05:13.520 Do you think those people just want to live out 1:05:13.520 --> 1:05:16.840 their lives in bunkers with private militaries and gas masks 1:05:16.880 --> 1:05:19.439 because they've created that world. I don't think they want 1:05:19.440 --> 1:05:21.680 that world. So reied Hoffman, who's the founder of LinkedIn, 1:05:21.840 --> 1:05:23.880 was early at PayPal and you know, was at some 1:05:23.880 --> 1:05:27.160 point a friend of Peter Thiel's, you know, has this 1:05:27.200 --> 1:05:30.000 proposal that we can tax companies based on the proportion 1:05:30.240 --> 1:05:33.440 ratio of how many employees they have relative to their revenue. 1:05:33.760 --> 1:05:39.840 So you want to basically disincentivize the single solo unicorn company. 1:05:39.840 --> 1:05:40.800 I don't want to add to it. 1:05:41.000 --> 1:05:41.240 Yeah. 1:05:41.280 --> 1:05:43.480 Well, and also just just to note that it doesn't 1:05:43.480 --> 1:05:47.400 stop with just single person unicorns. Yeah, with automating that 1:05:47.440 --> 1:05:49.840 final one person not so hard, So you're gonna end 1:05:49.920 --> 1:05:50.920 up with zero person. 1:05:50.920 --> 1:05:52.840 You can have a CEO, B and AI. Yeah, and 1:05:52.840 --> 1:05:54.760 that's actually happening. We're already getting ais that are on 1:05:54.800 --> 1:05:56.680 boards and things like this. Yeah, and so even if 1:05:56.680 --> 1:05:58.800 you might find it questionable that that person might have again, 1:05:58.840 --> 1:06:02.040 they can. They can earn their wealth, but you have 1:06:02.080 --> 1:06:04.880 to have some taxation to make sure this is being stipid. 1:06:04.920 --> 1:06:07.160 And we also need we need to find ways of 1:06:07.200 --> 1:06:10.240 having universal basic ownership, not just universal basic cash payments 1:06:10.240 --> 1:06:13.080 and UBI, but universal basic ownership. I think people need 1:06:13.120 --> 1:06:15.280 to have a stake in the success that's happening, like 1:06:15.280 --> 1:06:17.960 what Norway did with the Sovereign Wealth Fund. But oil 1:06:18.080 --> 1:06:20.440 didn't oil produce this kind of gravy on top for 1:06:20.480 --> 1:06:23.440 the civilization. People still had jobs. So what's different about 1:06:23.440 --> 1:06:25.320 this is we do need to find ways of doing 1:06:25.440 --> 1:06:29.200 universal basic work. We also need to find ways of 1:06:29.720 --> 1:06:33.280 having certain professions in which that embodied wisdom, like a 1:06:33.320 --> 1:06:36.360 surgeon or a senior lawyer or a senior judge. We 1:06:36.400 --> 1:06:39.600 need ways of training and apprenticing almost like minimum quotas 1:06:39.600 --> 1:06:42.160 of those kinds of occupations and roles in society to 1:06:42.200 --> 1:06:44.440 make sure that we have that ongoing knowledge. Because again 1:06:44.920 --> 1:06:47.080 the short term benefit of like no one needs lawyers 1:06:47.080 --> 1:06:49.800 and then the senior lawyers all die out, that world 1:06:49.840 --> 1:06:50.400 doesn't work. 1:06:50.960 --> 1:06:53.800 So just the last thing to sort of add here 1:06:53.880 --> 1:06:56.560 is that you instead of getting into arguments about how 1:06:56.680 --> 1:06:58.600 quickly exactly are people going to lose the jobs, well, 1:06:58.640 --> 1:07:01.200 people really do lose their jobs. Let's let's plan and say, okay, 1:07:01.280 --> 1:07:03.360 we think people are going to be out of livelihoods. 1:07:03.600 --> 1:07:05.360 This is this idea we really came to us from 1:07:05.400 --> 1:07:08.240 read haasting to see Pharmacy and Pharmacy Netflix, where we said, 1:07:08.440 --> 1:07:11.560 let's set up trigger point laws. If in unemployment hits 1:07:11.560 --> 1:07:12.800 ten percent, what are we going to do if it 1:07:12.880 --> 1:07:15.280 hits twenty percent? What are we going to do? You 1:07:15.320 --> 1:07:18.080 could pre set up those sort of conditions so we 1:07:18.080 --> 1:07:20.040 don't have to argue about whether it's going to happen, 1:07:20.080 --> 1:07:21.160 just what we should do when it does. 1:07:21.400 --> 1:07:23.560 And I know that you've been running with Engaged California, 1:07:23.720 --> 1:07:27.120 these citizen deliberations, these ways of you know, aggregating citizen assemblies, 1:07:27.360 --> 1:07:30.000 having citizens actually deal with and think about these issues 1:07:30.000 --> 1:07:32.640 and come to some you know, have their own input 1:07:32.640 --> 1:07:35.440 in this process. But by reckoning with these facts, and 1:07:35.640 --> 1:07:37.280 I think that these are all things that we need. 1:07:37.320 --> 1:07:40.040 You know, the countries that discover a natural resource that 1:07:40.120 --> 1:07:43.480 didn't have this engaged, well educated citizen kind of engage 1:07:43.560 --> 1:07:46.760 this in infrastructure, like Venezuela or Libya or something like that, 1:07:47.080 --> 1:07:49.400 they don't do so well. But countries like Norway where 1:07:49.440 --> 1:07:52.800 you did have the engaged citizens, with oversight of those funds, 1:07:53.080 --> 1:07:54.640 you end up with a healthier society. 1:07:54.720 --> 1:07:57.000 So it's sort of like an a California engagement fund 1:07:57.000 --> 1:08:00.000 because you want like people involved in the redistribution. 1:08:00.320 --> 1:08:00.680 That's right. 1:08:01.120 --> 1:08:04.600 So it's interesting that you mentioned mencome which the old 1:08:04.640 --> 1:08:09.919 Canadian construct ubi university basic income. You had Elon Musk 1:08:09.960 --> 1:08:12.840 the other day say he wants university basic high income. 1:08:12.960 --> 1:08:14.439 He said there is going to be univers of basic 1:08:14.480 --> 1:08:15.080 high income, and. 1:08:15.040 --> 1:08:18.679 Then was asked with Peter, right, exactly I saw him probably, 1:08:18.680 --> 1:08:20.040 and he was asked simple question, how are you going 1:08:20.080 --> 1:08:20.400 to do it? 1:08:20.439 --> 1:08:22.000 And he said, oh I did. I was just joking. 1:08:22.000 --> 1:08:22.559 I made it up. 1:08:22.760 --> 1:08:23.679 That was comforting. 1:08:23.840 --> 1:08:25.680 Yeah, I think people should really take note of that. 1:08:25.920 --> 1:08:29.160 Yes, it was just a mispronouncing of high not as 1:08:29.200 --> 1:08:31.200 in like a high income. It's just like he's he's 1:08:31.360 --> 1:08:32.400 not being able to think about it. 1:08:32.439 --> 1:08:34.880 I mean, it's uh, it's a it's a little alarming. 