1 00:00:21,183 --> 00:00:24,393 S1: All right. Welcome to unsupervised learning. This is Daniel Miessler. 2 00:00:24,423 --> 00:00:27,723 S1: All right, a few notes to start off. All better 3 00:00:27,723 --> 00:00:30,723 S1: from being sick. It was actually quite minor. Definitely the 4 00:00:30,723 --> 00:00:35,013 S1: most minor time I've been sick in recent history. Would 5 00:00:35,013 --> 00:00:37,203 S1: not even have known that I was sick if it 6 00:00:37,203 --> 00:00:41,373 S1: wasn't for testing. So yeah, good. Good that that's over. 7 00:00:41,373 --> 00:00:45,903 S1: We migrated fabric to go. So this is really major, 8 00:00:45,903 --> 00:00:49,923 S1: massive shout out to Jonathan Dunn who helped massively with this. 9 00:00:49,953 --> 00:00:54,363 S1: And project still looks largely the same. But it's all 10 00:00:54,393 --> 00:00:58,143 S1: go now. And if you check this out look at 11 00:00:58,143 --> 00:01:01,533 S1: the installation. Now this is just insane. Watch this. You 12 00:01:01,533 --> 00:01:05,043 S1: just run this, that's all. You run, okay? That's how 13 00:01:05,043 --> 00:01:08,073 S1: you install it and it just installs it. It just works. 14 00:01:08,073 --> 00:01:12,363 S1: So no more Python, no more Python dependency. Hell, you 15 00:01:12,363 --> 00:01:14,763 S1: just run this and you run setup and it works. 16 00:01:14,793 --> 00:01:19,803 S1: Now you might have to add some environmental variables or 17 00:01:19,833 --> 00:01:22,863 S1: your go install might do it for you. It depends 18 00:01:22,863 --> 00:01:25,683 S1: how you install go, but check this out. If you're 19 00:01:25,683 --> 00:01:29,043 S1: going to migrate, you just do pip uninstall fabric and 20 00:01:29,043 --> 00:01:31,143 S1: that will get rid of the Python version. And then 21 00:01:31,143 --> 00:01:35,073 S1: you want to check your bash RC and your zshrc 22 00:01:35,073 --> 00:01:37,923 S1: and basically clean those out. If you have any aliases 23 00:01:37,923 --> 00:01:39,963 S1: or anything like that in there. And then you do 24 00:01:39,963 --> 00:01:41,913 S1: this which is the same one as above, and you 25 00:01:41,913 --> 00:01:45,153 S1: do the setup. And again you could do the environmental variables. 26 00:01:45,153 --> 00:01:49,293 S1: But watch this. Let's say you haven't messed with it 27 00:01:49,293 --> 00:01:52,413 S1: in a while, which would be blasphemy. But let's say 28 00:01:52,413 --> 00:01:54,033 S1: you haven't messed with it in a while and you 29 00:01:54,063 --> 00:01:56,403 S1: come back, you went on vacation. Whatever, you come back 30 00:01:56,433 --> 00:01:58,923 S1: three weeks later and you want to install this thing 31 00:01:58,923 --> 00:02:00,933 S1: and you want to get the latest version, how do 32 00:02:00,933 --> 00:02:04,713 S1: you upgrade? You just run the go install again. That's 33 00:02:04,713 --> 00:02:06,993 S1: all you do. You just run the go install again 34 00:02:06,993 --> 00:02:12,393 S1: and you're completely updated and upgraded. And yeah, we kind 35 00:02:12,393 --> 00:02:16,323 S1: of kind of simplified up the the Readme here. It 36 00:02:16,323 --> 00:02:20,343 S1: should be pretty clean. And yeah, just really, really happy 37 00:02:20,343 --> 00:02:23,623 S1: with this. and it is so much faster. Yeah, it 38 00:02:23,623 --> 00:02:27,913 S1: is so much faster now. Yeah. I don't know what, um, 39 00:02:27,943 --> 00:02:32,533 S1: I could show you here. All right. Yeah. That's basically, uh, 40 00:02:32,533 --> 00:02:35,473 S1: getting some aphorisms. So I think I might have used 41 00:02:35,473 --> 00:02:37,333 S1: one of these. Yeah, I used one of these for 42 00:02:37,333 --> 00:02:41,113 S1: the newsletter this week. Oh, actually. And I'm recording the 43 00:02:41,113 --> 00:02:44,173 S1: newsletter right now. Of course. So we'll actually see this. 44 00:02:44,173 --> 00:02:47,443 S1: This is the command that I used to actually come 45 00:02:47,443 --> 00:02:50,743 S1: up with some ideas for aphorisms. So I just echo 46 00:02:50,743 --> 00:02:54,103 S1: Red Queen concept from Ridley's book. So I knew I 47 00:02:54,103 --> 00:02:58,633 S1: wanted aphorisms around that. And um, look at this. I 48 00:02:58,633 --> 00:03:02,113 S1: get back some really, really cool different aphorisms that I 49 00:03:02,113 --> 00:03:05,803 S1: could potentially use, and depending on the model, I was 50 00:03:05,803 --> 00:03:09,463 S1: using one model and it only gave me Matt Ridley quotes. 51 00:03:09,463 --> 00:03:12,133 S1: So that was pretty cool. So really depends on the 52 00:03:12,133 --> 00:03:15,013 S1: model that you use. But I basically use fabric all 53 00:03:15,013 --> 00:03:18,613 S1: the time and the go version is just unbelievably good. 54 00:03:18,613 --> 00:03:22,213 S1: So that's that. um, highly recommend you check out this 55 00:03:22,213 --> 00:03:27,223 S1: conversation between Peter Thiel and Joe Rogan. Um, Joe's an 56 00:03:27,223 --> 00:03:31,003 S1: interesting character. He he had Katt Williams on, and I 57 00:03:31,033 --> 00:03:33,493 S1: watched as much as I could of that episode, and 58 00:03:33,493 --> 00:03:38,173 S1: it was just like these crazy conspiracy theories. And Katt 59 00:03:38,173 --> 00:03:41,443 S1: Williams was quite serious, and Joe was kind of pretending 60 00:03:41,473 --> 00:03:44,113 S1: to be serious. But you also don't know with Joe, like, 61 00:03:44,113 --> 00:03:46,213 S1: what he believes and what he doesn't believe sometimes when 62 00:03:46,213 --> 00:03:48,523 S1: he's going off on those things. But then you see 63 00:03:48,553 --> 00:03:52,423 S1: him with someone like Peter Thiel and he's way more balanced. 64 00:03:52,453 --> 00:03:54,343 S1: He's like, he's not coming up with a bunch of 65 00:03:54,343 --> 00:03:58,123 S1: crazy stuff. He's just having a normal conversation with someone 66 00:03:58,123 --> 00:04:01,423 S1: he knows is smart. And here's what's crazy. I actually 67 00:04:01,423 --> 00:04:05,443 S1: thought his his description of the future of humanity was 68 00:04:05,443 --> 00:04:08,743 S1: better than Peter Thiel's. Now, maybe just me and Joe 69 00:04:08,773 --> 00:04:13,063 S1: are wrong, but Thiel seemed unwilling to sort of go 70 00:04:13,063 --> 00:04:16,453 S1: in the direction that the Joe was taking it, which 71 00:04:16,453 --> 00:04:19,033 S1: happens to be the exact same direction that I always 72 00:04:19,033 --> 00:04:21,493 S1: take it in, which is like this is inevitable. We 73 00:04:21,493 --> 00:04:24,613 S1: have to merge with them. Basically, it's like AI is 74 00:04:24,613 --> 00:04:26,683 S1: going to get better and better, and either we're going 75 00:04:26,713 --> 00:04:29,113 S1: to merge with it or we're going to be completely 76 00:04:29,113 --> 00:04:32,713 S1: replaced and it would be better to merge. And that 77 00:04:32,713 --> 00:04:36,733 S1: just becomes the new US. And that was Joe's perspective 78 00:04:36,733 --> 00:04:38,833 S1: on it. And it's always been my perspective on it. 79 00:04:38,833 --> 00:04:42,343 S1: And I thought that was really interesting. So basically you 80 00:04:42,343 --> 00:04:46,993 S1: see this kind of elevated really smart Joe on here. 81 00:04:46,993 --> 00:04:49,303 S1: And then you kind of see the opposite when he's 82 00:04:49,303 --> 00:04:52,603 S1: talking to someone else. He kind of just basically adapts 83 00:04:52,603 --> 00:04:55,663 S1: to whoever he's talking to. If he's talking to some fighter, 84 00:04:55,663 --> 00:04:58,363 S1: he'll just talk about fights and he'll just vibe out 85 00:04:58,363 --> 00:05:00,313 S1: on that. If he's talking to some hunter, he'll talk 86 00:05:00,313 --> 00:05:03,553 S1: about hunting like he just blends in with whoever he 87 00:05:03,583 --> 00:05:07,963 S1: kind of is empathic with whoever he's talking to. And 88 00:05:07,963 --> 00:05:10,093 S1: you might want to say, well, that's just a gimmick 89 00:05:10,093 --> 00:05:13,453 S1: and he's just trash and blah, blah, blah. He did 90 00:05:13,453 --> 00:05:18,403 S1: start podcasting, basically. He did make podcasting a thing by 91 00:05:18,403 --> 00:05:21,793 S1: doing the thing that he's doing. So if you don't 92 00:05:21,793 --> 00:05:24,793 S1: like his politics or whatever, you still have to objectively 93 00:05:24,793 --> 00:05:28,273 S1: recognize that he is good at doing what he does. 94 00:05:28,303 --> 00:05:31,963 S1: And I believe it's because of this empathy thing. So anyway, 95 00:05:31,963 --> 00:05:35,053 S1: recommend checking it out. Most of the cool thoughts were 96 00:05:35,053 --> 00:05:38,023 S1: coming from teal one because he's super smart and two 97 00:05:38,023 --> 00:05:40,813 S1: because he was the guest. So that's Joe's job is 98 00:05:40,813 --> 00:05:44,893 S1: to extract that out of the guest. But I'm I'm 99 00:05:44,893 --> 00:05:50,233 S1: still trying to solve Peter Thiel because I have always 100 00:05:50,233 --> 00:05:52,843 S1: thought he was kind of like this right wing, uh, 101 00:05:52,843 --> 00:05:57,163 S1: crazy person. Uh, I really, really was mad at him 102 00:05:57,163 --> 00:06:01,813 S1: because he was supporting Trump back in, what, 2020 or 2016, 103 00:06:01,813 --> 00:06:04,363 S1: I can't remember. And I was like, yeah, this is 104 00:06:04,363 --> 00:06:06,943 S1: I just don't like this person because he's supporting Trump. 