1 00:00:00,800 --> 00:00:04,939 S1: Unsupervised Learning is a podcast about trends and ideas in cybersecurity, 2 00:00:04,970 --> 00:00:10,010 S1: national security, AI, technology and society, and how best to 3 00:00:10,039 --> 00:00:17,660 S1: upgrade ourselves to be ready for what's coming. All right, 4 00:00:17,660 --> 00:00:22,369 S1: welcome to episode 472. Hope your week is going well. 5 00:00:22,400 --> 00:00:26,180 S1: Updates on this side I've got an RSS feed which 6 00:00:26,180 --> 00:00:29,180 S1: I encourage you to go sign up for. RSS is 7 00:00:29,180 --> 00:00:32,449 S1: a real thing and it's awesome. And I wish it 8 00:00:32,450 --> 00:00:37,430 S1: didn't get knocked off the radar by Google Reader going under, 9 00:00:37,760 --> 00:00:41,660 S1: but I use Feedly. A lot of people use Netnewswire. 10 00:00:41,690 --> 00:00:45,530 S1: Whatever your RSS reader is, go check it out. It's 11 00:00:45,530 --> 00:00:49,310 S1: just slash feed RSS, I think at the end of 12 00:00:49,310 --> 00:00:52,130 S1: the URL. Thanks for all the cooking advice. Around 500 13 00:00:52,130 --> 00:00:57,650 S1: people responded with experience, encouragement, recommendations. It was just insane. 14 00:00:57,650 --> 00:01:01,760 S1: So really appreciate all that. I tried windsurf. A lot 15 00:01:01,790 --> 00:01:03,560 S1: of people are talking about it. I basically try all 16 00:01:03,560 --> 00:01:07,310 S1: these tools that people talk about. I've been using Klein 17 00:01:07,310 --> 00:01:11,060 S1: for a while. I was using cursor before that and 18 00:01:11,060 --> 00:01:13,640 S1: I keep trying cursor every time they have a big 19 00:01:13,640 --> 00:01:17,419 S1: release or when windsurf has a big release, I'll try 20 00:01:17,420 --> 00:01:20,570 S1: it again. The stuff is changing so fast that I 21 00:01:20,569 --> 00:01:24,740 S1: just constantly try different stuff. But Klein has been my 22 00:01:24,770 --> 00:01:29,030 S1: go to. I feel like it manages the functionality of 23 00:01:29,060 --> 00:01:33,140 S1: Lmms better and the UI. I just like better, so 24 00:01:33,170 --> 00:01:35,840 S1: that's what I've been using. It is just an extension 25 00:01:35,840 --> 00:01:38,929 S1: into VS code. I wish it was actually a full editor, 26 00:01:38,930 --> 00:01:44,390 S1: just like cursor and windsurf are. So Klein people. If 27 00:01:44,390 --> 00:01:48,230 S1: you're listening, please make an editor. And my personal AI 28 00:01:48,230 --> 00:01:52,160 S1: infrastructure is getting absolutely insane right now. I'm going to 29 00:01:52,160 --> 00:01:54,500 S1: have something to show on this soon. I'm going to 30 00:01:54,500 --> 00:01:57,080 S1: do a demo of like how this all works together, 31 00:01:57,080 --> 00:01:59,880 S1: but basically it's one main agent and then all my 32 00:01:59,880 --> 00:02:05,040 S1: personal services are sitting behind it that do different things, 33 00:02:05,040 --> 00:02:09,330 S1: and they're all able to access different services that do 34 00:02:09,360 --> 00:02:13,049 S1: output as well. So there's like you send the request 35 00:02:13,080 --> 00:02:17,850 S1: into the agent. The agent, if I tell it, send me, 36 00:02:17,880 --> 00:02:21,180 S1: you know, test a website and then send me an 37 00:02:21,180 --> 00:02:25,050 S1: email for it, it'll do the two different services. One 38 00:02:25,050 --> 00:02:27,480 S1: is to test the website and the other one is 39 00:02:27,480 --> 00:02:31,050 S1: to send an email. And again it just naturally parsed that. 40 00:02:31,050 --> 00:02:33,660 S1: And I just sent that from my phone. Right. So 41 00:02:33,660 --> 00:02:37,440 S1: the whole idea is I'm just using my regular tech infrastructure, 42 00:02:37,440 --> 00:02:40,440 S1: which is my iPhone, and I send a request into 43 00:02:40,470 --> 00:02:42,240 S1: my agent, and this is what it's going to be 44 00:02:42,240 --> 00:02:45,600 S1: in the future too. I'm just sort of hacking it 45 00:02:45,600 --> 00:02:48,450 S1: together with my own services. Pretty soon it'll just be 46 00:02:48,450 --> 00:02:52,770 S1: built into Google and built into the iPhone and stuff 47 00:02:52,770 --> 00:02:55,200 S1: like that. But right now I'm just making it happen 48 00:02:55,200 --> 00:03:00,359 S1: through like Shortcuts and a bunch of AI services, but 49 00:03:00,360 --> 00:03:04,800 S1: the overall effect is quite awesome and you have to 50 00:03:04,800 --> 00:03:08,550 S1: see this thing actually working. It's very much in line 51 00:03:08,550 --> 00:03:12,570 S1: with what I've been talking about, uh, in the predictable 52 00:03:12,600 --> 00:03:16,590 S1: AI post, which is this link here where it's all going, 53 00:03:16,620 --> 00:03:21,269 S1: put out a concise explanation of model context protocol, MCP 54 00:03:21,270 --> 00:03:24,570 S1: servers and why everyone's excited about them. So I've got 55 00:03:24,570 --> 00:03:28,440 S1: a full thread on that and you should go check out. And, uh, 56 00:03:28,440 --> 00:03:32,310 S1: I'm enjoying the fact that this is starting to happen, 57 00:03:32,340 --> 00:03:37,140 S1: specifically the MCP server, because in that predictable AI post, 58 00:03:37,140 --> 00:03:40,440 S1: which actually comes from the book in 2016, I talked 59 00:03:40,440 --> 00:03:45,870 S1: about the API ification of everything and specifically how businesses 60 00:03:45,870 --> 00:03:49,950 S1: would all become APIs. So this is text. From that, 61 00:03:49,980 --> 00:03:53,820 S1: most software businesses will become algorithms presented to others through 62 00:03:53,820 --> 00:03:58,470 S1: their business daemons. Many traditional businesses will continue to become 63 00:03:58,470 --> 00:04:04,170 S1: software businesses. So basically, Andreessen talked about software as eating 64 00:04:04,170 --> 00:04:06,870 S1: the world. I think AI is going to eat the world. 65 00:04:06,900 --> 00:04:10,440 S1: Like that's pretty. Everyone's saying that it doesn't really mean anything. 66 00:04:10,440 --> 00:04:13,260 S1: What I think is actually eating the world is APIs. 67 00:04:13,260 --> 00:04:16,650 S1: And the fact that AI is what's going to interact 68 00:04:16,650 --> 00:04:21,270 S1: with all these APIs. The primary interface for doing anything 69 00:04:21,270 --> 00:04:25,109 S1: is going to be APIs, because it's the AI that's 70 00:04:25,110 --> 00:04:30,030 S1: parsing them, right? What humans need to look at a 71 00:04:30,029 --> 00:04:32,549 S1: website or a list of services, or a menu of 72 00:04:32,550 --> 00:04:35,729 S1: food or whatever. What humans need for that is way 73 00:04:35,730 --> 00:04:39,540 S1: different than what a computer needs. A computer can parse 74 00:04:39,570 --> 00:04:46,080 S1: an API or an MCP service directory, you know, basically 75 00:04:46,080 --> 00:04:49,440 S1: instantly and know exactly how to use it. So if 76 00:04:49,470 --> 00:04:52,440 S1: you if you think about SEO, if you think about UI, UX, 77 00:04:52,440 --> 00:04:55,170 S1: if you think about advertisements, if you think about websites, 78 00:04:55,170 --> 00:04:59,400 S1: if you think about news media, all of that, it's 79 00:04:59,400 --> 00:05:02,970 S1: all right now focused on presenting that to humans. The 80 00:05:02,970 --> 00:05:06,270 S1: fundamental change that's happening right now is that's all switching 81 00:05:06,270 --> 00:05:11,250 S1: away from being primarily designed for humans to being primarily 82 00:05:11,250 --> 00:05:16,140 S1: designed for AI. Primarily, these interfaces will be for AI 83 00:05:16,170 --> 00:05:20,070 S1: to find a business, AI to find a service AI 84 00:05:20,100 --> 00:05:23,669 S1: to rank those services. Right. So examples will include and 85 00:05:23,670 --> 00:05:26,580 S1: this is keep in mind this text right here. This 86 00:05:26,580 --> 00:05:30,090 S1: is from 2016 when I wrote that stupid book, which 87 00:05:30,089 --> 00:05:34,110 S1: you probably don't need to read anymore because it's all 88 00:05:34,110 --> 00:05:37,500 S1: in that, that post. And it's got pictures which took 89 00:05:37,500 --> 00:05:40,169 S1: me like hours to make. So it's a better version 90 00:05:40,170 --> 00:05:42,930 S1: of the book that blog post is. But the key 91 00:05:42,930 --> 00:05:45,570 S1: point here is I was talking about this in 2016, 92 00:05:45,570 --> 00:05:48,510 S1: so I'm looking for a little like, I don't know, 93 00:05:48,510 --> 00:05:50,820 S1: somebody to send me a pat on the back. That 94 00:05:50,820 --> 00:05:52,830 S1: was kind of the whole purpose of writing that stupid 95 00:05:52,830 --> 00:05:56,220 