1:08:34.920 --> 1:08:37.559 So look, what do you quickly Look, I'm we're thinking 1:08:37.560 --> 1:08:39.640 about all these things down to the parochial on the 1:08:39.640 --> 1:08:42.639 warn Act, which is how we actually warn the public 1:08:42.720 --> 1:08:46.719 the social impacts of large scale displacement and job loss. 1:08:47.280 --> 1:08:52.000 Having more capacity to see earlier. That's after the fact 1:08:52.000 --> 1:08:53.920 the warneck comes out, those jobs are already going to 1:08:53.960 --> 1:08:56.439 be lost. What are the science to show us what 1:08:56.479 --> 1:08:58.559 the impacts are happening in the job market in real time? 1:08:58.760 --> 1:09:03.160 Issues of unemployment and sure it's becoming employment insurance so 1:09:03.200 --> 1:09:05.200 that you can keep people employed for a period of time. 1:09:05.520 --> 1:09:07.639 The Dutch do it about ninety percent of the wage. 1:09:07.840 --> 1:09:10.600 The ultimate training is a job, the dignity of a 1:09:10.680 --> 1:09:14.440 job back to those Voltaire constructs, and then the opportunity 1:09:14.479 --> 1:09:18.240 then potentially transition by using the federal government to help you. 1:09:18.280 --> 1:09:22.080 The backstop portable benefits then become fundamental. This notion of 1:09:22.200 --> 1:09:26.599 UBC university basic capital, this notion of sovereign wealth fund 1:09:26.760 --> 1:09:31.839 or equity public equity with dividends and somehow we get shares. 1:09:32.120 --> 1:09:36.240 There's equity shares contributions from these large companies. That's not 1:09:36.320 --> 1:09:39.040 just taxes, it's actual equity in the company. So there's 1:09:39.040 --> 1:09:41.839 a notion of an ownership and a larger ownership society. 1:09:41.960 --> 1:09:46.120 That we don't tax jobs with payroll taxes and then 1:09:46.280 --> 1:09:50.200 subsidize automation through tax credits. We do the inverse in 1:09:50.240 --> 1:09:52.960 that context. So all of that. So that's the stuff 1:09:53.080 --> 1:09:55.120 I'm playing around with a lot of us are thinking 1:09:55.160 --> 1:09:58.800 about right now. But how from your vantage point, how 1:09:58.920 --> 1:10:02.360 quickly is is happening. There's a lot of headlines, but 1:10:02.400 --> 1:10:04.400 there's a lot of debate around these headlines of all 1:10:04.479 --> 1:10:07.080 these cuts of jobs. But then people say, well, that 1:10:07.160 --> 1:10:09.759 was a lot of coverd over iroing there's some business 1:10:09.760 --> 1:10:13.080 model issues there that they're sort of hiding and suggesting 1:10:13.160 --> 1:10:19.920 it's AI. We even necessarily seeing massive job destruction yet 1:10:20.200 --> 1:10:22.120 or have we with AI? 1:10:22.320 --> 1:10:24.559 I think the point people should get here is that 1:10:24.960 --> 1:10:27.760 obviously it's complicated, and there's jobs are shuffling throughout the 1:10:27.800 --> 1:10:30.639 economy a little bit right now. But the long term 1:10:30.680 --> 1:10:32.360 goal of these companies is. 1:10:32.479 --> 1:10:36.360 Not the original Open AI. 1:10:36.280 --> 1:10:39.720 Missions mission statement exactly. Opening Eyes mission statement was not 1:10:39.760 --> 1:10:41.760 to give people helpful tools so that you can do 1:10:41.800 --> 1:10:44.280 your job slightly better. Their mission statement is we do 1:10:44.320 --> 1:10:47.160 your job. And actually, you know govin in in La 1:10:47.240 --> 1:10:48.920 there's an article in the La Times about this that 1:10:49.640 --> 1:10:51.720 is one of these new popular gig worker jobs in 1:10:51.840 --> 1:10:55.400 La is everybody straps a GoPro camera to their head 1:10:56.040 --> 1:10:58.880 and they look down and then they do laundry, They cook, 1:10:58.960 --> 1:11:02.000 they eat, they do all these things. So basically the 1:11:02.120 --> 1:11:04.959 number one job soon in the world will be training 1:11:05.080 --> 1:11:07.200 the replacement for that job. Think of it like being 1:11:07.200 --> 1:11:09.360 a coffin builder, Like you're designing the coffin and then 1:11:09.360 --> 1:11:11.360 you put yourself in it. And you know, if you 1:11:11.360 --> 1:11:13.240 think I'm lying, by the way, just think about Meta 1:11:14.439 --> 1:11:18.439 Mettal Instagram creators. You know, here we love creators, we 1:11:18.479 --> 1:11:20.840 love creativity. We want you to be successful. We want 1:11:20.840 --> 1:11:23.240 to you know, our whole mission statement is making creators 1:11:23.439 --> 1:11:26.720 super successful. That's before they were an AI company. When 1:11:26.800 --> 1:11:29.080 they do the second that they were an AI company, 1:11:28.840 --> 1:11:32.160 they trained on all the videos of their creators and 1:11:32.200 --> 1:11:35.439 then created generative videos that now basically suck up like 1:11:35.439 --> 1:11:38.240 a vampire, your essence, your life, force, your creativity, and 1:11:38.240 --> 1:11:40.080 then hit button and now they have a digital copy 1:11:40.080 --> 1:11:42.439 of you that you didn't consent to that can generate 1:11:42.479 --> 1:11:44.840 all these things. If you don't believe me, Just a 1:11:44.880 --> 1:11:47.000 few weeks ago, there was an article Meta is now 1:11:47.000 --> 1:11:50.719 forcing their employees to basically track all of their movements, 1:11:50.760 --> 1:11:52.040 all their clicks, all the things are doing on a 1:11:52.040 --> 1:11:54.439 computer to train AI agents to do all their jobs. 1:11:54.840 --> 1:11:57.519 This is not a conspiracy theory. All you have to 1:11:57.520 --> 1:11:59.839 do to know where we're going is understand the incentives 1:11:59.840 --> 1:12:02.280 and look at the early warning signs they're telling you 1:12:02.320 --> 1:12:04.720 who they really are. You know, Open AI says we're 1:12:04.760 --> 1:12:06.960 all here to make the world a better place. And 1:12:07.000 --> 1:12:09.960 then they release this AI slop app Sora, which was 1:12:10.280 --> 1:12:12.200 basically an infinitely they killed it now, but it was 1:12:12.240 --> 1:12:16.120 an infinitely scrolling, deep fake generated content of like, you know, 1:12:16.240 --> 1:12:19.360 funny videos of Stephen Hawking, you know, you know, going 1:12:19.360 --> 1:12:21.880 