WEBVTT - UL NO. 436: Thoughts on the Future of AI & Societal Stability 0:00:00.110 --> 0:00:02.240 What do you call an endpoint security product that works 0:00:02.240 --> 0:00:05.600 perfectly but makes users miserable a failure? The old approach 0:00:05.600 --> 0:00:08.420 to endpoint security is to lock down employee devices and 0:00:08.420 --> 0:00:11.870 roll out changes through forced restarts, but it just doesn't work. 0:00:11.900 --> 0:00:14.720 It is miserable because they've got a mountain of support tickets. 0:00:14.720 --> 0:00:17.570 Employees start using personal devices just to get their work done, 0:00:17.570 --> 0:00:19.880 and executives opt out the first time it makes them 0:00:19.880 --> 0:00:22.190 late for a meeting. You can have a successful security 0:00:22.190 --> 0:00:24.920 implementation unless you work with end users, and that's where 0:00:24.920 --> 0:00:28.100 Kaleid comes in. Their user first device trust solution notifies 0:00:28.100 --> 0:00:30.590 users as soon as it detects an issue on their device, 0:00:30.590 --> 0:00:32.720 and teaches them how to solve it without needing help 0:00:32.720 --> 0:00:35.930 from it. That way, untrusted devices are blocked from authenticating, 0:00:35.930 --> 0:00:38.810 but users don't stay blocked. Kaleidos is designed for companies 0:00:38.810 --> 0:00:42.080 with Okta, and it works on macOS, windows, Linux, and 0:00:42.080 --> 0:00:44.240 mobile devices. So if you have Okta and you're looking 0:00:44.240 --> 0:00:47.690 for a device trust solution that respects your team, visit collider.com. 0:00:47.960 --> 0:00:50.600 Unsupervised learning to watch a demo and see how it works. 0:00:50.600 --> 0:00:58.280 That's collider.com unsupervised learning. Welcome to Unsupervised Learning, a security, AI, 0:00:58.280 --> 0:01:00.830 and meaning focused podcast that looks at how best to 0:01:00.830 --> 0:01:04.459 thrive as humans in a post AI world. It combines 0:01:04.459 --> 0:01:08.300 original ideas, analysis, and mental models to bring not just 0:01:08.300 --> 0:01:11.209 the news, but why it matters and how to respond. 0:01:15.540 --> 0:01:18.330 All right. Welcome to unsupervised learning. This is Daniel Meisler. 0:01:18.360 --> 0:01:24.119 All right. 436. In about a bunch of AI stuff here. So, uh, 0:01:24.569 --> 0:01:27.780 big religious day for me on Monday. It was, uh, 0:01:27.780 --> 0:01:31.800 WWDC by Apple. And what I said after it was Apple, 0:01:31.800 --> 0:01:35.190 just one. I they caught up and passed everyone basically 0:01:35.190 --> 0:01:38.190 in one shot. And I think this is because the 0:01:38.190 --> 0:01:40.229 thing that they get that most people don't is it 0:01:40.230 --> 0:01:42.690 it's all about the platform. It's all about the integrations. 0:01:42.690 --> 0:01:45.809 It's all about the ecosystem. And they're essentially the only 0:01:45.810 --> 0:01:49.200 one who understands that and is also good at tech. 0:01:49.200 --> 0:01:52.380 And they also have this massive platform in the iPhone. Right. 0:01:52.380 --> 0:01:55.260 So it's just they're just winning. They're so far ahead 0:01:55.260 --> 0:01:57.960 of everyone in terms of long term thinking. Like it's 0:01:57.960 --> 0:02:00.660 not even close now. This one has since been cleared up, 0:02:00.660 --> 0:02:02.820 but at the time it wasn't clear what part of 0:02:02.820 --> 0:02:06.870 what they were doing was actually Apple stuff. And what 0:02:06.870 --> 0:02:09.420 part was OpenAI. I feel like they could have done 0:02:09.419 --> 0:02:11.820 a better job explaining that because I was paying attention, 0:02:11.820 --> 0:02:14.940 and I still kind of didn't get it. Looking back, 0:02:14.940 --> 0:02:18.030 it was pretty clear somehow I missed it. And I, 0:02:18.030 --> 0:02:20.190 I mean, I used to work there, I've worked on 0:02:20.190 --> 0:02:23.460 AI quite a bit and I like, I love AI 0:02:23.460 --> 0:02:25.620 and I love Apple, and I still didn't get it. 0:02:25.620 --> 0:02:27.510 So I feel like they could have been more clear. 0:02:27.510 --> 0:02:30.120 So the answer to this question is that most of 0:02:30.120 --> 0:02:33.060 what they showed was actually all Apple stuff. It was 0:02:33.060 --> 0:02:37.440 not pure OpenAI. And that that's a huge difference, right? 0:02:37.440 --> 0:02:40.980 So so essentially what the situation is, is, I would 0:02:40.980 --> 0:02:43.980 say probably 90% of what they showed. I don't know, 0:02:43.980 --> 0:02:46.170 you can figure it out for yourself. Like if you 0:02:46.169 --> 0:02:48.510 did the the math or whatever, it doesn't really matter. 0:02:48.510 --> 0:02:51.840 The point is like 90% roughly of everything they talked 0:02:51.840 --> 0:02:56.610 about on device, the cloud stuff, the whole new cloud infrastructure, 0:02:56.610 --> 0:03:00.870 which is absolutely insane. They basically built a new AI 0:03:00.900 --> 0:03:05.250 cloud infrastructure that's completely secure, where, you know, they don't 0:03:05.250 --> 0:03:08.160 even have access to the content that's on there. They 0:03:08.160 --> 0:03:11.220 don't give that access to any third parties. It's essentially 0:03:11.220 --> 0:03:15.270 like a cloud version of the local, very secure enclave. 0:03:15.270 --> 0:03:17.580 I'm not sure it's quite that level, but it's pretty 0:03:17.580 --> 0:03:20.670 much that level. It is better than anything that exists 0:03:20.669 --> 0:03:23.280 in the industry right now in terms of like secure 0:03:23.280 --> 0:03:27.180 cloud compute, and it's just extraordinary. And they generated it 0:03:27.180 --> 0:03:30.239 just so that they could do these AI features. So 0:03:30.240 --> 0:03:33.840 that is essentially their AI story is they're going to 0:03:33.840 --> 0:03:37.140 have deep context about you because they're Apple. And this 0:03:37.140 --> 0:03:40.050 is essentially life OS is what we're talking about. And 0:03:40.050 --> 0:03:43.620 so they have all that information about you and they 0:03:43.620 --> 0:03:46.290 have the ability to bring that to AI. And what 0:03:46.290 --> 0:03:48.090 they've done is they've made it so that if you 0:03:48.090 --> 0:03:50.610 could do it locally, it does it locally, and if 0:03:50.610 --> 0:03:52.470 you can't do it locally, it goes up to the 0:03:52.470 --> 0:03:56.820 cloud using AI models by Apple. So these are Apple 0:03:56.820 --> 0:04:00.660 models running on Apple silicon right. It's not even it's 0:04:00.660 --> 0:04:04.140 not even Nvidia stuff. They're using their own chips also 0:04:04.140 --> 0:04:07.350 in the cloud and obviously on device. So that's the 0:04:07.350 --> 0:04:12.480 whole ecosystem. Now that is for all the context data okay. 0:04:12.480 --> 0:04:16.590 That's for basically all of the personal data that you have, 0:04:16.589 --> 0:04:18.810 which they don't want to share with other people and 0:04:18.810 --> 0:04:21.210 they want to be very, very private. Now in the 0:04:21.210 --> 0:04:23.880 case of like looking things up like, oh, get the 0:04:23.880 --> 0:04:27.270 current weather and stuff like that. That's where they partnered 0:04:27.270 --> 0:04:31.919 with the new Siri, with the new with OpenAI. Right. 0:04:31.920 --> 0:04:34.529 So that's that's what they did. So essentially what they 0:04:34.529 --> 0:04:38.820 did was they said, okay for live searches, OpenAI is 0:04:38.820 --> 0:04:42.120 really good at that. You know, fact checking, stuff like that. 0:04:42.120 --> 0:04:44.700 Just just calling the internet. And maybe they're not great 0:04:44.700 --> 0:04:46.620 at fact checking, but you know what I mean? So 0:04:46.620 --> 0:04:51.420 they're leveraging third party AI for a very specific thing, 0:04:51.420 --> 0:04:54.539 but they're not like uploading all this stuff to OpenAI. 0:04:54.540 --> 0:04:57.599 It's like the exact opposite of what they're doing. They 0:04:57.600 --> 0:05:00.570 built this whole new infrastructure just to keep it all 0:05:00.570 --> 0:05:03.299 very secure and very private. So it's very much two 0:05:03.300 --> 0:05:06.030 different things. Apple AI, which is like 90% of what 0:05:06.029 --> 0:05:08.580 they talked about, which has the on prem and the 0:05:08.580 --> 0:05:12.060 cloud component. And then this other third party integration, which 0:05:12.060 --> 0:05:14.640 is with ChatGPT. And they even talked about yeah, they're 0:05:14.640 --> 0:05:17.670 going to do like they're looking at possibly doing Gemini. 0:05:17.670 --> 0:05:20.609 But the point is other third party services for those 0:05:20.610 --> 0:05:24.539 kind of integrations. But don't confuse it. That is not 0:05:24.570 --> 0:05:28.920 what they're doing with our personal data. Like so Live Bovary. 