105 00:06:06,943 --> 00:06:08,713 S1: And how could you possibly do that? So I was 106 00:06:08,713 --> 00:06:11,983 S1: very triggered by that. He has since been like, yeah, 107 00:06:11,983 --> 00:06:14,833 S1: that was a horrible mistake. Like he's way more chaotic 108 00:06:14,863 --> 00:06:16,843 S1: than I thought he was going to be. And I 109 00:06:16,843 --> 00:06:20,413 S1: think he's Supporting Trump again, but in a very sort 110 00:06:20,443 --> 00:06:22,993 S1: of distant kind of way, almost kind of like Elon 111 00:06:23,023 --> 00:06:25,933 S1: seems to be doing for the purposes of. Same with 112 00:06:25,933 --> 00:06:29,323 S1: Andreesen and same with Horowitz for the purpose of saying 113 00:06:29,323 --> 00:06:32,683 S1: I think the other one is worse. You're not getting 114 00:06:32,713 --> 00:06:36,463 S1: like strong endorsements from any of these. I would call 115 00:06:36,463 --> 00:06:39,673 S1: centre right people. You're not getting strong endorsements of like, 116 00:06:39,703 --> 00:06:43,513 S1: oh yeah, Trump is great, right? And I don't mean 117 00:06:43,513 --> 00:06:46,573 S1: to be talking about politics here, but I'm actually not 118 00:06:46,573 --> 00:06:49,153 S1: talking about politics. I'm talking about what it means to 119 00:06:49,183 --> 00:06:55,633 S1: be an intellectual and have opinions that are counter to Orthodoxy. Right. 120 00:06:55,663 --> 00:06:58,453 S1: So essentially what I'm saying here is I'm trying to 121 00:06:58,453 --> 00:07:02,323 S1: figure teal out. And the thing that I'm realizing about 122 00:07:02,353 --> 00:07:05,563 S1: teal is, and what I'm really liking about teal is 123 00:07:05,563 --> 00:07:08,773 S1: he and sort of dislike in some ways, it's like 124 00:07:08,773 --> 00:07:11,533 S1: he's not really committing to anything. He's not. I don't 125 00:07:11,533 --> 00:07:14,053 S1: see him saying, I think we should do this, which 126 00:07:14,053 --> 00:07:17,053 S1: I would really appreciate. Um, it's something that I try 127 00:07:17,083 --> 00:07:19,973 S1: to do. I try to do the objective thing, but 128 00:07:19,973 --> 00:07:22,703 S1: then come up with a direction to move in. I 129 00:07:22,703 --> 00:07:24,233 S1: feel like that's what you have to do if you 130 00:07:24,233 --> 00:07:27,053 S1: want to try to lead or to motivate or inspire. 131 00:07:27,053 --> 00:07:29,423 S1: But I feel like what teal does is just he's 132 00:07:29,423 --> 00:07:32,993 S1: just got this giant table full of hats, and he's 133 00:07:32,993 --> 00:07:35,663 S1: just always putting these different hats on. And he's like, well, 134 00:07:35,663 --> 00:07:40,253 S1: let's let's steel man that let's beat that up. Let's 135 00:07:40,253 --> 00:07:42,413 S1: steel man that, let's beat that up. So he's just 136 00:07:42,413 --> 00:07:45,473 S1: like constantly looking at pros and cons of different ways 137 00:07:45,473 --> 00:07:48,803 S1: of thinking. And I think he has some general tendencies right. 138 00:07:48,833 --> 00:07:52,613 S1: He's obviously generally libertarian but he'll just straight up tell 139 00:07:52,613 --> 00:07:55,223 S1: you where he's not libertarian. And so what I like 140 00:07:55,253 --> 00:07:58,373 S1: about him is he's not fitting into any sort of 141 00:07:58,373 --> 00:08:01,553 S1: right wing mold. Right. A lot of right people, right 142 00:08:01,553 --> 00:08:03,923 S1: wing people probably think he is. But if you actually 143 00:08:03,923 --> 00:08:07,133 S1: listen to him, he's not. He's actually going down his 144 00:08:07,133 --> 00:08:11,513 S1: own path from first principles. From all this reading, if 145 00:08:11,513 --> 00:08:14,903 S1: you watch his conversation with Tyler Cowen. Oh my God, 146 00:08:14,933 --> 00:08:17,103 S1: I mean, it's like the Joe Rogan one times a 147 00:08:17,103 --> 00:08:20,213 S1: 1000 because they are just riffing on all this history 148 00:08:20,213 --> 00:08:22,703 S1: and all this different stuff that they're talking about, and 149 00:08:22,703 --> 00:08:25,133 S1: they're just jumping from one topic to another. And so 150 00:08:25,133 --> 00:08:28,553 S1: he's using all that knowledge to walk his own path 151 00:08:28,553 --> 00:08:32,723 S1: using first principles. And sometimes that takes him like censure, right. 152 00:08:32,753 --> 00:08:35,933 S1: Sometimes that takes some center left, I imagine sometimes he 153 00:08:35,933 --> 00:08:39,353 S1: goes more extreme left and more extreme right. But I 154 00:08:39,353 --> 00:08:42,233 S1: don't think he honestly cares about left and right. I 155 00:08:42,233 --> 00:08:45,773 S1: think what he cares about is first principles and and 156 00:08:45,773 --> 00:08:48,983 S1: studying and learning and trying to just constantly improve his 157 00:08:48,983 --> 00:08:52,193 S1: models so that he can have all these different hats. 158 00:08:52,223 --> 00:08:54,803 S1: Because ultimately, I think he's an investor, and I think 159 00:08:54,833 --> 00:08:57,203 S1: ultimately his whole game is trying to figure out how 160 00:08:57,203 --> 00:09:01,013 S1: to predict the future to some degree, some limited degree. Oh, 161 00:09:01,013 --> 00:09:04,403 S1: that's the other thing. He's extremely humble. He's extremely cautious. 162 00:09:04,403 --> 00:09:05,963 S1: They'll be like, hey, what do you think of this? 163 00:09:05,993 --> 00:09:08,843 S1: What's the future going to be? He's like, yeah, nobody knows. 164 00:09:08,843 --> 00:09:11,843 S1: It's a dumb question. So it's like the opposite of 165 00:09:11,843 --> 00:09:15,083 S1: the right. Okay. You get a typical right wing pundit. 166 00:09:15,083 --> 00:09:18,053 S1: So let me give you a perfect example. Peter Thiel 167 00:09:18,053 --> 00:09:23,213 S1: is the absolute opposite from Tucker Carlson. Tucker Carlson is 168 00:09:23,213 --> 00:09:27,443 S1: extremely sure about things and extremely wrong. And Peter Thiel 169 00:09:27,443 --> 00:09:31,343 S1: is extremely unsure about things and tends to be very right. 170 00:09:31,343 --> 00:09:34,943 S1: And he's very cautious and he's just constantly consuming and learning. 171 00:09:34,943 --> 00:09:38,303 S1: And it's just yeah, so I'm still trying to lock 172 00:09:38,303 --> 00:09:40,823 S1: in on Peter Thiel. I'm still trying to look for 173 00:09:40,823 --> 00:09:44,213 S1: skeletons or look for biases, but I feel like he's 174 00:09:44,213 --> 00:09:47,483 S1: a pretty good model of how to learn about the world. 175 00:09:47,513 --> 00:09:49,193 S1: As far as I can tell so far, I've got 176 00:09:49,223 --> 00:09:52,703 S1: like a 75% read. I would love to see if 177 00:09:52,703 --> 00:09:56,093 S1: anyone has a different read, or has knowledge that can 178 00:09:56,123 --> 00:09:59,753 S1: sort of shape my model of him based on this, 179 00:09:59,753 --> 00:10:01,973 S1: and I bought one of those mini libraries to put 180 00:10:01,973 --> 00:10:03,683 S1: up in my neighborhood. I don't know if you've ever 181 00:10:03,683 --> 00:10:07,523 S1: seen these. I've seen them all over, definitely in Palo Alto. 182 00:10:07,523 --> 00:10:10,013 S1: So if you drive through one of these neighborhoods in 183 00:10:10,013 --> 00:10:13,523 S1: Palo Alto, like kind of near Stanford, um, not Atherton, 184 00:10:13,523 --> 00:10:16,383 S1: but one of the nicer neighborhoods in Palo Alto. You'll 185 00:10:16,383 --> 00:10:19,203 S1: see like these little it looks like a birdhouse, kind of. 186 00:10:19,233 --> 00:10:22,293 S1: But it's like on the curb, like right against the street. 