through a raceway or something like that. It's just deep 1:12:21.920 --> 1:12:25.040 fake AI slop. Why are they doing that because they 1:12:25.040 --> 1:12:26.960 want to increase their market dominance, because they want to 1:12:27.000 --> 1:12:29.280 get users, because they want to get training data, and 1:12:29.320 --> 1:12:31.519 it gets more people using open ai, so their numbers 1:12:31.520 --> 1:12:31.840 going up. 1:12:31.840 --> 1:12:33.599 And if your incentives tell you everything, and if you're 1:12:33.600 --> 1:12:37.400 a company and you're choosing do I hire a real 1:12:37.479 --> 1:12:42.360 human paralegal or GPT seven that works for less than 1:12:42.400 --> 1:12:45.479 minimum wage twenty four to seven doesn't whistle blow, doesn't 1:12:45.479 --> 1:12:48.600 a plane, doesn't have cultural issues, doesn't have paid time off? Like, 1:12:48.680 --> 1:12:49.519 which one are you going to do? 1:12:49.800 --> 1:12:50.000 Well? 1:12:50.040 --> 1:12:52.360 Your business incentive is just very very clear, and everyone's 1:12:52.360 --> 1:12:53.000 going to be trapped in this. 1:12:53.080 --> 1:12:54.880 And if your competitor you may say no, I'm going 1:12:54.920 --> 1:12:57.200 to keep the human, and then your competitor goes stops 1:12:57.200 --> 1:12:59.240 the direction, You're out of business. Yeah, I have no. 1:12:59.280 --> 1:13:01.880 Choice back to the But with those set of interventions 1:13:01.880 --> 1:13:03.599 that you just mentioned, and you just mentioned a whole 1:13:03.640 --> 1:13:05.759 slew of things we can be doing. You just mentioned 1:13:05.800 --> 1:13:07.720 so many things that we could be doing. And I 1:13:07.720 --> 1:13:10.040 think what it represents is I think people think, but 1:13:10.080 --> 1:13:12.200 if we don't race to automate every job as fast 1:13:12.200 --> 1:13:14.439 as possible, we're going to lose to China. But what 1:13:14.920 --> 1:13:17.519 this is showing and revealing is we're not in a 1:13:17.600 --> 1:13:20.600 race just the technology. We're actually in a race for 1:13:20.640 --> 1:13:23.240 a different currency. The currency is not who has the 1:13:23.240 --> 1:13:25.680 power first, but who is better at governing, steering and 1:13:25.720 --> 1:13:29.400 integrating that power in a healthy and sustainable and strengthening 1:13:29.439 --> 1:13:32.240 way into your society. And we saw this with social 1:13:32.280 --> 1:13:36.120 media because the US beat China to the psychological bazooka 1:13:36.280 --> 1:13:39.439 behavior modification machine of social media, and then we had 1:13:39.439 --> 1:13:41.559 no idea how to govern it. So we flipped around. 1:13:41.600 --> 1:13:43.759 We blew off our own brain with the brain row economy. 1:13:45.080 --> 1:13:47.519 So and by the way China regulates social media, they 1:13:47.520 --> 1:13:49.040 do a whole bunch of suffing. When you open up 1:13:49.080 --> 1:13:52.479 a douyen TikTok in China, it's their version of TikTok, 1:13:52.720 --> 1:13:55.280 and you scroll, you get videos about who won the 1:13:55.280 --> 1:13:59.360 Nobel Prize, financial advice, here's the new quantum physics theory. 1:14:00.040 --> 1:14:02.320 You know, here's patriotism videos. And obviously there's problems with that. 1:14:02.360 --> 1:14:04.000 We don't want to do it that way, but the 1:14:04.040 --> 1:14:05.880 point is you don't have to do it in the 1:14:05.920 --> 1:14:08.360 wild west. Blow off your own brain. And now if 1:14:08.400 --> 1:14:10.360 we release it in a way that automates all the 1:14:10.400 --> 1:14:13.400 labor with no transition plan. It's like, great, we pumped 1:14:13.479 --> 1:14:16.080 up our steroids, but we just burst our lungs. Yeah, right, 1:14:16.120 --> 1:14:19.360 the societal body, we do that. So what you just 1:14:19.400 --> 1:14:22.040 outlined was a set of interventions that we can be 1:14:22.080 --> 1:14:24.760 exploring to help smooth this transition. And we're in a 1:14:24.760 --> 1:14:28.120 competition with China for who's better at making this transition 1:14:28.520 --> 1:14:30.559 to an AI integrated world. 1:14:31.280 --> 1:14:34.080 Where are you on the transition curve? I mean, I mean, 1:14:34.080 --> 1:14:36.960 we talk about flattering it, but how quick? I mean, honestly, 1:14:37.080 --> 1:14:40.040 if we're sitting here a year from now having this conversation, 1:14:40.640 --> 1:14:43.320 are we looking at that you know, ten percent unemployment, 1:14:43.680 --> 1:14:47.360 not that twenty necessarily, which is that threshold for fascism 1:14:47.479 --> 1:14:49.320 and a whole other societal collap. 1:14:49.040 --> 1:14:51.720 But it's going to be confusing and spiking hard to 1:14:51.720 --> 1:14:55.960 predict exactly because just think about your own experience with 1:14:56.040 --> 1:14:57.960 AI so far two years ago, it can barely write 1:14:57.960 --> 1:15:00.040 an essay, and now it can do some part so 1:15:00.120 --> 1:15:02.000 of your work pretty well and other parts like that 1:15:02.080 --> 1:15:05.360 is a really dumb error. And so we're going to 1:15:05.439 --> 1:15:08.040 see not that much job, not that much job LOFs 1:15:08.400 --> 1:15:11.280 like entry level stuff getting sort of switched around and 1:15:11.320 --> 1:15:13.400 then it's going to hit really really quickly when it 1:15:13.439 --> 1:15:15.400 crosses the next spreshold, just like Mythos did. 1:15:15.439 --> 1:15:18.240 There's another confusing aspect about AI that people in our 1:15:18.280 --> 1:15:21.880 space called AI jaggedness, which is that there's certain capabilities 1:15:21.920 --> 1:15:25.480 like in cyber hacking, that it is already superhuman, already superhumans, 1:15:25.960 --> 1:15:28.840 while it'll still make a very basic dumb mistake on 1:15:28.880 --> 1:15:31.600 something else. And I think part of what's confusing for 1:15:31.640 --> 1:15:33.840 people that naturally has them say is this just hype 1:15:33.840 --> 1:15:35.720 and these companies are trying to hype this stuff is 1:15:35.760 --> 1:15:37.760 if you just look at the dumb examples where it's 1:15:37.800 --> 1:15:40.720 messing up, you're like, this thing isn't that powerful. We 1:15:40.840 --> 1:15:44.880 have never been confronted with the technology that is simultaneously 1:15:45.240 --> 1:15:49.320 sci Fi level superhuman that makes a person who's very 1:15:49.360 --> 1:15:52.479 deep in security call it like seeing Jesus. At the 1:15:52.520 --> 1:15:55.280 same time that that same technology can like mistake how 1:15:55.280 --> 1:15:58.000 many rs are in the word strawberry, like, we have 1:15:58.080 --> 1:15:59.599 just not seen that. And so it's just I want 1:15:59.600 --> 1:16:01.840 people to not for their own psychology that this is 1:16:01.880 --> 1:16:05.920 a new psychological object that our normal intuitions about how 1:16:05.960 --> 1:16:08.559 to evaluate something we have to get more nuanced. 