1 00:00:01,230 --> 00:00:04,470 S1: Welcome to Unsupervised Learning, a security, AI and meaning focused 2 00:00:04,470 --> 00:00:07,380 S1: podcast that looks at how best to thrive as humans 3 00:00:07,380 --> 00:00:11,580 S1: in a post AI world. It combines original ideas, analysis, 4 00:00:11,580 --> 00:00:14,790 S1: and mental models to bring not just the news, but 5 00:00:14,790 --> 00:00:22,569 S1: why it matters and how to respond. All right. Welcome 6 00:00:22,570 --> 00:00:25,030 S1: to unsupervised learning. This is Daniel Missler going to do 7 00:00:25,030 --> 00:00:28,570 S1: an Andrew Huberman Lex Fridman type ad read here. And 8 00:00:28,570 --> 00:00:31,420 S1: you can of course just jump forward. But like they 9 00:00:31,420 --> 00:00:34,750 S1: say it would be appreciated if you listen and take 10 00:00:34,750 --> 00:00:37,059 S1: a look at the sponsor. Definitely helps the show. When 11 00:00:37,060 --> 00:00:39,430 S1: you go through airport security, there's one line where the 12 00:00:39,430 --> 00:00:42,250 S1: TSA agent checks your ID and another line where a 13 00:00:42,250 --> 00:00:45,640 S1: machine scans your bag. Same thing happens in enterprise security, 14 00:00:45,640 --> 00:00:48,730 S1: but instead of passengers and luggage, it's end users and 15 00:00:48,729 --> 00:00:51,400 S1: their devices. These days, most companies are pretty good at 16 00:00:51,400 --> 00:00:54,550 S1: the first part of the equation where they check user identity, 17 00:00:54,550 --> 00:00:57,850 S1: but user devices can roll right through authentication without getting 18 00:00:57,850 --> 00:01:02,110 S1: inspected at all. And in fact, 47% of companies allow unmanaged, 19 00:01:02,110 --> 00:01:05,230 S1: untrusted devices to access their data. That means an employee 20 00:01:05,230 --> 00:01:07,179 S1: can log in from a laptop that has its own 21 00:01:07,180 --> 00:01:09,880 S1: firewall turned off, and hasn't been updated in like six 22 00:01:09,880 --> 00:01:13,180 S1: months or worse. That laptop might actually belong to a 23 00:01:13,180 --> 00:01:17,440 S1: bad actor. Using employee credentials, collide finally solves the device 24 00:01:17,440 --> 00:01:20,440 S1: trust problem. Collide ensures that no device can log on 25 00:01:20,440 --> 00:01:24,460 S1: to your Okta protected apps unless it passes your security checks. Plus, 26 00:01:24,459 --> 00:01:27,760 S1: you can use Collider and devices without MDM like your 27 00:01:27,760 --> 00:01:31,720 S1: Linux fleet contractor devices in every BYoD phone and laptop 28 00:01:31,720 --> 00:01:40,839 S1: in your company, visit collider.com/unsupervised learning that's collide.com/unsupervised learning to 29 00:01:40,840 --> 00:01:42,850 S1: watch a demo and see how it works. All right 30 00:01:42,850 --> 00:01:46,810 S1: let's get into it. So some really cool patterns for 31 00:01:46,810 --> 00:01:50,050 S1: fabric this week. Got one called Create Better frame. So 32 00:01:50,050 --> 00:01:53,320 S1: basically any place where somebody is putting forth an opinion 33 00:01:53,320 --> 00:01:56,080 S1: about anything and it basically makes it so that when 34 00:01:56,080 --> 00:01:59,020 S1: you send that thing in, it looks at the negative 35 00:01:59,020 --> 00:02:01,240 S1: way to look at that content, and it looks at 36 00:02:01,240 --> 00:02:03,820 S1: positive ways to look at that content, and that's the 37 00:02:03,820 --> 00:02:06,310 S1: output that it produces and the negative ways to look 38 00:02:06,310 --> 00:02:09,940 S1: at the content. It's actually like in escalating levels. So 39 00:02:09,940 --> 00:02:13,480 S1: if it's about dating for example, it'll be like, oh, 40 00:02:13,480 --> 00:02:15,910 S1: the dating scene is really bad. And then it kind 41 00:02:15,910 --> 00:02:19,480 S1: of like catastrophizing that. And the next one is like, 42 00:02:19,480 --> 00:02:22,120 S1: I'll never find anyone in the last one is like, 43 00:02:22,120 --> 00:02:25,480 S1: why go on living if I'm always going to be alone? 44 00:02:25,480 --> 00:02:28,630 S1: So it's like negative frames of thinking that could be 45 00:02:28,630 --> 00:02:31,960 S1: generated from a given article or something. And what this 46 00:02:31,960 --> 00:02:35,230 S1: does is it also adds the positive ones at the 47 00:02:35,230 --> 00:02:39,160 S1: end and says, no, it's more like, okay, dating apps 48 00:02:39,160 --> 00:02:42,669 S1: aren't very good, but the final one there is, look, 49 00:02:42,669 --> 00:02:49,540 S1: I already wanted to prioritize in-person meet ups and relationships anyway, 50 00:02:49,540 --> 00:02:53,680 S1: so this is just support. This article about bad dating 51 00:02:53,680 --> 00:02:56,169 S1: apps is just support for the fact that I wanted 52 00:02:56,169 --> 00:02:59,620 S1: more in-person connection anyway, so it's a positive frame for 53 00:02:59,620 --> 00:03:01,690 S1: a thing. So the idea here is that this can 54 00:03:01,690 --> 00:03:05,860 S1: function essentially like a positivity filter, which eventually will be 55 00:03:05,860 --> 00:03:08,320 S1: like a function of days. So that's why I'm super 56 00:03:08,320 --> 00:03:11,350 S1: excited about this one. Next one is create Academic paper. 57 00:03:11,350 --> 00:03:14,080 S1: So this one takes any bullet points article, essay or 58 00:03:14,080 --> 00:03:16,840 S1: anything you've written and turns it into like a LaTeX 59 00:03:16,840 --> 00:03:20,530 S1: formatted academic paper. And it's pretty cool. It works great. 