1 00:00:21,393 --> 00:00:24,333 S1: All right. Welcome to unsupervised learning. This is Daniel Miessler. 2 00:00:24,333 --> 00:00:27,813 S1: I am in a hotel room in Vegas recording because 3 00:00:27,813 --> 00:00:31,923 S1: I have to, but it is Hacker Week and, uh, 4 00:00:31,923 --> 00:00:35,073 S1: here for Defcon and Black Hat and all that good stuff, 5 00:00:35,073 --> 00:00:38,493 S1: but I wanted to get the episode out. So first 6 00:00:38,493 --> 00:00:41,463 S1: note here is that Osint is one of my favorite hobbies, 7 00:00:41,463 --> 00:00:44,583 S1: and there's something called a pizza index. That's one of 8 00:00:44,583 --> 00:00:48,033 S1: my favorite examples of this, which is how much pizza 9 00:00:48,033 --> 00:00:51,843 S1: essentially the neighborhood around the Pentagon is ordering, which really 10 00:00:51,843 --> 00:00:55,443 S1: means the Pentagon. And there's another index related to that, 11 00:00:55,443 --> 00:00:59,013 S1: which is how many people are in the bars and 12 00:00:59,013 --> 00:01:04,233 S1: this person real. Ben Geller posted a tweet about this. 13 00:01:04,233 --> 00:01:07,923 S1: And essentially it says it shows that the people in 14 00:01:07,923 --> 00:01:11,943 S1: the bars is like extremely low and the pizza meter 15 00:01:11,943 --> 00:01:15,023 S1: is off the charts. And I just love this so 16 00:01:15,023 --> 00:01:19,613 S1: much because it indicates pretty strongly that something is about 17 00:01:19,613 --> 00:01:21,563 S1: to go down. And I've got a friend who used 18 00:01:21,563 --> 00:01:24,443 S1: to be an analyst at the Pentagon, and he says, 19 00:01:24,443 --> 00:01:27,503 S1: this is absolutely true. When people are ordering in pizza 20 00:01:27,503 --> 00:01:31,373 S1: and nobody's going home, it's obviously because something is going down. 21 00:01:31,373 --> 00:01:33,593 S1: And in this case, we kind of know what's going down, 22 00:01:33,593 --> 00:01:37,853 S1: which is Iran is preparing to attack Israel and or 23 00:01:37,853 --> 00:01:41,893 S1: whoever else. So that's what that is. But definitely check 24 00:01:41,893 --> 00:01:44,863 S1: out this tweet. It's pretty interesting. So this is also 25 00:01:44,863 --> 00:01:47,983 S1: why I can't wait to fully build out my agent framework, 26 00:01:47,983 --> 00:01:51,853 S1: and for agent framework to become more tightly integrated with 27 00:01:51,853 --> 00:01:54,793 S1: models and platforms, because it's going to allow a lot 28 00:01:54,793 --> 00:01:57,313 S1: more people to do things like this. What I love 29 00:01:57,313 --> 00:02:01,213 S1: about it is you could track all the different experts, right? 30 00:02:01,213 --> 00:02:03,033 S1: I'm going to use a whole bunch of different stuff 31 00:02:03,033 --> 00:02:05,973 S1: for this, but there will be some agent functionality in 32 00:02:05,973 --> 00:02:09,693 S1: the middle to sort of handle, like orchestration and summarization 33 00:02:09,693 --> 00:02:12,423 S1: and creating like an Intel report. But I love the 34 00:02:12,423 --> 00:02:17,283 S1: idea of like gathering all these individual, hopefully standalone, uh, 35 00:02:17,283 --> 00:02:22,863 S1: intelligence sources, aggregating them together, but also keeping them separate 36 00:02:22,863 --> 00:02:27,263 S1: and then triangulating on truth based on that. And I 37 00:02:27,263 --> 00:02:30,773 S1: heard some pretty cool ideas from, um, it was actually 38 00:02:30,773 --> 00:02:33,533 S1: a friend of mine named John, uh, who was talking 39 00:02:33,533 --> 00:02:36,203 S1: about how you want to rate those different sources in 40 00:02:36,203 --> 00:02:39,533 S1: different ways. One way to rate them is to rate 41 00:02:39,533 --> 00:02:43,343 S1: them based on their difference in their uniqueness of ideas 42 00:02:43,343 --> 00:02:46,223 S1: relative to other people, because you don't know if they're 43 00:02:46,223 --> 00:02:50,213 S1: actually just reading other people's Intel and following along, and 44 00:02:50,213 --> 00:02:52,773 S1: you don't want to use eight of those people who 45 00:02:52,773 --> 00:02:57,393 S1: are all following the same thing as eight different ends, right? 46 00:02:57,393 --> 00:03:01,413 S1: Eight different sources of, of data or sources of signal. 