1 00:00:01,230 --> 00:00:04,500 S1: Welcome to Unsupervised Learning, a security, AI and meaning focused 2 00:00:04,500 --> 00:00:07,380 S1: podcast that looks at how best to thrive as humans 3 00:00:07,380 --> 00:00:11,610 S1: in a post AI world. It combines original ideas, analysis, 4 00:00:11,610 --> 00:00:14,820 S1: and mental models to bring not just the news, but 5 00:00:14,820 --> 00:00:21,939 S1: why it matters and how to respond. All right. Welcome 6 00:00:21,940 --> 00:00:25,150 S1: to supervised learning. This is Daniel Missler. All right. In 7 00:00:25,150 --> 00:00:28,630 S1: this one. What have we got here. This is episode 429. 8 00:00:28,780 --> 00:00:34,450 S1: So llama three came out last week and it was 9 00:00:34,450 --> 00:00:38,230 S1: super impressive. I'm getting more and more use out of it, actually. 10 00:00:38,229 --> 00:00:41,379 S1: I'm finding it's doing a lot of stuff as well 11 00:00:41,380 --> 00:00:46,540 S1: as GPT four, and sometimes it just completely fails. And 12 00:00:46,540 --> 00:00:49,600 S1: it's worse than like GPT three. So I think that's 13 00:00:49,600 --> 00:00:52,210 S1: the case with all these different models. Like, you never know, 14 00:00:52,210 --> 00:00:54,280 S1: like how good it's going to be at certain things 15 00:00:54,280 --> 00:00:56,410 S1: and how bad it's going to be at other things. 16 00:00:56,410 --> 00:00:59,410 S1: And what I really hope we get before too long 17 00:00:59,410 --> 00:01:03,790 S1: is like this really robust testing framework. So we know 18 00:01:03,790 --> 00:01:06,580 S1: what things are good at and what they're not good at. 19 00:01:06,580 --> 00:01:08,980 S1: And we could just have this continuous testing because a 20 00:01:08,980 --> 00:01:12,490 S1: lot of the benchmarks that are used to test, you know, 21 00:01:12,490 --> 00:01:15,400 S1: to show the quality of models, I feel like people 22 00:01:15,400 --> 00:01:18,250 S1: are actually building their models to score good on them, 23 00:01:18,459 --> 00:01:24,640 S1: and they're pretty useless. There's also this cool bolt AI application, 24 00:01:24,640 --> 00:01:27,190 S1: which this link is actually bad, I think. I think 25 00:01:27,190 --> 00:01:31,360 S1: this is the wrong thing, but if we do bolt AI, 26 00:01:31,630 --> 00:01:35,260 S1: this should actually pull us up into it. Yeah, this 27 00:01:35,260 --> 00:01:37,720 S1: is the one. So this is a pretty cool interface 28 00:01:37,720 --> 00:01:41,920 S1: for actually dealing with working with LMS. It looks almost 29 00:01:41,920 --> 00:01:45,910 S1: like ChatGPT, but it's a standalone app and you could 30 00:01:45,910 --> 00:01:48,970 S1: use all your different models, right? So you got like 31 00:01:49,000 --> 00:01:52,240 S1: you could use GPT four, you could use local models, 32 00:01:52,390 --> 00:01:54,310 S1: you could put in your own model. This is just 33 00:01:54,310 --> 00:01:58,000 S1: really good. And see here you got different commands AI 34 00:01:58,000 --> 00:02:01,360 S1: and line. Yeah, it's just really powerful. I like this 35 00:02:01,360 --> 00:02:05,080 S1: little lap here. So that's one one nice app to 36 00:02:05,080 --> 00:02:08,380 S1: take a look at. Again it's called bolt AI. That 37 00:02:08,380 --> 00:02:11,320 S1: link I have here, like I said it's not the 38 00:02:11,320 --> 00:02:16,029 S1: correct link but whatever. Let's see here. Yeah llama three 39 00:02:16,030 --> 00:02:18,550 S1: really impressive. At the seven B level there's also an 40 00:02:18,550 --> 00:02:22,089 S1: eight B version which I don't really mess with. I 41 00:02:22,090 --> 00:02:24,820 S1: only messed with the 70 B version. By the way, 42 00:02:24,820 --> 00:02:28,090 S1: the 70 B version actually all the different versions of 43 00:02:28,090 --> 00:02:31,480 S1: llama three are also in fabric, so you can use 44 00:02:31,480 --> 00:02:35,470 S1: fabric with local models of course, and that includes llama three. 45 00:02:35,470 --> 00:02:38,650 S1: And one thing I've noticed is that it's way less restricted. 