1 00:00:00,080 --> 00:00:03,410 S1: Wiz is the number one cloud security platform based on 2 00:00:03,410 --> 00:00:07,490 S1: independent customer reviews that helps security leaders and developers protect 3 00:00:07,490 --> 00:00:11,300 S1: everything they build and run in the cloud. Leading organizations 4 00:00:11,300 --> 00:00:14,840 S1: from the fortune 100 to high growth companies all trust 5 00:00:14,840 --> 00:00:18,530 S1: wiz to innovate at the pace of cloud while staying secure. 6 00:00:18,560 --> 00:00:21,980 S1: Click the link in the show notes or head to wizfolio. 7 00:00:21,980 --> 00:00:26,479 S1: Slash lowercase ul to request a personalized product tour with 8 00:00:26,480 --> 00:00:30,890 S1: a Wiz expert. Unsupervised Learning Podcast listeners also get a 9 00:00:30,890 --> 00:00:35,990 S1: free Cloud Security Health scan by going to Wiz IOL. 10 00:00:36,050 --> 00:00:44,510 S1: That's wizard Tolowercase ul. Welcome to Unsupervised Learning, a security, 11 00:00:44,540 --> 00:00:47,240 S1: AI and meaning focused podcast that looks at how best 12 00:00:47,240 --> 00:00:50,630 S1: to thrive as humans in a post AI world. It 13 00:00:50,630 --> 00:00:54,620 S1: combines original ideas, analysis and mental models to bring not 14 00:00:54,620 --> 00:00:57,830 S1: just the news, but why it matters and how to respond. 15 00:01:02,370 --> 00:01:05,250 S1: All right. Welcome to unsupervised learning. This is Daniel Miessler. 16 00:01:05,280 --> 00:01:09,209 S1: This is episode 459 of the podcast. All right. So 17 00:01:09,209 --> 00:01:12,840 S1: had a great conversation with Rob Allen from Threatlocker about 18 00:01:12,840 --> 00:01:15,300 S1: their approach to zero Trust. And it turned out to 19 00:01:15,300 --> 00:01:19,319 S1: be a really interesting conversation. I always like to dive 20 00:01:19,319 --> 00:01:21,929 S1: into what somebody means when they say zero trust, because 21 00:01:21,930 --> 00:01:26,850 S1: it could be anywhere from total horsecrap to pretty decent, 22 00:01:26,850 --> 00:01:31,260 S1: and I really liked Rob's approach in responses there, so 23 00:01:31,260 --> 00:01:34,620 S1: definitely check that out. Got you all membership. Black Friday 24 00:01:34,650 --> 00:01:39,809 S1: sale just started 20% off the first month, so or 25 00:01:39,810 --> 00:01:43,770 S1: the first year. So go check that out. And jumping 26 00:01:43,770 --> 00:01:46,920 S1: right into security. Actually, first of all, um, did a 27 00:01:46,920 --> 00:01:50,250 S1: huge upgrade to my ubiquiti gear now moving towards a 28 00:01:50,250 --> 00:01:53,820 S1: ten gigabit world, which I talked a bit about last week, 29 00:01:53,820 --> 00:01:59,130 S1: and going to Saudi to speak at Blackhat EMEA. Mia 30 00:01:59,160 --> 00:02:04,620 S1: Blackhat Mia. Yeah Saudi Arabia next week. Going to be awesome. 31 00:02:04,650 --> 00:02:07,680 S1: All right. Jumping into security. This one didn't get near 32 00:02:07,680 --> 00:02:10,530 S1: enough coverage. I don't think ChatGPT has basically a new 33 00:02:10,530 --> 00:02:14,490 S1: feature that could read code from Mac OS. So apps 34 00:02:14,490 --> 00:02:19,620 S1: like VSCode, Xcode, terminal, it could basically read the app 35 00:02:19,620 --> 00:02:23,430 S1: and understand what it's saying and doing, and then give 36 00:02:23,430 --> 00:02:26,070 S1: you analysis based on that. And I think this is 37 00:02:26,070 --> 00:02:28,740 S1: getting closer to what a whole bunch of people in 38 00:02:28,740 --> 00:02:32,970 S1: AI are actually working on, which is what I believe 39 00:02:33,000 --> 00:02:37,020 S1: to be like an inevitable future, which is you can 40 00:02:37,020 --> 00:02:39,420 S1: see all the screens, right? I'm looking at three screens 41 00:02:39,419 --> 00:02:42,930 S1: right now. Ideally, my eye would be looking from inside 42 00:02:42,930 --> 00:02:45,420 S1: the computer because it would have like agent access, like 43 00:02:45,419 --> 00:02:47,820 S1: we already talked about. But also I wanted to be 44 00:02:47,820 --> 00:02:50,250 S1: able to see my surroundings. I wanted to be able 45 00:02:50,250 --> 00:02:52,200 S1: to see my cameras. I wanted to be able to 46 00:02:52,200 --> 00:02:55,440 S1: see around me, behind me, especially behind me. If you're 47 00:02:55,440 --> 00:02:59,220 S1: out in the city walking around and also cameras or 48 00:02:59,220 --> 00:03:04,240 S1: also Microphones and other types of sensor, but mostly visual 49 00:03:04,240 --> 00:03:08,290 S1: and audio. And that way it can be consuming, you know, 50 00:03:08,320 --> 00:03:12,519 S1: the environment all the time, constantly parsing what's going on 51 00:03:12,520 --> 