S1: book was to get this stuff locked in because I 96 00:05:56,220 --> 00:05:59,940 S1: saw this happening so clearly, even though that was way 97 00:05:59,940 --> 00:06:03,809 S1: before this AI stuff popped. Right. So what is 22 98 00:06:03,839 --> 00:06:07,920 S1: the end of 22 versus the end of 2016? That's 99 00:06:07,920 --> 00:06:11,040 S1: six years. Six years. I basically said, we're going to 100 00:06:11,040 --> 00:06:15,870 S1: have digital assistants, which are AI powered. I talked about 101 00:06:15,870 --> 00:06:18,420 S1: the AI component right there in there. It's going to 102 00:06:18,450 --> 00:06:21,450 S1: know everything about us. We're going to talk to our Da. 103 00:06:21,480 --> 00:06:24,810 S1: Our Da is going to read all the services. Every 104 00:06:24,839 --> 00:06:26,730 S1: human is going to have a demon. Not every human, 105 00:06:26,730 --> 00:06:29,070 S1: but most humans. A lot of a lot of humans 106 00:06:29,070 --> 00:06:31,469 S1: are going to have demons. But most importantly, all the 107 00:06:31,470 --> 00:06:35,790 S1: businesses have demons. We talk to our Da. The Da 108 00:06:36,000 --> 00:06:39,570 S1: infers what we probably want. We don't even need to 109 00:06:39,600 --> 00:06:43,560 S1: talk to it. It's the one interacting with the world, right? 110 00:06:43,560 --> 00:06:47,250 S1: And there's a millions upon millions of other DA's out 111 00:06:47,250 --> 00:06:50,760 S1: there working for their principles. Some are working for businesses, 112 00:06:50,760 --> 00:06:56,400 S1: some are working for Individuals. Some are working for organizations, entities, whatever. 113 00:06:56,460 --> 00:06:58,500 S1: But the point is, they're the ones doing the hard 114 00:06:58,500 --> 00:07:01,710 S1: work of parsing all these millions of demons. Right. So 115 00:07:01,740 --> 00:07:06,180 S1: examples finding gifts customized to the exact individual organizing person. 116 00:07:06,240 --> 00:07:10,290 S1: Perfect vacation based on the four people going, navigating all 117 00:07:10,290 --> 00:07:12,630 S1: the logistics and the best way possible of that four 118 00:07:12,660 --> 00:07:16,020 S1: person vacation. Determining the current mood of a person or 119 00:07:16,020 --> 00:07:19,350 S1: location based on what is known about them. Finding the 120 00:07:19,350 --> 00:07:23,010 S1: optimal route from one place to another. Determining the best 121 00:07:23,010 --> 00:07:26,520 S1: way to charge a customer based on evolving competition, logistics, 122 00:07:26,520 --> 00:07:29,250 S1: and conditions on the ground. All these tasks will be 123 00:07:29,250 --> 00:07:33,120 S1: services available online. There will be several or thousands of 124 00:07:33,120 --> 00:07:37,740 S1: competitors in any particular space. There will even be services 125 00:07:37,740 --> 00:07:40,620 S1: that consume and rate those services and present their output 126 00:07:40,620 --> 00:07:44,760 S1: through their own demons. Right. So that's another demon. So 127 00:07:44,790 --> 00:07:49,290 S1: my digital assistant, my personal I will be using those 128 00:07:49,290 --> 00:07:53,110 S1: services to find the best service to organize a trip, 129 00:07:53,110 --> 00:07:56,770 S1: or to play the perfect song, or to do basically anything. 130 00:07:56,770 --> 00:08:02,260 S1: All these SaaS services, all these B2C services companies. Forget 131 00:08:02,290 --> 00:08:05,320 S1: a human going to the website and clicking around. That's 132 00:08:05,320 --> 00:08:08,470 S1: all done. That is all done. Our Das will be 133 00:08:08,470 --> 00:08:13,000 S1: doing that for us through their demons, through their business demons, 134 00:08:13,000 --> 00:08:16,989 S1: which maybe MCP is that. And that's where we are. 135 00:08:17,020 --> 00:08:21,070 S1: This is essentially what MSPs are finally starting to happen. 136 00:08:21,220 --> 00:08:24,970 S1: MSPs are a way to take any application and turn 137 00:08:24,970 --> 00:08:30,400 S1: it into globally available services. That's why this MCP thing 138 00:08:30,400 --> 00:08:35,080 S1: is so exciting. It's taking any application and presenting it 139 00:08:35,470 --> 00:08:39,640 S1: to I. So I like what we're doing now is 140 00:08:39,670 --> 00:08:43,060 S1: we're taking IDs and we're loading in these marketplaces. So 141 00:08:43,059 --> 00:08:46,420 S1: we have the ability to manually choose these things. But 142 00:08:46,420 --> 00:08:50,080 S1: pretty soon our IDs and more importantly, our digital assistants 143 00:08:50,110 --> 00:08:53,439 S1: are just going to there's going to be natural, like 144 00:08:53,470 --> 00:08:58,240 S1: vetted high quality lists of these MSPs of these demons 145 00:08:58,240 --> 00:09:00,790 S1: out there, and they're going to be super high quality. 146 00:09:00,820 --> 00:09:02,619 S1: They're going to be rated for quality. They're going to 147 00:09:02,620 --> 00:09:05,290 S1: be rated for uptime. They're going to be rated for, 148 00:09:05,320 --> 00:09:08,650 S1: you know, how good the results are and everything. So 149 00:09:08,650 --> 00:09:11,710 S1: my Da is just going to be constantly using all 150 00:09:11,710 --> 00:09:15,459 S1: of those, right. So when I'm like hey I've got sniffles. 151 00:09:15,460 --> 00:09:19,660 S1: It's going to be find the best, uh, pharmacy, find 152 00:09:19,660 --> 00:09:23,710 S1: the best medication, do all the research. Um, find the 153 00:09:23,710 --> 00:09:27,459 S1: best delivery thing and within, you know, like six minutes, 154 00:09:27,550 --> 00:09:31,090 S1: somebody drops it off on my porch, right. It's it's 155 00:09:31,090 --> 00:09:34,240 S1: going to be that simple because it made 400 different 156 00:09:34,240 --> 00:09:38,620 S1: requests to multiple different demons to be able to do 157 00:09:38,620 --> 00:09:41,080 S1: the research and figure that out and find the best one. 158 00:09:41,110 --> 00:09:43,870 S1: Turns out I paid like a tiny amount of money 159 00:09:43,870 --> 00:09:47,650 S1: for that, or whatever the the amount was probably won't 160 00:09:47,650 --> 00:09:51,190 S1: even be a human delivery. It'll be like a Waymo delivery. Anyway, 161 00:09:51,280 --> 00:09:54,850 S1: bottom bottom line is we all get digital assistants. Everything 162 00:09:54,850 --> 00:09:58,450 S1: gets an API, including people, objects and businesses. Our Das 163 00:09:58,480 --> 00:10:01,569 S1: are the ones interacting with the world on our behalf, 164 00:10:01,570 --> 00:10:04,600 S1: because there will be billions of APIs or demons out 165 00:10:04,600 --> 00:10:08,679 S1: there to interact with, and our Das will. Ultimately, this 166 00:10:08,679 --> 00:10:11,680 S1: is the AR piece, which is not quite happening yet, 167 00:10:11,679 --> 00:10:14,770 S1: but everyone's building it. It will display in our are 168 00:10:14,800 --> 00:10:17,890 S1: the best version of of the world to look at, right? 169 00:10:17,920 --> 00:10:20,500 S1: If we're in a security situation, it'll turn to like 170 00:10:20,530 --> 00:10:23,439 S1: a security filter for in like a dating situation, it'll 171 00:10:23,440 --> 00:10:25,990 S1: turn into like a dating filter. If we're in like 172 00:10:26,020 --> 00:10:31,810 S1: a library, it'll like. Overlay interesting topics based on the 173 00:10:31,809 --> 00:10:34,000 S1: books that we're currently looking at. It'll be like, oh, 174 00:10:34,030 --> 00:10:37,569 S1: you know, Jason really liked this book and Chris really 175 00:10:37,570 --> 00:10:41,440 S1: liked that book or whatever. Or, hey, you were reading 176 00:10:41,440 --> 00:10:44,410 S1: something about this. This book will go into it in 177 00:10:44,410 --> 00:10:49,030 S1: more detail, right? So contextual display of information within your 178 00:10:49,030 --> 00:10:52,480 S1: are everyone's currently building Das mch and lots of other 179 00:10:52,480 --> 00:10:54,939 S1: projects are APIs for everything. And lots of people are 180 00:10:54,940 --> 00:10:58,569 S1: working on the AR stuff as well, like meta and uh, 181 00:10:58,570 --> 00:11:04,870 S1: I'm sure Apple eventually. All right. Security. Cybersecurity. Critical. PHP. CGI. 182 00:11:04,900 --> 00:11:09,670 S1: Remote code execution. Vulnerability is being mass exploited against windows systems. 