46 00:02:38,650 --> 00:02:41,170 S1: It actually will tell you a lot of stuff that 47 00:02:41,169 --> 00:02:43,930 S1: llama two would not, and a lot of the other 48 00:02:43,930 --> 00:02:46,750 S1: pinnacle models will not talk to you about at all. 49 00:02:46,750 --> 00:02:51,130 S1: It's not quite uncensored, but it is a lot more 50 00:02:51,130 --> 00:02:54,220 S1: similar to an uncensored, I would say 50% there or something. 51 00:02:54,220 --> 00:02:58,329 S1: And of course, there's already dolphin versions of llama three 52 00:02:58,330 --> 00:03:02,919 S1: that are uncensored. So, um, not just dolphin, but other versions. 53 00:03:02,919 --> 00:03:06,250 S1: So definitely take a look at that if you need that. 54 00:03:06,250 --> 00:03:08,710 S1: And one thing that's really interesting to me is like 55 00:03:08,710 --> 00:03:12,489 S1: how good llama three is getting or local models in general, 56 00:03:12,490 --> 00:03:16,420 S1: because Apple is now releasing open source models. Snowflake just 57 00:03:16,419 --> 00:03:19,540 S1: released an open source model. And here's one thing I 58 00:03:19,540 --> 00:03:22,960 S1: really think is interesting about this. If you think about 59 00:03:22,960 --> 00:03:26,290 S1: the daily tasks that humans go through every day with like, oh, 60 00:03:26,290 --> 00:03:29,170 S1: find me a car, get me a grocery list, give 61 00:03:29,169 --> 00:03:31,690 S1: me a really good workout, play the best song for 62 00:03:31,690 --> 00:03:34,210 S1: right now. You know, get a present for my partner 63 00:03:34,210 --> 00:03:37,690 S1: because it's our anniversary or whatever. All these tasks that 64 00:03:37,690 --> 00:03:41,770 S1: are like everyday things they have, this is a hypothesis, 65 00:03:41,770 --> 00:03:44,560 S1: but they have a limit to how good they need 66 00:03:44,560 --> 00:03:48,610 S1: to be. Right? So if you exceed that limit of like, 67 00:03:48,610 --> 00:03:51,820 S1: they just get the most amazing flowers every single time 68 00:03:51,820 --> 00:03:55,030 S1: they pick the best song or music every single time 69 00:03:55,030 --> 00:03:59,290 S1: and it's really good. Then if that was GPT five 70 00:03:59,290 --> 00:04:02,830 S1: that made it that good, then GPT seven isn't really needed. 71 00:04:02,830 --> 00:04:06,340 S1: And if local models exceed what GPT five could do, 72 00:04:06,340 --> 00:04:08,350 S1: which of course I think they will within a year 73 00:04:08,350 --> 00:04:11,170 S1: or two or whatever, however long that takes. The point 74 00:04:11,170 --> 00:04:16,180 S1: is the models keep getting better, including the local models, 75 00:04:16,300 --> 00:04:20,290 S1: and that a task that a human wants to have 76 00:04:20,290 --> 00:04:25,090 S1: done kind of hits a level of diminishing return. And 77 00:04:25,089 --> 00:04:27,760 S1: I think for a lot of human tasks, we're about 78 00:04:27,760 --> 00:04:31,419 S1: to hit that level with our pinnacle models. And then 79 00:04:31,420 --> 00:04:34,630 S1: you have your local models that are like lagging behind it. 80 00:04:34,630 --> 00:04:39,580 S1: Once that exceeds it, you have local models everywhere that 81 00:04:39,580 --> 00:04:44,770 S1: could do most tasks. So for example, home automation, home management, 82 00:04:44,770 --> 00:04:48,339 S1: automation of tasks around the home. Like I said, groceries, 83 00:04:48,339 --> 00:04:52,600 S1: everyday things, those things can all be done almost instantly, 84 00:04:52,600 --> 00:04:56,920 S1: basically instantly and locally on the device because the local 85 00:04:56,920 --> 00:04:59,919 S1: chipset will be good enough, or there'll be like a 86 00:04:59,920 --> 00:05:02,980 S1: local chipset in the house somewhere that you outsource to 87 00:05:02,980 --> 00:05:07,090 S1: or whatever, but you don't need pinnacle models for those. Now, 88 00:05:07,089 --> 00:05:08,950 S1: it could be that some of those tasks would actually 89 00:05:09,250 --> 00:05:12,909 S1: benefit from a much bigger model, but I think so 90 00:05:12,910 --> 00:05:16,779 S1: many of them actually won't. You won't need them. And 91 00:05:16,779 --> 00:05:20,450 S1: that's really exciting because when local models can. You most 92 00:05:20,450 --> 00:05:23,630 S1: of what we need. That's an insane world. And my 93 00:05:23,630 --> 00:05:26,270 S1: buddy Joseph Thacker keeps talking about this thing. It's like, 94 00:05:26,270 --> 00:05:28,580 S1: why not just have it in everything? And