1 00:00:00,000 --> 00:00:03,150 S1: Welcome to Unsupervised Learning, a security, AI and meaning focused 2 00:00:03,150 --> 00:00:06,000 S1: podcast that looks at how best to thrive as humans 3 00:00:06,000 --> 00:00:10,230 S1: in a post AI world. It combines original ideas, analysis, 4 00:00:10,260 --> 00:00:13,470 S1: and mental models to bring not just the news, but 5 00:00:13,470 --> 00:00:21,419 S1: why it matters and how to respond. All right, welcome 6 00:00:21,420 --> 00:00:24,780 S1: to unsupervised learning. This is Daniel. All right. So settling 7 00:00:24,780 --> 00:00:28,290 S1: in for a couple of light holiday weeks, trying to 8 00:00:28,320 --> 00:00:30,720 S1: chill out as much as possible, reading a bunch of 9 00:00:30,720 --> 00:00:32,909 S1: lit RPG. I don't think I'm going to play too 10 00:00:32,909 --> 00:00:36,900 S1: much path of Exile. It's like a religion. I've already 11 00:00:36,900 --> 00:00:40,260 S1: got too many religions. I've got reading as a religion, 12 00:00:40,260 --> 00:00:43,440 S1: I've got AI as a religion, got security work. I'm 13 00:00:43,440 --> 00:00:46,260 S1: building a bunch of tech. I just I don't have 14 00:00:46,260 --> 00:00:49,710 S1: time to invest in getting good at Poe. It's just 15 00:00:49,710 --> 00:00:53,339 S1: too intimidating. Plus, the little bit that I did play 16 00:00:53,340 --> 00:00:57,330 S1: was mostly like hitting the spacebar and dodging way too 17 00:00:57,330 --> 00:00:59,850 S1: much dodging. So don't think I'm going to be playing 18 00:00:59,860 --> 00:01:02,560 S1: that any time soon, but I do watch other people 19 00:01:02,560 --> 00:01:05,199 S1: play it, so that's kind of fun. Please go subscribe 20 00:01:05,200 --> 00:01:07,420 S1: to the YouTube channel. If you're watching this on YouTube. 21 00:01:07,420 --> 00:01:10,240 S1: That's probably kind of silly because you're probably already on 22 00:01:10,240 --> 00:01:13,840 S1: the channel and hopefully you're subscribed, but always forget to 23 00:01:13,870 --> 00:01:16,990 S1: promote that. So trying to remember wrote a piece for 24 00:01:16,990 --> 00:01:19,899 S1: dad's been one to get out for a while. It's 25 00:01:19,900 --> 00:01:24,160 S1: basically about how updated context is going to be like 26 00:01:24,190 --> 00:01:27,940 S1: the centerpiece of vulnerability management, because the problem isn't finding vulns, 27 00:01:27,940 --> 00:01:32,050 S1: it's actually fixing them. So go check that out and 28 00:01:32,170 --> 00:01:35,679 S1: let's get into security. So full face masks for facial 29 00:01:35,680 --> 00:01:41,050 S1: recognition evasion. So realistic masks that you could put on 30 00:01:41,050 --> 00:01:42,670 S1: I'm going to open this up. This is why we 31 00:01:42,670 --> 00:01:45,279 S1: do it live. Look at this thing. So it's like 32 00:01:45,280 --> 00:01:47,320 S1: a stocking or whatever. But you put it over your 33 00:01:47,319 --> 00:01:50,290 S1: face and then it kind of pulls to contort to 34 00:01:50,320 --> 00:01:53,140 S1: your face and look at that from a distance. That 35 00:01:53,140 --> 00:01:55,660 S1: is not so bad, not so bad at all. And 36 00:01:55,660 --> 00:01:58,360 S1: look at all these different faces you got. So pretty cool. Yeah. 37 00:01:58,390 --> 00:02:00,610 S1: And Bruce Schneier is the one I think I saw 38 00:02:00,640 --> 00:02:04,210 S1: that from. But yeah, I'm curious. I'd love to see 39 00:02:04,210 --> 00:02:06,850 S1: some testing done from the security community to buy a 40 00:02:06,850 --> 00:02:08,920 S1: bunch of these. And then maybe if we have access 41 00:02:08,919 --> 00:02:12,010 S1: to like a facial recognition system to see if we 42 00:02:12,010 --> 00:02:14,350 S1: could get it to match someone else. I think the 43 00:02:14,350 --> 00:02:17,770 S1: big thing is like, is it only visual or is 44 00:02:17,770 --> 00:02:22,150 S1: there like an active component? I think, and I don't 45 00:02:22,150 --> 00:02:25,210 S1: think most do have an active component. I know face 46 00:02:25,210 --> 00:02:28,570 S1: ID does, but I think most facial recognition does not 47 00:02:28,570 --> 00:02:31,630 S1: have an active component. And what that would mean is 48 00:02:31,660 --> 00:02:35,080 S1: like is something actually bouncing off your face, which would 49 00:02:35,080 --> 00:02:37,480 S1: go through the stocking so it'd be able to see 50 00:02:37,510 --> 00:02:40,390 S1: like where your cheekbones are, where your chin is, and, 51 00:02:40,419 --> 00:02:45,339 S1: you know, eyes, foreheads, stuff like that. That would seem 52 00:02:45,340 --> 00:02:49,690 S1: to break the system or break the masks. But yeah, 53 00:02:49,720 --> 00:02:52,570 S1: I don't I don't think most are doing that, but 54 00:02:52,570 --> 00:02:56,470 S1: I'm not sure. Visual Studio Code remote tunnel hacks. So 55 00:02:56,470 --> 00:03:00,380 S1: some Chinese attackers are caught stealing V's codes, remote tunnel 56 00:03:00,380 --> 00:03:03,799 S1: feature or using the feature to hack it. Service providers 57 00:03:03,800 --> 00:03:06,590 S1: in South Europe and China