1 00:00:00,110 --> 00:00:03,800 S1: Say goodbye to countless spreadsheets and endless email threads when 2 00:00:03,800 --> 00:00:07,700 S1: you automate compliance with Fanta. Fanta automates up to 90% 3 00:00:07,700 --> 00:00:11,930 S1: of the work for Soc2, ISO 27,001, HIPAA, and more. 4 00:00:11,930 --> 00:00:15,500 S1: Getting you audit ready results in weeks instead of months. 5 00:00:15,500 --> 00:00:18,920 S1: Customers report saving over 300 hours of manual work and 6 00:00:18,920 --> 00:00:23,720 S1: up to 85% of associated costs. And with 200 plus integrations, 7 00:00:23,720 --> 00:00:26,540 S1: you can easily monitor and secure the tools your business 8 00:00:26,540 --> 00:00:30,560 S1: relies on. Join the 5000 fast growing companies that leverage 9 00:00:30,560 --> 00:00:34,340 S1: Fanta to manage risks and prove security in real time. 10 00:00:34,340 --> 00:00:38,060 S1: Vanta scales with your business, helping you successfully enter new markets, 11 00:00:38,060 --> 00:00:42,590 S1: land bigger deals, and earn customer trust. Unsupervised learning listeners 12 00:00:42,590 --> 00:00:52,040 S1: get $1,000 off Vanta at Vanta comm slash unsupervised. That's vanta.com/unsupervised. 13 00:00:52,940 --> 00:00:56,360 S1: Welcome to Unsupervised Learning, a security, AI, and meaning focused 14 00:00:56,360 --> 00:00:59,240 S1: podcast that looks at how best to thrive as humans. 15 00:00:59,240 --> 00:01:03,440 S1: In a post AI world. It combines original ideas, analysis, 16 00:01:03,440 --> 00:01:06,709 S1: and mental models to bring not just the news, but 17 00:01:06,709 --> 00:01:13,580 S1: why it matters and how to respond. All right. Welcome 18 00:01:13,580 --> 00:01:17,480 S1: to unsupervised Learning. This is Daniel, episode 432. Can you 19 00:01:17,480 --> 00:01:20,480 S1: summarize your work in a sentence? All right. Lots of 20 00:01:20,480 --> 00:01:24,440 S1: stuff here. Got a new fabric pattern called git wow 21 00:01:24,440 --> 00:01:28,220 S1: per minute. This is basically like an estimation of the 22 00:01:28,220 --> 00:01:31,670 S1: value density of any piece of content. So what you 23 00:01:31,670 --> 00:01:34,490 S1: do is you essentially take the content and you send 24 00:01:34,490 --> 00:01:38,150 S1: it to this fabric pattern, and it'll give you a 25 00:01:38,150 --> 00:01:40,640 S1: score from 0 to 10. And if it's anything above 26 00:01:40,640 --> 00:01:44,509 S1: like a five, it's got really it's got decent value. Okay. 27 00:01:44,510 --> 00:01:47,930 S1: So 5 to 7 is like decent value and then 28 00:01:47,930 --> 00:01:52,040 S1: like eight, 9 or 10, 8 or 9 even is 29 00:01:52,040 --> 00:01:56,630 S1: like top. I've not seen a ten yet. So just 30 00:01:56,630 --> 00:01:59,030 S1: consider nine to be like the top. So it's like 31 00:01:59,030 --> 00:02:01,220 S1: how much value am I going to get for this thing? 32 00:02:01,220 --> 00:02:04,370 S1: And basically what it's doing is it's looking at novelty, 33 00:02:04,370 --> 00:02:10,190 S1: it's looking at insights, ideas, novelty, surprise and wisdom and 34 00:02:10,190 --> 00:02:13,220 S1: basically collecting all those together and saying those are all 35 00:02:13,220 --> 00:02:16,730 S1: instances of value. And then it's kind of putting all 36 00:02:16,730 --> 00:02:19,880 S1: those together into a score between 0 and 10, and 37 00:02:19,880 --> 00:02:22,460 S1: it is really working well. I put it through some 38 00:02:22,460 --> 00:02:24,170 S1: stuff that I knew was going to be weak sauce, 39 00:02:24,169 --> 00:02:26,120 S1: and it scored like a two or a three, and 40 00:02:26,120 --> 00:02:28,130 S1: I put it through a couple things that I knew 41 00:02:28,160 --> 00:02:31,040 S1: were super high value, and they scored eights and nines. 42 00:02:31,040 --> 00:02:34,190 S1: So highly recommend this if you want to like, test 43 00:02:34,190 --> 00:02:38,299 S1: your gut or use it programmatically for something to rate value. 44 00:02:38,300 --> 00:02:41,210 S1: And importantly, it's doing it value per time, right? So 45 00:02:41,210 --> 00:02:44,960 S1: that was key to me. So OpenAI did some cool 46 00:02:44,960 --> 00:02:49,280 S1: stuff on Monday. So new model the O stands for 47 00:02:49,280 --> 00:02:52,580 S1: Omni GPT four O. It's not a zero. It's not 48 00:02:52,580 --> 00:02:56,600 S1: GPT 40. And it's about as smart as four, but 49 00:02:56,600 --> 00:03:00,019 S1: four times faster and twice as cheap. So half the cost. 50 00:03:00,020 --> 00:03:04,070 S1: And the big thing is that it has vision capabilities 51 00:03:04,070 --> 00:03:07,520 S1: and better audio. And this is not fully released yet, 52 00:03:07,520 --> 00:03:09,830 S1: but what it's going to allow you to do is 53 00:03:09,830 --> 00:03:13,669 S1: basically just talk with it like a normal person. And 54 00:03:13,669 --> 00:03:17,030 S1: even better, they released a desktop app which will soon 55 00:03:17,030 --> 00:03:21,440 S1: have the capability of monitoring your screen. So I could 56 00:03:21,440 --> 00:03:23,870 S1: be like scrolling through this right now and just like 57 00:03:23,870 --> 00:03:26,810 S1: looking at stuff and it'll be like, oh yeah, so 58 00:03:26,810 --> 00:03:30,140 S1: here's this. Don't forget to mention this. I could say, 59 00:03:30,139 --> 00:03:32,330 S1: give me a summary of what's on the screen right now, 60 00:03:32,330 --> 00:03:35,600 S1: and it'll just talk to