1 00:00:01,129 --> 00:00:04,490 S1: Welcome to Unsupervised Learning, a security, AI and meaning focused 2 00:00:04,490 --> 00:00:07,370 S1: podcast that looks at how best to thrive as humans 3 00:00:07,370 --> 00:00:11,570 S1: in a post AI world. It combines original ideas, analysis, 4 00:00:11,570 --> 00:00:14,810 S1: and mental models to bring not just the news, but 5 00:00:14,810 --> 00:00:21,980 S1: why it matters and how to respond. All right. Welcome 6 00:00:21,980 --> 00:00:25,100 S1: to unsupervised learning. This is Daniel Meisler. This episode is. 7 00:00:25,100 --> 00:00:30,710 S1: Can you articulate yourself in 50 words? Let's jump in here. Yeah. 8 00:00:30,710 --> 00:00:33,140 S1: One of the most important things for surviving AI is 9 00:00:33,140 --> 00:00:36,560 S1: being able to articulate what you do, or even better, 10 00:00:36,560 --> 00:00:39,290 S1: who you are. So I was at a party with 11 00:00:39,290 --> 00:00:42,410 S1: a bunch of like, high net worth people, like muckety 12 00:00:42,409 --> 00:00:45,290 S1: mucks like important people. And somebody asked me, I was 13 00:00:45,290 --> 00:00:48,290 S1: at RSA, so people have asked me this like 20 14 00:00:48,290 --> 00:00:50,360 S1: times over the course of the week. They asked me 15 00:00:50,360 --> 00:00:52,190 S1: what I did and I told them I built AI 16 00:00:52,190 --> 00:00:55,850 S1: stuff and they just looked at me like, oh, you're 17 00:00:55,850 --> 00:00:58,790 S1: an asshole. Basically, like I was trying to be cool. 18 00:00:58,790 --> 00:01:01,430 S1: And it's like, I am seriously not trying to be cool. 19 00:01:01,430 --> 00:01:04,490 S1: I'm trying to say a small amount, which maybe spawns 20 00:01:04,490 --> 00:01:07,399 S1: a conversation if they're like, oh yeah, really? What kind 21 00:01:07,400 --> 00:01:09,350 S1: of AI or something? But I don't want to go 22 00:01:09,350 --> 00:01:11,690 S1: into some sales pitch of like, oh, I'm building AI 23 00:01:11,690 --> 00:01:13,760 S1: that blah blah, blah blah does this. But it was 24 00:01:13,760 --> 00:01:16,850 S1: just like not landing. It didn't land for me. It 25 00:01:16,850 --> 00:01:19,250 S1: didn't land for them. It just wasn't good. So I 26 00:01:19,250 --> 00:01:22,790 S1: think I'm basically going to answer differently from now on. 27 00:01:22,790 --> 00:01:25,460 S1: I'm going to start with the problems. So I'm basically 28 00:01:25,459 --> 00:01:28,009 S1: worried about two problems people having a lack of meaning 29 00:01:28,010 --> 00:01:32,240 S1: in their lives and AI making that worse when their 30 00:01:32,240 --> 00:01:34,700 S1: jobs go away. So what I do is I use 31 00:01:34,700 --> 00:01:37,460 S1: AI to build products and services that help people and 32 00:01:37,459 --> 00:01:41,360 S1: companies create a version of themselves that will not just survive, 33 00:01:41,390 --> 00:01:45,620 S1: but will thrive after AI is everywhere. And I could 34 00:01:45,620 --> 00:01:48,080 S1: say that in multiple different ways, but it starts with 35 00:01:48,080 --> 00:01:51,110 S1: I'm worried about two problems, so I'm building these things 36 00:01:51,110 --> 00:01:53,780 S1: to solve them. So that's basically like my new spiel. 37 00:01:53,780 --> 00:01:56,450 S1: And I like this for a few different reasons. One, 38 00:01:56,450 --> 00:01:59,960 S1: it starts with the problems. Two, it highlights that I'm building. 39 00:01:59,960 --> 00:02:03,260 S1: It mentions AI, which is good to mention, and it 40 00:02:03,470 --> 00:02:06,680 S1: is actually what I'm doing, and it puts meaning as 41 00:02:06,680 --> 00:02:09,380 S1: the most important piece even above AI. So I like 42 00:02:09,380 --> 00:02:12,080 S1: it for all these different reasons. But I think the 43 00:02:12,080 --> 00:02:16,880 S1: most screwed people right now are those who don't know themselves, right? 44 00:02:16,880 --> 00:02:19,640 S1: They don't know what they're about, they don't have a 45 00:02:19,639 --> 00:02:22,520 S1: purpose they can't articulate, like why they get up in 46 00:02:22,520 --> 00:02:25,130 S1: the morning and what they're driving towards, or why they 47 00:02:25,130 --> 00:02:27,980 S1: do the job that they're doing other than, well, I 48 00:02:27,980 --> 00:02:31,190 S1: it's just the job I have or whatever. So my 49 00:02:31,190 --> 00:02:35,060 S1: recommendation know yourself and be able to articulate that thing. 50 00:02:35,060 --> 00:02:39,890 S1: So I ran this through the create five sentence summary 51 00:02:39,889 --> 00:02:43,130 S1: fabric pattern. And it ended up with this helping people 52 00:02:43,130 --> 00:02:47,600 S1: find meaning in the AI era. Empowering human potential with AI, 53 00:02:47,630 --> 00:02:52,880 S1: AI for human flourishing, meaningful AI solutions, and AI powered meaning. 54 00:02:52,880 --> 00:02:56,000 S1: Not too bad. So my work network Chuck, my buddy 55 00:02:56,000 --> 00:03:00,530 S1: Chuck here just posted the most amazing video about fabric. 56 00:03:00,530 --> 00:03:04,640 S1: It was just absolutely, insanely good. I've been waiting for 57 00:03:04,639 --> 00:03:06,380 S1: this thing to drop for a couple of months now, 58 00:03:06,380 --> 00:03:11,329 S1: and so he basically captures this spirit of the project 59 00:03:11,330 --> 00:03:15,260 S1: in such a fantastic way. He just really gets the 60 00:03:15,260 --> 00:03:18,080 S1: concept of it how it's human. First, he mentions it 61 00:03:18,080 --> 00:03:21,980 S1: multiple times, he pulls multiple quotes out. He's got us 62 00:03:21,980 --> 00:03:25,579 S1: talking about it in an interactive way there. It's just 63 00:03:25,580 --> 00:03:28,400 S1: unbelievably good. And it's a pretty long video. And he 64 00:03:28,400 --> 00:03:32,419 S1: walks through all sorts of use cases. Highly recommend that one. Security. 