183 00:11:09,700 --> 00:11:12,730 S1: It was originally a bunch of Japanese targets. Now it's gone. 184 00:11:12,730 --> 00:11:17,380 S1: Global help Net Security put together a list of 28 185 00:11:17,500 --> 00:11:23,500 S1: cybersecurity positions across the US. President is nominated Sean Plinky, 186 00:11:23,530 --> 00:11:27,640 S1: Plonky veteran with experience at US Cyber Command in the 187 00:11:27,640 --> 00:11:31,689 S1: previous Trump administration to head Cisa. So Cisa has a 188 00:11:31,690 --> 00:11:35,739 S1: new leader. Malicious poisoning of AI narratives. So TechCrunch put 189 00:11:35,740 --> 00:11:40,630 S1: out a report that Russian propaganda was there's a basically 190 00:11:40,660 --> 00:11:45,040 S1: a network called Pravda, which is a Russian propaganda network, 191 00:11:45,040 --> 00:11:49,700 S1: is flooding the web with fake news about. I assume 192 00:11:49,700 --> 00:11:53,390 S1: it's like, uh, Russian narratives around Ukraine and a whole 193 00:11:53,390 --> 00:11:56,329 S1: bunch of stuff that just Russia wants everyone to believe. 194 00:11:56,360 --> 00:12:00,380 S1: This is crazy. This is an actual AI poisoning attack. 195 00:12:00,380 --> 00:12:04,760 S1: Real world because everyone was worried about you. Poison the 196 00:12:04,760 --> 00:12:07,790 S1: training data. Oh, but it's like, well, how are you 197 00:12:07,790 --> 00:12:10,130 S1: going to break into the company and poison the training data? 198 00:12:10,160 --> 00:12:14,929 S1: Turns out these bots, like OpenAI, like anthropic, they're all 199 00:12:14,929 --> 00:12:19,099 S1: constantly parsing the internet. You don't have to poison the 200 00:12:19,100 --> 00:12:23,329 S1: original pre-training stuff. Well, actually, that's being poisoned as well, 201 00:12:23,360 --> 00:12:26,359 S1: because every time you do a pre-training, you have to 202 00:12:26,390 --> 00:12:30,020 S1: re harvest from the internet. If you change the internet, 203 00:12:30,020 --> 00:12:34,309 S1: then you have poisoned it, right? So, for example, if 204 00:12:34,340 --> 00:12:37,819 S1: this this group, Pravda is trying to change the narrative 205 00:12:37,820 --> 00:12:44,960 S1: around something. So let's say this company Pravda successfully saturates 206 00:12:44,960 --> 00:12:49,250 S1: the entire world with the idea that Ukraine attacked first. Okay, 207 00:12:49,280 --> 00:12:51,800 S1: Ukraine came over the border and killed a bunch of 208 00:12:51,800 --> 00:12:55,760 S1: innocent people. And what Russia did was they did a counterattack. 209 00:12:55,760 --> 00:12:58,910 S1: And this is purely defensive. I'm not saying that's that's 210 00:12:58,910 --> 00:13:02,690 S1: what the propaganda said. Let's just take it as an example. Well, 211 00:13:02,690 --> 00:13:06,410 S1: if they put out millions of articles like that across 212 00:13:06,410 --> 00:13:09,140 S1: the entire internet, and then you parse the internet and 213 00:13:09,140 --> 00:13:12,770 S1: you put it into your training data for the next 214 00:13:12,770 --> 00:13:16,400 S1: version of sonnet or the next version of, uh, you know, 215 00:13:16,429 --> 00:13:22,670 S1: O4 or GPT five or whatever it is. Well, how 216 00:13:22,670 --> 00:13:24,710 S1: is the thing? How was I going to know that's 217 00:13:24,710 --> 00:13:28,100 S1: not real? It's it's parsing the internet to figure out 218 00:13:28,100 --> 00:13:32,990 S1: what's real. So this is an actual AI data poisoning attack. 219 00:13:33,140 --> 00:13:36,890 S1: Ten leading chat bots were tested. Chatbots were tested for 220 00:13:36,890 --> 00:13:44,510 S1: susceptibility to Russian disinformation. Pravda published 3.6 million misleading articles 221 00:13:44,510 --> 00:13:50,210 S1: just in 2024. 3.6 million misleading articles and news guards 222 00:13:50,210 --> 00:13:55,819 S1: analysis found that the chatbots collectively repeated false information a 223 00:13:55,820 --> 00:14:00,860 S1: third of the time. That's insane. Real world AI data poisoning. 224 00:14:01,130 --> 00:14:03,590 S1: So it used to be about SEO poisoning. Now it's 225 00:14:03,590 --> 00:14:06,829 S1: about flooding the internet with marketing or propaganda that will 226 00:14:06,830 --> 00:14:10,010 S1: get you picked up by up by bots. Now I 227 00:14:10,010 --> 00:14:13,250 S1: say marketing and propaganda. It's the same thing, right? Let's 228 00:14:13,250 --> 00:14:17,060 S1: say you have something. You have a an endpoint product 229 00:14:17,059 --> 00:14:20,810 S1: that's nowhere near as good as CrowdStrike, but you hire 230 00:14:20,810 --> 00:14:25,610 S1: Pravda and you blast out to the internet that, you know, 231 00:14:25,640 --> 00:14:30,560 S1: secure point is way, way better than CrowdStrike. And all 232 00:14:30,560 --> 00:14:33,290 S1: the bots parse this thing. And a third of the time, 233 00:14:33,290 --> 00:14:38,120 S1: Secure Point comes back as better than CrowdStrike. That would 234 00:14:38,120 --> 00:14:40,700 S1: be massive. I mean, that that would get you promoted 235 00:14:40,700 --> 00:14:43,950 S1: on a marketing team because guess what? The new search 236 00:14:43,950 --> 00:14:47,010 S1: engine is actually AI. A lot. A lot of people 237 00:14:47,010 --> 00:14:53,640 S1: are actually asking AI to compare products and to come 238 00:14:53,670 --> 00:14:56,670 S1: back because. Because it knows that the AI is read 239 00:14:56,670 --> 00:15:01,920 S1: everything and is is summarizing the results of all of that. Well, 240 00:15:01,920 --> 00:15:06,540 S1: that changes that changes significantly when now everyone's going to 241 00:15:06,540 --> 00:15:10,530 S1: switch to this. So, so all these all these LLM 242 00:15:10,530 --> 00:15:13,740 S1: providers are going to have to figure out a way 243 00:15:13,740 --> 00:15:17,040 S1: to filter out poisoning. I don't see an easy way 244 00:15:17,040 --> 00:15:19,110 S1: to do it. You're going to need a lot of 245 00:15:19,110 --> 00:15:22,860 S1: really smart I just to do the data cleansing for 246 00:15:22,860 --> 00:15:26,640 S1: this type of poisoning. All right. Poland's space agency Polska 247 00:15:26,670 --> 00:15:33,510 S1: has taken its systems offline after detecting unauthorized network access. And, uh. 248 00:15:33,510 --> 00:15:37,260 S1: And they're blaming Russian cyber campaigns against the country. Japanese 249 00:15:37,260 --> 00:15:40,530 S1: telecom giant NTT got breached and exposed corporate data for 250 00:15:40,530 --> 00:15:45,800 S1: around 18,000 companies. A's targeting of enemies US is using 251 00:15:45,830 --> 00:15:49,730 S1: AI to scan social media and revoke visas of foreign 252 00:15:49,730 --> 00:15:54,470 S1: students it believes support Hamas or other terrorist groups. The 253 00:15:54,470 --> 00:15:57,380 S1: reason I think this is worrying because this this might 254 00:15:57,380 --> 00:15:59,359 S1: be a good use of it, right? But the problem is, 255 00:15:59,360 --> 00:16:03,620 S1: once you use it once and everyone's like, oh, that's okay. 256 00:16:03,650 --> 00:16:05,930 S1: I mean, there's a saying that says first they come 257 00:16:05,930 --> 00:16:09,200 S1: for so-and-so, then they come for so-and-so, and you say 258 00:16:09,200 --> 00:16:12,170 S1: it's okay, and then they come for so-and-so, and you say, oh, 259 00:16:12,200 --> 00:16:14,810 S1: that's still okay, because it's not me. But eventually they 260 00:16:14,810 --> 00:16:17,479 S1: come for you. And I'm not saying that's going to happen. 261 00:16:17,480 --> 00:16:21,290 S1: I'm just saying whenever you have a thing that's like, hey, 262 00:16:21,290 --> 00:16:25,160 S1: we're looking for people like this because they're bad and 263 00:16:25,160 --> 00:16:28,130 S1: the people are like, well, that's okay. And everyone like 264 00:16:28,160 --> 00:16:33,860 S1: allows that type of Gestapo behavior. It's slippery, very slippery, 265 00:16:33,890 --> 00:16:37,220 S1: and it's way worse when it would have taken a 266 00:16:37,220 --> 00:16:39,560 S1: lot of research in the past to actually go and 267 00:16:39,560 --> 00:16:42,150 S1: find these people, right. You got you got to parse 268 00:16:42,150 --> 00:16:45,300 S1: all these videos. You gotta you gotta, you know, go 269 00:16:45,300 --> 00:16:47,610 S1: through all the blog posts. You got to find people. 270 00:16:47,610 --> 00:16:50,970 S1: You basically have to hire a bunch of researchers, researchers 271 00:16:50,970 --> 00:16:53,250 S1: to do this. Well, what happens if you just spin 272 00:16:53,250 --> 00:16:56,280 S1: up a really powerful AI with a bunch of tools 273 00:16:56,280 --> 00:16:59,010 S1: and say, hey, go parse the internet for the last 274 00:16:59,010 --> 00:17:02,520 S1: two years or during this six month period and find 275 00:17:02,520 --> 00:17:06,900 S1: everybody who in the crowd was waving a Hamas flag, 276 00:17:06,900 --> 00:17:10,710 S1: or in the crowd who was waving whatever flag for 277 00:17:10,710 --> 00:17:12,930 S1: the thing that I currently don't like. And this could 278 00:17:12,930 --> 00:17:14,610 S1: be a left leaning thing. This could be a right 279 00:17:14,640 --> 00:17:18,480 S1: leaning thing. Doesn't matter. You could target an enemy and 280 00:17:18,480 --> 00:17:22,710 S1: basically go after them. That is that is very serious stuff. 