00:03:16,870 S1: and constantly doing analysis and giving you like, views inside 52 00:03:16,870 --> 00:03:20,620 S1: of a heads up display or, you know, glasses or 53 00:03:20,650 --> 00:03:23,919 S1: lenses or however that interface is going to be, depending 54 00:03:23,919 --> 00:03:26,560 S1: on how far the tech is. But the most important 55 00:03:26,560 --> 00:03:28,870 S1: thing is it has to be seeing what you're seeing 56 00:03:28,870 --> 00:03:31,270 S1: and even more than what you're seeing. Big problem right 57 00:03:31,270 --> 00:03:35,050 S1: now is the on ramp to using AI. It's not 58 00:03:35,050 --> 00:03:37,330 S1: a problem of the models themselves. The models are getting 59 00:03:37,330 --> 00:03:39,310 S1: really smart. The problem is how do you get stuff 60 00:03:39,310 --> 00:03:41,350 S1: into the model and how do you get it out. 61 00:03:41,350 --> 00:03:44,380 S1: That's the problem. So this type of thing is a 62 00:03:44,380 --> 00:03:47,890 S1: really huge advance because it's able to see inside your apps. 63 00:03:47,890 --> 00:03:51,400 S1: And I'm really excited about this future. I've talked about 64 00:03:51,400 --> 00:03:53,860 S1: it a lot. I've talked about it in the predictable 65 00:03:53,860 --> 00:03:57,550 S1: path of AI. It's just that I'm not really going 66 00:03:57,580 --> 00:04:00,100 S1: to trust this. There's already a couple of companies who 67 00:04:00,100 --> 00:04:03,680 S1: will watch your entire screen and will upload that stuff, 68 00:04:03,680 --> 00:04:05,960 S1: and they will do the parsing and they will send 69 00:04:05,960 --> 00:04:10,070 S1: you back some really cool stuff. Like there's one particular company, 70 00:04:10,220 --> 00:04:12,620 S1: I think they changed their name. It's called rewind. Or 71 00:04:12,620 --> 00:04:16,070 S1: maybe rewind is the new name. But anyway, I signed 72 00:04:16,070 --> 00:04:20,000 S1: up for their new, uh, gadget, which should be coming soon. 73 00:04:20,000 --> 00:04:22,609 S1: I did not give them access to my whole computer 74 00:04:22,610 --> 00:04:26,000 S1: because I could have the calendar open or a message 75 00:04:26,000 --> 00:04:28,789 S1: or something, and it's like you could be doxing your friends. 76 00:04:28,790 --> 00:04:31,099 S1: You don't know what's going to happen. Most importantly, you 77 00:04:31,100 --> 00:04:35,030 S1: don't know how secure that startup is. And having consulted 78 00:04:35,029 --> 00:04:40,430 S1: for startups in security specifically for decades, I'm not sending 79 00:04:40,430 --> 00:04:43,580 S1: my data to those startups. I basically only trust like 80 00:04:43,610 --> 00:04:46,969 S1: one company that much, both on the security side and 81 00:04:46,970 --> 00:04:50,000 S1: also on the privacy side, and just how seriously they 82 00:04:50,000 --> 00:04:53,420 S1: take this stuff, combined with a lack of conflicts of interest. 83 00:04:53,420 --> 00:04:57,740 S1: And that's Apple. I would say Google is just as competent, 84 00:04:57,770 --> 00:05:01,560 S1: maybe even more competent on the security side, I would say. 85 00:05:01,589 --> 00:05:05,400 S1: But the problem is they also make their money off 86 00:05:05,400 --> 00:05:08,100 S1: of ads or a lot of their money off of ads. 87 00:05:08,100 --> 00:05:11,430 S1: So I just don't like the idea of like them 88 00:05:11,430 --> 00:05:13,650 S1: having all that data and then being able to sell 89 00:05:13,650 --> 00:05:16,380 S1: it or sell things to me. I don't like the 90 00:05:16,380 --> 00:05:19,560 S1: conflict of interest there, even though I like the security. 91 00:05:19,560 --> 00:05:22,650 S1: So that's why I mostly only trust two companies and 92 00:05:22,650 --> 00:05:26,100 S1: mostly only Apple. But this is where this is going. 93 00:05:26,100 --> 00:05:31,650 S1: Continuous monitoring all the time with as many sensors as possible. Front, back, 94 00:05:31,650 --> 00:05:35,010 S1: inside your apps, inside your phone. That's where everyone is 95 00:05:35,010 --> 00:05:37,500 S1: rushing to. Even if they don't know that they're rushing 96 00:05:37,529 --> 00:05:40,770 S1: to that location. And that's why I think Google and 97 00:05:40,770 --> 00:05:45,000 S1: Apple have a massive advantage because they have the device, right? 