65 00:03:32,419 --> 00:03:34,549 S1: Looks like we're going to have a cyber force similar 66 00:03:34,550 --> 00:03:38,150 S1: to the Space Force. Chinese linked hackers are using things 67 00:03:38,150 --> 00:03:42,020 S1: called orbs, which are basically these hop networks that they 68 00:03:42,020 --> 00:03:44,960 S1: can bounce through for hiding, spying, and a bunch of 69 00:03:44,960 --> 00:03:50,420 S1: other shenanigans. Fixed problems instead of doing risk measurement theatre. 70 00:03:50,420 --> 00:03:53,270 S1: I came up with that title, but Andy Ellis wrote 71 00:03:53,270 --> 00:03:58,130 S1: this piece. Quite cool. This piece actually. Short. Sweet. Yeah. 72 00:03:58,160 --> 00:04:01,490 S1: Really good. Highly recommend it. Run some AI against it 73 00:04:01,490 --> 00:04:04,010 S1: if you don't want to read that. But yeah, really good. 74 00:04:04,130 --> 00:04:10,610 S1: Got a bug cloud? No, a cloud related bug. Linguistic lumberjack. 75 00:04:10,640 --> 00:04:14,930 S1: Pretty cool alliteration there. Issue with Fluentbit. And basically it's 76 00:04:14,930 --> 00:04:18,590 S1: got 3 billion downloads and it's yeah all sorts of systems. 77 00:04:18,589 --> 00:04:23,120 S1: So another supply chain issue American criminal records exposed 70 78 00:04:23,120 --> 00:04:28,580 S1: million rows of American criminal record in a database somewhere. 79 00:04:28,850 --> 00:04:33,740 S1: So Russia evidently launched some sort of space weapon capable 80 00:04:33,740 --> 00:04:36,470 S1: of attacking satellites. And they're saying that it's close to 81 00:04:36,470 --> 00:04:39,830 S1: a particular one, same orbit as a US government satellite 82 00:04:39,830 --> 00:04:42,740 S1: posing a direct threat. And of course, Russian said Russia 83 00:04:42,740 --> 00:04:45,710 S1: said fake news. Not true, which I believe. And why 84 00:04:45,710 --> 00:04:48,500 S1: wouldn't you believe him? Like why would they lie? That's 85 00:04:48,500 --> 00:04:52,820 S1: the thing. I breaks into a tech giant. So X-Force 86 00:04:52,820 --> 00:04:56,300 S1: from IBM, they've been around forever. They've been around like 87 00:04:56,300 --> 00:04:58,729 S1: they were around when I got started, I think. And 88 00:04:58,730 --> 00:05:00,919 S1: they got a new AI hacking platform, and they use 89 00:05:00,920 --> 00:05:04,430 S1: it to attack some major tech manufacturers network, and they 90 00:05:04,430 --> 00:05:07,310 S1: got in around eight hours. Question is, how much human 91 00:05:07,310 --> 00:05:10,460 S1: involvement was there or was it like completely automated? That 92 00:05:10,460 --> 00:05:13,520 S1: would be impressive. They haven't released this tool yet, but 93 00:05:13,520 --> 00:05:16,400 S1: eager to hear more about it. Tesla's can still be hacked. 94 00:05:16,400 --> 00:05:20,549 S1: I believe these were model threes, but. A buddy, Malwaretech 95 00:05:20,550 --> 00:05:24,839 S1: Marcus Hutchins did a demo of this on his TikTok. 96 00:05:24,870 --> 00:05:27,599 S1: Pretty easy to get into these cars still, even though 97 00:05:27,600 --> 00:05:30,210 S1: they switched to wideband to try to get around this, 98 00:05:30,210 --> 00:05:33,060 S1: I believe. But yeah, it still works. Not sure if 99 00:05:33,060 --> 00:05:37,050 S1: it works on model wise. Preliminary report on the Baltimore 100 00:05:37,050 --> 00:05:42,540 S1: Bridge looks like unexpected structural weaknesses. So not cyber like 101 00:05:42,540 --> 00:05:46,830 S1: so many people said immediately. Core issue with Wi-Fi with 102 00:05:46,830 --> 00:05:51,090 S1: lack of Ssid authentication during beaconing. Okay, got a US 103 00:05:51,089 --> 00:05:54,060 S1: woman and four others charged in a scheme using fake 104 00:05:54,060 --> 00:05:57,450 S1: IT jobs to get lots of money into North Korea's 105 00:05:57,450 --> 00:06:03,060 S1: nuclear program. 60 identities and 300 companies US companies. Insane. 106 00:06:03,330 --> 00:06:07,050 S1: NYPD is rolling out drones to respond to 911 calls. 107 00:06:07,080 --> 00:06:09,660 S1: I am happy about this, but only because I like 108 00:06:09,660 --> 00:06:14,190 S1: sci fi technology. You. The EU passed its AI act, 109 00:06:14,190 --> 00:06:16,650 S1: and I've got a summary here. I'm not going to 110 00:06:16,650 --> 00:06:19,800 S1: read all ten of these points, but the fines up 111 00:06:19,800 --> 00:06:22,919 S1: to 7% of worldwide turnover, that's a pretty hefty fine. 112 00:06:23,040 --> 00:06:27,870 S1: Probably not going to be ignored as much as other regulation. Oh, 113 00:06:27,870 --> 00:06:30,330 S1: the other key point, because I read a whole bunch 114 00:06:30,330 --> 00:06:34,380 S1: of this thing, was that they tried to support innovation, 115 00:06:34,380 --> 00:06:37,380 S1: so they didn't go like crazy Europe on it with 116 00:06:37,380 --> 00:06:41,310 S1: like just more and more regulation. They actually have a 117 00:06:41,310 --> 00:06:44,700 S1: regulation in here that says you can't impose other regulation. 