1 00:00:01,280 --> 00:00:04,609 S1: Welcome to Unsupervised Learning, a security, AI and meaning focused 2 00:00:04,610 --> 00:00:07,460 S1: podcast that looks at how best to thrive as humans 3 00:00:07,460 --> 00:00:11,719 S1: in a post AI world. It combines original ideas, analysis, 4 00:00:11,720 --> 00:00:14,930 S1: and mental models to bring not just the news, but 5 00:00:14,930 --> 00:00:22,279 S1: why it matters and how to respond. All right, welcome 6 00:00:22,280 --> 00:00:25,400 S1: to unsupervised learning. This is Daniel Miessler. This is episode 7 00:00:25,400 --> 00:00:28,670 S1: 459 of the podcast. All right. So had a great 8 00:00:28,670 --> 00:00:32,600 S1: conversation with Rob Allen from Threatlocker about their approach to 9 00:00:32,630 --> 00:00:34,820 S1: Zero Trust. And it turned out to be a really 10 00:00:34,820 --> 00:00:39,290 S1: interesting conversation. I always like to dive into what somebody 11 00:00:39,290 --> 00:00:41,630 S1: means when they say zero trust, because it could be 12 00:00:41,630 --> 00:00:46,370 S1: anywhere from total horse crap to pretty decent, and I 13 00:00:46,370 --> 00:00:50,990 S1: really liked Rob's approach in responses there, so definitely check 14 00:00:50,990 --> 00:00:54,530 S1: that out. Got you all membership. Black Friday sale just 15 00:00:54,530 --> 00:01:00,120 S1: started 20% off the first month, so or the first year, 16 00:01:00,120 --> 00:01:04,620 S1: so go check that out. And jumping right into security. Actually, 17 00:01:04,620 --> 00:01:06,930 S1: first of all, um, did a huge upgrade to my 18 00:01:06,930 --> 00:01:11,100 S1: ubiquiti gear now moving towards a ten gigabit world, which 19 00:01:11,100 --> 00:01:14,309 S1: I talked a bit about last week, and going to 20 00:01:14,340 --> 00:01:19,860 S1: Saudi to speak at Blackhat EMEA. Mia Blackhat Mia. Yeah 21 00:01:19,890 --> 00:01:24,030 S1: Saudi Arabia next week going to be awesome. All right. 22 00:01:24,030 --> 00:01:27,420 S1: Jumping into security. This one didn't get near enough coverage. 23 00:01:27,420 --> 00:01:30,660 S1: I don't think ChatGPT has basically a new feature that 24 00:01:30,660 --> 00:01:35,820 S1: could read code from macOS. So apps like VSCode, Xcode, terminal, 25 00:01:35,850 --> 00:01:40,440 S1: it could basically read the app and understand what it's 26 00:01:40,440 --> 00:01:44,040 S1: saying and doing, and then give you analysis based on that. 27 00:01:44,040 --> 00:01:46,680 S1: And I think this is getting closer to what a 28 00:01:46,680 --> 00:01:50,190 S1: whole bunch of people in AI are actually working on, 29 00:01:50,220 --> 00:01:54,660 S1: which is what I believe to be like an inevitable future, 30 00:01:54,660 --> 00:01:57,610 S1: which is you can see all the screens, right? I'm 31 00:01:57,610 --> 00:02:00,610 S1: looking at three screens right now. Ideally my I would 32 00:02:00,610 --> 00:02:03,100 S1: be looking from inside the computer because it would have 33 00:02:03,130 --> 00:02:06,309 S1: like agent access like we already talked about. But also 34 00:02:06,310 --> 00:02:08,290 S1: I wanted to be able to see my surroundings. I 35 00:02:08,290 --> 00:02:10,750 S1: wanted to be able to see my cameras. I wanted 36 00:02:10,750 --> 00:02:13,840 S1: to be able to see around me, behind me, especially 37 00:02:13,840 --> 00:02:15,910 S1: behind me. If you're out in the city walking around 38 00:02:15,910 --> 00:02:20,770 S1: and also cameras, um, or also microphones, right. And other 39 00:02:20,770 --> 00:02:25,270 S1: types of sensor, but mostly visual and audio. And that 40 00:02:25,270 --> 00:02:28,869 S1: way it can be consuming, you know, the environment all 41 00:02:28,870 --> 00:02:32,920 S1: the time, constantly parsing what's going on and constantly doing 42 00:02:32,919 --> 00:02:36,280 S1: analysis and giving you like, views inside of a heads 43 00:02:36,310 --> 00:02:41,260 S1: up display or, you know, glasses or lenses or however 44 00:02:41,260 --> 00:02:43,510 S1: that interface is going to be, depending on how far 45 00:02:43,510 --> 00:02:46,300 S1: the tech is. But the most important thing is it 46 00:02:46,300 --> 00:02:48,850 S1: has to be seeing what you're seeing and even more 47 00:02:48,880 --> 00:02:52,149 S1: than what you're seeing. Big problem right now is the 48 00:02:52,150 --> 00:02:54,700 S1: on ramp to using AI. It's not a problem of 49 00:02:54,700 --> 00:02:57,090 S1: the models themselves. The models are getting really smart. The 50 00:02:57,090 --> 00:02:59,130 S1: problem is, how do you get stuff into the model 51 00:02:59,130 --> 00:03:01,470 S1: and how do you get it out? That's the problem. 