47 00:03:01,413 --> 00:03:03,873 S1: So there's a whole bunch of cool stuff that you 48 00:03:03,873 --> 00:03:06,873 S1: could do once you have the discrete signal coming in 49 00:03:06,873 --> 00:03:09,903 S1: from all these different places. And then you could factor 50 00:03:09,903 --> 00:03:13,713 S1: in things like prediction markets and stuff like that and just, uh, 51 00:03:14,133 --> 00:03:16,603 S1: lots of different stuff you could do. But ultimately what 52 00:03:16,603 --> 00:03:20,323 S1: I want is a daily Intel report, which is as 53 00:03:20,323 --> 00:03:22,783 S1: good or better than what you would get from, like 54 00:03:22,783 --> 00:03:26,113 S1: Stratfor back in the day. Or you know what a 55 00:03:26,113 --> 00:03:29,983 S1: lot of these paid platforms would do, um, or even like, 56 00:03:29,983 --> 00:03:32,653 S1: you know, a high level government thing, I think we 57 00:03:32,653 --> 00:03:37,513 S1: could build something really, really good that leverages the intelligence 58 00:03:37,513 --> 00:03:40,383 S1: of all these different, really smart people who are posting 59 00:03:40,383 --> 00:03:42,693 S1: their stuff online. We're not talking about private stuff. We're 60 00:03:42,693 --> 00:03:46,353 S1: talking about people on Twitter. We're talking about people on 61 00:03:46,353 --> 00:03:50,403 S1: different platforms, blogs. They're writing their stuff out there. And 62 00:03:50,403 --> 00:03:52,803 S1: a lot of times nobody's reading it. But if you 63 00:03:52,803 --> 00:03:54,963 S1: put the effort in, you could find all those signals 64 00:03:54,963 --> 00:03:59,283 S1: and start triangulating. So really excited about that. Okay. The 65 00:03:59,283 --> 00:04:03,093 S1: state of things. Yeah. I wrote a long piece about, uh, 66 00:04:03,093 --> 00:04:06,123 S1: this is I posted it on X, a fairly long piece. 67 00:04:06,123 --> 00:04:08,253 S1: I should probably turn it into a full blog, but 68 00:04:08,253 --> 00:04:12,093 S1: it's a little bit, uh, long winded and it's got 69 00:04:12,093 --> 00:04:13,653 S1: some politics in it, so I think I'm going to 70 00:04:13,653 --> 00:04:17,553 S1: skip it, but I recommend going and checking it out 71 00:04:17,553 --> 00:04:20,013 S1: if you're into that kind of stuff. And I spoke 72 00:04:20,013 --> 00:04:24,933 S1: with Christine Gadsby, the head of product security operations at BlackBerry, 73 00:04:24,933 --> 00:04:27,753 S1: and we talked about the role of AI in cybersecurity 74 00:04:27,753 --> 00:04:31,023 S1: and a whole bunch of different topics. The topic list 75 00:04:31,023 --> 00:04:34,773 S1: for this episode is quite large, and you should absolutely 76 00:04:34,773 --> 00:04:37,623 S1: check it out. So so go check that out on 77 00:04:37,623 --> 00:04:39,903 S1: the YouTube or you can click it in the newsletter 78 00:04:39,903 --> 00:04:43,683 S1: as well. So for security, two critical ServiceNow vulnerabilities were 79 00:04:43,683 --> 00:04:47,853 S1: reported by Asset Note. Company has reportedly paid a new 80 00:04:47,853 --> 00:04:51,603 S1: record high 75 million to a ransomware group. And that 81 00:04:51,603 --> 00:04:53,683 S1: seems like a lot of money. But it's not a 82 00:04:53,683 --> 00:04:56,263 S1: lot compared to not being able to do business at all. 83 00:04:56,263 --> 00:04:58,633 S1: So a lot of people kind of beat up people 84 00:04:58,633 --> 00:05:01,603 S1: for paying ransoms, and it really is kind of similar 85 00:05:01,603 --> 00:05:04,123 S1: to like, your kid gets stolen. It's like all the 86 00:05:04,123 --> 00:05:07,993 S1: philosophy goes away when somebody has your kid and it's 87 00:05:07,993 --> 00:05:11,203 S1: the same as a CEO or whoever. When you have 88 00:05:11,203 --> 00:05:14,503 S1: the ability to pay some money and get business back online, sure, 89 00:05:14,503 --> 00:05:18,423 S1: they might just ransom you again. Sure they might do whatever. Sure, 90 00:05:18,423 --> 00:05:21,393 S1: it might be bad for other people. But when business 91 00:05:21,393 --> 00:05:25,623 S1: has stopped, things become quite clear to you in terms 92 00:05:25,623 --> 00:05:28,203 S1: of what you need to do. So I'm not saying 93 00:05:28,203 --> 00:05:30,903 S1: people should pay or anything like that. I'm not making 94 00:05:30,903 --> 00:05:34,593 S1: any judgments. I'm just saying, well, essentially that don't make judgments. 