1:16:09.240 --> 1:16:11.479 And part of the nuance is a deeper understanding that 1:16:11.600 --> 1:16:13.679 we're not just talking about apps here, we're talking about 1:16:13.680 --> 1:16:14.880 the physical world as well. 1:16:15.160 --> 1:16:15.519 That's right. 1:16:15.960 --> 1:16:19.000 Was a big headline that I hope people paid attention 1:16:19.080 --> 1:16:24.080 to when Elon announced that in his original factory here 1:16:24.080 --> 1:16:26.920 in Fremont, California, that he's converting the s and the 1:16:26.920 --> 1:16:31.280 ex Tesla cars now to humanoid robotics and his goal 1:16:31.360 --> 1:16:34.040 is ultimately a million You know, that's Elon who they 1:16:34.080 --> 1:16:36.719 all knows, you know, his goal setting. But the notion 1:16:37.080 --> 1:16:40.719 that he's converting that factory from cars to humanoid robotics 1:16:41.000 --> 1:16:44.320 and AI now into the physical world is that. I mean, 1:16:45.280 --> 1:16:48.080 we're seeing driverless cars and if you haven't seen them 1:16:48.080 --> 1:16:50.640 in you know, your home state, you're going about to 1:16:50.760 --> 1:16:54.000 you're going to see flying cars or they're just quad 1:16:54.000 --> 1:16:56.680 copters basically, but they're coming soon. We're going to be 1:16:56.760 --> 1:16:58.240 doing a lot more of that run by the way. 1:16:58.560 --> 1:17:01.360 That's great. We can have a world of quad copters 1:17:01.439 --> 1:17:05.760 and you know, innovation while not racing to replace us economically, 1:17:05.840 --> 1:17:08.439 replace us socially. So Mark Zuckerberg, you know, designs your 1:17:08.479 --> 1:17:10.599 kids friends rather than you had them having actual friends 1:17:10.840 --> 1:17:13.040 replace us politically and not having political power and then 1:17:13.040 --> 1:17:16.400 replace us physically by owning our physical presence in our 1:17:16.439 --> 1:17:19.800 in robots. So again, I think people might hear this 1:17:19.840 --> 1:17:22.519 as an anti technology conversation. Yeah, you know, you talked 1:17:22.520 --> 1:17:25.840 about your legacy and your father and grandfather and we 1:17:25.840 --> 1:17:30.160 were talking backstage. Ye A's's father started the Macintosh projected 1:17:30.200 --> 1:17:32.559 at Yeah. Yeah, you know, we come from a legacy 1:17:32.640 --> 1:17:35.519 where the word humane is about an inspiring vision about 1:17:35.520 --> 1:17:38.879 technology that's actually integrated and in service of our humanity, 1:17:39.280 --> 1:17:41.360 of a pro human future. That's what all of this 1:17:41.400 --> 1:17:44.240 is motivated for. So I get excited about you know, flying, 1:17:44.439 --> 1:17:46.160 you know, cars that have cool you know. 1:17:46.120 --> 1:17:48.200 Aif And just to say too, just as my own 1:17:48.200 --> 1:17:51.000 personal experiences, you know, I spend a big portion of 1:17:51.040 --> 1:17:53.519 my life. I found it a thing called Earth Species Project. 1:17:54.000 --> 1:17:56.000 We're now around forty people and you're. 1:17:55.840 --> 1:17:56.799 The biggest consumer. 1:17:57.600 --> 1:18:00.679 I wouldn't say biggest consumer. Yeah exactly, it's it's men. Sorry, guys, 1:18:02.000 --> 1:18:05.320 first infinite scroll now this, but you know, we were 1:18:05.400 --> 1:18:08.600 using AI to translate animal language animal communication. One of 1:18:08.600 --> 1:18:12.360 our researchers just discovered just before she joined us that like, 1:18:12.439 --> 1:18:14.840 not only do dolphins have names that they call each 1:18:14.840 --> 1:18:17.439 other by that their mothers teach them. They will talk 1:18:17.439 --> 1:18:19.320 about each other in the third person, so they'll talk 1:18:19.360 --> 1:18:21.720 about another dolphin that isn't here. So it's like one 1:18:21.760 --> 1:18:23.280 of the biggest hallmarks of language to be able to 1:18:23.280 --> 1:18:25.720 talk about something that's not here and not now. But 1:18:25.760 --> 1:18:29.400 they will continue to use their mother's name even after 1:18:29.439 --> 1:18:34.240 she's died. Sweet right, it's beautiful. These are the kinds 1:18:34.240 --> 1:18:37.040 of things that AI can show us and teach us 1:18:37.040 --> 1:18:39.479 about the world and connect us with the natural world, 1:18:39.560 --> 1:18:42.240 the world around us, and show things we couldn't possibly imagine. 1:18:42.320 --> 1:18:44.200 So I just want everyone to hear it's not that 1:18:44.240 --> 1:18:47.519 it's just here saying no AI. It's just saying not AI. 1:18:47.840 --> 1:18:51.640 In this way that technological progress might be inevitable, but 1:18:51.720 --> 1:18:55.280 the way that AI rolls out is not. And every 1:18:55.320 --> 1:18:58.200 time somebody says it's inevitable, it's like casting a spell. 1:18:58.280 --> 1:19:01.120 That's where because if if it's inevitable, then there's nothing 1:19:01.160 --> 1:19:02.840 we can do, then you shouldn't possibly act. 1:19:02.880 --> 1:19:05.320 There's a button we're hitting called commit suicide. Commits suicide? 1:19:05.520 --> 1:19:07.200 Is it inevitable that we all hit that button? 1:19:07.240 --> 1:19:07.360 No? 1:19:07.400 --> 1:19:09.560 When it's called you know, ten percent GDP growth And 1:19:09.680 --> 1:19:12.519 like an innovation, we pushed the button. If we could 1:19:12.520 --> 1:19:15.280 collectively see that the button we're pushing is more nuanced 1:19:15.280 --> 1:19:17.720 than that, that there's some threshold by which we are 1:19:17.960 --> 1:19:20.880 essentially committing civilizational suicide. We're not going to have a 1:19:20.920 --> 1:19:24.879 human future. If you ask anybody what gives me hope? 1:19:25.000 --> 1:19:27.240 We've been on the road with this film. You walk 1:19:27.280 --> 1:19:29.519 people through the basic facts we've talked through, and you 1:19:29.560 --> 1:19:31.599 ask who here is stoked about the future that we're 1:19:31.640 --> 1:19:34.479 headed to. I was even at the Miami Tech Summit 1:19:34.520 --> 1:19:36.479 in front of a pro business pro AI. A lot 1:19:36.520 --> 1:19:39.479 of people invested in it. Again, we're invested in in 1:19:39.520 --> 1:19:42.040 positive technology too, but I was able to see that 1:19:42.200 --> 1:19:44.040 entire audience is like, no, I don't want that either. 