starting to limit the sale 58 00:03:06,590 --> 00:03:10,820 S1: of critical drone drone components to the US and Europe 59 00:03:10,820 --> 00:03:14,390 S1: that are needed for Ukraine's fighting efforts. So this is 60 00:03:14,389 --> 00:03:17,809 S1: all part of the trade war of us limiting their 61 00:03:17,810 --> 00:03:21,590 S1: access to chips. So now they're limiting our access to 62 00:03:21,620 --> 00:03:25,639 S1: key materials to actually build stuff, and now they're limiting 63 00:03:25,669 --> 00:03:28,550 S1: drone stuff. I really can't wait to get our drone 64 00:03:28,550 --> 00:03:33,380 S1: systems up and running. I'm really worried that any war, 65 00:03:33,410 --> 00:03:36,590 S1: any conflict, hopefully not a major war, but any conflict 66 00:03:36,590 --> 00:03:39,290 S1: that we have with China, it's going to be about drones, 67 00:03:39,290 --> 00:03:42,080 S1: like aircraft carriers I don't think are going to be 68 00:03:42,080 --> 00:03:44,870 S1: as big of a deal. I remember seeing a piece 69 00:03:44,870 --> 00:03:47,720 S1: or like a series or something. I was in the 70 00:03:47,720 --> 00:03:50,180 S1: Army at the time, so this was like mid 90s 71 00:03:50,180 --> 00:03:53,120 S1: and I was reading about this thing called the Assassin's 72 00:03:53,120 --> 00:03:57,360 S1: Mace weapon, which is a Chinese concept, I believe where 73 00:03:57,360 --> 00:03:59,580 S1: it's like when you have an enemy that is much 74 00:03:59,580 --> 00:04:03,990 S1: too big and powerful for you, you don't bring force 75 00:04:03,990 --> 00:04:06,360 S1: against force. What you do is you bring some kind 76 00:04:06,360 --> 00:04:11,070 S1: of weapon that is highly asymmetrical, that is capable of 77 00:04:11,070 --> 00:04:16,080 S1: hurting this thing when regular force has no chance whatsoever. 78 00:04:16,110 --> 00:04:20,820 S1: And drones are that. And once again, I've recommended this 79 00:04:20,820 --> 00:04:24,060 S1: multiple times, but I recommend you read the book Kill 80 00:04:24,060 --> 00:04:28,830 S1: Decision by Daniel Suarez. It is fantastic. And it's all 81 00:04:28,830 --> 00:04:33,060 S1: about mini drones and implications and stuff like that. FBI 82 00:04:33,060 --> 00:04:38,490 S1: is shut down Redux, a cybercrime marketplace operating since 2016, 83 00:04:38,490 --> 00:04:41,880 S1: and they arrested three Kosovo nationals who were running it. 84 00:04:41,910 --> 00:04:47,310 S1: Russian state backed APT Kameraden has been targeting Russian speaking 85 00:04:47,310 --> 00:04:51,750 S1: individuals with two Android spyware families called Bone Spy and 86 00:04:51,750 --> 00:04:55,409 S1: Plain Gnome, and it can record calls, capture photos and 87 00:04:55,450 --> 00:05:00,430 S1: collect SMS messages. Yahoo cut 25% of its cybersecurity team, 88 00:05:00,430 --> 00:05:03,219 S1: which is also known as the paranoids, over the last year. 89 00:05:03,250 --> 00:05:06,700 S1: Paranoids are kind of og famous in the security community, 90 00:05:06,700 --> 00:05:09,940 S1: so kind of sad to see them getting cut down. 91 00:05:10,060 --> 00:05:12,880 S1: And I know the attrition has happened for a long 92 00:05:12,880 --> 00:05:17,950 S1: time over the years, but I presented and taking classes 93 00:05:17,980 --> 00:05:20,500 S1: over there in the facility that they were based out of. 94 00:05:20,500 --> 00:05:24,219 S1: And yeah, sad to see kind of this old school 95 00:05:24,250 --> 00:05:28,390 S1: kind of organization going away. Another run of augmented happening 96 00:05:28,390 --> 00:05:32,680 S1: in February. It's actually February 3rd of this coming year, 97 00:05:32,680 --> 00:05:34,750 S1: and this is going to be a full discussion and 98 00:05:34,750 --> 00:05:39,039 S1: workshop on building out your personal Telos files and using 99 00:05:39,070 --> 00:05:43,990 S1: AI to access them and use them, manipulate them. So 100 00:05:43,990 --> 00:05:48,429 S1: it is $495, but you get 25% off if you 101 00:05:48,460 --> 00:05:51,880 S1: are a member and you can go and reserve a slot. 102 00:05:51,910 --> 00:05:56,200 S1: I think they're still open, but you definitely want to 103 00:05:56,200 --> 00:06:00,430 S1: check it out. Probably will be full soon. And if 104 00:06:00,460 --> 00:06:03,550 S1: you become a member to get the discount, the discount 105 00:06:03,550 --> 00:06:07,600 S1: is actually more than the cost of the membership. So 106 00:06:07,600 --> 00:06:10,420 S1: really good time to become a member if you've been 107 00:06:10,420 --> 00:06:14,529 S1: thinking about it. Super annoyed about the fact that ChatGPT Pro, 108 00:06:14,560 --> 00:06:20,890 S1: which I use constantly, is not as good, unfortunately, as 109 00:06:20,890 --> 00:06:25,450 S1: Claude AI. I have more problems with it. It sends 110 00:06:25,450 --> 00:06:29,740 S1: me down more rat holes. It doesn't follow my instructions 111 00:06:29,740 --> 00:06:34,930 S1: as closely as Claude does, and it generally produces code 112 00:06:34,930 --> 00:06:37,479 S1: that's not as good. The code quality is a little 113 00:06:37,510 --> 00:06:41,230 