118 00:06:44,700 --> 00:06:48,000 S1: So they wanted to avoid this getting out of control, 119 00:06:48,000 --> 00:06:50,969 S1: which I thought was smart of them. And when your 120 00:06:50,970 --> 00:06:55,050 S1: AI sounds too much like Scarlett Johansson. So I have 121 00:06:55,050 --> 00:06:58,920 S1: opinions on this. So basically Sam tweeted out her, okay, 122 00:06:58,920 --> 00:07:01,080 S1: so it was pretty obvious that they wanted this thing 123 00:07:01,110 --> 00:07:03,419 S1: to sound like the movie, right? So they reached out 124 00:07:03,420 --> 00:07:06,690 S1: to a voice actress who sounded like Scarlett, and then 125 00:07:06,690 --> 00:07:10,920 S1: they reached out to actual Scarlett Johansson. Real Scarlett Johansson 126 00:07:10,920 --> 00:07:12,630 S1: said no. So they're like, okay, we're going to go 127 00:07:12,630 --> 00:07:14,700 S1: with the other one. They already I don't know if 128 00:07:14,700 --> 00:07:17,070 S1: they had tweeted out her before or after that, but 129 00:07:17,070 --> 00:07:19,920 S1: it was obviously they were going in that direction. I 130 00:07:19,920 --> 00:07:22,410 S1: don't think this is horribly malicious. Like if you've ever 131 00:07:22,410 --> 00:07:24,630 S1: worked inside of one of these companies, especially at like 132 00:07:24,630 --> 00:07:28,200 S1: the leadership level, there is mass chaos. You might have 133 00:07:28,200 --> 00:07:30,660 S1: marketing talking about something. You might have someone on your 134 00:07:30,660 --> 00:07:34,530 S1: leadership team who, like, was approved to say something. But 135 00:07:34,530 --> 00:07:37,170 S1: they said the thing that was different than the actual message. 136 00:07:37,170 --> 00:07:40,470 S1: There could be miscommunication here. So yeah, I don't think 137 00:07:40,470 --> 00:07:44,100 S1: it's malicious, but somebody was bringing up. Look, it's not 138 00:07:44,100 --> 00:07:48,060 S1: Scarlett Johansson. You can hire somebody who sounds like somebody else, 139 00:07:48,060 --> 00:07:50,280 S1: no big deal. And I had a tweet somewhere. It 140 00:07:50,280 --> 00:07:54,090 S1: was like, you're not allowed to deepfake people using real humans. 141 00:07:54,090 --> 00:07:57,540 S1: That's still a deepfake. Not technically. It's not technically a 142 00:07:57,540 --> 00:07:59,880 S1: deepfake because it's not AI. It's a real person. But 143 00:07:59,880 --> 00:08:02,400 S1: the point is to copy someone else. And I'm sure 144 00:08:02,400 --> 00:08:05,400 S1: there's precedent for this all over the place. Already on 145 00:08:05,400 --> 00:08:08,280 S1: the books. It's like if I hire somebody who looks 146 00:08:08,280 --> 00:08:13,620 S1: exactly like whoever, um, the Rock, Dwayne Johnson or whatever, 147 00:08:13,620 --> 00:08:17,640 S1: and I'm like, yeah, I'm claiming to be him or no, 148 00:08:17,640 --> 00:08:20,550 S1: I just call the app the Stone, and it's like 149 00:08:20,550 --> 00:08:22,710 S1: the Rock's voice, but it's like, no, I didn't know 150 00:08:22,710 --> 00:08:24,690 S1: it was the Rock, and I. What are you talking about? 151 00:08:24,690 --> 00:08:26,550 S1: I don't even know who the rock is. But you 152 00:08:26,550 --> 00:08:28,650 S1: hire a person who sounds just like him, and the 153 00:08:28,650 --> 00:08:31,860 S1: app is called the Stone. Then you tweet out, rocks 154 00:08:31,860 --> 00:08:34,950 S1: are cool, that come on, there's got to be a 155 00:08:34,950 --> 00:08:37,650 S1: precedent for that. So they're going to get smashed if 156 00:08:37,650 --> 00:08:40,800 S1: that does go to court. I wish it just wouldn't. Right? 157 00:08:40,800 --> 00:08:43,890 S1: I mean, I'm trying to look at positives here. Like 158 00:08:43,890 --> 00:08:46,830 S1: Sam is trying to do a cool thing. Okay. It's 159 00:08:46,830 --> 00:08:50,010 S1: like her is awesome, I love her. I've got a 160 00:08:50,010 --> 00:08:53,250 S1: video about this where I'm like, hey, look, you could 161 00:08:53,250 --> 00:08:57,630 S1: talk to Samantha from the movie her and I'm breaking down, 162 00:08:57,630 --> 00:08:59,729 S1: like how you can use that with Siri and Shortcuts 163 00:08:59,730 --> 00:09:03,420 S1: and everything. So it's like so obvious and it's so cool. 164 00:09:03,420 --> 00:09:06,210 S1: And it's also cool for Scarlett to be able to say, yeah, 165 00:09:06,210 --> 00:09:09,809 S1: you know what? Not me. Can't do that. So anyway, 166 00:09:09,809 --> 00:09:12,480 S1: it's a legal issue. I'm not sure what the actual 167 00:09:12,480 --> 00:09:16,140 S1: precedent says, but yeah, hiring someone who sounds just like 168 00:09:16,140 --> 00:09:19,949 S1: Scarlett and pretending it didn't happen, that's not the game. 169 00:09:20,130 --> 00:09:23,069 S1: Looks like Apple tried to get all of TSMC's two 170 00:09:23,070 --> 00:09:30,420 S1: nanometer chips stock skyrocketing 262%, $26 billion. They're about to 171 00:09:30,420 --> 00:09:34,109 S1: be worth more than Apple, which is ridiculous. And I 172 00:09:34,110 --> 00:09:37,590 S1: think Scott Galloway was quoting some stats of like, it's 173 00:09:37,590 --> 00:09:40,110 S1: Nvidia at this point is worth more than like the 174 00:09:40,110 --> 00:09:44,819 S1: economy of Germany or something. Just ridiculous numbers. And I 175 00:09:44,820 --> 00:09:47,340 S1: think we're just starting with AI, so I don't see 176 00:09:47,340 --> 00:09:51,240 S1: how someone's going to catch up before, who knows how 177 00:09:51,240 --> 00:09:53,280 S1: high Nvidia is going to go, but I feel like 178 00:09:53,280 --> 