0:05:29.070 --> 0:05:31.290 A whole bunch of people were like, well, how do 0:05:31.290 --> 0:05:34.409 I find an android? Like, I can't believe they're sending 0:05:34.410 --> 0:05:36.510 all my data to OpenAI. And it's like, that is 0:05:36.510 --> 0:05:38.729 not what they said. It's not what they said. It's 0:05:38.730 --> 0:05:42.270 like the opposite of what they said. So, um, already 0:05:42.270 --> 0:05:46.950 running the demos or the betas for iOS 18, uh, 0:05:46.950 --> 0:05:50.490 remarkably stable, but largely because they didn't really change that 0:05:50.490 --> 0:05:54.060 much yet, and they definitely haven't added like the AI 0:05:54.089 --> 0:05:57.600 services yet. So like, Siri is still the same. That 0:05:57.600 --> 0:06:00.570 stuff is coming out like later this year, they said. 0:06:00.570 --> 0:06:03.029 So I feel like I didn't really get the AI 0:06:03.029 --> 0:06:06.300 features that I was looking for immediately in the beta, 0:06:06.300 --> 0:06:09.480 which I was hoping to get. But whatever it happens 0:06:09.480 --> 0:06:13.560 when it happens went and trained kickboxing, I was. Bad 0:06:13.560 --> 0:06:16.800 at it. It was quite sad. I was like five 0:06:16.800 --> 0:06:19.950 times worse than a beginner, I think, because it's hard 0:06:19.950 --> 0:06:23.070 for me to like, do things slowly and I'm like 0:06:23.070 --> 0:06:25.529 imagining things in my brain, but my body won't do 0:06:25.529 --> 0:06:28.350 that thing that's in my brain. I'm just very sort 0:06:28.350 --> 0:06:30.810 of confused. So I think I'm going to do some more, 0:06:30.810 --> 0:06:34.380 like learning the basics of the body coordination, like on 0:06:34.380 --> 0:06:37.890 a bag, and then do it with live people after that. 0:06:37.890 --> 0:06:41.039 That's my current plan. Plus I'm going to be doing 0:06:41.040 --> 0:06:43.320 a Gits as well. And uh, one of the most 0:06:43.320 --> 0:06:47.010 important conversations ever on AI safety. Absolutely. Got to watch 0:06:47.010 --> 0:06:51.960 this thing. It's, uh, Leopold Aschenbrenner and Dwarkesh Patel. I've 0:06:51.960 --> 0:06:54.630 talked about this a million times, but it is really, 0:06:54.630 --> 0:06:58.890 really good. Wrote a new fabric pattern called Capture Thinkers Work. 0:06:58.890 --> 0:07:03.150 So you basically send in, like Hayek or somebody into, 0:07:03.150 --> 0:07:07.170 you know, fabric SP capture thinkers work. And I mean, 0:07:07.170 --> 0:07:10.800 it's extraordinary. In fact, I can show you, okay, so 0:07:10.800 --> 0:07:15.180 if we do a capture thinkers work here, let's do this. 0:07:15.180 --> 0:07:18.030 Let's first let's blow it up some and then we'll 0:07:18.030 --> 0:07:22.740 do who's a philosopher? Hume. We'll send that to capture 0:07:22.740 --> 0:07:26.310 thinkers work. So comes up with a single line sentence 0:07:26.310 --> 0:07:30.900 up here. Most impactful ideas primary advice or teachings. The 0:07:30.900 --> 0:07:33.960 works their main works. So you can go build a 0:07:33.960 --> 0:07:37.830 reading list based on this top quotes. Something is human 0:07:37.830 --> 0:07:41.460 if it emphasizes the role of experienced custom and sentiment 0:07:41.460 --> 0:07:45.540 over abstract reason and innate ideas, and like advice of 0:07:45.540 --> 0:07:50.760 how to like implement human concepts into your life. So yeah, 0:07:50.760 --> 0:07:53.100 love this pattern. It's a great way of like thinking 0:07:53.100 --> 0:07:56.430 about someone's work. So for example, if you do like 0:07:56.460 --> 0:07:58.800 S or Perl, which it doesn't matter if you spell 0:07:58.800 --> 0:08:01.290 it wrong, for example, it looks like maybe I spelled 0:08:01.290 --> 0:08:03.930 it wrong because it corrected it to one R and 0:08:03.930 --> 0:08:07.290 one L, but yeah. Background where they were born, the 0:08:07.290 --> 0:08:11.580 school psychotherapy just really, really powerful. Got a thread here 0:08:11.580 --> 0:08:14.280 on how to fight a good mentor. I, I put 0:08:14.280 --> 0:08:16.230 this out a very long time ago, but I think 0:08:16.230 --> 0:08:18.239 it's worth sharing. I'm going to go ahead and open 0:08:18.240 --> 0:08:22.260 it up. So it's like yeah, being mentored by someone 0:08:22.260 --> 0:08:25.560 ahead of you can really change your life. Don't overuse flattery. 0:08:25.560 --> 0:08:30.600 Ask something very specific. Behave like a future peer like don't. 0:08:30.690 --> 0:08:34.980 Don't grovel. Basically, show your you've already done a bunch 0:08:34.980 --> 0:08:38.040 of work and, you know, don't ask basic questions that 0:08:38.040 --> 0:08:41.370 you could have got by reading their blog or their book. 0:08:41.370 --> 0:08:43.709 I mean, to some extent you can, but you don't 0:08:43.710 --> 0:08:46.380 want to overdo that and ask for an opinion on 0:08:46.380 --> 0:08:48.900 something you've created. So it's like, look, I've done work, 0:08:48.900 --> 0:08:52.979 I've created something. I'm curious about a very specific question 0:08:52.980 --> 0:08:55.320 about what you think about this thing you don't want 0:08:55.320 --> 0:08:58.469 to ask open ended things. Kind of a general rule 0:08:58.470 --> 0:09:01.980 here is like, don't give them work. You're you're being 0:09:01.980 --> 0:09:05.760 respectful of their time by asking a pretty small thing 0:09:05.760 --> 0:09:08.640 and sort of like presenting yourself as a peer and 0:09:08.640 --> 0:09:10.770 you're like, hey, look, I'm looking for feedback on this. 0:09:10.770 --> 0:09:12.630 I know you've done it a lot, and that's kind 0:09:12.630 --> 0:09:15.030 of the vibe. Or you can offer an improvement or 0:09:15.030 --> 0:09:19.800 construction a constructive like comment or feedback on something they've 0:09:19.800 --> 0:09:23.400 done and most importantly, like produce something really cool and 0:09:23.400 --> 0:09:26.189 show it to them. And then they're likely to just like, 0:09:26.190 --> 0:09:29.040 treat you like a future peer and interact with you. 0:09:29.040 --> 0:09:31.829 And that's the summary here. So and it got a 0:09:31.830 --> 0:09:35.670 blog post around it okay. Let's see. Yeah. Adding source 0:09:35.670 --> 0:09:38.460 names to stories now which we'll see here in a second. 0:09:38.460 --> 0:09:41.190 So basically you'll know what the source is before you 0:09:41.220 --> 0:09:44.820 click it. New member piece on the future of AI okay. 0:09:44.820 --> 0:09:46.980 Yeah I'm going to go into this one. This one's 0:09:46.980 --> 0:09:51.690 pretty major okay. Let's dive into this. All right. So 0:09:51.690 --> 0:09:57.210 first thing is like my definition of AGI and ASI. 0:09:57.240 --> 0:10:02.640 So my definition of AGI is can functionally replace an 0:10:02.640 --> 0:10:05.970 average white collar worker in the US making 80 K 0:10:05.970 --> 0:10:09.540 in 2023. And I say 2023 because it's going to 0:10:09.540 --> 0:10:12.570 get weird. This definition will get more weird if you 0:10:12.570 --> 0:10:15.719 don't lock it into a previous time before. Like AGI. 0:10:15.750 --> 0:10:19.500 We're basically saying like, look, a white collar worker making 0:10:19.500 --> 0:10:22.469 an average salary, right? Like 80 K in the US 0:10:22.470 --> 0:10:28.470 in 2023. And ASI, I'm defining as an AGI that's 0:10:28.470 --> 0:10:32.040 smarter and more capable than any human that's ever lived. 0:10:32.040 --> 0:10:34.980 And what what this is essentially saying is like, if 0:10:34.980 --> 0:10:37.560 you can functionally replace an average white collar worker in 0:10:37.559 --> 0:10:42.780 the US, what that is referring to is generality, because 0:10:42.780 --> 0:10:45.150 what I'm talking about is the ability for the boss 0:10:45.179 --> 0:10:47.610 to say, you know what, you're actually not on that 0:10:47.610 --> 0:10:51.690 project anymore. Go get with Julie and Lisa, and you're 0:10:51.690 --> 0:10:55.110 being completely retasked to this other project, which is kind 0:10:55.110 --> 0:10:57.840 of unrelated to that. And there is some issue of 0:10:57.840 --> 0:11:01.860 like specialization, but we're assuming it's still within the specialization 0:11:01.860 --> 0:11:05.189 of that person, and it's general enough where you could 0:11:05.190 --> 0:11:08.250 be like, look, it's a completely different task. It's a 0:11:08.250 --> 0:11:10.860 completely different project. It's got a different time frame, it's 0:11:10.860 --> 0:11:14.250 got different goals, it's got different metrics. And what a 0:11:14.250 --> 0:11:17.640 human does is basically say, okay, cool, I will erase 0:11:17.640 --> 0:11:20.970 what I was thinking before. I will take the new requirements, 0:11:20.970 --> 0:11:22.560 I will come up with a new plan and I 0:11:22.559 --> 0:11:24.270 will start working on it, which means you got to 0:11:24.270 --> 0:11:26.610 break it into tasks. You got to have a timeline, 0:11:26.640 --> 0:11:28.110 you got to have a schedule, you got to have 0:11:28.110 --> 0:11:30.720 all that. So in other words, it's not just narrowly 0:11:30.720 --> 0:11:35.310 executing small tasks, it's also the planning stage. It's also 0:11:35.309 --> 0:11:39.030 constant readjustment, constant evolution of like how do I best 0:11:39.030 --> 0:11:41.880 do this. And obviously there's different tiers of quality of 0:11:41.880 --> 0:11:44.760 people who can do that as well. But that's why 0:11:44.760 --> 0:11:49.410 I say average white collar worker making $80,000. So it's 0:11:49.410 --> 0:11:53.400 not like exceptional. And it's not like the worst. So 0:11:53.400 --> 0:11:57.000 so given that as a definition of AGI, this general 0:11:57.000 --> 0:12:01.020 intelligence that can like be retasked as well as do 0:12:01.020 --> 0:12:05.130 specific tasks underneath plus answer emails plus be, you know, 0:12:05.130 --> 0:12:09.059 communicative and, you know, basically function as a good employee. 