S1: bit closer than than the other aspects. What Claude does really, 114 00:06:41,230 --> 00:06:45,279 S1: really well, and let's not frame it as ChatGPT pro 115 00:06:45,279 --> 00:06:48,640 S1: being bad. It's more like what Claude does really well 116 00:06:48,640 --> 00:06:52,630 S1: is it follows your instructions and it responds like a human, 117 00:06:52,630 --> 00:06:56,420 S1: and it's like you're actually working with a coworker. So 118 00:06:56,420 --> 00:06:58,909 S1: if you tell it, look, I only want one instruction 119 00:06:58,910 --> 00:07:01,099 S1: at a time. This is one of my instructions. I 120 00:07:01,100 --> 00:07:04,820 S1: only want one instruction at a time. Because oftentimes we're 121 00:07:04,820 --> 00:07:06,830 S1: going to need to troubleshoot that. I have to go 122 00:07:06,830 --> 00:07:08,780 S1: in and run it and see if it works, and 123 00:07:08,779 --> 00:07:11,210 S1: then let you know if it doesn't work. Claude follows 124 00:07:11,210 --> 00:07:17,660 S1: that almost perfectly. And OpenAI or ChatGPT, it'll be like, okay, boom, 125 00:07:17,660 --> 00:07:21,590 S1: here's seven more steps. So I'm scrolling to get back 126 00:07:21,590 --> 00:07:25,010 S1: up to whatever. And I'm just like, why did you 127 00:07:25,010 --> 00:07:27,739 S1: give me seven steps? The thing before didn't work. I 128 00:07:27,740 --> 00:07:30,470 S1: told you not to do this. And then we give 129 00:07:30,470 --> 00:07:33,380 S1: me code and it's like, here's a little bit of code. 130 00:07:33,380 --> 00:07:36,110 S1: And then at the bottom it's like, here's the rest 131 00:07:36,110 --> 00:07:38,870 S1: of the code or insert the rest of the code. 132 00:07:38,870 --> 00:07:43,160 S1: I said explicitly in the instructions not to do that. Also, 133 00:07:43,160 --> 00:07:46,100 S1: here's the other instruction I gave it. And also to Claude, 134 00:07:46,100 --> 00:07:50,780 S1: write out files using echo and EOF, which is end 135 00:07:50,780 --> 00:07:53,190 S1: of file. That way I could just copy the thing 136 00:07:53,190 --> 00:07:56,760 S1: and paste it, and it writes the file in my directory, 137 00:07:56,760 --> 00:07:59,130 S1: as opposed to just giving me the raw code. Then 138 00:07:59,130 --> 00:08:04,800 S1: I have to open the file delete paste close, which saves. 139 00:08:04,980 --> 00:08:11,490 S1: So those extra steps Claude follows perfectly hardly ever misses that. 140 00:08:11,520 --> 00:08:17,130 S1: And ChatGPT like 60% of the time misses it. So 141 00:08:17,130 --> 00:08:19,860 S1: I am really hopeful that ChatGPT is going to catch 142 00:08:19,860 --> 00:08:24,150 S1: up and hopefully even get better. But it's really annoying 143 00:08:24,180 --> 00:08:26,790 S1: to pay $200 for a thing and have that thing 144 00:08:26,790 --> 00:08:28,890 S1: not be as good as the thing that's free. Or 145 00:08:28,890 --> 00:08:32,699 S1: I guess I'm paying the $20 subscription for anthropic. Not 146 00:08:32,700 --> 00:08:35,760 S1: even sure. But anyway, I'm not paying $200 for it 147 00:08:35,760 --> 00:08:39,900 S1: and it's better. And this is Claude not even having 148 00:08:39,900 --> 00:08:43,829 S1: shipped anything recently. I have a feeling when Claude ships 149 00:08:43,830 --> 00:08:47,250 S1: its new thing, it's going to jump way ahead of 150 00:08:47,250 --> 00:08:51,579 S1: Google and way ahead of ChatGPT. But I'm sure again, 151 00:08:51,610 --> 00:08:55,960 S1: it's just leapfrogs. Constant leapfrogs. And I think 2025 is 152 00:08:55,960 --> 00:08:59,470 S1: going to be even more. It's going to accelerate in 153 00:08:59,470 --> 00:09:02,170 S1: the difference between one to the other. So I think 154 00:09:02,559 --> 00:09:06,429 S1: Apple is actually doing exceptionally well on the AI front. 155 00:09:06,429 --> 00:09:09,190 S1: Most people think that they're kind of way behind, and 156 00:09:09,190 --> 00:09:12,130 S1: Apple intelligence isn't very good. The one thing I agree 157 00:09:12,130 --> 00:09:15,040 S1: with is Siri is still massively broken, but it is 158 00:09:15,040 --> 00:09:19,300 S1: getting better. But the thing I think that I realized 159 00:09:19,300 --> 00:09:21,400 S1: that a lot of people don't. And keep in mind, 160 00:09:21,429 --> 00:09:24,400 S1: I'm an Apple fanboy, so you got to take that 161 00:09:24,400 --> 00:09:28,120 S1: bias into account. I don't think it actually applies. In fact, 162 00:09:28,120 --> 00:09:32,530 S1: I think it gives me a benefit here. But I 163 00:09:32,530 --> 00:09:34,540 S1: do want to make it very clear that I am 164 00:09:34,540 --> 00:09:37,930 S1: an Apple fanboy. So if I were to be wrong, 165 00:09:37,929 --> 00:09:42,400 S1: that would definitely be part of the reason. Um, so 166 00:09:42,400 --> 00:09:47,110 S1: I believe that Apple is building life OS and that 167 00:09:47,110 --> 00:09:51,980 S1: they've been slowly and quietly adding AI to that. And 168 00:09:51,980 --> 00:09:53,960 S1: they kind of have been for a long time. But 169 00:09:53,960 --> 00:09:57,650 S1: with Apple intelligence, you should never expect them to jump 170 00:09:57,650 --> 00:10:00,349 S1: out in front and do something crazy with their AI. 171 00:10:00,380 --> 00:10:03,199 S1: That is just not the way the apple works. What 172 00:10:03,200 --> 00:10:07,130 S1: they're doing, though, is they're adding education, they're adding health, 173 00:10:07,130 --> 00:10:10,850 S1: they're adding, you know, the ability to see more and 174 00:10:10,850 --> 00:10:15,650 S1: hear more with like the AirPods. They're just they're slowly 175 00:10:15,650 --> 00:10:19,429 S1: incorporating more and more of our lives into their ecosystem. 