00:09:55,560 S1: they got a massive lead and everyone's going to be 179 00:09:55,559 --> 00:09:58,800 S1: trying to get in. AI is just starting. People are 180 00:09:58,800 --> 00:10:02,880 S1: talking about plateaus. I think they're absolutely insane. So yeah, 181 00:10:02,880 --> 00:10:06,480 S1: exciting times. Looks like we found some lithium in the US. 182 00:10:06,750 --> 00:10:11,610 S1: Massive untapped lithium source Microsoft is reimagining windows with a 183 00:10:11,610 --> 00:10:15,900 S1: deep dive into AI, introducing copilot NPCs that blend cutting 184 00:10:15,900 --> 00:10:19,800 S1: edge AI features directly into the operating system. People are saying, hey, 185 00:10:19,800 --> 00:10:24,179 S1: nobody wants this. Well, yeah, they actually do. Once everyone 186 00:10:24,179 --> 00:10:28,350 S1: has the ability to remember everything they've done on their computer, 187 00:10:28,350 --> 00:10:30,840 S1: remember every screen, and they could just talk to their 188 00:10:30,840 --> 00:10:33,930 S1: computer and be like, hey, where was that one email? Hey, 189 00:10:33,929 --> 00:10:36,900 S1: am I missing a thing? Hey, keep me up to date. Hey, 190 00:10:36,900 --> 00:10:38,940 S1: make a list of this. Hey, send a list of 191 00:10:38,940 --> 00:10:41,939 S1: this to all my friends. Hey, show me the most 192 00:10:41,940 --> 00:10:45,060 S1: important things I should be taking a look at. Hey, 193 00:10:45,059 --> 00:10:49,110 S1: reprioritize my work. Hey, rewrite this document for me. And 194 00:10:49,110 --> 00:10:51,959 S1: when I say rewrite this document for me, it knows 195 00:10:51,960 --> 00:10:54,570 S1: that I'm looking at the screen and it knows what 196 00:10:54,570 --> 00:10:57,510 S1: document I'm looking at. So it just rewrites it and 197 00:10:57,510 --> 00:10:59,910 S1: it's like, ah, not so much. Make it less formal. 198 00:10:59,910 --> 00:11:03,630 S1: Boom rewrites it less formal. Okay, send it out to 199 00:11:03,630 --> 00:11:06,930 S1: all my directs or whatever. Boom sends it out. Once 200 00:11:06,929 --> 00:11:09,270 S1: you're interacting with a computer like that and you have 201 00:11:09,270 --> 00:11:12,180 S1: all of your history available, this is like I said 202 00:11:12,179 --> 00:11:15,540 S1: in my AI prediction video, everyone's going to want this, 203 00:11:15,540 --> 00:11:18,930 S1: and they're going to demand it. And what it requires 204 00:11:18,950 --> 00:11:22,820 S1: is everything being recorded all the time, all your screens, 205 00:11:22,820 --> 00:11:26,960 S1: all your conversations. Trust me, it's happening. It's going to happen. 206 00:11:26,960 --> 00:11:29,089 S1: There's nothing that's going to stop that. There's nothing's going 207 00:11:29,090 --> 00:11:33,020 S1: to slow it down. It is too powerful and too awesome. Now, 208 00:11:33,020 --> 00:11:38,569 S1: will there be security issues? Oh yes. So many security issues. 209 00:11:38,570 --> 00:11:41,420 S1: And they will be horrible because it will be your 210 00:11:41,420 --> 00:11:45,679 S1: whole entire soul getting compromised when that stuff gets leaked. Right. 211 00:11:45,679 --> 00:11:48,200 S1: It's every text message that came through everything. A little 212 00:11:48,200 --> 00:11:51,470 S1: pop pop up, someone sent you a dirty joke or whatever, 213 00:11:51,470 --> 00:11:54,470 S1: or you sent somebody else a dirty joke that's also recorded. 214 00:11:54,470 --> 00:11:57,770 S1: That's also captured by AI. That's somewhere. Now it's in 215 00:11:57,770 --> 00:12:01,400 S1: some law document somewhere. Because guess what? Lawyers get access 216 00:12:01,400 --> 00:12:03,860 S1: to that and everything. I mean, it's going to be 217 00:12:03,860 --> 00:12:07,190 S1: crazy world we're about to move into. But I tell you, 218 00:12:07,190 --> 00:12:09,589 S1: the downsides are not going to slow it down in 219 00:12:09,590 --> 00:12:13,219 S1: the slightest because the upsides are going to be so powerful. 220 00:12:13,490 --> 00:12:17,330 S1: Mapping AI's inner universe. This is really cool. Anthropic released 221 00:12:17,330 --> 00:12:22,430 S1: a paper on sonnet, which is their medium level model, 222 00:12:22,429 --> 00:12:26,569 S1: and basically showed how it thought essentially basically a lot 223 00:12:26,570 --> 00:12:30,350 S1: more transparency for AI, which has been traditionally opaque with 224 00:12:30,350 --> 00:12:35,060 S1: neural nets. China released a chatbot called Ask President XI, 225 00:12:35,420 --> 00:12:39,679 S1: and it allows citizens to ask Prachi how to be 226 00:12:39,679 --> 00:12:43,339 S1: better socialists. Fantastic. I bet that's flying off the shelf. 