1:19:44.200 --> 1:19:45.000 Wow. Yeah. 1:19:45.040 --> 1:19:47.080 So again I think it's like, as long as you 1:19:47.120 --> 1:19:49.000 give people the off ramp, there are ways we can 1:19:49.080 --> 1:19:52.360 have technology that's in service of making life better. We 1:19:52.400 --> 1:19:55.040 need to be funding and innovating in that way, and 1:19:55.080 --> 1:19:56.680 we need to be blocking off the parts that are 1:19:56.920 --> 1:20:00.479 hurting our children, that are, you know, diminishing people cognitive 1:20:00.479 --> 1:20:04.160 capacities or replacing their kids' relationships with AI companions. And 1:20:04.200 --> 1:20:05.479 we can do that. And you've done some of that 1:20:05.479 --> 1:20:08.439 with here in California. So you know, it's not that 1:20:08.520 --> 1:20:12.400 I am or weird by default optimistic about the default trajectory, 1:20:12.439 --> 1:20:15.120 not at all. It's that if we have the clarity, 1:20:15.680 --> 1:20:17.800 we can put our hand in the steering wheel and 1:20:17.840 --> 1:20:18.880 we can steer it somewhere else. 1:20:19.400 --> 1:20:21.960 This notion of accelerate and steer. 1:20:21.880 --> 1:20:24.439 Yeah, yes, right, exactly what happens when you accelerate and 1:20:24.479 --> 1:20:27.519 you don't steer, you obviously crash. It's just it's like 1:20:27.560 --> 1:20:30.679 not rocket scient there, it's one hundred percent the likely outcome. 1:20:32.680 --> 1:20:36.160 What just as we wrap up this notion of governance, 1:20:36.240 --> 1:20:38.679 going back to that is it's it's the foundation here 1:20:39.240 --> 1:20:43.320 which the regulations and and and the relationships are formed, 1:20:43.320 --> 1:20:46.680 the partnerships to begin to address the common humanity and 1:20:46.680 --> 1:20:49.600 the common threat and the common cause, common opportunities that 1:20:49.680 --> 1:20:53.559 present themselves. This notion of catcher, I mean, you've got 1:20:53.640 --> 1:20:57.720 these packs, You've got so much concentrated wealth. Yeah, we're 1:20:57.720 --> 1:21:02.160 going to likely have the first few trillionaires this year. Yeah, 1:21:02.200 --> 1:21:06.519 this calendar years right, yeah, yep, these IPOs, I mean 1:21:06.560 --> 1:21:09.760 it's just I'm a now, it's a big surplus in 1:21:09.800 --> 1:21:13.480 the state, the abundance, the GDP I mean, it's frothy, 1:21:14.040 --> 1:21:15.840 as they say, but a lot of that now is 1:21:15.880 --> 1:21:18.800 going to you know, make sure that you know we're 1:21:18.800 --> 1:21:24.120 protecting incumbents against innovation. You know, some incumbent capitalism, not 1:21:24.320 --> 1:21:28.960 entrepreneurial capitalism, necessarily innovation capitalism. There's that friction that's always 1:21:28.960 --> 1:21:31.280 ongoing and those that are just going to do everything 1:21:31.320 --> 1:21:33.680 to hammer as they did with social media, to make 1:21:33.720 --> 1:21:37.240 sure there's no regular we're still debating section two thirty 1:21:36.920 --> 1:21:39.880 that Christ yeah in this country. So how do we 1:21:39.920 --> 1:21:42.720 start to break that reality? 1:21:42.840 --> 1:21:45.760 I think money in politics, that's just that's right, that's 1:21:46.320 --> 1:21:48.320 at the anti just just a little bit even further, 1:21:48.360 --> 1:21:51.200 which is that as we get more trillionaires, like they 1:21:51.200 --> 1:21:54.800 can hire private security, but that still requires relying on 1:21:54.880 --> 1:21:55.479 human beings. 1:21:55.479 --> 1:21:58.080 But you're just pointing out that we're heading into like 1:21:58.240 --> 1:22:01.120 a world we have drone army, we have human eight 1:22:01.200 --> 1:22:05.920 rowood armies. When trillionaires can just buy like their drone 1:22:06.040 --> 1:22:09.160 armies to fight, like, we enter into techno feudalism, right, 1:22:09.200 --> 1:22:11.599 And so that is the world that if we see, 1:22:12.520 --> 1:22:14.519 if we don't do something, we're going to end up into. 1:22:14.680 --> 1:22:17.519 But to answer your question, the it's all about the 1:22:17.560 --> 1:22:20.639 campaign Loving, it's all about that, And one hundred million, 1:22:20.720 --> 1:22:24.320 one hundred and ninety million dollars has gone into basically 1:22:24.400 --> 1:22:29.240 AI accelerationist packs and funding is for this midterm, for 1:22:29.280 --> 1:22:32.240 this mintrum electional Love interm just the midterms. That's so 1:22:32.240 --> 1:22:34.719 one hundred ninety million. I believe the presidential is two billions. 1:22:34.760 --> 1:22:37.120 So basically ten percent of the presidential is going into 1:22:37.320 --> 1:22:39.200 just the midtrums for AI alone, not even for the 1:22:39.240 --> 1:22:39.600 rest of it. 1:22:39.880 --> 1:22:41.160 And it's not to regulate. 1:22:41.360 --> 1:22:43.920 It's not to regulate it. It's it's to say, remove 1:22:43.960 --> 1:22:45.160 everything and go as fast. 1:22:45.000 --> 1:22:46.479 As possible, right, let it rip. 1:22:46.760 --> 1:22:49.880 If you had everyone in the world, I think here 1:22:49.960 --> 1:22:52.679 the conversation we just had and at a basic level 1:22:52.840 --> 1:22:54.680 common sense looking at your children in the eye and 1:22:54.720 --> 1:22:57.920 say are you stoked about that future? No one wants that. 1:22:58.680 --> 1:23:01.519 So the key piece of agency is going into the 1:23:01.520 --> 1:23:05.719 midterm elections not voting for people who have taken money 1:23:05.720 --> 1:23:09.759 from those AI accelerationist groups or don't have a position 1:23:09.800 --> 1:23:12.559 on AI. That's trying to steer away from these outcomes. Now, 1:23:12.560 --> 1:23:14.720 we obviously have to articulate that in a clearer way. 1:23:14.800 --> 1:23:16.639 What does it mean to have kind of a pro 1:23:16.760 --> 1:23:20.200 human platform in future, and the companies try to make 1:23:20.200 --> 1:23:22.680 the conversation inaccessible, like oh, well, you don't understand a 1:23:22.840 --> 1:23:24.080 They're trying to make it seem like you don't know 1:23:24.160 --> 1:23:26.120 how to regulate live to put us in charge. 