52 00:03:01,500 --> 00:03:04,680 S1: So this type of thing is a really huge advance 53 00:03:04,680 --> 00:03:07,950 S1: because it's able to see inside your apps. And I'm 54 00:03:07,950 --> 00:03:10,950 S1: really excited about this future. I've talked about it a lot. 55 00:03:10,980 --> 00:03:14,100 S1: I've talked about it in the predictable path of AI. 56 00:03:14,130 --> 00:03:17,520 S1: It's just that I'm not really going to trust this. 57 00:03:17,520 --> 00:03:20,250 S1: There's already a couple of companies who will watch your 58 00:03:20,250 --> 00:03:23,190 S1: entire screen and will upload that stuff, and they will 59 00:03:23,190 --> 00:03:25,560 S1: do the parsing and they will send you back some 60 00:03:25,560 --> 00:03:29,760 S1: really cool stuff. Like there's one particular company, I think 61 00:03:29,760 --> 00:03:32,459 S1: they changed their name. It's called rewind. Or maybe rewind 62 00:03:32,490 --> 00:03:35,670 S1: is the new name. But anyway, I signed up for 63 00:03:35,670 --> 00:03:39,480 S1: their new gadget, which should be coming soon. I did 64 00:03:39,480 --> 00:03:43,080 S1: not give them access to my whole computer because I 65 00:03:43,080 --> 00:03:45,690 S1: could have the calendar open or a message or something, 66 00:03:45,690 --> 00:03:48,030 S1: and it's like you could be doxing your friends. You 67 00:03:48,030 --> 00:03:50,400 S1: don't know what's going to happen. Most importantly, you don't 68 00:03:50,430 --> 00:03:54,580 S1: know how secure that startup is and having consulted for 69 00:03:54,580 --> 00:03:59,680 S1: startups in security specifically for decades. I'm not sending my 70 00:03:59,680 --> 00:04:02,950 S1: data to those startups. I basically only trust like one 71 00:04:02,950 --> 00:04:06,340 S1: company that much, both on the security side and also 72 00:04:06,340 --> 00:04:09,310 S1: on the privacy side, and just how seriously they take 73 00:04:09,310 --> 00:04:12,490 S1: this stuff, combined with a lack of conflicts of interest. 74 00:04:12,490 --> 00:04:16,839 S1: And that's Apple. I would say Google is just as competent, 75 00:04:16,870 --> 00:04:20,650 S1: maybe even more competent on the security side, I would say. 76 00:04:20,680 --> 00:04:24,490 S1: But the problem is they also make their money off 77 00:04:24,490 --> 00:04:27,190 S1: of ads or a lot of their money off of ads. 78 00:04:27,190 --> 00:04:30,490 S1: So I just don't like the idea of like them 79 00:04:30,490 --> 00:04:32,710 S1: having all that data and then being able to sell 80 00:04:32,710 --> 00:04:35,470 S1: it or sell things to me. I don't like the 81 00:04:35,470 --> 00:04:38,650 S1: conflict of interest there, even though I like the security. 82 00:04:38,650 --> 00:04:41,740 S1: So that's why I mostly only trust two companies and 83 00:04:41,740 --> 00:04:45,190 S1: mostly only Apple. But this is where this is going. 84 00:04:45,190 --> 00:04:50,200 S1: Continuous monitoring all the time with as many sensors as possible. Front. 85 00:04:50,200 --> 00:04:53,960 S1: Back inside your apps. Inside your phone. That's where everyone 86 00:04:53,960 --> 00:04:56,270 S1: is rushing to. Even if they don't know that they're 87 00:04:56,270 --> 00:04:59,720 S1: rushing to that location. And that's why I think Google 88 00:04:59,720 --> 00:05:04,070 S1: and Apple have a massive advantage because they have the device, right? 89 00:05:04,100 --> 00:05:08,480 S1: I'm on Apple for iPhone, I'm on Apple for Mac 90 00:05:08,480 --> 00:05:13,580 S1: and desktop and laptop. So because they are the OS, 91 00:05:13,610 --> 00:05:17,210 S1: they have a massive advantage here okay. Palo Alto Networks 92 00:05:17,210 --> 00:05:20,690 S1: has released indicators of compromise IOCs for a new zero 93 00:05:20,720 --> 00:05:25,310 S1: day vulnerability affecting firewalls. VMware confirmed that threat actors are 94 00:05:25,310 --> 00:05:29,660 S1: exploiting two vCenter server vulnerabilities, and ones like a 9.8 95 00:05:29,690 --> 00:05:31,820 S1: and the other ones like a seven something I. And 96 00:05:31,820 --> 00:05:34,880 S1: Tech Anthropic has a new prompt improver that takes a 97 00:05:34,880 --> 00:05:37,580 S1: given prompt and writes a better one. Really cool to 98 00:05:37,610 --> 00:05:40,669 S1: see people getting in on this. This is part of 99 00:05:40,670 --> 00:05:44,930 S1: the overall ecosystem, right? Um, and we got another one here. Uh, 100 00:05:44,930 --> 00:05:49,010 S1: OpenAI might launch an AI