me. And even cooler is like, 61 00:03:35,600 --> 00:03:38,120 S1: if I'm looking at a data visualization or some math 62 00:03:38,120 --> 00:03:41,030 S1: or something, it could, like, walk me through that math. 63 00:03:41,030 --> 00:03:45,080 S1: It's just unbelievable. I've seen a couple of demos of this. 64 00:03:45,080 --> 00:03:47,240 S1: I'm not sure if they were internally built or if 65 00:03:47,240 --> 00:03:50,150 S1: they were third parties. Actually, I saw one by Khan 66 00:03:50,150 --> 00:03:53,750 S1: from the Khan Academy and he was walking his son 67 00:03:53,750 --> 00:03:58,100 S1: through like teaching him geometry. It's literally you're just talking 68 00:03:58,100 --> 00:04:00,950 S1: to this thing. And what's really crazy about this release 69 00:04:00,950 --> 00:04:04,460 S1: is that this is going to be for free users 70 00:04:04,460 --> 00:04:08,000 S1: as well. So they opened GPT four to free users, 71 00:04:08,000 --> 00:04:12,410 S1: which is absolutely incredible. And I'm just super excited about 72 00:04:12,410 --> 00:04:16,310 S1: this whole thing. It's it's really, really exciting. Again, one 73 00:04:16,310 --> 00:04:18,320 S1: of the things I've been talking about here with AI 74 00:04:18,320 --> 00:04:23,779 S1: stuff is that it's extraordinary when AI is used to 75 00:04:23,779 --> 00:04:28,130 S1: do things that everybody needs, but very few people have 76 00:04:28,130 --> 00:04:33,170 S1: access to, so that's creating movies, that's creating ideas, writing books, 77 00:04:33,170 --> 00:04:36,290 S1: but it's also things like getting a mole looked at, 78 00:04:36,290 --> 00:04:41,420 S1: most importantly, education. I mean, the education use case for 79 00:04:41,420 --> 00:04:46,429 S1: AI currently is insane. And it's really, really incredible when 80 00:04:46,430 --> 00:04:51,380 S1: you start doing this for Zero Vibe, where one they 81 00:04:51,380 --> 00:04:53,900 S1: made the whole model free. I'm not sure what the 82 00:04:53,900 --> 00:04:57,230 S1: usage is going to be for that. Hopefully it'll be decent, 83 00:04:57,230 --> 00:05:00,770 S1: but not only do they make it free, but it's interactive. 84 00:05:00,770 --> 00:05:04,279 S1: So you could literally just say teach me calculus, teach 85 00:05:04,279 --> 00:05:09,140 S1: me this infinite patience. It doesn't get mad at you. 86 00:05:09,140 --> 00:05:11,930 S1: In fact, it now detects emotions, so if it hears 87 00:05:11,930 --> 00:05:14,270 S1: you getting frustrated, it's going to be like, hey, let's 88 00:05:14,270 --> 00:05:16,729 S1: slow down. Let me explain it in a different way. 89 00:05:16,730 --> 00:05:20,720 S1: Let me use an analogy that is unbelievable. We're talking 90 00:05:20,720 --> 00:05:25,850 S1: about massively multiplying the percentage of people on this planet 91 00:05:25,850 --> 00:05:29,780 S1: who are educated because they have access to an extremely 92 00:05:29,779 --> 00:05:36,529 S1: high quality superhuman intelligence, superhuman patience, tutor everyone in the 93 00:05:36,529 --> 00:05:39,169 S1: world having access to this. And I'm not sure everyone 94 00:05:39,170 --> 00:05:42,380 S1: in the world has access to GPT yet. And of course, 95 00:05:42,380 --> 00:05:45,710 S1: you need a computer and you need internet. So that's 96 00:05:45,710 --> 00:05:50,960 S1: not exactly true, but the barrier is massively lower as 97 00:05:50,960 --> 00:05:56,000 S1: a result of this. It's just unbelievably positive and awesome. 98 00:05:56,000 --> 00:05:58,820 S1: And obviously that doesn't take away any of the negatives 99 00:05:58,820 --> 00:06:02,300 S1: or the dangers or the alignment problem or superintelligence. Like 100 00:06:02,300 --> 00:06:06,560 S1: we've got issues coming up, no question. But the positives 101 00:06:06,560 --> 00:06:11,620 S1: are just undeniable. And this model becoming. Free. I mean, 102 00:06:11,620 --> 00:06:19,630 S1: and having these interactive conversation capabilities and emotion detection. Just unbelievable. 103 00:06:19,630 --> 00:06:22,270 S1: I predicted there would be more agent stuff, which I 104 00:06:22,270 --> 00:06:25,810 S1: think that will be coming soon. I still believe in 105 00:06:25,810 --> 00:06:29,650 S1: these predictions, not so much about like what OpenAI specifically 106 00:06:29,650 --> 00:06:33,190 S1: is going to do. I'm making broader predictions about what 107 00:06:33,190 --> 00:06:37,390 S1: OpenAI and anthropic and probably Google and all other people 108 00:06:37,390 --> 00:06:41,830 S1: will do, which is essentially moving prompts into or no 109 00:06:41,830 --> 00:06:45,340 S1: moving agents into prompts. So basically, prompting is the most 110 00:06:45,339 --> 00:06:48,669 S1: natural way to interact with an AI, and I believe 111 00:06:48,670 --> 00:06:52,720 S1: that it's the most natural place to build agents as well. 112 00:06:52,720 --> 00:06:55,479 S1: In fact, you'll just describe what you want. Like I 113 00:06:55,480 --> 00:06:58,300 S1: want a team of 100 people creating ideas. I want 114 00:06:58,300 --> 00:07:01,690 S1: a team of ten people filtering through those ideas to 115 00:07:01,690 --> 00:07:03,850 S1: figure out the best ones. In fact, you won't even 116 00:07:03,850 --> 00:07:06,610 S1: have to say 100 or 10. It'll just figure out 117 00:07:06,610 --> 00:07:09,219 S1: based on how much money you have to spend, based 118 00:07:09,220 --> 00:07:11,920 S1: on how much time you have and the time constraints 119 00:07:11,920 --> 00:07:15,700 S1: and your deadline or whatever. It'll spin up more, spin 120 00:07:15,700 --> 00:07:20,590 S1: up more or fewer based on all those constraints, right? 