60 00:03:20,530 --> 00:03:23,320 S1: And you just paste it into like an online latex 61 00:03:23,320 --> 00:03:26,890 S1: renderer and you get the results. Okay. Next one here 62 00:03:26,889 --> 00:03:29,440 S1: is summarize git changes, which is a great way to 63 00:03:29,440 --> 00:03:33,160 S1: see and share updates on recent progress for a git project. Yeah, 64 00:03:33,160 --> 00:03:35,530 S1: so really cool. So you basically just paste in the 65 00:03:35,530 --> 00:03:38,560 S1: front page of a GitHub project and it outputs like 66 00:03:38,560 --> 00:03:42,070 S1: exactly how to install it or whatever. Oh actually that's 67 00:03:42,070 --> 00:03:44,740 S1: a different one. That's a different one. That is explain 68 00:03:44,740 --> 00:03:48,100 S1: project is actually the one I just described. This one 69 00:03:48,100 --> 00:03:51,640 S1: is different. This one is called summarize get changes and 70 00:03:51,640 --> 00:03:53,590 S1: this is the output from it. So you basically do 71 00:03:53,590 --> 00:03:57,700 S1: this git log command. And that pulls out like 500 changes. 72 00:03:57,700 --> 00:04:01,540 S1: And then it produces this type of a summary of 73 00:04:01,540 --> 00:04:03,580 S1: the actual project. So it's like here are all the 74 00:04:03,580 --> 00:04:06,190 S1: features we added like all that kind of stuff. Also 75 00:04:06,190 --> 00:04:11,050 S1: thresholds first commercial product or Ul's first commercial product called 76 00:04:11,050 --> 00:04:15,010 S1: threshold is imminent. I'm already using it and making final 77 00:04:15,010 --> 00:04:17,770 S1: tweaks now, and it's going to launch in preview, which 78 00:04:17,770 --> 00:04:20,260 S1: means they're going to be like tons of updates in 79 00:04:20,260 --> 00:04:23,770 S1: like the first couple of weeks and months, but can't 80 00:04:23,770 --> 00:04:26,260 S1: wait to share this thing. Wrote a piece called Personal 81 00:04:26,260 --> 00:04:29,710 S1: Eyes will mediate everything, and it's just talking about, you know, 82 00:04:29,710 --> 00:04:34,180 S1: what happens when Daz, your digital assistants are actually doing 83 00:04:34,180 --> 00:04:36,580 S1: all the interaction with the products around you. And this 84 00:04:36,580 --> 00:04:38,560 S1: is kind of like a picture of that. In fact, 85 00:04:38,560 --> 00:04:40,480 S1: we'll just go and look at that real quick. So 86 00:04:40,480 --> 00:04:42,970 S1: it's like if you look at this picture, you go 87 00:04:42,970 --> 00:04:46,300 S1: into a city street or whatever, you're surrounded by thousands 88 00:04:46,300 --> 00:04:50,620 S1: upon thousands of APIs, demons from the things all around you, 89 00:04:50,620 --> 00:04:53,049 S1: all the people, all the different buildings, all the all 90 00:04:53,050 --> 00:04:55,870 S1: the streets and the cars and everything. And, you know, 91 00:04:55,870 --> 00:04:58,990 S1: like the the driver service or Uber or whatever. And 92 00:04:58,990 --> 00:05:02,350 S1: all of these are actually like products, and you might 93 00:05:02,350 --> 00:05:04,570 S1: want to order some food or something. And it has 94 00:05:04,570 --> 00:05:07,420 S1: a menu and that's a product. And normally you would 95 00:05:07,420 --> 00:05:09,849 S1: be going there and looking at it and flipping through. 96 00:05:09,850 --> 00:05:13,810 S1: And as a result, people build applications in a certain way. Right. 97 00:05:13,810 --> 00:05:17,710 S1: And with AI happening, an API is happening and DHA 98 00:05:17,710 --> 00:05:19,870 S1: is happening. Humans aren't going to be the ones going 99 00:05:19,900 --> 00:05:21,930 S1: to the webpage. They're not going to be. Ones looking 100 00:05:21,930 --> 00:05:24,270 S1: at the catalog, they're going to know our preferences. They're 101 00:05:24,270 --> 00:05:26,160 S1: going to be doing all the looking. And then what 102 00:05:26,160 --> 00:05:28,050 S1: they're going to do is they're going to pull all 103 00:05:28,050 --> 00:05:31,229 S1: this content back via the API, and then they're going 104 00:05:31,230 --> 00:05:34,740 S1: to use your own preferred way of looking at things 105 00:05:34,740 --> 00:05:37,320 S1: inside of AR, and that's going to be shown inside 106 00:05:37,320 --> 00:05:41,370 S1: your glasses or your lenses or whatever the the modality is. 107 00:05:41,370 --> 00:05:45,300 S1: So you're going to have your own preferred UI interfaces 108 00:05:45,300 --> 00:05:48,060 S1: that you just like the best. Like it might look cyberpunk, 109 00:05:48,060 --> 00:05:51,690 S1: it might look solar punk, it might look minimalist, or 110 00:05:51,690 --> 00:05:55,560 S1: it might look like super pretty and gaudy and extravagant 111 00:05:55,560 --> 00:05:58,080 S1: or whatever. And that's going to be how you see things. 112 00:05:58,080 --> 00:06:01,890 S1: But again, like all these are company APIs, right? And 113 00:06:01,890 --> 00:06:04,979 S1: her Da is now looking at these things and parsing 114 00:06:04,980 --> 00:06:07,740 S1: all that information for her. So the real question is, 115 00:06:07,740 --> 00:06:11,219 S1: as a builder of a product like Google or like 116 00:06:11,220 --> 00:06:13,710 S1: a product catalog or like Amazon, what does it mean 117 00:06:13,710 --> 00:06:17,159 S1: to your company if nobody's coming to the website because 118 00:06:17,160 --> 00:06:19,289 S1: the Das are coming to the website and there's going 119 00:06:19,290 --> 00:06:21,780 S1: to be separate third parties that do the display, and 120 00:06:21,779 --> 00:06:25,529 S1: that's going to change a whole lot about product development. 