98 00:05:45,029 --> 00:05:49,410 S1: I'm on Apple for phone, I'm on Apple for Mac 99 00:05:49,410 --> 00:05:54,479 S1: and desktop and laptop. So because they are the OS, 100 00:05:54,510 --> 00:05:58,140 S1: they have a massive advantage here. Okay. Palo Alto Networks 101 00:05:58,140 --> 00:06:01,620 S1: has released indicators of compromise IOCs for a new zero 102 00:06:01,650 --> 00:06:06,240 S1: day vulnerability affecting firewalls. VMware confirmed that threat actors are 103 00:06:06,240 --> 00:06:10,589 S1: exploiting two vCenter server vulnerabilities, and ones like a 9.8 104 00:06:10,620 --> 00:06:12,719 S1: and the other ones like a seven something I and 105 00:06:12,720 --> 00:06:15,810 S1: Tech Anthropic has a new prompt improver that takes the 106 00:06:15,810 --> 00:06:18,510 S1: given prompt and writes a better one. Really cool to 107 00:06:18,510 --> 00:06:21,570 S1: see people getting in on this. This is part of 108 00:06:21,570 --> 00:06:25,860 S1: the overall ecosystem, right? Um, and we got another one here. Uh, 109 00:06:25,860 --> 00:06:29,910 S1: OpenAI might launch an AI agent tool called Operator in January, 110 00:06:29,910 --> 00:06:33,090 S1: and it will compete with anthropic computer use. I think 111 00:06:33,120 --> 00:06:35,430 S1: agents is going to be the biggest thing that happens 112 00:06:35,430 --> 00:06:38,640 S1: in 2025 for AI. And this is part of a 113 00:06:38,640 --> 00:06:41,250 S1: bigger trend that I've been talking about, where it's more 114 00:06:41,250 --> 00:06:44,130 S1: about the ecosystem. Okay, so if I if I actually 115 00:06:44,130 --> 00:06:48,719 S1: click and open this thing here, it's not just about 116 00:06:48,720 --> 00:06:52,830 S1: the models. Okay. So the big pieces that I think 117 00:06:52,830 --> 00:06:56,789 S1: are part of these four pillars of an AI ecosystem, 118 00:06:56,790 --> 00:07:00,580 S1: the model itself, the post training of the model internal 119 00:07:00,610 --> 00:07:06,729 S1: tooling and agent functionality. So agent or model self-explanatory. Post-training 120 00:07:06,760 --> 00:07:10,120 S1: a set of highly proprietary tricks that magnify the overall 121 00:07:10,120 --> 00:07:12,940 S1: quality of the Raw model. Way to think of this 122 00:07:12,940 --> 00:07:14,770 S1: is to say that it's a way to connect model 123 00:07:14,770 --> 00:07:18,730 S1: weights to human problems. Okay. Internal tooling. All right. So 124 00:07:18,730 --> 00:07:23,110 S1: look at this list. High quality APIs, larger context sizes, 125 00:07:23,140 --> 00:07:29,830 S1: simple fine tuning. Haystack performance. Strict output control. External tooling 126 00:07:29,830 --> 00:07:34,510 S1: like function calling. Trust and safety features. Mobile apps. Prompt testing. 127 00:07:34,510 --> 00:07:38,710 S1: Voice mode and apps OS integration. Integrations with things like 128 00:07:38,710 --> 00:07:43,180 S1: make Zapier end to end and things like caching mode. 129 00:07:43,210 --> 00:07:46,300 S1: These are like all the internal tooling stuff that just 130 00:07:46,300 --> 00:07:49,000 S1: makes it easier. Think of it this way the problem 131 00:07:49,000 --> 00:07:51,790 S1: isn't the models. The problem is the on ramps onto 132 00:07:51,790 --> 00:07:54,489 S1: the models and the output out of them back into 133 00:07:54,490 --> 00:07:57,850 S1: your life. Okay, we are humans. We have human problems. 134 00:07:57,850 --> 00:08:00,680 S1: We have business problems. We have personal problems. whatever we 135 00:08:00,680 --> 00:08:03,890 S1: need to get that content, the content of that problem 136 00:08:03,890 --> 00:08:07,100 S1: into an AI and then back out into our lives, 137 00:08:07,100 --> 00:08:10,040 S1: into our actual brains, into the real world. That's what 138 00:08:10,040 --> 00:08:13,310 S1: this internal tooling piece is, right? Because if you can't 139 00:08:13,310 --> 00:08:16,760 S1: do this, well, then it doesn't matter if your model 140 00:08:16,760 --> 00:08:21,920 S1: is 13% better on some random benchmark, right? So the 141 00:08:21,920 --> 00:08:26,480 S1: models have to get better, but not as important as 142 00:08:26,510 --> 00:08:30,590 S1: actually getting the interfaces to the models better. So uh 143 00:08:30,620 --> 00:08:33,320 S1: oh yeah. And the next one relevant to the next 144 00:08:33,320 --> 00:08:38,000 S1: story there is agents. Right. So an AI component that 145 00:08:38,000 --> 00:08:40,429 S1: interprets instructions and takes on more of the work in 146 00:08:40,429 --> 00:08:46,490 S1: total AI workflows than just LLM response, for example, executing functions, 147 00:08:46,490 --> 00:08:50,569 S1: performing data lookups, etc. before passing on results. And I 148 00:08:50,570 --> 00:08:53,660 S1: actually have a improved version of that, which I won't 149 00:08:53,690 --> 00:08:56,390 S1: I won't go find it, but it's, uh, it's updated 150 00:08:56,390 --> 00:09:02,030 S1: on the AGI definition inside of raid the real world 151 00:09:02,059 --> 00:09:05,150 S1: I definitions if you want to go check