121 00:07:20,590 --> 00:07:23,020 S1: So when you say, I need a team of people 122 00:07:23,020 --> 00:07:24,790 S1: to do this, a team of people to do that, 123 00:07:24,790 --> 00:07:26,800 S1: team of people to do this, then I need the 124 00:07:26,800 --> 00:07:32,020 S1: final result to be rated and tested and validated. It'll 125 00:07:32,020 --> 00:07:35,560 S1: just go and build all those pieces, which are actually 126 00:07:35,560 --> 00:07:39,280 S1: teams of agents all collaborating together. And all you had 127 00:07:39,280 --> 00:07:41,590 S1: to do was put that in the prompt. And that's 128 00:07:41,590 --> 00:07:44,739 S1: why I think prompting is everything. It's an article I 129 00:07:44,740 --> 00:07:49,480 S1: wrote earlier called prompting. Most of AI is prompting or 130 00:07:49,690 --> 00:07:51,400 S1: something like that a couple of weeks ago or a 131 00:07:51,400 --> 00:07:54,730 S1: week ago. So I would check that one out. And yeah, 132 00:07:54,730 --> 00:07:58,120 S1: they got a desktop app. And yeah, they basically created 133 00:07:58,120 --> 00:08:02,200 S1: digital assistants, which I've got a book out from 2016 134 00:08:02,200 --> 00:08:04,840 S1: that was basically talking about this. And it's also in 135 00:08:04,840 --> 00:08:08,860 S1: the Predictable Path video, which you might have seen, which 136 00:08:08,860 --> 00:08:12,640 S1: is kind of like an updated visual version of that book, 137 00:08:12,640 --> 00:08:14,530 S1: but with lots of new stuff, because I didn't know 138 00:08:14,530 --> 00:08:18,910 S1: about Gen AI back in 2016, obviously, but they basically 139 00:08:18,910 --> 00:08:22,630 S1: built her. Not only did they built her from the movie, 140 00:08:22,630 --> 00:08:26,050 S1: but they're also using for the demos the voice that 141 00:08:26,050 --> 00:08:30,790 S1: I use for my personal AI with OpenAI, which is Skylar, 142 00:08:30,790 --> 00:08:35,920 S1: which is the voice of Scarlett Johansson, essentially. I doubt 143 00:08:35,920 --> 00:08:37,840 S1: that it actually is, but maybe it is. Maybe she 144 00:08:37,840 --> 00:08:41,650 S1: got paid for this. RSA was fantastic. Really good. Caught 145 00:08:41,650 --> 00:08:43,750 S1: up with a lot of people, did a couple of 146 00:08:43,750 --> 00:08:47,679 S1: talks and panels. Probably should have brought something to sell. 147 00:08:47,679 --> 00:08:50,980 S1: I didn't really like go in there trying to sell anything, 148 00:08:50,980 --> 00:08:53,440 S1: which I guess I should have. But I did talk 149 00:08:53,440 --> 00:08:57,070 S1: a little bit about threshold, so that was cool. The 150 00:08:57,070 --> 00:09:03,100 S1: energy and optimism around RSA was extraordinary. I think this 151 00:09:03,100 --> 00:09:05,500 S1: is largely in part to AI, and there were tons 152 00:09:05,500 --> 00:09:08,560 S1: of products talking about how much they're using AI and 153 00:09:08,559 --> 00:09:11,980 S1: everything like that. So we've all made those jokes already. 154 00:09:12,160 --> 00:09:17,830 S1: But speaking of threshold, holy crap, I am absolutely in 155 00:09:17,830 --> 00:09:21,189 S1: love with this product. Okay, I am in love with 156 00:09:21,190 --> 00:09:23,950 S1: this product. And yes, it is a paid product and 157 00:09:23,950 --> 00:09:27,220 S1: it is my product. But look at this. This is 158 00:09:27,220 --> 00:09:29,679 S1: my current feed right now. I didn't plan on showing this, 159 00:09:29,679 --> 00:09:32,170 S1: but whatever. So how do I blow this up? Yeah. 160 00:09:32,170 --> 00:09:34,390 S1: So how to cope with the fear of aging? I 161 00:09:34,390 --> 00:09:37,660 S1: haven't looked at that one yet. 2025 models will be 162 00:09:37,660 --> 00:09:41,920 S1: more like coworkers than than search engines. Again, this I 163 00:09:41,920 --> 00:09:45,400 S1: did not explicitly go follow this person. What is this? 164 00:09:45,400 --> 00:09:46,780 S1: This is. Uh dwarkesh. 165 00:09:46,780 --> 00:09:49,930 S2: I think in 1 or 2 years we'll find that 166 00:09:49,929 --> 00:09:50,410 S2: you can use. 167 00:09:50,410 --> 00:09:53,980 S1: So this is Dwarkesh Patel, who's doing these interviews, and 168 00:09:53,980 --> 00:09:57,699 S1: he's talking to really smart people about AI or whatever. 169 00:09:57,700 --> 00:10:00,940 S1: He's talking about lots of different things. But this particular 170 00:10:00,940 --> 00:10:05,979 S1: conversation is about coworkers and search engines and AGI and 171 00:10:05,980 --> 00:10:10,210 S1: like all this super interesting topics. But here's the thing, okay? 172 00:10:10,210 --> 00:10:12,790 S1: In order to show this to me, it came into 173 00:10:12,790 --> 00:10:15,940 S1: the top of threshold and all my different AI stuff 174 00:10:15,940 --> 00:10:19,450 S1: ran against it to assess it and determine what the 175 00:10:19,450 --> 00:10:22,990 S1: content was and how high quality the content was, and 176 00:10:22,990 --> 00:10:26,650 S1: it had to be above a certain threshold in order 177 00:10:26,650 --> 00:10:28,810 S1: for it to even show up. Now I have it 178 00:10:28,809 --> 00:10:31,449 S1: set to 50, which is the minimum because I want 179 00:10:31,450 --> 