187 00:10:22,293 --> 00:10:25,173 S1: And it says like mini library or community library or 188 00:10:25,173 --> 00:10:27,903 S1: whatever on it. And it's just like this little birdhouse thing, 189 00:10:27,903 --> 00:10:29,983 S1: except for when you open it up. It's like 1 190 00:10:29,983 --> 00:10:33,513 S1: or 2 shelves of books. And the idea is you 191 00:10:33,513 --> 00:10:35,733 S1: take a book, you just walk away. You just walk 192 00:10:35,733 --> 00:10:37,713 S1: away with it. Maybe you never bring it back, but 193 00:10:37,713 --> 00:10:41,283 S1: the idea is community, karma. So yes, you do bring 194 00:10:41,283 --> 00:10:43,173 S1: it back. Or maybe you bring back an extra book. 195 00:10:43,173 --> 00:10:45,573 S1: So it's just like this. It's a combination of community 196 00:10:45,573 --> 00:10:50,223 S1: with books. Love it obviously. All right. So my work 197 00:10:50,253 --> 00:10:53,703 S1: got this new piece on the four components of what's 198 00:10:53,703 --> 00:10:57,753 S1: going to make models when I models win. And it's 199 00:10:57,753 --> 00:11:01,383 S1: basically I'll go into it here. It's basically the model itself, 200 00:11:01,383 --> 00:11:05,643 S1: post-training internal tooling and then agents. And I'm actually going 201 00:11:05,673 --> 00:11:08,313 S1: to add some extra stuff here. But these are the four. 202 00:11:08,343 --> 00:11:10,503 S1: And then the summary is just the four of those. 203 00:11:10,503 --> 00:11:14,043 S1: So we should start thinking about top models as model 204 00:11:14,043 --> 00:11:17,123 S1: ecosystems Systems instead of just models, because it's not just 205 00:11:17,123 --> 00:11:19,283 S1: the neural net weights doing the work. There are four 206 00:11:19,283 --> 00:11:23,273 S1: main components, and the company that wins the AI model 207 00:11:23,273 --> 00:11:25,643 S1: wars will need to excel at all four of these. 208 00:11:25,643 --> 00:11:28,043 S1: Not just spending lots of money to have the neural 209 00:11:28,043 --> 00:11:31,283 S1: net with the most parameters. That's that piece. Next one 210 00:11:31,283 --> 00:11:34,403 S1: here is the link between free will and LLM denial. 211 00:11:34,403 --> 00:11:38,003 S1: That one's a bit deep and esoteric, and if you 212 00:11:38,003 --> 00:11:40,493 S1: don't like free will, it won't interest you. So I'm 213 00:11:40,493 --> 00:11:43,613 S1: going to skip that one. Okay. Security. Microsoft just released 214 00:11:43,613 --> 00:11:48,323 S1: a bunch of patches 90 security flaws, including ten zero days, 215 00:11:48,323 --> 00:11:52,433 S1: six of which being actively exploited. I think the biggest one, 216 00:11:52,433 --> 00:11:56,483 S1: if I'm not incorrect about this, is wasn't there a 217 00:11:56,483 --> 00:12:01,463 S1: really nasty IPv6 one? Just pull this up real quick. IPv6. 218 00:12:01,493 --> 00:12:07,343 S1: Microsoft repeatedly sent IPv6 packets include specially crafted packets to 219 00:12:07,343 --> 00:12:12,923 S1: windows machine enable remote code execution. Yeah. So pretty nasty. 220 00:12:12,923 --> 00:12:15,753 S1: And I think it's before the firewall. So somebody was 221 00:12:15,753 --> 00:12:19,023 S1: saying it might have been Marcus Hutchins saying this. Who? 222 00:12:19,023 --> 00:12:21,093 S1: He got to meet in person for the first time 223 00:12:21,423 --> 00:12:25,143 S1: after talking to him for a number of years just online, 224 00:12:25,143 --> 00:12:28,023 S1: but got to meet him at this crater meetup that 225 00:12:28,023 --> 00:12:31,143 S1: we do in Vegas. Probably got Covid there. Um, lots 226 00:12:31,143 --> 00:12:33,123 S1: of people probably gave me Covid there. Lots of people 227 00:12:33,123 --> 00:12:35,433 S1: probably got Covid there anyway. Got to meet him in 228 00:12:35,433 --> 00:12:39,213 S1: person there in Vegas. That was cool. But I think 229 00:12:39,213 --> 00:12:42,723 S1: he might have been saying somewhere that basically the IPv6 230 00:12:42,723 --> 00:12:46,113 S1: packets come in and don't even make it to the firewall. 231 00:12:46,113 --> 00:12:50,373 S1: It just goes straight through and gets parsed and produces 232 00:12:50,373 --> 00:12:54,723 S1: the ability to create RC. So that's great. Uh, Russian 233 00:12:54,723 --> 00:12:59,793 S1: cyber spies from the FSB, cold War trail, cold west trail. 234 00:12:59,823 --> 00:13:02,613 S1: Cold west trail. Who knows how to say that? Uh, 235 00:13:02,613 --> 00:13:05,793 S1: been running a massive phishing campaign dubbed River of Phish, 236 00:13:05,793 --> 00:13:10,623 S1: targeting U.S. and European entities since 2022, going after high 237 00:13:10,653 --> 00:13:15,603 S1: risk individuals, NGOs, media outlets, government people, and the Pentagon 238 00:13:15,603 --> 00:13:18,513 S1: is talking about their plan to flood the Taiwan Strait 239 00:13:18,513 --> 00:13:21,543 S1: with thousands of drones in the event of a Chinese invasion. 240 00:13:21,543 --> 00:13:26,403 S1: So this guy, Admiral Sam Samuel Paparo, described the strategy 241 00:13:26,403 --> 00:13:31,293 S1: as creating an unmanned hellscape to delay Chinese forces. Why 242 00:13:31,323 --> 00:13:33,993 S1: are we talking about this? Unless this is part of 243 00:13:33,993 --> 00:13:36,633 S1: the strategy to make them think we're doing this, and 244 00:13:36,633 --> 00:13:38,463 S1: then they try to adjust for that, but that's not 245 00:13:38,463 --> 00:13:41,463 S1: actually what we're doing. Like, are we playing games here? 246 00:13:41,493 --> 00:13:44,943 S1: Like what exactly? Why are we talking about this? Evidently, 247 00:13:44,943 --> 00:13:48,963 S1: it's not meant to be useful as a surprise because 248 00:13:48,963 --> 00:13:52,113 S1: it's no longer a surprise. Whatever. That'd be cool if 249 00:13:52,113 --> 00:13:54,093 S1: it was like, yeah, we're not really going to do that. 250 00:13:54,093 --> 00:13:56,253 S1: We're just going to make them do a bunch of 251 00:13:56,253 --> 00:13:59,253 S1: anti-drone stuff. And actually our strategy is going to be 252 00:13:59,253 --> 00:14:02,883 S1: completely different. Whatever. We'd like to know more. Someone published 253 00:14:02,883 --> 00:14:06,873 S1: a timeline of his research on offensive AI agents, three 254 00:14:06,873 --> 00:14:11,943 S1: distinct types of AI offensive systems. SolarWinds patched a critical 255 00:14:11,943 --> 00:14:17,103 S1: Deserialization Realization vulnerability 9.8 on the Richter scale. Iranian banks 256 00:14:17,103 --> 00:14:19,893 S1: have been hit by a massive cyber attack, probably tied 257 00:14:19,893 --> 00:14:23,073 S1: to Israel. I'm just guessing Trump shared a fake image. 258 00:14:23,103 --> 00:14:26,253 S1: Oh man, I didn't put the link. I just got 259 00:14:26,253 --> 00:14:30,093 S1: an email about this. Didn't put the link. Oh man. Whatever. 260 00:14:30,093 --> 00:14:33,603 S1: Fake image of Harris speaking at a communist event. All right, 261 00:14:33,633 --> 00:14:37,683 S1: let me let me find this thing. Trump share. Yeah, 262 00:14:37,713 --> 00:14:41,943 S1: look at that. Look at that. I mean, does anyone 263 00:14:41,973 --> 00:14:44,523 S1: think she would actually go to an event like that? Well, 264 00:14:44,553 --> 00:14:46,683 S1: first of all, yeah, this this is what I said 265 00:14:46,713 --> 00:14:49,713 S1: actually in the piece. Right. Um, one lots of people 266 00:14:49,713 --> 00:14:52,713 S1: will believe it's real. And two, current tech can already 267 00:14:52,713 --> 00:14:55,743 S1: make more believable ones than this. So this isn't the 268 00:14:55,743 --> 00:15:00,003 S1: actual problem. The actual problem is when it gets much worse. Um. Oh, 269 00:15:00,033 --> 00:15:02,703 S1: here's the other one that they were sharing. Uh, watch 270 00:15:02,703 --> 00:15:08,043 S1: this one. Trump share Taylor Swift should be enough, right? 