agent tool called Operator in January, 101 00:05:49,010 --> 00:05:52,170 S1: and it will compete with Anthropic's computer use. I think 102 00:05:52,200 --> 00:05:54,510 S1: agents is going to be the biggest thing that happens 103 00:05:54,510 --> 00:05:57,719 S1: in 2025 for AI, and this is part of a 104 00:05:57,720 --> 00:06:00,330 S1: bigger trend that I've been talking about, where it's more 105 00:06:00,330 --> 00:06:03,210 S1: about the ecosystem. Okay, so if I if I actually 106 00:06:03,210 --> 00:06:07,800 S1: click and open this thing here, it's not just about 107 00:06:07,800 --> 00:06:11,910 S1: the models. Okay. So the big pieces that I think 108 00:06:11,910 --> 00:06:15,900 S1: are part of these four pillars of an AI ecosystem, 109 00:06:15,900 --> 00:06:19,650 S1: the model itself, the post training of the model internal 110 00:06:19,680 --> 00:06:25,200 S1: tooling and agent functionality. So agent or model self-explanatory. Post 111 00:06:25,230 --> 00:06:28,800 S1: training a set of highly proprietary tricks that magnify the 112 00:06:28,800 --> 00:06:31,770 S1: overall quality of the raw model. Way to think of 113 00:06:31,770 --> 00:06:33,510 S1: this is to say that it's a way to connect 114 00:06:33,540 --> 00:06:37,590 S1: model weights to human problems. Okay. Internal tooling. All right. 115 00:06:37,620 --> 00:06:42,180 S1: So look at this list. High quality APIs, larger context sizes, 116 00:06:42,210 --> 00:06:48,920 S1: simple fine tuning, haystack performance, strict output control. External tooling 117 00:06:48,920 --> 00:06:53,600 S1: like function calling. Trust and safety features. Mobile apps. Prompt testing. 118 00:06:53,600 --> 00:06:57,800 S1: Voice mode and apps OS integration integrations with things like 119 00:06:57,800 --> 00:07:02,270 S1: make Zapier end to end and things like caching mode. 120 00:07:02,270 --> 00:07:05,360 S1: These are like all the internal tooling stuff that just 121 00:07:05,360 --> 00:07:08,090 S1: makes it easier. Think of it this way the problem 122 00:07:08,089 --> 00:07:10,880 S1: isn't the models. The problem is the on ramps onto 123 00:07:10,880 --> 00:07:13,550 S1: the models and the output out of them back into 124 00:07:13,550 --> 00:07:16,910 S1: your life. Okay, we are humans. We have human problems. 125 00:07:16,910 --> 00:07:19,730 S1: We have business problems, we have personal problems, whatever. We 126 00:07:19,730 --> 00:07:22,970 S1: need to get that content, the content of that problem 127 00:07:22,970 --> 00:07:26,180 S1: into an AI and then back out into our lives, 128 00:07:26,180 --> 00:07:29,120 S1: into our actual brains, into the real world. That's what 129 00:07:29,120 --> 00:07:32,390 S1: this internal tooling piece is, right? Because if you can't 130 00:07:32,390 --> 00:07:35,840 S1: do this, well, then it doesn't matter if your model 131 00:07:35,840 --> 00:07:41,000 S1: is 13% better on some random benchmark, right? So the 132 00:07:41,000 --> 00:07:45,560 S1: models have to get better, but not as important as 133 00:07:45,590 --> 00:07:50,250 S1: actually getting the interfaces to the models better. So, uh. Oh, yeah. 134 00:07:50,280 --> 00:07:53,070 S1: And the next one relevant to the next story there 135 00:07:53,070 --> 00:07:58,170 S1: is agents. Right. So an AI component that interprets instructions 136 00:07:58,170 --> 00:08:00,930 S1: and takes on more of the work in total AI 137 00:08:00,960 --> 00:08:06,090 S1: workflows than just LLM response, for example, executing functions, performing 138 00:08:06,090 --> 00:08:09,930 S1: data lookups, etc. before passing on results. And I actually 139 00:08:09,930 --> 00:08:12,900 S1: have a improved version of that, which I won't I 140 00:08:12,900 --> 00:08:15,810 S1: won't go find it, but it's, uh, it's updated on 141 00:08:15,810 --> 00:08:22,620 S1: the AGI definition inside of Raid. The real world AI definitions, 142 00:08:22,620 --> 00:08:24,900 S1: if you want to go check that out. Okay, so 143 00:08:24,990 --> 00:08:29,160 S1: Sam Altman and Arianna Huffington have a thrive AI health 144 00:08:29,160 --> 00:08:34,470 S1: company and, uh, looking at doing personalized advice on sleep, food, 145 00:08:34,470 --> 00:08:38,910 S1: fitness and more, Google.org is putting 20 million in cash 146 00:08:38,910 --> 00:08:41,939 S1: and 2 million in cloud credits into a new initiative 147 00:08:41,970 --> 00:08:46,540 S1: to help researchers use AI for scientific breakthroughs. One of 148 00:08:46,540 --> 00:08:50,800 S1: the most important things that I think I could possibly 