0:12:09.059 --> 0:12:13.860 That's essentially AGI. So given that what who is able 0:12:13.860 --> 0:12:16.319 to do that and basically be a thinker and be 0:12:16.320 --> 0:12:20.429 a producer of new ideas, basically smarter and more capable 0:12:20.429 --> 0:12:22.800 than any human that's ever lived. And a good way 0:12:22.800 --> 0:12:26.400 of thinking about this is like and Aschenbrenner talked about 0:12:26.400 --> 0:12:29.850 this as well in the Türkische caste. So while using 0:12:29.850 --> 0:12:33.150 their first names, Leopold talked about that in the Dwarkesh 0:12:33.150 --> 0:12:37.170 podcast conversation. It's essentially like, what if you could just 0:12:37.170 --> 0:12:42.570 make like 10,000 Einsteins or John von Neumann, right? That 0:12:42.570 --> 0:12:45.810 would be insane. So just imagine John von Neumann here, 0:12:45.809 --> 0:12:48.989 or Einstein or whoever you want to put as like 0:12:48.990 --> 0:12:51.630 the smartest person. And just imagine you can have a 0:12:51.630 --> 0:12:54.390 whole bunch of them who could go and do these tasks, 0:12:54.390 --> 0:12:58.110 like AGI tasks like, you know, most importantly, come up 0:12:58.110 --> 0:13:00.720 with new science, come up with new innovations, come up 0:13:00.720 --> 0:13:04.050 with new medicines, but also, yeah, go do all this work, 0:13:04.050 --> 0:13:06.870 figure out how how to do it better figure out 0:13:06.870 --> 0:13:08.429 better ways to do all the work we have to 0:13:08.429 --> 0:13:10.830 do and then go do it. So that's ASI. And 0:13:10.830 --> 0:13:14.339 I'm saying essentially there's a big divide here between conscious 0:13:14.340 --> 0:13:17.580 and unconscious. And it actually doesn't matter all that much. 0:13:17.610 --> 0:13:20.580 A lot of people think that consciousness is on this 0:13:20.580 --> 0:13:22.890 path over here. So it's like, oh, we're going to 0:13:22.890 --> 0:13:26.310 get artificial narrow intelligence and then we're going to get AGI. 0:13:26.309 --> 0:13:28.440 And then after that it becomes conscious and then it 0:13:28.440 --> 0:13:31.950 becomes super intelligent or some combination thereof. But I think 0:13:31.950 --> 0:13:34.200 the better way to think about this is that we 0:13:34.200 --> 0:13:37.200 could stay completely over here and not even get consciousness, 0:13:37.200 --> 0:13:41.730 but still have artificial super intelligence, right? So one does 0:13:41.730 --> 0:13:45.300 not require the other. And yeah, I've got prediction for 0:13:45.630 --> 0:13:50.100 a date later on. We are now inside the post itself. Yeah. 0:13:50.100 --> 0:13:52.560 So how I see the paths of development and the 0:13:52.559 --> 0:13:54.959 thing that got me talking about this, or thinking about 0:13:54.960 --> 0:13:58.350 this a lot more, is just this whole Leopold and 0:13:58.350 --> 0:14:02.220 Dwarkesh thing. Yeah. So I talked about that path and 0:14:02.220 --> 0:14:06.600 that's the conversation. Yeah. Okay. So this is basically my 0:14:06.600 --> 0:14:09.420 argument of like how we're going to get to AGI 0:14:09.420 --> 0:14:14.429 and ASI. So I think this is a hypothesis right. 0:14:14.520 --> 0:14:17.190 It's it's an idea. It's a thought only so many 0:14:17.190 --> 0:14:20.640 components to understand the world. Some of these are easily 0:14:20.640 --> 0:14:23.850 attainable and some are like impossible. So an impossible one 0:14:23.850 --> 0:14:26.790 is like you can't calculate all the combinations of things 0:14:26.790 --> 0:14:31.770 like numbers get crazy very quickly with combinations. So unless 0:14:31.770 --> 0:14:35.280 you can brute force everything like like God or something, 0:14:35.280 --> 0:14:39.060 it's really hard. But I think there are really realistic 0:14:39.060 --> 0:14:44.489 and approachable abstractions of that complexity. So my intuition is 0:14:44.490 --> 0:14:47.430 that there are likely abstractions of those that we haven't 0:14:47.430 --> 0:14:51.090 yet attained. I think that's pretty obvious. And then an 0:14:51.090 --> 0:14:54.750 analogy I use here is like fluid dynamics representing all 0:14:54.750 --> 0:14:58.680 the molecules in a liquid, or Newtonian physics being an 0:14:58.680 --> 0:15:01.680 abstraction for like quantum or whatever the bottom layer is. 0:15:01.680 --> 0:15:05.100 So my friend's argument is I was walking with him 0:15:05.100 --> 0:15:08.190 recently and like, I think it's really smart. He basically 0:15:08.190 --> 0:15:11.310 said there's limitations to what we can do, even with 0:15:11.310 --> 0:15:14.700 super intelligence, because of the one way nature of calculation. 0:15:14.700 --> 0:15:18.060 So basically P and P, you could try lots of things, 0:15:18.060 --> 0:15:20.130 but it's a one way function. You can see if 0:15:20.130 --> 0:15:23.430 it works, but you can't like immediately see all the options. 0:15:23.430 --> 0:15:26.640 And that's kind of like P and P. So he 0:15:26.640 --> 0:15:29.190 came up with a great definition of intelligence, which I've 0:15:29.190 --> 0:15:32.730 since seen like other people talking about as well, which 0:15:32.730 --> 0:15:35.880 is like the ability to quickly reduce a search space 0:15:35.880 --> 0:15:39.060 for a problem, like like pruning trees. So there's like 0:15:39.060 --> 0:15:42.510 100 trillion options. What can we do to get that 0:15:42.510 --> 0:15:45.990 down to like 10 or 100 or 1000 or whatever 0:15:45.990 --> 0:15:49.050 as fast as possible. And the faster you could do that, 0:15:49.050 --> 0:15:51.150 the more you could do that, the more intelligent you are. 0:15:51.150 --> 0:15:54.330 And actually reminds me a lot of quantum computing for 0:15:54.330 --> 0:15:57.120 the first time. I've never seen like an intersection there, 0:15:57.120 --> 0:16:00.990 but imagine that you could test multiple options all at once, 0:16:00.990 --> 0:16:03.960 which is kind of the whole thing with quantum anyway, 0:16:03.960 --> 0:16:07.530 so here are the components I see leading to superintelligence. 0:16:07.530 --> 0:16:10.500 One is a super complex model of the world, like 0:16:10.500 --> 0:16:14.480 how the cell works, how medicines. Interact with cells, you know, 0:16:14.480 --> 0:16:19.400 molecular interaction with different body parts. So it's essentially like 0:16:19.400 --> 0:16:23.870 testing drugs, but doing it with models instead of like 0:16:23.870 --> 0:16:27.590 having to actually try it in labs and people and mice. 0:16:27.590 --> 0:16:31.010 So think of that at like all levels of physics. 0:16:31.010 --> 0:16:35.600 So like quantum atomic molecules, cells and bigger and bigger, 0:16:35.600 --> 0:16:40.010 eventually you're at like psychology, sociology, whatever. And in some 0:16:40.010 --> 0:16:43.700 cases it's just feeding our current understanding as text. But 0:16:43.700 --> 0:16:47.060 even better, if you like, have actual recordings of these 0:16:47.060 --> 0:16:51.710 things happening, similar to how Tesla taught its new full Self-Driving. 0:16:51.710 --> 0:16:55.040 It's just like, yeah, this is here. Here are the 0:16:55.040 --> 0:16:58.220 actual interactions. Learn from that. So the use case I 0:16:58.220 --> 0:17:02.510 kept using for this was solving human aging. So let's 0:17:02.510 --> 0:17:04.939 say it's like the telomeres or whatever the length of 0:17:04.940 --> 0:17:07.430 the telomeres. And there's a whole bunch of different possible 0:17:07.430 --> 0:17:10.490 mechanisms of how aging could be happening, and therefore how 0:17:10.490 --> 0:17:12.320 we might be able to turn it off or reverse 0:17:12.320 --> 0:17:15.320 it or whatever. But the most important thing is like 0:17:15.320 --> 0:17:18.320 have a deep understanding of how everything works, so you 0:17:18.320 --> 0:17:20.690 know where you could actually start to go and look. 0:17:20.690 --> 0:17:23.330 So the next big thing for this, in my opinion, 0:17:23.330 --> 0:17:27.740 is the patterns that analogies and metaphors. So links between things. 0:17:27.740 --> 0:17:30.679 And we're already seeing like models get really good at that. 0:17:30.680 --> 0:17:32.719 And I think the next one is the size of 0:17:32.720 --> 0:17:36.680 the working memory. Right. So not exactly sure what this means, 0:17:36.680 --> 0:17:39.800 but it's like it's it's like your Ram or your 0:17:39.800 --> 0:17:43.040 functional memory. It's like, how much of this map can 0:17:43.040 --> 0:17:45.439 you fit in your brain at once without having to 0:17:45.440 --> 0:17:47.300 slide it over? And parts of the map fall off 0:17:47.300 --> 0:17:49.520 the table and you can't see them. So maybe that's 0:17:49.520 --> 0:17:52.220 a combination of multiple tiers. So you have like you 0:17:52.220 --> 0:17:55.160 have like Ram, then you have like L1, L2 inside 0:17:55.160 --> 0:17:57.980 of the CPU, which is really the brain, not really 0:17:57.980 --> 0:18:00.709 memory but kind of all the same. Right. And then 0:18:00.710 --> 0:18:03.770 you've got disk which is much slower, which has to 0:18:03.770 --> 0:18:07.400 go into memory to get loaded up. And when there's 0:18:07.400 --> 0:18:10.580 too much on disk and not enough Ram, like I said, 0:18:10.580 --> 0:18:13.490 you got to swap. Right. So these are the types 0:18:13.490 --> 0:18:16.370 of things that I think the overall size of that 0:18:16.369 --> 0:18:21.290 memory that can do the patterns based on the world 0:18:21.290 --> 0:18:24.800 model is essentially what gets us there. That's my theory. 0:18:24.800 --> 0:18:28.250 And essentially this theory is that with scale inside of 0:18:28.250 --> 0:18:32.300 neural networks, those three things just keep getting better and better, 0:18:32.300 --> 0:18:35.840 plus some post-processing which we'll talk about, and a whole 0:18:35.840 --> 0:18:38.000 bunch of hacks and tricks that are basically just going 0:18:38.000 --> 0:18:40.369 to come with the industry, because my friend pointed out 0:18:40.369 --> 0:18:44.239 that just the size of the model enough alone is 0:18:44.240 --> 0:18:47.270 not going to be enough, and that Post-training is super critical. 0:18:47.270 --> 0:18:50.420 And once I got what he was saying, like, we 0:18:50.420 --> 0:18:52.790 definitely agree on that now, it was just something that 0:18:52.790 --> 0:18:55.520 I wasn't aware of and that got us kind of 0:18:55.520 --> 0:18:58.010 in line. So this is my way of capturing this. 0:18:58.010 --> 0:19:01.100 So my hypothesis is that there are very few fundamental 0:19:01.100 --> 0:19:04.520 components that allow AI to scale up from narrow AI 0:19:04.640 --> 0:19:08.720 to AGI to ASI. So it's a world model, sufficient 0:19:08.720 --> 0:19:12.680 training examples to allow deep enough ability to find patterns 0:19:12.680 --> 0:19:17.600 and similarities between phenomenon within that world model, sufficient working memory, 0:19:17.600 --> 0:19:21.530 and then ultimately the ability to like, model the scientific 0:19:21.530 --> 0:19:26.450 method and test things. Ideally not having to actually do 0:19:26.450 --> 0:19:31.430 that test, but being able to test things in an abstraction. Right. 