271 00:15:08,073 --> 00:15:11,853 S1: No links. Yeah. Look at this. Look at this. So 272 00:15:11,853 --> 00:15:13,983 S1: we've been talking about for years. I went looking for 273 00:15:13,983 --> 00:15:15,753 S1: this post where I first started talking about this. I 274 00:15:15,753 --> 00:15:19,953 S1: couldn't find it. So beehive, fix your search, please. Couldn't 275 00:15:19,953 --> 00:15:22,893 S1: find it with Google either, so kind of annoying. Anyway, 276 00:15:22,893 --> 00:15:26,793 S1: I wrote about basically deep fakes becoming an issue because 277 00:15:26,793 --> 00:15:29,103 S1: we're not going to be able to believe things. It's 278 00:15:29,103 --> 00:15:31,623 S1: so obvious now. It's not even a good post, but 279 00:15:31,623 --> 00:15:34,743 S1: it was very early. This was like two years ago. 280 00:15:35,103 --> 00:15:39,603 S1: I was talking about this. Um, but anyway, everyone knows now, uh, 281 00:15:39,603 --> 00:15:44,883 S1: and here is the actual. It's just now starting to matter. Okay, 282 00:15:44,973 --> 00:15:47,523 S1: look at this picture. The top right one that looks 283 00:15:47,523 --> 00:15:50,403 S1: like an actual, like event. This person's got his phone. 284 00:15:50,403 --> 00:15:53,433 S1: He's got his foot sticking out. That looks super real. 285 00:15:53,433 --> 00:15:56,793 S1: Super real Swifties for Trump. Now, you know, be funny 286 00:15:56,793 --> 00:16:01,173 S1: is if that was if that was actually real. Like some. Okay. 287 00:16:01,203 --> 00:16:05,493 S1: The bottom left definitely not real okay. They're probably none 288 00:16:05,493 --> 00:16:08,223 S1: of these are probably real. Just to be very clear about, 289 00:16:08,253 --> 00:16:12,373 S1: you know, disinformation or whatever. But um, the point is, 290 00:16:12,373 --> 00:16:14,383 S1: you can make these. Oh, yeah. This bottom right one 291 00:16:14,383 --> 00:16:17,443 S1: also looks super real. The bottom left one does not. 292 00:16:17,443 --> 00:16:22,543 S1: Because that Swifties for Trump looks like too clear. Like 293 00:16:22,543 --> 00:16:25,093 S1: the white of the text just looks a little bit 294 00:16:25,093 --> 00:16:28,603 S1: too artificial. But this one Swifties for Trump. It's got 295 00:16:28,633 --> 00:16:31,933 S1: shadow on it. The T is hidden because it's on 296 00:16:31,933 --> 00:16:34,903 S1: the other side of her body, so that one looks real. 297 00:16:34,933 --> 00:16:38,383 S1: It's actually the same person as this. Honestly, these two 298 00:16:38,383 --> 00:16:42,073 S1: look really, really good that that would trick basically anybody, 299 00:16:42,133 --> 00:16:45,493 S1: including me right now. So the point is, these are 300 00:16:45,493 --> 00:16:48,223 S1: getting so good that you can't tell the difference between 301 00:16:48,223 --> 00:16:52,093 S1: reality and doesn't really matter for this right now. This 302 00:16:52,093 --> 00:16:55,843 S1: one maybe kind of does. The question is what happens 303 00:16:55,843 --> 00:16:58,543 S1: when it's a thing that really matters? Like did a 304 00:16:58,543 --> 00:17:03,403 S1: crime actually occur? Um, who is the assailant? Evidence of 305 00:17:03,403 --> 00:17:06,793 S1: something bad happening where where we're going to use pictures 306 00:17:06,793 --> 00:17:10,033 S1: or video as part of the evidence for that crime. 307 00:17:10,033 --> 00:17:13,063 S1: That's when it starts to really, really matter. And also 308 00:17:13,063 --> 00:17:16,903 S1: for politics as well. Um, the worst possible one is, 309 00:17:16,933 --> 00:17:22,633 S1: is like the absolute worst one is where you have, um, 310 00:17:22,663 --> 00:17:27,253 S1: a video of Trump saying, you know, it looks like 311 00:17:27,253 --> 00:17:30,403 S1: we're going to have to attack Russia, pre-emptive nuclear strike. 312 00:17:30,403 --> 00:17:33,703 S1: And it's very believable. And it freaked somebody out. And 313 00:17:33,703 --> 00:17:36,853 S1: that person freaks out Putin and something goes off or 314 00:17:36,853 --> 00:17:39,103 S1: something similar to that. Maybe it's not a nuclear war, 315 00:17:39,103 --> 00:17:43,303 S1: but something where there's an actual threat being made and 316 00:17:43,303 --> 00:17:46,243 S1: it causes the other person to respond. So it's a 317 00:17:46,243 --> 00:17:51,583 S1: trigger point for kinetic action that that's pretty serious. And 318 00:17:51,583 --> 00:17:55,783 S1: we have the tech now to be able to do that. Um, 319 00:17:55,783 --> 00:17:59,023 S1: not everyone could do it, but it's getting pretty close. 320 00:17:59,023 --> 00:18:01,213 S1: And it's much easier to do with images than it 321 00:18:01,213 --> 00:18:04,363 S1: is for video. But the point is, this is getting very, 322 00:18:04,363 --> 00:18:08,923 S1: very real. No longer theoretical Iranian hacking group Abt 42 323 00:18:08,953 --> 00:18:13,663 S1: has targeted both Trump and Biden campaigns, according to Google's 324 00:18:13,663 --> 00:18:18,343 S1: Tag group. They look like they work for Iran's Revolutionary 325 00:18:18,373 --> 00:18:21,373 S1: Guard Corps, which is like their elite group. And it 326 00:18:21,373 --> 00:18:24,733 S1: looks like they targeted both campaigns. But only Trump's campaign 327 00:18:24,733 --> 00:18:29,113 S1: had sensitive files leaked. Kind of interesting. I mean, maybe 328 00:18:29,113 --> 00:18:32,683 S1: there was nothing there. On the Democratic side, maybe, I 329 00:18:32,683 --> 00:18:37,423 S1: don't know. Why are they why are they anti-Trump? Um. Oh, maybe. 330 00:18:37,753 --> 00:18:42,193 S1: Maybe because Trump is about to massively support Israel and 331 00:18:42,193 --> 00:18:45,583 S1: massively crush Iran. That would be a good explanation. Trump 332 00:18:45,583 --> 00:18:49,063 S1: corroborated this by pointing the finger at Iran for hacking 333 00:18:49,063 --> 00:18:52,963 S1: his campaign, praising the FBI's investigation into the breach. This 334 00:18:52,963 --> 00:18:56,113 S1: is one thing that I like decently about Trump when 335 00:18:56,113 --> 00:18:58,213 S1: he does it, and it's only because the bar is 336 00:18:58,213 --> 00:19:00,493 S1: so low. But if he gets on a phone call 337 00:19:00,493 --> 00:19:03,163 S1: with the FBI and the FBI is like, yes sir, 338 00:19:03,193 --> 00:19:05,593 S1: no sir, blah, blah, blah, you know, we did this 339 00:19:05,593 --> 00:19:08,053 S1: or we did that. If he just got done railing 340 00:19:08,083 --> 00:19:10,753 S1: against the FBI for like the last six months, and 341 00:19:10,753 --> 00:19:12,703 S1: he has that call. He'll go and talk to the 342 00:19:12,703 --> 00:19:14,383 S1: press and be like, yeah, I just had a great 343 00:19:14,383 --> 00:19:17,563 S1: call with the FBI. Very professional, very cool. It's like 344 00:19:17,563 --> 00:19:20,323 S1: the thing that happened most recently is like his view 345 00:19:20,323 --> 00:19:22,723 S1: of reality, but at least he's willing to, like, be 346 00:19:22,723 --> 00:19:25,543 S1: positive if a positive thing happened. I mean, I can't 347 00:19:25,543 --> 00:19:29,293 S1: believe I'm saying that as a positive thing, since that's 348 00:19:29,293 --> 00:19:33,793 S1: called normal. China linked cyber spies have infected dozens of 349 00:19:33,793 --> 00:19:37,393 S1: Russian government and IT sector computers with backdoors and trojans 350 00:19:37,393 --> 00:19:41,563 S1: since late July, according to Kaspersky. So, China v Russia 351 00:19:41,593 --> 00:19:44,983 S1: scammers are targeting young Chinese job seekers in a tough economy. 352 00:19:45,013 --> 00:19:49,483 S1: AI and tech grok chatbot now lets users create images 353 00:19:49,483 --> 00:19:52,423 S1: from text prompts and publish them to X, leading to 354 00:19:52,453 --> 00:19:56,173 S1: chaotic results like Barack Obama doing cocaine and Donald Trump 355 00:19:56,173 --> 00:19:59,923 S1: and a Nazi uniform. Yeah, the one that I saw. Yeah, 356 00:19:59,923 --> 00:20:02,743 S1: I wrote this right here. Um, the one I saw 357 00:20:02,743 --> 00:20:06,133 S1: was Trump basically had his arm around Elon, and Elon 358 00:20:06,133 --> 00:20:08,893 S1: was pregnant, and Elon posted that. And he was like 359 00:20:08,893 --> 00:20:13,513 S1: live by the sword. Die by the sword. Yeah. Okay. 360 00:20:13,543 --> 00:20:16,633 S1: Got someone who's on a mission to make you love 361 00:20:16,633 --> 00:20:19,843 S1: reading again. So he's using AI tools like ChatGPT to 362 00:20:19,873 --> 00:20:26,203 S1: recommend books, understand deeper themes, and novels like Hermann Hesse's Siddhartha, 363 00:20:26,203 --> 00:20:30,253 S1: and create actionable strategies from business books like Alex Hormozi 364 00:20:30,283 --> 00:20:34,303 S1: $100 Million Offers. I've already got that built into fabric. Um, 365 00:20:34,303 --> 00:20:38,293 S1: I've got one called create $100 million offer. You give 366 00:20:38,293 --> 00:20:42,013 S1: it any product, concept or idea and it'll build you 367 00:20:42,043 --> 00:20:45,373 S1: a hormozi offer based on it. Comedians are increasingly using 368 00:20:45,403 --> 00:20:49,123 S1: AI to help write jokes and brainstorm ideas with mixed results. 