149 00:08:50,800 --> 00:08:56,710 S1: do is actively going to just invent new things, make 150 00:08:56,710 --> 00:09:03,250 S1: new research right. Discover new things right. You take the 151 00:09:03,250 --> 00:09:05,890 S1: smartest people in the world who who are capable of 152 00:09:05,890 --> 00:09:09,280 S1: doing this, and there's very few of them. So if 153 00:09:09,280 --> 00:09:12,820 S1: we can actually scale that, that's where the real take 154 00:09:12,850 --> 00:09:16,810 S1: off starts to happen. Apple's M4 Max CPU transcribes audio 155 00:09:16,809 --> 00:09:22,569 S1: twice as fast as Nvidia's RTX A5000 GPU while using 156 00:09:22,570 --> 00:09:25,959 S1: significantly less power. I really want to get one of 157 00:09:25,960 --> 00:09:31,960 S1: these clusters, like a m4 Mac mini cluster would be 158 00:09:31,960 --> 00:09:34,929 S1: super cool. But yeah, a lot of people have been 159 00:09:34,960 --> 00:09:38,020 S1: asking me, should I do the cluster thing with smaller boxes, 160 00:09:38,020 --> 00:09:40,420 S1: or should I just get one big rig? I feel 161 00:09:40,420 --> 00:09:43,819 S1: like the next generation or the current next the current 162 00:09:43,820 --> 00:09:47,570 S1: generation in like another year or so, it's still going 163 00:09:47,600 --> 00:09:49,430 S1: to be better to have one big box. It really 164 00:09:49,429 --> 00:09:51,650 S1: depends on what you need it for, but I would 165 00:09:51,650 --> 00:09:55,430 S1: say we're not quite there yet. With the cluster of 166 00:09:55,429 --> 00:09:59,330 S1: smaller boxes, this exo lab and things like that, they're 167 00:09:59,330 --> 00:10:02,900 S1: a little bit experimental. The best thing to do still 168 00:10:02,900 --> 00:10:06,440 S1: is to buy a box. I bought a premade one 169 00:10:06,440 --> 00:10:09,560 S1: and it's fantastic. It's got two 49 seconds in it. 170 00:10:09,590 --> 00:10:13,820 S1: It's got tons of memory. It's very, very fast. And 171 00:10:13,820 --> 00:10:16,880 S1: it's just like plug and play. Easy to do. I 172 00:10:16,880 --> 00:10:21,950 S1: would say if this clustering technology gets better and better 173 00:10:21,950 --> 00:10:25,130 S1: and there's more and more devices like Mac minis and 174 00:10:25,130 --> 00:10:27,980 S1: stuff like that that you could piece together, then it 175 00:10:27,980 --> 00:10:31,219 S1: starts to be a compelling alternative to actually buying a 176 00:10:31,220 --> 00:10:35,660 S1: giant AI box. But until then, I think the giant 177 00:10:35,660 --> 00:10:39,140 S1: box is probably still going to be better. Okay, iOS 178 00:10:39,140 --> 00:10:43,150 S1: 18.2 is music recognition feature. Now logs where you were 179 00:10:43,150 --> 00:10:45,910 S1: when you actually heard the song. So that's part of 180 00:10:45,910 --> 00:10:48,849 S1: the metadata. Now pharma stocks have crashed. This is under 181 00:10:48,850 --> 00:10:53,410 S1: the humans label. Pharma stocks have crashed. After RFK Jr 182 00:10:53,470 --> 00:10:56,440 S1: was announced to be taking over Health and Human Services. 183 00:10:56,440 --> 00:10:59,380 S1: Moderna is down like 40%. Yeah, I don't know what 184 00:10:59,380 --> 00:11:02,679 S1: this is currently at uh, down five, six, two, but 185 00:11:02,679 --> 00:11:05,470 S1: that's for the day. So if we go to three months, 186 00:11:05,590 --> 00:11:12,790 S1: down 57% in the last three months, Moderna. Okay. They're 187 00:11:12,820 --> 00:11:15,010 S1: they're one of the people who came out with the 188 00:11:15,010 --> 00:11:20,559 S1: best vaccine for Covid, like 57% because of RFK. I 189 00:11:20,559 --> 00:11:23,439 S1: almost feel like this is a buy opportunity. How could 190 00:11:23,440 --> 00:11:26,350 S1: this not be a buy opportunity? It's not like they 191 00:11:26,380 --> 00:11:30,370 S1: suddenly stopped being able to make things. And I just 192 00:11:30,370 --> 00:11:33,550 S1: don't believe that people are going to let RFK just 193 00:11:33,550 --> 00:11:38,020 S1: destroy these companies. I just don't see that happening. Hopefully 194 00:11:38,020 --> 00:11:40,460 S1: they're going to be able to figure out because RFK 195 00:11:40,460 --> 00:11:43,100 S1: is going to do some cool stuff. Like he's right 196 00:11:43,100 --> 00:11:45,020 S1: about a lot of stuff. That's what's most scary about 197 00:11:45,020 --> 00:11:47,510 S1: a lot of these people. They're so right about so 198 00:11:47,510 --> 00:11:51,050 S1: many things, and the things they're wrong about are mixed 199 00:11:51,050 --> 00:11:53,960 S1: in with it. Right. And