281 00:17:22,800 --> 00:17:25,680 S1: And the better the AI gets, the better it's the 282 00:17:25,710 --> 00:17:28,199 S1: faster it's going to be able to do that. I mean, 283 00:17:28,200 --> 00:17:30,629 S1: you could be like, look, hey, we got a Swat 284 00:17:30,660 --> 00:17:33,570 S1: team that's ready to go. Find me the best possible 285 00:17:33,570 --> 00:17:37,619 S1: enemy of me and my friends and find me a 286 00:17:37,619 --> 00:17:41,820 S1: reason in all of their. In the last six months 287 00:17:41,820 --> 00:17:44,010 S1: to go after them and boom, you send the Swat 288 00:17:44,010 --> 00:17:48,960 S1: team like that. That's a very powerful weapon. And you 289 00:17:48,960 --> 00:17:52,560 S1: have to be very careful who is actually wielding that thing. 290 00:17:52,590 --> 00:17:56,879 S1: DOJ charged 12 Chinese nationals for state backed hacking operations, 291 00:17:56,880 --> 00:18:00,149 S1: including two government officers, eight employees of a hacking company 292 00:18:00,150 --> 00:18:05,070 S1: called I-soon and a few members of APT 27. Microsoft 293 00:18:05,100 --> 00:18:08,760 S1: Threat Intelligence says China linked Silk Typhoon hacking group is 294 00:18:08,760 --> 00:18:15,750 S1: shifting from compromise to actually going after it. Supply chains 295 00:18:15,780 --> 00:18:20,550 S1: as an entry point to corporate networks, national security, Russia 296 00:18:20,550 --> 00:18:25,260 S1: and China and probably others are targeting Doge fired US 297 00:18:25,260 --> 00:18:29,160 S1: federal workers with special focus on those with security clearances. 298 00:18:29,160 --> 00:18:31,740 S1: So basically, a bunch of people with clearances who know 299 00:18:31,770 --> 00:18:35,580 S1: things get laid off and they're going after them, especially 300 00:18:35,580 --> 00:18:38,700 S1: if they're angry. Right? That's that's a lot of vulnerability. 301 00:18:38,730 --> 00:18:44,640 S1: China announced another 7.2% increase to its military budget, taking 302 00:18:44,640 --> 00:18:51,240 S1: it to $245 billion a year. The Polish president, Andrzej Duda, 303 00:18:51,240 --> 00:18:56,310 S1: is pushing NATO, NATO members to increase defense spending to 3% 304 00:18:56,310 --> 00:18:59,760 S1: of GDP. Yeah, China is trying to bump it by 305 00:18:59,760 --> 00:19:04,770 S1: another 7.2%, and we're trying to get NATO members to 3%. 306 00:19:04,800 --> 00:19:11,850 S1: Oliver Carroll reports that the US has cut off critical Himars, intelligence, Himars, 307 00:19:11,850 --> 00:19:16,230 S1: Himars intelligence to Ukraine, reducing their ability to defend against 308 00:19:16,230 --> 00:19:21,510 S1: Russian forces. I so Manas broke the internet last week. 309 00:19:21,540 --> 00:19:27,869 S1: Big hype is uh, it's kind of like AGI similar. Um, 310 00:19:27,900 --> 00:19:31,650 S1: they didn't say it was AGI. The video that they 311 00:19:31,650 --> 00:19:34,770 S1: put out was like, hey, it's leaning towards AGI. It's 312 00:19:34,770 --> 00:19:38,160 S1: a little bit like AGI. We're heading in that direction, 313 00:19:38,160 --> 00:19:41,310 S1: but it's basically like Claude or OpenAI, whatever they have 314 00:19:41,340 --> 00:19:44,400 S1: behind the thing, but it's just way smarter. It's doing 315 00:19:44,400 --> 00:19:48,449 S1: a lot more manual work that that wouldn't have needed 316 00:19:48,450 --> 00:19:51,000 S1: to be done by a human in the past. And 317 00:19:51,000 --> 00:19:52,920 S1: it's very much in line with what I was talking 318 00:19:52,920 --> 00:19:55,770 S1: about the other day where, um, I don't think that 319 00:19:55,770 --> 00:19:59,430 S1: we need smarter models, actually, to get to AGI. I 320 00:19:59,430 --> 00:20:04,050 S1: think we just need better orchestration of tooling and more 321 00:20:04,050 --> 00:20:07,050 S1: working memory for the models we already have. And it 322 00:20:07,050 --> 00:20:11,250 S1: seems to me like Manus is just is just that right? 323 00:20:11,280 --> 00:20:15,690 S1: It's probably Claude underneath. And a lot of people have been, uh, 324 00:20:15,690 --> 00:20:19,770 S1: you know, basically hypothesizing that that's the case. We don't 325 00:20:19,770 --> 00:20:23,040 S1: know for sure, but it could just be Claude with 326 00:20:23,040 --> 00:20:26,460 S1: better orchestration. Really want people to think deeply about this 327 00:20:26,460 --> 00:20:31,050 S1: general intelligence competence might be already possible, given how smart 328 00:20:31,050 --> 00:20:36,060 S1: something like O1 Pro or Sonnet 37 already are. it 329 00:20:36,060 --> 00:20:40,350 S1: already has the intelligence to understand stuff like millions or 330 00:20:40,350 --> 00:20:43,050 S1: billions of questions that you could ask it. It's smart 331 00:20:43,050 --> 00:20:46,530 S1: enough to generalize answers, or you can give it situations 332 00:20:46,530 --> 00:20:50,790 S1: like brand new situation and based on everything it knows 333 00:20:50,790 --> 00:20:54,300 S1: about these other things, it can give you the direction 334 00:20:54,300 --> 00:20:57,270 S1: or a solution, you know, for that new problem that 335 00:20:57,270 --> 00:21:01,020 S1: you have. That's generalization, right? There might be smart enough. 336 00:21:01,020 --> 00:21:04,770 S1: The problem is it's eyes and it's hands and it's 337 00:21:04,770 --> 00:21:09,900 S1: getting confused after doing too many things because of context windows. Right. 338 00:21:09,900 --> 00:21:13,950 S1: And just overall coordination and orchestration of the thing that 339 00:21:13,950 --> 00:21:16,680 S1: it's doing to try to accomplish a task. Seems like 340 00:21:16,680 --> 00:21:19,800 S1: Manas did that better, but I think that's all that's 341 00:21:19,830 --> 00:21:24,450 S1: needed is that orchestration. So the point is people are like, well, 342 00:21:24,540 --> 00:21:29,730 S1: AI is not getting smart enough fast enough to actually 343 00:21:29,730 --> 00:21:33,150 S1: move us towards AGI. We don't need that. We don't 344 00:21:33,150 --> 00:21:39,100 S1: need smarter models. We need larger context windows, more tools 345 00:21:39,100 --> 00:21:43,870 S1: for the agents to have. And better orchestration. That by 346 00:21:43,869 --> 00:21:47,860 S1: itself I think gets us there to AGI. And I've 347 00:21:47,859 --> 00:21:50,560 S1: got another post which yeah I'm not going to pull 348 00:21:50,560 --> 00:21:52,480 S1: it up, but I've got another post that basically talks 349 00:21:52,480 --> 00:21:56,230 S1: about exactly what AGI is going to look like under 350 00:21:56,230 --> 00:22:00,880 S1: my definition, which is um, my, my definition is basically 351 00:22:00,910 --> 00:22:03,610 S1: can it do the work of a knowledge worker making 352 00:22:03,640 --> 00:22:08,409 S1: $80,000 in the year 2022? And I say 22 because 353 00:22:08,410 --> 00:22:12,970 S1: that was like pre AI or whatever. So the marketing 354 00:22:12,970 --> 00:22:17,109 S1: for the first AGI is going to be something like 355 00:22:17,350 --> 00:22:22,420 S1: welcome to new hire. New hire will join your team 356 00:22:22,420 --> 00:22:25,600 S1: on Monday morning. The day that they start they will 357 00:22:25,600 --> 00:22:29,290 S1: do the onboarding. They will read all the slack messages. 