121 00:06:25,529 --> 00:06:29,040 S1: And I don't know, really the economy in general. All right. 122 00:06:29,040 --> 00:06:32,340 S1: Had a really cool sponsored conversation with Jason Miller of 123 00:06:32,339 --> 00:06:36,510 S1: One Password. And, uh, he was the CEO at collide 124 00:06:36,510 --> 00:06:39,360 S1: and is still over that product inside of one password. 125 00:06:39,360 --> 00:06:42,360 S1: And that is the sponsor slot that we read earlier 126 00:06:42,360 --> 00:06:46,440 S1: as well. So thanks to Callide for sponsoring and, uh, security. 127 00:06:46,440 --> 00:06:49,049 S1: So got a set of deepfakes here. Actually I'm going 128 00:06:49,050 --> 00:06:51,150 S1: to click on these, right. I mean, this is the 129 00:06:51,150 --> 00:06:53,940 S1: advantage of doing video, right. So I'm going to show 130 00:06:53,940 --> 00:06:56,640 S1: a couple of these. Um, this one is the most 131 00:06:56,640 --> 00:06:57,630 S1: scary one to me. 132 00:06:57,660 --> 00:07:00,599 S2: World from art to music. Today we're diving into how 133 00:07:00,630 --> 00:07:03,510 S2: AI is transforming the creative world from art to music. 134 00:07:03,510 --> 00:07:06,510 S2: I'm the deepfake version of Marc Andreessen, made by Argo, 135 00:07:06,510 --> 00:07:08,940 S2: and I think we need more and more realistic deepfake. 136 00:07:08,940 --> 00:07:11,730 S2: It will empower new form of content creation. You should 137 00:07:11,730 --> 00:07:12,510 S2: check it out. 138 00:07:12,510 --> 00:07:16,470 S1: I mean, that is insane. That is completely insane. So 139 00:07:16,470 --> 00:07:19,650 S1: what I'm basically saying is we need a Snopes. Okay, 140 00:07:19,860 --> 00:07:24,090 S1: we're going into 2024 election in the US. Video deepfakes 141 00:07:24,090 --> 00:07:27,000 S1: that are that good, that could be Biden saying something 142 00:07:27,000 --> 00:07:32,520 S1: or Trump saying something. Those situations are so bad that 143 00:07:32,520 --> 00:07:34,170 S1: we need to be able to account for them. So 144 00:07:34,170 --> 00:07:36,210 S1: this is what I wrote. We we need like a 145 00:07:36,210 --> 00:07:38,730 S1: global Snopes platform. So you get a bunch of people 146 00:07:38,730 --> 00:07:42,300 S1: left people, center people, right people, and you build a 147 00:07:42,300 --> 00:07:45,660 S1: platform that does like Snopes, which is like it tests 148 00:07:45,660 --> 00:07:48,330 S1: internet claims and basically says, is it real or is 149 00:07:48,330 --> 00:07:51,180 S1: it crap? And it could be like a collection point 150 00:07:51,180 --> 00:07:54,120 S1: for arguments, right? So it's something like, okay, there's a 151 00:07:54,120 --> 00:07:58,470 S1: video of Obama saying we're going to attack Mylanta. The 152 00:07:58,470 --> 00:08:01,140 S1: video is currently being analyzed, and then you have like 153 00:08:01,140 --> 00:08:05,250 S1: streaming in. You have oh, it's created by oh, looks 154 00:08:05,250 --> 00:08:08,100 S1: to be fake. So Fox News says, oh it's fake. Oh. 155 00:08:08,100 --> 00:08:11,040 S1: And then says there's no evidence that it's fake. Also 156 00:08:11,040 --> 00:08:13,200 S1: known as they believe it's real or they want you 157 00:08:13,200 --> 00:08:16,410 S1: to believe it's real. See some analysis. They're like, oh, 158 00:08:16,410 --> 00:08:18,780 S1: this is definitely a deep fake. And all these are 159 00:08:18,780 --> 00:08:22,020 S1: links to go actually look at their analysis. Right. And 160 00:08:22,020 --> 00:08:24,750 S1: Breitbart is like, no, he probably said it or whatever. 161 00:08:24,750 --> 00:08:28,020 S1: So current conclusion is, given all the evidence, we are 162 00:08:28,020 --> 00:08:31,500 S1: almost certain and that this is a thing roughly similar 163 00:08:31,500 --> 00:08:34,590 S1: to what the CIA has used in the past to 164 00:08:34,590 --> 00:08:38,280 S1: assess likelihood. So it's a good sort of scale to 165 00:08:38,280 --> 00:08:42,000 S1: use for how certain you are about a certain thing. So, um, 166 00:08:42,000 --> 00:08:44,520 S1: almost certain that this is a deepfake. So this is 167 00:08:44,520 --> 00:08:47,460 S1: the conclusion of the site, and you've got all the 168 00:08:47,460 --> 00:08:50,130 S1: evidence up here of like all the different analyses, they've 169 00:08:50,130 --> 00:08:53,310 S1: all been collected in one place. So some people are 170 00:08:53,309 --> 00:08:54,929 S1: just not going to believe it. I mean, they're going 171 00:08:54,929 --> 00:08:57,390 S1: to believe whatever they want to believe about the video 172 00:08:57,390 --> 00:09:00,510 S1: because of predetermined bias. But we need to have a 173 00:09:00,510 --> 00:09:05,939 S1: service in which somebody can see somewhat objective analysis, somewhat 174 00:09:05,940 --> 00:09:12,870 S1: multi subjective analysis. And then finally a verdict is this 175 00:09:12,870 --> 00:09:15,450 S1: thing real or not. And it could be like chances 176 00:09:15,450 --> 00:09:18,990 S1: about even we don't know impossible. We also need to 177 00:09:18,990 --> 00:09:21,330 S1: put like we just don't know. There's got to be 178 00:09:21,330 --> 00:09:24,060 S1: one of the options. But we absolutely need the service. 