that's not just RFK. That's 200 00:11:53,960 --> 00:11:58,400 S1: a lot of people, including probably myself. So, yeah, I 201 00:11:58,400 --> 00:12:03,350 S1: don't see how it doesn't bounce back from a 60% 202 00:12:03,380 --> 00:12:08,059 S1: cut from three months. Right. Personally, I'm not investing, but 203 00:12:08,059 --> 00:12:10,520 S1: I think it would probably be smart, at least in 204 00:12:10,520 --> 00:12:15,470 S1: the long term. Netflix had a record 65 million concurrent 205 00:12:15,470 --> 00:12:19,250 S1: streams during Mike Tyson versus Jake Paul fight. It did 206 00:12:19,250 --> 00:12:21,679 S1: have a bunch of connection problems, though. Everyone I know 207 00:12:21,679 --> 00:12:25,400 S1: at Netflix got massively spammed with like, all their friends 208 00:12:25,400 --> 00:12:27,440 S1: texting and was like, hey, what's going on? It's like, yeah, 209 00:12:27,440 --> 00:12:30,800 S1: you think I'm in charge of actual throughput during a 210 00:12:30,800 --> 00:12:34,220 S1: Tyson fight? Like, stop texting me. New study shows that 211 00:12:34,220 --> 00:12:37,370 S1: treating bullying as a collective issue rather than an individual 212 00:12:37,370 --> 00:12:41,280 S1: one could significantly reduce its occurrence in primary schools. I 213 00:12:41,280 --> 00:12:44,790 S1: love the concept. It's kind of like how the johns 214 00:12:44,820 --> 00:12:48,750 S1: get in trouble in some European countries for prostitution instead 215 00:12:48,750 --> 00:12:52,350 S1: of the prostitute, because it's like the ecosystem that's the 216 00:12:52,350 --> 00:12:56,610 S1: actual problem, right? So I like this idea of with bullying, 217 00:12:56,610 --> 00:12:59,309 S1: I like the idea of shaming the people around who 218 00:12:59,309 --> 00:13:01,679 S1: didn't say anything or do anything. And of course, we 219 00:13:01,679 --> 00:13:03,630 S1: have to be careful about shaming. We're talking about kids 220 00:13:03,630 --> 00:13:05,460 S1: here in a lot of cases or in most cases. 221 00:13:05,460 --> 00:13:09,689 S1: But like, I feel like all the marketing needs to 222 00:13:09,690 --> 00:13:12,209 S1: be heading in the direction of like if you were 223 00:13:12,210 --> 00:13:14,550 S1: one of these people watching and not saying anything, not 224 00:13:14,550 --> 00:13:18,120 S1: reporting it, not intervening like you don't want to be 225 00:13:18,120 --> 00:13:20,429 S1: unsafe or whatever, but you could go report it. You 226 00:13:20,429 --> 00:13:24,660 S1: could do something to prevent this from happening again. And 227 00:13:24,660 --> 00:13:27,960 S1: if you're not something like you are the bully or 228 00:13:27,960 --> 00:13:30,329 S1: you are enabling this and this is really bad, and 229 00:13:30,330 --> 00:13:32,640 S1: I know this is already an aspect of like a 230 00:13:32,670 --> 00:13:34,980 S1: lot of these programs, but I think it could be 231 00:13:34,980 --> 00:13:40,640 S1: significantly magnified ideas. Reboot I oh, I absolutely love this one. 232 00:13:40,670 --> 00:13:43,579 S1: Absolutely love this one. I can't remember where I got 233 00:13:43,580 --> 00:13:46,370 S1: this idea. I think it's been hit me from multiple places, 234 00:13:46,370 --> 00:13:49,670 S1: but I want to build a local AI that can 235 00:13:49,670 --> 00:13:52,850 S1: run offline. Oh, I know where I first heard this idea. 236 00:13:52,850 --> 00:13:56,390 S1: It was actually from Joseph Thacker, like a year and 237 00:13:56,390 --> 00:13:58,730 S1: a half ago. He's like, oh, I just want a 238 00:13:58,730 --> 00:14:02,600 S1: thing that I can use offline. But, um, more recently 239 00:14:02,600 --> 00:14:05,180 S1: I got this from somewhere else I can't was it, 240 00:14:05,210 --> 00:14:09,680 S1: was it X or Instagram? I don't know, but the 241 00:14:09,679 --> 00:14:12,740 S1: idea is let's say all the power is out, or 242 00:14:12,770 --> 00:14:16,880 S1: let's say the internet is out and let's say, um, oh, 243 00:14:16,910 --> 00:14:20,300 S1: I know what it was. The initial conversation I was 244 00:14:20,300 --> 00:14:24,140 S1: having with, uh, Joseph, and he actually brought it to me. 245 00:14:24,170 --> 00:14:27,020 S1: It was his idea. He was like, what if you 246 00:14:27,020 --> 00:14:30,020 S1: could go back in time? What could you bring with 247 00:14:30,020 --> 00:14:33,560 S1: you to actually move society forward or something like that? 248 00:14:33,560 --> 00:14:35,850 S1: And it's like, I think about this a lot way 249 00:14:35,850 --> 00:14:38,820 S1: more than the Roman Empire. I think a lot about like, 250 00:14:38,850 --> 00:14:41,790 S1: could I actually move science forward if I was put 251 00:14:41,790 --> 00:14:44,729 S1: 200 years in the past or 2000 years in the past? 252 00:14:44,760 --> 00:14:47,400 S1: What could I actually offer to them? I think someone 253 00:14:47,400 --> 00:14:50,010 S1: makes a joke about this in current stand up comedy. 