1 00:00:01,100 --> 00:00:04,490 S1: Welcome to Unsupervised Learning, a security, AI and meaning focused 2 00:00:04,490 --> 00:00:07,370 S1: podcast that looks at how best to thrive as humans 3 00:00:07,370 --> 00:00:11,570 S1: in a post AI world. It combines original ideas, analysis, 4 00:00:11,570 --> 00:00:14,810 S1: and mental models to bring not just the news, but 5 00:00:14,810 --> 00:00:21,939 S1: why it matters and how to respond. All right. Welcome 6 00:00:21,940 --> 00:00:24,459 S1: to unsupervised Learning. This is Daniel Meisler. I'm going to 7 00:00:24,460 --> 00:00:27,880 S1: jump into this episode this week. So someone asked me 8 00:00:27,880 --> 00:00:30,130 S1: last week about the best way to run AI models 9 00:00:30,130 --> 00:00:32,559 S1: I think they were specifically talking about based on the question. 10 00:00:32,560 --> 00:00:35,380 S1: They were talking about local models. So here's a breakdown. 11 00:00:35,380 --> 00:00:38,740 S1: So and this is for AI in general. So I 12 00:00:38,740 --> 00:00:41,080 S1: prefer using APIs for any sort of thing that I'm 13 00:00:41,080 --> 00:00:43,480 S1: doing over and over. I don't like to use front 14 00:00:43,479 --> 00:00:47,590 S1: ends like ChatGPT or Claude's version of ChatGPT. I like 15 00:00:47,590 --> 00:00:50,260 S1: to go directly to the API using my own code 16 00:00:50,260 --> 00:00:53,379 S1: for day to day use. I prefer using fabric, which 17 00:00:53,380 --> 00:00:57,430 S1: is my own project, and that is because it has 18 00:00:57,430 --> 00:01:03,070 S1: specific use cases for dozens, almost 100 different regular human 19 00:01:03,070 --> 00:01:06,700 S1: problems that we do every day. So I would recommend 20 00:01:06,700 --> 00:01:10,390 S1: using fabric for that. And again, the project is here. 21 00:01:10,390 --> 00:01:13,000 S1: And if you look at patterns, these are all the 22 00:01:13,000 --> 00:01:15,039 S1: different things that you could do with it. Right. So 23 00:01:15,040 --> 00:01:20,530 S1: it's like analyzing text, doing job analysis, creating presentations, finding 24 00:01:20,530 --> 00:01:26,169 S1: flaws in arguments. You know, taking exquisite manual like notes 25 00:01:26,170 --> 00:01:29,620 S1: on a four hour video and doing it in 30s, 26 00:01:29,620 --> 00:01:33,580 S1: extracting the ideas from a book, pulling out predictions that 27 00:01:33,580 --> 00:01:36,670 S1: a person made, like all sorts of stuff you could do. 28 00:01:36,670 --> 00:01:40,180 S1: So that is like my number one way of interacting 29 00:01:40,180 --> 00:01:44,169 S1: with AI. If you just want to show off what 30 00:01:44,170 --> 00:01:48,490 S1: you could do with local models, I would recommend Olama, 31 00:01:48,490 --> 00:01:50,500 S1: which is really cool. I'll show you that real quick. 32 00:01:50,500 --> 00:01:53,410 S1: So you just go to models and if you click 33 00:01:53,410 --> 00:01:55,750 S1: on Lama three, you can just run this model right 34 00:01:55,750 --> 00:01:58,750 S1: here and you just literally copy this. And once you 35 00:01:58,750 --> 00:02:02,440 S1: have Al-ummah installed and running, which takes a few seconds, 36 00:02:02,440 --> 00:02:05,680 S1: you basically run this and it will download the model 37 00:02:05,680 --> 00:02:08,230 S1: and put you into a shell that allows you to 38 00:02:08,230 --> 00:02:11,290 S1: talk directly to it. So it's really cool. So let 39 00:02:11,290 --> 00:02:13,900 S1: me just show you what that looks like if I 40 00:02:13,900 --> 00:02:17,500 S1: am over here. Okay. This one is almost done. I 41 00:02:17,500 --> 00:02:19,690 S1: was doing it like the pie is in the oven 42 00:02:19,690 --> 00:02:23,769 S1: type thing. We still got a little bit more time 99%. 43 00:02:23,770 --> 00:02:26,470 S1: This one is the 70 B version. So this is 44 00:02:26,470 --> 00:02:30,310 S1: a giant model. This is like 40 something gigs and 45 00:02:30,310 --> 00:02:33,520 S1: have it running here. You see the command I put 46 00:02:33,520 --> 00:02:37,810 S1: in Olama run Al-umma three colon 70 B. All right. 47 00:02:37,810 --> 00:02:41,200 S1: So that's all you basically do. And then you get 48 00:02:41,200 --> 00:02:44,620 S1: a shell and you are talking to the model completely 49 00:02:44,620 --> 00:02:48,280 S1: locally just on your own machine. So that's really cool. Okay. 50 00:02:48,280 --> 00:02:53,109 S1: So if you go to back to here. So that 51 00:02:53,110 --> 00:02:56,170 S1: is a way to like demo models and show them 52 00:02:56,169 --> 00:02:58,660 S1: off and show what you could potentially do with them. 53 00:02:58,660 --> 00:03:00,940 S1: To me it's a little bit gimmicky because I would 54 00:03:00,940 --> 00:03:05,200 S1: rather use fabric or the next one I'm going to 55 00:03:05,200 --> 00:03:10,419 S1: talk about, which is this one. This is a new tool. 56 00:03:10,419 --> 00:03:12,250 S1: It's not new. It's been out for a while, but 57 00:03:12,250 --> 00:03:15,520 S1: it's growing in popularity. It's called LM studio. This is 58 00:03:15,520 --> 00:03:19,450 S1: a full Mac application. It also runs on Windows and 59 00:03:19,450 --> 00:03:23,530 S1: Linux and essentially what you do is you install this 60 00:03:23,530 --> 00:03:27,370 S1: thing and you just type into this, the search bar here, 61 00:03:27,370 --> 00:03:29,200 S1: the name of the model you want to get, and 62 00:03:29,200 --> 00:03:32,650 S1: it pulls down models from like hugging face and like 63 00:03:32,650 --> 00:03:36,280 S1: all sorts of different places, it downloads it for you, 64 00:03:36,610 --> 00:03:39,130 S1: loads it for you, which could be really difficult if 65 00:03:39,130 --> 00:03:41,140 S1: you try to do it manually. And then when you 66 00:03:41,140 --> 00:03:43,450 S1: click on this icon here, you're actually in a chat 67 00:03:43,450 --> 00:03:47,500 S1: interface very similar to ChatGPT, except for it's a local 68 00:03:47,500 --> 00:03:52,240 S1: thing and it's working completely local for local models, and 69 00:03:52,240 --> 00:03:55,210 S1: it's just really powerful. And there's actually even a comparison, 70 00:03:55,210 --> 00:03:58,390 S1: one where you can actually compare different models. But the 71 00:03:58,390 --> 00:04:00,760 S1: most important thing about this is that you could try 72 00:04:00,760 --> 00:04:03,160 S1: new things. You could try things that just came out 73 00:04:03,160 --> 00:04:07,510 S1: like yesterday or the day before or today. Whereas with 74 00:04:07,510 --> 00:04:10,720 S1: Olama this one, if you go to models, it's really 75 00:04:10,720 --> 00:04:13,390 S1: just kind of like the major ones. There's only a 76 00:04:13,390 --> 00:04:15,730 S1: few here. I mean, I say a few, but it's 77 00:04:15,730 --> 00:04:18,040 S1: like a couple of dozen. But if you look at 78 00:04:18,040 --> 00:04:24,310 S1: like hugging face, let's look at that number. Yeah 629,000. 79 00:04:24,310 --> 00:04:28,930 S1: So it's pretty different versus a couple dozen. So that's 80 00:04:28,930 --> 00:04:31,630 S1: the type of thing you could do. If you're using 81 00:04:32,020 --> 00:04:35,050 S1: LM studio, you just have access to way more models 82 00:04:35,050 --> 00:04:37,990 S1: to mess around and play with and even compare the models. 83 00:04:37,990 --> 00:04:40,630 S1: So I would say this is like what I would 84 00:04:40,630 --> 00:04:43,720 S1: recommend for someone who is gooey oriented and wants to 85 00:04:43,720 --> 00:04:46,450 S1: do lots of tinkering. And if you want to just 86 00:04:46,450 --> 00:04:48,970 S1: be a workhorse and get lots of things done, I 87 00:04:48,970 --> 00:04:53,469 S1: would recommend Fabrique, which is a command line interface and 88 00:04:53,470 --> 00:04:58,390 S1: also a GUI. Now, like I said before, using other models, 89 00:04:58,390 --> 00:05:01,630 S1: lots of cutting edge models. That's kind of the advantage 90 00:05:01,630 --> 00:05:03,700 S1: of LM studio. And you could do that with fabric 91 00:05:03,700 --> 00:05:07,089 S1: as well. We're adding the ability to do that. But 92 00:05:07,089 --> 00:05:11,590 S1: my current favorite one is this llama 370 B instruct. 93 00:05:11,589 --> 00:05:15,280 S1: This one's really cool from Quant Factory. It's really powerful. 94 00:05:15,279 --> 00:05:17,739 S1: One I've been messing with some different ones. Lots of 95 00:05:17,740 --> 00:05:21,979 S1: different groups put out their modified. Versions of Lama essentially, 96 00:05:21,980 --> 00:05:26,360 S1: and this is a version of 70. Be fairly fast, 97 00:05:26,360 --> 00:05:30,560 S1: really high quality. And yeah, just really cool way to 98 00:05:30,560 --> 00:05:32,450 S1: interact with stuff. And again, the name of the tool 99 00:05:32,450 --> 00:05:35,659 S1: is LM studio. We had a really great book club 100 00:05:35,660 --> 00:05:39,110 S1: this week. We talked for like an hour and ten minutes, ten, 101 00:05:39,110 --> 00:05:42,740 S1: 15 minutes. And then we pick the next book. And yeah, 102 00:05:42,740 --> 00:05:45,020 S1: really thankful to be part of the UL community and 103 00:05:45,020 --> 00:05:48,680 S1: doing these book clubs. We do book clubs once a month. 104 00:05:48,680 --> 00:05:52,489 S1: We do a weekly or a monthly meetup halfway through 105 00:05:52,490 --> 00:05:55,099 S1: the month where we talk about we share tips. We 106 00:05:55,100 --> 00:06:01,159 S1: share like automation, hacking tips like security tools, productivity tools. 107 00:06:01,160 --> 00:06:04,700 S1: We talk about life. It's a very sort of intimate 108 00:06:04,700 --> 00:06:08,659 S1: and sharing and uplifting sort of group. And you should 109 00:06:08,660 --> 00:06:11,600 S1: come join us if that appeals to you. It's just 110 00:06:11,600 --> 00:06:15,469 S1: becoming a premium member of unsupervised learning. And I just 111 00:06:15,470 --> 00:06:18,229 S1: signed up for my first full body MRI, which is 112 00:06:18,230 --> 00:06:21,320 S1: going to be super fun. Not looking forward to being 113 00:06:21,320 --> 00:06:24,560 S1: inside of a tube. I guess I'll just meditate to 114 00:06:24,560 --> 00:06:27,560 S1: get through that, but I'm not overly claustrophobic. It just 115 00:06:27,560 --> 00:06:29,659 S1: doesn't sound like a good time, especially since the full 116 00:06:29,690 --> 00:06:33,380 S1: body one is 60 minutes and RSA is next week, 117 00:06:33,380 --> 00:06:37,219 S1: so that's going to be exciting to see everybody. So 118 00:06:37,220 --> 00:06:41,330 S1: I found a good reason for college and specifically good colleges. 119 00:06:41,330 --> 00:06:43,550 S1: And this is thanks to Paul Graham who I read 120 00:06:43,550 --> 00:06:46,400 S1: a bunch of his essays where he talked about college. 121 00:06:46,400 --> 00:06:48,470 S1: And I think there was one where he talked about 122 00:06:48,470 --> 00:06:50,930 S1: this particular concept. So I just wrote it up in 123 00:06:50,930 --> 00:06:54,380 S1: a separate post because he didn't do that. He basically 124 00:06:54,380 --> 00:06:57,200 S1: just mentioned it kind of as an aside. So this 125 00:06:57,200 --> 00:06:59,540 S1: is basically the reason you want to go to elite 126 00:06:59,540 --> 00:07:03,080 S1: colleges or send your kids to elite colleges is because 127 00:07:03,080 --> 00:07:06,650 S1: there's very few things that are more valuable than your 128 00:07:06,650 --> 00:07:13,070 S1: kid being surrounded by highly resourced, highly motivated, highly intelligent 129 00:07:13,070 --> 00:07:17,270 S1: and highly disciplined people. And that is what you find 130 00:07:17,270 --> 00:07:20,330 S1: at the higher and higher tiers of college. All the 131 00:07:20,330 --> 00:07:23,480 S1: way up to like Stanford and Harvard. It's just like 132 00:07:23,480 --> 00:07:27,740 S1: their parents were rich, their parents were probably smart, they're 133 00:07:27,740 --> 00:07:31,730 S1: probably smart, and they had to be disciplined to get 134 00:07:31,730 --> 00:07:34,910 S1: decent grades. So it's like the combination of things just 135 00:07:34,910 --> 00:07:37,310 S1: stacks up. And if you look at all these different 136 00:07:37,310 --> 00:07:40,940 S1: companies that are being founded, they're usually being founded by companies, 137 00:07:41,060 --> 00:07:43,910 S1: by people who went to these schools, even if they 138 00:07:43,910 --> 00:07:46,610 S1: dropped out. So it's like, don't think of it as 139 00:07:46,610 --> 00:07:50,270 S1: the education because the education is very similar at like 140 00:07:50,270 --> 00:07:54,170 S1: a decent state school compared to like an Ivy. But 141 00:07:54,170 --> 00:07:57,470 S1: it's not about the education, it's about the connections. So 142 00:07:57,470 --> 00:08:00,950 S1: just kind of an interesting little tip there way to 143 00:08:00,950 --> 00:08:05,750 S1: frame education. Oh, I will say that the thing that 144 00:08:05,750 --> 00:08:07,880 S1: this makes me think of is how can we get 145 00:08:07,880 --> 00:08:12,770 S1: that somewhere else? So let's disambiguate or let's break apart 146 00:08:12,770 --> 00:08:17,330 S1: these two things education and the networking. Let's find other 147 00:08:17,330 --> 00:08:19,190 S1: ways to do the networking. So we don't have to 148 00:08:19,190 --> 00:08:22,460 S1: have this weird college experience. Now, maybe that's not possible 149 00:08:22,460 --> 00:08:25,040 S1: because the college experience is actually what gives you the 150 00:08:25,040 --> 00:08:28,280 S1: connection in the networking with all the drinking and all 151 00:08:28,280 --> 00:08:31,550 S1: the parties and everything, and then all the just forced 152 00:08:31,550 --> 00:08:35,179 S1: time together in these dorms. Maybe that's the magic sauce, 153 00:08:35,179 --> 00:08:37,640 S1: but I bet you there is a better way to 154 00:08:37,640 --> 00:08:41,209 S1: do that. You could have organized off sites more like 155 00:08:41,210 --> 00:08:47,360 S1: hacking competitions or hackathons or summer camp or something like that, 156 00:08:47,360 --> 00:08:52,040 S1: where you're actually forging these connections separately from doing the 157 00:08:52,040 --> 00:08:56,390 S1: education piece, which arguably could be done better with YouTube 158 00:08:56,390 --> 00:09:01,640 S1: and online courses combined with like in-person workshops. So let's 159 00:09:01,640 --> 00:09:04,220 S1: take the best of education and let's take the best 160 00:09:04,220 --> 00:09:08,750 S1: of networking and maybe do them separately, or even mix them, 161 00:09:08,750 --> 00:09:11,780 S1: but not be confused about thinking that they're the same. 162 00:09:12,110 --> 00:09:15,740 S1: Because that's the problem currently with college. And the result 163 00:09:15,740 --> 00:09:19,729 S1: for most people is worse networking and worse education. Okay, 164 00:09:19,730 --> 00:09:24,440 S1: so security MI5 is starting to vet key researchers going 165 00:09:24,440 --> 00:09:27,770 S1: to top universities in Britain to make sure that they're 166 00:09:27,770 --> 00:09:31,880 S1: not essentially from a country like China, where they're just 167 00:09:31,880 --> 00:09:34,820 S1: coming in to steal all the stuff. And I really 168 00:09:34,820 --> 00:09:36,740 S1: think the US needs to start doing something like this 169 00:09:36,740 --> 00:09:40,280 S1: as well. And speaking of that, Germany just arrested three 170 00:09:40,280 --> 00:09:44,870 S1: people who are suspected Chinese spies for stealing secrets. And 171 00:09:44,870 --> 00:09:48,380 S1: this is right after their chancellor was giving crap to 172 00:09:48,380 --> 00:09:52,700 S1: the to China publicly basically talking about IP theft. And 173 00:09:52,700 --> 00:09:55,100 S1: then they made those arrests. I don't know if that 174 00:09:55,100 --> 00:09:58,219 S1: was coordinated or not. After moving forward with a possible 175 00:09:58,220 --> 00:10:02,720 S1: ban on TikTok, the US is now looking at DJI drones, 176 00:10:02,720 --> 00:10:05,240 S1: basically saying, look, they have so much of the market 177 00:10:05,240 --> 00:10:08,059 S1: and they're flying all over the country. Who knows what 178 00:10:08,059 --> 00:10:11,900 S1: they're doing with those pictures. And it's a state, partially 179 00:10:11,900 --> 00:10:16,850 S1: state owned Chinese company, which means that China runs it. 180 00:10:16,850 --> 00:10:19,610 S1: I mean, they have full control if they want it. 181 00:10:19,780 --> 00:10:23,890 S1: Of any company. So yeah, this makes sense. This one 182 00:10:23,890 --> 00:10:28,330 S1: is insane. Former athletic director arrested for using AI to 183 00:10:28,330 --> 00:10:32,109 S1: mimic a principal's voice, and basically had him saying a 184 00:10:32,110 --> 00:10:35,980 S1: bunch of racist and anti-Semitic things in the voice of 185 00:10:35,980 --> 00:10:39,339 S1: that person he was attacking. And then that started a 186 00:10:39,340 --> 00:10:41,950 S1: whole investigation, and it turned out it was a deep fake. 187 00:10:41,950 --> 00:10:46,929 S1: But yeah, yet another abuse case for deepfakes, character assassination. 