176 00:10:19,429 --> 00:10:22,760 S1: And the whole advantage of Apple, what they have over 177 00:10:22,760 --> 00:10:28,580 S1: everyone else is how cleanly and in a unified way, 178 00:10:28,580 --> 00:10:34,069 S1: they unify all their different ecosystem components to work together. Okay, 179 00:10:34,070 --> 00:10:37,550 S1: so that's what they're known for. That's what they're best at. 180 00:10:37,550 --> 00:10:40,309 S1: That's what makes them so compelling. That's why everyone switches 181 00:10:40,309 --> 00:10:44,150 S1: off of Android. Because when you have Apple stuff, the 182 00:10:44,150 --> 00:10:47,390 S1: more Apple stuff you use, the better everything gets. Now 183 00:10:47,420 --> 00:10:50,760 S1: add AI on top of that, which runs securely inside 184 00:10:50,760 --> 00:10:55,020 S1: of the Secure Enclave, or in their new secure cloud 185 00:10:55,020 --> 00:10:59,280 S1: infrastructure that they built, it's going to be insane. Absolutely insane. 186 00:10:59,280 --> 00:11:03,330 S1: I mean, imagine an AI agent having full access to 187 00:11:03,360 --> 00:11:06,179 S1: all your health data, and if you're a security person, 188 00:11:06,179 --> 00:11:09,750 S1: you're probably freaking out about it. But with Apple, you 189 00:11:09,750 --> 00:11:12,690 S1: you actually don't have to freak out about it. I 190 00:11:12,690 --> 00:11:16,050 S1: worked there for three years in security. I've been watching 191 00:11:16,050 --> 00:11:20,400 S1: them security wise for over a decade before that, and 192 00:11:20,400 --> 00:11:24,450 S1: I have never seen a company more crazy about doing 193 00:11:24,450 --> 00:11:28,829 S1: security and specifically privacy really, really well than them. And 194 00:11:28,830 --> 00:11:31,710 S1: that's why also it takes them a little bit longer 195 00:11:31,710 --> 00:11:34,859 S1: to do things, because they have to be very particular 196 00:11:34,860 --> 00:11:37,530 S1: and meticulous about it. So people are saying, oh, they're 197 00:11:37,530 --> 00:11:41,250 S1: going slow, they're being left behind. I assure you, they 198 00:11:41,280 --> 00:11:44,700 S1: are not being left behind. And here's what I'm telling 199 00:11:44,700 --> 00:11:48,760 S1: you to do. Let's have this conversation At the end 200 00:11:48,760 --> 00:11:53,559 S1: of 2026 or even the end of 2025. What happens 201 00:11:53,590 --> 00:11:57,429 S1: is they're small little winds in the ecosystem world and 202 00:11:57,429 --> 00:11:59,980 S1: also in the AI world on top of the ecosystem, 203 00:11:59,980 --> 00:12:04,870 S1: they're small. Little winds slowly build. They build and build 204 00:12:04,870 --> 00:12:07,990 S1: and build. And then one day some some YouTuber is 205 00:12:07,990 --> 00:12:10,780 S1: going to be like, wow, Apple just came out with 206 00:12:10,780 --> 00:12:14,590 S1: all this amazing stuff. They really surprised us. Didn't surprise me, 207 00:12:14,620 --> 00:12:17,800 S1: didn't surprise a lot of people who were watching Apple closely. 208 00:12:17,800 --> 00:12:24,729 S1: It's slow, incremental, methodical improvements. Every little OS update that 209 00:12:24,730 --> 00:12:27,730 S1: comes out, they tweak one little thing, which helps it 210 00:12:27,730 --> 00:12:31,179 S1: work with these nine other things. Then they tweak nine 211 00:12:31,179 --> 00:12:34,329 S1: other things, helps it work with this other thing. And 212 00:12:34,330 --> 00:12:38,500 S1: slowly it's it's not just improving each component, it's pulling 213 00:12:38,500 --> 00:12:42,429 S1: the components tighter to work better together. So Apple is 214 00:12:42,429 --> 00:12:47,929 S1: building life OS okay, not work os not, you know, 215 00:12:47,960 --> 00:12:53,060 S1: personal os. A unified life OS. That's mental health. It's health, 216 00:12:53,059 --> 00:12:57,319 S1: it's fitness, it's career, it's creativity. It's all those things 217 00:12:57,320 --> 00:13:00,170 S1: unified together. The way I like to think about it 218 00:13:00,170 --> 00:13:04,640 S1: is if you watch the most advanced sci fi about 219 00:13:04,640 --> 00:13:08,420 S1: an AI agent, something like her, that is essentially what 220 00:13:08,420 --> 00:13:12,440 S1: Apple is building. Okay. It's building something like her. But 221 00:13:12,470 --> 00:13:15,530 S1: imagine it more with like AR in addition, because that 222 00:13:15,530 --> 00:13:19,490 S1: was mostly just voice. But imagine, you know, glasses or 223 00:13:19,490 --> 00:13:23,600 S1: contact lenses. So