227 00:12:43,580 --> 00:12:47,449 S1: Astronomers are preparing to use AI to tackle 300PB of 228 00:12:47,450 --> 00:12:51,110 S1: data annually. Yeah. Watch this guy. Find some stuff. Make 229 00:12:51,110 --> 00:12:54,980 S1: sure rocks don't hit us. Windows 11 Powertoys allows you 230 00:12:54,980 --> 00:12:59,870 S1: to transform AI transform clipboard content into different formats or 231 00:12:59,870 --> 00:13:04,670 S1: languages with a simple shortcut. Love it and it uses OpenAI. Cool. 232 00:13:04,670 --> 00:13:08,750 S1: No problem. Microservices might be all the rage, but this 233 00:13:08,750 --> 00:13:13,339 S1: thing basically says use modules instead of microservices. Interesting. X 234 00:13:13,340 --> 00:13:15,830 S1: is getting rid of public likes, so you can't get 235 00:13:15,830 --> 00:13:20,120 S1: in trouble for liking somebody who's controversial. You'll still be 236 00:13:20,120 --> 00:13:22,010 S1: able to see when someone likes your stuff, and you'll 237 00:13:22,010 --> 00:13:24,260 S1: still be able to see your own likes, but you 238 00:13:24,260 --> 00:13:27,290 S1: won't be able to see other people's. You won't be 239 00:13:27,290 --> 00:13:30,350 S1: able to see someone who you hate and see a 240 00:13:30,350 --> 00:13:33,079 S1: third person who liked their stuff. That's what you won't 241 00:13:33,080 --> 00:13:36,050 S1: be able to see. And Elon basically thinks that this 242 00:13:36,050 --> 00:13:38,599 S1: is going to increase the number of likes. And I 243 00:13:38,600 --> 00:13:41,750 S1: think he's probably right about that. Focusing on fundamentals over 244 00:13:41,750 --> 00:13:45,620 S1: frameworks could be the secret sauce for long lasting tech skills. 245 00:13:45,620 --> 00:13:48,080 S1: Agree with this. This was a good essay. Shows you 246 00:13:48,080 --> 00:13:52,459 S1: how to use ghidra for crafting gdb scripts that trace 247 00:13:52,460 --> 00:13:58,430 S1: function calls helping with vulnerability, hunting D3 in-depth guides. Love 248 00:13:58,429 --> 00:14:03,740 S1: me some D3. Really powerful library experts JS simplifies creating 249 00:14:03,740 --> 00:14:07,970 S1: multi-agent AI systems. Starting to get pretty turned off by 250 00:14:08,030 --> 00:14:11,270 S1: AI agent systems. Honestly, I'm heading down the path of 251 00:14:11,270 --> 00:14:13,400 S1: just building all my own stuff, and that's what I'm 252 00:14:13,400 --> 00:14:16,730 S1: doing with substrate. So getting away from the agent frameworks, 253 00:14:16,730 --> 00:14:20,450 S1: I think just a lot of really cool stuff in 254 00:14:20,450 --> 00:14:24,410 S1: demos and not usable in production. That's what I'm finding. 255 00:14:24,410 --> 00:14:27,980 S1: And I think everyone's finding that who's actually building user 256 00:14:27,980 --> 00:14:33,590 S1: persona generator prompts simulates user interviews to create detailed user personas, 257 00:14:33,830 --> 00:14:38,600 S1: saving you from conducting dozens of real interviews. Yeah, wait 258 00:14:38,630 --> 00:14:43,010 S1: till you see what AI is going to hiring. Oh yeah, 259 00:14:43,010 --> 00:14:45,560 S1: that's what we're about to talk about. Job seekers are 260 00:14:45,560 --> 00:14:49,760 S1: struggling to find work and employers can't seem to fill positions. Paradox. 261 00:14:49,760 --> 00:14:53,240 S1: It is not a mystery. The problem is hiring managers 262 00:14:53,240 --> 00:14:55,640 S1: need people ready to do the job and they don't 263 00:14:55,640 --> 00:14:58,370 S1: want to train anyone. This is why degrees are mattering less, 264 00:14:58,370 --> 00:15:01,190 S1: and why somebody with an active tech blog or YouTube 265 00:15:01,190 --> 00:15:04,820 S1: channel who's actually building stuff and doing the thing can 266 00:15:04,820 --> 00:15:07,580 S1: get hired with no credentials over someone with a degree 267 00:15:07,580 --> 00:15:11,390 S1: in computer science, or two degrees, or a master's degree 268 00:15:11,390 --> 00:15:15,080 S1: or a PhD. The ability to actually do the thing 269 00:15:15,080 --> 00:15:18,770 S1: is all that matters now. Expect to see AI services soon. 270 00:15:19,240 --> 00:15:22,900 S1: Now look at a person's social media blog, YouTube, whatever, 271 00:15:22,900 --> 00:15:26,710 S1: and then do an interactive interview with them and give 272 00:15:26,710 --> 00:15:29,830 S1: a score of how likely they are to do a 273 00:15:29,830 --> 00:15:32,710 S1: good job at this job. This will cut through all 274 00:15:32,710 --> 00:15:35,320 S1: the credentials, all the fluff, all sorts of other BS 275 00:15:35,320 --> 00:15:38,260 S1: and get at the real thing. The value of traditional 276 00:15:38,260 --> 00:15:42,940 S1: education will drastically fall because of this. Like pay attention 277 00:15:42,940 --> 00:15:47,110 S1: to this. The value of traditional education will drastically fall 278 00:15:47,110 --> 00:15:50,980 S1: as a result of this, because degrees won't significantly raise 279 00:15:50,980 --> 00:15:55,780 S1: these scores for most people. This is something this controversial point. 280 00:15:55,780 --> 00:15:58,360 S1: I used to think that it was all about expose 281 00:15:58,360 --> 00:16:00,700 S1: people to the training and they will get better, and 282 00:16:00,700 --> 00:16:03,400 S1: you just need better training. If they're not moving, you 283 00:16:03,400 --> 00:16:07,120 S1: give them better training. Turns out that's not it. It's 284 00:16:07,120 --> 00:16:10,780 S1: actually something inside the person that if they are that 285 00:16:10,780 --> 00:16:14,560 S1: person who's going to be awesome, it doesn't take much 286 00:16:14,560 --> 00:16:20,320 S1: to prompt them or to stimulate them into their curiosity, 287 00:16:20,320 --> 00:16:24,700 S1: which just makes them a voracious reader and learner and studier. 