254 00:14:50,010 --> 00:14:55,140 S1: It's like, you know, oh, it's Nate Bathgate, Bargatze, whatever 255 00:14:55,170 --> 00:14:58,740 S1: his name is. Um, it's just like, yeah, um, there's 256 00:14:58,740 --> 00:15:02,610 S1: going to be phones and they're going to have satellite technology. Oh, really? 257 00:15:02,610 --> 00:15:05,130 S1: What's a satellite? Well, it's this thing that goes around. 258 00:15:05,310 --> 00:15:07,140 S1: How do you make one? What does it do? You're 259 00:15:07,140 --> 00:15:10,110 S1: saying there's thousands of satellites in the air and the 260 00:15:10,110 --> 00:15:14,040 S1: Earth is actually round? Like, can you prove anything about this? No, 261 00:15:14,040 --> 00:15:18,030 S1: I can't. So we we don't know how any of 262 00:15:18,030 --> 00:15:22,530 S1: this is working, right? And, uh, and we can't describe 263 00:15:22,530 --> 00:15:25,260 S1: it to anyone. And more importantly, if you lose the 264 00:15:25,260 --> 00:15:29,100 S1: internet and there's, like, say, a meteor hits or whatever, 265 00:15:29,100 --> 00:15:31,710 S1: I'm not going to go into negativity right now too 266 00:15:31,710 --> 00:15:34,060 S1: early in the show for that. But let's say something 267 00:15:34,060 --> 00:15:36,940 S1: bad happens and you are stuck in your house and 268 00:15:36,940 --> 00:15:40,060 S1: there's no internet. Let's say you're not dying of hunger 269 00:15:40,060 --> 00:15:43,240 S1: or thirst, but you don't have the internet. Okay, so 270 00:15:43,240 --> 00:15:49,330 S1: watch this. Tourniquets, sterilizing water, building shelters, identifying edible plants 271 00:15:49,330 --> 00:15:53,080 S1: which mushroom is actually will kill you. And which mushroom 272 00:15:53,080 --> 00:15:56,440 S1: can you put on a salad? These are important distinctions 273 00:15:56,440 --> 00:15:59,500 S1: to make. So check this out. What if you had 274 00:15:59,500 --> 00:16:03,640 S1: an AI that ran? We got to assume solar power, right? 275 00:16:03,670 --> 00:16:06,670 S1: Or maybe the grid actually works, but there's no internet. Whatever. 276 00:16:06,670 --> 00:16:09,520 S1: Just work with me. You can show it. Pictures. You 277 00:16:09,520 --> 00:16:12,190 S1: could take pictures, or you could show it an actual 278 00:16:12,190 --> 00:16:14,140 S1: live plant or whatever. All you have to do is 279 00:16:14,140 --> 00:16:17,560 S1: show this AI the particular thing or you describe, hey, 280 00:16:17,560 --> 00:16:20,440 S1: I need a shelter that does this. I've got this 281 00:16:20,440 --> 00:16:22,750 S1: much water. I got it from this kind of creek. 282 00:16:22,780 --> 00:16:25,330 S1: I live in this kind of area. What kind of 283 00:16:25,330 --> 00:16:27,340 S1: toxins are likely to be in it? How can I 284 00:16:27,340 --> 00:16:30,070 S1: get those toxins out? What kind of filter can I build? 285 00:16:30,070 --> 00:16:34,100 S1: Will these iodine tablets actually work? I have these symptoms. 286 00:16:34,100 --> 00:16:38,300 S1: Which drug should I take? You have a local model running. 287 00:16:38,300 --> 00:16:40,910 S1: All that needs is power. Doesn't need the internet. Doesn't 288 00:16:40,910 --> 00:16:44,960 S1: need OpenAI. Think of how much knowledge is inside of 289 00:16:44,960 --> 00:16:48,350 S1: a llama two or a llama three or a llama 290 00:16:48,350 --> 00:16:52,040 S1: four or whatever. Local model. And as long as you 291 00:16:52,040 --> 00:16:54,590 S1: have power to run it and it can actually see 292 00:16:54,620 --> 00:16:57,590 S1: and you can type to it, it can answer tons 293 00:16:57,590 --> 00:17:01,760 S1: of stuff that can actually keep you alive. Even better, 294 00:17:01,790 --> 00:17:04,760 S1: let's go a little sci fi. Okay. It's just you 295 00:17:04,760 --> 00:17:08,990 S1: and 10,000 other people and the rest of the, let's say, 296 00:17:09,020 --> 00:17:11,060 S1: the rest of the planet got hit by a meteor 297 00:17:11,090 --> 00:17:16,040 S1: or whatever. You have to rebuild all of society. Okay. Irrigation. 