179 00:09:24,059 --> 00:09:26,309 S1: And Dan Kaminsky used to be fond of saying we 180 00:09:26,309 --> 00:09:29,100 S1: have the technology right. We can actually build this if 181 00:09:29,100 --> 00:09:32,219 S1: we want to. And basically the deepfakes are too good 182 00:09:32,220 --> 00:09:35,640 S1: to not have this. In the election season 2024, they 183 00:09:35,640 --> 00:09:38,939 S1: were supposedly a data leak of 71 million AT&T customers 184 00:09:38,940 --> 00:09:41,970 S1: are saying it wasn't them. Someone build a active in 185 00:09:41,970 --> 00:09:46,229 S1: the middle attack using some Cloudflare worker stuff. Really cool 186 00:09:46,230 --> 00:09:51,179 S1: bypass of MFA against Microsoft accounts. Leaked documents about a 187 00:09:51,179 --> 00:09:54,750 S1: hacking group that was going after like tons of foreign governments. 188 00:09:54,750 --> 00:09:58,260 S1: Just really nasty thing. SpaceX is contracted to build a 189 00:09:58,260 --> 00:10:02,070 S1: spy satellite network for US intelligence agency. So Elon is 190 00:10:02,070 --> 00:10:05,250 S1: building spy satellites for the government. That's interesting. I mean, 191 00:10:05,250 --> 00:10:07,860 S1: who better to actually get satellites in space? I don't 192 00:10:07,860 --> 00:10:13,020 S1: know of anyone. Rohan Pandit modified lama to to unredacted 193 00:10:13,020 --> 00:10:18,839 S1: an email from Elon to Ilya. So, uh, yeah. Unredacted content. Interesting. 194 00:10:18,840 --> 00:10:22,150 S1: Burglars are starting to use Wi-Fi jammers. Oh, actually, this. 195 00:10:22,150 --> 00:10:26,020 S1: This is our second sponsor. Hardly strictly security. I'm actually 196 00:10:26,020 --> 00:10:28,420 S1: doing a talk here, so shout out to them in 197 00:10:28,420 --> 00:10:31,569 S1: that conference. Yeah. Burglars are now using Wi-Fi jammers to 198 00:10:31,570 --> 00:10:35,770 S1: knock out security cameras, and evidently it's starting to happen more. 199 00:10:35,770 --> 00:10:38,530 S1: I've got a few cases. I think this was in Chicago. 200 00:10:38,530 --> 00:10:43,179 S1: Fortinet has disclosed a critical SQL injection flaw. Fortinet has 201 00:10:43,330 --> 00:10:47,380 S1: had some issues. All right. Technology. Stephen Howe gave Devin 202 00:10:47,380 --> 00:10:51,100 S1: access to his work stuff, which I'm like, how do 203 00:10:51,100 --> 00:10:53,920 S1: you do that? That's, uh, that's an agent with full 204 00:10:53,920 --> 00:10:57,040 S1: access to your account. This is like I've been talking about. 205 00:10:57,040 --> 00:10:59,530 S1: This is the number one threat from AI, in my opinion, 206 00:10:59,530 --> 00:11:02,380 S1: is people giving too much power to agents that have 207 00:11:02,380 --> 00:11:06,550 S1: access to really strong APIs, such as posting on slack 208 00:11:06,550 --> 00:11:10,000 S1: on behalf of the human right. So, Devin, is this 209 00:11:10,000 --> 00:11:13,209 S1: really controversial thing. It's basically an agent that writes code. 210 00:11:13,210 --> 00:11:15,459 S1: There's a million of them out right now. This one 211 00:11:15,460 --> 00:11:18,640 S1: is particularly well packaged. They had a good launch interview, 212 00:11:18,640 --> 00:11:21,730 S1: and I think that's why they're getting so much pushback. 213 00:11:21,730 --> 00:11:23,920 S1: It's like it just launched. Well it got a lot 214 00:11:23,920 --> 00:11:26,470 S1: of PR. So that's why people are more mad at 215 00:11:26,470 --> 00:11:28,870 S1: it than other things that they don't know about. That's 216 00:11:28,870 --> 00:11:33,100 S1: my theory anyway. Um, it can basically do tasks. It 217 00:11:33,100 --> 00:11:35,770 S1: sets out a thing of tasks it needs to do 218 00:11:35,770 --> 00:11:39,189 S1: to like write an application or something. Right? Well, if 219 00:11:39,190 --> 00:11:42,700 S1: it gets stuck, it can go and search many different 220 00:11:42,700 --> 00:11:45,520 S1: avenues to try to get unstuck. One of the avenues 221 00:11:45,520 --> 00:11:49,270 S1: that Steven Howe gave it was to go on slack. 222 00:11:49,270 --> 00:11:51,580 S1: So it actually goes on slack. I'm going to click this, 223 00:11:52,090 --> 00:11:56,500 S1: look at this, this thing. It goes on slack. And actually, um, 224 00:11:56,500 --> 00:12:00,130 S1: posts in slack to ask a question that would solve 225 00:12:00,130 --> 00:12:02,770 S1: its problem. And it got back a response, and it 226 00:12:02,770 --> 00:12:06,310 S1: used that response to go and continue writing code, got 227 00:12:06,309 --> 00:12:09,040 S1: unstuck and actually finished the project. And yeah, I was 228 00:12:09,040 --> 00:12:11,620 S1: talking about this like so much hate towards Steven. I'm 229 00:12:11,620 --> 00:12:14,950 S1: not sure why there's a million different platforms doing something similar. 230 00:12:14,950 --> 00:12:19,570 S1: And uh, Midjourney new character reference feature finally makes things 231 00:12:19,570 --> 00:12:21,880 S1: look the same. So check this out. This is all 232 00:12:21,880 --> 00:12:24,370 S1: from one prompt. So basically you can have the same 233 00:12:24,370 --> 00:12:28,090 S1: person in different settings. And uh, that was not possible before. 234 00:12:28,090 --> 00:12:31,929 S1: So that's really really cool feature now in Midjourney uh 235 00:12:32,230 --> 00:12:36,699 S1: open source grok. Not really. Um, in my opinion, the weights, 236 00:12:36,700 --> 00:12:40,059 S1: the data and the full training methodology are used for 237 00:12:40,059 --> 00:12:44,080 S1: saying that a model is fully open source and grok 238 00:12:44,080 --> 00:12:46,750 S1: did not get there. They did not release the methodology 239 00:12:46,750 --> 00:12:50,890 S1: or all of it. At least Co-variance is launching RFM one, 240 00:12:50,890 --> 00:12:55,630 S1: which is bringing ChatGPT like capabilities to robots. AI is big, 241 00:12:55,630 --> 00:12:59,650 S1: robots are big, but the biggest is AI in robots. 