you're seeing overlays. If you look at my, um, 224 00:13:23,630 --> 00:13:28,850 S1: my peace AI's predictable path, that is what AI is 225 00:13:28,850 --> 00:13:31,280 S1: going to look like. And I think Apple's going to 226 00:13:31,280 --> 00:13:34,699 S1: get there first. Each component, they won't get there first, 227 00:13:34,700 --> 00:13:38,840 S1: but having it all unified into a platform, into an ecosystem, 228 00:13:38,840 --> 00:13:41,900 S1: they will get to that first. And then even when 229 00:13:41,900 --> 00:13:44,929 S1: other people arrive, it'll be cleaner with Apple, which is 230 00:13:44,929 --> 00:13:48,020 S1: why everyone will continue to use it. All right. Former 231 00:13:48,050 --> 00:13:54,110 S1: OpenAI researcher who publicly criticized the criticized the company's data practices, 232 00:13:54,110 --> 00:13:58,670 S1: was found dead in his San Francisco apartment. Suchir Balaji, 233 00:13:58,700 --> 00:14:02,510 S1: 26 years old, dead of suicide and a lot of 234 00:14:02,510 --> 00:14:04,850 S1: conspiracy going around this. The one thing I will say 235 00:14:04,850 --> 00:14:07,910 S1: about this I don't know anything about the case, you know, 236 00:14:07,940 --> 00:14:11,900 S1: probably wasn't foul play. Just just from the math of it. 237 00:14:11,929 --> 00:14:16,010 S1: When people are whistleblowers, which is what he was, what 238 00:14:16,040 --> 00:14:18,199 S1: a lot of people don't realize is you are you 239 00:14:18,230 --> 00:14:22,250 S1: are shunning your community. You're taking somebody who's usually pretty 240 00:14:22,250 --> 00:14:25,640 S1: powerful inside of your community, and you're basically blowing the 241 00:14:25,640 --> 00:14:30,200 S1: whistle on them, which means they that big company or 242 00:14:30,200 --> 00:14:33,470 S1: that big person kind of. I'm not saying OpenAI did this. 243 00:14:33,470 --> 00:14:37,580 S1: I'm saying in general going back, you know, hundreds of years, 244 00:14:37,790 --> 00:14:42,320 S1: this concept is that the community turns against you. You 245 00:14:42,320 --> 00:14:46,020 S1: get looked at as like a rat or a tattletale 246 00:14:46,020 --> 00:14:48,480 S1: or a narc or something. And it's just like it's 247 00:14:48,480 --> 00:14:53,310 S1: very lonely. Like, the most dangerous thing right now is loneliness, right? 248 00:14:53,310 --> 00:14:56,430 S1: And and I'm just projecting here. I'm just saying this 249 00:14:56,430 --> 00:14:59,880 S1: is a common thing that happens with whistleblowers is they 250 00:14:59,880 --> 00:15:04,740 S1: get ostracized, which is isolating. And it's stressful. That is, 251 00:15:04,740 --> 00:15:09,240 S1: in my opinion, the most likely cause of something like this. 252 00:15:09,240 --> 00:15:14,910 S1: But who knows? I mean, some conspiracies sometimes, uh, whistleblowers 253 00:15:14,940 --> 00:15:17,550 S1: are actually attacked. Doesn't mean it needs to be open. 254 00:15:17,550 --> 00:15:21,030 S1: I it could have been someone who's friendly to OpenAI. Who? 255 00:15:21,060 --> 00:15:24,960 S1: Who knows? I don't know, but I just wanted to 256 00:15:24,960 --> 00:15:28,740 S1: add that dynamic there. Y Combinator backed startup called artisan 257 00:15:28,740 --> 00:15:31,590 S1: is running ads all over San Francisco with slogans like 258 00:15:31,590 --> 00:15:36,660 S1: stop hiring humans, and artisans won't complain about work life balance. 259 00:15:36,660 --> 00:15:39,750 S1: This is brilliant. Okay, first of all, the CEO says 260 00:15:39,750 --> 00:15:43,570 S1: I did that on purpose. It was dystopian. On purpose. 261 00:15:44,020 --> 00:15:48,640 S1: We're not actually dystopian. Uh, don't believe you. I remember 262 00:15:48,640 --> 00:15:51,460 S1: that thing from, uh. What was it? I can't remember 263 00:15:51,460 --> 00:15:55,240 S1: the name of it. It's, um, not Woody Harrelson. Damn it. 264 00:15:55,270 --> 00:16:00,430 S1: What is this guy's name? Will Ferrell. Will Ferrell on. Um. Damn. 265 00:16:00,430 --> 00:16:02,800 S1: What is the name of that? I can't remember the 266 00:16:02,800 --> 00:16:04,960 S1: name of that damn show. Anyway, it's the one where 267 00:16:04,960 --> 00:16:08,620 S1: he's a newscaster and he's just like, I don't believe you. 268 00:16:08,620 --> 00:16:11,110 S1: And that's exactly how I feel about this. I don't 269 00:16:11,110 --> 00:16:14,380 S1: believe you. So I think he thought that was going 270 00:16:14,410 --> 00:16:17,170 S1: to be cool. I don't know, maybe I'm not giving 271 00:16:17,170 --> 00:16:20,860 S1: him enough benefit of the doubt. Maybe he actually did it. 272 00:16:21,010 --> 00:16:24,610 S1: It has dual purpose, right? When you do this, it 273 00:16:24,610 --> 00:16:29,620 S1: makes people look and go, Holy crap, artisans won't complain 274 00:16:29,620 --> 00:16:32,590 S1: about work life balance. That's going to make a whole 275 00:16:32,590 --> 00:16:35,260 S1: lot of people mad. All press is good press, right? 276 00:16:35,290 --> 00:16:38,200 S1: So bad press. Look, we're talking about it. A lot 277 00:16:38,230 --> 00:16:41,920 S1: of people are talking about it. So automatically they got attention. 