369 00:20:49,123 --> 00:20:52,933 S1: I think this is as big or bigger, not bigger 370 00:20:52,933 --> 00:20:55,633 S1: as big, maybe as the Turing test in terms of 371 00:20:55,633 --> 00:20:58,933 S1: the importance of AI progress. Because if I can write 372 00:20:58,963 --> 00:21:03,493 S1: a full comedy set and make humans laugh, that is 373 00:21:03,493 --> 00:21:07,903 S1: massively huge. Even better if it can be deepfake performed 374 00:21:07,903 --> 00:21:11,903 S1: right So it's actually I writing it and I performing it, 375 00:21:11,903 --> 00:21:15,773 S1: and all the human did was tell an agent to 376 00:21:15,803 --> 00:21:21,053 S1: write a comedy set around a topic. Okay. That is this. 377 00:21:21,053 --> 00:21:25,763 S1: This is an unbelievable metric. Okay. A milestone metric. So basically, 378 00:21:25,793 --> 00:21:29,693 S1: here's the test. Can I give an agent or any model? 379 00:21:29,693 --> 00:21:33,413 S1: Let's say, for example, I hand this to I write 380 00:21:33,443 --> 00:21:38,753 S1: a comedy sketch skit or set around, um, good coffee 381 00:21:38,783 --> 00:21:43,073 S1: versus bad coffee or how baristas don't care about you 382 00:21:43,073 --> 00:21:46,283 S1: or how customer services people don't care about you. Do 383 00:21:46,313 --> 00:21:49,433 S1: write a 20 minute set just about this topic. Okay. 384 00:21:49,463 --> 00:21:52,103 S1: So now I want to see on a screen either 385 00:21:52,103 --> 00:21:55,433 S1: a stand up view or a face view. Yeah. A 386 00:21:55,433 --> 00:21:57,203 S1: stand up view would be better. Actually, I want to 387 00:21:57,203 --> 00:21:59,093 S1: see a stage. I want to see a microphone. I 388 00:21:59,093 --> 00:22:02,153 S1: want to see a person standing there talking, and I 389 00:22:02,153 --> 00:22:03,893 S1: want to see them moving. I want to see them 390 00:22:03,893 --> 00:22:06,953 S1: moving their hands. I want to see them gesturing a full, 391 00:22:06,953 --> 00:22:10,463 S1: complete deepfake of the content that they wrote. And this 392 00:22:10,463 --> 00:22:12,563 S1: has to be funny. Okay. And it doesn't even have 393 00:22:12,563 --> 00:22:14,543 S1: to be like, the funniest thing. It doesn't have to 394 00:22:14,543 --> 00:22:18,173 S1: be Dave Chappelle. It has to be decently funny, like 395 00:22:18,173 --> 00:22:21,563 S1: good enough to get gigs on a comedy tour. And 396 00:22:21,563 --> 00:22:23,753 S1: if you've been on a comedy, tours like The Bar 397 00:22:23,753 --> 00:22:27,383 S1: is not super high for the average comic that I 398 00:22:27,383 --> 00:22:29,603 S1: don't know what I skeptics are going to say to that. 399 00:22:29,603 --> 00:22:31,553 S1: And I would say I'm going to I'm going to 400 00:22:31,553 --> 00:22:34,403 S1: call it right now. We are in the middle of 2024. 401 00:22:34,433 --> 00:22:39,983 S1: I'm going to call that as 2025, 2026 at the latest. 402 00:22:39,983 --> 00:22:45,533 S1: Full deepfake integration 2026 at the latest. So both 2026, 403 00:22:45,533 --> 00:22:48,233 S1: both in 2026. That's two years from now, a year 404 00:22:48,233 --> 00:22:50,303 S1: and a half from now, we will have the ability 405 00:22:50,303 --> 00:22:53,633 S1: to have a deepfake comic, write a funny set and 406 00:22:53,633 --> 00:22:55,763 S1: perform it in a way that if a regular person 407 00:22:55,763 --> 00:22:58,343 S1: who enjoys comedy watches it, they'll be like, yeah, that's 408 00:22:58,343 --> 00:23:01,613 S1: pretty funny. That's decently funny. Now, could they get their 409 00:23:01,613 --> 00:23:05,543 S1: own Netflix Netflix special based on that content? That would 410 00:23:05,543 --> 00:23:07,763 S1: be like the top, top tier, but keep in mind 411 00:23:07,763 --> 00:23:11,903 S1: how few comics actually have Netflix specials. So. Yeah. Interesting. 412 00:23:11,933 --> 00:23:14,813 S1: What did I say? 27. No, I said 26. Yeah. 413 00:23:14,843 --> 00:23:17,303 S1: All right. I'm holding myself to that. San Francisco is 414 00:23:17,303 --> 00:23:19,883 S1: looking to ban software that critics claim is being used 415 00:23:19,883 --> 00:23:25,703 S1: to artificially inflate rates, so it allegedly helps landlords coordinate 416 00:23:25,703 --> 00:23:29,453 S1: rent increases. You might be over using vim visual mode. 417 00:23:29,453 --> 00:23:31,793 S1: I've been worried about this for quite some time, actually. 418 00:23:31,793 --> 00:23:34,523 S1: Oh yeah, instead of doing GG takes you to the 419 00:23:34,523 --> 00:23:38,753 S1: top V which switched to visual mode, and then capital 420 00:23:38,783 --> 00:23:42,293 S1: G which goes to the bottom of the file. Instead 421 00:23:42,293 --> 00:23:46,673 S1: you do GG quote plus Y capital. That's a lot 422 00:23:46,703 --> 00:23:48,953 S1: of characters, man. Anyway, I'm going to try to learn 423 00:23:48,953 --> 00:23:51,053 S1: how to do this and do it more with normal 424 00:23:51,053 --> 00:23:54,203 S1: commands instead of visual. All right humans, uh, we're going 425 00:23:54,233 --> 00:23:59,243 S1: to have driver's licenses in the Apple Wallet before too long, 426 00:23:59,273 --> 00:24:01,763 S1: and you'll be able to go through LAX and SFO. 427 00:24:01,793 --> 00:24:05,663 S1: China's manufacturers are facing a financial crisis, with many going 428 00:24:05,663 --> 00:24:09,083 S1: bankrupt due to a combination of weak demand, rising costs, 429 00:24:09,113 --> 00:24:12,863 S1: and increased competition. Scientists at Fermilab have detected the first 430 00:24:12,863 --> 00:24:18,593 S1: neutrinos using a prototype detector. Deep Underground Neutrino Experiment. Dune. 431 00:24:18,593 --> 00:24:21,683 S1: You gotta land, you gotta land. The acronym. They landed it. 432 00:24:21,683 --> 00:24:25,673 S1: Venture capitalists aren't looking for nice founders. They want risk takers. 433 00:24:25,703 --> 00:24:29,333 S1: Nate Silver highlights that 70% of the billionaires on the 434 00:24:29,333 --> 00:24:36,293 S1: 2023 Forbes 400 list are self-made, 70% self-made, modest backgrounds. 435 00:24:36,503 --> 00:24:39,953 S1: So I saw Scott Galloway talking to Nate Silver about this. 436 00:24:39,953 --> 00:24:43,283 S1: I really love this idea of basically Nate Silver was 437 00:24:43,283 --> 00:24:45,203 S1: talking about this thing I've been talking about for a while, 438 00:24:45,203 --> 00:24:49,493 S1: which is they want crazy religious people who are just like, yep, 439 00:24:49,493 --> 00:24:52,523 S1: I have no life whatsoever. I will sleep under my desk. 440 00:24:52,523 --> 00:24:54,953 S1: That is what people are looking for. They're looking for 441 00:24:54,953 --> 00:24:57,653 S1: a crazy vision. And when I say people, I'm talking 442 00:24:57,653 --> 00:25:01,583 S1: about VCs and also startup founders who are hiring people. 443 00:25:01,583 --> 00:25:04,673 S1: And I would argue this is the future of all work, 444 00:25:04,673 --> 00:25:07,523 S1: the future of all work is you only hire these people. 445 00:25:07,523 --> 00:25:11,243 S1: You only hire the most talented and the most crazy 446 00:25:11,273 --> 00:25:16,073 S1: dedicated because to hire a worker who just does normal work, 447 00:25:16,103 --> 00:25:18,833 S1: they show up 9 to 5 and they do average 448 00:25:18,833 --> 00:25:22,853 S1: level work. That's called I, except for it doesn't go home, 449 00:25:22,853 --> 00:25:26,273 S1: it doesn't commute, it doesn't draw a salary and you 450 00:25:26,273 --> 00:25:30,023 S1: can upgrade it instantly. Okay. So 9 to 5 workers 451 00:25:30,023 --> 00:25:33,773 S1: who just do average work that gets replaced. The only 452 00:25:33,773 --> 00:25:36,893 S1: people I think that people are going to want going forward. 