that out. Okay, 152 00:09:05,179 --> 00:09:09,500 S1: so Sam Altman and Arianna Huffington have a thrive AI 153 00:09:09,530 --> 00:09:15,380 S1: health company and looking at doing personalized advice on sleep, food, 154 00:09:15,380 --> 00:09:19,850 S1: fitness and more, Google.org is putting 20 million in cash 155 00:09:19,850 --> 00:09:22,880 S1: and 2 million in cloud credits into a new initiative 156 00:09:22,910 --> 00:09:27,439 S1: to help researchers use AI for scientific breakthroughs. One of 157 00:09:27,440 --> 00:09:31,730 S1: the most important things that I think I could possibly 158 00:09:31,730 --> 00:09:41,119 S1: do is actively going to just invent new things, make new, uh, research, right? 159 00:09:41,150 --> 00:09:45,020 S1: Discover new things. Right? You take the smartest people in 160 00:09:45,020 --> 00:09:48,140 S1: the world who who are capable of doing this, and 161 00:09:48,140 --> 00:09:50,840 S1: there's very few of them. So if we can actually 162 00:09:50,840 --> 00:09:54,710 S1: scale that, that's where the real takeoff starts to happen. 163 00:09:54,740 --> 00:09:58,910 S1: Apple's M4 Max CPU transcribes audio twice as fast as 164 00:09:59,059 --> 00:10:05,849 S1: Nvidia's RTX A5000 GPU while using significantly less power. I 165 00:10:05,850 --> 00:10:09,150 S1: really want to get one of these clusters, like a 166 00:10:09,150 --> 00:10:15,120 S1: m4 Mac mini cluster would be super cool. But yeah, 167 00:10:15,150 --> 00:10:16,860 S1: a lot of people have been asking me, should I 168 00:10:16,860 --> 00:10:19,410 S1: do the cluster thing with smaller boxes? Or should I 169 00:10:19,410 --> 00:10:22,140 S1: just get one big rig? I feel like the next 170 00:10:22,140 --> 00:10:26,100 S1: generation or the current next the current generation in like 171 00:10:26,130 --> 00:10:28,950 S1: another year or so, it's still going to be better 172 00:10:28,950 --> 00:10:30,900 S1: to have one big box. It really depends on what 173 00:10:30,900 --> 00:10:33,150 S1: you need it for, but I would say we're not 174 00:10:33,150 --> 00:10:38,400 S1: quite there yet. With the cluster of smaller boxes, this 175 00:10:38,520 --> 00:10:41,939 S1: exo lab and things like that, they're a little bit experimental. 176 00:10:42,030 --> 00:10:45,569 S1: The best thing to do still is to buy a box. 177 00:10:45,570 --> 00:10:49,260 S1: I bought a pre-made one and it's fantastic. It's got 178 00:10:49,260 --> 00:10:52,530 S1: two 49 seconds in it. It's got tons of memory. 179 00:10:52,679 --> 00:10:56,370 S1: It's very, very fast. And it's just like plug and play. 180 00:10:56,370 --> 00:11:01,750 S1: Easy to do. I would say if this clustering technology 181 00:11:01,750 --> 00:11:04,929 S1: gets better and better, and there's more and more devices 182 00:11:04,929 --> 00:11:07,030 S1: like Mac minis and stuff like that that you could 183 00:11:07,030 --> 00:11:10,569 S1: piece together. Then it starts to be a compelling alternative 184 00:11:10,570 --> 00:11:15,130 S1: to actually buying a giant AI box. But until then, 185 00:11:15,130 --> 00:11:18,520 S1: I think the giant box is probably still going to 186 00:11:18,520 --> 00:11:22,720 S1: be better. Okay. iOS 18.2 is music recognition feature. Now 187 00:11:22,750 --> 00:11:25,719 S1: logs where you were when you actually heard the song. 188 00:11:25,720 --> 00:11:28,990 S1: So that's part of the metadata. Now pharma stocks have crashed. 189 00:11:28,990 --> 00:11:32,050 S1: This is under the humans label. Pharma stocks have crashed 190 00:11:32,050 --> 00:11:36,310 S1: after RFK Jr was announced to be taking over Health 191 00:11:36,309 --> 00:11:39,940 S1: and Human Services. Moderna is down like 40%. Yeah, I 192 00:11:39,940 --> 00:11:43,449 S1: don't know what this is currently at. Uh, down 562, 193 00:11:43,480 --> 00:11:45,160 S1: but that's for the day. So if we go to 194 00:11:45,190 --> 00:11:52,600 S1: three months, down 57% in the last three months, Moderna. Okay. 195 00:11:52,929 --> 00:11:55,630 S1: They're they're one of the people who came out with 196 00:11:55,630 --> 00:12:01,130 S1: the best vaccine for Covid, like 57% because of RFK. 197 00:12:01,160 --> 00:12:04,219 S1: I almost feel like this is a buy opportunity. How 198 00:12:04,220 --> 00:12:06,979 S1: could this not be a buy opportunity? It's not like 199 00:12:06,980 --> 00:12:11,060 S1: they suddenly stopped being able to make things. And I 200 00:12:11,059 --> 00:12:14,090 S1: just don't believe that people are going to let RFK 201 00:12:14,090 --> 00:12:18,110 S1: just destroy these companies. I just don't see that happening. 