278 00:16:41,930 --> 00:16:45,410 S1: But even more sinister than that, there are a whole 279 00:16:45,410 --> 00:16:48,890 S1: lot of hiring managers who are tired of humans complaining 280 00:16:48,890 --> 00:16:52,100 S1: about work life balance, and they're tired of people not 281 00:16:52,100 --> 00:16:54,080 S1: showing up to work, and they're tired of tired of 282 00:16:54,080 --> 00:16:57,260 S1: people doing half the work that they're supposed to be doing. 283 00:16:57,260 --> 00:17:00,530 S1: And so this is actually an appeal to them directly 284 00:17:00,530 --> 00:17:04,880 S1: as well, to fire your humans or stop hiring humans. 285 00:17:04,880 --> 00:17:08,360 S1: So it's brilliant in two different ways. And I'm curious 286 00:17:08,359 --> 00:17:10,850 S1: if they're actually going to do well. ChatGPT can now 287 00:17:10,850 --> 00:17:14,570 S1: analyze real time video through your phone's camera. Okay. ChatGPT 288 00:17:14,600 --> 00:17:18,080 S1: has this, OpenAI has this. I don't know if anthropic 289 00:17:18,080 --> 00:17:20,660 S1: has it. I haven't messed with the Claude app that much. 290 00:17:20,660 --> 00:17:25,399 S1: The actual static app, but, um, or the desktop app 291 00:17:25,400 --> 00:17:30,560 S1: or phone app, it is unbelievable to pull up an 292 00:17:30,560 --> 00:17:33,950 S1: I have a screen showing like I'm showing right now. 293 00:17:33,950 --> 00:17:36,770 S1: And then to say to it, um, hey, do you 294 00:17:36,770 --> 00:17:38,869 S1: see what I'm looking at? Hey, do you see that code? 295 00:17:38,869 --> 00:17:42,060 S1: Help me troubleshoot this thing over here. It's over here. 296 00:17:42,060 --> 00:17:45,570 S1: Working and returning code for you. But what? But watch this. 297 00:17:45,570 --> 00:17:49,050 S1: You didn't type anything. You're just talking. You're just talking. 298 00:17:49,050 --> 00:17:53,130 S1: And guess what? You're actually hearing it. Come back and 299 00:17:53,130 --> 00:17:56,250 S1: say things to you. So you are literally having a 300 00:17:56,250 --> 00:18:00,840 S1: voice conversation essentially with a colleague. And that colleague knows 301 00:18:00,840 --> 00:18:03,900 S1: everything about coding and is helping you code. But it's 302 00:18:03,900 --> 00:18:07,500 S1: not just coding. It can help you write or research 303 00:18:07,500 --> 00:18:11,070 S1: or do whatever. And this is the end of 2024 304 00:18:11,100 --> 00:18:13,680 S1: that that shipped and that was from Google. This one 305 00:18:13,680 --> 00:18:16,830 S1: is ChatGPT works the same. I only got it working 306 00:18:16,830 --> 00:18:19,770 S1: on my phone, though I haven't seen it on the 307 00:18:19,770 --> 00:18:21,869 S1: desktop app because it's a lot more powerful on a 308 00:18:21,869 --> 00:18:25,680 S1: desktop when you can see all your screens. And I'm 309 00:18:25,680 --> 00:18:28,620 S1: telling you, it is kind of like a ChatGPT moment 310 00:18:28,619 --> 00:18:32,760 S1: from late 2022, when you could literally have a regular 311 00:18:32,760 --> 00:18:36,060 S1: conversation with a thing that sees it's like it's over 312 00:18:36,090 --> 00:18:39,149 S1: your shoulder, it's sitting next to you. This is where 313 00:18:39,180 --> 00:18:43,090 S1: AGI starts to happen where? Not quite, but where you 314 00:18:43,090 --> 00:18:47,350 S1: have a colleague that understands everything you're saying and can 315 00:18:47,350 --> 00:18:51,760 S1: just go and do general tasks. So hugely impressive. I 316 00:18:51,760 --> 00:18:54,280 S1: recommend you go check it out. The one that's really 317 00:18:54,280 --> 00:19:00,220 S1: impressive is the Google AI studio. Okay. Gemini two and agents. Yeah, 318 00:19:00,220 --> 00:19:02,410 S1: this is what I was just talking about. Completely surreal 319 00:19:02,440 --> 00:19:05,920 S1: to just talk to an AI that you can basically 320 00:19:05,950 --> 00:19:08,109 S1: see your screen. So let's see if this one actually 321 00:19:08,109 --> 00:19:11,380 S1: does it. Um, do do they usually put it at 322 00:19:11,380 --> 00:19:13,810 S1: the bottom? Damn it. Can't find it. All right. So 323 00:19:13,810 --> 00:19:17,139 S1: this is basically what I was talking about. Uh, Gemini two. Okay. 324 00:19:17,170 --> 00:19:19,449 S1: Now I really have to go find this thing. It's 325 00:19:19,450 --> 00:19:22,240 S1: annoying me. Yeah, hopefully you can hear that. I can't 326 00:19:22,240 --> 00:19:25,300 S1: hear it, but hopefully you can. Yeah. So you're basically 327 00:19:25,300 --> 00:19:28,900 S1: just talking to it. It's really cool. All right, next story. 328 00:19:28,930 --> 00:19:32,410 S1: Exxon is building its first ever external power plant focused 329 00:19:32,410 --> 00:19:37,450 S1: on AI data centers. This is insane, Exxon. They are 330 00:19:37,450 --> 00:19:39,770 S1: so good at seeing the future. and I'm very mad 331 00:19:39,770 --> 00:19:42,980 S1: at them for many reasons. Well, like in the 1970s, 332 00:19:43,010 --> 00:19:46,250 S1: they saw the future of carbon going up and they 333 00:19:46,250 --> 00:19:49,070 S1: chose to bury that research. That's the thing that most 334 00:19:49,070 --> 00:19:51,679 S1: makes me angry. But in this case, they're seeing the 335 00:19:51,680 --> 00:19:56,810 S1: future of AI needing needing power. So they are going 336 00:19:56,840 --> 00:20:01,070 S1: into the business of providing AI data centers with power. 