188 00:10:46,929 --> 00:10:49,510 S1: And I think it's going to be super cool to 189 00:10:49,510 --> 00:10:52,210 S1: watch all these different attacks. But I think it's also 190 00:10:52,210 --> 00:10:54,490 S1: going to make it harder. This is kind of a 191 00:10:54,490 --> 00:10:57,100 S1: sad thing. It's going to make it harder to believe 192 00:10:57,100 --> 00:11:00,189 S1: people when they bring forth evidence. It's like, hey, my 193 00:11:00,190 --> 00:11:03,520 S1: boyfriend did this, my boss did this, and you show 194 00:11:03,520 --> 00:11:06,729 S1: him a video of it, of them doing it, or 195 00:11:06,730 --> 00:11:09,850 S1: you show them a recording, have them listen to a 196 00:11:09,850 --> 00:11:13,300 S1: recording and they're like, yeah, I mean, I hear them 197 00:11:13,300 --> 00:11:15,850 S1: saying it, but like, it's easy to make anybody say 198 00:11:15,850 --> 00:11:18,640 S1: anything and you go ask the actual person. They're like, yeah, 199 00:11:18,640 --> 00:11:21,490 S1: I never said that. That was a deepfake. So it's 200 00:11:21,490 --> 00:11:24,309 S1: going to take some power out of the hands of, 201 00:11:24,309 --> 00:11:27,430 S1: I think, victims. And so we're going to have to, 202 00:11:27,429 --> 00:11:30,820 S1: as a society, sort of figure that out. Nation state 203 00:11:30,820 --> 00:11:34,360 S1: attackers have been using two zero days in Cisco firewalls. 204 00:11:34,360 --> 00:11:37,330 S1: And this is not the first time that's happened to Cisco. 205 00:11:37,330 --> 00:11:41,290 S1: And it's also happening to other firewall providers as well. 206 00:11:41,290 --> 00:11:46,150 S1: And US lawmakers gave the okay to extend warrantless surveillance 207 00:11:46,150 --> 00:11:52,179 S1: under FISA, section 702, these laws that allow the government 208 00:11:52,179 --> 00:11:56,050 S1: to do these things, they kind of get perpetually extended. 209 00:11:56,200 --> 00:11:59,440 S1: I am mostly okay with it. I feel like as 210 00:11:59,440 --> 00:12:05,199 S1: long as there's transparency and oversight, then it's probably necessary. 211 00:12:05,200 --> 00:12:07,420 S1: But I don't like when they sort of become secret 212 00:12:07,420 --> 00:12:11,590 S1: programs and don't have oversight. Then you start worrying about 213 00:12:11,590 --> 00:12:15,580 S1: what they can actually do to Americans and change healthcare. 214 00:12:15,580 --> 00:12:19,690 S1: Paid 22 million in Bitcoin to ransomware attackers, but the 215 00:12:19,690 --> 00:12:22,179 S1: files were already compromised, so they're not sure exactly what 216 00:12:22,179 --> 00:12:24,400 S1: that fallout is going to be. And the Air Force 217 00:12:24,400 --> 00:12:28,030 S1: is moving forward with Anduril. I'm guessing that's how you 218 00:12:28,030 --> 00:12:32,380 S1: pronounce it. And General Atomics to develop unmanned fighter jets, 219 00:12:32,380 --> 00:12:37,840 S1: sidelining giants like Boeing and Lockheed Martin. So Anduril and 220 00:12:37,840 --> 00:12:42,910 S1: General Atomics, interesting unmanned fighter jets, just like we've always 221 00:12:42,910 --> 00:12:47,920 S1: predicted in sci fi and Detective Fi's external attack surface 222 00:12:47,920 --> 00:12:51,850 S1: management solution is now in AWS and that is un. 223 00:12:51,850 --> 00:12:54,910 S1: Oh yeah, defend defy versus Detectify. I was going to 224 00:12:54,910 --> 00:12:59,350 S1: say I think that was an accident. Technology. Amazon's robot 225 00:12:59,350 --> 00:13:05,350 S1: workforce has expanded to over 750,000 robots. And it says 226 00:13:05,350 --> 00:13:09,040 S1: overtaking 100,000 jobs. It's not super clear how they're getting 227 00:13:09,070 --> 00:13:12,820 S1: to that number, but whatever. I mean, we all know 228 00:13:12,820 --> 00:13:15,160 S1: where this is going, right? And like I say, don't 229 00:13:15,309 --> 00:13:18,670 S1: be surprised about this. Expect it. Companies will do what 230 00:13:18,670 --> 00:13:22,360 S1: makes sense for margins, not for people. The responsibility is 231 00:13:22,360 --> 00:13:27,130 S1: to shareholders. And human employees are like, I hate to 232 00:13:27,130 --> 00:13:30,670 S1: say it, they're the worst, right? It they're to be avoided. 233 00:13:30,670 --> 00:13:33,280 S1: You don't want to have to hire somebody if you 234 00:13:33,280 --> 00:13:35,380 S1: have a hot dog stand and it works perfectly and 235 00:13:35,380 --> 00:13:39,730 S1: it's just you're the sole proprietor of this establishment and 236 00:13:39,730 --> 00:13:42,640 S1: you have to hire people, that is not good. Now, 237 00:13:42,640 --> 00:13:46,120 S1: of course, if you have to hire because you want 238 00:13:46,120 --> 00:13:49,600 S1: to sell more hot dogs, that's fine. But it's much 239 00:13:49,600 --> 00:13:53,080 S1: better if you have some robots that you can buy, 240 00:13:53,080 --> 00:13:55,150 S1: and they can make more hot dogs and make more 241 00:13:55,150 --> 00:13:57,730 S1: people happy, and you get to share all the stuff, 242 00:13:57,730 --> 00:14:00,670 S1: and the robots don't get sick and they don't complain. 243 00:14:00,670 --> 00:14:05,199 S1: So that is the way companies see this, and we 244 00:14:05,200 --> 00:14:08,319 S1: just have to be ready for that. Voyager one is back. 245 00:14:08,320 --> 00:14:12,310 S1: Sharing information like it kind of went silent for a while. 246 00:14:12,309 --> 00:14:15,730 S1: We were worried about it. This thing is an interstellar space. 247 00:14:15,730 --> 00:14:19,690 S1: That means it's outside of our solar system. Like iPhones 248 00:14:19,690 --> 00:14:22,750 S1: don't perform this well, like nothing performs this well that 249 00:14:22,750 --> 00:14:25,870 S1: we have ever made. This thing was launched. What? When 250 00:14:25,870 --> 00:14:29,560 S1: I was a kid, I mean, decades ago. And it's 251 00:14:29,560 --> 00:14:32,950 S1: still working. I just can't believe this thing. And. Okay, 252 00:14:33,010 --> 00:14:37,630 S1: someone at Google senior executive is pushing for speed, basically 253 00:14:37,630 --> 00:14:42,400 S1: saying the 25,000 person strong knowledge and information team has 254 00:14:42,400 --> 00:14:48,310 S1: to adapt to a new operating reality with tighter project timelines. Yeah, 255 00:14:48,310 --> 00:14:51,670 S1: I wouldn't characterize it as tighter timelines. I would say 256 00:14:51,670 --> 00:14:56,680 S1: have vision, right. And like I said, this is progress. 