202 00:12:18,140 --> 00:12:20,690 S1: Hopefully they're going to be able to figure it out, 203 00:12:20,690 --> 00:12:23,540 S1: because RFK is going to do some cool stuff. Like 204 00:12:23,540 --> 00:12:25,460 S1: he's right about a lot of stuff. That's what's most 205 00:12:25,460 --> 00:12:28,040 S1: scary about a lot of these people. They're so right 206 00:12:28,040 --> 00:12:31,400 S1: about so many things, and the things they're wrong about 207 00:12:31,400 --> 00:12:34,640 S1: are mixed in with it. Right? And that's not just RFK. 208 00:12:34,670 --> 00:12:38,540 S1: That's a lot of people, including probably myself. So yeah, 209 00:12:38,570 --> 00:12:44,300 S1: I don't see how it doesn't bounce back from a 60% 210 00:12:44,300 --> 00:12:49,010 S1: cut from three months. Right. Personally, I'm not investing, but 211 00:12:49,010 --> 00:12:51,470 S1: I think it would probably be smart, at least in 212 00:12:51,470 --> 00:12:56,390 S1: the long term. Netflix had a record 65 million concurrent 213 00:12:56,390 --> 00:13:00,200 S1: streams during Mike Tyson versus Jake Paul fight. it did 214 00:13:00,200 --> 00:13:02,599 S1: have a bunch of connection problems though. Everyone I know 215 00:13:02,600 --> 00:13:06,320 S1: at Netflix got massively spammed with like, all their friends 216 00:13:06,320 --> 00:13:08,360 S1: texting and was like, hey, what's going on? It's like, yeah, 217 00:13:08,360 --> 00:13:11,720 S1: you think I'm in charge of actual throughput during a 218 00:13:11,720 --> 00:13:15,140 S1: Tyson fight? Like, stop texting me. New study shows that 219 00:13:15,140 --> 00:13:18,500 S1: treating bullying as a collective issue rather than an individual one, 220 00:13:18,500 --> 00:13:22,400 S1: can significantly reduce its occurrence in primary schools. I love 221 00:13:22,400 --> 00:13:26,000 S1: the concept. It's kind of like how the johns get 222 00:13:26,000 --> 00:13:29,750 S1: in trouble in some European countries for prostitution instead of 223 00:13:29,750 --> 00:13:34,520 S1: the prostitute, because it's like the ecosystem that's the actual problem, right? 224 00:13:34,610 --> 00:13:38,059 S1: So I like this idea of with bullying, I like 225 00:13:38,090 --> 00:13:40,699 S1: the idea of shaming the people around who didn't say 226 00:13:40,700 --> 00:13:42,800 S1: anything or do anything. And of course, we have to 227 00:13:42,830 --> 00:13:44,810 S1: be careful about shaming. We're talking about kids here in 228 00:13:44,809 --> 00:13:48,470 S1: a lot of cases or in most cases. But like, 229 00:13:48,530 --> 00:13:50,990 S1: I feel like all the marketing needs to be heading 230 00:13:50,990 --> 00:13:53,359 S1: in the direction of like if you are one of 231 00:13:53,360 --> 00:13:56,270 S1: these people watching and not saying anything, not reporting it, 232 00:13:56,300 --> 00:13:59,880 S1: not intervening like you don't want to be unsafe or whatever, 233 00:13:59,880 --> 00:14:02,339 S1: but you could go report it. You could do something 234 00:14:02,370 --> 00:14:06,449 S1: to prevent this from happening again. And if you're not 235 00:14:06,480 --> 00:14:09,780 S1: something like you are the bully or you are enabling 236 00:14:09,780 --> 00:14:11,699 S1: this and this is really bad, and I know this 237 00:14:11,700 --> 00:14:14,430 S1: is already an aspect of like a lot of these programs, 238 00:14:14,429 --> 00:14:19,860 S1: but I think it could be significantly magnified. Ideas reboot I, 239 00:14:20,070 --> 00:14:23,609 S1: I absolutely love this one. Absolutely love this one. I 240 00:14:23,610 --> 00:14:25,710 S1: can't remember where I got this idea. I think it's 241 00:14:25,710 --> 00:14:28,680 S1: been hit me from multiple places, but I want to 242 00:14:28,680 --> 00:14:32,040 S1: build a local AI that can run offline. Oh, I 243 00:14:32,040 --> 00:14:34,440 S1: know where I first heard this idea. It was actually 244 00:14:34,440 --> 00:14:38,190 S1: from Joseph Thacker, like a year and a half ago. 245 00:14:38,220 --> 00:14:40,110 S1: He's like, oh, I just want a thing that I 246 00:14:40,110 --> 00:14:43,920 S1: can use offline. But, um, more recently I got this 247 00:14:43,920 --> 00:14:47,190 S1: from somewhere else. I can't was it? Was it X 248 00:14:47,190 --> 00:14:51,870 S1: or Instagram? I don't know, but the idea is let's 249 00:14:51,870 --> 00:14:54,300 S1: say all the power is out, or let's say the 250 00:14:54,300 --> 00:14:58,290 S1: internet is out and let's say, um, oh, I know 251 00:14:58,290 --> 00:15:03,670 S1: what it was. The initial conversation I was having with Joseph, 252 00:15:03,700 --> 00:15:05,890 S1: and he actually brought it to me. It was his idea. 253 00:15:05,890 --> 00:15:09,460 S1: He was like, what if you could go back in time? 254 00:15:09,490 --> 00:15:12,850 S1: What could you bring with you to actually move society 255 00:15:12,880 --> 00:15:15,790 S1: forward or something like that? And it's like, I think 256 00:15:15,790 --> 00:15:18,100 S1: about this a lot way more than the Roman Empire. 