337 00:20:01,100 --> 00:20:03,320 S1: They're starting a whole new branch of business. I mean, 338 00:20:03,320 --> 00:20:06,919 S1: that is just absolutely brilliant. And like what? What is 339 00:20:06,920 --> 00:20:09,530 S1: their stock? Yeah. Let's look at like I don't like 340 00:20:09,530 --> 00:20:14,840 S1: this graph one day okay. All. Yeah that's ExxonMobil. Look 341 00:20:14,840 --> 00:20:20,990 S1: at this. Since October 1st 2020 it was $33. And 342 00:20:20,990 --> 00:20:26,119 S1: today it's $120. So they're doing okay. So yeah, really 343 00:20:26,119 --> 00:20:30,920 S1: really smart basically providing power to AI. Yeah. So it's 344 00:20:31,070 --> 00:20:34,639 S1: tracking health data in markdown files instead of apps. I 345 00:20:34,640 --> 00:20:38,730 S1: love doing everything in text. So really really enjoyed this article. 346 00:20:38,730 --> 00:20:42,629 S1: Center click has released a line of GPS based server appliances. 347 00:20:42,630 --> 00:20:45,119 S1: You know I got one. Even though I already have 348 00:20:45,119 --> 00:20:47,820 S1: a time server, I got another one because I am 349 00:20:47,820 --> 00:20:50,639 S1: stupid and this is actually fairly cheap. It's only like 350 00:20:50,640 --> 00:20:54,330 S1: 250 bucks. Yeah, it was like 250 bucks. So I 351 00:20:54,330 --> 00:20:56,940 S1: will probably give the other one away. Whichever one I 352 00:20:56,940 --> 00:20:59,790 S1: like the least, I will probably give away and I 353 00:20:59,790 --> 00:21:01,770 S1: might end up keeping this one. It seems like the 354 00:21:01,770 --> 00:21:04,680 S1: metrics might be a little bit nicer, but anyway, what 355 00:21:04,680 --> 00:21:08,700 S1: it is, it's a time server. It's stratum one, which 356 00:21:08,700 --> 00:21:11,790 S1: means instead, okay, stratum two is you're pulling from a 357 00:21:11,790 --> 00:21:17,100 S1: stratum one. Stratum one is or no. Is it stratum zero? 358 00:21:17,130 --> 00:21:18,720 S1: I'm going to pull mine up and have my IP 359 00:21:18,720 --> 00:21:21,810 S1: address in there. Um, I can't remember if it's stratum 360 00:21:21,810 --> 00:21:24,929 S1: one is the one that goes to satellites. No, it's 361 00:21:24,930 --> 00:21:28,920 S1: stratum zero is stratum zero I think. So basically the 362 00:21:28,920 --> 00:21:32,190 S1: most direct one you can have okay. The absolute most 363 00:21:32,190 --> 00:21:37,000 S1: direct one is you actually on on board these satellites? 364 00:21:37,030 --> 00:21:40,689 S1: This is the GPS system manages. It's the most accurate 365 00:21:40,720 --> 00:21:46,210 S1: time system in the world. It's 24 different satellites who have, um. 366 00:21:46,210 --> 00:21:49,990 S1: I believe cesium clocks. Either cesium or the other one. 367 00:21:49,990 --> 00:21:52,570 S1: I can't remember the other atom, but cesium is the 368 00:21:52,570 --> 00:21:57,550 S1: most accurate one. Their atomic clocks on the satellites, 24 369 00:21:57,580 --> 00:22:02,950 S1: of them. And they're in different orbit patterns. These 24 satellites, 370 00:22:02,950 --> 00:22:05,710 S1: if you connect to them, give you the most accurate 371 00:22:05,740 --> 00:22:09,010 S1: time that you can get on the planet, essentially, um, 372 00:22:09,190 --> 00:22:13,390 S1: because all their clocks are synced and they're all working 373 00:22:13,390 --> 00:22:16,060 S1: in conjunction and they agree on what the time is, 374 00:22:16,090 --> 00:22:19,120 S1: and then they provide that time to the world. And 375 00:22:19,119 --> 00:22:22,120 S1: this is how we have accuracy with GPS and everything. 376 00:22:22,119 --> 00:22:26,290 S1: So if you have a time server like time.com or whatever, 377 00:22:26,290 --> 00:22:29,890 S1: that is usually a I can't remember if that's stratum one. 378 00:22:29,920 --> 00:22:34,510 S1: Let me see. Okay. Stratum one. Yeah. Stratum one is 379 00:22:34,510 --> 00:22:38,530 S1: the original one. It's the one that has access to 380 00:22:38,560 --> 00:22:42,100 S1: actual satellites. So you see that S1 right there. That 381 00:22:42,100 --> 00:22:45,399 S1: S1 means it's the stratum one server. So look at this. 382 00:22:45,400 --> 00:22:53,229 S1: I am 0.001680 milliseconds off of the true time on 383 00:22:53,230 --> 00:22:57,159 S1: the satellites. And if I do mine, which is time, 384 00:22:57,160 --> 00:22:58,419 S1: I don't want to do it because it's got my 385 00:22:58,420 --> 00:23:00,879 S1: IP address in there. Not really a security risk, just 386 00:23:00,910 --> 00:23:03,370 S1: kind of stupid to do on a video. But um, 387 00:23:03,369 --> 00:23:07,300 S1: mine is even more accurate because it's actually running from 388 00:23:07,300 --> 00:23:11,350 S1: my server room and it's got an actual antenna talking 389 00:23:11,380 --> 00:23:14,440 S1: to actual satellites. I just love that I've always been 390 00:23:14,440 --> 00:23:17,649 S1: obsessed with time. So bottom line, if you want to 391 00:23:17,650 --> 00:23:22,300 S1: get into some super geeky time stuff for only like 392 00:23:22,330 --> 00:23:25,270 S1: 2 or 300 bucks, you should get one of these things. 