95 00:05:34,683 --> 00:05:37,533 S1: Try to avoid making judgments because it's really hard to 96 00:05:37,533 --> 00:05:42,573 S1: be in that position. Digicert is revoking 83,000 TLS certificates 97 00:05:42,573 --> 00:05:46,653 S1: due to a domain validation bug. China is getting around 98 00:05:46,653 --> 00:05:50,403 S1: US bans on advanced AI chips through smuggling front companies 99 00:05:50,403 --> 00:05:55,323 S1: and loopholes, basically finding ways to get the chips that 100 00:05:55,323 --> 00:05:58,023 S1: are not supposed to be getting. Ransomware attacks are rising, 101 00:05:58,023 --> 00:06:01,683 S1: with an 18% year on year increase reported by Zscaler, 102 00:06:01,683 --> 00:06:05,963 S1: and I've always considered ransomware attacks to be something that 103 00:06:05,963 --> 00:06:08,333 S1: we'd have to invent as a government. It would have 104 00:06:08,333 --> 00:06:10,703 S1: to be like a government service if if it didn't 105 00:06:10,703 --> 00:06:13,193 S1: exist in the marketplace, like as a way to test 106 00:06:13,193 --> 00:06:16,133 S1: for bad security. And maybe you give like a fine 107 00:06:16,133 --> 00:06:18,863 S1: or something if people keep having the mistake. But my 108 00:06:18,863 --> 00:06:21,413 S1: intuition was that after a number of years, it would 109 00:06:21,413 --> 00:06:25,733 S1: get harder and harder because security would increase. So if 110 00:06:25,733 --> 00:06:28,603 S1: these attacks are still increasing, I wonder what the reason is. 111 00:06:28,633 --> 00:06:31,663 S1: Is it because attackers are moving to like the more 112 00:06:31,663 --> 00:06:34,723 S1: vulnerable targets, or are they just getting better at finding 113 00:06:34,723 --> 00:06:37,513 S1: the holes or something else? Or all of the above? 114 00:06:37,513 --> 00:06:39,883 S1: Probably all of the above. But if somebody has more 115 00:06:39,883 --> 00:06:43,693 S1: insight on why things aren't getting tighter or see, that's 116 00:06:43,693 --> 00:06:46,543 S1: the trick is it doesn't mean things aren't getting harder. 117 00:06:46,543 --> 00:06:49,123 S1: Just because the number of attacks are going up doesn't 118 00:06:49,123 --> 00:06:52,783 S1: mean things aren't getting harder. They might just be getting better, faster. 119 00:06:52,813 --> 00:06:56,293 S1: Got a great analysis here of securing secrets in AWS. 120 00:06:56,323 --> 00:07:00,943 S1: Saw the blog post discussing creating custom implants for evasion 121 00:07:00,943 --> 00:07:04,273 S1: by building them in C and this thing details server setup, 122 00:07:04,273 --> 00:07:08,833 S1: client functionality, and testing against security tools. The average cost 123 00:07:08,833 --> 00:07:14,353 S1: of a data breach jumped 10% to 4.88 million. In 23, 124 00:07:14,353 --> 00:07:18,393 S1: China is tightening its civilian drone export rules starting September 125 00:07:18,393 --> 00:07:23,373 S1: 1st to prevent use in military or terrorist activities. Yeah, 126 00:07:23,373 --> 00:07:27,063 S1: I'm trying to figure out if this is CCP trying 127 00:07:27,063 --> 00:07:30,423 S1: to keep it their stuff from being used against them, 128 00:07:30,423 --> 00:07:33,813 S1: or if they're trying to make it easier to sell 129 00:07:33,813 --> 00:07:38,943 S1: their products because they're playing nice and they're appearing to 130 00:07:38,943 --> 00:07:42,713 S1: be good guys. AI and tech OpenAI has started rolling 131 00:07:42,713 --> 00:07:47,873 S1: out its new ChatGPT voice feature for ChatGPT plus users, 132 00:07:47,873 --> 00:07:50,663 S1: and it's quite good. It's it's quite a bit different. 133 00:07:50,663 --> 00:07:53,753 S1: You can basically interrupt it. It sounds a lot more natural. 134 00:07:53,753 --> 00:07:56,213 S1: I am getting a lot of voice artifacts though, like 135 00:07:56,213 --> 00:07:59,003 S1: it'll sound like choppy and broken and a lot of 136 00:07:59,003 --> 00:08:01,763 S1: weird pauses. Not like in a human way, but I 137 00:08:01,763 --> 00:08:04,793 S1: think the platform might be overwhelmed. Or maybe, I don't know, 138 00:08:04,793 --> 00:08:07,273 S1: maybe I need to restart the app. Maybe it was buggy. 139 00:08:07,273 --> 00:08:11,053 S1: Not sure. Lots of I talk at Blackhat, which, uh, yeah, 140 00:08:11,053 --> 00:08:13,963 S1: already here and it's already happening. Another thing to mention 141 00:08:13,963 --> 00:08:17,683 S1: about the ChatGPT stuff is, uh, Greg Brockman is taking 142 00:08:17,683 --> 00:08:22,843 S1: a sabbatical. Uh, John Schulman, I think, is leaving the company. 143 00:08:22,843 --> 00:08:25,423 S1: Is he the one that went to to anthropic? I 144 00:08:25,423 --> 00:08:28,993 S1: can't remember. Another leader went to anthropic and another one 145 00:08:28,993 --> 00:08:31,933 S1: left as well, so the three people left all at once. 146 00:08:31,933 --> 00:08:36,103 S1: But it's not like a mass exodus all to one 147 00:08:36,103 --> 00:08:40,693 S1: place they're not mad at. OpenAI seems to be fairly benign. Um, 148 00:08:40,693 --> 00:08:42,943 S1: but it does look kind of weird to have an 149 00:08:42,943 --> 00:08:45,463 S1: announcement where three people leave at the same time. A 150 00:08:45,463 --> 00:08:48,043 S1: funniest joke I saw about this was that Sam Altman 151 00:08:48,043 --> 00:08:52,093 S1: predicted that soon there would be a one person unicorn company, 152 00:08:52,093 --> 00:08:55,443 S1: and the joke was, yeah, it might be your company. 