358 00:22:29,290 --> 00:22:31,780 S1: They will read all the documentation. They will meet with 359 00:22:31,780 --> 00:22:34,690 S1: their manager. They will take direction. They will reach out 360 00:22:34,690 --> 00:22:37,840 S1: to the team members and introduce themselves, and they will 361 00:22:37,840 --> 00:22:40,719 S1: start to do the work that they are assigned. And 362 00:22:40,720 --> 00:22:44,409 S1: most importantly, if that work changes because of a reorg 363 00:22:44,410 --> 00:22:46,930 S1: or because they bought some new company or whatever, they 364 00:22:46,930 --> 00:22:50,470 S1: get new tasks, new goals, new work they have to do. 365 00:22:50,590 --> 00:22:53,859 S1: They will discard their previous stuff, they will learn the 366 00:22:53,859 --> 00:22:56,440 S1: new stuff, they will read the new docs, they will 367 00:22:56,440 --> 00:22:58,899 S1: take new direction. They will learn what they have to, 368 00:22:58,930 --> 00:23:01,030 S1: and they'll start doing that work. And if their manager 369 00:23:01,060 --> 00:23:03,070 S1: is like, hey, you're doing pretty good here, but you 370 00:23:03,070 --> 00:23:05,680 S1: need to improve on this just like a human. They'll 371 00:23:05,680 --> 00:23:09,189 S1: be like, okay, thanks, boss. I'll make that adjustment. I'll 372 00:23:09,190 --> 00:23:12,490 S1: make sure you get a status report every day and 373 00:23:12,790 --> 00:23:16,000 S1: whatever it is they were asked for this company, new hire, 374 00:23:16,000 --> 00:23:17,980 S1: which is not a real company, by the way, but 375 00:23:17,980 --> 00:23:20,650 S1: it's going to be something exactly like this. That's how 376 00:23:20,650 --> 00:23:24,460 S1: they're going to market it. Treat it exactly like a 377 00:23:24,460 --> 00:23:28,960 S1: regular employee. That is going to be the moment, in 378 00:23:28,960 --> 00:23:33,129 S1: my opinion, that we've reached AGI, because if that tool, 379 00:23:33,160 --> 00:23:34,419 S1: by the way, there's going to be a bunch of 380 00:23:34,420 --> 00:23:37,210 S1: tools that say they can do this, and then people 381 00:23:37,240 --> 00:23:39,040 S1: are going to implement them and they're going to be 382 00:23:39,040 --> 00:23:41,859 S1: super stupid. They're going to be more waste than it's worth. 383 00:23:41,859 --> 00:23:45,190 S1: It's just not going to work. But when this happens 384 00:23:45,190 --> 00:23:47,950 S1: and let's say it's this company called New Hire, um, 385 00:23:47,980 --> 00:23:51,580 S1: when that happens, let's say new hire takes off and 386 00:23:51,609 --> 00:23:54,369 S1: it just it goes slow for a second and all 387 00:23:54,369 --> 00:23:57,490 S1: of a sudden everyone starts saying, oh my God, I 388 00:23:57,490 --> 00:24:02,020 S1: employed ten new hires, and they're doing the work of 389 00:24:02,050 --> 00:24:07,150 S1: like 50 regular employees. And I just increased my subscription 390 00:24:07,150 --> 00:24:10,990 S1: because I'm growing my business faster than ever. I employed 391 00:24:10,990 --> 00:24:13,600 S1: my new hires in marketing. I employed them in sales, 392 00:24:13,600 --> 00:24:17,050 S1: I employed them in administration. I employed them in security 393 00:24:17,050 --> 00:24:22,030 S1: administration in my SOC. Whatever the moment that happens, I'm 394 00:24:22,030 --> 00:24:25,120 S1: calling that the AGI moment. I believe that is the 395 00:24:25,119 --> 00:24:30,189 S1: actual AGI moment, because there is no better definition of 396 00:24:30,190 --> 00:24:33,520 S1: general intelligence than humans. That's what we're basing this whole 397 00:24:33,520 --> 00:24:38,260 S1: entire thing on, is humans ability to generalize, generalize when 398 00:24:38,260 --> 00:24:43,300 S1: it's working. Right. The fact that you could change up context, say, hey, 399 00:24:43,300 --> 00:24:46,510 S1: that's not your job anymore. Here's your new job. That's 400 00:24:46,510 --> 00:24:50,200 S1: the best definition we have is a knowledge worker constantly 401 00:24:50,200 --> 00:24:54,609 S1: dealing with change, constantly dealing with one offs and asks 402 00:24:54,609 --> 00:24:58,630 S1: and adjustments. So that's the perfect definition also for actual 403 00:24:58,660 --> 00:25:03,159 S1: AGI artificial general intelligence. And if it could do the 404 00:25:03,160 --> 00:25:07,090 S1: job of an average knowledge worker I would say that 405 00:25:07,119 --> 00:25:09,639 S1: hits the bar. So that's why I'm using it for 406 00:25:09,670 --> 00:25:12,580 S1: a definition. And I'm saying that when a product comes 407 00:25:12,580 --> 00:25:15,160 S1: out that can do this and it's actually doing it, 408 00:25:15,160 --> 00:25:19,030 S1: that will be the moment. And my prediction is late 409 00:25:19,060 --> 00:25:24,280 S1: 2025 at the earliest. This is my modified prediction. My 410 00:25:24,280 --> 00:25:29,859 S1: overall prediction, which I said in 23 was 25 to 28. 411 00:25:29,869 --> 00:25:35,570 S1: so my modified prediction is late 25 to early 27. 412 00:25:35,900 --> 00:25:38,300 S1: So it's still right there inside that window. But I 413 00:25:38,300 --> 00:25:43,340 S1: think it's getting pretty close. And things like Manus are 414 00:25:43,340 --> 00:25:46,160 S1: starting to show signs of it. All right. New model 415 00:25:46,190 --> 00:25:51,710 S1: q w q 32 billion parameters. This thing is outperforming 416 00:25:51,710 --> 00:25:57,050 S1: or performing just as good as 671 billion parameter deep 417 00:25:57,050 --> 00:26:02,359 S1: seek 32 billion versus 671 billion. And they are right 418 00:26:02,390 --> 00:26:07,340 S1: on par. This is the absolute most impressive local model 419 00:26:07,340 --> 00:26:10,520 S1: I've ever used. I test out all these big ones 420 00:26:10,580 --> 00:26:13,700 S1: when they come out. I've got right in my other room. 421 00:26:13,700 --> 00:26:16,400 S1: I've got two 49 seconds over there and I fire 422 00:26:16,430 --> 00:26:20,149 S1: them up and I test this thing. This I mean, 423 00:26:20,150 --> 00:26:27,140 S1: it performs honestly very similar to like, a clod or 424 00:26:27,140 --> 00:26:31,070 S1: an OpenAI. um, like an oh one or an oh 425 00:26:31,070 --> 00:26:35,330 S1: one mini. It's really, really good. Um, I used it 426 00:26:35,330 --> 00:26:38,120 S1: with some, uh, story writing stuff. I'm trying to write 427 00:26:38,150 --> 00:26:41,929 S1: a fiction story about some, like, an AI future. Uh, 428 00:26:41,930 --> 00:26:44,030 S1: I'm getting pretty excited about that, but I was using 429 00:26:44,030 --> 00:26:46,970 S1: it to help me write, and it's doing things that 430 00:26:46,970 --> 00:26:52,070 S1: no previous model could possibly do. Um, just vast jumps 431 00:26:52,070 --> 00:26:56,000 S1: in quality in the output that I'm getting. So really 432 00:26:56,000 --> 00:26:58,100 S1: excited about that. You should go check it out. You 433 00:26:58,100 --> 00:27:01,100 S1: could run it with a llama llama run q w q. 434 00:27:01,130 --> 00:27:03,650 S1: That's all you have to do. It'll download it, it'll 435 00:27:03,650 --> 00:27:05,840 S1: run it, and you could test it out. And it's 436 00:27:05,840 --> 00:27:09,680 S1: a thinking model, uh, built with reinforcement learning, which means 437 00:27:09,680 --> 00:27:12,080 S1: you're going to see the thinking instructions as part of 438 00:27:12,080 --> 00:27:15,380 S1: the output before it starts outputting the real thing. Myo 439 00:27:15,410 --> 00:27:20,540 S1: addresses AI hallucinations, so doctors are very upset and worried 440 00:27:20,540 --> 00:27:23,660 S1: about hallucinations. For good reason, because it's life and death 441 00:27:23,660 --> 00:27:26,660 S1: for them, right? So reverse rag is a technique that 442 00:27:26,660 --> 00:27:31,790 S1: they're using to basically figure out the sources for the 443 00:27:31,790 --> 00:27:36,350 S1: data that they return. So they don't just take hallucinations, 444 00:27:36,350 --> 00:27:38,630 S1: or they don't just take outputs and assume it's good. 445 00:27:38,630 --> 00:27:41,180 S1: What they do is they take each particular one and 446 00:27:41,180 --> 00:27:44,210 S1: they go and research it and find sources to it. 447 00:27:44,210 --> 00:27:47,660 S1: And then they say, okay, this is now trustworthy. And 448 00:27:47,660 --> 00:27:49,670 S1: I've been talking about this for, I don't know, six 449 00:27:49,670 --> 00:27:52,850 S1: months or now or however long. I did a talk 450 00:27:52,880 --> 00:27:56,750 S1: about this at some conference. And basically what we're talking 451 00:27:56,750 --> 00:28:00,590 S1: about here is a pipeline of trust that has to 452 00:28:00,590 --> 00:28:05,720 S1: happen for high trust output. So if you're in medical 453 00:28:05,720 --> 00:28:10,429 S1: or you're in military or you're in, uh, law or something, 454 00:28:10,430 --> 00:28:14,689 S1: if you're in something where results actually matter, you have 455 00:28:14,690 --> 00:28:18,080 S1: to have a system that comes after the AI output, 456 00:28:18,109 --> 00:28:21,170 S1: which is like a series of steps for validation, and 457 00:28:21,170 --> 00:28:23,510 S1: only after it goes through those steps and gets a 458 00:28:23,510 --> 00:28:26,720 S1: green check mark, then you can trust the thing, right? 