00:10:33,339 S1: to see a lot of content, but I can move 180 00:10:33,340 --> 00:10:36,189 S1: this up and I could turn off a lot of 181 00:10:36,190 --> 00:10:38,740 S1: these things here. This is the first time I've actually 182 00:10:38,740 --> 00:10:43,059 S1: showing threshold kind of interesting. So I can move those 183 00:10:43,059 --> 00:10:45,370 S1: and I can say save and then look at this 184 00:10:45,610 --> 00:10:48,910 S1: completely change the feed. But now it has to be 185 00:10:48,910 --> 00:10:52,540 S1: above an 80 and it has to be in these 186 00:10:52,540 --> 00:10:58,000 S1: categories or these categories have to apply. So just I 187 00:10:58,000 --> 00:11:02,410 S1: mean every single thing listen to me, every single thing 188 00:11:02,410 --> 00:11:05,650 S1: I click on here, I love I absolutely love it. 189 00:11:05,650 --> 00:11:09,219 S1: And I don't have to stress about oh did I 190 00:11:09,220 --> 00:11:12,230 S1: watch all the. Tests, from leks and from Huberman and 191 00:11:12,230 --> 00:11:15,290 S1: all these different people and all these different I people. No, 192 00:11:15,290 --> 00:11:17,360 S1: I just put them into the top of the funnel. 193 00:11:17,360 --> 00:11:20,060 S1: I put them into the top of the funnel for threshold, 194 00:11:20,059 --> 00:11:23,660 S1: and the whole I pipeline runs against it and I 195 00:11:23,660 --> 00:11:26,630 S1: get the analysis. Oh, and I forgot to mention watch this. Okay, 196 00:11:26,630 --> 00:11:29,030 S1: this is the sickest video. I want to talk about 197 00:11:29,030 --> 00:11:32,750 S1: this anyway. So this is the sickest business video that 198 00:11:32,750 --> 00:11:36,170 S1: I've ever seen and I forget how long it is. 199 00:11:36,170 --> 00:11:38,210 S1: Soon we're going to have duration things in here. But 200 00:11:38,210 --> 00:11:40,760 S1: watch this. If I go click on the video, obviously 201 00:11:40,760 --> 00:11:41,360 S1: it's going to open up. 202 00:11:41,360 --> 00:11:43,700 S3: In the engineering room and said these. 203 00:11:43,700 --> 00:11:46,790 S1: Basically this guy Alex talking about stuff for an hour 204 00:11:46,790 --> 00:11:50,570 S1: and 29 minutes okay. It is fantastic. But watch this. 205 00:11:50,570 --> 00:11:53,900 S1: I could go like this. This is a major feature 206 00:11:53,900 --> 00:11:57,229 S1: of threshold that no other platform has. You click on 207 00:11:57,230 --> 00:12:00,560 S1: it and it tells you a summary of it, the 208 00:12:00,559 --> 00:12:03,830 S1: ideas and a recommendation, and it gives you a review 209 00:12:03,830 --> 00:12:05,570 S1: of it so you don't even have to go watch 210 00:12:05,570 --> 00:12:07,849 S1: it if you don't want. But watch this. That was 211 00:12:07,850 --> 00:12:10,850 S1: surface level. Watch this. If I go to middle level, 212 00:12:10,850 --> 00:12:14,240 S1: gives me a deeper summary. It breaks out more ideas 213 00:12:14,240 --> 00:12:17,390 S1: and more recommendations. And if I go to deep level 214 00:12:17,390 --> 00:12:23,390 S1: auto updates, summary, core ideas, core recommendations. Unbelievable. I mean, 215 00:12:23,390 --> 00:12:26,780 S1: I've always wanted this and that's why we actually created it. 216 00:12:26,780 --> 00:12:30,199 S1: I've always wanted this as a tool and the stuff 217 00:12:30,200 --> 00:12:32,960 S1: that we're about to be adding to this, the I'm 218 00:12:32,960 --> 00:12:35,390 S1: not even going to talk about that yet, but insane 219 00:12:35,390 --> 00:12:38,630 S1: stuff along the lines of everything I've been talking about 220 00:12:38,630 --> 00:12:41,120 S1: for this last year and a half with with AI 221 00:12:41,120 --> 00:12:43,940 S1: content rating, all that stuff like the new features that 222 00:12:43,940 --> 00:12:46,370 S1: are coming are just going to be insane. And again, 223 00:12:46,370 --> 00:12:49,040 S1: the guiding light that I'm using here, like I haven't 224 00:12:49,040 --> 00:12:52,309 S1: raised any VC money. I'm not worried about investors. I'm 225 00:12:52,309 --> 00:12:54,949 S1: worried about making the best possible tool for me. And 226 00:12:54,950 --> 00:12:58,100 S1: this is the tool that I've always needed, and that's 227 00:12:58,100 --> 00:13:00,620 S1: why it exists. And of course, when people are like, hey, 228 00:13:00,620 --> 00:13:02,960 S1: I wish I had this. Well, if it's something that 229 00:13:02,960 --> 00:13:05,870 S1: really resonates with me and I think everyone's going to love, 230 00:13:05,870 --> 00:13:07,790 S1: then yeah, we put that into the pipeline and we 231 00:13:07,790 --> 00:13:11,090 S1: actually build that. But ultimately it's it's a thing that 232 00:13:11,090 --> 00:13:14,390 S1: solves a problem that I've had of collecting as much 233 00:13:14,390 --> 00:13:18,229 S1: information as possible, as high quality as possible, but at 234 00:13:18,230 --> 00:13:21,380 S1: a threshold level that I could control the flow. So 235 00:13:21,380 --> 00:13:23,120 S1: I could basically go in here and say, you know what, 236 00:13:23,120 --> 00:13:25,309 S1: I need it to be a 90 or a higher, 237 00:13:25,309 --> 00:13:28,309 S1: and I want it to be only about AI. So 238 00:13:28,309 --> 00:13:32,120 S1: watch this. And that pulls it down even further. It's 239 00:13:32,120 --> 00:13:36,410 S1: just it's just unbelievably useful. So that's my pitch for that. 240 00:13:36,410 --> 00:13:40,280 S1: It is a paid thing. So yeah. Whatever. Is that 241 00:13:40,280 --> 00:13:42,500 S1: even called a sponsor post. Is it a sponsor if 242 00:13:42,500 --> 00:13:44,990 S1: it's your own product? I have no idea. But anyway, 243 00:13:44,990 --> 00:13:47,780 S1: highly worth it. I would pay thousands of dollars per 244 00:13:47,780 --> 00:13:51,920 S1: year to buy this product. No joke, I absolutely would. 