288 00:16:24,700 --> 00:16:26,440 S1: And all of a sudden, boom, they shoot out of 289 00:16:26,440 --> 00:16:28,960 S1: the thing. That's why the help desk is so good 290 00:16:28,960 --> 00:16:33,370 S1: at finding talent, because it's a wide filter. There's so 291 00:16:33,370 --> 00:16:36,160 S1: many people on Help Desk, but there's somebody on help 292 00:16:36,160 --> 00:16:40,270 S1: desk who, the moment they sat on the tier one floor, 293 00:16:40,300 --> 00:16:44,680 S1: they immediately like started answering all the questions for everyone else. Boom! 294 00:16:44,680 --> 00:16:47,530 S1: They become tier two, boom. They become tier three. And 295 00:16:47,530 --> 00:16:50,410 S1: before long, if they have the right networking or whatever, 296 00:16:50,410 --> 00:16:52,990 S1: they go into networking, they go into sysadmin, or they 297 00:16:52,990 --> 00:16:56,110 S1: go into programming or they go into security or something, 298 00:16:56,110 --> 00:16:59,950 S1: and they shoot off and they become a star. And 299 00:16:59,950 --> 00:17:03,460 S1: an experienced person could have seen them at tier one 300 00:17:03,460 --> 00:17:05,980 S1: in tech and just been like, oh, that's a star 301 00:17:05,980 --> 00:17:10,700 S1: right there. First of that is that there are people 302 00:17:10,700 --> 00:17:14,119 S1: who come in with two degrees, go to tier one, 303 00:17:14,119 --> 00:17:17,990 S1: you send them to Harvard, the best training, the best everything, 304 00:17:17,990 --> 00:17:21,530 S1: and they can't even move into tier two. They're there 305 00:17:21,530 --> 00:17:24,140 S1: for six years. And when they see a problem that 306 00:17:24,140 --> 00:17:26,659 S1: they've seen a million times, they call their manager or 307 00:17:26,660 --> 00:17:28,970 S1: they ask the person next to them, and that is 308 00:17:28,970 --> 00:17:32,150 S1: a function of them versus the other person. Now. And 309 00:17:32,150 --> 00:17:35,149 S1: there's lots of people in between what this is going 310 00:17:35,150 --> 00:17:38,480 S1: to do. And this is here's two extremes here. Lots 311 00:17:38,480 --> 00:17:43,550 S1: of education, little ability to do the thing practically and 312 00:17:43,550 --> 00:17:48,649 S1: technically versus somebody who doesn't have much training at all. Right. 313 00:17:48,680 --> 00:17:52,460 S1: No exposure to the stuff, no networking, no connections. And 314 00:17:52,460 --> 00:17:55,520 S1: they're just voracious. They just what? Oh, how do you 315 00:17:55,520 --> 00:17:57,380 S1: do that? Whoa whoa whoa whoa. And they just jump 316 00:17:57,380 --> 00:17:59,930 S1: into it and jump all over it. And they learn 317 00:17:59,930 --> 00:18:02,570 S1: the stuff instantly and they're like, oh, you can just 318 00:18:02,570 --> 00:18:06,050 S1: code that. Oh, well, what's the name of that Python? Oh, okay. 319 00:18:06,050 --> 00:18:08,689 S1: I guess I'll go learn Python. They come back, they're like, hey, 320 00:18:08,690 --> 00:18:11,240 S1: I wrote a program to do that. It's like, when 321 00:18:11,240 --> 00:18:13,610 S1: did you learn Python? Oh, I didn't, I just looked on. 322 00:18:13,609 --> 00:18:16,010 S1: I watched a couple of YouTube videos. I wrote a program. 323 00:18:16,010 --> 00:18:18,439 S1: Some people are like that and some people are the 324 00:18:18,440 --> 00:18:22,609 S1: opposite of that. This is what I is good at. 325 00:18:22,609 --> 00:18:24,890 S1: It's going to be able to tell you the difference 326 00:18:24,890 --> 00:18:27,680 S1: between those. So guess what? You're going to be able 327 00:18:27,680 --> 00:18:34,340 S1: to filter through tens, hundreds, thousands, tens of thousands of resumes, 328 00:18:34,340 --> 00:18:38,090 S1: find people like that, send them into an additional filter, 329 00:18:38,090 --> 00:18:41,090 S1: which is like a live interview with like an AI 330 00:18:41,090 --> 00:18:44,930 S1: avatar or whatever. It runs them through scenarios, sees how 331 00:18:44,930 --> 00:18:48,080 S1: curious they are. It's going to find the magic sauce, 332 00:18:48,080 --> 00:18:50,390 S1: the same magic sauce that I'm talking about right now. 333 00:18:50,390 --> 00:18:53,510 S1: The AI is going to find that magic sauce. It's 334 00:18:53,510 --> 00:18:55,340 S1: going to give them a score, and they're going to 335 00:18:55,340 --> 00:18:57,560 S1: shoot to the top. And guess what? Everyone's going to 336 00:18:57,560 --> 00:19:02,120 S1: go hire those people. Nobody wants to hire a new person. 337 00:19:02,119 --> 00:19:05,210 S1: Nobody wants to train new people. This is why there 338 00:19:05,210 --> 00:19:08,090 S1: is the gap in hiring. This is why so many 339 00:19:08,090 --> 00:19:10,130 S1: people are getting out of college, can't find a job. 340 00:19:10,130 --> 00:19:13,399 S1: It's because they're like, hey, look, I've got my degrees. 341 00:19:13,670 --> 00:19:15,710 S1: Can't wait to do my training so you can teach 342 00:19:15,710 --> 00:19:18,080 S1: me how to do my job. And the hiring manager 343 00:19:18,080 --> 00:19:20,870 S1: is like, yeah, so about that, we're busy. We don't 344 00:19:20,869 --> 00:19:25,370 S1: have time to train anybody, so we won't be hiring you. 345 00:19:25,400 --> 00:19:28,940 S1: We will be hiring Caleb, who is a 15 year 346 00:19:28,940 --> 00:19:32,929 S1: old who is doing this job already and has been 347 00:19:32,930 --> 00:19:36,080 S1: for three years in between playing video games. And that's 348 00:19:36,080 --> 00:19:37,610 S1: who we're going to hire instead of you with your 349 00:19:37,609 --> 00:19:43,639 S1: master's degrees. So it'll filter for drive, curiosity, obsession, and 350 00:19:43,640 --> 00:19:47,420 S1: proof of work like actual builders and coders and people 351 00:19:47,420 --> 00:19:50,450 S1: who raise insightful questions that are curious. This is going 352 00:19:50,450 --> 