459 00:28:26,750 --> 00:28:29,240 S1: I mean, humans need that, by the way. I mean, 460 00:28:29,660 --> 00:28:33,530 S1: humans don't put out stuff that's trustworthy, right? So I 461 00:28:33,530 --> 00:28:35,929 S1: think AI is going to be actually much better at 462 00:28:35,930 --> 00:28:40,310 S1: doing this than, uh, most human doctors. Um, you still 463 00:28:40,310 --> 00:28:42,740 S1: don't get that. Like the doctor touch and the doctor 464 00:28:42,740 --> 00:28:46,280 S1: intuition and the the human aspect. So it's it's not 465 00:28:46,280 --> 00:28:48,740 S1: better in all ways, but in terms of, like, putting 466 00:28:48,740 --> 00:28:52,490 S1: out the best possible diagnosis and stuff like that, I 467 00:28:52,490 --> 00:28:55,880 S1: think it's going to be fantastic. So kudos to Mayo 468 00:28:55,880 --> 00:28:59,209 S1: for doing that. Tyler Cowen told Dwarkesh Patel that he 469 00:28:59,210 --> 00:29:04,460 S1: predicts AI will only boost economic growth by 0.5% per year, 470 00:29:04,460 --> 00:29:08,690 S1: half a percent. I massively disagree with this and so 471 00:29:08,690 --> 00:29:13,370 S1: does Dwarkesh. Black Forest Labs, also known as replicate, is 472 00:29:13,370 --> 00:29:18,590 S1: dominating image generation. I'm using this myself. Um, I've been 473 00:29:18,590 --> 00:29:22,670 S1: using it in conjunction more with, uh, Midjourney. Midjourney is 474 00:29:22,670 --> 00:29:24,680 S1: like my go to, but I've been using a lot 475 00:29:24,680 --> 00:29:28,500 S1: more replicate and it is absolutely fantastic. Larry Page is 476 00:29:28,500 --> 00:29:31,620 S1: working on a new AI company that uses artificial intelligence 477 00:29:31,620 --> 00:29:36,480 S1: to design and manufacture optimized products. Ben Buchanan, Biden's AI advisor, 478 00:29:36,480 --> 00:29:41,130 S1: says Washington now believes AGI is possible and realizes they're 479 00:29:41,130 --> 00:29:44,610 S1: competing with tech companies for talent to regulate it. Nirvana 480 00:29:44,610 --> 00:29:48,960 S1: raised $80 million for the AI based trucking insurance platform 481 00:29:48,960 --> 00:29:53,250 S1: that uses telematics and 20 billion miles of driving data 482 00:29:53,250 --> 00:29:56,460 S1: to create better policies for drivers. And I've got a 483 00:29:56,460 --> 00:30:01,170 S1: post here on real time insurance premiums from 2016. Yeah, 484 00:30:01,200 --> 00:30:05,490 S1: I've been talking about this for a while. 2016 essentially, again, 485 00:30:05,490 --> 00:30:08,910 S1: everything with AI comes down to context. So if I'm 486 00:30:08,910 --> 00:30:12,570 S1: wearing a microphone and a camera and I give access 487 00:30:12,570 --> 00:30:15,180 S1: to that, there's like that progressive thing. Remember that progressive 488 00:30:15,210 --> 00:30:17,010 S1: box you put in your car and it can see 489 00:30:17,010 --> 00:30:19,680 S1: how you're driving? Well, imagine that it could see that 490 00:30:19,680 --> 00:30:25,080 S1: you're cussing or yelling at your spouse or kicking garbage 491 00:30:25,110 --> 00:30:30,180 S1: cans or, um, you just came from the bar, and 492 00:30:30,180 --> 00:30:33,780 S1: it knows that because it has the cameras on and monitoring. Well, 493 00:30:33,780 --> 00:30:38,340 S1: guess what? Your insurance premiums might be getting modified in 494 00:30:38,340 --> 00:30:41,070 S1: real time or close to real time. Or you might 495 00:30:41,070 --> 00:30:44,400 S1: be like, hey, if you get in this car right now, 496 00:30:44,430 --> 00:30:49,230 S1: having come from, you know, o'shanaghgan, then, um, you your 497 00:30:49,230 --> 00:30:52,080 S1: policy will be instantly cancelled, right? It's going to be 498 00:30:52,080 --> 00:30:56,700 S1: things like that. Um, all possible because of AI and context. 499 00:30:56,910 --> 00:31:01,740 S1: Simon Willison taught a Nykaa 2025 workshop on advanced web 500 00:31:01,770 --> 00:31:04,890 S1: scraping techniques. Really, really cool talk. You ought to check 501 00:31:04,890 --> 00:31:07,590 S1: that one out. Okay, McDonald's is rolling out a massive 502 00:31:07,590 --> 00:31:11,730 S1: AI upgrade for its 43,000 locations, including AI drive through 503 00:31:11,760 --> 00:31:15,930 S1: smart kitchen equipment and generative AI Virtual Manager, one of 504 00:31:15,930 --> 00:31:19,890 S1: the biggest visible AI rollouts so far. Can't wait to 505 00:31:19,890 --> 00:31:24,100 S1: see how this works. Um, I heard somebody say, I 506 00:31:24,100 --> 00:31:26,170 S1: don't know who because I don't. I wouldn't eat there 507 00:31:26,170 --> 00:31:29,230 S1: because it's really bad. But Domino's and Carl's Jr has 508 00:31:29,230 --> 00:31:33,790 S1: really good eye. So I heard technology. Apple is pushing back. 509 00:31:33,790 --> 00:31:38,890 S1: It's more personalized Siri features. So basically the big unified 510 00:31:38,890 --> 00:31:42,100 S1: Siri that has access to all the context is being 511 00:31:42,100 --> 00:31:45,760 S1: pushed back. And a lot of people are speculating that 512 00:31:45,760 --> 00:31:50,470 S1: this is because of security issues, uh, prompt injection against 513 00:31:50,470 --> 00:31:53,290 S1: that entire context, which what do you think is going 514 00:31:53,320 --> 00:31:57,580 S1: to happen when your assistant that you call up with 515 00:31:57,580 --> 00:32:02,110 S1: this keyword can do anything across your entire context, including 516 00:32:02,110 --> 00:32:04,990 S1: all of your applications, all your health data and everything 517 00:32:04,990 --> 00:32:09,190 S1: and your interface into that is is the English language 518 00:32:09,190 --> 00:32:12,580 S1: and other languages as well, but especially English. That is 519 00:32:12,580 --> 00:32:16,840 S1: a massive attack surface to go after. And I think Apple, 520 00:32:16,840 --> 00:32:19,690 S1: rightly so, is being very careful about that. So they 521 00:32:19,690 --> 00:32:22,320 S1: pushed that back. One thing to keep in mind, though, 522 00:32:22,350 --> 00:32:25,470 S1: is they already rolled out tons of Apple intelligence. I 523 00:32:25,470 --> 00:32:32,580 S1: already use my Siri to actually, um, do live queries 524 00:32:32,580 --> 00:32:36,900 S1: because it's connected to ChatGPT. So don't think that we're 525 00:32:36,900 --> 00:32:39,240 S1: waiting for this thing to happen. And you don't have 526 00:32:39,240 --> 00:32:42,450 S1: Apple intelligence before that. It's already out there. Like they 527 00:32:42,450 --> 00:32:46,110 S1: already rolled out like dozens of these features, including, most importantly, 528 00:32:46,110 --> 00:32:49,200 S1: the ChatGPT one. And the fact that if you hold 529 00:32:49,200 --> 00:32:52,590 S1: down your camera button, the new camera button on the side, 530 00:32:52,590 --> 00:32:56,490 S1: you can have ChatGPT identify things with the camera. So 531 00:32:56,610 --> 00:32:59,850 S1: all this to say, Apple intelligence is like already live 532 00:32:59,850 --> 00:33:03,719 S1: and already working. It's just this unified big context, one 533 00:33:03,720 --> 00:33:07,650 S1: that they haven't enabled yet because of the security concern. 534 00:33:07,680 --> 00:33:11,070 S1: Waymo is expanding its self-driving car service in Palo Alto 535 00:33:11,070 --> 00:33:15,060 S1: and some surrounding peninsula areas. Cannot wait to get it 536 00:33:15,060 --> 00:33:21,180 S1: around here. Waymo is just just unbelievably good. New Zealand 537 00:33:21,210 --> 00:33:24,480 S1: appears to be managing its $16 billion health budget. Using 538 00:33:24,510 --> 00:33:28,260 S1: a single Excel spreadsheet probably makes Doge pretty happy. Apple 539 00:33:28,260 --> 00:33:32,760 S1: is evidently planning a major redesign of iOS, iPadOS, and 540 00:33:32,760 --> 00:33:35,130 S1: macOS for the fall. Can't wait to see what that 541 00:33:35,130 --> 00:33:37,200 S1: looks like. I'll be running it on day one because 542 00:33:37,200 --> 00:33:41,280 S1: I'm stupid. Tesla's global sales are tanking pretty hard right now, 543 00:33:41,280 --> 00:33:45,780 S1: with significant drops across Germany, Australia and China. Yeah, you 544 00:33:45,780 --> 00:33:49,860 S1: do a Hitler salute and sales fall in Germany. Big surprise. 545 00:33:49,860 --> 00:33:55,080 S1: New Mac studio has 32. This is the M3 ultra 546 00:33:55,110 --> 00:34:01,080 S1: 32 CPU cores and enough unified memory, 512GB to load 547 00:34:01,080 --> 00:34:05,970 S1: the 600 billion parameter models. I was going to get one, 548 00:34:05,970 --> 00:34:08,610 S1: but I already have my eye box. I'm going to 549 00:34:08,610 --> 00:34:11,609 S1: skip this generation, maybe even another one because I already 550 00:34:11,610 --> 00:34:14,910 S1: have an M2 ultra. Not worth it to bump up 551 00:34:14,910 --> 00:34:20,339 S1: to the 512 especially. That's like a $15,000 Desktop. All right. Humans, 552 00:34:20,340 --> 00:34:25,950 S1: multiple indicators, including nowcast models and the inverted yield curve, 553 00:34:25,980 --> 00:34:30,450 S1: suggests we might face higher than expected recession risk going 554 00:34:30,480 --> 00:34:34,620 S1: into 2025. Brad Sigmund is set to become the first 555 00:34:34,620 --> 00:34:38,430 S1: person executed by firing squad, unless the governor or the 556 00:34:38,670 --> 00:34:43,350 S1: Supreme Court gets involved. Tiago Forte says traditional goal setting 557 00:34:43,380 --> 00:34:48,930 S1: may be killing. Our creativity suggests optimizing for novelty and interestingness. 