he keeps 95 00:05:28,580 --> 00:05:30,890 S1: talking about this cool idea, which I want to touch 96 00:05:30,890 --> 00:05:33,800 S1: on too, which is this concept of an oracle, which 97 00:05:33,800 --> 00:05:35,719 S1: I've always wondered about too. It's like if you went 98 00:05:35,720 --> 00:05:38,510 S1: back in time, how useful would you actually be? Could 99 00:05:38,510 --> 00:05:42,320 S1: you actually describe, you know, chemistry or, you know, satellites 100 00:05:42,320 --> 00:05:45,560 S1: or combustion engines or any of this stuff? It'd be 101 00:05:45,560 --> 00:05:47,780 S1: amazing to have like a little soccer ball that talks 102 00:05:47,779 --> 00:05:50,719 S1: to you. And inside of the soccer ball is this model, 103 00:05:50,720 --> 00:05:53,570 S1: and you could somehow get power into it or whatever. 104 00:05:53,570 --> 00:05:57,620 S1: But this one little thing, this tiny little device has 105 00:05:57,620 --> 00:06:01,070 S1: something like Lama three on it, which is a large 106 00:06:01,070 --> 00:06:05,570 S1: percentage of all human knowledge ever. And it's local, doesn't 107 00:06:05,570 --> 00:06:07,700 S1: need to call out to any cloud. And you can 108 00:06:07,700 --> 00:06:10,339 S1: put that in a soccer ball or a park bench 109 00:06:10,339 --> 00:06:13,910 S1: or anything like that. It's insane to me that you 110 00:06:13,910 --> 00:06:18,470 S1: could basically bootstrap humanity to some degree with something that's 111 00:06:18,470 --> 00:06:20,960 S1: completely local on a chip. All right, more stuff going 112 00:06:20,960 --> 00:06:24,650 S1: on prepping for RSA. Definitely come by and say hi. 113 00:06:24,650 --> 00:06:28,070 S1: I might be crazy busy or distracted or socially awkward 114 00:06:28,070 --> 00:06:31,520 S1: at the time. Whatever. Just come say hi. What have 115 00:06:31,520 --> 00:06:35,779 S1: we got here? We got hugs, waves, finger guns, always appropriate. 116 00:06:35,779 --> 00:06:39,890 S1: Fist bumps, whatever. Whatever you got I should be comfortable with. 117 00:06:39,890 --> 00:06:42,950 S1: Last few talks have gone really well. There's something to 118 00:06:42,950 --> 00:06:46,640 S1: be said for speaking, to share an idea rather than 119 00:06:46,640 --> 00:06:50,960 S1: trying to, like, do a presentation. I've crossed over for 120 00:06:50,960 --> 00:06:53,570 S1: most of my talks. At least I'm trying to cross 121 00:06:53,570 --> 00:06:57,169 S1: over completely to this thing where I'm just I have 122 00:06:57,170 --> 00:06:59,690 S1: this idea that I want to share, and the goal 123 00:06:59,690 --> 00:07:02,240 S1: of the talk has nothing to do with the talk 124 00:07:02,240 --> 00:07:04,969 S1: or the slides. It has to do with how many 125 00:07:04,970 --> 00:07:08,659 S1: people actually get converted over to this way of thinking 126 00:07:08,660 --> 00:07:12,020 S1: at the end. And when you are trying to do that, 127 00:07:12,020 --> 00:07:15,800 S1: it makes the slides not matter. It makes like if 128 00:07:15,800 --> 00:07:19,430 S1: whether you stumble, your words not matter. Nervousness kind of 129 00:07:19,430 --> 00:07:22,550 S1: goes away. And I've got a couple of talks already 130 00:07:22,580 --> 00:07:25,700 S1: or essays about how to make that happen. Anyway, there's 131 00:07:25,700 --> 00:07:28,370 S1: a breathing technique for like the tactical version, and then 132 00:07:28,370 --> 00:07:32,420 S1: there's like this concept for getting rid of the overall 133 00:07:32,420 --> 00:07:36,050 S1: problem of stage fright. If you simply have something to 134 00:07:36,050 --> 00:07:39,320 S1: share and you are sharing it, that is not a 135 00:07:39,320 --> 00:07:42,739 S1: stage fright issue. It's very strange. It's like, what are 136 00:07:42,740 --> 00:07:45,860 S1: you actually looking at? If you go up on stage 137 00:07:45,860 --> 00:07:49,790 S1: and you are thinking about your talk in terms of 138 00:07:49,790 --> 00:07:54,290 S1: like the slides, the slide notes, where are the cameras, 139 00:07:54,290 --> 00:07:56,929 S1: where's the monitor, where are the people, who are they 140 00:07:56,930 --> 00:08:00,680 S1: looking at? If you pull into yourself like that, you're 141 00:08:00,680 --> 00:08:03,080 S1: going to get stressed. You're going to get nervous. The 142 00:08:03,080 --> 00:08:07,400 S1: way to fix that is to go outside of yourself. 