257 00:14:56,680 --> 00:14:59,290 S1: But they need a whole new way of thinking about 258 00:14:59,290 --> 00:15:01,990 S1: making products at Google. And I think that requires a 259 00:15:01,990 --> 00:15:05,200 S1: new leader. Police are now using GPT four powered body 260 00:15:05,200 --> 00:15:10,060 S1: cams to turn audio into reports. That's great for transparency, 261 00:15:10,060 --> 00:15:13,720 S1: I think. How about this for every stop, you basically 262 00:15:13,720 --> 00:15:16,900 S1: publish the transcripts and that goes on an open website. 263 00:15:16,900 --> 00:15:19,570 S1: That would be fantastic. And they could clean it up. 264 00:15:19,700 --> 00:15:23,390 S1: For privacy reasons, but you can see the whole interaction. 265 00:15:23,390 --> 00:15:26,540 S1: And then you also publish the video and the transcript. 266 00:15:26,540 --> 00:15:30,410 S1: That would be community policing, especially in AI, because you 267 00:15:30,410 --> 00:15:34,130 S1: can have AI monitoring the city, looking at every transcript 268 00:15:34,130 --> 00:15:38,210 S1: of every stop. And you could characterize and actually have, 269 00:15:38,210 --> 00:15:41,120 S1: you know, a way to profile like, how is this 270 00:15:41,120 --> 00:15:44,450 S1: police department doing, are they doing more community management or 271 00:15:44,450 --> 00:15:50,210 S1: more like predictive enforcement, or are they rude? Are they racist? Whatever. Okay. 272 00:15:50,210 --> 00:15:55,340 S1: Apple eight open source LMS pushing the envelope for on 273 00:15:55,340 --> 00:16:00,020 S1: device text generation with models up to 3 billion parameters. 274 00:16:00,020 --> 00:16:03,980 S1: Apple doing crazy stuff in open source. Cannot wait for 275 00:16:03,980 --> 00:16:10,010 S1: iOS 18 and llama three. Llama three is huge. Yeah 276 00:16:10,010 --> 00:16:15,170 S1: 800 llama three variations on hugging face insane. It is 277 00:16:15,170 --> 00:16:19,550 S1: really good. It is really good. And I think it's 278 00:16:19,550 --> 00:16:24,320 S1: getting so good that the people like OpenAI and anthropic, 279 00:16:24,320 --> 00:16:26,600 S1: they need to be afraid. And Google need to be 280 00:16:26,600 --> 00:16:29,869 S1: very afraid of this. Because what I like about the 281 00:16:29,870 --> 00:16:33,560 S1: open source is it empowers the whole world to compete 282 00:16:33,560 --> 00:16:37,880 S1: with OpenAI and Google, right? Because if they're like, look, 283 00:16:37,880 --> 00:16:40,760 S1: this is how good it was for us. I mean, 284 00:16:40,760 --> 00:16:44,570 S1: the barriers are essentially new techniques and the size of 285 00:16:44,570 --> 00:16:48,620 S1: your infrastructure and the quality of the data that you have. Right. 286 00:16:48,620 --> 00:16:53,420 S1: So new techniques, the crowd, the group of public researchers, 287 00:16:53,420 --> 00:16:57,110 S1: they can jump ahead really quickly there. It's harder for 288 00:16:57,110 --> 00:17:00,290 S1: them to compete with one of these big companies, of course, 289 00:17:00,290 --> 00:17:03,170 S1: in the size of the cluster and how long you 290 00:17:03,170 --> 00:17:05,869 S1: could train and stuff like that, because that's super expensive. 291 00:17:05,869 --> 00:17:08,030 S1: And then the other one is data. It's also hard 292 00:17:08,030 --> 00:17:11,930 S1: to compete there, but I think the technique part might 293 00:17:11,930 --> 00:17:16,129 S1: be such a big lever that it allows some companies 294 00:17:16,130 --> 00:17:18,980 S1: to really catch up and maybe even jump ahead in 295 00:17:18,980 --> 00:17:21,800 S1: some cases. And then of course they'll get lapped the 296 00:17:21,800 --> 00:17:24,350 S1: next time the big company moves. But I feel like 297 00:17:24,560 --> 00:17:30,619 S1: community AI with open source models will move so often 298 00:17:30,619 --> 00:17:36,790 S1: and iterate so often that speed it might. Get very 299 00:17:36,790 --> 00:17:41,710 S1: close or even equal with or in some cases surpass 300 00:17:41,890 --> 00:17:44,500 S1: the big models. Now, I do think the big models 301 00:17:44,500 --> 00:17:47,590 S1: will have bigger leaps and they'll go way ahead, but 302 00:17:47,590 --> 00:17:50,379 S1: it could be that open source catches up in the 303 00:17:50,380 --> 00:17:54,130 S1: meantime just because of the speed of iteration. And I'm 304 00:17:54,160 --> 00:17:57,160 S1: guessing here, but it's kind of just an intuition. Open 305 00:17:57,160 --> 00:18:01,990 S1: voice is crushing voice cloning with the ability to mimic tone, emotion, 306 00:18:01,990 --> 00:18:07,780 S1: and even cross-lingual speech. Snowflake just shipped another LM. Simone 307 00:18:07,780 --> 00:18:11,590 S1: turns YouTube videos into blog posts US is finally getting 308 00:18:11,590 --> 00:18:14,619 S1: its first high speed rail. Elon wants to turn Tesla 309 00:18:14,619 --> 00:18:18,820 S1: cars into a distributed AI compute network, kind of like 310 00:18:19,180 --> 00:18:23,950 S1: AWS for AI, but on wheels, and Tesla's Autopilot and 311 00:18:24,070 --> 00:18:29,590 S1: self-driving are under scrutiny because there's been some deaths, and 312 00:18:29,710 --> 00:18:31,840 S1: I think this is probably going to be the case. 313 00:18:31,840 --> 00:18:35,679 S1: But I think over time, autonomous driving is going to 314 00:18:35,680 --> 00:18:38,859 S1: be better than human driving because humans get tired, humans 315 00:18:38,859 --> 00:18:42,610 S1: get distracted, humans are texting while driving or watching TikTok 316 00:18:42,609 --> 00:18:47,859 S1: while driving, and that's probably worse. Again, this is an 317 00:18:47,859 --> 00:18:51,159 S1: empirical question, so we'll have to see the actual data 318 00:18:51,160 --> 00:18:53,770 S1: and how it plays out. But that that again, is 319 00:18:53,770 --> 00:18:57,460 S1: my intuition. And yeah, example of this. We had someone 320 00:18:57,460 --> 00:19:03,700 S1: arrested for vehicular homicide after distraction with his iPhone or yeah, 321 00:19:03,700 --> 00:19:06,580 S1: I assume iPhone, but