00:19:54,110 S1: to spawn a completely new type of education dedicated to 353 00:19:54,109 --> 00:19:58,220 S1: making more people like the stars that I'm talking about, 354 00:19:58,220 --> 00:20:00,590 S1: because a lot of the people who are dull in 355 00:20:00,590 --> 00:20:02,990 S1: this way that I'm talking about, they were just let 356 00:20:02,990 --> 00:20:07,520 S1: down by their maybe it's genetic, maybe they're just not 357 00:20:07,520 --> 00:20:11,090 S1: as smart. Of course, there are smarter, less smart people, 358 00:20:11,090 --> 00:20:14,150 S1: but I think a lot of this is cultural where 359 00:20:14,150 --> 00:20:18,080 S1: you have they were just not given curiosity as a 360 00:20:18,080 --> 00:20:21,110 S1: thing as they were growing up. And this is another 361 00:20:21,109 --> 00:20:25,160 S1: opportunity for AI to change that education and not teach them. Oh, 362 00:20:25,160 --> 00:20:27,950 S1: it's about what you can remember and instead teach them. 363 00:20:27,950 --> 00:20:31,190 S1: It's about what you find interesting, right? Get that curiosity 364 00:20:31,190 --> 00:20:34,670 S1: into them early, then expose them to modules of training 365 00:20:34,670 --> 00:20:38,359 S1: instead of stupid curricula. Tough times for new grads. Okay, 366 00:20:38,359 --> 00:20:41,419 S1: same point. Yeah, see above. And of course this is 367 00:20:41,420 --> 00:20:44,060 S1: not the only factor, but yeah, hitting a wall with 368 00:20:44,060 --> 00:20:49,520 S1: entry level hiring projected to drop by 5.8% in 24. Yeah, 369 00:20:49,520 --> 00:20:53,389 S1: wait till you see 25 and 26 and 27. It's 370 00:20:53,390 --> 00:20:57,050 S1: going to get nastier. Software engineer positions on indeed have 371 00:20:57,050 --> 00:21:02,780 S1: plummeted 30% from pre-pandemic levels. Even with all this stuff considered, 372 00:21:02,780 --> 00:21:05,689 S1: I still don't understand that number. That's ridiculous. That's got 373 00:21:05,690 --> 00:21:08,239 S1: to be some macroeconomic stuff. I'll have to talk to 374 00:21:08,240 --> 00:21:11,000 S1: Mike about that. I don't know all the inputs coming in. 375 00:21:11,000 --> 00:21:14,359 S1: I think the thing I'm describing is one of those inputs, 376 00:21:14,359 --> 00:21:16,880 S1: but I don't know how that accounts for 30%. So 377 00:21:16,880 --> 00:21:22,430 S1: obviously multifactorial is college overrated? Nearly 30% of Americans are 378 00:21:22,430 --> 00:21:25,910 S1: questioning the value of college, according to a recent Pew poll. 379 00:21:25,910 --> 00:21:32,990 S1: Once again, see above. Job hopping not just flaky, but strategic. Yeah, yeah, 380 00:21:32,990 --> 00:21:34,699 S1: I wrote this here. You're wondering why I say I 381 00:21:34,700 --> 00:21:38,540 S1: wrote this. I helps with this sometimes. Write some of 382 00:21:38,540 --> 00:21:43,010 S1: these sentences. Some portion of this could be I written 383 00:21:43,010 --> 00:21:45,980 S1: a lot of these I write myself or I have 384 00:21:45,980 --> 00:21:48,110 S1: to go and edit them just because I don't like 385 00:21:48,109 --> 00:21:50,690 S1: what the AI writes. So I have to do it myself. 386 00:21:50,690 --> 00:21:53,180 S1: Like I had to write this whole thing. I think 387 00:21:53,180 --> 00:21:55,489 S1: not because they're flaky, but because they're in search of 388 00:21:55,490 --> 00:21:58,580 S1: a culture that matches their drive. Sticking it out in 389 00:21:58,580 --> 00:22:01,730 S1: a toxic work environment can dole a high performance edge. 390 00:22:01,730 --> 00:22:05,870 S1: I think that might be I right there. Can't actually remember. 391 00:22:05,869 --> 00:22:08,810 S1: My AI is getting so good that. A it actually 392 00:22:08,810 --> 00:22:14,209 S1: copies me pretty well. I definitely write everything in the blue. 100% yeah. 393 00:22:14,210 --> 00:22:18,200 S1: So here's my main point for truly top performers. It's 394 00:22:18,200 --> 00:22:21,080 S1: not bad to see them jump every 3 to 4 years. 395 00:22:21,080 --> 00:22:24,560 S1: Hiring managers know they get bored quickly and that they 396 00:22:24,560 --> 00:22:28,100 S1: can probably solve a bunch of problems, and then they 397 00:22:28,100 --> 00:22:29,899 S1: just leave because they want to solve a new challenge. 398 00:22:29,900 --> 00:22:32,030 S1: And plus they got a little bit of their stock payout. 399 00:22:32,450 --> 00:22:35,419 S1: I know tons of people like this, like star programmers, 400 00:22:35,420 --> 00:22:38,149 S1: and they just bounce around and nobody really cares because 401 00:22:38,150 --> 00:22:40,040 S1: they're just happy to have them when they show up. 402 00:22:40,340 --> 00:22:44,450 S1: It doesn't work for the whatever. 