0:19:31.430 --> 0:19:35.000 And I'm not sure how smart or how good that is, 0:19:35.000 --> 0:19:38.869 because if you're not testing, you're not really testing. I 0:19:38.869 --> 0:19:41.690 think this might be possible. And yeah, like I said, 0:19:41.690 --> 0:19:45.770 I'm least confident that number four is required or that 0:19:45.770 --> 0:19:49.129 it's possible. However, I would say that these three are 0:19:49.130 --> 0:19:53.030 amazing and possibly enough. And this one, if it were possible, 0:19:53.030 --> 0:19:55.640 to whatever degree, it would massively enhance the other three. 0:19:55.640 --> 0:19:58.490 And I think these are the main components. I mean, 0:19:58.490 --> 0:20:02.150 I've spent relatively little time thinking about this. I could 0:20:02.150 --> 0:20:04.430 revise it in the future. The other thing to say 0:20:04.430 --> 0:20:08.210 about this is that number four, which is the testing piece, 0:20:08.210 --> 0:20:13.550 might just be implied somehow if number one is sufficiently advanced, 0:20:13.550 --> 0:20:17.090 if the world model is good enough, maybe we start 0:20:17.090 --> 0:20:20.090 getting to this, I don't know. And maybe it requires 0:20:20.090 --> 0:20:24.080 the patterns and similarities. And obviously I think a lot 0:20:24.080 --> 0:20:26.690 of this requires the memory. So I guess what I'm 0:20:26.690 --> 0:20:29.270 saying is the whole thing might be two big things 0:20:29.270 --> 0:20:31.909 the complexity of the world model and the size of 0:20:31.910 --> 0:20:34.669 the active memory that it's able to use. And maybe 0:20:34.670 --> 0:20:37.910 the patterns and the testing stuff might come along with 0:20:37.910 --> 0:20:42.050 those two. And I guess the most controversial thing is that, 0:20:42.050 --> 0:20:45.920 and this kind of undergirds everything I'm saying is the 0:20:45.920 --> 0:20:50.270 hypothesis that the universe has like an actual complexity that 0:20:50.270 --> 0:20:54.560 is approachable or finite, and that models only need to 0:20:54.560 --> 0:21:00.410 cross a certain threshold of depth of understanding or quality 0:21:00.410 --> 0:21:05.330 of abstractions or depth of abstractions or height of abstractions 0:21:05.330 --> 0:21:10.459 to become very similar to like a Laplacian demon in 0:21:10.460 --> 0:21:13.810 the sense of like full knowledge. So I don't think 0:21:13.810 --> 0:21:16.869 they can have full knowledge. I think that's God level 0:21:16.869 --> 0:21:21.189 stuff that's supernatural. At least according to our current understanding 0:21:21.190 --> 0:21:24.760 and my belief. But that if you hit a certain 0:21:24.760 --> 0:21:29.830 threshold or depth of abstraction quality, it's functionally very similar. 0:21:29.830 --> 0:21:34.900 That is what I'm basing my intuition, my hypothesis of 0:21:34.900 --> 0:21:39.490 AGI and ASI on. I got another concept called human 3.0. 0:21:39.490 --> 0:21:43.060 So it's essentially about how many people believe that they 0:21:43.060 --> 0:21:45.370 have ideas that are useful to the world. How many 0:21:45.369 --> 0:21:48.280 people believe they could write a short book, percentage of 0:21:48.280 --> 0:21:51.609 a person's creativity that they're actively using in their life, 0:21:51.609 --> 0:21:55.419 percentage of a person's total capabilities that they actually broadcast 0:21:55.420 --> 0:21:59.200 the world as some sort of value, and the degree 0:21:59.200 --> 0:22:02.710 to which someone lives as their full spectrum self. So 0:22:02.710 --> 0:22:07.869 in human 2.0, we're essentially living and working for capitalism. 0:22:07.869 --> 0:22:10.780 Capitalism is like the most important thing. So we're kind 0:22:10.780 --> 0:22:14.950 of like 10%. We're putting like 10% of our value 0:22:14.950 --> 0:22:17.980 in what we can offer to the world, I would 0:22:17.980 --> 0:22:20.560 argue in our CV because it's like, oh, I'm really 0:22:20.560 --> 0:22:23.830 good with Microsoft Excel, and I can send emails or whatever. 0:22:23.830 --> 0:22:26.380 And it's like, is that really you? I don't think so. 0:22:26.380 --> 0:22:29.950 The other aspect is that human 2.0, I would say 0:22:29.950 --> 0:22:34.570 current model is essentially most people believing that they're not special, 0:22:34.570 --> 0:22:39.219 most people believing that only special people write books, write blogs, 0:22:39.220 --> 0:22:44.380 give presentations, do things like that, and regular people don't. 0:22:44.380 --> 0:22:49.060 So two main things bifurcated lives broken into personal and professional. 0:22:49.060 --> 0:22:52.960 This is 2.0, and the percentage of people who think 0:22:52.960 --> 0:22:54.939 that they have something to offer the world, which I 0:22:54.940 --> 0:22:59.109 would argue is extremely low and human 3.0 is the 0:22:59.109 --> 0:23:02.710 transition to a world where people understand that everyone has 0:23:02.710 --> 0:23:04.900 something to offer to the world, and they live as 0:23:04.900 --> 0:23:09.280 full spectrum people. So it's like living for other humans, 0:23:09.400 --> 0:23:14.530 living as humans, for humans, as opposed to living as 0:23:14.530 --> 0:23:19.540 humans who like it's like living to work instead of 0:23:19.540 --> 0:23:22.810 working to live and wouldn't even be working to live. 0:23:22.810 --> 0:23:25.270 Human 3.0 would even be more advanced than that. It 0:23:25.270 --> 0:23:27.609 would be. So it means like your public profile is 0:23:27.609 --> 0:23:30.850 everything about you, right? How caring you are, how smart 0:23:30.850 --> 0:23:34.300 you are, how funny you are, your favorite things in life, 0:23:34.300 --> 0:23:37.420 your best ideas, the projects you like to work on, 0:23:37.420 --> 0:23:40.180 the problems you think are important and meaningful in the world, 0:23:40.180 --> 0:23:44.380 plus your technical skills, of course. And they know that 0:23:44.380 --> 0:23:48.070 they're valuable because they are all these things, not just 0:23:48.070 --> 0:23:51.250 what they can do for a corporation. And so my 0:23:51.250 --> 0:23:54.070 thoughts on this are like, because you have to start wondering, 0:23:54.070 --> 0:23:57.700 when you listen to the Leopold and Aakash conversation, it's like, 0:23:57.700 --> 0:24:01.420 is human 3.0 in danger? Are we in danger of 0:24:01.420 --> 0:24:05.890 losing a world where humans exist and do what they 0:24:05.890 --> 0:24:08.860 do for other humans? Because AC is going to take 0:24:08.859 --> 0:24:12.940 over and whatever turn us into whatever. Maybe, hopefully? Well, 0:24:12.940 --> 0:24:16.810 I'm pushing for a possible AGI or a that allows 0:24:16.810 --> 0:24:20.410 us to do human 3.0 either. It would be kind 0:24:20.410 --> 0:24:22.780 of scary if it just took over and implemented it. 0:24:22.780 --> 0:24:25.359 That would be scary. Hopefully it would arrive at something 0:24:25.359 --> 0:24:29.200 like human 3.0 and keep us alive and thriving. Or 0:24:29.200 --> 0:24:31.690 it doesn't take over. It's still used as a tool, 0:24:31.690 --> 0:24:35.260 but it's in the hands of someone somewhat benign like 0:24:35.260 --> 0:24:40.540 the US, hopefully remains benign. And at that point we 0:24:40.570 --> 0:24:44.920 we basically give it the criteria of help us maintain 0:24:44.920 --> 0:24:47.859 a society in which we move towards and maintain something 0:24:47.859 --> 0:24:51.010 like human 3.0. Another thing I'm really worried about is 0:24:51.010 --> 0:24:54.910 like how immigration will affect UBI. So if the government 0:24:54.910 --> 0:24:58.570 is basically the jobs are gone, no, very few people 0:24:58.570 --> 0:25:01.660 can do something that I can't. And that's like, whatever 10% 0:25:01.660 --> 0:25:04.270 of the world or 1% of the world, or 25% 0:25:04.270 --> 0:25:08.230 of the world, whatever. Different times, different numbers, different ways 0:25:08.230 --> 0:25:11.710 of measuring. But let's say it's 10% and the other 90% 0:25:11.710 --> 0:25:15.670 just aren't employable because, like Harari said, you know, they're 0:25:15.670 --> 0:25:18.070 a useless class. They have nothing that they can offer 0:25:18.070 --> 0:25:20.979 the economy. And this is human 2.0. Still, this is 0:25:20.980 --> 0:25:25.030 not crossed over into this hopeful land of human 3.0. 0:25:25.030 --> 0:25:28.180 In the meantime, I mean, because some people, like Leopold 0:25:28.180 --> 0:25:31.570 are worried about ASI happening in like a number of 0:25:31.570 --> 0:25:36.520 years and just like completely redoing society and the government's 0:25:36.520 --> 0:25:38.320 obviously going to have to be involved there. They're going 0:25:38.320 --> 0:25:42.040 to have to pay people to basically not riot, not 0:25:42.040 --> 0:25:48.100 kill themselves and, you know, provide education, provide gaming, provide entertainment, 0:25:48.100 --> 0:25:52.210 provide something some sort of basis for meaning that's going 0:25:52.210 --> 0:25:55.270 to have to be included. But what does that look like? 0:25:55.270 --> 0:25:58.960 I mean, there's like a welfare state and there's like 0:25:58.960 --> 0:26:01.750 the the government's giving out too much money, and then 0:26:01.750 --> 0:26:04.540 there's a whole nother level when it's actually UBI, because 0:26:04.540 --> 0:26:07.630 no one's even able to work. Right. It's not that 0:26:07.630 --> 0:26:11.170 they can't find a job when when people are actually hiring, 0:26:11.170 --> 0:26:13.780 it's like, no. Nobody's hiring. All the work's being done 0:26:13.780 --> 0:26:18.490 by AI or whatever, 90%, and AI is producing trillions 0:26:18.490 --> 0:26:22.060 of dollars in value. And that raises another question. Okay, 0:26:22.060 --> 0:26:24.730 you're making all this new stuff. You have all these 0:26:24.730 --> 0:26:29.020 new creators, you have all this new AI generated productivity 0:26:29.140 --> 0:26:34.020 and products and services. For who? Who's buying the stuff? 