distraction with his phone laid to 322 00:19:06,580 --> 00:19:10,600 S1: a led to a fatal collision with a motorcyclist while 323 00:19:10,600 --> 00:19:14,710 S1: he was on autopilot, and Volvo brought in a Chinese 324 00:19:14,710 --> 00:19:17,949 S1: EV to the US market. Okay, Stanford found out you 325 00:19:17,950 --> 00:19:21,820 S1: could predict political leanings by looking at someone's face and 326 00:19:21,820 --> 00:19:25,510 S1: specifically the size of your face. The bigger your face, 327 00:19:25,510 --> 00:19:28,929 S1: the more likely you are to be a conservative. Evidently, 328 00:19:28,930 --> 00:19:32,469 S1: according to this paper, and the smaller your face, the 329 00:19:32,470 --> 00:19:35,440 S1: more likely you are to be a liberal. And I'm 330 00:19:35,440 --> 00:19:38,889 S1: trying to figure out like what exactly that means. How 331 00:19:38,890 --> 00:19:40,960 S1: do you measure the size of your face? I don't 332 00:19:40,960 --> 00:19:43,899 S1: get it, but whatever. I'm sure it's in the paper. 333 00:19:43,900 --> 00:19:47,230 S1: Article claims that 30% of kids 5 to 7 are 334 00:19:47,230 --> 00:19:51,790 S1: on TikTok. 5 to 7. No bueno. I am very 335 00:19:51,790 --> 00:19:55,210 S1: much in Jonathan Heights camp and you need to go 336 00:19:55,210 --> 00:19:58,030 S1: read his book. I think he is touching on some 337 00:19:58,030 --> 00:20:02,919 S1: really good stuff. FTC has banned nearly all non-compete agreements. 338 00:20:02,920 --> 00:20:05,950 S1: This is great news. I think it's going to spawn 339 00:20:05,950 --> 00:20:09,460 S1: even more innovation in AI and tech, and some are 340 00:20:09,460 --> 00:20:12,160 S1: worried about IP theft, but if you move, you'll just 341 00:20:12,160 --> 00:20:15,490 S1: steal the content. It's already illegal to steal from companies, 342 00:20:15,490 --> 00:20:18,010 S1: so you don't need a law to keep them there 343 00:20:18,010 --> 00:20:21,820 S1: so they don't steal. Like again, should be two separate laws, 344 00:20:21,820 --> 00:20:25,270 S1: one for leaving and moving and one for stealing. Stealing 345 00:20:25,270 --> 00:20:28,419 S1: is already illegal, so let's just stick with that. Supreme 346 00:20:28,420 --> 00:20:33,340 S1: court is set to decide if cities can penalize the homeless. 347 00:20:33,340 --> 00:20:36,880 S1: And I've got a vibe here. So I've been thinking 348 00:20:36,880 --> 00:20:40,389 S1: about this for a while. I think if you're mentally ill, 349 00:20:40,390 --> 00:20:43,810 S1: let's get you help or get you housed or whatever. 350 00:20:43,810 --> 00:20:46,150 S1: If you're on drugs, let's get you treated and then 351 00:20:46,180 --> 00:20:48,880 S1: see if you're mentally ill after that. If you're not 352 00:20:48,880 --> 00:20:51,580 S1: one of those, let's figure out if you actually want 353 00:20:51,580 --> 00:20:54,280 S1: to work. And if you do want to work and 354 00:20:54,280 --> 00:20:57,490 S1: you're not one of these two, then let's get you help. 355 00:20:57,490 --> 00:21:01,780 S1: Let's get you job support, training, all these different things 356 00:21:01,780 --> 00:21:05,649 S1: like interview help, right. Try to get you on your feet. 357 00:21:05,650 --> 00:21:07,960 S1: And if you're not any one of those, so you're 358 00:21:07,960 --> 00:21:11,020 S1: not on drugs, you're not mentally ill, but you don't 359 00:21:11,020 --> 00:21:14,500 S1: want to work and you don't want to participate in society. Well, 360 00:21:14,500 --> 00:21:17,560 S1: let's get you somewhere where you're not bothering other people 361 00:21:17,560 --> 00:21:21,190 S1: who are trying to be useful. And I think most 362 00:21:21,190 --> 00:21:24,520 S1: homeless are actually dealing with one and two, so let's 363 00:21:24,520 --> 00:21:28,900 S1: help them. That's our responsibility. And if it's not that 364 00:21:28,900 --> 00:21:32,170 S1: situation and a person just really doesn't want to participate 365 00:21:32,170 --> 00:21:36,340 S1: in society, they're like a Ted Kaczynski but not violent. Cool. 366 00:21:36,340 --> 00:21:38,890 S1: Let's help them go do that somewhere else, you know, 367 00:21:38,890 --> 00:21:43,390 S1: in a peaceful way. All right. Using AI for actual science, 368 00:21:43,390 --> 00:21:47,320 S1: innovation and research. So AI is now helping physicists explore 369 00:21:47,320 --> 00:21:51,610 S1: the vast possibilities of string theory. So this is great. 370 00:21:51,609 --> 00:21:55,210 S1: I mean, think about what Einstein did. Einstein went from 371 00:21:55,210 --> 00:21:59,860 S1: there not being a curved space time to just thinking 372 00:21:59,859 --> 00:22:03,250 S1: of curved space time. I think about this all the 373 00:22:03,250 --> 00:22:06,940 S1: time because I'm reading a lot of experimental physics, theoretical 374 00:22:06,940 --> 00:22:13,060 S1: physics lately, and I really think that with my prompting skills, 375 00:22:13,060 --> 00:22:15,490 S1: if I were better at the science, which I'm not, 376 00:22:15,490 --> 00:22:18,490 S1: but if I were actually, I could probably just do 377 00:22:18,490 --> 00:22:21,250 S1: this with AI without even knowing the science. If I 378 00:22:21,250 --> 00:22:24,850 S1: had a big enough sort of resources to work with 379 00:22:24,850 --> 00:22:27,700 S1: in the AI was a little bit further along. What 380 00:22:27,700 --> 00:22:30,250 S1: you do is you essentially say, hey, look, here's all 381 00:22:30,250 --> 00:22:34,480 S1: the different ways that people have crossed over and had 382 00:22:34,480 --> 00:22:38,619 S1: these creative breakthroughs. You describe the science before Einstein, you 383 00:22:38,619 --> 00:22:42,250 S1: describe what Einstein did, and you do that for multiple 384 00:22:42,250 --> 00:22:45,760 S1: different places inside of science. Right. And what this does 385 00:22:45,760 --> 00:22:49,330 S1: is it teaches it what it looks like to be creative. 