30 years ago IBM employee, 403 00:22:44,450 --> 00:22:47,630 S1: they'll be like, oh, you left after seven years, why 404 00:22:47,630 --> 00:22:50,119 S1: are your pants on fire or whatever? All right. So 405 00:22:50,119 --> 00:22:52,969 S1: now they're saying avoiding the sun might actually be a 406 00:22:52,970 --> 00:22:56,389 S1: risk factor. Yeah. So get too much sun. You'll die. 407 00:22:56,390 --> 00:22:59,690 S1: Don't get enough sun. You'll die. Punchline. You'll die. I 408 00:22:59,690 --> 00:23:02,840 S1: definitely believe this. I think there's value to this whole like, 409 00:23:02,840 --> 00:23:07,610 S1: get sun touch grass thing grounding. Get in touch with nature, whatever. Yeah, 410 00:23:07,609 --> 00:23:11,929 S1: definitely true for me, anecdotally and possibly anecdotally as well. 411 00:23:11,930 --> 00:23:14,420 S1: I think we need more science on this. More people 412 00:23:14,420 --> 00:23:17,420 S1: are doing marijuana than drinking now. I guess we could 413 00:23:17,420 --> 00:23:19,910 S1: take that as a win. That drinking is down because 414 00:23:19,910 --> 00:23:23,390 S1: Uberman said alcohol is poison. He also went pretty hard 415 00:23:23,390 --> 00:23:27,710 S1: on cannabis as far as I can remember. Psychedelics as painkillers. Yeah, 416 00:23:27,710 --> 00:23:30,920 S1: this is getting a lot of press. One third of 417 00:23:30,920 --> 00:23:35,570 S1: Amazon warehouse workers are on food stamps or Medicaid. That 418 00:23:35,570 --> 00:23:38,690 S1: is not a good number one file to rule them all. 419 00:23:38,690 --> 00:23:44,900 S1: One massive, sprawling text file I am in favor of. Text. 420 00:23:44,900 --> 00:23:49,040 S1: Text is my favorite. I love large files, broken down, 421 00:23:49,040 --> 00:23:53,900 S1: markdown bullets. Super clear, easy to read. I love this 422 00:23:53,900 --> 00:23:57,170 S1: whole vibe, and I'm using it for fabric and managing 423 00:23:57,170 --> 00:24:01,010 S1: my own internal context and everything. US Justice Department is 424 00:24:01,010 --> 00:24:05,840 S1: gearing up to dismantle the Live Nation Ticketmaster giant. Hallelujah! 425 00:24:05,840 --> 00:24:10,220 S1: So overdue. And thank you for Taylor Swift, because I 426 00:24:10,220 --> 00:24:12,740 S1: think she had something to do with this. California is 427 00:24:12,740 --> 00:24:16,909 S1: getting over a quarter of its energy from sun soared. 428 00:24:16,910 --> 00:24:19,609 S1: Finding is not what you think. An argument from one 429 00:24:19,609 --> 00:24:23,150 S1: expert claiming everything we know about sword fighting is bad. 430 00:24:23,330 --> 00:24:26,450 S1: It's my favorite Paul Graham essay. Great work happens by 431 00:24:26,450 --> 00:24:32,180 S1: focusing consistently on something you're genuinely interested in. I love 432 00:24:32,180 --> 00:24:36,590 S1: this sentence. When you pause to take stock, you're surprised 433 00:24:36,590 --> 00:24:39,590 S1: how far you've come. Paul Graham is one of my 434 00:24:39,590 --> 00:24:43,550 S1: favorite essayists ever, and this is one of my favorite ones. 435 00:24:43,550 --> 00:24:45,590 S1: I think it's called Great Work or How to Do 436 00:24:45,590 --> 00:24:47,570 S1: Great Work is the name of the essay should have 437 00:24:47,570 --> 00:24:50,239 S1: linked it here. This rhymes with a realization I've been 438 00:24:50,240 --> 00:24:52,879 S1: having lately that the best leaders and founders have something 439 00:24:52,880 --> 00:24:55,220 S1: in common, which is like we were talking about before, 440 00:24:55,220 --> 00:24:58,760 S1: actually obsession. It's not that they're disciplined, or they might be, 441 00:24:58,760 --> 00:25:01,550 S1: and they probably are, but they just might be obsessed, 442 00:25:01,550 --> 00:25:04,610 S1: which tends to look a lot like discipline because they 443 00:25:04,609 --> 00:25:08,720 S1: both produce consistent effort in a uniform direction. I find 444 00:25:08,720 --> 00:25:12,200 S1: myself increasingly looking for people who are obsessed with things, 445 00:25:12,200 --> 00:25:15,709 S1: life and idea, whatever. They can't stop thinking about it. 446 00:25:15,710 --> 00:25:18,680 S1: And the million different ideas they have tend to revolve 447 00:25:18,680 --> 00:25:22,070 S1: around that same concept. It's one that my buddy Joel 448 00:25:22,070 --> 00:25:25,399 S1: Paris sent me the link to. This is cool. Nostalgia 449 00:25:25,400 --> 00:25:27,260 S1: tends to peak at a single age, and I think 450 00:25:27,260 --> 00:25:30,350 S1: this is either Pew or Gallup. I mix the two up, 451 00:25:30,350 --> 00:25:33,290 S1: but if you draw the line, it's somewhere like right here, 452 00:25:33,290 --> 00:25:37,550 S1: and that's around 15. Yeah, 15 years old, 18 years 453 00:25:37,550 --> 00:25:40,910 S1: old or whatever. And guess what? The most close knit 454 00:25:40,910 --> 00:25:44,450 S1: community is the most moral society, the least political division, 455 00:25:44,450 --> 00:25:48,740 S1: the happiest families, the most reliable news reporting. It all 456 00:25:48,740 --> 00:25:53,480 S1: happens when you're 18 years old. So regardless of your age, 457 00:25:53,480 --> 00:25:56,899 S1: whenever you were 18, that's when you think the Golden 458 00:25:56,900 --> 00:26:02,389 S1: Age was brilliant. Data visualization recommendation of the week. Become 459 00:26:02,390 --> 00:26:05,990 S1: able to articulate who you are and what you're about 460 00:26:05,990 --> 00:26:09,379 S1: in 25 to 50 words. And the aphorism of the 461 00:26:09,380 --> 00:26:14,030 S1: week fall seven times. Stand up eight, fall seven times. 462 00:26:14,030 --> 00:26:19,590 S1: Stand up eight. Japanese proverb. Unsupervised Learning is produced and 463 00:26:19,590 --> 00:26:23,130 S1: edited by Daniel Meisler on a Neumann U87 AI microphone 464 00:26:23,130 --> 00:26:27,119 S1: using Hindenburg. Intro and outro music is by zombie with 465 00:26:27,119 --> 00:26:29,940 S1: the why and to get the text and links from 466 00:26:29,940 --> 00:26:32,070 S1: this episode, sign up for the newsletter version of the 467 00:26:32,070 --> 00:26:37,610 S1: show at Daniel meisler.com/newsletter. We'll see you next time.