245 00:13:51,920 --> 00:13:54,470 S1: Oh yeah, and this is the email. So I have 246 00:13:54,470 --> 00:13:56,780 S1: it set to a daily email. So you get this 247 00:13:56,809 --> 00:13:59,720 S1: email that basically shows you the stuff. Because sometimes like 248 00:13:59,720 --> 00:14:01,610 S1: I'll forget to check, I'll just be super busy in 249 00:14:01,610 --> 00:14:04,190 S1: meetings or whatever, and I'll be super busy. And then 250 00:14:04,190 --> 00:14:06,410 S1: at the end of the day I get this, look 251 00:14:06,410 --> 00:14:08,990 S1: at these. How strong is AI as a tool, as 252 00:14:08,990 --> 00:14:12,800 S1: a successor? Deferred happiness syndrome? This thing is unbelievable. And 253 00:14:12,800 --> 00:14:15,050 S1: then I click on view right here and it goes 254 00:14:15,050 --> 00:14:17,179 S1: yeah this is the business truth one. That's how I 255 00:14:17,179 --> 00:14:19,940 S1: found that video. Also from what's it called. This is 256 00:14:19,940 --> 00:14:23,060 S1: from Peter Attia on VO2 max and muscle mass. Anyway 257 00:14:23,060 --> 00:14:26,690 S1: it's it's fantastic. I just love it. I got a 258 00:14:26,690 --> 00:14:32,180 S1: sponsored conversation had with BlackBerry and Cary Ransom about maritime security. 259 00:14:32,180 --> 00:14:35,720 S1: My dad just went on a podcast in, uh, in 260 00:14:35,720 --> 00:14:38,600 S1: a studio in San Francisco. I was actually giving a 261 00:14:38,600 --> 00:14:41,060 S1: talk a couple of blocks away. My dad's in the 262 00:14:41,060 --> 00:14:45,170 S1: studio talking about his music, showing his music, just really, 263 00:14:45,170 --> 00:14:47,510 S1: really cool. And I got a link here to the 264 00:14:47,510 --> 00:14:50,690 S1: full show, which is this one or no, this one. 265 00:14:50,690 --> 00:14:53,360 S1: And then this one is him playing one of my 266 00:14:53,360 --> 00:14:56,990 S1: favorite songs that he's ever written called children of the nights, 267 00:14:56,990 --> 00:15:00,680 S1: and you definitely want to check that one out. It's fantastic. 268 00:15:00,680 --> 00:15:04,130 S1: And security Dell got hacked real bad through an API. 269 00:15:04,130 --> 00:15:06,620 S1: I think if I could invest in any sort of 270 00:15:06,620 --> 00:15:11,540 S1: legacy non-ai tech, which I guess it's quite AI related 271 00:15:11,540 --> 00:15:13,310 S1: at this point. And that's kind of the point, is 272 00:15:13,310 --> 00:15:17,390 S1: that API security is like everything, because everything is becoming 273 00:15:17,390 --> 00:15:22,550 S1: APIs that get wielded by AI, and your company is 274 00:15:22,550 --> 00:15:26,810 S1: essentially about to be mostly your API. So you've got 275 00:15:26,810 --> 00:15:28,760 S1: to lock that down. So I feel like of all 276 00:15:28,760 --> 00:15:31,430 S1: the traditional security spaces that I would want to be 277 00:15:31,430 --> 00:15:33,260 S1: in right now, which I don't want to be in 278 00:15:33,260 --> 00:15:35,180 S1: any of them, but if I did, it would probably 279 00:15:35,180 --> 00:15:38,630 S1: be API security. And speaking of that, my buddy Joseph 280 00:15:38,630 --> 00:15:41,720 S1: Thacker has put out a couple of different pieces about 281 00:15:41,720 --> 00:15:47,390 S1: API security assumptions about AI, and also one about agents 282 00:15:47,390 --> 00:15:51,830 S1: and authentication. So check that out. Both of those and 283 00:15:51,830 --> 00:15:54,920 S1: CSA has a new alert system. They're basically allowing companies 284 00:15:54,920 --> 00:15:59,210 S1: to sign up. And then they send them Kev alerts 285 00:15:59,210 --> 00:16:02,510 S1: if they see something associated with their company. Attackers are 286 00:16:02,510 --> 00:16:07,220 S1: using Microsoft Graphs API for malware comms. I love these 287 00:16:07,220 --> 00:16:11,270 S1: side channel things, I just. Absolutely love them where you could, like, 288 00:16:11,270 --> 00:16:14,840 S1: use something in a way that's not expected. The best 289 00:16:14,840 --> 00:16:19,100 S1: example of this that I remember is Gmail drafts. You actually, 290 00:16:19,100 --> 00:16:22,400 S1: you you have a shared email account. You create a 291 00:16:22,400 --> 00:16:25,910 S1: draft on one side, you don't send the email, and 292 00:16:25,910 --> 00:16:28,700 S1: then someone else has access to that email account and 293 00:16:28,700 --> 00:16:32,030 S1: they just log in and read the draft so the 294 00:16:32,030 --> 00:16:35,450 S1: email never gets sent. So that's another example of just 295 00:16:35,450 --> 00:16:39,050 S1: like using a a platform in a way that's sneaky. 