386 00:22:49,330 --> 00:22:52,570 S1: And you say, look, you're looking for more things like this. 387 00:22:52,570 --> 00:22:55,690 S1: Then you feed it the current state of thinking with 388 00:22:55,690 --> 00:22:59,619 S1: quantum physics and multiple universes and all these different things. 389 00:22:59,619 --> 00:23:02,859 S1: You feed it all the math and everything, and you say, 390 00:23:02,859 --> 00:23:05,619 S1: now that you have all the current do the same 391 00:23:05,619 --> 00:23:09,100 S1: thing Einstein did, but do it for the current state now. 392 00:23:09,100 --> 00:23:11,710 S1: And the idea is it should come up with really 393 00:23:11,710 --> 00:23:15,010 S1: weird stuff. It's like, what if we're inside of a fishbowl, 394 00:23:15,010 --> 00:23:19,630 S1: inside of a mormon, you know, simulation cluster running on 395 00:23:19,630 --> 00:23:22,900 S1: Linux or whatever? It just starts making up things and 396 00:23:22,900 --> 00:23:26,229 S1: then the humans can be like, oh, well, actually, I 397 00:23:26,230 --> 00:23:28,330 S1: never thought of that. Let's go mess with that. And 398 00:23:28,330 --> 00:23:31,780 S1: then even better, and this is something that Joseph Thacker 399 00:23:31,780 --> 00:23:35,020 S1: and I talked about multiple times. It was his idea 400 00:23:35,020 --> 00:23:38,710 S1: initially was like, how do you automate not just the 401 00:23:38,710 --> 00:23:41,290 S1: having of the idea but the testing. And so we've 402 00:23:41,290 --> 00:23:43,840 S1: been working on like an infrastructure that you could potentially 403 00:23:43,840 --> 00:23:47,140 S1: do that with. And this is like the most incredible 404 00:23:47,140 --> 00:23:50,440 S1: thing I think about AI because think about this, okay. 405 00:23:50,440 --> 00:23:55,149 S1: Global warming, cancer, aging okay. People have to die at 406 00:23:55,150 --> 00:23:59,350 S1: 90 at 180, 90, 100. We describe all the reasons 407 00:23:59,350 --> 00:24:02,139 S1: that we're dying at 89 to 100. And we say, look, 408 00:24:02,140 --> 00:24:04,959 S1: we want to live longer. We want to live to 150. 409 00:24:04,990 --> 00:24:07,959 S1: How can we live to 150? And we tell it 410 00:24:07,960 --> 00:24:10,510 S1: to go and crunch and do what Einstein did, but 411 00:24:10,510 --> 00:24:12,940 S1: for aging, and you give it all the genomes, you 412 00:24:12,940 --> 00:24:15,129 S1: give it all the illnesses, you give it all the 413 00:24:15,130 --> 00:24:19,240 S1: doctor reports, you give it all the autopsies, and it's like, okay, cool. Well, 414 00:24:19,240 --> 00:24:20,830 S1: have you thought of this? If you do this to 415 00:24:20,830 --> 00:24:23,619 S1: the mitochondria, it'll slow this down and you'll live to 416 00:24:23,619 --> 00:24:26,590 S1: 180 years old. And it could just churn out these 417 00:24:26,590 --> 00:24:30,280 S1: things and even better, churn out testing for those things. 418 00:24:30,280 --> 00:24:32,260 S1: So it could be like, look, I need a lab 419 00:24:32,260 --> 00:24:36,400 S1: with the following, you know, beakers and petri dishes and 420 00:24:36,400 --> 00:24:38,350 S1: we need to combine these different things. And I think 421 00:24:38,350 --> 00:24:40,810 S1: I could make a drug that'll do this. You have 422 00:24:40,810 --> 00:24:43,750 S1: to be careful, though, when you start building things that 423 00:24:43,750 --> 00:24:46,750 S1: the AI asks you to build. But that's a separate 424 00:24:46,750 --> 00:24:51,550 S1: talk show. But just think of that automated thinking combined 425 00:24:51,550 --> 00:24:55,270 S1: with experimentation to to yield a result. I think it's 426 00:24:55,270 --> 00:24:59,740 S1: super exciting. Okay, meticulous planning and luck led to capturing 427 00:24:59,740 --> 00:25:03,550 S1: a breathtaking eclipse photo. This is the advantage of doing video. 428 00:25:03,550 --> 00:25:07,119 S1: I'm going to open it up. Look at this thing. 429 00:25:07,119 --> 00:25:11,409 S1: Look at this thing. Is that ridiculous? Like the wing 430 00:25:11,410 --> 00:25:13,870 S1: is just like perfectly there and you've got the flare 431 00:25:13,869 --> 00:25:19,000 S1: around it. I mean, that is that's gorgeous. Margaret MacDonald 432 00:25:19,000 --> 00:25:23,320 S1: argues philosophical theories are more like artful stories than scientific facts. 433 00:25:23,320 --> 00:25:25,810 S1: Quite a good essay. Here you are, what you read, 434 00:25:25,810 --> 00:25:28,510 S1: even if you don't always remember it. I call this 435 00:25:28,540 --> 00:25:33,370 S1: osmotic learning, which is basically just it absorbs into you 436 00:25:33,369 --> 00:25:36,760 S1: in a way that's not attributable to the book or 437 00:25:36,760 --> 00:25:39,430 S1: the video that you got it from. It just it 438 00:25:39,430 --> 00:25:42,070 S1: soaks into you. Right. And I think I've read close 439 00:25:42,070 --> 00:25:45,250 S1: to like 800 books and I feel myself getting smarter, 440 00:25:45,250 --> 00:25:49,420 S1: but I can't point to why my models are being updated, 441 00:25:49,420 --> 00:25:53,139 S1: which I call algorithmic learning. If you actually say Huberman 442 00:25:53,140 --> 00:25:57,310 S1: said this, therefore I'm putting that into an algorithm to 443 00:25:57,310 --> 00:26:00,010 S1: update what I do in the morning for my routine. 