1:23:26.400 --> 1:23:28.920 What they call this the under the hood bias, where 1:23:28.920 --> 1:23:30.400 it's as if people who know how to make the 1:23:30.439 --> 1:23:33.160 biggest engines know how to lay out cities and traffic lights, 1:23:33.360 --> 1:23:35.559 and it's just not true for people who know how 1:23:35.600 --> 1:23:38.040 to make car engines are not the best people for 1:23:38.120 --> 1:23:39.840 knowing how to make cars safe and. 1:23:39.840 --> 1:23:43.880 Prevent recidents work exactly. So it's pretty simple. It's like, 1:23:44.080 --> 1:23:46.880 do you want an anti human future in which you 1:23:46.880 --> 1:23:49.519 will be permanently disempowered where no one has an incentive 1:23:49.520 --> 1:23:52.080 except for charity to pay your bills for you? And 1:23:52.520 --> 1:23:55.519 you you want the companies that took your job, you 1:23:55.560 --> 1:23:56.960 want to be dependent on them for the rest of 1:23:57.000 --> 1:23:59.240 your life to pay your bills for you with no 1:23:59.320 --> 1:24:03.320 economic leverle no. So this is the final window. You 1:24:03.360 --> 1:24:04.920 know you want to vote for people who are going 1:24:04.960 --> 1:24:08.160 to protect you economically, protect you, socially, protect you, politically 1:24:08.160 --> 1:24:11.880 meaning protect our jobs, protect our vote, voting pro human 1:24:12.320 --> 1:24:14.839 and obviously that has to get articulated even more clearly. 1:24:15.240 --> 1:24:17.280 But that is the number one way that people can 1:24:17.320 --> 1:24:19.679 make a difference in the short term. There's other things too, 1:24:19.720 --> 1:24:23.000 like boycotting companies that are enabling mass surveillance. You know, 1:24:23.040 --> 1:24:25.800 when the company's subscriptions you know, go down by a lot, 1:24:25.840 --> 1:24:27.400 they really need their numbers to be going up. So 1:24:27.439 --> 1:24:29.720 the companies are more vulnerable than you think, and you're 1:24:29.760 --> 1:24:31.960 more powerful than you think. Not just if you unsubscribe 1:24:31.960 --> 1:24:34.280 and boycott them, but get your company, get your church 1:24:34.280 --> 1:24:37.240 group to do that too, and when those numbers start 1:24:37.240 --> 1:24:38.920 to change, it actually has a difference. 1:24:39.040 --> 1:24:41.479 Scott Galloway has been talking a lot about that as well. 1:24:42.479 --> 1:24:46.600 Absolutely got to use whatever power your disposal. Let's just 1:24:46.680 --> 1:24:51.720 briefly talk then about the power I mean, free and 1:24:51.760 --> 1:24:56.080 fair elections. We talk about truth trust more broadly, deep fakes, 1:24:56.560 --> 1:25:00.479 political ads. I mean I've seen stuff, you know, meetings, 1:25:00.520 --> 1:25:03.840 conversations I've had that are I mean next level what's 1:25:03.880 --> 1:25:06.920 out the you yes, just the bs, it's already out there. 1:25:07.080 --> 1:25:10.479 The ability to manipulate the crowd and the context of 1:25:10.680 --> 1:25:13.280 you know, social media, the algorithms I mean, now you've 1:25:13.280 --> 1:25:15.240 got you know, concentrated in the hands of a few 1:25:15.560 --> 1:25:17.519 in that respect, I mean that whole thing. I mean 1:25:17.760 --> 1:25:20.680 you you guys have talked about free and fair elections. 1:25:21.479 --> 1:25:24.800 We talk about the timelines not only in job displacement, 1:25:25.040 --> 1:25:30.439 but timelines to get this right domestically globally. I mean 1:25:32.280 --> 1:25:35.800 twenty twenty six, you're talking mid terms. I mean this 1:25:35.880 --> 1:25:37.519 is I mean, we only have a few more at 1:25:37.520 --> 1:25:40.160 bats to get this right. That's right? Or is that overstated? No, 1:25:41.000 --> 1:25:41.639 it's a few more. 1:25:42.280 --> 1:25:44.639 Certainly by twenty twenty eight, Like that'll be the last 1:25:45.120 --> 1:25:47.639 human election. Like it's going to be AIS running all 1:25:47.640 --> 1:25:50.200 of the election, election campaigns, the ads, doing all the 1:25:50.320 --> 1:25:53.240 both the information and disinformation, because human beings just can't 1:25:53.240 --> 1:25:55.240 operate at that speed and are not that effective. 1:25:55.400 --> 1:25:56.360 Yeah, so this is it. 1:25:56.439 --> 1:25:58.559 This is the window. But and I know, I just 1:25:58.560 --> 1:26:00.439 want to like get the human experience level. We've been 1:26:00.479 --> 1:26:03.680 talking about some hard stuff, the last little level, and 1:26:03.840 --> 1:26:05.040 you know, I just want to say, you know, we 1:26:05.080 --> 1:26:08.160 struggle with how to communicate this in a way that's responsible, 1:26:08.200 --> 1:26:11.559 because here's the trade. Right, it's hard to face this. 1:26:11.680 --> 1:26:13.799 But if we don't face it, we just like, look away, 1:26:14.760 --> 1:26:16.040 what are we going to get? We're going to get 1:26:16.080 --> 1:26:19.240 the default anti human path, and so there's this trade 1:26:19.280 --> 1:26:21.960 where we the only way out is through like it is. 1:26:22.000 --> 1:26:23.559 We call it kind of like a rite of passage, 1:26:23.560 --> 1:26:28.000 like our ability to confront basically a shadow of a 1:26:28.040 --> 1:26:31.960 technology and the default future that that brings. If we 1:26:32.000 --> 1:26:34.280 can see that clearly, and if we can know that, 1:26:34.400 --> 1:26:36.240 you know, and I know that we don't want that, 1:26:36.479 --> 1:26:38.519 and if g and Trump and you know, the people 1:26:38.520 --> 1:26:40.599 at the highest levels of these governments, Because you ask 1:26:40.640 --> 1:26:43.360 any reasonable person at a very high level of national 1:26:43.400 --> 1:26:45.679 security on any side, and you say, do you want 1:26:45.720 --> 1:26:47.960 AIS that are going road can hack into any computer 1:26:48.040 --> 1:26:51.439 system and are already mining cryptocurrency? Does that sound good 1:26:51.479 --> 1:26:53.519 to you? Does that sound safe to you? At a 1:26:53.640 --> 1:26:57.639 universal level, it's not. So there's actually much more common ground. 