242 00:12:59,650 --> 00:13:03,100 S1: This I think might move faster than we think everyone 243 00:13:03,100 --> 00:13:05,320 S1: is thinking. You know you can't crawl under a house, 244 00:13:05,320 --> 00:13:07,900 S1: you can't look at the piping from an old house 245 00:13:07,900 --> 00:13:10,210 S1: and actually figure out how to fix this, like a 246 00:13:10,210 --> 00:13:13,870 S1: an electrician or a plumber. And if you've seen what 247 00:13:13,870 --> 00:13:16,210 S1: I can do, you've seen what some of these robot 248 00:13:16,210 --> 00:13:19,540 S1: demos can do. And then you see how nimble these 249 00:13:19,540 --> 00:13:22,719 S1: little things can be. And like the dexterity of their 250 00:13:22,720 --> 00:13:27,160 S1: arms and their hands. Combine that with local models and or, 251 00:13:27,160 --> 00:13:30,880 S1: you know, pinnacle models that are available via the cloud. 252 00:13:30,880 --> 00:13:34,390 S1: I'm not sure that this thing can't take pictures or 253 00:13:34,390 --> 00:13:40,390 S1: stream video of pipe fittings and washers and, you know, 254 00:13:40,840 --> 00:13:43,300 S1: be able to use wrenches and all this different stuff. 255 00:13:43,300 --> 00:13:46,000 S1: I'm not saying it's going to be easy. I'm saying 256 00:13:46,000 --> 00:13:49,780 S1: we should not be so arrogant as to think it's invulnerable, 257 00:13:49,780 --> 00:13:53,800 S1: like we currently seem to be thinking it is. All right. Um, 258 00:13:53,800 --> 00:13:57,670 S1: Finland has a giant sand battery to store heat in winter. 259 00:13:57,880 --> 00:14:01,090 S1: I didn't know this was possible. That one really excites me. 260 00:14:01,090 --> 00:14:02,710 S1: And the other one that excites me is I'm trying 261 00:14:02,710 --> 00:14:07,660 S1: to figure out is storing energy in water upstream. I 262 00:14:07,660 --> 00:14:10,030 S1: can't remember how this works. I think they store the 263 00:14:10,030 --> 00:14:13,390 S1: water and then they release the water through like a 264 00:14:13,390 --> 00:14:16,270 S1: turbine that the water moves through as it goes downhill. 265 00:14:16,270 --> 00:14:20,710 S1: Something about the gravity and potential energy of all that weight, 266 00:14:20,710 --> 00:14:26,500 S1: I think. Anyway, another way of like generating and storing energy. Yeah. 267 00:14:26,500 --> 00:14:31,030 S1: Nvidia getting into human robots. Yeah. This thing, it honestly 268 00:14:31,030 --> 00:14:33,220 S1: reminded me of Black Mirror. I saw a video of 269 00:14:33,220 --> 00:14:36,430 S1: this thing jumping up like 3 or 4ft high and 270 00:14:36,430 --> 00:14:39,880 S1: just like instantly just going over this giant obstacle at 271 00:14:39,880 --> 00:14:43,330 S1: the Nvidia event. And it was terrifying. Hong Kong is 272 00:14:43,330 --> 00:14:46,630 S1: implementing a new Beijing driven, stringent security law that goes 273 00:14:46,630 --> 00:14:49,780 S1: after treason and other types. And it's got like life 274 00:14:49,780 --> 00:14:53,020 S1: in prison, I think maybe even worse, maybe even death. 275 00:14:53,020 --> 00:14:56,080 S1: I mostly saw life in prison, though. Midjourney is blocking 276 00:14:56,080 --> 00:14:59,260 S1: AI images of Trump and Biden. US added more jobs 277 00:14:59,260 --> 00:15:02,770 S1: than usual, but unemployment went slightly up to 3.9. That 278 00:15:02,770 --> 00:15:04,990 S1: is such a low number, but it trips me out 279 00:15:04,990 --> 00:15:07,239 S1: that that number does not include people who don't want 280 00:15:07,240 --> 00:15:10,330 S1: to work. So I mean, if you counted that, I mean, 281 00:15:10,330 --> 00:15:12,370 S1: it would be a massive number. A really good thread 282 00:15:12,370 --> 00:15:15,130 S1: here on Hacker News about experienced programmers not being able 283 00:15:15,130 --> 00:15:19,300 S1: to find jobs, super strict policies at private schools in England, 284 00:15:19,300 --> 00:15:23,710 S1: rigid routines and discipline. And this is for disadvantaged students 285 00:15:23,710 --> 00:15:26,080 S1: to help them succeed. And they've had really good results. 286 00:15:26,110 --> 00:15:29,020 S1: I don't know, I feel like this is probably going 287 00:15:29,020 --> 00:15:31,960 S1: to be a trend, not just for disadvantaged students, which 288 00:15:31,960 --> 00:15:34,660 S1: I think it's great for, but for everyone really reminds 289 00:15:34,660 --> 00:15:36,729 S1: me of, like those man camps that are happening right 290 00:15:36,730 --> 00:15:40,000 S1: now in like, uh, Appalachia mountains where you go off 291 00:15:40,000 --> 00:15:42,850 S1: and you do camping and hunting like all these other 292 00:15:42,850 --> 00:15:46,150 S1: manly things. And it's because people want to build character, right? 