444 00:26:00,010 --> 00:26:03,430 S1: That's algorithmic. But if you just consume the video, you 445 00:26:03,430 --> 00:26:07,690 S1: don't write anything down, but you find yourself changing over time. 446 00:26:07,750 --> 00:26:11,859 S1: I call that osmotic for like osmosis seeps in. And 447 00:26:11,859 --> 00:26:13,989 S1: turns out the reason you're not landing that job in 448 00:26:13,990 --> 00:26:17,890 S1: 2024 might be because of ghost jobs, which are like 449 00:26:17,890 --> 00:26:20,860 S1: basically fake jobs that nobody's applying for, and they're not 450 00:26:20,859 --> 00:26:23,890 S1: real jobs. And you can actually get it. And got 451 00:26:23,890 --> 00:26:28,090 S1: an argument here that your personality and creative application strategies 452 00:26:28,090 --> 00:26:31,960 S1: can actually land you a job. So they're basically saying 453 00:26:31,960 --> 00:26:36,160 S1: it's not about skills, it's about personality and creativity. And 454 00:26:36,160 --> 00:26:38,770 S1: I think I've been talking about this quite a bit. 455 00:26:38,770 --> 00:26:42,189 S1: It's about how you broadcast yourself. It's about how you 456 00:26:42,190 --> 00:26:44,710 S1: present yourself. This is why I'm doing the whole human 457 00:26:44,710 --> 00:26:48,490 S1: 3.0 thing. So I definitely agree with this. I would 458 00:26:48,490 --> 00:26:52,090 S1: characterize it a little differently, but yeah, pretty much right. 459 00:26:52,090 --> 00:26:55,990 S1: All right. Continuous recording within ideas and analysis. So I 460 00:26:55,990 --> 00:26:58,960 S1: think we're heading towards a place where everything is recorded, 461 00:26:58,960 --> 00:27:01,480 S1: or at least where you can expect someone in a 462 00:27:01,480 --> 00:27:05,260 S1: public space is probably recording you. And I like what 463 00:27:05,260 --> 00:27:08,889 S1: that does for getting value out of conversations. But oftentimes 464 00:27:08,890 --> 00:27:11,230 S1: the value is just having the talk in the first 465 00:27:11,230 --> 00:27:14,230 S1: place and having it with somebody who you actually trust. 466 00:27:14,230 --> 00:27:16,990 S1: So I worry a lot about the fact that people 467 00:27:16,990 --> 00:27:22,180 S1: might be more guarded now when having conversations, just assuming 468 00:27:22,180 --> 00:27:24,730 S1: that it is being recorded. So I expect there to 469 00:27:24,730 --> 00:27:28,390 S1: be some sort of conversation indicator, like a red light 470 00:27:28,390 --> 00:27:30,820 S1: or a blue light or something, or a green light, 471 00:27:30,820 --> 00:27:35,080 S1: or a visual indicator on somebody's person that displays one that. 472 00:27:35,310 --> 00:27:39,389 S1: They are recording or two, that they're okay with being recorded, 473 00:27:39,390 --> 00:27:43,290 S1: or that they are not okay with being recorded. And 474 00:27:43,290 --> 00:27:45,240 S1: I think we're going to have some sort of visual 475 00:27:45,240 --> 00:27:50,100 S1: protocol like this indicates someone's comfort level. And I think 476 00:27:50,100 --> 00:27:52,379 S1: venues will also have that, like you'll walk in and 477 00:27:52,380 --> 00:27:54,270 S1: it'll be like a red light, like a stop sign 478 00:27:54,270 --> 00:27:57,149 S1: or something, and it'll be like, not cool to record 479 00:27:57,150 --> 00:28:01,320 S1: inside this building or Starbucks maybe does that for all internal, 480 00:28:01,320 --> 00:28:05,969 S1: you know, Starbucks or whatever. Now question of enforcement that's separate. 481 00:28:05,970 --> 00:28:09,389 S1: But the point is, the social contract I think is 482 00:28:09,390 --> 00:28:11,070 S1: going to be there. It's going to need to be. 483 00:28:11,070 --> 00:28:14,070 S1: So I think it's going to be interesting to see 484 00:28:14,070 --> 00:28:17,280 S1: how that all affects how we interact with each other. 485 00:28:17,280 --> 00:28:22,200 S1: One option is we stop talking and being like fresh 486 00:28:22,200 --> 00:28:25,500 S1: and open with each other. Another option is we stop 487 00:28:25,500 --> 00:28:28,980 S1: caring because there's no other ways to be embarrassed or canceled, 488 00:28:28,980 --> 00:28:32,250 S1: because it's already happened to everyone already. And another option 489 00:28:32,250 --> 00:28:36,030 S1: is that we break into pockets. So some places, like 490 00:28:36,030 --> 00:28:40,500 S1: nobody's embarrassed about anything. Everyone's recording, everything is being transcribed. 491 00:28:40,500 --> 00:28:44,760 S1: It's just radical transparency. And in other places it'll be 492 00:28:44,760 --> 00:28:48,630 S1: like maybe Germany or like the Amish or something. It's like, 493 00:28:48,630 --> 00:28:53,940 S1: no tech, no recording, complete privacy. And, you know, just 494 00:28:53,940 --> 00:28:56,310 S1: like the polar opposite. And you'll have lots of different 495 00:28:56,310 --> 00:29:01,260 S1: societies in between. Recommendation of the week. I recommend the 496 00:29:01,260 --> 00:29:03,720 S1: book of the week or actually book of the month, 497 00:29:03,720 --> 00:29:06,960 S1: actually The Courage to Be Disliked. This was the book 498 00:29:06,960 --> 00:29:09,360 S1: that we just did the book club on. It was 499 00:29:09,360 --> 00:29:15,060 S1: the book for April and it was wonderful. Absolutely loved it. 500 00:29:15,060 --> 00:29:20,610 S1: Courage to be disliked. And it's about the teachings of Adler. 501 00:29:20,970 --> 00:29:25,560 S1: It's called Adlerian philosophy or Adlerian psychology. Highly recommend the 502 00:29:25,560 --> 00:29:28,770 S1: book and the aphorism of the week be who you 503 00:29:28,770 --> 00:29:31,500 S1: are and say what you feel, because those who mind 504 00:29:31,530 --> 00:29:35,250 S1: don't matter and those who matter don't mind. Be who 505 00:29:35,280 --> 00:29:37,530 S1: you are and say what you feel, because those who 506 00:29:37,530 --> 00:29:42,540 S1: mind don't matter and those who matter don't mind. Unsupervised 507 00:29:42,540 --> 00:29:45,060 S1: learning is produced and edited by Daniel Meisler on a 508 00:29:45,060 --> 00:29:49,950 S1: Neumann U87 AI microphone using Hindenburg. Intro and outro music 509 00:29:49,950 --> 00:29:53,010 S1: is by zombie with the why and to get the 510 00:29:53,010 --> 00:29:55,080 S1: text and links from this episode, sign up for the 511 00:29:55,080 --> 00:30:00,750 S1: newsletter version of the show at Daniel meisler.com/newsletter. We'll see 512 00:30:00,750 --> 00:30:01,470 S1: you next time.