153 00:08:55,443 --> 00:08:58,203 S1: You might be the only one left. Um, I thought 154 00:08:58,203 --> 00:09:02,703 S1: that was kind of clever. California's SB 1047 safe and 155 00:09:02,703 --> 00:09:07,113 S1: Secure Innovation for Frontier Artificial Intelligence Models Act. That's a 156 00:09:07,113 --> 00:09:10,683 S1: long name. It's looking to regulate large AI models by 157 00:09:10,683 --> 00:09:16,773 S1: mandating safety features to prevent catastrophic incidents. Use risk based 158 00:09:16,773 --> 00:09:20,993 S1: AI regulation began on August 1st, and it's got staggered 159 00:09:21,413 --> 00:09:25,373 S1: deadlines based on low or no risk versus high risk 160 00:09:25,373 --> 00:09:29,093 S1: and limited risk tiers. So that's starting to roll out. 161 00:09:29,093 --> 00:09:33,083 S1: And OpenAI has launched the GPT four long output model. 162 00:09:33,083 --> 00:09:36,053 S1: I've already switched all or at least a lot of 163 00:09:36,053 --> 00:09:39,113 S1: my stuff. I switched my fabric prompt over to this. 164 00:09:39,113 --> 00:09:45,043 S1: So it's got 64 output tokens, 64,000 output tokens, which 165 00:09:45,043 --> 00:09:49,333 S1: is 16 times more than the previous one, and it's 50% 166 00:09:49,333 --> 00:09:53,053 S1: cheaper for most things. And a lot of people are 167 00:09:53,053 --> 00:09:55,393 S1: saying that the benchmarks, it's actually much better than the 168 00:09:55,393 --> 00:09:58,303 S1: previous one. So I consider it just a straight across 169 00:09:58,303 --> 00:10:02,173 S1: upgrade plus being cheaper. So yeah, I already made that change. 170 00:10:02,173 --> 00:10:06,943 S1: Google's experimental Gemini 1.5 Pro has claimed top spot on 171 00:10:06,943 --> 00:10:12,223 S1: a bunch of leaderboards, surpassing GPT four and, uh, sonnet 172 00:10:12,223 --> 00:10:15,433 S1: 3.5 with a score of 1300. I've not used it yet, 173 00:10:15,433 --> 00:10:18,223 S1: because every time I try to use a Google product, 174 00:10:18,223 --> 00:10:20,653 S1: I have to vomit. But I am going to try 175 00:10:20,653 --> 00:10:24,643 S1: again soon to see if it's usable. Meta says it'll 176 00:10:24,643 --> 00:10:27,763 S1: need ten times more computing power to train llama four 177 00:10:27,763 --> 00:10:31,173 S1: compared to llama three. Elliott Management is calling Nvidia a 178 00:10:31,173 --> 00:10:35,493 S1: bubble and says AI is overhyped. They mark. They argue 179 00:10:35,493 --> 00:10:39,333 S1: that the market is overly optimistic about AI's potential and 180 00:10:39,333 --> 00:10:41,943 S1: Nvidia's role in it. I think it's a bubble, but 181 00:10:41,943 --> 00:10:45,393 S1: it's a bubble like the internet in 1995. In other words, 182 00:10:45,393 --> 00:10:48,873 S1: there absolutely will be a burst of lots and lots 183 00:10:48,873 --> 00:10:53,993 S1: of companies, right? Pets.com and companies like that. The AI equivalents, 184 00:10:53,993 --> 00:10:57,173 S1: thousands of those companies are going to fail. Lots of 185 00:10:57,173 --> 00:10:59,603 S1: investors are going to be very sad about this, but 186 00:10:59,603 --> 00:11:03,683 S1: that's completely unrelated to what AI is about to do 187 00:11:03,683 --> 00:11:06,053 S1: to the world. Right? So I think people shouldn't be 188 00:11:06,053 --> 00:11:08,723 S1: confused about those two things. One happening doesn't mean that 189 00:11:08,723 --> 00:11:11,243 S1: the other one is not going to happen. Bellingcat has 190 00:11:11,243 --> 00:11:16,163 S1: put together a guide on identifying explosive ordnance in social 191 00:11:16,163 --> 00:11:20,603 S1: media imagery. CrowdStrike is facing a massive lawsuit after Blue 192 00:11:20,603 --> 00:11:24,383 S1: Friday crashed over 8 million computers globally. Intel is laying 193 00:11:24,383 --> 00:11:26,993 S1: off over 15% of its workforce as part of a 194 00:11:26,993 --> 00:11:30,833 S1: $10 billion cost reduction plan. Apple just posted a record 195 00:11:30,833 --> 00:11:35,993 S1: breaking Q3 2024, $86 billion in revenue. And one thing 196 00:11:35,993 --> 00:11:39,503 S1: that's interesting about this is Berkshire Hathaway just sold a 197 00:11:39,503 --> 00:11:41,903 S1: whole bunch of stuff, uh, a whole bunch of Apple. 198 00:11:41,903 --> 00:11:45,133 S1: And they sold it right before this crash happened. The 199 00:11:45,133 --> 00:11:47,983 S1: crash happened. There was a giant recession that hit the 200 00:11:47,983 --> 00:11:50,143 S1: United States, and then it went away the next day. 201 00:11:50,173 --> 00:11:55,813 S1: Today was like a lot of that money came back. But, um. Yeah. Strange. 