0:26:34.050 --> 0:26:36.929 Who's buying the stuff? If the top 10% are the 0:26:36.930 --> 0:26:40.320 only ones who can make things and sell things. Who's 0:26:40.320 --> 0:26:44.010 buying the stuff? And the natural answer looks like, which 0:26:44.010 --> 0:26:49.140 is quite gross. It's like, I guess the government taxes 0:26:49.140 --> 0:26:52.439 all the income that's being made. Again, where's the income 0:26:52.440 --> 0:26:55.470 coming from? It's almost like there's an exchange between the 0:26:55.470 --> 0:26:59.220 government and the producers, and the producers are making all 0:26:59.220 --> 0:27:01.679 this awesome stuff. The government is using all the awesome 0:27:01.680 --> 0:27:04.770 stuff to generate all these awesome things. But then the 0:27:04.770 --> 0:27:07.590 government has to take this money, which is very strange 0:27:07.590 --> 0:27:09.990 word for it now, because it's not being generated by 0:27:09.990 --> 0:27:13.320 the people and being taxed by the government. In fact, 0:27:13.320 --> 0:27:16.860 it's like it's just a money idea. The money idea 0:27:16.859 --> 0:27:21.389 goes to the government. The government then gives the population money, 0:27:21.390 --> 0:27:24.179 the ability to purchase these goods and services, which are 0:27:24.180 --> 0:27:27.210 being generated by the small percentage of the population to 0:27:27.210 --> 0:27:30.870 the tune of whatever trillions of trillions upon trillions of dollars. 0:27:30.869 --> 0:27:33.150 And then they give that money out to the people 0:27:33.150 --> 0:27:35.700 and the people that can then use that to buy 0:27:35.700 --> 0:27:38.399 the services that are being made by this 10%. And 0:27:38.400 --> 0:27:41.970 that feels very not quite Ponzi, but like shell game, 0:27:41.970 --> 0:27:44.940 it's like, oh, it's fake money. We're moving it around. 0:27:44.940 --> 0:27:48.149 It just feels very strange when when the, the human 0:27:48.150 --> 0:27:53.700 population is not producing not participating in that creation. Now 0:27:53.700 --> 0:27:57.510 in human 3.0, I think they will be I'm not 0:27:57.510 --> 0:27:59.970 sure exactly what that economy will look like. I'm sure 0:27:59.970 --> 0:28:03.090 much smarter people can help me think about that. It's 0:28:03.090 --> 0:28:07.080 essentially more of a human to human exchange barter system. 0:28:07.080 --> 0:28:10.470 Obviously a currency like a maybe like a doctor, a 0:28:10.590 --> 0:28:13.740 coffee type of currency or something. But you can also 0:28:13.740 --> 0:28:16.710 buy regular goods and services, but you could also pay 0:28:16.710 --> 0:28:20.190 people for making you happy or whatever. Like a lot 0:28:20.190 --> 0:28:22.200 of people will be creating art and doing lots of 0:28:22.200 --> 0:28:25.169 stuff like that. So I'm hoping there's a large economy 0:28:25.170 --> 0:28:28.080 for sharing on this full spectrum that we talked about 0:28:28.080 --> 0:28:31.650 with human 3.0, and that will involve money and stuff 0:28:31.650 --> 0:28:34.740 like that. But what I'm talking about here is really 0:28:34.740 --> 0:28:38.040 human 2.0 stuff. This is like this year, next year, 0:28:38.040 --> 0:28:40.530 the next five years, the next ten years, you know, 0:28:40.530 --> 0:28:42.360 what do we do in the meantime with all these 0:28:42.360 --> 0:28:45.240 people who are going to be needing money? I mean, 0:28:45.240 --> 0:28:49.200 what happens with an economy with millions of immigrants who 0:28:49.200 --> 0:28:54.780 are undocumented, who have families, they need to feed them, 0:28:54.780 --> 0:28:57.120 they need to house them. And what if the government 0:28:57.120 --> 0:28:59.790 is like, well, what if the jobs go away? Well, 0:28:59.790 --> 0:29:02.430 maybe they leave. Maybe they don't leave, though, and they're 0:29:02.430 --> 0:29:05.340 very angry because they can no longer get work, and 0:29:05.340 --> 0:29:09.180 they're also not being paid like the local citizens. I 0:29:09.180 --> 0:29:12.270 don't know what the solution here is, but it's I'm 0:29:12.270 --> 0:29:14.130 saying we need to think about it. So my thoughts 0:29:14.130 --> 0:29:18.810 on conscious AI are, I think, fairly simple actually. I 0:29:18.810 --> 0:29:21.960 think the way humans got consciousness is that it was 0:29:21.960 --> 0:29:25.979 adaptive for helping us accelerate evolution. So basically winning and 0:29:25.980 --> 0:29:29.790 losing is what powers, evolution, blame and praise are smaller 0:29:29.790 --> 0:29:33.330 versions of winning and losing plants and insects win and 0:29:33.330 --> 0:29:37.770 lose as well, evolutionarily, but it seems like only humans 0:29:37.770 --> 0:29:40.380 and a few other species. And maybe it's a lot, 0:29:40.380 --> 0:29:44.820 but it's a subset. Very small subset, have subjective experiences, 0:29:44.820 --> 0:29:47.760 and it's not clear that anyone else other than humans 0:29:47.760 --> 0:29:50.970 has like blame and praise. Well, maybe dogs do. I 0:29:50.970 --> 0:29:55.020 feel like they understand, blame and praise. Good boy, bad girl, right? 0:29:55.020 --> 0:29:57.870 I feel like that works based on their facial expressions. 0:29:57.870 --> 0:30:00.960 Who knows if that's right. But anyway, bottom line is 0:30:00.960 --> 0:30:05.940 blame and praise might empower evolution if we experience winning 0:30:05.940 --> 0:30:10.020 and losing, being blamed and praised and feeling that at 0:30:10.020 --> 0:30:14.970 a visceral level versus having no inter like immediate repercussions 0:30:14.970 --> 0:30:17.790 of not doing well. So maybe evolution still gets you 0:30:17.790 --> 0:30:22.380 there if you don't have blame and praise and subjective experience. 0:30:22.380 --> 0:30:26.340 But I think this whole kind of container is how 0:30:26.340 --> 0:30:29.430 we ended up being conscious, because it enables this, which 0:30:29.430 --> 0:30:33.300 makes evolution better. That that's my guess. If that's right, 0:30:33.300 --> 0:30:37.440 then consciousness might have simply emerged from evolution as our 0:30:37.440 --> 0:30:41.040 brains got bigger and we became better at getting better, 0:30:41.040 --> 0:30:47.670 which means it could emerge naturally through like self similar loops, 0:30:47.670 --> 0:30:51.030 definitely some RL related stuff. Or we could kind of 0:30:51.030 --> 0:30:53.340 just like hack the system and like add it on 0:30:53.340 --> 0:30:56.610 and be like, hey, develop this so that you can 0:30:56.610 --> 0:30:59.280 have this advantage or whatever. We can try to shortcut 0:30:59.280 --> 0:31:02.250 what evolution did over millions of years. Yeah. And then 0:31:02.250 --> 0:31:05.490 one question is like, well, if we hacked it that way, 0:31:05.490 --> 0:31:08.250 would it be real consciousness? And me as a materialist, 0:31:08.250 --> 0:31:11.670 I'm like, well, what is real consciousness like? I think 0:31:11.670 --> 0:31:15.330 at some level it's all an illusion anyway. But practically, 0:31:15.330 --> 0:31:19.980 if you feel yourself feeling and it hurts, that's real, right? 0:31:19.980 --> 0:31:24.420 It doesn't matter how mechanistic the substrate below actually is, 0:31:24.420 --> 0:31:27.570 because at some point you can pick along the way 0:31:27.570 --> 0:31:30.780 and just be like, this is mechanical, this is mechanical. This. 0:31:30.920 --> 0:31:35.360 A mechanical. Therefore it's all mechanical. Doesn't matter. Friendship still 0:31:35.360 --> 0:31:38.630 is awesome. Love is awesome. Ice cream is awesome. Doesn't 0:31:38.630 --> 0:31:41.360 make it any less awesome than it's made out of atoms, right? 0:31:41.360 --> 0:31:45.800 So my AC prediction I'm ferm on my AGI prediction 0:31:45.800 --> 0:31:51.890 of 2025 and 2028. Somewhere in there, 2026 2027 is 0:31:51.890 --> 0:31:54.590 like ideal, the sweet spot for me. I'm going to 0:31:54.590 --> 0:32:01.100 give soft prediction, soft prediction of AC 2027 to 31. 0:32:01.100 --> 0:32:04.760 So what is that. That's three four plus two. So 0:32:04.760 --> 0:32:08.210 that's roughly 2029. Don't feel great about it. That's like 0:32:08.210 --> 0:32:12.440 a 30 to 40% prediction like confidence level. This one 0:32:12.440 --> 0:32:17.810 I'm saying 90% by 2028. End of 2028 I'm saying 0:32:17.810 --> 0:32:21.080 like 60% by end of 2025. So I feel a 0:32:21.080 --> 0:32:24.200 lot more confident about that one. But who knows. Predictions 0:32:24.200 --> 0:32:27.080 are hard, especially when they're about the future. Niels Bohr 0:32:27.080 --> 0:32:30.560 said that. All right, unifying thoughts. Leopold, you want to 0:32:30.560 --> 0:32:35.390 watch him on AI safety? Completely. Absolutely impressive guy. And 0:32:35.390 --> 0:32:37.190 you want to go read his essays that are released 0:32:37.190 --> 0:32:40.520 on this as well. Talked to Aakash Patel is becoming 0:32:40.520 --> 0:32:43.910 one of my favorite podcasters ever right now. He is 0:32:43.910 --> 0:32:49.430 basically an autodidact. He's a crazy reader. When Dwarkesh talks 0:32:49.430 --> 0:32:53.420 to Cowan or Cowan, that's how he pronounces it Tyler Cowan. 0:32:53.420 --> 0:32:57.410 When they talk, they jump around, they're like, oh yeah, 0:32:57.410 --> 0:32:59.360 you know, I was I was knitting this thing the 0:32:59.360 --> 0:33:01.820 other day and I was working. I was, you know, 0:33:01.820 --> 0:33:06.230 making wood figurines on my lathe in the back backyard. 0:33:06.230 --> 0:33:09.230 And I was thinking about, you know, whatever, Roman history. 0:33:09.230 --> 0:33:12.260 And then I was thinking about, you know, Adam's economics. 