393 00:23:25,270 --> 00:23:28,570 S1: They are super cool. Look at this United Airlines, which 394 00:23:28,570 --> 00:23:34,300 S1: is my favorite airline. They are adding share item location 395 00:23:34,310 --> 00:23:38,000 S1: into the mobile app. So if you lose your luggage, 396 00:23:38,000 --> 00:23:42,050 S1: you can actually add your link to your AirTag and 397 00:23:42,050 --> 00:23:44,840 S1: help United find your bag. That is the coolest thing ever. 398 00:23:44,869 --> 00:23:47,929 S1: YouTube is seeing massive growth in TV viewing, with sports 399 00:23:47,930 --> 00:23:53,300 S1: content up 30%. Users watching over 400,000,000 hours of podcasts 400 00:23:53,300 --> 00:23:58,010 S1: on TV's monthly YouTube. Just absolutely killing it. Someone in 401 00:23:58,010 --> 00:24:01,310 S1: Louisiana got a bird flu and they are hospitalized. I 402 00:24:01,310 --> 00:24:04,910 S1: actually heard a conflicting report. Somebody was saying or some 403 00:24:04,910 --> 00:24:07,970 S1: report was saying 60% of people die. I've heard a 404 00:24:07,970 --> 00:24:10,609 S1: lot of other people. It might have been this article 405 00:24:10,609 --> 00:24:12,379 S1: that said a lot of people got better from it. 406 00:24:12,380 --> 00:24:17,300 S1: So it wasn't like killing everybody. So maybe we don't 407 00:24:17,300 --> 00:24:20,360 S1: have a pandemic, maybe we do, and it's more like 408 00:24:20,359 --> 00:24:22,850 S1: Covid and it's not so bad because we have, you know, 409 00:24:22,880 --> 00:24:25,910 S1: vaccines come out. I mean, we already have flu vaccines, 410 00:24:25,910 --> 00:24:28,610 S1: so maybe, maybe it won't be so bad. I don't 411 00:24:28,640 --> 00:24:32,720 S1: know don't really want to think about it, honestly. Tokyo 412 00:24:32,720 --> 00:24:37,240 S1: is trying to increase births by giving its 160,000 government 413 00:24:37,240 --> 00:24:41,740 S1: workers a four day workweek starting in April. I mean 414 00:24:41,770 --> 00:24:46,270 S1: that the birth rate is down to 1.2 babies per woman, 415 00:24:46,270 --> 00:24:49,600 S1: and they're trying to get that that back up. Um, 416 00:24:49,630 --> 00:24:55,030 S1: expecting fewer than 700,000 newborns this year, lowest since records 417 00:24:55,030 --> 00:24:58,930 S1: started in 1899. So they're trying to give people a 418 00:24:58,930 --> 00:25:03,790 S1: four day work weeks. So basically a three day weekend. Hopefully, 419 00:25:03,820 --> 00:25:06,639 S1: you know, go out, have a good time, uh, meet 420 00:25:06,640 --> 00:25:10,030 S1: somebody and make a baby. Research shows ketones don't just 421 00:25:10,030 --> 00:25:15,100 S1: provide energy to the brain. They actually help remove misfolded proteins. Okay, 422 00:25:15,130 --> 00:25:19,810 S1: I actually have my buddy Danny gave me some ketone drinks. 423 00:25:19,810 --> 00:25:22,149 S1: I don't know, a little bit cautious of that. I'm 424 00:25:22,150 --> 00:25:24,220 S1: going to wait till the science is a little bit better. 425 00:25:24,250 --> 00:25:28,540 S1: All right. Discovery Stack Analyzer detects more than 500 technologies 426 00:25:28,540 --> 00:25:30,520 S1: in your code base. You basically point it to a 427 00:25:30,550 --> 00:25:33,070 S1: GitHub repo, and it tells you all the different tech 428 00:25:33,100 --> 00:25:36,220 S1: that it uses, all the different security talks at Reinvent. 429 00:25:36,250 --> 00:25:40,210 S1: Thanks to Clint for showing me this link. And in 430 00:25:40,210 --> 00:25:45,910 S1: a series of compression tests, Zstd consistently outperformed gzip and 431 00:25:45,910 --> 00:25:52,150 S1: zlib across speed, compression and decompression efficiency. Metrics and recommendation 432 00:25:52,150 --> 00:25:58,660 S1: of the week downtime, fiction, family, friends, downtime, fiction, family, friends, 433 00:25:58,660 --> 00:26:02,980 S1: and the aphorism of the week. Almost everything will work 434 00:26:03,010 --> 00:26:06,490 S1: again if you unplug it for a few minutes, including you. 435 00:26:06,520 --> 00:26:10,540 S1: Almost everything will work again if you unplug it for 436 00:26:10,540 --> 00:26:15,520 S1: a few minutes, including you, Anne Lamott. Unsupervised learning is 437 00:26:15,520 --> 00:26:18,610 S1: produced and edited by Daniel Miessler on a Neumann U87 438 00:26:18,640 --> 00:26:21,940 S1: AI microphone using Hindenburg. Intro and outro music is by 439 00:26:21,940 --> 00:26:24,189 S1: Zomby with a Y, and to get the text and 440 00:26:24,190 --> 00:26:26,560 S1: links from this episode, sign up for the newsletter version 441 00:26:26,560 --> 00:26:31,090 S1: of the show at Daniel miessler.com/newsletter. We'll see you next time.