298 00:17:16,040 --> 00:17:18,889 S1: How to how does the stoplight work? How does a 299 00:17:18,890 --> 00:17:22,399 S1: combustion engine work? What kind of metal do you need 300 00:17:22,400 --> 00:17:25,639 S1: to build in order to make a combustion engine? What 301 00:17:25,640 --> 00:17:30,850 S1: are all these different alloys you can bootstrap a society 302 00:17:30,850 --> 00:17:34,479 S1: with one box. Isn't that crazy? You can bootstrap a 303 00:17:34,480 --> 00:17:38,500 S1: society with one box that you can have, like sitting 304 00:17:38,500 --> 00:17:41,859 S1: next to your MNAs over here. You just need power. 305 00:17:41,859 --> 00:17:43,840 S1: You need power. Well, I guess you need the peripherals 306 00:17:43,840 --> 00:17:46,000 S1: to be able to talk to it and everything, but 307 00:17:46,000 --> 00:17:51,070 S1: you don't need that much. It's way more impressive than 308 00:17:51,070 --> 00:17:53,770 S1: trying to collect all of Wikipedia. I mean, that was 309 00:17:53,770 --> 00:17:56,080 S1: the other model, right? You just download Wikipedia and you 310 00:17:56,080 --> 00:17:58,480 S1: have a bunch of things you can look for. Not 311 00:17:58,480 --> 00:18:01,030 S1: nearly as good as a chat bot that you can 312 00:18:01,030 --> 00:18:05,590 S1: ask questions. And if it's visually oriented, like you could 313 00:18:05,590 --> 00:18:08,620 S1: just these new models, they're getting amazing, right? It's like, 314 00:18:08,650 --> 00:18:11,470 S1: show me the design. Show me a picture of it, 315 00:18:12,010 --> 00:18:16,570 S1: draw me a picture, design me whatever a compound that 316 00:18:16,570 --> 00:18:19,600 S1: I could defend my plants with because people are probably 317 00:18:19,600 --> 00:18:22,780 S1: going to come and get us whatever. Yeah, I don't 318 00:18:22,780 --> 00:18:25,900 S1: have any guns. How do I defend myself from these 319 00:18:25,900 --> 00:18:27,889 S1: roving people who were going to come try to get 320 00:18:27,890 --> 00:18:30,260 S1: our stuff. You could ask it anything. Ideally, it would 321 00:18:30,290 --> 00:18:32,270 S1: be uncensored for that reason, right? You want to be 322 00:18:32,270 --> 00:18:35,720 S1: able to ask security questions. So all that to say, 323 00:18:35,720 --> 00:18:39,439 S1: I'm going to do a project called reboot I with 324 00:18:39,440 --> 00:18:44,000 S1: reboot being like Reboot Society or whatever, an offline oracle 325 00:18:44,000 --> 00:18:48,470 S1: for emergencies. Um, now it's such a cool idea. I'm 326 00:18:48,470 --> 00:18:50,570 S1: sure a million people have already had it, so they're 327 00:18:50,570 --> 00:18:52,909 S1: probably already working on it. So if anyone is hearing 328 00:18:52,910 --> 00:18:56,960 S1: this or they're reading this, um, send me a link 329 00:18:56,960 --> 00:19:00,020 S1: and I'll just like, go buy one and or procure one. 330 00:19:00,020 --> 00:19:03,860 S1: If not, and someone wants to help build one, I'm 331 00:19:03,859 --> 00:19:05,990 S1: going to go build this. I want this running in 332 00:19:05,990 --> 00:19:08,330 S1: my house. I've already got solar. I've got lots of 333 00:19:08,330 --> 00:19:12,200 S1: ways to gather energy and store it in batteries, and 334 00:19:12,200 --> 00:19:15,110 S1: it would be super nice without internet to be able 335 00:19:15,109 --> 00:19:18,919 S1: to ask all those sorts of questions. So cool. Discovery 336 00:19:18,950 --> 00:19:23,750 S1: Cloudflare's robots.txt file. It's a mix of Ascii art and 337 00:19:23,750 --> 00:19:28,920 S1: directives for web crawlers. Obviously, that allows Twitter bot and 338 00:19:28,920 --> 00:19:33,659 S1: demand based website preview to access specific pages blocks many 339 00:19:33,660 --> 00:19:36,420 S1: other from accessing. Actually, you know what I'm going to do? 340 00:19:36,780 --> 00:19:40,020 S1: Let me just oh bam, look at that. This is 341 00:19:40,020 --> 00:19:42,960 S1: why you do a video podcast. You can just zoom 342 00:19:42,960 --> 00:19:46,590 S1: in to stuff. Look at this. Why are they. That's 343 00:19:46,590 --> 00:19:49,260 S1: a lot of that's that's a lot of bandwidth. Depending 344 00:19:49,260 --> 00:19:50,729 S1: on how many people are pulling it. It's a lot 345 00:19:50,730 --> 00:19:53,669 S1: of bandwidth. Our tree is a redwood. Cool. This is 346 00:19:53,670 --> 00:19:56,399 S1: cool though. Look at all these allows I'm looking at 347 00:19:56,400 --> 00:20:00,030 S1: the scroll bar. How long is this thing okay okay. 348 00:20:00,119 --> 00:20:02,940 S1: It's a lot of disallows. Are those languages. No those 349 00:20:02,940 --> 00:20:06,510 S1: are directories. Yeah. These are all directories okay. Why do 350 00:20:06,510 --> 00:20:08,760 S1: they