143 00:08:07,400 --> 00:08:10,550 S1: What's bigger than yourself? What? What is outside of yourself? 144 00:08:10,550 --> 00:08:14,450 S1: The idea is the thing that matters. You focus on 145 00:08:14,450 --> 00:08:17,150 S1: that and suddenly you're very calm because what you want 146 00:08:17,150 --> 00:08:19,550 S1: to do is you want to convert your excitement away 147 00:08:19,550 --> 00:08:24,380 S1: from anxiousness and into excitement around sharing the idea. And 148 00:08:24,380 --> 00:08:26,960 S1: you could view this as a hack, or you could 149 00:08:26,960 --> 00:08:30,800 S1: view it as just a better way of thinking about presenting. Period. 150 00:08:30,800 --> 00:08:33,260 S1: And I think it is both, but I frame it 151 00:08:33,260 --> 00:08:35,660 S1: as the second one. Updated the intro to the newsletter 152 00:08:35,660 --> 00:08:40,190 S1: focusing on human to 2.0 transitioning to 3.0. Let me 153 00:08:40,190 --> 00:08:43,040 S1: know what you think of that. I put discovery actually 154 00:08:43,040 --> 00:08:46,880 S1: into each section instead of having a dedicated discovery this time. 155 00:08:46,880 --> 00:08:51,469 S1: And if you have time, you've got to go listen 156 00:08:51,470 --> 00:08:55,340 S1: to this conversation between Tyler Cohen and Peter Thiel. I'm 157 00:08:55,340 --> 00:08:58,580 S1: not a Peter Thiel fan in general. I didn't like 158 00:08:58,580 --> 00:09:01,429 S1: a lot of his political views. While back they had 159 00:09:01,429 --> 00:09:05,330 S1: this wide ranging conversation from like Star Wars to the 160 00:09:05,330 --> 00:09:11,060 S1: Antichrist to the Bible to Shakespeare to economics, political theory, 161 00:09:11,179 --> 00:09:15,589 S1: just theology. It was an insane conversation. What I like 162 00:09:15,590 --> 00:09:17,510 S1: most about it, because I've shown a few people and 163 00:09:17,510 --> 00:09:19,550 S1: they were like, yeah, I don't know why you liked 164 00:09:19,550 --> 00:09:22,040 S1: it so much. And a few people said, look, they 165 00:09:22,040 --> 00:09:23,750 S1: just talked about a whole bunch of stuff that I 166 00:09:23,750 --> 00:09:27,290 S1: didn't understand. Here's my trick for you, and this is 167 00:09:27,290 --> 00:09:29,750 S1: why I was so excited about this talk. I don't 168 00:09:29,750 --> 00:09:34,910 S1: often find anyone lately because I read so much who 169 00:09:34,910 --> 00:09:37,550 S1: says a string of words and actually just a string 170 00:09:37,550 --> 00:09:42,199 S1: of answers to questions. And in those answers I'm like, 171 00:09:42,200 --> 00:09:46,280 S1: what does that mean? Like Strachan, for example, Strachan was 172 00:09:46,370 --> 00:09:48,770 S1: Stross was one of the people that he was talking 173 00:09:48,770 --> 00:09:51,830 S1: about in this thing. And I'm like, I don't know 174 00:09:51,830 --> 00:09:54,950 S1: Stross yet, I don't what does that mean? So I 175 00:09:54,950 --> 00:09:57,740 S1: ended up having to go in Google and look up 176 00:09:57,740 --> 00:10:00,830 S1: all these different things. Google more like talk to my 177 00:10:00,830 --> 00:10:03,890 S1: Da and have her explain the stuff to me. And 178 00:10:03,890 --> 00:10:07,309 S1: then I also put a set of books into audible 179 00:10:07,309 --> 00:10:09,590 S1: as well. So now I'm going to go read these 180 00:10:09,590 --> 00:10:13,310 S1: canonical books that came from these thinkers. So I don't 181 00:10:13,309 --> 00:10:17,630 S1: accept when somebody talks over my head because I essentially say, well, 182 00:10:17,630 --> 00:10:19,860 S1: you can only do that once because. Now I'm going 183 00:10:19,860 --> 00:10:22,650 S1: to research all that stuff. And first of all, figure 184 00:10:22,650 --> 00:10:25,500 S1: out what you were actually saying just so it doesn't 185 00:10:25,500 --> 00:10:28,800 S1: appear smart when it's actually not smart. But what I 186 00:10:28,800 --> 00:10:32,069 S1: came away with after deciphering all the stuff and learning 187 00:10:32,070 --> 00:10:34,650 S1: like a quick primer on this stuff, is I came 188 00:10:34,650 --> 00:10:38,429 S1: away realizing that Peter Thiel