0:33:12.260 --> 0:33:15.470 And that reminds me of this one poem by Rilke 0:33:15.470 --> 0:33:18.020 or whatever. And they just jump from here to there 0:33:18.020 --> 0:33:20.090 and they're like, oh, that reminds me of this. Hey, 0:33:20.090 --> 0:33:22.490 what about World War Two? And this thing happened and 0:33:22.490 --> 0:33:24.860 they're just jumping from place to place and like, I 0:33:24.860 --> 0:33:26.810 feel like I'm pretty good at that. I feel like 0:33:26.810 --> 0:33:31.070 that's one of my superpowers. But Cowan, seeing Cowan do this, 0:33:31.070 --> 0:33:35.030 it is insane. And Dwarkesh is like a mini Cowan. 0:33:35.030 --> 0:33:38.210 He's like coming up so fast. It's the reason that 0:33:38.210 --> 0:33:42.740 his interviews are so good. He's, like, fiercely curious and 0:33:42.740 --> 0:33:46.880 an autodidact, reading, like, whatever, 12 books a second, whatever 0:33:46.880 --> 0:33:51.980 his current metric is, it's insane. Anyway. Follow them. It's really, 0:33:51.980 --> 0:33:56.720 really good stuff. And human 3.0. My best optimism here 0:33:56.720 --> 0:33:59.840 is we get AGI, but we don't get AC for 0:33:59.840 --> 0:34:02.180 a while, which will give us a chance to hopefully 0:34:02.180 --> 0:34:05.810 move towards 3.0 or we get AC, but it's like 0:34:05.810 --> 0:34:09.320 controlled by the US and or benign US actors and 0:34:09.320 --> 0:34:14.720 we move towards human 3.0 or AC comes online, it's 0:34:14.719 --> 0:34:17.960 benign and it kind of takes over and it builds 0:34:17.960 --> 0:34:21.260 human 3.0 because that's the best future for humanity or 0:34:21.260 --> 0:34:23.900 something that looks similar. Right. And this is what I 0:34:23.900 --> 0:34:27.020 was saying before. I don't actually care how unlikely these 0:34:27.020 --> 0:34:31.610 scenarios are, because it's kind of the only good path 0:34:31.610 --> 0:34:34.220 that I could see happening from all this AI stuff 0:34:34.219 --> 0:34:39.330 is we eventually get to human 3.0. A human world 0:34:39.330 --> 0:34:42.930 for humans, and I would say future versions of humans, 0:34:42.930 --> 0:34:46.350 which might be hybrids with AI or whatever. But you 0:34:46.350 --> 0:34:49.109 got to maintain the humanity. We got to be doing 0:34:49.110 --> 0:34:52.440 what we're doing for humans, not for the sake of tech. 0:34:52.440 --> 0:34:55.770 So I'm optimistic about that. I am trying to help 0:34:55.770 --> 0:34:57.989 build that. That's my whole purpose in, like doing all 0:34:57.989 --> 0:35:01.680 of this so that that's what I'm shooting for. And 0:35:01.680 --> 0:35:04.050 I don't care if it's a 20% chance of us 0:35:04.050 --> 0:35:06.120 getting there. First of all, nobody can know what this 0:35:06.120 --> 0:35:09.239 number is, so it doesn't really matter. But second of all, 0:35:09.239 --> 0:35:12.060 if it's a 10% chance, guess what? I'm shooting for 0:35:12.060 --> 0:35:14.280 the 10%. I'm trying to make it 11. Try to 0:35:14.280 --> 0:35:17.040 make it a 50, whatever. Try to make it a 90. 0:35:17.070 --> 0:35:19.740 If it's a 60% chance and it's a 40% chance, 0:35:19.739 --> 0:35:21.690 we're all going to die. Whatever. I'm not going to 0:35:21.690 --> 0:35:25.049 sit around thinking about the 40%. I mean, that's my vibe. 0:35:25.050 --> 0:35:29.160 I am shooting for human 3.0. And if someone's like, yeah, well, 0:35:29.160 --> 0:35:31.200 you don't know, China can take it over and blah 0:35:31.200 --> 0:35:33.570 blah blah, AC and it can kill us all and 0:35:33.570 --> 0:35:38.190 whatever paper clips and I just don't see the value 0:35:38.190 --> 0:35:42.630 of going hard down the path of like, the doomers 0:35:42.630 --> 0:35:46.590 because let's say they're right. Their answer is turn it off, 0:35:46.590 --> 0:35:50.760 don't build it. That's not going to happen. Next answer. 0:35:50.760 --> 0:35:52.890 They're like, I don't have a next answer. We should 0:35:52.890 --> 0:35:55.050 stop talking about this. We should stop building it. You 0:35:55.050 --> 0:35:57.510 have exited yourself from the adult table. You are no 0:35:57.510 --> 0:36:01.799 longer in the conversation because of Moloch. We are building this. 0:36:01.800 --> 0:36:05.640 It is happening. We are running full speed with scissors 0:36:05.640 --> 0:36:08.520 in both hands. And there's nothing we can do to 0:36:08.520 --> 0:36:11.550 stop that. My argument is, the only thing we could 0:36:11.550 --> 0:36:14.280 do is try to aim it in a good direction. 0:36:14.280 --> 0:36:17.400 Try to point towards human 3.0 and that's what I'm 0:36:17.400 --> 0:36:22.319 trying to do. All right, huge diversion, but worth talking about. 0:36:22.320 --> 0:36:25.530 That was the AI piece collection of thoughts and predictions 0:36:25.530 --> 0:36:29.669 about AI. June of 2024. All right. Security got a 0:36:29.670 --> 0:36:35.070 whole bunch of misinformation sneaking into like, little man stuffed news. 0:36:35.070 --> 0:36:38.460 So this is like manosphere, like podcasts or whatever, or 0:36:38.460 --> 0:36:44.790 news sites or whatever. They're actually republishing Russian propaganda network content. 0:36:44.790 --> 0:36:47.490 So it's like, yeah, it can seep in and go 0:36:47.489 --> 0:36:49.710 lots of different places. This is why you actually need 0:36:49.739 --> 0:36:53.580 AI to be watching everything. And basically when a new 0:36:53.580 --> 0:36:59.069 information or disinformation or misinformation propaganda campaign comes out, when 0:36:59.070 --> 0:37:02.940 it comes out, we should categorize, okay, what the core 0:37:02.940 --> 0:37:06.630 concept is, the core idea. It's trying to spread with AI, 0:37:06.630 --> 0:37:09.900 and then you monitor all the different stories coming out. 0:37:09.900 --> 0:37:14.580 It might be like Saskatchewan news from South Parish, you know, 0:37:14.580 --> 0:37:19.080 Wisconsin or whatever. And they're spreading the same story, because 0:37:19.080 --> 0:37:21.480 what the attackers will do is they'll figure out, okay, 0:37:21.480 --> 0:37:24.000 where can we get it in? Where can we push 0:37:24.000 --> 0:37:26.550 this narrative? And maybe, maybe the front door is closed, 0:37:26.550 --> 0:37:29.730 but maybe there's a thousand different backdoors. That was that one. 0:37:29.730 --> 0:37:32.610 TikTok had an issue with DMs. There was a zero 0:37:32.640 --> 0:37:34.980 day in DMs where you click it and open it. 0:37:34.980 --> 0:37:39.090 It's pretty bad. The current tactic for what? You basically 0:37:39.090 --> 0:37:43.170 get compromised for compromise of your TikTok. If you did that, 0:37:43.170 --> 0:37:47.100 and a new tactic TikTok is using to avoid the 0:37:47.100 --> 0:37:50.219 lawsuit is basically to say they're going to build out 0:37:50.219 --> 0:37:53.310 a separate algorithm and separate data just for us. So 0:37:53.310 --> 0:37:55.830 we'll see if that works. New paper claims that GPT 0:37:55.830 --> 0:38:00.480 four can autonomously hack zero day vulnerabilities. 53% success rate 0:38:00.510 --> 0:38:01.920 going to be a lot of these papers. A lot 0:38:01.920 --> 0:38:04.380 of these papers are not very good. A lot of 0:38:04.380 --> 0:38:07.440 them are decent. I will keep reading and posting these 0:38:07.440 --> 0:38:10.770 when the claims are interesting and they're from decent outfits, 0:38:10.770 --> 0:38:13.680 but you want to watch the quality of the paper 0:38:13.680 --> 0:38:17.580 very closely. Look, especially for how much of their testing 0:38:17.580 --> 0:38:20.250 methodology that they publish. Did they show exactly what they 0:38:20.250 --> 0:38:22.380 tried against these different things? Did they show all the 0:38:22.380 --> 0:38:25.890 different examples? Did they, you know, give you all the 0:38:25.890 --> 0:38:29.220 stuff to do the replication? If not, then they're kind 0:38:29.219 --> 0:38:31.410 of like hiding the ball and you can't really trust 0:38:31.410 --> 0:38:35.550 their score. And we got to analyze paper fabric pattern 0:38:35.550 --> 0:38:37.200 that you can use for this. And when I ran 0:38:37.200 --> 0:38:39.840 it against this one it got a six on rigor 0:38:39.840 --> 0:38:44.969 benchmark and Ablations look good but some details missing and 0:38:44.969 --> 0:38:48.750 stats lacking. That's why that's why we have fabric. Okay 0:38:48.750 --> 0:38:53.339 Chinese drone photographer got snagged by the US Espionage Act 0:38:53.340 --> 0:38:58.410 taking pics of military shipyards. So there's another thing that's 0:38:58.410 --> 0:39:00.690 being talked about right now with China buying up a 0:39:00.690 --> 0:39:04.410 whole bunch of land right next to military bases, they're 0:39:04.410 --> 0:39:07.680 buying like massive amounts of farmland, and they start setting 0:39:07.680 --> 0:39:09.629 up all this gear. It's like, oh, don't pay any 0:39:09.630 --> 0:39:12.600 attention to us. We're just building some stuff over here. 0:39:12.600 --> 0:39:16.350 So I hope Biden or Trump or whoever just like, 0:39:16.350 --> 0:39:20.009 finds a list of these. And just like search and seizure, 0:39:20.040 --> 0:39:23.279 good bye. Do not pass go. And maybe some of 0:39:23.280 --> 0:39:27.270 those are benign, right. But I have to basically be like, well, 0:39:27.270 --> 0:39:33.210 based on the, you know, 9566, you know, fucking billion 0:39:33.210 --> 0:39:37.210 times that you've actually tried to steal from us. And 0:39:37.210 --> 0:39:40.779 actually hacked us, then you know you're going to lose 0:39:40.780 --> 0:39:43.180 some for the home team when you're actually doing something 0:39:43.180 --> 0:39:46.239 benign and actually just setting up a farm next to 0:39:46.239 --> 0:39:49.239 a military base on accident. And they didn't actually mean 0:39:49.239 --> 0:39:52.480 anything by that. Too bad you ruined your reputation. Therefore 0:39:52.480 --> 0:39:55.480 get out. Okay. OpenAI revealed that its AI tech was 0:39:55.480 --> 0:39:59.290 used by Russia and China for sneaky influence ops. Yeah, 0:39:59.290 --> 0:40:02.109 they've said this a few times and I like it. 0:40:02.110 --> 0:40:05.890 I like the transparency. They're basically like, look, this keeps happening. 