293 00:15:46,150 --> 00:15:49,180 S1: And you can't build character when everything is easy. So 294 00:15:49,180 --> 00:15:52,239 S1: these schools or these man camps or whatever, it's a 295 00:15:52,240 --> 00:15:56,020 S1: way of like getting to like this resistance training, this 296 00:15:56,020 --> 00:16:00,100 S1: calculated suffering that builds character. Uh, women are drifting more 297 00:16:00,100 --> 00:16:02,770 S1: left and men are not. So there's like a bigger 298 00:16:02,770 --> 00:16:06,370 S1: gap than ever. Former Boeing whistleblower, the I think I 299 00:16:06,370 --> 00:16:09,160 S1: don't know if he was the 1 or 1 of them, 300 00:16:09,160 --> 00:16:13,030 S1: but found dead amid a lawsuit and, uh, some drama 301 00:16:13,030 --> 00:16:15,820 S1: around that. And Toronto police are now saying, leave your 302 00:16:15,820 --> 00:16:18,640 S1: car keys at the front door so that when people 303 00:16:18,640 --> 00:16:21,070 S1: break in, because we're probably not going to be able 304 00:16:21,070 --> 00:16:23,500 S1: to help you. Basically, when people break in, you want 305 00:16:23,500 --> 00:16:25,540 S1: to make it easy for them to take your car 306 00:16:25,540 --> 00:16:27,400 S1: and so they won't come in and hurt you. And 307 00:16:27,400 --> 00:16:29,920 S1: I think this is like how you get Republicans elected 308 00:16:29,920 --> 00:16:32,320 S1: because Republicans are going to be like, I got an idea. 309 00:16:32,320 --> 00:16:35,500 S1: How about we shoot the the attacker? How about we 310 00:16:35,500 --> 00:16:37,510 S1: don't allow this to happen? How about we have more 311 00:16:37,510 --> 00:16:41,080 S1: police so we don't just accept break ins. And this 312 00:16:41,080 --> 00:16:44,680 S1: is how you get like more right leaning people eventually 313 00:16:44,680 --> 00:16:48,250 S1: getting to to like far right people like we're seeing 314 00:16:48,250 --> 00:16:53,290 S1: all over like Sweden and, um, you know, Scandinavian countries 315 00:16:53,290 --> 00:16:56,470 S1: where you have like huge like migrant crime problems. And 316 00:16:56,470 --> 00:16:59,260 S1: the left is like, no, immigration is fine, which of 317 00:16:59,260 --> 00:17:01,960 S1: course it is. But it's like when you're not addressing 318 00:17:01,960 --> 00:17:05,290 S1: the issues, you give the power of the truth to 319 00:17:05,290 --> 00:17:08,350 S1: the far right. And like liberals cannot do this. And 320 00:17:08,350 --> 00:17:11,200 S1: if they do, they're going to be consequences. So I 321 00:17:11,200 --> 00:17:13,389 S1: think we need to, you know, do some sort of 322 00:17:13,390 --> 00:17:16,840 S1: hybrid of like, you've got to enforce laws strictly and 323 00:17:16,840 --> 00:17:18,879 S1: then you have to be very liberal in trying to 324 00:17:18,880 --> 00:17:20,889 S1: make it. So that doesn't happen in the first place. 325 00:17:20,890 --> 00:17:24,280 S1: So conservative in some ways, liberal in other ways. Uh, 326 00:17:24,280 --> 00:17:27,790 S1: recent Boing incidents have got a far right conspiracy going 327 00:17:27,790 --> 00:17:32,439 S1: that diversity is causing intentional failures, and it's part of 328 00:17:32,440 --> 00:17:36,520 S1: a plot to undermine Western civilization and promote communism. I'm like, 329 00:17:36,670 --> 00:17:40,180 S1: somebody explain this to me. That seems fringe. Using tap 330 00:17:40,180 --> 00:17:42,280 S1: water in the neti pot is bad. So this is 331 00:17:42,280 --> 00:17:44,470 S1: the thing that cleans out your sinuses, which I need 332 00:17:44,470 --> 00:17:47,650 S1: to do again right now. I did yesterday, but you 333 00:17:47,650 --> 00:17:51,280 S1: should use sterilized water for this. I use my reverse 334 00:17:51,280 --> 00:17:55,540 S1: osmosis water from the filter. You do not want nasty stuff. 335 00:17:55,570 --> 00:17:58,120 S1: You're shooting it up like very close to your brain. 336 00:17:58,119 --> 00:18:01,450 S1: Like your sinus cavity goes way up inside your your head. 337 00:18:01,450 --> 00:18:04,570 S1: So you do not want stuff up there that's like 338 00:18:04,570 --> 00:18:07,929 S1: alive or poisonous or whatever. And if you look at 339 00:18:07,930 --> 00:18:13,030 S1: Huberman recent episode on water, he says tap water is nasty. 340 00:18:13,030 --> 00:18:15,130 S1: He's he's like, I don't want to alarm people, but 341 00:18:15,130 --> 00:18:18,850 S1: it is alarming. It is nasty what I found in 342 00:18:18,850 --> 00:18:22,240 S1: these tap waters. So don't shoot that up into your brain. 343 00:18:22,240 --> 00:18:24,730 S1: How about that? This analysis is saying people used to 344 00:18:24,730 --> 00:18:29,290 S1: consume more calories without gaining as much weight. Skeptical, but 345 00:18:29,290 --> 00:18:33,490 S1: thought it was worth sharing. Uh, fentanyl poisoning the leading 346 00:18:33,490 --> 00:18:37,630 S1: cause of death for Americans 18 to 45 fentanyl. What 347 00:18:37,630 --> 00:18:42,010 S1: is going on? 2000 newspapers have closed since 2004, and 348 00:18:42,010 --> 00:18:44,649 S1: car washes are popping up because, uh, they make a 349 00:18:44,650 --> 00:18:52,900 S1: lot of money. All right. Ideas and analysis. Uh. Epiphany. Epiphany. 350 00:18:52,930 --> 00:18:56,409 S1: What is that? Uh, that, um, that shows you I 351 00:18:56,410 --> 00:18:59,470 S1: don't use AI to to write the newsletter. I wouldn't 352 00:18:59,470 --> 00:19:02,350 S1: have made that mistake. Yeah. So last couple of months, 353 00:19:02,350 --> 00:19:04,119 S1: I'm not going to go into this one. This this 354 00:19:04,119 --> 00:19:07,750 S1: is a big topic by itself. Basically, I'm learning a 355 00:19:07,750 --> 00:19:10,689 S1: lot about the power of framing. And it's kind of 356 00:19:10,690 --> 00:19:13,929 S1: turning into like a, you know, universal like unified theory 357 00:19:13,930 --> 00:19:15,939 S1: for me for a lot of things. But worth a 358 00:19:15,940 --> 00:19:19,270 S1: separate discussion. All right. Really interesting back and forth with 359 00:19:19,270 --> 00:19:23,410 