453 00:25:36,893 --> 00:25:39,413 S1: And this is an exaggeration. There's still going to be 454 00:25:39,413 --> 00:25:42,053 S1: a need for some some of the other type. But 455 00:25:42,053 --> 00:25:45,683 S1: I would say in general the motion towards is towards 456 00:25:45,683 --> 00:25:49,793 S1: this dedicated zealot crazy person which happens to be young, 457 00:25:49,823 --> 00:25:54,053 S1: happens to be male, usually, um, and just obsessed with tech, 458 00:25:54,083 --> 00:25:57,443 S1: obsessed with building things, obsessed with like, and you got 459 00:25:57,443 --> 00:26:00,503 S1: all sorts of hang ups around this. You got God complexes, 460 00:26:00,503 --> 00:26:04,583 S1: you got narcissism, you got like, just ambition. Just all 461 00:26:04,583 --> 00:26:08,403 S1: sorts of crazy testosterone powered things going on there. And 462 00:26:08,403 --> 00:26:12,423 S1: the upside is taking risk there. Like, oh yeah, just 463 00:26:12,423 --> 00:26:14,553 S1: I'll take all my money and put it into this 464 00:26:14,553 --> 00:26:17,703 S1: idea and sleep under my desk for four years. And 465 00:26:17,703 --> 00:26:19,923 S1: you don't see many older people doing that. You don't 466 00:26:19,923 --> 00:26:23,193 S1: see many women doing that. There are some, uh, a 467 00:26:23,193 --> 00:26:25,653 S1: decent amount, right? If you look at the entire world, 468 00:26:25,653 --> 00:26:29,313 S1: but in general it starts to look like one particular demographic. 469 00:26:29,313 --> 00:26:33,153 S1: And silver is basically saying, yeah, that's what gets it done. 470 00:26:33,183 --> 00:26:36,543 S1: That's what VCs are looking for and not people who 471 00:26:36,543 --> 00:26:41,073 S1: are cautious. And I find that really interesting. This negativity, 472 00:26:41,073 --> 00:26:45,273 S1: this risk, acceptance. It feels very Atlas Shrugged to me, 473 00:26:45,273 --> 00:26:48,183 S1: which is kind of awesome and kind of gross for 474 00:26:48,183 --> 00:26:51,993 S1: different reasons. But I think that's one of the premises 475 00:26:51,993 --> 00:26:55,503 S1: of Atlas Shrugged is that the risk takers are the 476 00:26:55,503 --> 00:26:58,233 S1: ones who get the benefits, like the awesome people are 477 00:26:58,233 --> 00:27:00,513 S1: the ones who are actually building. And I feel like 478 00:27:00,513 --> 00:27:03,633 S1: Silicon Valley is based around this entire concept, is that 479 00:27:03,633 --> 00:27:05,793 S1: those people who sleep under the desk. They are the 480 00:27:05,793 --> 00:27:09,063 S1: special people. And of course, very few of them actually succeed, 481 00:27:09,213 --> 00:27:12,903 S1: you know? Especially the first time. But anyway, really cool 482 00:27:12,903 --> 00:27:15,483 S1: concept coming out of Nate Silver. And by the way, 483 00:27:15,483 --> 00:27:18,603 S1: he's the one who did FiveThirtyEight, although he does not 484 00:27:18,603 --> 00:27:21,543 S1: run it anymore, someone else is running it and they're 485 00:27:21,543 --> 00:27:26,283 S1: actually competing using Nate Silver's brand of FiveThirtyEight. So do 486 00:27:26,283 --> 00:27:28,803 S1: not tie yourself to a brand and then sell it, 487 00:27:28,803 --> 00:27:31,143 S1: because everyone will think you still have it. Growing trend 488 00:27:31,173 --> 00:27:34,893 S1: of Gen Z men becoming NEETs not in employment, education 489 00:27:34,893 --> 00:27:38,493 S1: or training. 1 in 5 young men under 25 unemployed 490 00:27:38,493 --> 00:27:42,303 S1: and not actively looking for work, 1 in 5 unemployed 491 00:27:42,303 --> 00:27:45,483 S1: and not looking for work. What the eff is going on? 492 00:27:45,483 --> 00:27:48,783 S1: Slow is smooth. Smooth is fast. I've always loved this. Um, 493 00:27:48,783 --> 00:27:52,113 S1: I'm ex-military myself. I've heard this a million times. It's 494 00:27:52,113 --> 00:27:54,963 S1: not just Navy Seals. Lots of military talk about this, 495 00:27:54,963 --> 00:27:58,203 S1: but cool idea. No one wants kids anymore and it's 496 00:27:58,203 --> 00:28:02,073 S1: not you. Cool video talks about declining birth rates, touching 497 00:28:02,073 --> 00:28:06,813 S1: on economic pressures changing societal values and personal choices. Imposter 498 00:28:06,813 --> 00:28:11,133 S1: syndrome often stems from systemic biases, not just self-doubt, and 499 00:28:11,133 --> 00:28:16,143 S1: highlights that many women experience this due to real exclusionary practices, 500 00:28:16,143 --> 00:28:19,053 S1: which is 100% true. Guy got fired and replaced by 501 00:28:19,083 --> 00:28:22,743 S1: AI at cosmos magazine, and the management didn't tell anyone. 502 00:28:22,743 --> 00:28:26,493 S1: They're using generative AI to write articles, possibly trained on 503 00:28:26,523 --> 00:28:30,003 S1: their own authors work, and basically firing all the humans 504 00:28:30,003 --> 00:28:33,033 S1: who could have possibly predicted this. I give my kids 505 00:28:33,063 --> 00:28:36,573 S1: a summer like mine. In the 1980s, this parent decided 506 00:28:36,573 --> 00:28:39,453 S1: to give her ten and five year old daughters a 507 00:28:39,453 --> 00:28:43,593 S1: taste of a 1980s summer holiday where the bedroom where 508 00:28:43,623 --> 00:28:50,313 S1: boredom was common and self entertainment was key. 1980s summer holiday. 509 00:28:50,313 --> 00:28:51,783 S1: This is when I grew up. I grew up in 510 00:28:51,783 --> 00:28:57,273 S1: the 80s where boredom was common and self-entertainment was key. Boredom. 511 00:28:57,273 --> 00:28:59,643 S1: You know how awesome it is. I'm thinking about this 512 00:28:59,643 --> 00:29:03,903 S1: right now. To be a kid outside on a weekend. 513 00:29:03,903 --> 00:29:06,603 S1: you had some sort of breakfast which your parents cooked 514 00:29:06,603 --> 00:29:09,363 S1: for you. You are now out playing. That's what it's called. 515 00:29:09,363 --> 00:29:11,913 S1: You're out playing. I don't know what age this is. 516 00:29:11,913 --> 00:29:15,453 S1: Let's call it 12 or something like that or 14. 517 00:29:15,483 --> 00:29:19,083 S1: You then get together with your buddies and obviously I'm 518 00:29:19,083 --> 00:29:22,173 S1: a boy, so I'm going to have a boy perspective. 519 00:29:22,173 --> 00:29:24,843 S1: But you would get together with your guy friends all 520 00:29:24,843 --> 00:29:26,733 S1: the same age and you're just like, what are we 521 00:29:26,763 --> 00:29:28,563 S1: going to do? I don't know, I don't know, what 522 00:29:28,563 --> 00:29:30,213 S1: are you going to do? And you're just like, maybe 523 00:29:30,243 --> 00:29:32,283 S1: we should go over here. No. That's stupid. We just 524 00:29:32,283 --> 00:29:34,473 S1: did that. Uh, maybe we should go over here. Hey, 525 00:29:34,473 --> 00:29:35,973 S1: you want to, like, see if we can lift up 526 00:29:35,973 --> 00:29:38,703 S1: these manhole covers and, like, roll them around? Yeah, that 527 00:29:38,703 --> 00:29:41,223 S1: sounds cool. So you go over, you try to stick 528 00:29:41,253 --> 00:29:43,593 S1: your finger in the hole. You try to, like, wedge 529 00:29:43,593 --> 00:29:46,053 S1: some sort of piece of metal in there and yank 530 00:29:46,083 --> 00:29:47,943 S1: it up. Turns out the metal was sharp and you 531 00:29:47,943 --> 00:29:50,793 S1: just gouged your hand open. You're bleeding all over the place, 532 00:29:50,793 --> 00:29:52,983 S1: so someone, like, rips the sleeve off of their shirt 533 00:29:52,983 --> 00:29:55,083 S1: and wraps it around your hand, and then you just 534 00:29:55,083 --> 00:29:58,323 S1: keep doing it, and then you just, like, roll this thing. 535 00:29:58,323 --> 00:30:01,023 S1: Then you realize, well, someone walks by here and falls 536 00:30:01,023 --> 00:30:04,123 S1: in that hole like you're super screwed. You know you're 537 00:30:04,123 --> 00:30:06,193 S1: going to hurt yourself. Maybe one of your friends actually 538 00:30:06,193 --> 00:30:08,713 S1: do it. Maybe you decide to climb down there because 539 00:30:08,713 --> 00:30:11,113 S1: that's stupid also, but also kind of fun. Do you 540 00:30:11,113 --> 00:30:14,593 S1: know how many hundreds of these moments that I've had 541 00:30:14,593 --> 00:30:17,353 S1: that Gen X people have had that are just magical? 542 00:30:17,353 --> 00:30:19,423 S1: I was talking to my girl about this as well. 543 00:30:19,453 --> 00:30:22,723 S1: She loves this. The same exact thing. She would get 544 00:30:22,753 --> 00:30:26,143 S1: on her bike and go riding all over the frickin town, 545 00:30:26,143 --> 00:30:29,083 S1: riding all over. Just go visit her friends and hang out. 546 00:30:29,083 --> 00:30:31,603 S1: And she'd have her bike with a banana seat and 547 00:30:31,603 --> 00:30:34,183 S1: just like, take it everywhere. And then she would start 548 00:30:34,183 --> 00:30:36,853 S1: getting hungry. The sun would start going down. And if 549 00:30:36,853 --> 00:30:39,553 S1: she didn't get home once that happened, that was an 550 00:30:39,553 --> 00:30:44,863 S1: ass whippin. And just everyone understood that you are free, 1,000% 551 00:30:44,863 --> 00:30:47,353 S1: free until the street lights come on. Then you got 552 00:30:47,353 --> 00:30:50,173 S1: to get home. Then there's dinner there for you. You 553 00:30:50,173 --> 00:30:53,653 S1: have dinner and it's just magical. There is no boredom now. 554 00:30:53,683 --> 00:30:57,493 S1: Kids do not get bored because TikTok or Facebook or 555 00:30:57,493 --> 00:31:01,663 S1: Instagram or whatever it is, texting is so compelling and 556 00:31:01,663 --> 00:31:06,073 S1: arguably more compelling than manhole covers. And that's a problem. 