knows a lot of stuff 189 00:10:38,429 --> 00:10:41,280 S1: and what I'm really attracted to. And I don't know 190 00:10:41,280 --> 00:10:43,980 S1: if this is something where he changed or where I changed, 191 00:10:43,980 --> 00:10:48,420 S1: or maybe both or something, but he doesn't seem bound 192 00:10:48,420 --> 00:10:52,350 S1: to any sort of side. He seems bound to principles 193 00:10:52,350 --> 00:10:54,810 S1: of like what he wants to see happen. And I 194 00:10:54,809 --> 00:10:57,300 S1: think that is really powerful. And I talked about that. 195 00:10:57,300 --> 00:11:00,090 S1: I think in a previous show. I can't remember where 196 00:11:00,090 --> 00:11:03,359 S1: I talked about that recently, but I think it's really oh, 197 00:11:03,360 --> 00:11:07,410 S1: it's actually in this newsletter actually. So I would say 198 00:11:07,410 --> 00:11:10,140 S1: definitely check this out no matter what. Check it out. 199 00:11:10,140 --> 00:11:12,750 S1: You might not like it, but listen to what I 200 00:11:12,750 --> 00:11:14,880 S1: was saying before and see if that helps you like 201 00:11:14,880 --> 00:11:17,250 S1: it more. All right, so I wrote a new essay 202 00:11:17,250 --> 00:11:20,969 S1: on the old paradigm of planning a career no longer working, 203 00:11:20,970 --> 00:11:25,620 S1: and basically saying you should plan your career around problems instead. 204 00:11:25,620 --> 00:11:30,030 S1: And I actually have released a standalone video on this 205 00:11:30,030 --> 00:11:31,860 S1: as well, which is also going to come out as 206 00:11:31,860 --> 00:11:35,160 S1: a podcast, so you can check that out on YouTube. 207 00:11:35,280 --> 00:11:38,340 S1: It should already be live. All right. Security, the House 208 00:11:38,340 --> 00:11:40,860 S1: just passed a bill making it illegal for the government 209 00:11:40,860 --> 00:11:44,610 S1: to buy your data without a warrant. Calling it the 210 00:11:44,610 --> 00:11:47,910 S1: Fourth amendment is not for sale. This is really cool. 211 00:11:48,150 --> 00:11:51,360 S1: That's bipartisan. It seems like it would. It would have 212 00:11:51,360 --> 00:11:53,790 S1: had to have been. Not sure about that, but I'm 213 00:11:53,790 --> 00:11:58,410 S1: pretty sure that's right. Sandworm. Notorious Russian hacking group linked 214 00:11:58,410 --> 00:12:02,820 S1: to cyber attack on a water facility in Texas. The 215 00:12:02,820 --> 00:12:06,210 S1: ban on TikTok looks like it's actually going forward. That 216 00:12:06,210 --> 00:12:09,780 S1: is pretty cool. Biden signed it. So that's bipartisan as well. 217 00:12:09,780 --> 00:12:13,590 S1: A little bit of light in the darkness there might 218 00:12:13,590 --> 00:12:17,010 S1: have got compromised among a bunch of other companies by 219 00:12:17,010 --> 00:12:20,280 S1: another China based attacker, and a flaw and putty lets 220 00:12:20,280 --> 00:12:24,120 S1: people look at a whole bunch of different keys or 221 00:12:24,120 --> 00:12:28,290 S1: cryptographic signatures. Basically figure out your private keys, which is 222 00:12:28,290 --> 00:12:33,900 S1: no bueno. FBI Director Christopher Wray says that China is 223 00:12:33,900 --> 00:12:39,300 S1: basically shifting to figuring out how to attack US critical infrastructure, 224 00:12:39,300 --> 00:12:44,340 S1: possibly to go in concert with a move in 2027. 225 00:12:44,340 --> 00:12:46,800 S1: So it would be like, you know, do these cyber 226 00:12:46,800 --> 00:12:49,830 S1: or infrastructure things at the same time to kind of 227 00:12:49,830 --> 00:12:55,260 S1: keep the enemy off balance? Marlinspike moxie basically says he's 228 00:12:55,260 --> 00:12:58,230 S1: no longer affiliated with signal. I've not been a huge 229 00:12:58,230 --> 00:12:59,939 S1: fan of signal. I don't like the fact that you 230 00:12:59,940 --> 00:13:03,180 S1: need to restart it every 45 minutes. I don't like 231 00:13:03,179 --> 00:13:06,689 S1: being outside of Apple Messages. I prefer to have as 232 00:13:06,690 --> 00:13:10,530 S1: few messaging apps as possible because I'm on Apple. That 233 00:13:10,530 --> 00:13:13,439 S1: tends to be my favorite. Not a fan of WhatsApp, 234 00:13:13,440 --> 00:13:15,900 S1: not a fan of telegram, so I tend to just 235 00:13:15,900 --> 00:13:19,350 S1: prefer messages. I was okay with signal a little bit 236 00:13:19,350 --> 00:13:23,160 S1: because it was moxie, but now moxie is out. I'm 237 00:13:23,160 --> 00:13:25,350 S1: going to try to get people to talk to me 238 00:13:25,350 --> 00:13:30,840 S1: on messages again. Sacramento International Airport had to stop flights 239 00:13:30,840 --> 00:13:36,480 S1: because somebody deliberately cut the AT&T internet cable that provided 240 00:13:36,480 --> 00:13:40,890 S1: internet to the airport. Tale scale now has SSH. DeepMind's 241 00:13:40,890 --> 00:13:44,460 S1: boss says Google is going to outspend everyone in AI. 242 00:13:44,490 --> 00:13:47,910 S1: You know, it's not about outspending guys. You've been able 243 00:13:47,910 --> 00:13:51,060 S1: to outspend and have been outspending people for a long time. 244 00:13:51,480 --> 00:13:55,860 S1: It's you've lost the ability to ship good products because 245 00:13:55,860 --> 00:13:58,590 S1: you don't have vision as your primary guide. You have 246 00:13:58,590 --> 00:14:02,100 S1: basically engineers making stuff and throw them at the walls. 247 00:14:02,100 --> 00:14:05,850 S1: And then like three years later, if it launches, you 248 00:14:05,850 --> 00:14:10,170 S1: put it in the graveyard. It's just complete joke. I 249 00:14:10,170 --> 00:14:13,590 S1: think you need new leadership, honestly, is what you need. 250 00:14:13,710 --> 00:14:17,819 S1: And Stanford released a quality report on AI models and 251 00:14:17,820 --> 00:14:21,479 S1: we did a micro summary from fabric. So basically I 252 00:14:21,480 --> 00:14:26,130 S1: surpasses humans in specific tasks, not in complex reasoning. I 253 00:14:26,130 --> 00:14:30,720 S1: think that'll probably change soon. US is ahead of model development. 254 00:14:30,720 --> 00:14:34,170 S1: And yeah, pretty good breakdown here. Interesting argument about how 255 00:14:34,170 --> 00:14:39,030 S1: search engines, especially Google, can actually move elections. Google fired 256 00:14:39,030 --> 00:14:42,660 S1: 28 employees for protesting and they actually fired some more 257 00:14:42,660 --> 00:14:46,680 S1: people as well recently. More recently, Google has merged its 258 00:14:46,680 --> 00:14:51,210 S1: Android and hardware teams. Netflix is using FreeBSD current for 259 00:14:51,210 --> 00:14:53,790 S1: its edge network. That is surprising to me. I would 260 00:14:53,790 --> 00:14:57,990 S1: have thought those packages would be old, not well maintained, 261 00:14:58,380 --> 00:15:03,570 S1: and definitely behind. Yeah, so that's surprising in a good way. 262 00:15:03,720 --> 00:15:06,300 S1: Reddit showing up a lot more in Google results. They 263 00:15:06,300 --> 00:15:09,180 S1: actually have some sort of contract together. I think they 264 00:15:09,270 --> 00:15:13,560 S1: sent Reddit like $30 million or something. So something around there. 265 00:15:13,680 --> 00:15:16,260 S1: Apple's Airplay is starting to come to hotel rooms. Hopefully 266 00:15:16,260 --> 00:15:21,340 S1: Marriott's soon tiny S.A. is a budget. Friendly spectrum analyzer 267 00:15:21,790 --> 00:15:25,690 S1: programming is mostly thinking this it is. This was a 268 00:15:25,690 --> 00:15:30,310 S1: good essay, I enjoyed it. Broad introduction to AWS log 269 00:15:30,310 --> 00:15:35,650 S1: sources and events and Gen Z is outperforming previous generations 270 00:15:35,650 --> 00:15:40,120 S1: at their age. This was also surprising to me. This 271 00:15:40,120 --> 00:15:44,380 S1: article says that societal decline mirrors the death spiral seen 272 00:15:44,380 --> 00:15:49,220 S1: in ants. Why everything is becoming a game. This blog 273 00:15:49,220 --> 00:15:54,320 S1: by the way. Really cool. I love this blog. Gurinder. Yeah, 274 00:15:54,320 --> 00:15:58,220 S1: this guy is awesome. Definitely want to subscribe. Their study 275 00:15:58,220 --> 00:16:00,680 S1: found that jobs that require you to think a lot 276 00:16:00,680 --> 00:16:04,640 S1: or protective against Alzheimer's. I think that needs an H. 277 00:16:04,790 --> 00:16:08,360 S1: Bayer is doing an experiment where they remove most of 278 00:16:08,360 --> 00:16:14,060 S1: middle management and let basically like 100,000 employees mostly self-organize. 