0:40:05.890 --> 0:40:07.960 We keep catching it and we're going to try to 0:40:07.960 --> 0:40:13.210 do better. All right. WWDC keynote got an analysis here. 0:40:13.210 --> 0:40:16.660 This is a fabric analysis from a TechCrunch article. As 0:40:16.660 --> 0:40:19.450 I talked about before, Apple basically killed the AI thing. 0:40:19.450 --> 0:40:21.460 They also did a bunch of really cool stuff. One 0:40:21.460 --> 0:40:23.170 of the things is you bring your phone next to 0:40:23.170 --> 0:40:26.140 someone else, you can transfer them cash. I've been waiting 0:40:26.140 --> 0:40:29.020 that for that forever. Problem is, a lot of people 0:40:29.020 --> 0:40:32.470 won't have that feature turned on, or they'll have an 0:40:32.500 --> 0:40:35.110 Android and you won't be able to do it like so. 0:40:35.110 --> 0:40:38.080 A lot of people are switching to like Zelle or Venmo. 0:40:38.110 --> 0:40:42.400 I don't like Venmo, and Zelle is slow. Whatever. I 0:40:42.400 --> 0:40:45.430 wish more people had the ability to do these swipey 0:40:45.430 --> 0:40:49.330 things and Apple things, but kind of rude to require 0:40:49.330 --> 0:40:52.149 that everyone has an iPhone, right? I mean, that's just 0:40:52.150 --> 0:40:54.700 not realistic and also not cool. All right. All the 0:40:54.700 --> 0:41:00.190 different startups that Apple killed today always fun. After WWDC 0:41:00.190 --> 0:41:05.620 or an OpenAI event, US needs 225,000 more cybersecurity workers. 0:41:05.620 --> 0:41:09.489 I never trust numbers like this because it's hard to know. Like, 0:41:09.489 --> 0:41:12.610 what are they asking? Are they asking it right? What 0:41:12.610 --> 0:41:15.190 are the questions that they accept? Like, I just feel 0:41:15.190 --> 0:41:17.470 like the numbers are bad. I feel like that number 0:41:17.469 --> 0:41:19.720 is going to go down because of AI, and I 0:41:19.719 --> 0:41:21.790 feel like it's going to go up because of AI. 0:41:21.820 --> 0:41:23.860 So it's going to go down because of AI, because 0:41:23.860 --> 0:41:26.320 the jobs are going to get easier to do without people, 0:41:26.320 --> 0:41:29.859 but so much more stuff is going to get made 0:41:29.860 --> 0:41:33.340 that we're just going to be injecting insecurity into the world, 0:41:33.340 --> 0:41:36.010 especially when it's like getting made with AI. So it's 0:41:36.010 --> 0:41:38.439 like AI going in, building a bunch of stuff and 0:41:38.440 --> 0:41:40.750 have that stuff's going to be really nasty in terms 0:41:40.750 --> 0:41:43.810 of security vulnerabilities. So it's like still going to need 0:41:43.810 --> 0:41:48.489 people for a while. Like assuming all this AI replacement 0:41:48.489 --> 0:41:51.759 stuff goes really, really fast. We're still talking about years 0:41:51.760 --> 0:41:55.719 and years. I mean, super fast for like an 80% 0:41:55.719 --> 0:41:59.770 replacement of all jobs, super fast. And these are crazy 0:41:59.770 --> 0:42:03.370 numbers I'm just throwing out. But but like Nuclear fast 0:42:03.370 --> 0:42:06.969 is like ten years now as far as starting and 0:42:06.969 --> 0:42:10.089 actually having major impact, I think it's already starting and 0:42:10.090 --> 0:42:12.820 it's already like a year or two from now we're 0:42:12.820 --> 0:42:14.800 going to start feeling it. So I think that's going 0:42:14.800 --> 0:42:18.100 to be really fast. There's going to be a regulatory response. 0:42:18.100 --> 0:42:21.790 It's slow to make any change in any organization, and 0:42:21.790 --> 0:42:24.550 the bigger and more bureaucratic it is, the slower it 0:42:24.550 --> 0:42:27.940 will be. So just because something could happen in a 0:42:27.940 --> 0:42:31.000 year or six months doesn't mean it will. And it's 0:42:31.000 --> 0:42:33.730 more likely to happen in 2 or 3 or 4 0:42:33.730 --> 0:42:37.030 or 10 years. Got to keep that in mind. It's 0:42:37.030 --> 0:42:39.400 going to go very fast. It's going to go faster 0:42:39.400 --> 0:42:41.830 than you think and slower than you think both at 0:42:41.830 --> 0:42:45.160 the same time. Apple just hit the three. $3 trillion 0:42:45.160 --> 0:42:48.399 mark again. But Nvidia jumped ahead. And now Nvidia is 0:42:48.400 --> 0:42:52.750 even more valuable than Apple. And Microsoft is still number 0:42:52.750 --> 0:42:57.520 one insane. Such as killing it over there. Cartwheel tool 0:42:57.550 --> 0:43:02.320 olama CLE wire JS for us for UI design shade 0:43:02.320 --> 0:43:06.040 map project that turns every mountain building and tree shadow 0:43:06.040 --> 0:43:08.800 for any date and time. How to think like a 0:43:08.800 --> 0:43:13.510 computer scientist. Pretty cool. Uh, interactive edition here. Starship just 0:43:13.510 --> 0:43:16.810 landed its first one. Like all the other ones either 0:43:16.810 --> 0:43:19.300 blew up on the way out or, uh, blew up 0:43:19.300 --> 0:43:23.080 on the way in, but he finally landed one very impressive, 0:43:23.080 --> 0:43:29.830 I think 5000 tons. So that's what, £20,000? Five times two, £2,000. 0:43:29.830 --> 0:43:33.730 2000 times 5000. Holy crap, that's a lot of pounds. 0:43:33.730 --> 0:43:38.290 World's largest solar farm just went online. Three point gigawatt 0:43:38.290 --> 0:43:42.219 Marty just went live in China. Okay, here's my question. 0:43:42.219 --> 0:43:45.969 Why can't we build like ten of these in the US? Okay, 0:43:45.969 --> 0:43:48.880 we know we need all this power for AI. So 0:43:48.880 --> 0:43:52.510 I say we go Biden. Trump does this. You basically 0:43:52.510 --> 0:43:54.430 go to each state and be like, look, I need 0:43:54.430 --> 0:43:59.860 you to build 3.5GW in your state within five years. 0:43:59.860 --> 0:44:02.080 And they're like, yeah, that's impossible. We'd have to hire 0:44:02.080 --> 0:44:04.930 all these people. And the government's like, whatever. I mean, 0:44:04.930 --> 0:44:07.839 we have money printers, so we're just going to print 0:44:07.840 --> 0:44:10.900 all this money, we're going to inject it into the economy. 0:44:10.900 --> 0:44:15.100 Most importantly, giving it to these training programs will, which 0:44:15.100 --> 0:44:17.140 will train the people how to build the wind farms 0:44:17.140 --> 0:44:20.739 and the solar farms, whatever the best renewable energy is 0:44:20.739 --> 0:44:24.520 for that particular state, like Nevada should build like 100 0:44:24.520 --> 0:44:27.430 of these things for for sun. I mean, I fly 0:44:27.430 --> 0:44:30.129 over it all the time. There's nothing there. In fact, 0:44:30.130 --> 0:44:33.070 you fly over the entire United States. There's nothing there. 0:44:33.070 --> 0:44:36.010 It's all empty. And meanwhile the sun just hits it 0:44:36.010 --> 0:44:39.920 every day. Like nine times out of ten. For every 0:44:39.920 --> 0:44:42.319 day of the year the sun comes up. In fact, 0:44:42.320 --> 0:44:44.480 I think it might be higher than 90% when the 0:44:44.480 --> 0:44:47.120 sun comes up. In fact, it might be 100%. The 0:44:47.120 --> 0:44:49.910 sun always comes up and it is hitting us with 0:44:49.910 --> 0:44:53.630 like so much energy. And we just it soaks in, 0:44:53.630 --> 0:44:57.890 you know, goes off into the atmosphere. Capture it. Okay, look, 0:44:57.890 --> 0:45:00.140 we've got people who don't have jobs and don't have 0:45:00.140 --> 0:45:02.510 meaning in their lives. We know that AI is going 0:45:02.510 --> 0:45:05.359 to require all this power. So millions of people have 0:45:05.360 --> 0:45:07.310 no meaning in their lives because they don't have work. 0:45:07.310 --> 0:45:10.790 We talk about the loss of manufacturing jobs. Go build 0:45:10.790 --> 0:45:13.940 all these solar panels, go build all these solar plants, 0:45:13.940 --> 0:45:16.700 have the people working there. Eventually they'll be replaced by 0:45:16.700 --> 0:45:19.460 robots in AI. That's fine. That'll take time like we 0:45:19.460 --> 0:45:23.210 talked about before. In the meantime, let's go build 100 0:45:23.210 --> 0:45:29.270 of these, whatever, 3.5GW facilities or ten gigawatt facilities, whatever. 0:45:29.270 --> 0:45:34.040 Giant solar farms, giant wind farms. And we put millions 0:45:34.040 --> 0:45:37.069 of people to work, which gives them meaning, and they're 0:45:37.070 --> 0:45:39.859 building towards a renewable future. So, like, even if I 0:45:39.860 --> 0:45:42.740 didn't happen whatever, we got renewable energy. Now we can 0:45:42.739 --> 0:45:44.960 export that energy, we can sell it other places. We 0:45:44.960 --> 0:45:47.810 just have it for whatever purpose. And you have to 0:45:47.810 --> 0:45:49.670 build the pipelines as well, right? You got to get 0:45:49.670 --> 0:45:51.890 that energy to where it needs to go. So that's 0:45:51.890 --> 0:45:56.360 like pipes, cables, like, you know, wires, all that stuff, 0:45:56.360 --> 0:46:00.020 all the infrastructure for moving energy. Where's the downside here? 0:46:00.020 --> 0:46:04.190 It's win win jobs and and tons of energy and 0:46:04.190 --> 0:46:07.460 no reliance on external energy. And we get to move 0:46:07.460 --> 0:46:10.969 on AI faster. So really bothers me when I see 0:46:10.969 --> 0:46:15.080 China understanding this and building so quickly. And we're like, man, 0:46:15.110 --> 0:46:18.350 let's fight with each other instead. Oh, in nuclear, nuclear 0:46:18.350 --> 0:46:21.590 is the other one. Okay, so USA's although I prefer solar, 0:46:21.590 --> 0:46:24.770 but someone knows better than me. USA's solar panel manufacturing 0:46:24.770 --> 0:46:28.610 capacity jumped by 71%. I like it, I like it. 0:46:28.610 --> 0:46:31.820 Florida and Texas are leading. Let's go. What about Nevada? 