558 00:34:48,960 --> 00:34:56,520 S1: I like that interestingness. Apple's hit dystopian show severance is fantastic. 559 00:34:56,550 --> 00:35:00,000 S1: You have to check it out. And having worked in 560 00:35:00,000 --> 00:35:03,330 S1: a bunch of big companies, especially Apple and management, a 561 00:35:03,330 --> 00:35:07,830 S1: lot of the stuff is very familiar. Very familiar. Not 562 00:35:07,830 --> 00:35:11,460 S1: in a good way. NASA's new sphere X Observatory is 563 00:35:11,460 --> 00:35:15,930 S1: heading to space to map over 450 million galaxies. Stanford 564 00:35:15,930 --> 00:35:21,189 S1: researchers have developed an antibody duo therapy that neutralizes all 565 00:35:21,190 --> 00:35:27,850 S1: SARS-CoV-2 variants by targeting two different parts of the virus simultaneously. 566 00:35:27,880 --> 00:35:30,399 S1: Can't wait to see if that comes to market research. 567 00:35:30,700 --> 00:35:34,239 S1: Of course, no one will take it because it's a vaccine. Probably. 568 00:35:34,239 --> 00:35:36,670 S1: I assume it's a vaccine, or maybe it's a after 569 00:35:36,700 --> 00:35:40,210 S1: after the fact treatment, I don't know. Researchers suggest autism 570 00:35:40,210 --> 00:35:46,750 S1: and ADHD frequently coexist as ADHD. ADHD. Yeah, up to 571 00:35:46,780 --> 00:35:51,610 S1: half of autistic people exhibit ADHD symptoms, and two thirds 572 00:35:51,610 --> 00:35:57,219 S1: of people with ADHD show autism characteristics. Scientists at Mass, 573 00:35:57,219 --> 00:36:00,759 S1: Eye and Ear have used patient's own stem cells to 574 00:36:00,790 --> 00:36:06,010 S1: repair cornea damage, previously thought to be permanently blinding. Physicists 575 00:36:06,010 --> 00:36:09,580 S1: have discovered a third limitation on our ability to predict 576 00:36:09,580 --> 00:36:14,410 S1: the universe. So in addition to quantum uncertainty and chaotic 577 00:36:14,410 --> 00:36:19,629 S1: butterfly effects, you've got this thing called undecidability, and they're 578 00:36:19,630 --> 00:36:26,170 S1: claiming that because of undecidability, some systems are fundamentally unknowable. 579 00:36:26,170 --> 00:36:28,270 S1: I would have thought you would have got that already 580 00:36:28,270 --> 00:36:32,980 S1: from quantum indeterminacy. But, uh, yeah, it was an interesting article. 581 00:36:33,010 --> 00:36:35,230 S1: Price of coffee has gone up so much that millions 582 00:36:35,230 --> 00:36:37,989 S1: of businesses in its supply chain are running out of 583 00:36:37,989 --> 00:36:41,680 S1: money to buy it. I was thinking, isn't the canary 584 00:36:41,680 --> 00:36:44,529 S1: in the coal mine here? Isn't it? Starbucks and companies 585 00:36:44,530 --> 00:36:46,870 S1: like that who have to buy at scale? Or do 586 00:36:46,870 --> 00:36:50,020 S1: they have their own supply chains? Uh, yeah. It'll be 587 00:36:50,020 --> 00:36:53,890 S1: interesting to watch this. And what happens there? Discovery a 588 00:36:54,160 --> 00:36:57,370 S1: machine learning engineer named Jasper Gillie quit his job because 589 00:36:57,370 --> 00:36:59,980 S1: he believes AI will automate his entire role by the 590 00:36:59,980 --> 00:37:04,029 S1: end of 2025. Keep in mind, this is an extremely 591 00:37:04,030 --> 00:37:10,030 S1: highly paid machine learning engineer, and he's saying his job 592 00:37:10,060 --> 00:37:12,910 S1: is going to be replaced by the end of 2025. 593 00:37:12,940 --> 00:37:16,600 S1: Robin Moffatt argues we should or we shouldn't just write 594 00:37:16,600 --> 00:37:19,600 S1: the perfectly structured blog post, but also share our messy 595 00:37:19,600 --> 00:37:23,410 S1: struggles and solutions that other people need too. I like this. 596 00:37:23,440 --> 00:37:26,169 S1: It's like learning in public. Show your work, you know. 597 00:37:26,200 --> 00:37:31,239 S1: Feel free to be vulnerable and imperfect online. I think 598 00:37:31,270 --> 00:37:34,390 S1: it's a it's endearing. Hopefully so since I've said edit 599 00:37:34,390 --> 00:37:37,600 S1: like nine times in this recording already. Small docs a 600 00:37:37,600 --> 00:37:41,350 S1: project around writing small docs instead of giant ones that 601 00:37:41,350 --> 00:37:45,430 S1: are hard to follow. Jamie Rumbelow created a slick hybrid 602 00:37:45,460 --> 00:37:50,020 S1: between databases, programming and UIs. Make scripts more user friendly 603 00:37:50,020 --> 00:37:52,660 S1: without needing a full web app. And basically the pitch 604 00:37:52,660 --> 00:37:57,100 S1: there was that the database, the UI and the app 605 00:37:57,100 --> 00:38:00,640 S1: are kind of unified. They're all the same thing. Astro 606 00:38:00,640 --> 00:38:03,130 S1: and vim, one of the better vim distros. I do 607 00:38:03,160 --> 00:38:06,160 S1: prefer lazy vim, although I hate the name. Reminds me 608 00:38:06,160 --> 00:38:09,280 S1: of the dummies books which I could never get myself 609 00:38:09,280 --> 00:38:13,600 S1: to buy. Too much self respect interview DB lets you 610 00:38:13,600 --> 00:38:18,730 S1: access Crowdsource tech interview questions from the community. Alexander Solzhenitsyn 611 00:38:18,760 --> 00:38:22,210 S1: on the future of the West. This was a crazy, 612 00:38:22,210 --> 00:38:27,640 S1: crazy piece of text there. Seneschal MCP. MCP. By a 613 00:38:27,640 --> 00:38:33,430 S1: brilliant Ull member named Matt, who, um, wrote an MCP 614 00:38:33,460 --> 00:38:38,920 S1: that helps Llms access standardized health data from the snarky, 615 00:38:39,400 --> 00:38:43,780 S1: cynical seneschal. I keep thinking seneschal. What am I thinking of? 616 00:38:43,810 --> 00:38:48,339 S1: Seneschal API? So basically this is an MCP that makes 617 00:38:48,340 --> 00:38:50,830 S1: your health data available so that you can have local 618 00:38:50,830 --> 00:38:54,040 S1: agents or any kind of agent really make your health 619 00:38:54,040 --> 00:38:58,180 S1: data available to an app. So if you're building like 620 00:38:58,210 --> 00:39:02,469 S1: a full life encompassing, you know, management app with like 621 00:39:02,500 --> 00:39:08,410 S1: workout routines, health, fitness, food, diet, money, all that stuff, 622 00:39:08,410 --> 00:39:12,880 S1: this is an API. This is an MCP that presents 623 00:39:12,890 --> 00:39:16,370 S1: the health side of it. Very very cool. From UL 624 00:39:16,400 --> 00:39:21,350 S1: member Matt Self-interview. I tool that runs invisibly during technical interviews, 625 00:39:21,350 --> 00:39:24,080 S1: helping you with coding questions while your screen is shared. 626 00:39:24,080 --> 00:39:27,109 S1: Also known as cheating. So this is what you have 627 00:39:27,110 --> 00:39:30,410 S1: to watch out for if you're doing interviews. Piano FM 628 00:39:30,410 --> 00:39:34,970 S1: generates endless beautiful AI piano music. I listened for a while. 629 00:39:34,969 --> 00:39:39,410 S1: It was not repetitive and it sounded. I'm not a 630 00:39:39,410 --> 00:39:44,540 S1: classical music expert or a piano expert and it sounded 631 00:39:44,540 --> 00:39:48,470 S1: pretty normal to me. It sounded like regular piano, whatever 632 00:39:48,500 --> 00:39:51,830 S1: that's worth. Matrix hacking game. Someone built a matrix themed 633 00:39:51,860 --> 00:39:55,490 S1: AI hacking game lets users learn about prompt engineering and 634 00:39:55,489 --> 00:40:00,710 S1: security by defeating an Agent Smith AI character. Send new 635 00:40:00,739 --> 00:40:04,790 S1: app connects your iPhone camera to multimodal llms for real 636 00:40:04,820 --> 00:40:09,020 S1: time visual intelligence. Made with zero VC money and cursor. 