279 00:16:14,210 --> 00:16:18,110 S1: And they're hoping to save. I think that was 12 billion, 280 00:16:18,110 --> 00:16:22,490 S1: not 2 billion. That might be a typo. Unfortunate. I 281 00:16:22,490 --> 00:16:26,750 S1: think it was $12 billion. I think this is really cool. 282 00:16:26,750 --> 00:16:29,960 S1: I love the idea of 100,000 people with all these 283 00:16:29,960 --> 00:16:33,920 S1: layers of middle management, where it's just a giant waste machine. 284 00:16:34,040 --> 00:16:37,730 S1: And who better to figure that out then? AI who 285 00:16:37,730 --> 00:16:41,510 S1: figures out, okay, who's doing what, what functions actually need 286 00:16:41,510 --> 00:16:45,950 S1: to get done? What who whose teams are structured in 287 00:16:45,950 --> 00:16:48,890 S1: what way? How do they communicate with each other? What 288 00:16:48,890 --> 00:16:51,860 S1: are the work products that are actually being produced? What 289 00:16:51,860 --> 00:16:55,670 S1: are those pipelines look like for work being produced? You 290 00:16:55,670 --> 00:16:58,730 S1: overlay that on the org structure, and you look at 291 00:16:58,730 --> 00:17:00,980 S1: the budget and everything, and you look at the outcomes 292 00:17:00,980 --> 00:17:04,130 S1: and actually how fast things move. And AI is going 293 00:17:04,130 --> 00:17:06,740 S1: to look at it and just be like a horrible mess. 294 00:17:07,130 --> 00:17:09,740 S1: Why do you have all these managers? What are they 295 00:17:09,740 --> 00:17:13,340 S1: actually doing now? Humans have already known that we didn't 296 00:17:13,340 --> 00:17:16,040 S1: need all those layers of middle management like this. It's 297 00:17:16,040 --> 00:17:20,030 S1: so common. It's just a meme. It's common knowledge. But 298 00:17:20,030 --> 00:17:21,530 S1: AI is going to be able to look at it 299 00:17:21,530 --> 00:17:24,230 S1: for any given company and just be like, here's why 300 00:17:24,230 --> 00:17:28,220 S1: this is a giant waste. You know, Chris isn't doing 301 00:17:28,220 --> 00:17:31,430 S1: anything right. And he hands it to John, who's not 302 00:17:31,430 --> 00:17:34,070 S1: doing anything either. And what is Mary doing? Like this 303 00:17:34,070 --> 00:17:38,840 S1: makes no sense whatsoever. And it's going to recommend look, 304 00:17:38,869 --> 00:17:41,179 S1: you need your experts and you're going to need you 305 00:17:41,180 --> 00:17:44,840 S1: need your high level thinkers and you need very little 306 00:17:44,840 --> 00:17:48,169 S1: middle management, which really is probably going to be AI 307 00:17:48,170 --> 00:17:52,400 S1: agents very soon. And of course, at some point you 308 00:17:52,400 --> 00:17:56,060 S1: can even use AI for the top and bottom pieces 309 00:17:56,060 --> 00:17:59,060 S1: as well. But first they're going to come for those 310 00:17:59,060 --> 00:18:01,400 S1: middle managers. And I think it's a good thing the 311 00:18:01,400 --> 00:18:06,139 S1: term brainwashing morphed into a blanket term for unconventional behavior. 312 00:18:06,140 --> 00:18:09,830 S1: And this came from MKUltra seeing a lot of stuff 313 00:18:09,830 --> 00:18:13,609 S1: talking about MKUltra. I wonder what the reason for that is. Okay, 314 00:18:13,609 --> 00:18:18,530 S1: ideas and analysis. All right. I like I said, I 315 00:18:18,530 --> 00:18:21,260 S1: talked about Peter Thiel. I'm not going to read this 316 00:18:21,260 --> 00:18:23,240 S1: whole thing. This is a bit of an essay here. 317 00:18:23,240 --> 00:18:27,050 S1: I think I might turn this into a standalone essay. Okay. 318 00:18:27,050 --> 00:18:32,060 S1: Let's see here. Okay. Recommendation of the week. Establish your 319 00:18:32,060 --> 00:18:35,210 S1: ground truth in terms of morality and the society that 320 00:18:35,210 --> 00:18:37,100 S1: you want to live in. And this is going to 321 00:18:37,100 --> 00:18:40,879 S1: the essay above. Okay. Establish your ground truth in terms 322 00:18:40,880 --> 00:18:43,550 S1: of morality and the society that you want to live in. 323 00:18:43,550 --> 00:18:47,150 S1: Lock that in perfectly. You say, look, this is how 324 00:18:47,150 --> 00:18:52,280 S1: I think the world should work and that is your morality, okay? 325 00:18:52,280 --> 00:18:55,790 S1: And it's got no labels on it. Left, right, progressive, conservative. 