296 00:16:39,080 --> 00:16:43,430 S1: All right. So one password. Thank you for sponsoring. That's 297 00:16:43,430 --> 00:16:50,240 S1: Collidin one password and Russian influence campaign looking at magnifying 298 00:16:50,240 --> 00:16:55,190 S1: and taking advantage of the campus protests. Google's making MFA 299 00:16:55,190 --> 00:16:59,150 S1: setup smoother by letting you skip phone number for options 300 00:16:59,150 --> 00:17:03,109 S1: like authenticator or keys. Marines are testing robot dogs with 301 00:17:03,110 --> 00:17:10,040 S1: AI aimed rifles. Awesome awesome awesome awesome 95% of international 302 00:17:10,040 --> 00:17:14,179 S1: data travels through undersea cables. Okay, so this is one 303 00:17:14,180 --> 00:17:19,909 S1: that's really interesting. I've never understood why undersea cables weren't 304 00:17:19,910 --> 00:17:23,450 S1: more targeted and just like totally abused by terrorists. I 305 00:17:23,450 --> 00:17:26,209 S1: guess there is the small matter of like you're in 306 00:17:26,210 --> 00:17:29,540 S1: a boat, you're way, way above this cable. It takes 307 00:17:29,540 --> 00:17:32,450 S1: time and money and skill to get down there. So 308 00:17:32,450 --> 00:17:35,720 S1: I guess that's a barrier. But once you're down there, 309 00:17:35,720 --> 00:17:39,500 S1: Holy crap. I mean, it seems pretty easy to do 310 00:17:39,500 --> 00:17:42,590 S1: damage to these things, and the amount of disruption you 311 00:17:42,590 --> 00:17:47,359 S1: could do if you're disrupting international internet traffic is is 312 00:17:47,359 --> 00:17:51,650 S1: pretty serious. Plus, you think this has got to get easier. Like, 313 00:17:51,650 --> 00:17:54,949 S1: if you can send drones down there like underwater water, 314 00:17:54,950 --> 00:18:00,140 S1: drones with like explosives, this wouldn't be easy for low 315 00:18:00,140 --> 00:18:03,710 S1: level attackers to do. But for a state actor to 316 00:18:03,710 --> 00:18:08,060 S1: want to just continuously like send drones after it or 317 00:18:08,060 --> 00:18:11,330 S1: teams or whatever, it seems so easy for any state 318 00:18:11,330 --> 00:18:18,320 S1: actor to basically sever internet communications through undersea cables. And 319 00:18:18,320 --> 00:18:22,940 S1: the US just closed another door essentially for Huawei getting 320 00:18:22,940 --> 00:18:28,129 S1: AI stuff with Intel chips. So another attack against the 321 00:18:28,130 --> 00:18:33,980 S1: Chinese trying to get AI chips. And someone from Andreessen 322 00:18:33,980 --> 00:18:37,850 S1: Horowitz basically says half a Google staff or just pretending 323 00:18:37,880 --> 00:18:40,490 S1: to work, they're not actually doing anything. Yeah, and a 324 00:18:40,490 --> 00:18:43,940 S1: lot of people are talking about this trend of fake work, 325 00:18:43,940 --> 00:18:46,639 S1: and I think it's a giant mess. It's fake in 326 00:18:46,640 --> 00:18:49,850 S1: multiple ways. Right. So I talked about David Graeber's bullshit 327 00:18:49,850 --> 00:18:52,910 S1: Jobs book. Really, really good. Basically, we have all these 328 00:18:52,910 --> 00:18:56,150 S1: jobs that shouldn't exist. You have all these people with 329 00:18:56,150 --> 00:19:03,200 S1: massive salaries, hardly doing anything and somehow being justified, especially 330 00:19:03,200 --> 00:19:05,869 S1: in Google where like they're not making much new stuff 331 00:19:05,869 --> 00:19:10,450 S1: at all. Because they don't have a product management focused 332 00:19:10,450 --> 00:19:14,919 S1: product teams. They're just like engineering focused teams. It's a 333 00:19:14,920 --> 00:19:17,679 S1: total mess. And you add AI to this and it's 334 00:19:17,680 --> 00:19:21,010 S1: just it's going to be a total mess. The safest 335 00:19:21,010 --> 00:19:23,980 S1: thing to do, in my opinion, the safest place to 336 00:19:23,980 --> 00:19:28,480 S1: be is building new things, creating new things. Be a programmer, 337 00:19:28,480 --> 00:19:31,930 S1: definitely be a programmer. But more so be a thinker 338 00:19:31,930 --> 00:19:37,179 S1: who focuses on problems. Solve those problems. Have the tech skills. 339 00:19:37,180 --> 00:19:40,900 S1: Have the humanities skills. You want to be broad. You 340 00:19:40,900 --> 00:19:44,620 S1: want to be general. I recommend everyone has tech skills. 341 00:19:44,619 --> 00:19:49,150 S1: Everyone has programming skills. Everyone has AI skills. Everyone has 342 00:19:49,150 --> 00:19:53,350 S1: writing and speaking and presentation skills. These are like the universals, 343 00:19:53,350 --> 00:19:56,560 S1: and I'm actually working on a big post around this, 344 00:19:56,560 --> 00:19:59,470 S1: which is going to be a member post around this. 345 00:19:59,470 --> 00:20:01,870 S1: That's going to be super key of like how to 346 00:20:01,869 --> 00:20:05,830 S1: get ready for what's coming, what skills you need. So 347 00:20:05,830 --> 00:20:09,220 S1: that's pretty much a teaser for that. Mitra is partnering 348 00:20:09,220 --> 00:20:14,020 S1: with Nvidia to create a $20 million AI supercomputer to 349 00:20:14,020 --> 00:20:18,760 S1: to make US government operations more efficient, which is cool 350 00:20:18,760 --> 00:20:24,520 S1: and terrifying. Joe Biden is converting the Foxconn debacle, which 351 00:20:24,520 --> 00:20:28,870 S1: didn't go well in Wisconsin into a $3.3 billion Microsoft 352 00:20:28,869 --> 00:20:33,160 S1: AI center. I feel like this administration is doing well 353 00:20:33,190 --> 00:20:36,909 S1: on AI, both in countering China but also building ourselves up. 354 00:20:36,910 --> 00:20:41,169 S1: And the acquired podcast is basically like biographies of people, 355 00:20:41,170 --> 00:20:43,600 S1: except for it's for companies. So a lot of people 356 00:20:43,600 --> 00:20:46,090 S1: are super into that. In the Bay area, Biden is 357 00:20:46,090 --> 00:20:49,360 S1: quadrupling tariffs on Chinese EVs. These things are a real 358 00:20:49,359 --> 00:20:52,899 S1: threat to the American car industry. So he's going to 359 00:20:52,900 --> 00:20:58,210 S1: tariff them quite a bit. This one said 400% but quadrupling. Yeah, 360 00:20:58,210 --> 00:21:02,439 S1: this one said quadrupling. But I've seen it say doubling. 