1:26:57.960 --> 1:27:00.000 And even you know forty se I think it's already 1:27:00.080 --> 1:27:02.000 the case that fifty seven percent of Americans think that 1:27:02.040 --> 1:27:04.760 the risks of AI currently outweigh the benefits. I don't 1:27:04.760 --> 1:27:07.120 like that stat because it makes it too like it's 1:27:07.160 --> 1:27:09.599 all bad versus all good or something like that. There's 1:27:09.640 --> 1:27:12.719 already the pro human AI Declaration, where forty six groups 1:27:12.760 --> 1:27:15.559 came together and said, we agree on these five principles 1:27:15.600 --> 1:27:17.160 to make a pro human future. You can look it up. 1:27:17.160 --> 1:27:19.920 It's human Statement dot org. That's the one that also 1:27:19.920 --> 1:27:23.759 includes again Glen Back, Bernie Sanders, Steve Bannon, all these people. 1:27:24.160 --> 1:27:26.760 I believe it's sixty five percent of Americans believe we 1:27:26.760 --> 1:27:29.559 should not create superintelligence until we know how to do 1:27:29.600 --> 1:27:33.040 it provably, safely and controlled. Sounds like a pretty basic thing, 1:27:33.080 --> 1:27:35.160 like let's not do something. It's like, should we build 1:27:35.160 --> 1:27:36.880 a nuclear bomb that until we know how to do 1:27:36.920 --> 1:27:38.840 it safely or nobody won't set off of if we 1:27:38.840 --> 1:27:41.000 won't ignite the atmosphere, Probably we should wait to do that. 1:27:41.560 --> 1:27:43.799 So this is not a radical proposal. 1:27:43.840 --> 1:27:46.000 This is not do you want to know how many, 1:27:46.120 --> 1:27:48.680 like what percentage of Americans think that we should just 1:27:48.720 --> 1:27:53.320 go as fast as possible, unfettered, non regulated AI. What 1:27:53.439 --> 1:27:55.360 percent of Americans? Five percent? 1:27:55.720 --> 1:27:57.000 Literally, yeah, literally five percent. 1:27:57.880 --> 1:28:00.559 So actually it's the most popular to run on. That's 1:28:00.640 --> 1:28:02.639 right to do the like the safe thing, And. 1:28:02.600 --> 1:28:04.960 It's whether you're a Democrat or Republican, you don't want 1:28:04.960 --> 1:28:07.280 to be surveilled by AIS. Whether you're Democrat Republican, you 1:28:07.280 --> 1:28:09.120 don't want AIS taking your jobs, which it will do 1:28:09.160 --> 1:28:10.280 equally to both sides. 1:28:10.360 --> 1:28:13.320 But do you then subscribe to the burning AOC frame 1:28:13.600 --> 1:28:17.160 just shut down the data centers and moratorium. 1:28:16.479 --> 1:28:19.799 And tell I think of it as those data centers, 1:28:19.840 --> 1:28:22.599 and it's like I want a pro human data center policy. 1:28:22.800 --> 1:28:25.120 It's like, you get to build the data center when 1:28:25.360 --> 1:28:27.120 these conditions are met and we know that it's a 1:28:27.720 --> 1:28:29.479 pro usure. I'm not saying that's easy. I'm not saying 1:28:29.520 --> 1:28:32.360 no articulation that because I want people to hear. It's 1:28:32.400 --> 1:28:35.719 not just no to all of it. It's making sure 1:28:35.720 --> 1:28:38.560 that the conditions are the steering is built in. So 1:28:38.600 --> 1:28:40.519 when you see that data center, you should ask, is 1:28:40.520 --> 1:28:43.920 that data center here to basically enhance my life and 1:28:43.920 --> 1:28:45.240 strengthen my family. 1:28:45.479 --> 1:28:47.439 You were even just saying data center was solar. Mostly 1:28:47.520 --> 1:28:48.599 data centers aren't solar. 1:28:49.200 --> 1:28:49.559 That's right. 1:28:50.080 --> 1:28:53.000 We're turning back on cold plants. That's right, old natural 1:28:53.000 --> 1:28:54.680 gas plants that are exactly yea. 1:28:55.680 --> 1:28:58.519 Often in the sci fi movies, when is it the 1:28:58.560 --> 1:29:01.559 case that human beings act actually stop all their bickering 1:29:02.200 --> 1:29:05.320 and they start coordinating, it's when the aliens come, right, Yeah, 1:29:05.640 --> 1:29:07.960 we are summoning the demon. We are summoning the alien. 1:29:08.280 --> 1:29:10.439 And if we can see it that way, then it's 1:29:10.439 --> 1:29:12.679 sort of like a Game of Thrones, winter is coming. 1:29:13.120 --> 1:29:14.640 We have to understand winter is coming, then all the 1:29:14.680 --> 1:29:16.760 fighting in West Ros can like pause for just long 1:29:16.880 --> 1:29:19.200 enough that we can deal with it. That's this moment, 1:29:19.200 --> 1:29:22.160 because otherwise it just feels completely hopeless, like how are 1:29:22.200 --> 1:29:24.720 we going to deal with all the finance reforms when 1:29:24.720 --> 1:29:29.639 we can't when has Congress actually done anything? And yet 1:29:29.800 --> 1:29:33.560 there is this one moment where like all of humanity 1:29:33.760 --> 1:29:36.439 is on one side, there is a human movement. That's 1:29:36.439 --> 1:29:39.120 what I think the social media stuff shows. If we 1:29:39.160 --> 1:29:41.360 don't think of this as just an AI problem, but 1:29:41.600 --> 1:29:46.160 as a technology encroaching onto our humanity, overreaching into our 1:29:46.240 --> 1:29:49.639 humanity problem, then actually there is massive momentum, more than 1:29:49.640 --> 1:29:51.360 we ever thought was possible, because what we have to 1:29:51.439 --> 1:29:55.360 do is juice those like the momentum that's already there. 1:29:56.200 --> 1:30:00.000 Yeah. Well, look, in the absence of a federal leadership, 1:30:00.080 --> 1:30:02.240 California will continue to assert itself. I believe in the 1:30:02.240 --> 1:30:05.559 power of emulation successfullys clues will continue to try to 1:30:05.920 --> 1:30:09.040 iterate on this and lean in. But look this, you 1:30:09.080 --> 1:30:11.320 know the clarity of you guys bring to this conversation, 1:30:11.439 --> 1:30:14.560 The importance of this conversation being brought to scale and 1:30:15.320 --> 1:30:19.280 brought it into consciousness and the imperative of seeing this 1:30:19.439 --> 1:30:22.719 documentary again. The documentary is called. 1:30:22.600 --> 1:30:25.520 The AI doc or How I Became an apocaly optimism. 1:30:25.640 --> 1:30:29.960 And we can't let that slip twice because you've used 1:30:30.080 --> 1:30:33.720 a word that people are not familiar with, which is 1:30:34.080 --> 1:30:37.040 a good way to end, and that is this convergence 1:30:37.120 --> 1:30:41.719 of optimism and pessimism, a little more optimism than pessimism 1:30:41.960 --> 1:30:42.599 in a better place. 