326 00:18:55,790 --> 00:18:58,670 S1: Nobody cares. You don't put anything on that. You just say, 327 00:18:58,670 --> 00:19:02,240 S1: here's how I think the world should work. Then you say, okay, 328 00:19:02,240 --> 00:19:05,240 S1: I'm going to listen to any ideas, right? I'm curious 329 00:19:05,240 --> 00:19:09,500 S1: how those ideas or those structures or those religions or whatever. 330 00:19:09,920 --> 00:19:12,649 S1: How would those affect this world that I want to 331 00:19:12,650 --> 00:19:16,940 S1: see come to pass? And then you don't discard ideas 332 00:19:16,940 --> 00:19:19,969 S1: just because you disagree with something else that they say 333 00:19:19,970 --> 00:19:23,570 S1: somewhere else. So if people have like this idea, a 334 00:19:23,570 --> 00:19:27,800 S1: constellation or belief constellation, you are free to pick and 335 00:19:27,800 --> 00:19:31,909 S1: choose for where they they are smart in your mind, 336 00:19:31,910 --> 00:19:34,580 S1: at least where they're smart about things, where they're dumb 337 00:19:34,580 --> 00:19:38,629 S1: about things. I get so many comments where people are like, hey, listen, 338 00:19:38,630 --> 00:19:42,500 S1: I absolutely love your stuff. I don't agree with everything, 339 00:19:42,500 --> 00:19:45,320 S1: but I agree with you on most things, and I 340 00:19:45,320 --> 00:19:47,870 S1: kind of feel like I'm happy when I hear that 341 00:19:47,960 --> 00:19:52,160 S1: because I feel like they're parsing everything. They're deciding because 342 00:19:52,160 --> 00:19:55,610 S1: they've already done number one. They've decided what they believe. 343 00:19:55,760 --> 00:19:58,580 S1: And I've had other people also say that over time 344 00:19:58,580 --> 00:20:01,820 S1: they came to believe something that I also believed, which 345 00:20:01,820 --> 00:20:05,030 S1: is cool. And sometimes I'll switch and, you know, believe 346 00:20:05,030 --> 00:20:07,790 S1: what they believe. The point is, you don't have to 347 00:20:07,790 --> 00:20:11,989 S1: accept everything a person says to pull wisdom out of 348 00:20:11,990 --> 00:20:15,109 S1: what they say, right? Maybe they're only good at making 349 00:20:15,109 --> 00:20:19,100 S1: the best, you know, Reuben sandwiches. And maybe that's all 350 00:20:19,100 --> 00:20:21,680 S1: I get from them, because the person's name is Tucker 351 00:20:21,680 --> 00:20:25,939 S1: Carlson and they don't know anything other than sandwiches. That's fine, 352 00:20:25,940 --> 00:20:28,910 S1: I will listen. I probably won't spend time like trying 353 00:20:28,910 --> 00:20:31,070 S1: to seek out that content, but it doesn't matter. That's 354 00:20:31,070 --> 00:20:34,280 S1: the point of this. Feel free to label people as 355 00:20:34,280 --> 00:20:37,639 S1: bad overall or stupid, but realize it doesn't mean they're 356 00:20:37,640 --> 00:20:41,210 S1: wrong about everything. So again, same concept. Be willing to 357 00:20:41,210 --> 00:20:45,409 S1: take truth from anywhere. And basically you want to regularly 358 00:20:45,410 --> 00:20:49,280 S1: revisit your number one, which. Is your basis of how 359 00:20:49,280 --> 00:20:54,260 S1: you think the world should work and then regularly re-evaluate, right? 360 00:20:54,260 --> 00:20:58,610 S1: Pull in lots of different opinions and see if it changes. 361 00:20:58,609 --> 00:21:02,030 S1: Number one and the aphorism of the week, best way 362 00:21:02,030 --> 00:21:05,570 S1: to have good ideas is to have lots of ideas. 363 00:21:05,570 --> 00:21:07,910 S1: The best way to have good ideas is to have 364 00:21:07,910 --> 00:21:13,209 S1: lots of ideas. Linus Pauling. Unsupervised Learning is produced and 365 00:21:13,210 --> 00:21:16,810 S1: edited by Daniel Missler on a Neumann 87 AI microphone 366 00:21:16,810 --> 00:21:20,770 S1: using Hindenburg. Intro and outro music is by zombie with 367 00:21:20,770 --> 00:21:23,590 S1: a Y, and to get the text and links from 368 00:21:23,590 --> 00:21:25,720 S1: this episode, sign up for the newsletter version of the 369 00:21:25,720 --> 00:21:30,570 S1: show at Daniel Missler. Com slash newsletter. We'll see you 370 00:21:30,570 --> 00:21:31,139 S1: next time.