have these. Why do we need to say that okay. 351 00:20:08,790 --> 00:20:12,030 S1: And then we got our sitemaps okay. And more Ascii art. 352 00:20:12,060 --> 00:20:17,460 S1: Yeah cool I like it. Interesting. Managing high performers a 353 00:20:17,490 --> 00:20:21,659 S1: guide on how to effectively manage high performing employees. And 354 00:20:21,660 --> 00:20:25,900 S1: Ian's secure shoelace knot is the best shoelace. Not that 355 00:20:25,900 --> 00:20:30,669 S1: I know of. And no, there's no sponsorship. Because that 356 00:20:30,670 --> 00:20:33,730 S1: would be silly because no such thing exists. It's a 357 00:20:33,730 --> 00:20:38,679 S1: shoelace knot. I actually tie this for my sneakers. Actually, 358 00:20:38,680 --> 00:20:41,650 S1: the ones I'm wearing right now. Wish I could put 359 00:20:41,650 --> 00:20:44,950 S1: my foot up there. That would hurt. Uh, and I 360 00:20:44,950 --> 00:20:48,370 S1: mostly leave them that way. I literally, these are common projects. 361 00:20:48,369 --> 00:20:51,010 S1: I literally just slide my foot into there. Sometimes I 362 00:20:51,010 --> 00:20:53,830 S1: use a shoehorn to do that, and sometimes I just 363 00:20:53,830 --> 00:20:58,240 S1: use my finger. But I tied these with this knot 364 00:20:58,240 --> 00:21:01,690 S1: and it's the coolest looking knot. It's the most secure knot. 365 00:21:01,720 --> 00:21:04,659 S1: It's awesome. And because we're on video, I'm going to 366 00:21:04,660 --> 00:21:07,300 S1: go show you this knot. Look at this thing. So 367 00:21:07,300 --> 00:21:12,910 S1: you make two thingies. Two separate thingies. Then you cross 368 00:21:12,910 --> 00:21:16,720 S1: the front one over the left one, the right one 369 00:21:16,720 --> 00:21:21,250 S1: goes over the the left one. And you do this twice. 370 00:21:21,250 --> 00:21:25,290 S1: You do over and under for both of them, and 371 00:21:25,290 --> 00:21:28,500 S1: then you pull the thing and you end up with that. 372 00:21:28,500 --> 00:21:31,530 S1: You end up with this looking thing right here. And 373 00:21:31,530 --> 00:21:36,150 S1: it's super symmetrical and flat. It's not like twisted and 374 00:21:36,150 --> 00:21:39,449 S1: trying to go up and it's got like this really 375 00:21:39,450 --> 00:21:42,480 S1: cool looking box knot in the middle. And that's the 376 00:21:42,480 --> 00:21:46,290 S1: most you've ever heard about a shoelace knot on a 377 00:21:46,290 --> 00:21:50,070 S1: podcast probably. All right. Recommendation of the week. Check out 378 00:21:50,070 --> 00:21:53,280 S1: the aphorism of the week below. So we'll jump there. 379 00:21:53,400 --> 00:21:56,219 S1: If you hit a wrong note, it's the next note 380 00:21:56,220 --> 00:21:59,070 S1: you play that determines if it's good or bad. If 381 00:21:59,070 --> 00:22:01,320 S1: you hit a wrong note, it's the next note you 382 00:22:01,320 --> 00:22:05,220 S1: play that determines if it's good or bad. Okay, so 383 00:22:05,220 --> 00:22:07,920 S1: that was the aphorism of the week. Now focus your 384 00:22:07,920 --> 00:22:11,490 S1: efforts on being flexible after wrong notes, as opposed to 385 00:22:11,520 --> 00:22:14,280 S1: being able to play perfect notes all the time. That's 386 00:22:14,280 --> 00:22:17,280 S1: my recommendation of the week. 2020 five inches the next 387 00:22:17,280 --> 00:22:20,500 S1: few years are likely to be so crazy that we 388 00:22:20,500 --> 00:22:23,650 S1: won't be able to plan or play the right notes. 389 00:22:23,650 --> 00:22:26,770 S1: So we just have to be good playing the next 390 00:22:26,770 --> 00:22:30,159 S1: note afterwards. And I will read the aphorism of the 391 00:22:30,160 --> 00:22:33,400 S1: week again, because that's what we do. If you hit 392 00:22:33,400 --> 00:22:35,860 S1: a wrong note, it's the next note you play that 393 00:22:35,859 --> 00:22:38,860 S1: determines if it's good or bad. If you hit a 394 00:22:38,859 --> 00:22:42,399 S1: wrong note, it's the next note you play that determines 395 00:22:42,400 --> 00:22:46,810 S1: if it's good or bad. Miles Davis. Unsupervised learning is 396 00:22:46,810 --> 00:22:49,930 S1: produced and edited by Daniel Miessler on a Neumann U87 397 00:22:49,960 --> 00:22:54,070 S1: AI microphone using Hindenburg. Intro and outro music is by 398 00:22:54,070 --> 00:22:57,250 S1: zombie with a Y. And to get the text and 399 00:22:57,250 --> 00:22:59,619 S1: links from this episode, sign up for the newsletter version 400 00:22:59,619 --> 00:23:05,350 S1: of the show at Daniel miessler.com/newsletter. We'll see you next time.