257 00:15:18,100 --> 00:15:20,920 S1: I think a lot about like, could I actually move 258 00:15:20,950 --> 00:15:23,560 S1: science forward if I was put 200 years in the 259 00:15:23,560 --> 00:15:26,290 S1: past or 2000 years in the past, what could I 260 00:15:26,290 --> 00:15:28,840 S1: actually offer to them? I think someone makes a joke 261 00:15:28,840 --> 00:15:32,860 S1: about this in current stand up comedy. It's like, yeah, oh, 262 00:15:32,890 --> 00:15:37,870 S1: it's Nate Bathgate, Bargatze, whatever his name is. Um, it's 263 00:15:37,870 --> 00:15:41,350 S1: just like, yeah, um, there's going to be phones and 264 00:15:41,380 --> 00:15:44,770 S1: they're going to have satellite technology. Oh, really? What's the satellite? Well, 265 00:15:44,770 --> 00:15:47,020 S1: it's this thing that goes around. How do you make one? 266 00:15:47,050 --> 00:15:49,720 S1: What does it do? You're saying there's thousands of satellites 267 00:15:49,720 --> 00:15:52,690 S1: in the air and the Earth is actually round? Like, 268 00:15:52,720 --> 00:15:56,260 S1: can you prove anything about this? No, I can't. So 269 00:15:56,710 --> 00:16:01,960 S1: we we don't know how any of this is working, right? And, uh, 270 00:16:02,440 --> 00:16:05,380 S1: and we can't describe it to anyone. And more importantly, 271 00:16:05,380 --> 00:16:08,050 S1: if you lose the internet and there's like, say, a 272 00:16:08,050 --> 00:16:11,170 S1: meteor hits or whatever, I'm not going to go into 273 00:16:11,200 --> 00:16:13,630 S1: negativity right now too early in the show for that. 274 00:16:13,630 --> 00:16:17,140 S1: But let's say something bad happens and you are stuck 275 00:16:17,140 --> 00:16:19,600 S1: in your house and there's no internet. Let's say you're 276 00:16:19,600 --> 00:16:22,660 S1: not dying of hunger or thirst, but you don't have 277 00:16:22,660 --> 00:16:28,120 S1: the internet. Okay, so watch this. Tourniquets, sterilizing water, building shelters, 278 00:16:28,120 --> 00:16:33,100 S1: identifying edible plants. Which mushroom is actually will kill you. 279 00:16:33,100 --> 00:16:36,100 S1: And which mushroom can you put on a salad? These 280 00:16:36,100 --> 00:16:40,000 S1: are important distinctions to make. So check this out. What 281 00:16:40,000 --> 00:16:43,180 S1: if you had an AI that ran? We got to 282 00:16:43,180 --> 00:16:46,120 S1: assume solar power, right? Or maybe the grid actually works, 283 00:16:46,120 --> 00:16:48,850 S1: but there's no internet. Whatever. Just work with me. You 284 00:16:48,850 --> 00:16:51,670 S1: can show it. Pictures. You could take pictures, or you 285 00:16:51,670 --> 00:16:54,460 S1: could show it an actual live plant or whatever. All 286 00:16:54,460 --> 00:16:56,890 S1: you have to do is show this AI the particular 287 00:16:56,890 --> 00:16:59,470 S1: thing or you describe, hey, I need a shelter that 288 00:16:59,480 --> 00:17:02,270 S1: does this. I've got this much water. I got it 289 00:17:02,270 --> 00:17:04,790 S1: from this kind of creek. I live in this kind 290 00:17:04,790 --> 00:17:07,550 S1: of area. What kind of toxins are likely to be 291 00:17:07,550 --> 00:17:09,530 S1: in it? How can I get those toxins out? What 292 00:17:09,530 --> 00:17:12,530 S1: kind of filter can I build? Will these iodine tablets 293 00:17:12,560 --> 00:17:16,310 S1: actually work? I have these symptoms. Which drugs should I take? 294 00:17:16,340 --> 00:17:20,300 S1: You have a local model running. All it needs is power. 295 00:17:20,300 --> 00:17:23,330 S1: It doesn't need the internet. It doesn't need OpenAI. Think 296 00:17:23,330 --> 00:17:27,109 S1: of how much knowledge is inside of a llama two 297 00:17:27,140 --> 00:17:30,590 S1: or a llama three, or a llama four, or whatever. 298 00:17:30,590 --> 00:17:33,409 S1: Local model. And as long as you have power to 299 00:17:33,440 --> 00:17:35,990 S1: run it and it can actually see and you can 300 00:17:35,990 --> 00:17:39,109 S1: type to it, it can answer tons of stuff that 301 00:17:39,109 --> 00:17:43,550 S1: can actually keep you alive. Even better, let's go a 302 00:17:43,550 --> 00:17:46,550 S1: little sci fi, okay. It's just you and 10,000 other 303 00:17:46,550 --> 00:17:50,480 S1: people and the rest of the, let's say, the rest 304 00:17:50,480 --> 00:17:52,730 S1: of the planet got hit by a meteor or whatever. 