361 00:21:02,440 --> 00:21:05,320 S1: So I'm not sure the exact numbers. But he's definitely 362 00:21:05,320 --> 00:21:09,070 S1: increasing the tariffs. California is going to start charging your 363 00:21:09,070 --> 00:21:13,870 S1: bill for electricity based on your income starting in 2025. 364 00:21:13,960 --> 00:21:19,210 S1: Scientists have found all DNA and RNA bases in meteorites, 365 00:21:19,540 --> 00:21:23,919 S1: basically hinting that building blocks might be extraterrestrial. This is 366 00:21:23,920 --> 00:21:26,800 S1: not surprising in the slightest. To me, it just seems 367 00:21:26,800 --> 00:21:30,850 S1: logical and there's life floating around everywhere. I don't think 368 00:21:30,850 --> 00:21:36,670 S1: life is that special. In one characterization of special in 369 00:21:36,670 --> 00:21:40,179 S1: terms of special of like, oh, we're we're a snowflake 370 00:21:40,180 --> 00:21:42,100 S1: and we're the only ones who have it. I don't 371 00:21:42,100 --> 00:21:44,470 S1: think we're special in that way. That does not mean 372 00:21:44,470 --> 00:21:50,440 S1: it's not special in terms of having high, interesting, unique value. 373 00:21:50,440 --> 00:21:53,470 S1: I think we absolutely have. That doesn't mean it has 374 00:21:53,470 --> 00:21:56,140 S1: to be unique. Does it mean it has to be 375 00:21:56,140 --> 00:21:59,890 S1: one of a kind? Okay. Scientists constructed a one millimetre 376 00:21:59,890 --> 00:22:06,220 S1: square piece of the human cerebral cortex at a nanoscale resolution. 377 00:22:06,220 --> 00:22:09,939 S1: So they basically get to see the actual neurons and 378 00:22:09,940 --> 00:22:13,300 S1: I believe synapses as well. I haven't looked at this 379 00:22:13,300 --> 00:22:16,450 S1: document fully yet or this image fully yet. I'm not 380 00:22:16,450 --> 00:22:19,240 S1: sure I would even understand it, but I love the 381 00:22:19,240 --> 00:22:20,860 S1: fact that we could see that. I love the fact 382 00:22:20,859 --> 00:22:24,189 S1: that it's only one millimetre square and we're getting all that, 383 00:22:24,369 --> 00:22:28,180 S1: all those, you know, millions of different cells and neurons 384 00:22:28,180 --> 00:22:32,440 S1: and everything in there. The study says very large study, 385 00:22:32,470 --> 00:22:39,100 S1: 154 million deaths prevented due to vaccines. Streaming is basically cable. 386 00:22:39,100 --> 00:22:42,700 S1: Now there's a racket around emotional support animals. That was 387 00:22:42,700 --> 00:22:48,070 S1: quite hilarious and disturbing. And I think Mark Andreessen is 388 00:22:48,070 --> 00:22:54,369 S1: mistaken about paradox applying to AI. This this podcast here. 389 00:22:54,369 --> 00:22:57,970 S1: He basically says that AI will not make it easier 390 00:22:57,970 --> 00:23:02,770 S1: to build companies because of Jevons Paradox, which basically says 391 00:23:02,770 --> 00:23:07,540 S1: that you think if you make roads wider, it would 392 00:23:07,540 --> 00:23:10,240 S1: reduce the amount of traffic, but in fact it just 393 00:23:10,240 --> 00:23:14,140 S1: makes more people drive on them, which actually increases traffic. 394 00:23:14,140 --> 00:23:19,810 S1: Now Mark basically said, well, I think that means that 395 00:23:19,900 --> 00:23:23,830 S1: AI making it easier to start a company will make 396 00:23:23,830 --> 00:23:26,740 S1: it actually harder to start a company because more people 397 00:23:26,740 --> 00:23:29,770 S1: will be doing it and the standard will be higher. 398 00:23:29,770 --> 00:23:33,879 S1: And I think he is misunderstanding that. And so basically 399 00:23:33,880 --> 00:23:36,340 S1: what I say about that is it's only when it's 400 00:23:36,340 --> 00:23:41,650 S1: a resource. Jevons paradox applies to a resource. It basically 401 00:23:41,650 --> 00:23:46,510 S1: says if you make that thing cheaper, it won't reduce 402 00:23:46,510 --> 00:23:50,500 S1: the total amount of usage across the world because more 403 00:23:50,500 --> 00:23:53,409 S1: people will use it. So usage actually goes up. That's 404 00:23:53,410 --> 00:23:56,920 S1: the limit of the paradox, at least as I understand it. 405 00:23:56,920 --> 00:24:01,540 S1: And the situation with AI helping startups is actually a 406 00:24:01,540 --> 00:24:04,600 S1: friction issue. It's reducing friction to get into the market. 407 00:24:04,600 --> 00:24:07,130 S1: It's not about there's only so. Much of X or 408 00:24:07,130 --> 00:24:10,250 S1: so much of Y, and if it were, the thing 409 00:24:10,250 --> 00:24:13,400 S1: that would be talking about is there's only so much energy. 410 00:24:13,520 --> 00:24:17,000 S1: There's only so much compute. So that's a limited resource 411 00:24:17,000 --> 00:24:20,570 S1: that would actually apply to Jevons paradox, but not friction 412 00:24:20,570 --> 00:24:22,669 S1: of getting into a market. So I think he was 413 00:24:22,670 --> 00:24:25,850 S1: wrong about that. And that basically cleared that up. Courage 414 00:24:25,850 --> 00:24:29,960 S1: is everything I've been throwing around this three level system here. 415 00:24:29,960 --> 00:24:34,490 S1: Courage is action versus fear. Discipline is courage versus laziness. 