1:30:42.640 --> 1:30:44.600 And it's about agency. And I'll just leave you with 1:30:44.640 --> 1:30:47.200 a quote that I loved from a meditation teacher who 1:30:47.240 --> 1:30:50.840 talked who happened to be a meditation teacher, and it's 1:30:50.880 --> 1:30:54.080 from the Army Corps of Engineers, which is that the 1:30:54.240 --> 1:30:58.880 difficult we do today, the impossible takes just a little longer. 1:31:00.160 --> 1:31:04.880 Like it, and just end by saying it really actually 1:31:05.040 --> 1:31:07.960 isn't about being an optimist or a pessimist, because to 1:31:08.439 --> 1:31:11.120 choose that label for yourself, it's a sort of it's 1:31:11.160 --> 1:31:13.000 to take a backseat, like to sit down and just 1:31:13.040 --> 1:31:15.160 be like, I'm just going to accept that it'll either 1:31:15.200 --> 1:31:19.799 come out well or not, versus taking responsibility for trying 1:31:19.840 --> 1:31:22.639 to see clearly to shift the world to go well 1:31:24.200 --> 1:31:27.800 and I think that's what everyone listening can can do, 1:31:27.960 --> 1:31:31.880 is that this can all feel like too big? What 1:31:32.400 --> 1:31:34.519 can I do? And then you realize even like the 1:31:34.640 --> 1:31:36.840 like the CEOs of the company sort of feel a 1:31:36.920 --> 1:31:40.960 similar way. But this is not just about what we 1:31:41.080 --> 1:31:44.280 must do. This is fundamentally a question of like who 1:31:44.320 --> 1:31:46.760 we must be. That if we are the kind of 1:31:46.800 --> 1:31:49.400 people that aren't looking for a path, and the path 1:31:49.479 --> 1:31:52.759 is certainly not clear, doesn't seem obvious or even possible. 1:31:53.080 --> 1:31:55.240 But if we're the kind of people that don't look 1:31:55.240 --> 1:31:56.720 for the path, we definitely won't find it. 1:31:56.760 --> 1:31:57.240 If it's there. 1:31:57.640 --> 1:31:59.479 If we are the kinds of people that do look 1:31:59.520 --> 1:32:01.960 for the path, then if it's there, we have the 1:32:02.080 --> 1:32:03.599 opportunity to find it. 1:32:03.640 --> 1:32:07.639 And just maybe one last thing is people listening to this, 1:32:07.640 --> 1:32:10.439 this is a lot. Your role is not to take 1:32:10.439 --> 1:32:12.240 on this whole problem. You don't have to do that. 1:32:12.680 --> 1:32:14.280 There's some people who are soldiers and there's some people 1:32:14.320 --> 1:32:17.400 who are civilians. But your role is to be part 1:32:17.439 --> 1:32:20.800 of the collective immune system against this anti human future. 1:32:20.960 --> 1:32:22.960 One simple way you can do that is to share 1:32:23.080 --> 1:32:26.800 this conversation yeah, with literally the most powerful people that 1:32:26.800 --> 1:32:29.320 you know, and ask them to watch it. And to 1:32:29.320 --> 1:32:31.240 share it with the most powerful people that they know. 1:32:31.640 --> 1:32:33.960 And if you've done that, you can say, as long 1:32:34.000 --> 1:32:35.600 as you are spreading the word and being part of 1:32:35.600 --> 1:32:38.640 that immune system, you can rest at home, kiss your 1:32:38.720 --> 1:32:41.040 children at night, focus on the things that all of 1:32:41.040 --> 1:32:43.599 this is about anyway, which is what do we love 1:32:43.800 --> 1:32:45.840 about the world, what do we love about life? That 1:32:45.880 --> 1:32:48.400 we want to continue and come from that place because 1:32:48.439 --> 1:32:51.280 that is the energy that we will that will inspire 1:32:51.360 --> 1:32:53.599 other people to want to take those other actions too. 1:32:53.720 --> 1:32:55.719 And I know, Kevin, you ended up watching an earlier 1:32:55.760 --> 1:32:58.280 presentation that we did the Aidlemma. Yeah, I think somebody 1:32:58.320 --> 1:32:59.760 said that you watched it like three times and share 1:32:59.800 --> 1:33:02.320 with all your staff. How did you end up hearing 1:33:02.320 --> 1:33:02.760 about it? 1:33:03.640 --> 1:33:06.880 Well, I mean, come on, hearing about it from you guys. 1:33:07.920 --> 1:33:10.799 You guys. I was able to get the early early 1:33:10.880 --> 1:33:13.800 preview from the two of you, and I was able 1:33:14.000 --> 1:33:18.680 to devour it, took notes, and then shared it universally 1:33:18.720 --> 1:33:22.240 with everybody around me. Look, you know, the spirit of 1:33:22.320 --> 1:33:23.920 you guys just DoD it. I couldn't agree with you more. 1:33:23.960 --> 1:33:26.479 This notion of agency is so important, and we talk 1:33:26.520 --> 1:33:28.479 about that on the podcast all the time. But this 1:33:28.560 --> 1:33:30.760 notion of the future. It's not something to experience, something 1:33:30.760 --> 1:33:34.920 to manifest futures inside of us. And so it's decisions, 1:33:34.960 --> 1:33:36.000 not conditions. 1:33:35.600 --> 1:33:36.120 That's exactly. 1:33:36.200 --> 1:33:39.559 And so this idea that we are powerless, it's just bullshit. 1:33:40.040 --> 1:33:43.960 It's complete bullshit. Everything that we laid out is an 1:33:43.960 --> 1:33:46.920 opportunity to do better and be better. And I think 1:33:46.960 --> 1:33:49.479 the spirit of the calt arms for everybody, as we 1:33:49.520 --> 1:33:51.679 all have a role to play, and those roles are different, 1:33:51.720 --> 1:33:53.439 and no one has to be you know, you don't 1:33:53.479 --> 1:33:56.760 have to be overwhelmed. But this notion of just being 1:33:56.840 --> 1:34:01.160 present in the conversation and and and in the conversation 1:34:01.240 --> 1:34:05.000 and sharing it, I think is foundational. So, guys, this 1:34:05.040 --> 1:34:11.600 is really important. The timeliness of this conversation it cannot overstate. 1:34:12.439 --> 1:34:14.639 And so I'm very grateful for you to be out 1:34:14.960 --> 1:34:19.080 on the road all across this country sharing this remarkable documentary. 1:34:19.080 --> 1:34:22.480 I encourage everybody go out and watch it and more importantly, 1:34:22.840 --> 1:34:25.920 share it and not fall prey to any of the 1:34:26.000 --> 1:34:30.400 citizensm and negativity. Maintain your sense of optimism again, we 1:34:30.439 --> 1:34:32.040 can shape the future. Thank you both