305 00:17:52,730 --> 00:17:57,260 S1: You have to rebuild all of society. Okay? Irrigation. How 306 00:17:57,290 --> 00:18:00,570 S1: to how does the stoplight work? How does a combustion 307 00:18:00,570 --> 00:18:03,419 S1: engine work? What kind of metal do you need to 308 00:18:03,420 --> 00:18:06,629 S1: build in order to make a combustion engine? What are 309 00:18:06,630 --> 00:18:12,300 S1: all these different alloys? You can bootstrap a society with 310 00:18:12,300 --> 00:18:15,840 S1: one box. Isn't that crazy? You can bootstrap a society 311 00:18:15,840 --> 00:18:19,650 S1: with one box that you can have, like sitting next 312 00:18:19,650 --> 00:18:22,950 S1: to your NAS over here. You just need power. You 313 00:18:22,950 --> 00:18:24,810 S1: need power. Well, I guess you need the peripherals to 314 00:18:24,840 --> 00:18:27,270 S1: be able to talk to it and everything, but you 315 00:18:27,270 --> 00:18:32,310 S1: don't need that much. It's way more impressive than trying 316 00:18:32,310 --> 00:18:34,770 S1: to collect all of Wikipedia. I mean, that was the 317 00:18:34,770 --> 00:18:37,140 S1: other model, right? You just download Wikipedia and you have 318 00:18:37,140 --> 00:18:39,900 S1: a bunch of things you can look for. Not nearly 319 00:18:39,900 --> 00:18:43,020 S1: as good as a chatbot that you can ask questions. 320 00:18:43,020 --> 00:18:47,250 S1: And if it's visually oriented like you could just these 321 00:18:47,250 --> 00:18:50,070 S1: new models are getting amazing, right? It's like, show me 322 00:18:50,070 --> 00:18:53,520 S1: the design, show me a picture of it, draw me 323 00:18:53,520 --> 00:18:57,930 S1: a picture, design me whatever a compound that I could 324 00:18:57,930 --> 00:19:00,810 S1: defend my plants with because people are probably going to 325 00:19:00,810 --> 00:19:04,620 S1: come and get us whatever. Yeah. I don't have any guns. 326 00:19:04,770 --> 00:19:07,980 S1: How do I defend myself from these roving people who 327 00:19:07,980 --> 00:19:09,600 S1: are going to come try to get our stuff? You 328 00:19:09,600 --> 00:19:12,030 S1: could ask it anything. Ideally, it would be uncensored for 329 00:19:12,060 --> 00:19:13,830 S1: that reason. Right. You want to be able to ask 330 00:19:13,830 --> 00:19:17,310 S1: security questions. So all that to say, I'm going to 331 00:19:17,310 --> 00:19:21,360 S1: do a project called reboot I with reboot being like 332 00:19:21,359 --> 00:19:26,760 S1: Reboot Society or whatever, an offline oracle for emergencies. Um, 333 00:19:26,760 --> 00:19:29,940 S1: now it's such a cool idea. I'm sure a million 334 00:19:29,940 --> 00:19:32,340 S1: people have already had it, so they're probably already working 335 00:19:32,340 --> 00:19:34,770 S1: on it. So if anyone is hearing this or they're 336 00:19:34,770 --> 00:19:38,700 S1: reading this, um, send me a link and I'll just like, 337 00:19:38,730 --> 00:19:42,330 S1: go buy one and or procure one. If not, and 338 00:19:42,330 --> 00:19:45,030 S1: someone wants to help build one, I'm going to go 339 00:19:45,060 --> 00:19:47,790 S1: build this. I want this running in my house. I've 340 00:19:47,790 --> 00:19:50,730 S1: already got solar. I've got lots of ways to gather 341 00:19:50,730 --> 00:19:53,520 S1: energy and store it in batteries, and it would be 342 00:19:53,520 --> 00:19:56,580 S1: super nice without internet to be able to ask all 343 00:19:56,580 --> 00:20:02,260 S1: those sorts of questions. So cool. discovery. Cloudflare's robots.txt file. 344 00:20:02,260 --> 00:20:06,850 S1: It's a mix of Ascii art and directives for web crawlers. Obviously, 345 00:20:06,850 --> 00:20:12,340 S1: that allows Twitter bot and demon based website preview to 346 00:20:12,369 --> 00:20:16,419 S1: access specific pages blocks many other from accessing. Actually, you 347 00:20:16,420 --> 00:20:19,810 S1: know what I'm going to do? Let me just. Oh bam! 348 00:20:19,930 --> 00:20:22,930 S1: Look at that. This is why you do a video podcast. 349 00:20:22,960 --> 00:20:25,780 S1: You can just zoom in to stuff. Look at this. 350 00:20:25,869 --> 00:20:29,020 S1: Why are they. That's a lot of that's that's a 351 00:20:29,020 --> 00:20:31,300 S1: lot of bandwidth depending on how many people are pulling it. 352 00:20:31,330 --> 00:20:34,270 S1: It's a lot of bandwidth. Our tree is a redwood. Cool. 353 00:20:34,270 --> 00:20:36,970 S1: This is cool though. Look at all these allows I'm 354 00:20:36,970 --> 00:20:39,010 S1: looking at the scroll bar. How long is this thing. 