557 00:31:06,103 --> 00:31:07,933 S1: This is what I think a big problem is. I 558 00:31:07,933 --> 00:31:10,963 S1: don't know why I'm fucking rambling about this shit. I'm 559 00:31:10,963 --> 00:31:14,443 S1: supposed to be supposed to be delivering this podcast. I'm 560 00:31:14,443 --> 00:31:17,623 S1: just going to do whatever. When stimuli is too high, 561 00:31:17,623 --> 00:31:21,433 S1: the dopamine level is too high, and anything that drops 562 00:31:21,433 --> 00:31:25,303 S1: the dopamine level down. Let's say that currently in 2024, 563 00:31:25,333 --> 00:31:28,573 S1: things are so awesome in terms of stimuli and inputs. 564 00:31:28,573 --> 00:31:32,803 S1: The dopamine levels are like an 85, let's say in 565 00:31:32,803 --> 00:31:36,343 S1: the 1980s. As an average kid, they used to be 566 00:31:36,343 --> 00:31:39,493 S1: like a 35, okay. And now they're an 85 because 567 00:31:39,493 --> 00:31:43,813 S1: of like all this tech. Okay, better games, better game consoles, 568 00:31:43,813 --> 00:31:49,153 S1: better screens, better audio experiences, more video games. Okay. More 569 00:31:49,153 --> 00:31:53,353 S1: access to porn, all the social media, TikTok. I mean, 570 00:31:53,353 --> 00:31:56,323 S1: it's not even 85. It's like 285. And it used 571 00:31:56,323 --> 00:31:59,623 S1: to be a 35. Okay, now try to take somebody 572 00:31:59,623 --> 00:32:02,383 S1: like that and put him outside. They're like, are you 573 00:32:02,413 --> 00:32:05,263 S1: kidding me? It's dirty out here. You want me to 574 00:32:05,293 --> 00:32:07,783 S1: ride my bike? You want me to? Like, what am 575 00:32:07,783 --> 00:32:11,383 S1: I supposed to do out here? Nothing compares. Nothing in 576 00:32:11,413 --> 00:32:16,903 S1: that world compares to TikTok. TikTok is crack. It's crack cocaine. 577 00:32:16,903 --> 00:32:20,893 S1: And that's dangerous. Now what everyone seems to know and 578 00:32:20,893 --> 00:32:23,143 S1: the experts seem to know. And I think you can 579 00:32:23,143 --> 00:32:25,783 S1: learn this just by doing it, is you can actually 580 00:32:25,783 --> 00:32:30,253 S1: reset your dopamine level very quickly. So what this person 581 00:32:30,253 --> 00:32:34,273 S1: I think did this parents decided to give her ten 582 00:32:34,303 --> 00:32:36,883 S1: and five year old daughters a taste of the 1980s 583 00:32:36,913 --> 00:32:40,543 S1: summer holiday, where boredom was common and self entertainment was key. 584 00:32:40,573 --> 00:32:43,003 S1: So what that means is you have to take the 585 00:32:43,003 --> 00:32:45,763 S1: things away and throw them together and just be like, 586 00:32:45,793 --> 00:32:48,433 S1: figure out something to do and they'll be like, oh, well, 587 00:32:48,433 --> 00:32:50,533 S1: let's take this blanket and put it on an angle. 588 00:32:50,533 --> 00:32:52,963 S1: And now we have a fort, right? And pretty soon 589 00:32:52,963 --> 00:32:56,653 S1: you're making forts because you were bored. And that enjoyment 590 00:32:56,653 --> 00:32:59,863 S1: of coming up with something from boredom and getting your 591 00:32:59,893 --> 00:33:04,663 S1: dopamine level from a five to a 25 feels amazing. 592 00:33:04,663 --> 00:33:09,223 S1: And it involves more creativity, more innovation. I think it's 593 00:33:09,223 --> 00:33:12,223 S1: just healthier. It's just healthier. It's more natural, it's more healthy. 594 00:33:12,223 --> 00:33:14,863 S1: And I don't think there's a problem with all the 595 00:33:14,863 --> 00:33:19,963 S1: tech and hyper, you know, blasting with dopamine from a 596 00:33:19,963 --> 00:33:23,053 S1: TikTok or whatever. But it's you got to regulate it. 597 00:33:23,053 --> 00:33:25,663 S1: You got to regulate it and go back to this 598 00:33:25,663 --> 00:33:29,893 S1: dopamine fasting and, you know, lower and slower anyway. All right. 599 00:33:29,923 --> 00:33:33,793 S1: Few ideas I've had recently. The ultimate privilege, I think 600 00:33:33,793 --> 00:33:36,613 S1: the ultimate privilege might be growing up in a stable 601 00:33:36,613 --> 00:33:40,423 S1: household with two parents who give you a strong work ethic. 602 00:33:40,423 --> 00:33:42,583 S1: It trips me out how simple this is and how 603 00:33:42,583 --> 00:33:45,373 S1: the best advice is often like this. It's the exact 604 00:33:45,373 --> 00:33:49,243 S1: same with diet, exercise, relationships, and a million other things. 605 00:33:49,243 --> 00:33:54,223 S1: The best advice is concise, wise, and generally hard to do, 606 00:33:54,223 --> 00:33:57,433 S1: but it is not a mystery. I think the US 607 00:33:57,433 --> 00:34:00,823 S1: and the world should lock in on this one thing 608 00:34:00,823 --> 00:34:04,843 S1: stable two parent households that imbue a strong work ethic 609 00:34:04,843 --> 00:34:08,773 S1: and focus a lot of energy on getting to 100% 610 00:34:08,773 --> 00:34:11,893 S1: on that metric. All right, next idea. The biggest market 611 00:34:11,893 --> 00:34:14,983 S1: opening right now is for a product platform that validates 612 00:34:14,983 --> 00:34:19,273 S1: the authenticity of content coming from a creator or publisher. Yeah, 613 00:34:19,273 --> 00:34:22,093 S1: that's a tweet. I'm not going to do the whole thread. 614 00:34:22,123 --> 00:34:24,253 S1: I used to think there was a big difference between 615 00:34:24,253 --> 00:34:26,863 S1: somebody being weak and somebody being evil. I now treat 616 00:34:26,863 --> 00:34:29,743 S1: them mostly the same because the outcomes they manifest are 617 00:34:29,743 --> 00:34:31,903 S1: mostly the same. The only difference is that with a 618 00:34:31,903 --> 00:34:34,603 S1: weak person, I could try to make them strong. Someone 619 00:34:34,603 --> 00:34:38,143 S1: responded and said, don't you mean cowardice and not weakness? 620 00:34:38,143 --> 00:34:41,533 S1: I'm kind of equivocating those two. But I do see 621 00:34:41,533 --> 00:34:44,923 S1: their point. I'm not talking about like little children. I'm 622 00:34:44,953 --> 00:34:47,683 S1: talking about who I'm not talking about somebody who's supposed 623 00:34:47,683 --> 00:34:52,483 S1: to be weak because they're disabled or they're super young. They're, 624 00:34:52,513 --> 00:34:55,063 S1: you know, one and a half years old. Stop being weak. 625 00:34:55,063 --> 00:34:57,283 S1: That's not the point. The point is somebody who is 626 00:34:57,283 --> 00:35:01,103 S1: weak due to a flaw in character and who is 627 00:35:01,103 --> 00:35:05,813 S1: causing damage to people around them and themselves as a result. 628 00:35:05,843 --> 00:35:10,373 S1: And importantly, somebody, ideally who could potentially fix themselves. That's 629 00:35:10,373 --> 00:35:13,403 S1: what I'm talking about. All right. Discovery fabric plus raycast. 630 00:35:13,403 --> 00:35:16,463 S1: So this guy Wil Chen did a YouTube video on 631 00:35:16,463 --> 00:35:20,213 S1: how to integrate fabric into raycast. And guess what. I 632 00:35:20,243 --> 00:35:23,423 S1: am now using raycast. Look at that. Activate trial. I 633 00:35:23,423 --> 00:35:27,623 S1: need to do that. This is. Yeah, I switched back 634 00:35:27,623 --> 00:35:30,653 S1: to spotlight because I like to use native tools rather 635 00:35:30,653 --> 00:35:33,023 S1: than third party. I was on Alfred for years and 636 00:35:33,023 --> 00:35:35,873 S1: years and years and years. Then I switched back to 637 00:35:35,903 --> 00:35:38,903 S1: spotlight because there was some OS upgrade. I was like, oh, 638 00:35:38,933 --> 00:35:41,393 S1: let's see how good spotlight is. I prefer to use 639 00:35:41,393 --> 00:35:44,213 S1: the native tools anyway, unless there's a compelling reason to 640 00:35:44,243 --> 00:35:46,523 S1: use a third party tool, which is why I'm on 641 00:35:46,523 --> 00:35:49,583 S1: Safari instead of Chrome. Even though Chrome is way faster. 