416 00:24:34,490 --> 00:24:36,410 S1: You see how they build on each other. This one 417 00:24:36,410 --> 00:24:41,119 S1: is this one. And then success is discipline versus mediocrity. 418 00:24:41,119 --> 00:24:44,390 S1: So it goes from courage to discipline to success. So 419 00:24:44,390 --> 00:24:46,910 S1: everything you want is on the other side of courage, 420 00:24:46,910 --> 00:24:51,229 S1: whether it's courage against fear or courage against laziness. And 421 00:24:51,230 --> 00:24:53,930 S1: this applies to all sorts of real life situations. So 422 00:24:53,930 --> 00:24:58,430 S1: it's like hard conversations, you know, getting more healthy, eating right, 423 00:24:58,460 --> 00:25:01,939 S1: not wasting time doing the wrong things, or quitting a 424 00:25:01,940 --> 00:25:04,609 S1: soul crushing job. So if you're trying to quit a job, 425 00:25:04,609 --> 00:25:07,100 S1: it's like you're scared if you're eating right, it's a 426 00:25:07,100 --> 00:25:10,430 S1: matter of discipline becoming fit. It's a matter of discipline. 427 00:25:10,430 --> 00:25:12,530 S1: You got to go to the gym. You got to, 428 00:25:12,560 --> 00:25:15,500 S1: you know, eat right. You got to exercise. Not wasting 429 00:25:15,500 --> 00:25:18,470 S1: so much time with games. It also discipline, having a 430 00:25:18,470 --> 00:25:22,100 S1: hard conversation. That's courage. I guess what I'm saying here 431 00:25:22,100 --> 00:25:24,470 S1: is all of these can be thought of as courage. 432 00:25:24,470 --> 00:25:27,740 S1: And what I'm trying to do here is build a frame. 433 00:25:27,740 --> 00:25:29,570 S1: I want to build a frame that says when I'm 434 00:25:29,570 --> 00:25:32,930 S1: being lazy, I'm not being courageous. So I want to 435 00:25:32,930 --> 00:25:37,040 S1: frame this to myself as have the courage to not 436 00:25:37,040 --> 00:25:40,550 S1: be lazy, have courage to go to the gym. So 437 00:25:40,550 --> 00:25:44,570 S1: if I'm presenting that as there's somebody to go save 438 00:25:44,570 --> 00:25:47,149 S1: in a fire and I'm going to get burnt and 439 00:25:47,150 --> 00:25:51,530 S1: that inflames my my interest to be a moral person, well, 440 00:25:51,530 --> 00:25:54,200 S1: that can motivate me. I will walk through that fire 441 00:25:54,200 --> 00:25:57,380 S1: to go do that thing because that's what my identity is. 442 00:25:57,380 --> 00:26:00,080 S1: I want that to be my identity. And it is. 443 00:26:00,080 --> 00:26:03,620 S1: So that's the thing I want to do now. I 444 00:26:03,619 --> 00:26:06,740 S1: don't naturally, and most of us don't naturally have that 445 00:26:06,740 --> 00:26:09,320 S1: for going to the gym. It's not a moral issue. 446 00:26:09,320 --> 00:26:12,080 S1: I want to turn it into a moral issue. I 447 00:26:12,080 --> 00:26:14,389 S1: want to frame it as a moral issue so that 448 00:26:14,390 --> 00:26:16,490 S1: I'm like, look, oh, you don't want to go to 449 00:26:16,490 --> 00:26:20,149 S1: the gym. Okay, well, that's because I don't want to 450 00:26:20,150 --> 00:26:22,730 S1: use bad language, but I basically want to talk to 451 00:26:22,730 --> 00:26:26,000 S1: myself in this bad language of, you know, man up, 452 00:26:26,000 --> 00:26:29,209 S1: go do it, have the courage to do the right thing. 453 00:26:29,210 --> 00:26:32,510 S1: And I think that can massively help. Recommendation of the week. 454 00:26:32,510 --> 00:26:36,380 S1: Know yourself. Think about what someone should say when they 455 00:26:36,380 --> 00:26:41,060 S1: introduce you. So this is Sarah. Sarah does this do 456 00:26:41,060 --> 00:26:43,610 S1: you want them to say about Sarah? What do you 457 00:26:43,609 --> 00:26:46,730 S1: want them to say about you? Mine is something like 458 00:26:46,730 --> 00:26:49,070 S1: he has a company that builds products and services that 459 00:26:49,070 --> 00:26:53,780 S1: help people transition to something called human 3.0, so they 460 00:26:53,780 --> 00:26:58,700 S1: can survive what's happening and thrive with what's happening with AI. 461 00:26:58,730 --> 00:27:02,149 S1: It's something like that, and it's different for different audiences. 462 00:27:02,150 --> 00:27:04,699 S1: But that's the basic vibe. And you basically want to 463 00:27:04,700 --> 00:27:08,150 S1: know what this is for you, not only so that 464 00:27:08,150 --> 00:27:10,700 S1: you can deliver it, but that other people can do 465 00:27:10,700 --> 00:27:13,129 S1: it when you're not there. And the aphorism for the 466 00:27:13,130 --> 00:27:17,750 S1: week where your fear is, there's your task. Where your 467 00:27:17,750 --> 00:27:24,200 S1: fear is, there is your task. Carl Jung. Unsupervised Learning 468 00:27:24,200 --> 00:27:26,840 S1: is produced and edited by Daniel Meisler on a Neumann 469 00:27:26,840 --> 00:27:31,430 S1: U87 AI microphone using Hindenburg. Intro and outro music is 470 00:27:31,430 --> 00:27:34,669 S1: by zombie with the why and to get the text 471 00:27:34,670 --> 00:27:36,830 S1: and links from this episode, sign up for the newsletter 472 00:27:36,830 --> 00:27:42,199 S1: version of the show at Daniel meisler.com/newsletter. We'll see you 473 00:27:42,200 --> 00:27:42,830 S1: next time.