637 00:40:09,650 --> 00:40:11,210 S1: This is the type of thing that could have gone 638 00:40:11,239 --> 00:40:16,250 S1: for could have raised, you know, $20 million and been 639 00:40:16,250 --> 00:40:21,020 S1: sold for $100 million maybe a couple of years ago. 640 00:40:21,020 --> 00:40:24,140 S1: And this person just built it with cursor with no 641 00:40:24,170 --> 00:40:27,440 S1: money whatsoever. Since feedback Dev built an AI that reads 642 00:40:27,440 --> 00:40:30,080 S1: the emotional vibe of your product, social media mentions and 643 00:40:30,080 --> 00:40:34,370 S1: gives you instant feedback. And somebody posted their cursor rules, 644 00:40:34,370 --> 00:40:37,130 S1: which I thought were quite good. All right, member essay 645 00:40:37,160 --> 00:40:41,780 S1: is intelligence just interest? What if people being smart or 646 00:40:41,780 --> 00:40:46,069 S1: brilliant about something really just comes down to being obsessed 647 00:40:46,070 --> 00:40:48,500 S1: with it? When you think about most people I know 648 00:40:48,530 --> 00:40:51,680 S1: who are so-called geniuses about a given topic, it tends 649 00:40:51,680 --> 00:40:56,150 S1: to overlap 100% with people who won't shut up about 650 00:40:56,150 --> 00:40:59,509 S1: that topic. Kind of a weird coincidence, right? I think 651 00:40:59,510 --> 00:41:02,629 S1: back to training people, like in cybersecurity when I was 652 00:41:02,630 --> 00:41:05,270 S1: back at HP, this is where I was doing it 653 00:41:05,270 --> 00:41:07,669 S1: the most. And it turned out that people who ended 654 00:41:07,670 --> 00:41:10,070 S1: up thriving and being like the best testers and the 655 00:41:10,070 --> 00:41:14,150 S1: best hackers were the ones who simply loved it and 656 00:41:14,180 --> 00:41:17,300 S1: were curious about it and would not stop talking about it. 657 00:41:17,300 --> 00:41:19,489 S1: And that made them want to study more. It made 658 00:41:19,520 --> 00:41:21,860 S1: them play with more tools. It made them like interact 659 00:41:21,860 --> 00:41:27,020 S1: with everything, um, far more than the other people who 660 00:41:27,020 --> 00:41:29,989 S1: just like, didn't care and were doing it for work. 661 00:41:30,020 --> 00:41:34,970 S1: So the question is, is curiosity actually the superpower? I 662 00:41:34,969 --> 00:41:37,489 S1: think this is a powerful frame because it changes the 663 00:41:37,489 --> 00:41:41,720 S1: narrative from you're not smart enough. Instead, to the question 664 00:41:41,750 --> 00:41:45,020 S1: of what are you interested in or what can you 665 00:41:45,050 --> 00:41:47,750 S1: not shut up about? That's what you should be doing 666 00:41:47,750 --> 00:41:51,440 S1: and building your career around. For human 3.0. If we 667 00:41:51,440 --> 00:41:53,750 S1: think about it this way, the main driver for success 668 00:41:53,750 --> 00:41:56,840 S1: isn't how smart you are at solving particular problems, but 669 00:41:56,840 --> 00:42:00,410 S1: how many problems you put in front of yourself. Curious 670 00:42:00,410 --> 00:42:04,850 S1: people create a vast network of like paths of discovery. 671 00:42:04,850 --> 00:42:08,000 S1: Reading this book and that book, and then it mentions 672 00:42:08,000 --> 00:42:10,489 S1: another book. So you go read that. So you're like 673 00:42:10,489 --> 00:42:13,670 S1: branching out. You're creating all these paths and it shows 674 00:42:13,670 --> 00:42:17,780 S1: them more things to get excited about. And that produces 675 00:42:17,780 --> 00:42:20,299 S1: more curiosity. And pretty soon you have like these vast 676 00:42:20,300 --> 00:42:25,430 S1: networks of like paths and, and, you know, subgenres and everything. 677 00:42:25,460 --> 00:42:29,810 S1: People who aren't curious kind of stay in one place 678 00:42:29,810 --> 00:42:33,799 S1: and don't really expose themselves to many kinds of problems. 679 00:42:33,800 --> 00:42:36,830 S1: So if the question is how much you learn and accomplish, 680 00:42:36,830 --> 00:42:39,950 S1: which I think it should be, then expanding your paths 681 00:42:39,950 --> 00:42:44,060 S1: and exposing yourself to more problems is far higher value 682 00:42:44,090 --> 00:42:46,850 S1: than your speed or ability to solve the problems that 683 00:42:46,850 --> 00:42:50,720 S1: you find. So for me, the takeaway here is massive 684 00:42:50,719 --> 00:42:56,509 S1: maximize curiosity, maximize exposure, maximize your inputs with the high 685 00:42:56,510 --> 00:43:01,400 S1: quality thoughts and ideas, mostly reading. And don't judge people 686 00:43:01,400 --> 00:43:03,799 S1: as not smart until you can answer whether or not 687 00:43:03,800 --> 00:43:08,359 S1: they're curious. And for those who aren't curious, isn't that 688 00:43:08,360 --> 00:43:12,980 S1: likely because they weren't exposed to cool enough stuff as 689 00:43:12,980 --> 00:43:16,310 S1: a kid? And why can't we start exposing them now? Right, 690 00:43:16,310 --> 00:43:19,069 S1: so you don't know if somebody is like a closet 691 00:43:19,070 --> 00:43:25,400 S1: genius because they are closet curious, right? So let's start 692 00:43:25,430 --> 00:43:28,310 S1: now with that. Let's bombard them with things that they 693 00:43:28,310 --> 00:43:31,730 S1: might be curious about, and it might unlock them. It 694 00:43:31,730 --> 00:43:33,620 S1: might turn them from like a C player or a 695 00:43:33,650 --> 00:43:37,280 S1: B player into an absolute a player, because you didn't 696 00:43:37,280 --> 00:43:39,799 S1: know they were so curious, because they never got exposed 697 00:43:39,800 --> 00:43:42,440 S1: by their parents or by their teachers, or by they 698 00:43:42,440 --> 00:43:45,739 S1: didn't have any mentors that were like intellectually focused. So 699 00:43:45,770 --> 00:43:48,500 S1: execute the same routine you would for a curious person 700 00:43:48,500 --> 00:43:51,950 S1: to stay curious and read everything, watch all the YouTube 701 00:43:51,950 --> 00:43:57,380 S1: channels around interesting topics, whatever. Basically exposure. When I stop 702 00:43:57,380 --> 00:44:01,880 S1: consuming great books, I get dumb. Within weeks it happens 703 00:44:01,880 --> 00:44:06,259 S1: like almost instantly. Maybe that's because it neuters my curiosity 704 00:44:06,260 --> 00:44:09,030 S1: and interest in the world. In fact, I'm sure that's 705 00:44:09,030 --> 00:44:11,580 S1: the reason. So this might be the single best way 706 00:44:11,580 --> 00:44:17,520 S1: to magnify someone's functional intelligence, curiosity over IQ and the 707 00:44:17,520 --> 00:44:22,110 S1: recommendation of the week. If curiosity is more important than intelligence, 708 00:44:22,110 --> 00:44:25,740 S1: the best thing you could do is read more good books. 709 00:44:25,770 --> 00:44:29,339 S1: Nothing is a substitute for books, not movies. Not even 710 00:44:29,340 --> 00:44:34,080 S1: good YouTube content about nonfiction topics. Nothing is as high 711 00:44:34,080 --> 00:44:39,540 S1: quality and high value as books. My hard recommendation to 712 00:44:39,570 --> 00:44:44,339 S1: everyone listening at least two high quality books per month. 713 00:44:44,370 --> 00:44:48,029 S1: At least one of those should be nonfiction. And if 714 00:44:48,030 --> 00:44:50,160 S1: it if you have fiction as the other one, it 715 00:44:50,160 --> 00:44:52,980 S1: should be high quality fiction, which really talks about a 716 00:44:52,980 --> 00:44:57,660 S1: whole bunch of nonfiction topics. And it's fine if you 717 00:44:57,660 --> 00:45:02,339 S1: want to add additional garbage fiction or just purely fun fiction. 718 00:45:02,340 --> 00:45:04,710 S1: I would say go ahead, but do those on top 719 00:45:04,710 --> 00:45:09,090 S1: of these two books, no exceptions, two books. High quality. 720 00:45:09,239 --> 00:45:13,230 S1: Four would be nice, but it's a little bit harder. Uh, 721 00:45:13,230 --> 00:45:15,780 S1: do at least two. And I got a few here 722 00:45:15,780 --> 00:45:20,910 S1: to get you started. The Technological Republic, quality Land and 723 00:45:20,910 --> 00:45:25,020 S1: poor Charlie's Almanack. If you do even two books a month, 724 00:45:25,020 --> 00:45:29,310 S1: your brain will be flooded with ideas. Seriously, trust the 725 00:45:29,310 --> 00:45:32,969 S1: system and the aphorism of the week. The purpose of 726 00:45:32,969 --> 00:45:38,280 S1: knowledge is action, not knowledge. The purpose of knowledge is action, 727 00:45:38,280 --> 00:45:44,280 S1: not knowledge. Aristotle. Unsupervised learning is produced on Hindenburg Pro 728 00:45:44,310 --> 00:45:48,060 S1: using an SM seven B microphone. A video version of 729 00:45:48,060 --> 00:45:51,660 S1: the podcast is available on the Unsupervised Learning YouTube channel, 730 00:45:51,780 --> 00:45:54,420 S1: and the text version with full links and notes is 731 00:45:54,420 --> 00:45:59,790 S1: available at Daniel Mysa.com slash newsletter. We'll see you next time.