642 00:35:49,583 --> 00:35:54,563 S1: I like the inbuilt stuff in Safari. Anyway, fabric integration 643 00:35:54,563 --> 00:35:58,133 S1: into raycast. That is a big enough reason for me 644 00:35:58,133 --> 00:36:00,413 S1: to switch back So guess what? Now I'm about to 645 00:36:00,443 --> 00:36:04,493 S1: become a raycast guru and look at this. I might 646 00:36:04,493 --> 00:36:07,283 S1: integrate it more deeply by hosting a set of these 647 00:36:07,283 --> 00:36:10,703 S1: scripts within fabric. So basically these are Python scripts that 648 00:36:10,703 --> 00:36:15,023 S1: raycast reads and that's how raycast executes these things. So 649 00:36:15,023 --> 00:36:17,723 S1: I'm thinking about putting a whole directory of these inside 650 00:36:17,723 --> 00:36:19,883 S1: of fabric. So you could just point raycast to that 651 00:36:19,883 --> 00:36:22,673 S1: directory and boom you're ready to go. And so the 652 00:36:22,673 --> 00:36:24,863 S1: way he described this you know what. Frick it. Just 653 00:36:24,863 --> 00:36:25,943 S1: open it. Watch this. 654 00:36:25,943 --> 00:36:28,103 S2: Uh, okay. So, um, I don't. 655 00:36:28,103 --> 00:36:29,753 S1: Know if you can hear that, but. So watch this. 656 00:36:29,783 --> 00:36:31,673 S2: I, uh, look at this. 657 00:36:31,793 --> 00:36:32,483 S1: Like, going back. 658 00:36:32,513 --> 00:36:37,103 S2: Going back to these, uh, to these various patterns in fabric, um, like, 659 00:36:37,133 --> 00:36:39,713 S2: if you're if you're like, if you have some text 660 00:36:39,713 --> 00:36:42,083 S2: in your selection, like if you, if you have something, 661 00:36:42,113 --> 00:36:44,363 S2: you want something or you want a quick Paul Graham essay, 662 00:36:44,393 --> 00:36:47,423 S2: I can just go ahead and just type in extract wisdom. Right. 663 00:36:47,453 --> 00:36:50,363 S2: And let's say I can grab this one. So he 664 00:36:50,393 --> 00:36:52,253 S2: copied or I can I can grab the whole thing, 665 00:36:52,283 --> 00:36:55,073 S2: grabs the whole thing. Wisdom right. Extract wisdom. And then just. 666 00:36:55,073 --> 00:36:56,663 S1: Typed X and then paste. 667 00:36:56,663 --> 00:36:57,293 S2: It and it will run. 668 00:36:57,293 --> 00:36:57,863 S1: And look at this. 669 00:36:57,863 --> 00:37:00,923 S2: Here comes fabric That fabric prompt, I use it to 670 00:37:00,953 --> 00:37:05,453 S2: run it with a clod. Sonnet 3.5. And it's insane. 671 00:37:05,573 --> 00:37:09,533 S1: So yeah, I'm about to rig everything up in that way. 672 00:37:09,563 --> 00:37:12,533 S1: Eric Schmidt of Google did a crazy, honest interview at Stanford, 673 00:37:12,533 --> 00:37:14,693 S1: and it was so spicy that Stanford took it down. 674 00:37:14,693 --> 00:37:17,123 S1: But guess what? I got the link to the video, 675 00:37:17,123 --> 00:37:19,463 S1: and I pulled out the transcript and I made a 676 00:37:19,463 --> 00:37:22,523 S1: fabric summary of it. So there you go. Ideal founding 677 00:37:22,553 --> 00:37:25,493 S1: team Ben Horowitz lays out the perfect founding team in 678 00:37:25,493 --> 00:37:28,463 S1: the clearest way I've ever seen. It's just it's really good. 679 00:37:28,463 --> 00:37:32,693 S1: Scrape it now. New CLI tool for web scraping grok two. 680 00:37:32,723 --> 00:37:34,943 S1: This is the one that just does all sorts of 681 00:37:34,973 --> 00:37:39,083 S1: crazy stuff. Prompt caching with Clod flux AI by Black 682 00:37:39,083 --> 00:37:41,783 S1: Forest Labs. This one getting a lot of talk. I 683 00:37:41,783 --> 00:37:44,003 S1: haven't used it yet because I'm not doing a lot 684 00:37:44,003 --> 00:37:46,973 S1: of AI image stuff. When I need an AI image, 685 00:37:46,973 --> 00:37:50,333 S1: I usually use Midjourney. Still graphic info. New website lets 686 00:37:50,333 --> 00:37:54,113 S1: you generate infographics to make your articles more engaging. Agile 687 00:37:54,113 --> 00:37:56,783 S1: is for losers, a rant about the author's decade long 688 00:37:56,813 --> 00:38:01,283 S1: frustrations with the agile methodology infiltrating digital agencies and the 689 00:38:01,283 --> 00:38:05,153 S1: recommendation of the week. Stop accepting it when your loved ones, 690 00:38:05,153 --> 00:38:08,333 S1: especially the young people, are not AI literate. Here's the 691 00:38:08,333 --> 00:38:10,373 S1: way I recommend you think about this. Imagine that the 692 00:38:10,373 --> 00:38:13,883 S1: competition level for getting top jobs or mates or whatever 693 00:38:13,883 --> 00:38:18,083 S1: was 100 in 2022. The average person was at like 694 00:38:18,113 --> 00:38:22,893 S1: an 80. Okay. AI is augmentation technology. It adds 20 695 00:38:22,893 --> 00:38:25,433 S1: to 50 points to people who get good at it. 696 00:38:25,433 --> 00:38:28,493 S1: So now the person with an 85 learns AI and 697 00:38:28,493 --> 00:38:32,273 S1: they're like a 125. So the new standard is reset 698 00:38:32,273 --> 00:38:35,723 S1: from 80 to 120. So if you were a 90 699 00:38:35,753 --> 00:38:39,293 S1: before or a 110, you are now behind. Don't let 700 00:38:39,293 --> 00:38:43,523 S1: your people get left behind. AI is the new reading. 701 00:38:43,523 --> 00:38:46,373 S1: It's the new high school diploma. It's the new degree. 702 00:38:46,403 --> 00:38:50,123 S1: Make sure your people, the people that you love, have it. 703 00:38:50,123 --> 00:38:52,763 S1: And just to show you how real this thing is. Okay, 704 00:38:52,793 --> 00:38:55,613 S1: this is the best. This is the absolute best. And 705 00:38:55,613 --> 00:38:59,013 S1: to get you motivated here is an 18. No an 706 00:38:59,043 --> 00:39:02,763 S1: eight year old. Here's an eight year old doing live coding. 707 00:39:02,763 --> 00:39:04,833 S1: Watch this thing every one. 708 00:39:04,833 --> 00:39:10,113 S3: And today I'm well I'm. Yeah I'm I'm going to 709 00:39:10,143 --> 00:39:15,123 S3: show you how to code with cursor. I'm going to 710 00:39:15,123 --> 00:39:20,103 S3: go to terminal. I'm going to say new terminal and 711 00:39:20,133 --> 00:39:21,093 S3: new terminal. 712 00:39:21,123 --> 00:39:23,463 S1: Look at this. She's she's running. 713 00:39:23,493 --> 00:39:27,513 S3: Npm run npm. 714 00:39:27,513 --> 00:39:28,563 S1: Run dev there. 715 00:39:29,133 --> 00:39:32,193 S3: And I hit that. And then I go to be 716 00:39:32,223 --> 00:39:35,853 S3: and look I made it say hello I want this 717 00:39:35,853 --> 00:39:39,123 S3: to say say hey. 718 00:39:39,153 --> 00:39:40,053 S1: Okay. So we're gonna. 719 00:39:40,353 --> 00:39:40,683 S3: We're gonna. 720 00:39:40,683 --> 00:39:42,123 S1: Jump ahead here. 721 00:39:43,443 --> 00:39:47,793 S3: So, um, I want it to say chat with Harry 722 00:39:47,793 --> 00:39:51,573 S3: and stuff, and then maybe some Cloudflare workers I. And 723 00:39:51,573 --> 00:39:55,563 S3: then it will do all of the stuff. So you 724 00:39:55,563 --> 00:39:58,713 S3: can't do it if it's not finished, because then it 725 00:39:58,713 --> 00:40:05,493 S3: just won't work. So now I'm gonna save it. So 726 00:40:05,493 --> 00:40:10,293 S3: now I'm gonna say hi. Wicked. Hi there. I'm Harry 727 00:40:10,323 --> 00:40:14,223 S3: James Potter. Oh. 728 00:40:14,643 --> 00:40:17,313 S1: My goodness. She just built a chat bot to talk 729 00:40:17,313 --> 00:40:19,473 S1: to Harry Potter. She's eight years. 730 00:40:19,473 --> 00:40:20,703 S3: Old, right? 731 00:40:20,733 --> 00:40:23,643 S1: Okay, I'm telling you, that is an eight year old 732 00:40:23,643 --> 00:40:26,913 S1: who just built a chat bot to talk to Harry Potter. Okay? 733 00:40:26,943 --> 00:40:29,463 S1: And you and I have a lot of loved ones 734 00:40:29,463 --> 00:40:32,103 S1: who are like, oh, ChatGPT, is that still a thing? 735 00:40:32,103 --> 00:40:34,113 S1: How is that still a thing? I thought that blew 736 00:40:34,143 --> 00:40:37,893 S1: out like a year ago. I thought it got really important. 737 00:40:37,893 --> 00:40:40,413 S1: Then people start, stopped talking about it, and now I 738 00:40:40,413 --> 00:40:43,533 S1: don't care anymore. This is what you're competing with. Eight 739 00:40:43,563 --> 00:40:46,233 S1: year olds, building chat bots to talk to Harry Potter. 740 00:40:46,263 --> 00:40:49,143 S1: All right. And the aphorism of the week standing still 741 00:40:49,143 --> 00:40:53,073 S1: in evolution is equivalent to moving backwards. Standing still in 742 00:40:53,073 --> 00:40:57,063 S1: evolution is equivalent to moving backwards Matt Ridley.