202 00:11:55,813 --> 00:11:59,503 S1: Who knows? It could happen again tomorrow. But very volatile, 203 00:11:59,503 --> 00:12:02,893 S1: very emotional sort of time. I feel like in lots 204 00:12:02,893 --> 00:12:05,203 S1: of different ways and I feel like the stock market 205 00:12:05,203 --> 00:12:08,523 S1: is matching that. But the other thing to mention about 206 00:12:08,523 --> 00:12:13,413 S1: Apple is that their services money is now almost equal 207 00:12:13,413 --> 00:12:17,493 S1: to their devices money, which is a huge tipping point 208 00:12:17,493 --> 00:12:20,913 S1: or a milestone in terms of their growth. Apple is 209 00:12:20,913 --> 00:12:24,813 S1: ramping up spending to get Apple intelligence ready for launch 210 00:12:24,813 --> 00:12:27,603 S1: in the fall, and I'm already using the beta, and 211 00:12:27,603 --> 00:12:29,853 S1: it's pretty impressive, even though a lot of the features 212 00:12:29,853 --> 00:12:32,933 S1: aren't rolled out yet. All right, human news, A lot 213 00:12:32,933 --> 00:12:35,903 S1: of the world tried to push Huawei out of their infrastructure, 214 00:12:35,903 --> 00:12:41,003 S1: but they're actually getting more successful, not less. Software company 215 00:12:41,003 --> 00:12:45,893 S1: increased user engagement by eight times by drastically shortening their emails. 216 00:12:45,893 --> 00:12:50,813 S1: Netlify fees, is that it? Yeah. Netlify fees. Initial 150 217 00:12:50,813 --> 00:12:54,473 S1: word emails had a 1% reply rate, but by cutting 218 00:12:54,473 --> 00:12:57,863 S1: the text to 37 words, it went to 4%, and 219 00:12:57,863 --> 00:13:01,433 S1: when they went to 14 words, it went to 8% 220 00:13:01,433 --> 00:13:05,243 S1: 14 words. Last month, Shane Mack offered everyone at his 221 00:13:05,243 --> 00:13:08,903 S1: company $25,000 to quit and six people took it. Yeah, 222 00:13:08,903 --> 00:13:11,933 S1: I think this is part of the Alaskan fishing boat 223 00:13:11,933 --> 00:13:14,513 S1: thing that I wrote a while back. Companies basically want 224 00:13:14,513 --> 00:13:18,233 S1: fully dedicated murderers is all they want. They want people 225 00:13:18,233 --> 00:13:22,093 S1: who eat, live, sleep, think and are obsessed with the company. 226 00:13:22,093 --> 00:13:26,023 S1: That's why they want return to office. That's their way 227 00:13:26,023 --> 00:13:28,993 S1: of filtering for people who who think of the company 228 00:13:28,993 --> 00:13:31,483 S1: as a religion. I mean, they can't say that, but 229 00:13:31,483 --> 00:13:33,013 S1: they can say you have to come to the office 230 00:13:33,013 --> 00:13:36,253 S1: and that's an automatic filter for it. Right? So this 231 00:13:36,253 --> 00:13:38,773 S1: is the way that management and managers and the whole 232 00:13:38,773 --> 00:13:43,263 S1: whole system can basically look for these obsessed people, which 233 00:13:43,263 --> 00:13:47,193 S1: are likely to be in certain demographics, right? Certain ages, 234 00:13:47,193 --> 00:13:51,693 S1: you know, certain groups that are awfully likely to look 235 00:13:51,693 --> 00:13:55,383 S1: kind of similar to each other, probably young, probably without kids, 236 00:13:55,383 --> 00:13:58,803 S1: probably male, who are just grind, grind, grind, don't care 237 00:13:58,803 --> 00:14:02,283 S1: about anything else. Yeah, whatever. Work life balance don't care. 238 00:14:02,283 --> 00:14:04,803 S1: I just want a code or whatever it is. Right. 239 00:14:04,803 --> 00:14:07,893 S1: So that's what these companies are looking for more and more. 240 00:14:07,893 --> 00:14:10,653 S1: And that's why I think and this is just my 241 00:14:10,653 --> 00:14:14,163 S1: hypothesis here, I don't, you know, we need more data 242 00:14:14,163 --> 00:14:18,693 S1: for all this. But my pet hypothesis here is that 243 00:14:18,693 --> 00:14:21,753 S1: this is a factor for all of these layoffs. It's 244 00:14:21,753 --> 00:14:26,373 S1: like this awakening across all of business that you know 245 00:14:26,373 --> 00:14:31,953 S1: what I want hardcore crazy people, religious people about this company. 246 00:14:31,953 --> 00:14:34,233 S1: And I want them to be a players. And I 247 00:14:34,233 --> 00:14:37,323 S1: want them to be really good at AI, and they're 248 00:14:37,323 --> 00:14:39,603 S1: going to help us do even more with AI because 249 00:14:39,603 --> 00:14:41,853 S1: they're going to bring the AI on and blah, blah, blah. 250 00:14:41,853 --> 00:14:44,103 S1: So it's like, I'm going to hire a bunch of 251 00:14:44,103 --> 00:14:47,343 S1: these crazy people, and a team of ten of them 252 00:14:47,343 --> 00:14:49,803 S1: is going to be like having a team of 1000 253 00:14:49,803 --> 00:14:53,853 S1: or 2000 people sometime in the future, in the near future. 