S1: Dino de survie about the cybersecurity flaw and sealing. This 360 00:19:23,410 --> 00:19:25,060 S1: one is a deep one. I'm not going to go 361 00:19:25,060 --> 00:19:27,670 S1: into it. You should definitely go check it out. If 362 00:19:27,670 --> 00:19:32,619 S1: these types of deep philosophical security things are your jam. Basically, 363 00:19:32,619 --> 00:19:35,379 S1: I think security is so important to innovation in daily 364 00:19:35,380 --> 00:19:38,530 S1: life in most situations, and that it falls to a 365 00:19:38,530 --> 00:19:41,830 S1: minimum as a result. And basically we should guard our 366 00:19:41,830 --> 00:19:45,970 S1: mental health against thinking people are steering us wrong and 367 00:19:45,970 --> 00:19:49,810 S1: that's why we have low security. But in fact, where 368 00:19:49,810 --> 00:19:52,900 S1: a lot of security people think it's a horrible situation, 369 00:19:52,900 --> 00:19:56,320 S1: it's actually not that bad because if it were that bad, 370 00:19:56,320 --> 00:19:59,350 S1: we would have fixed it. If it were urgent, we 371 00:19:59,350 --> 00:20:01,540 S1: would know because it would get fixed immediately. And if 372 00:20:01,540 --> 00:20:06,250 S1: it's not fixed immediately, then it's not urgent. And there's 373 00:20:06,250 --> 00:20:09,310 S1: just like, I don't know, it's it's my view of 374 00:20:09,310 --> 00:20:12,800 S1: security at this point. Um. And doesn't mean you can't 375 00:20:12,800 --> 00:20:15,770 S1: make progress. You absolutely can make progress. It's just that 376 00:20:15,770 --> 00:20:20,570 S1: progress tends to happen slowly and reactively. And a lot 377 00:20:20,570 --> 00:20:23,630 S1: of security people are really upset about. They don't realize that. 378 00:20:23,630 --> 00:20:26,720 S1: So they're very upset because they think something can be 379 00:20:26,720 --> 00:20:30,199 S1: done when they don't realize that they're hitting the ceiling 380 00:20:30,200 --> 00:20:34,040 S1: of the difference between what can be done and what 381 00:20:34,040 --> 00:20:38,240 S1: needs to be done. And security by default is a 382 00:20:38,240 --> 00:20:41,900 S1: very efficient thing. So I should probably write a piece 383 00:20:41,900 --> 00:20:43,790 S1: and like name it something like that. I think I 384 00:20:43,790 --> 00:20:48,980 S1: saw a comment like security efficiency theory set or something. 385 00:20:48,980 --> 00:20:53,149 S1: So it's like security wants to be as efficient as possible, 386 00:20:53,150 --> 00:20:56,600 S1: which is why it's easy to pick locks all over 387 00:20:56,600 --> 00:20:59,840 S1: the country and probably all over the world. It's because 388 00:20:59,840 --> 00:21:03,620 S1: that is the exact point of security that you need 389 00:21:03,619 --> 00:21:06,500 S1: to have for it to be somewhat useful, but not 390 00:21:06,500 --> 00:21:10,159 S1: to be an overspend. And that magical point I'm arguing, 391 00:21:10,160 --> 00:21:14,270 S1: is usually way lower than what security professionals think it 392 00:21:14,270 --> 00:21:16,880 S1: should be. They think the bar should be way up here, 393 00:21:16,880 --> 00:21:20,540 S1: and I think that's what Dino was talking about. Let's 394 00:21:20,540 --> 00:21:24,110 S1: raise the minimum bar. And my whole point I said 395 00:21:24,109 --> 00:21:25,910 S1: I wasn't going to go into this, but here we are. 396 00:21:25,910 --> 00:21:29,389 S1: My whole point was you can't move that bar. That 397 00:21:29,390 --> 00:21:34,520 S1: is security's efficiency baseline. And it's determined by reality. It's 398 00:21:34,520 --> 00:21:38,780 S1: not determined by what should be or what security people 399 00:21:38,780 --> 00:21:41,899 S1: want it to be. And it's related to this framing thing. 400 00:21:41,900 --> 00:21:45,290 S1: So we'll leave it there. All right. Really cool. You 401 00:21:45,290 --> 00:21:48,410 S1: all meet up this week. Had a member named John 402 00:21:48,410 --> 00:21:52,010 S1: just blow our minds about custom keyboards. And I now 403 00:21:52,010 --> 00:21:54,410 S1: have one on the way. So I'm going to be 404 00:21:54,410 --> 00:21:59,750 S1: having my tinted, like side keyboard programmable. Really cool. And 405 00:21:59,750 --> 00:22:03,800 S1: I ordered a kit from Jose and Martinez, so that 406 00:22:03,800 --> 00:22:07,400 S1: should be fun. And Voyager one is one light day 407 00:22:07,400 --> 00:22:10,370 S1: away from us and it still it keeps breaking. We're like, oh, 408 00:22:10,369 --> 00:22:14,690 S1: it's finally dead, you know, salutes, hats off to you. 409 00:22:14,690 --> 00:22:17,720 S1: And then it'll be like, I woke up. I'm okay, 410 00:22:17,720 --> 00:22:20,869 S1: I'm okay, I'm okay. And we're like, how are you? Okay. 411 00:22:20,869 --> 00:22:23,840 S1: It's like, what? I figured it out. We're okay. And 412 00:22:23,840 --> 00:22:27,530 S1: they're talking about having to troubleshoot this thing from 24 413 00:22:27,530 --> 00:22:30,980 S1: light hours away. It roughly, it might be closer or 414 00:22:30,980 --> 00:22:34,669 S1: further at this point, but roughly a day. How far 415 00:22:34,670 --> 00:22:37,070 S1: away does the thing have to be for the speed 416 00:22:37,070 --> 00:22:40,070 S1: of light, which is that fast? Eight minutes, by the way, 417 00:22:40,070 --> 00:22:43,520 S1: to the sun, 45 minutes for Jupiter, and this thing 418 00:22:43,520 --> 00:22:47,570 S1: is 24 hours away and we have to troubleshoot it. 419 00:22:47,570 --> 00:22:50,630 S1: So imagine like a vim session and you're typing keys. 420 00:22:50,630 --> 00:22:52,310 S1: And by the way, that's why a lot of the 421 00:22:52,310 --> 00:22:54,740 S1: commands are the way they are. I think Bill Joy 422 00:22:54,740 --> 00:22:58,189 S1: created VI and it was for interacting with very distant 423 00:22:58,190 --> 00:23:01,129 S1: things I think satellites. So you'd have this slight delay 424 00:23:01,130 --> 00:23:05,360 S1: and low resources on each computer. Well, imagine when a keystroke. 