0:46:31.820 --> 0:46:34.580 Where are you at? Florida. Forget about it. It rains 0:46:34.580 --> 0:46:36.320 all the time and it's sinking. It's not going to 0:46:36.320 --> 0:46:39.530 work out. I guess Florida's energy generation would have to 0:46:39.530 --> 0:46:42.680 be based on, like being flooded or winds. No, you 0:46:42.680 --> 0:46:45.469 can't do can't do wind farms and hurricanes. That would 0:46:45.469 --> 0:46:48.500 be a bit rough. All right. Humans. David Brooks argues 0:46:48.500 --> 0:46:51.110 that progressive energy has shifted from the working class to 0:46:51.110 --> 0:46:55.340 elite universities, making them more left leaning than ever. Points 0:46:55.340 --> 0:46:59.239 to protests and progressive opinions. Yeah, it's all bad. Highly 0:46:59.239 --> 0:47:02.779 recommend you read David Brooks stuff The Road to Character. 0:47:02.780 --> 0:47:08.000 My favorite one is the Second Mountain. Absolutely unbelievable book. Unbelievable. 0:47:08.000 --> 0:47:10.129 And I love his stuff. The New York Times, he's 0:47:10.130 --> 0:47:14.540 like a centrist kind of conservative, but like, yeah, logical. 0:47:14.540 --> 0:47:16.700 I like him. Okay. This one turned out to be 0:47:16.700 --> 0:47:19.790 not great. A couple people. I looked at it again, 0:47:19.790 --> 0:47:22.700 people don't like me linking to Substack. I'm going to 0:47:22.700 --> 0:47:25.759 link to Substack. The question isn't is it Substack or 0:47:25.760 --> 0:47:29.360 is it whatever, New York Times, because all these different 0:47:29.360 --> 0:47:32.210 sources can be bad. What I care about is like 0:47:32.210 --> 0:47:36.050 factually being wrong or like making claims and not backing 0:47:36.050 --> 0:47:38.810 them up in some sort of in any sort of way, 0:47:38.810 --> 0:47:42.980 or just being ideologue and just like being hateful or 0:47:42.980 --> 0:47:46.100 rambling or whatever, like, well, I'm rambling, so not quite 0:47:46.100 --> 0:47:49.130 the same. But the point is, the best people in 0:47:49.130 --> 0:47:53.060 the world at thinking and being rigorous and having good 0:47:53.060 --> 0:47:57.950 ideas are going to Substack, okay, and beehive and all 0:47:57.950 --> 0:48:02.150 these other platforms and YouTube and TikTok. Even so, it's 0:48:02.150 --> 0:48:04.640 not the source that I'm linking to. That's the problem. 0:48:04.640 --> 0:48:07.820 The quality of the content is the problem because before 0:48:07.820 --> 0:48:11.300 too long, CNBC and like Fox, first of all, Fox News, 0:48:11.300 --> 0:48:13.520 I don't link to that usually. I'm sure they have 0:48:13.520 --> 0:48:16.610 good news sometimes, but you get the point. It's not 0:48:16.610 --> 0:48:20.719 about the type of source, okay? Blogs are just as 0:48:20.719 --> 0:48:25.549 authoritative as YouTube as a lot of the media networks 0:48:25.550 --> 0:48:28.040 at this point. It's more about the individual and the 0:48:28.040 --> 0:48:31.910 actual article. I am making these links available so that 0:48:31.910 --> 0:48:34.910 someone can my link numbers or my click numbers are 0:48:34.910 --> 0:48:38.750 probably going to go down because they will be like, well, 0:48:38.750 --> 0:48:40.850 I don't want to read from this guy, or I 0:48:40.850 --> 0:48:43.250 don't want to read from Hacker News or Financial Times. 0:48:43.250 --> 0:48:45.950 I hate that place or whatever. Hopefully you don't need 0:48:45.950 --> 0:48:48.620 it anyway. Hopefully this one sentence is good enough. That's 0:48:48.620 --> 0:48:50.540 the whole point of doing the summary. All right. New 0:48:50.540 --> 0:48:54.080 York Times deep dive into devastation of Ukraine. Oh, this 0:48:54.080 --> 0:48:58.009 story was absolutely nuts. And yeah, New York Times good stuff. 0:48:58.010 --> 0:49:00.740 They're staying around at a company more than two years 0:49:00.739 --> 0:49:05.300 could massively hurt your earning potential. US job openings hit 0:49:05.330 --> 0:49:10.070 a three year low. Layoffs are surprisingly low, it says CNN. 0:49:10.070 --> 0:49:14.540 See above for commentary on sources. Viagra improves blood flow 0:49:14.540 --> 0:49:19.520 and could help to prevent dementia. University of Oxford 49% 0:49:19.520 --> 0:49:23.180 of independents think Trump should drop out because of his 0:49:23.180 --> 0:49:26.480 guilty verdict. I'm making a prediction. I think this effect 0:49:26.480 --> 0:49:29.180 wears off and Trump will be pulling even with Biden 0:49:29.180 --> 0:49:31.040 within a month. You know what I'm going to do? 0:49:31.040 --> 0:49:36.910 Let's go look at 538 538 polls. Wow. Look at 0:49:36.910 --> 0:49:41.529 the impact. Look at the impact of the guilty verdict. Yeah, 0:49:41.530 --> 0:49:45.640 it happened right in here somewhere. Oh, look. No impact. 0:49:45.640 --> 0:49:48.549 And maybe I'm reading this wrong. Maybe this doesn't account 0:49:48.550 --> 0:49:52.460 for independence. I mean, this is this is a June 0:49:52.460 --> 0:49:57.489 12th 10th through 12th result. Echelon insights. I don't know 0:49:57.489 --> 0:50:01.839 how big this thing is. Okay, YouGov I know that's respectable. 0:50:01.840 --> 0:50:06.670 Biden 40, Trump 42. Okay. We got two different echelon 0:50:06.670 --> 0:50:10.930 ones daily cause what do we got? 4545 tide. So 0:50:10.930 --> 0:50:15.250 so I read this article and it's like, yeah, 49% 0:50:15.250 --> 0:50:18.310 of independents think Trump should drop out after his guilty verdict. 0:50:18.310 --> 0:50:22.480 Are these polls okay? That's one argument. Independents don't respond 0:50:22.480 --> 0:50:25.569 to polls. And supposedly these independents are going to switch 0:50:25.570 --> 0:50:28.060 the thing and they're going to be mad at Trump, 0:50:28.060 --> 0:50:30.730 and it's going to go for Biden. I don't buy it. 0:50:30.940 --> 0:50:33.700 If anything, in my opinion, this is just my opinion. 0:50:33.700 --> 0:50:37.600 This is politics. Like, who knows anything, right? In my opinion, 0:50:37.600 --> 0:50:41.500 these numbers are soft for Trump. I think Trump is 0:50:41.500 --> 0:50:44.380 probably winning by another 5 or 10 points right now. 0:50:44.380 --> 0:50:48.730 That's my guess. And I don't think him being convicted 0:50:48.730 --> 0:50:53.650 is going to do anything for his numbers if if not, 0:50:53.650 --> 0:50:56.920 make it better for him. Just ridiculous. This is why 0:50:56.920 --> 0:50:59.170 I think about I. I literally do not look at 0:50:59.170 --> 0:51:03.520 news anymore other other than like I brings me political 0:51:03.520 --> 0:51:06.009 news and analyses for me and then I have to 0:51:06.010 --> 0:51:08.200 go like read the story, I have to do my 0:51:08.200 --> 0:51:10.990 analysis or whatever. But like I do not pay attention 0:51:10.989 --> 0:51:13.629 to this stuff. I'm trying to build human 3.0 I 0:51:13.630 --> 0:51:17.080 don't like. I'm just so disillusioned with everything. I'm just 0:51:17.080 --> 0:51:20.380 like checked out. All right, X is doing adult content. 0:51:20.380 --> 0:51:24.009 Good for them. Ernest Hemingway wrote a deeply moving letter 0:51:24.010 --> 0:51:27.460 to friends who lost their son, saying, those who truly 0:51:27.460 --> 0:51:34.750 live never really die. I love Maria Popova. The marginalien. Yeah, 0:51:34.750 --> 0:51:36.640 I liked the old name that she had for that 0:51:36.640 --> 0:51:40.390 site before. I can't remember what it was. The yellow site. Anyway, 0:51:40.390 --> 0:51:42.880 one of the coolest life wisdom reads I've come across 0:51:42.880 --> 0:51:44.680 in a while. This is a really cool one. I'll 0:51:44.680 --> 0:51:47.319 put this one up. You can never expect honesty from 0:51:47.320 --> 0:51:50.230 people who even lie to themselves. You'll be ten times 0:51:50.230 --> 0:51:52.600 happier if you forgive your parents and stop blaming them 0:51:52.600 --> 0:51:56.230 for your misery for yourself. From society's advice, most of 0:51:56.230 --> 0:51:58.510 them have no idea of what they're doing. If you 0:51:58.510 --> 0:52:01.660 continue to wait for the right time, you'll waste your 0:52:01.660 --> 0:52:04.540 entire life and nothing will ever happen. It's a massive 0:52:04.540 --> 0:52:07.870 list of these. They're quite, quite good. You'll lose 99% 0:52:07.870 --> 0:52:10.779 of your close friends when you start to improve your life. 0:52:10.780 --> 0:52:14.020 That's bleak. I wouldn't go with 99%, but I get 0:52:14.020 --> 0:52:17.920 the point. You don't need 100 self-help books. All you 0:52:17.920 --> 0:52:21.280 need is action and self-discipline. It's a great list. It's 0:52:21.280 --> 0:52:23.710 a great list. You should check it out. Ideas and 0:52:23.710 --> 0:52:27.940 analysis I think I figured out the simplest possible answer 0:52:27.940 --> 0:52:31.660 for why Trump is still doing so well. Oh my god, 0:52:31.660 --> 0:52:35.230 I didn't include the link. Oh man, I didn't include 0:52:35.230 --> 0:52:38.169 the link in here. It's like, oh really? You did great. 0:52:38.170 --> 0:52:40.870 Thanks for not sharing it. And recommendation for the week 0:52:40.870 --> 0:52:44.650 is Tyler Cowen and Aakash Patel. As I talked about 0:52:44.650 --> 0:52:47.109 in the aphorism of the week, live as if you 0:52:47.110 --> 0:52:49.810 were to die tomorrow. Learn as if you were to 0:52:49.810 --> 0:52:52.960 live forever. Live as if you were to die tomorrow. 0:52:52.960 --> 0:52:56.589 Learn as if you were to live forever. Mahatma Gandhi. 0:52:57.719 --> 0:53:00.839 Unsupervised Learning is produced and edited by Daniel Meisler on 0:53:00.840 --> 0:53:05.460 a Neumann U87 AI microphone using Hindenburg. Intro and outro 0:53:05.460 --> 0:53:08.790 music is by zombie with the why and to get 0:53:08.790 --> 0:53:10.859 the text and links from this episode, sign up for 0:53:10.860 --> 0:53:16.509 the newsletter version of the show at Daniel meisler.com/newsletter. We'll 0:53:16.510 --> 0:53:17.350 see you next time.