254 00:14:53,853 --> 00:14:55,633 S1: Whereas if you get a bunch of people who are 255 00:14:55,633 --> 00:14:59,023 S1: just straight out of college, they're entitled. They think they 256 00:14:59,023 --> 00:15:01,963 S1: are owed something even worse. They think that they're about 257 00:15:01,963 --> 00:15:05,323 S1: to receive training on the job because they don't know 258 00:15:05,323 --> 00:15:07,783 S1: how to do the job. And it's like, okay, well, 259 00:15:07,783 --> 00:15:10,243 S1: now train me. Now, teach me how to do this job. 260 00:15:10,243 --> 00:15:13,243 S1: And all these leaders at these companies are like, I 261 00:15:13,243 --> 00:15:16,123 S1: do not want you. I don't care what degrees you have. 262 00:15:16,123 --> 00:15:18,553 S1: If you can't do the job on day one or 263 00:15:18,553 --> 00:15:22,203 S1: you can't learn instantly like just by seeing it once, 264 00:15:22,203 --> 00:15:23,913 S1: and if you're not obsessed about it and want to 265 00:15:23,913 --> 00:15:26,913 S1: sleep under your desk, we have no use for you. 266 00:15:26,913 --> 00:15:31,263 S1: And unfortunately, that's like 80% of the workforce, I'm guessing, right? 267 00:15:31,263 --> 00:15:35,193 S1: It's like 52, 90% of the workforce, let's call it that. 268 00:15:35,193 --> 00:15:38,703 S1: And what that means is they are looking for that 10%. 269 00:15:38,703 --> 00:15:42,153 S1: They're looking for that 5%. They're looking for the A 270 00:15:42,153 --> 00:15:47,073 S1: players who are dedicated like religious people. And I believe 271 00:15:47,073 --> 00:15:50,403 S1: this is what we're seeing more than anything now. You 272 00:15:50,403 --> 00:15:52,653 S1: add AI on top of that. Now you see why 273 00:15:52,653 --> 00:15:54,693 S1: there's so many layoffs. Now you see why there's so 274 00:15:54,693 --> 00:15:57,933 S1: many open positions, but nobody's hiring for them because they're 275 00:15:57,933 --> 00:16:01,833 S1: kind of like fake positions. And this is multiple hypotheses 276 00:16:01,833 --> 00:16:04,983 S1: all rolled into one. But you get the vibe. This 277 00:16:04,983 --> 00:16:07,623 S1: is the basic vibe of what I think is happening. 278 00:16:07,653 --> 00:16:12,753 S1: Journalist Evan Gershkovich was among a group of Americans and 279 00:16:12,753 --> 00:16:17,343 S1: Russian dissidents released from Russia in a seven nation prisoner swap, 280 00:16:17,343 --> 00:16:21,153 S1: largest ever since the Cold War. Researchers at the University 281 00:16:21,153 --> 00:16:24,513 S1: of California, Santa Barbara have developed an AI model called 282 00:16:24,513 --> 00:16:27,843 S1: shark AI to help prevent shark attacks. The model uses 283 00:16:27,843 --> 00:16:32,143 S1: drones to detect sharks with greater accuracy than humans. I 284 00:16:32,143 --> 00:16:35,323 S1: love this, I love this every time I go to Maui, 285 00:16:35,323 --> 00:16:38,083 S1: I'm stupid and I read the stats about like, shark 286 00:16:38,083 --> 00:16:40,543 S1: attacks and they're like, oh, actually, right next to you 287 00:16:40,543 --> 00:16:42,913 S1: is the most dangerous place. And I'm like, cool. I 288 00:16:42,913 --> 00:16:44,893 S1: didn't want to go in the water anyway. Why did 289 00:16:44,893 --> 00:16:47,803 S1: I read that right before I went on vacation, where 290 00:16:47,803 --> 00:16:50,503 S1: I'm supposed to swim in the water? But anyway, if 291 00:16:50,503 --> 00:16:52,933 S1: I were able to look up and maybe they're so 292 00:16:52,933 --> 00:16:55,813 S1: high up you can't hear them, maybe it's not super annoying. 293 00:16:55,813 --> 00:16:58,513 S1: But anyway, if I know that there's ten of these 294 00:16:58,513 --> 00:17:03,763 S1: drones sweeping back and forth and you know they're being recharged, 295 00:17:03,763 --> 00:17:06,283 S1: they go back on rotation and they look down, they 296 00:17:06,283 --> 00:17:09,223 S1: could see very clearly, if there's a shark in the water, 297 00:17:09,223 --> 00:17:13,093 S1: I assume it wouldn't work. Maybe if the water was muddy, 298 00:17:13,093 --> 00:17:15,313 S1: but maybe you wouldn't be swimming anyway because it would 299 00:17:15,313 --> 00:17:18,703 S1: be dangerous water anyway. Usually in a lot of places 300 00:17:18,703 --> 00:17:21,153 S1: you could see right through the water. It's very easy 301 00:17:21,153 --> 00:17:24,783 S1: to see a shark from above, and they just