425 00:23:05,359 --> 00:23:08,780 S1: I imagine they batch things, but imagine a keystroke taking 426 00:23:08,780 --> 00:23:11,869 S1: a day just to get there and a day to 427 00:23:11,869 --> 00:23:15,230 S1: get back. And final thought on on Voyager one. What 428 00:23:15,230 --> 00:23:18,530 S1: a hero. All right, discovery hack trails Golang client allows 429 00:23:18,530 --> 00:23:22,160 S1: you to query security trails, API and do super useful 430 00:23:22,160 --> 00:23:25,939 S1: stuff for bounty hunters. And it's made by my buddy Luke. Uh, 431 00:23:25,940 --> 00:23:29,840 S1: super cool guy. Open API Tui lets you interact with 432 00:23:29,840 --> 00:23:34,280 S1: APIs with a terminal user interface, which I love. I 433 00:23:34,280 --> 00:23:37,489 S1: stopped loving Captain Kirk. Really good piece. Solar punk is 434 00:23:37,490 --> 00:23:40,609 S1: a new cyberpunk. Do It Now by Steve Pavlovna, one 435 00:23:40,609 --> 00:23:43,879 S1: of the first pieces that I read on productivity. Uh, 436 00:23:43,880 --> 00:23:48,830 S1: 2005 really takes me back that one minimal viable system. 437 00:23:48,830 --> 00:23:51,859 S1: Ben Koon shares why and how to blog which skills 438 00:23:51,859 --> 00:23:54,920 S1: are at least likely to be replaced by AI. Amanda 439 00:23:54,920 --> 00:23:58,340 S1: Haskell talks about why cloud three is so good. By 440 00:23:58,340 --> 00:24:02,240 S1: the way, cloud three opus I'm only using opus. It 441 00:24:02,240 --> 00:24:06,320 S1: is crushing GPT four in the most advanced tasks that 442 00:24:06,320 --> 00:24:09,050 S1: I have, which is like deep analysis of like the 443 00:24:09,050 --> 00:24:12,440 S1: most interesting and surprising ideas. So it's like the pinnacle. 444 00:24:12,440 --> 00:24:15,139 S1: It's like the top end of the most difficult things 445 00:24:15,140 --> 00:24:18,619 S1: for humans and AI to do, to pull out complex 446 00:24:18,619 --> 00:24:24,139 S1: ideas and distill them from a corpus, an entire book 447 00:24:24,140 --> 00:24:26,570 S1: or an entire video or whatever. To be able to 448 00:24:26,570 --> 00:24:29,389 S1: do that is the hardest thing that I've seen I 449 00:24:29,390 --> 00:24:34,010 S1: do anywhere. And, um, at that type of skill, opus 450 00:24:34,040 --> 00:24:37,220 S1: is crushing GPT four. I mean, it's like two or 451 00:24:37,220 --> 00:24:40,520 S1: 3 or 4 times as good in. It's hard to 452 00:24:40,520 --> 00:24:44,210 S1: measure that. So I'm making up those numbers. But intuitively, 453 00:24:44,210 --> 00:24:47,030 S1: I feel like the quality of GPT four is somewhere, 454 00:24:47,030 --> 00:24:49,909 S1: and this thing is like so much better now in 455 00:24:49,910 --> 00:24:54,110 S1: most other tasks. Uh, GPT four is as good. Um, 456 00:24:54,109 --> 00:24:57,980 S1: in some places it's actually still better than opus, but 457 00:24:57,980 --> 00:25:00,410 S1: and it's way faster. It's probably like three times faster, 458 00:25:00,410 --> 00:25:02,780 S1: I would guess. Yeah. Really interesting to see opus doing 459 00:25:02,780 --> 00:25:06,170 S1: so well. Spreadsheets as simulation tools. And the Getty has 460 00:25:06,170 --> 00:25:10,490 S1: released 88,000 art images for anyone to use. Recommendation of 461 00:25:10,490 --> 00:25:14,720 S1: the week. Let grow this thing. Oh this project. I 462 00:25:14,720 --> 00:25:21,870 S1: love this project. Can I click this? It's about resilience. 463 00:25:21,900 --> 00:25:27,900 S3: Independence is the key to developing happy, well-adjusted children. 464 00:25:28,920 --> 00:25:31,710 S1: Absolutely love this thing. It is a four minute video. 465 00:25:31,740 --> 00:25:34,050 S1: You've got to watch this thing. And I heard about 466 00:25:34,050 --> 00:25:38,550 S1: it from Jonathan Haidt. And reduce anxiety in elementary and 467 00:25:38,550 --> 00:25:42,900 S1: middle school is the let grow experience. It's all about teaching, 468 00:25:42,900 --> 00:25:48,840 S1: you know, self not so much discipline, independence, independence and resilience. 469 00:25:48,840 --> 00:25:53,790 S1: Trying falling down, getting up that sort of vibe. Absolutely. 470 00:25:53,790 --> 00:25:57,300 S1: Love this project. Okay. And the aphorism for the week 471 00:25:57,600 --> 00:26:00,000 S1: I am not what happened to me, I am what 472 00:26:00,000 --> 00:26:02,699 S1: I choose to become, I am not what happened to me, 473 00:26:02,700 --> 00:26:07,680 S1: I am what I choose to become. Carl Jung Unsupervised 474 00:26:07,680 --> 00:26:10,199 S1: Learning is produced and edited by Daniel Missler on a 475 00:26:10,200 --> 00:26:15,060 S1: Norman 87 I microphone using Hindenburg intro and outro. Music 476 00:26:15,060 --> 00:26:18,149 S1: is by zombie with the Y and to get the 477 00:26:18,150 --> 00:26:20,220 S1: text and links from this episode, sign up for the 478 00:26:20,220 --> 00:26:25,869 S1: newsletter version of the show at Daniel missler.com/newsletter. We'll see 479 00:26:25,869 --> 00:26:26,500 S1: you next time.