355 00:20:39,040 --> 00:20:43,630 S1: Okay okay. It's a lot of disallows. Are those languages. No. 356 00:20:43,630 --> 00:20:47,290 S1: Those are directories. Yeah. These are all directories okay. Why 357 00:20:47,290 --> 00:20:49,060 S1: do they have these. Why do we need to say 358 00:20:49,060 --> 00:20:52,090 S1: that okay. And then we got our sitemaps okay. And 359 00:20:52,090 --> 00:20:57,580 S1: more Ascii art. Yeah. Cool I like it. Interesting. Managing 360 00:20:57,580 --> 00:21:00,530 S1: high performers a guide on how to effectively manage high 361 00:21:00,530 --> 00:21:05,960 S1: performing employees and Eon's secure shoelace knot is the best shoelace. 362 00:21:05,960 --> 00:21:10,760 S1: Not that I know of. And no, there's no sponsorship 363 00:21:10,760 --> 00:21:13,940 S1: because that would be silly because no such thing exists. 364 00:21:13,940 --> 00:21:19,580 S1: It's a shoelace knot. I actually tie this for my sneakers, actually, 365 00:21:19,580 --> 00:21:22,550 S1: the ones I'm wearing right now. Wish I could put 366 00:21:22,550 --> 00:21:25,850 S1: my foot up there. That would hurt. Uh, and I 367 00:21:25,850 --> 00:21:29,300 S1: mostly leave them that way. I literally, these are common projects. 368 00:21:29,300 --> 00:21:31,909 S1: I literally just slide my foot into there. Sometimes I 369 00:21:31,910 --> 00:21:34,730 S1: use a shoehorn to do that, and sometimes I just 370 00:21:34,730 --> 00:21:39,169 S1: use my finger. But I tied these with this knot 371 00:21:39,170 --> 00:21:42,620 S1: and it's the coolest looking knot. It's the most secure knot. 372 00:21:42,650 --> 00:21:45,560 S1: It's awesome. And because we're on video, I'm going to 373 00:21:45,560 --> 00:21:48,230 S1: go show you this knot. Look at this thing. So 374 00:21:48,230 --> 00:21:53,840 S1: you make two thingies. Two separate thingies. Then you cross 375 00:21:53,840 --> 00:21:57,650 S1: the front one over the left one, the right one 376 00:21:57,650 --> 00:22:02,160 S1: goes over the the left one. And you do this twice. 377 00:22:02,160 --> 00:22:06,210 S1: You do over and under for both of them. And 378 00:22:06,210 --> 00:22:09,420 S1: then you pull the thing and you end up with that. 379 00:22:09,420 --> 00:22:12,450 S1: You end up with this looking thing right here. And 380 00:22:12,450 --> 00:22:17,070 S1: it's super symmetrical and flat. It's not like twisted and 381 00:22:17,070 --> 00:22:20,369 S1: trying to go up. And it's got like this really 382 00:22:20,369 --> 00:22:23,399 S1: cool looking box not in the middle. And that's the 383 00:22:23,400 --> 00:22:27,210 S1: most you've ever heard about a shoelace. Not on a 384 00:22:27,210 --> 00:22:30,960 S1: podcast probably. All right. Recommendation of the week. Check out 385 00:22:30,960 --> 00:22:34,200 S1: the aphorism of the week below. So we'll jump there. 386 00:22:34,290 --> 00:22:37,139 S1: If you hit a wrong note, it's the next note 387 00:22:37,140 --> 00:22:39,960 S1: you play that determines if it's good or bad. If 388 00:22:39,960 --> 00:22:42,210 S1: you hit a wrong note, it's the next note you 389 00:22:42,210 --> 00:22:46,109 S1: play that determines if it's good or bad. Okay, so 390 00:22:46,109 --> 00:22:48,840 S1: that was the aphorism of the week. Now focus your 391 00:22:48,840 --> 00:22:52,379 S1: efforts on being flexible after wrong notes, as opposed to 392 00:22:52,410 --> 00:22:55,200 S1: being able to play perfect notes all the time. That's 393 00:22:55,200 --> 00:22:58,170 S1: my recommendation of the week. 2020 five inches the next 394 00:22:58,170 --> 00:23:01,410 S1: few years are likely to be so crazy that we 395 00:23:01,410 --> 00:23:04,590 S1: won't be able to plan or play the right notes, 396 00:23:04,590 --> 00:23:07,680 S1: so we just have to be good playing the next 397 00:23:07,680 --> 00:23:11,070 S1: note afterwards. And I will read the aphorism of the 398 00:23:11,070 --> 00:23:14,310 S1: week again, because that's what we do. If you hit 399 00:23:14,340 --> 00:23:16,770 S1: a wrong note, it's the next note you play that 400 00:23:16,770 --> 00:23:19,800 S1: determines if it's good or bad. If you hit a 401 00:23:19,800 --> 00:23:23,310 S1: wrong note, it's the next note you play that determines 402 00:23:23,310 --> 00:23:27,719 S1: if it's good or bad. Miles Davis. Unsupervised learning is 403 00:23:27,720 --> 00:23:30,870 S1: produced and edited by Daniel Miessler on a Neumann U87 404 00:23:30,900 --> 00:23:34,980 S1: AI microphone using Hindenburg. Intro and outro music is by 405 00:23:34,980 --> 00:23:38,159 S1: zombie with a Y. And to get the text and 406 00:23:38,160 --> 00:23:40,530 S1: links from this episode, sign up for the newsletter version 407 00:23:40,530 --> 00:23:46,290 S1: of the show at Daniel miessler.com/newsletter. We'll see you next time.