call 302 00:17:24,783 --> 00:17:29,643 S1: the lifeguard station and trigger an alert, and they blow 303 00:17:29,643 --> 00:17:31,743 S1: the whistle and everyone gets out of the water, like, 304 00:17:31,743 --> 00:17:34,893 S1: that's going to be amazing. Love it. Treating failing eyesight 305 00:17:34,893 --> 00:17:37,173 S1: and high cholesterol are two new ways to lower the 306 00:17:37,173 --> 00:17:40,863 S1: risk of developing dementia, according to a major report. The 307 00:17:40,863 --> 00:17:46,283 S1: Lancet Commission's latest findings suggest that addressing 14 health issues 308 00:17:46,283 --> 00:17:50,573 S1: could theoretically prevent nearly half of all dementia cases worldwide. 309 00:17:50,573 --> 00:17:53,813 S1: And I believe from reading this that essentially they're talking 310 00:17:53,813 --> 00:17:57,293 S1: about things that just exacerbate it and make it worse. So, 311 00:17:57,293 --> 00:18:00,413 S1: for example, if you can't really see things, you generally 312 00:18:00,413 --> 00:18:02,813 S1: maybe you don't go out a lot. If you can't 313 00:18:02,813 --> 00:18:05,783 S1: hear conversations, you're not involved in conversations. So I think 314 00:18:05,783 --> 00:18:08,893 S1: a lot of this might be related to social interaction, 315 00:18:08,893 --> 00:18:12,013 S1: which once you start to get isolated and you're not 316 00:18:12,013 --> 00:18:15,163 S1: consuming media, you're not reading, you're not like, there's no 317 00:18:15,163 --> 00:18:18,583 S1: new inputs. Um, again, this is my hypothesis. I believe 318 00:18:18,583 --> 00:18:20,893 S1: this is based on some solid science I've already read, 319 00:18:20,893 --> 00:18:24,523 S1: though is basically, once you get isolated in that way, 320 00:18:24,523 --> 00:18:28,093 S1: your brain starts like shutting down and it really accelerates 321 00:18:28,093 --> 00:18:31,933 S1: the dementia. So that that would make sense if, uh, 322 00:18:31,933 --> 00:18:36,503 S1: that's what they're saying in this paper. Self-control is about 60% heritable, 323 00:18:36,503 --> 00:18:40,433 S1: meaning genes explain roughly 60% of the differences in self-control 324 00:18:40,433 --> 00:18:44,213 S1: among individuals. So I think this could be devastating if 325 00:18:44,213 --> 00:18:46,913 S1: it's supported in further studies. I worry about the narrative 326 00:18:46,913 --> 00:18:51,953 S1: that both IQ and self-discipline are mostly genetic, thus giving 327 00:18:51,953 --> 00:18:55,523 S1: people an easy ramp to write off individuals or even 328 00:18:55,523 --> 00:18:59,973 S1: groups if they have lower averages of these things. And 329 00:19:00,423 --> 00:19:03,723 S1: I think even if it were true, the groups don't 330 00:19:03,723 --> 00:19:07,893 S1: define the individuals and the study mentioned individuals here. It's 331 00:19:07,893 --> 00:19:10,263 S1: not talking about groups, but, you know, people are going 332 00:19:10,263 --> 00:19:13,683 S1: to people. Right? So the the other thing is there's 333 00:19:13,683 --> 00:19:17,433 S1: likely a lot of slack in, say, the 40%, which 334 00:19:17,433 --> 00:19:21,053 S1: is environmental, assuming those numbers are correct, like we're probably 335 00:19:21,053 --> 00:19:24,683 S1: getting whatever 10 or 20% of the 40% we're supposed 336 00:19:24,683 --> 00:19:28,373 S1: to be doing. So if we were to increase, you know, 337 00:19:28,373 --> 00:19:31,823 S1: the efforts of, you know, training and culture and all 338 00:19:31,823 --> 00:19:35,873 S1: the environmental things we can control, I think that would raise, 339 00:19:35,873 --> 00:19:39,323 S1: you know, the bar for, well, everyone, but especially the 340 00:19:39,323 --> 00:19:41,873 S1: bottom quite a bit. So I'm not sure this is 341 00:19:41,873 --> 00:19:45,493 S1: really anything too much to despair about other than making 342 00:19:45,493 --> 00:19:49,603 S1: it easier for certain negative narratives. A new study reveals 343 00:19:49,603 --> 00:19:52,963 S1: that people tend to alter their appearance to match their names. 344 00:19:53,173 --> 00:19:56,773 S1: Researchers found that adults faces often align with a social 345 00:19:56,773 --> 00:20:01,423 S1: stereotype associated with their name, while children's faces do not 346 00:20:01,423 --> 00:20:05,113 S1: show this pattern. A key protein called reelin may help 347 00:20:05,113 --> 00:20:08,473 S1: stave off Alzheimer's disease. A number of new studies suggest 348 00:20:08,473 --> 00:20:13,093 S1: that reelin helps maintain thinking and memory in aging brains, 349 00:20:13,093 --> 00:20:17,383 S1: and when its levels fall off, neurons become more vulnerable, 350 00:20:17,383 --> 00:20:21,073 S1: and people are starting to obviously work on drugs for this. 351 00:20:21,073 --> 00:20:24,883 S1: Wizards of the coast will release the 2024 Dungeons and 352 00:20:24,883 --> 00:20:30,643 S1: Dragons rulebooks under a Creative Commons license, which is fulfilling 353 00:20:30,643 --> 00:20:33,393 S1: a promise they made after the backlash over attempts to 354 00:20:33,393 --> 00:20:36,693 S1: change the Open Gaming License. If novelists wrote Your bug 355 00:20:36,693 --> 00:20:40,293 S1: reports imagines how famous authors would describe software bugs in 356 00:20:40,293 --> 00:20:44,943 S1: their unique styles. Ernest Klein likens a screen flicker to 357 00:20:44,943 --> 00:20:48,603 S1: scenes from back to the Future and Ghostbusters, while Ursula 358 00:20:48,603 --> 00:20:55,023 S1: K Le Guin philosophizes about the existential pain of coding errors. Ideas. 359 00:20:55,053 --> 00:20:58,373 S1: More analysis on how bad the results were of the 360 00:20:58,373 --> 00:21:03,233 S1: recent UBI study done by Sam Altman. It looks pretty bad, 361 00:21:03,233 --> 00:21:05,813 S1: just like we talked about last week, and got a 362 00:21:05,813 --> 00:21:08,903 S1: link here to go into that in depth. And really 363 00:21:08,903 --> 00:21:13,463 S1: cool idea from Jonathan Haidt about free range kids. And 364 00:21:13,463 --> 00:21:16,523 S1: a cool idea for giving them freedom is to create 365 00:21:16,523 --> 00:21:19,313 S1: a play street once a month where you close off 366 00:21:19,313 --> 00:21:22,913 S1: a street for two hours, give time for kids to 367 00:21:22,913 --> 00:21:25,703 S1: play in the street safely, and then the whole time 368 00:21:25,703 --> 00:21:29,933 S1: that the parents are there watching, like around the edges and, 369 00:21:29,933 --> 00:21:33,263 S1: you know, whatever. But the neighbors are also meeting and talking, 370 00:21:33,263 --> 00:21:37,043 S1: and he's saying it has transformative effects on the neighborhood 371 00:21:37,043 --> 00:21:40,463 S1: and just good times all around. I really love ideas 372 00:21:40,463 --> 00:21:44,423 S1: like this. Discovery Farmbot is an open source farming machine 373 00:21:44,423 --> 00:21:47,413 S1: for growing food in your own backyard. Super memory an 374 00:21:47,413 --> 00:21:51,583 S1: AI powered platform to organize, search and utilize saved information 375 00:21:51,583 --> 00:21:56,563 S1: acting as a digital second brain. Friend is Avi Shiffman's 376 00:21:56,563 --> 00:22:00,583 S1: new AI pendant, and it's designed to combat loneliness by 377 00:22:00,583 --> 00:22:04,483 S1: sending you reassuring or playful text based on what it 378 00:22:04,483 --> 00:22:09,073 S1: overhears so it doesn't have a speaker. It actually sends 379 00:22:09,073 --> 00:22:12,573 S1: you notifications. Kind of interesting way to do that. Daniel 380 00:22:12,573 --> 00:22:15,903 S1: Cosman walks you through installing fabric and open source AI 381 00:22:15,933 --> 00:22:20,283 S1: framework by Daniel Missler. That's weird. I wonder if I 382 00:22:20,283 --> 00:22:22,473 S1: wrote that because I don't talk about myself in the 383 00:22:22,473 --> 00:22:25,383 S1: third person. Fleet is an open source version of fleet 384 00:22:25,383 --> 00:22:30,303 S1: DMs tool built on OS query for vulnerability monitoring, MDM 385 00:22:30,303 --> 00:22:35,903 S1: detection engineering and more. Soc2 policy Templates collection of templates 386 00:22:35,903 --> 00:22:40,373 S1: for Soc2 policies and procedures. Clutch security is a platform 387 00:22:40,373 --> 00:22:46,643 S1: providing visibility into all non-human identities within an organization, helping 388 00:22:46,643 --> 00:22:50,723 S1: them identify associated risks and the recommendation of the week. 389 00:22:50,723 --> 00:22:53,243 S1: If you're at Blackhat this week, remember that ten and 390 00:22:53,243 --> 00:22:56,213 S1: 20 years from now, you will not remember the talks 391 00:22:56,213 --> 00:22:59,903 S1: that you saw this year, but you will remember spending 392 00:22:59,903 --> 00:23:04,073 S1: that time with your friends. So prioritize friend time over 393 00:23:04,073 --> 00:23:07,433 S1: presentation time. Not only is the friend time more precious 394 00:23:07,433 --> 00:23:10,433 S1: and valuable, but you can get the talks later if 395 00:23:10,433 --> 00:23:12,923 S1: you really want to. In the aphorism of the week, 396 00:23:12,923 --> 00:23:16,133 S1: friends show their love in times of trouble, not in happiness. 397 00:23:16,133 --> 00:23:22,553 S1: Friends show their love in times of trouble, not in happiness. Euripides.