1 00:00:01,129 --> 00:00:04,460 S1: Welcome to Unsupervised Learning, a security, AI and meaning focused 2 00:00:04,460 --> 00:00:07,310 S1: podcast that looks at how best to thrive as humans 3 00:00:07,310 --> 00:00:11,540 S1: in a post AI world. It combines original ideas, analysis, 4 00:00:11,570 --> 00:00:14,780 S1: and mental models to bring not just the news, but 5 00:00:14,780 --> 00:00:22,790 S1: why it matters and how to respond. All right, welcome 6 00:00:22,790 --> 00:00:25,790 S1: to unsupervised learning. This is Daniel. Okay. I'm going to 7 00:00:25,790 --> 00:00:31,430 S1: start off with something that just happened. So strawberry just launched. 8 00:00:31,460 --> 00:00:34,670 S1: It is being called zero one, which I assume the 9 00:00:34,670 --> 00:00:38,420 S1: O might mean Orion because people were saying that it 10 00:00:38,420 --> 00:00:41,030 S1: might have been called Orion. So this is the new 11 00:00:41,030 --> 00:00:44,030 S1: model from OpenAI. And I've been messing with it for 12 00:00:44,030 --> 00:00:47,599 S1: a couple hours already. So, uh, first thing is I 13 00:00:47,600 --> 00:00:49,760 S1: gave it a task of building a business plan for 14 00:00:49,760 --> 00:00:52,370 S1: something I'm working on, and it produced output that was 15 00:00:52,370 --> 00:00:55,790 S1: far and above better than Ford or Sonnet 3.5. Yeah, 16 00:00:55,790 --> 00:01:00,360 S1: it was really quite, quite good. Very detailed. It took 17 00:01:00,360 --> 00:01:03,300 S1: quite a while. There's no streaming in the API, so 18 00:01:03,330 --> 00:01:06,450 S1: it feels a little rough compared to the current models. 19 00:01:06,450 --> 00:01:10,890 S1: But whatever that that will come with time. It's quite expensive. 20 00:01:10,890 --> 00:01:16,380 S1: So basically I did a couple of conversation analysis analyses 21 00:01:16,380 --> 00:01:21,929 S1: by passing in, um, you know, conversations like transcripts from podcasts. 22 00:01:21,930 --> 00:01:24,180 S1: And I think I did 2 or 3 of those 23 00:01:24,180 --> 00:01:27,450 S1: and it was almost a dollar. And there's also a 24 00:01:27,450 --> 00:01:31,259 S1: mini version which is way less expensive, but I'm trying 25 00:01:31,260 --> 00:01:34,050 S1: to test the capabilities, so I'm using the full model. 26 00:01:34,110 --> 00:01:38,280 S1: But yeah, a few requests for a dollar, whereas I 27 00:01:38,280 --> 00:01:43,350 S1: would say probably many dozen or a couple of hundred 28 00:01:43,350 --> 00:01:49,170 S1: requests are normally like a few dollars. So it's many 29 00:01:49,200 --> 00:01:53,580 S1: factors more expensive. So just something to consider. As with 30 00:01:53,580 --> 00:01:56,820 S1: most models, you don't need the biggest, best or latest. 31 00:01:56,820 --> 00:01:58,980 S1: This is a tweet I just put out, so I'm 32 00:01:58,980 --> 00:02:03,750 S1: going through it. So this does one particular thing well, 33 00:02:03,750 --> 00:02:08,160 S1: which is in better than anything else, which is pausing 34 00:02:08,190 --> 00:02:11,040 S1: to think and actually going step by step. That's kind 35 00:02:11,040 --> 00:02:13,440 S1: of like the magic sauce. Here is the chain of 36 00:02:13,440 --> 00:02:16,860 S1: thought reasoning. So if you don't need that for what 37 00:02:16,860 --> 00:02:19,410 S1: you're trying to do, you definitely shouldn't use this because 38 00:02:19,410 --> 00:02:22,770 S1: it's more expensive, takes longer to run. All those sorts 39 00:02:22,770 --> 00:02:25,740 S1: of reasons, this type of model and similar ones going 40 00:02:25,740 --> 00:02:30,780 S1: forward are going to massively benefit from high quality prompting. 41 00:02:30,780 --> 00:02:34,530 S1: So things like we use with fabric, which is open 42 00:02:34,530 --> 00:02:37,500 S1: source on GitHub, if you're not familiar, but you probably 43 00:02:37,500 --> 00:02:40,890 S1: are if you're listening to this. But essentially, the more 44 00:02:40,889 --> 00:02:42,570 S1: you know what you want and the better you can 45 00:02:42,570 --> 00:02:45,750 S1: articulate that, the better this is going to perform, because 46 00:02:45,780 --> 00:02:49,620 S1: it is a chain of thought sort of concept. So 47 00:02:49,620 --> 00:02:52,350 S1: the more you give it to help with that, the better. Okay, 48 00:02:52,380 --> 00:02:54,090 S1: sorry about that. I was just checking to make sure 49 00:02:54,090 --> 00:02:57,580 S1: I wasn't doxxing Doxing anyone by showing my messages, but 50 00:02:57,580 --> 00:03:02,019 S1: I was not, so I don't have to rerecord. Okay, so, um, 51 00:03:02,050 --> 00:03:07,899 S1: continuing on here and going to expand this window fully. Okay. So, um, yeah, 52 00:03:07,930 --> 00:03:10,660 S1: the better you can articulate all of this. And by 53 00:03:10,660 --> 00:03:12,730 S1: the way, I want to do an edit there for 54 00:03:12,730 --> 00:03:15,760 S1: the team. So the better you can articulate this stuff 55 00:03:15,790 --> 00:03:19,359 S1: in exactly what you want, the better things are. That's 56 00:03:19,360 --> 00:03:21,370 S1: the bottom line here. So a lot of people are 57 00:03:21,400 --> 00:03:25,720 S1: going to question is this AGI or not? Uh, Sam 58 00:03:25,720 --> 00:03:29,320 S1: Altman already responded. He's like, yeah, this absolutely is not. 59 00:03:29,530 --> 00:03:33,430 S1: So that that should end it in terms of the 60 00:03:33,430 --> 00:03:36,760 S1: actual creator of this thing saying it's not. I also 61 00:03:36,760 --> 00:03:40,120 S1: don't think it is either, whatever that matters for. But 62 00:03:40,330 --> 00:03:42,910 S1: bottom line is, anyone who's making the claim of like 63 00:03:42,940 --> 00:03:46,660 S1: this is or isn't AGI. Here's my request to the internet. Basically, 64 00:03:46,660 --> 00:03:50,710 S1: anyone claiming something is or is not should also provide 65 00:03:50,710 --> 00:03:54,589 S1: a concise and achievable definition of what that means. And 66 00:03:54,590 --> 00:03:56,960 S1: I have one here, of course, which is I've talked 67 00:03:56,960 --> 00:04:01,190 S1: about before, whether the ability of an AI, whether a 68 00:04:01,190 --> 00:04:03,740 S1: model or a product or a system to perform the 69 00:04:03,740 --> 00:04:08,750 S1: work of an average US based knowledge worker in 2002, 70 00:04:08,750 --> 00:04:12,770 S1: and I say 2002 because that's pre GPT four, right. 71 00:04:12,800 --> 00:04:16,670 S1: So basically pre AI in these terms anyway. So yeah 72 00:04:16,700 --> 00:04:20,089 S1: anyone who's talking about AGI make sure they have a definition. 73 00:04:20,089 --> 00:04:23,930 S1: Otherwise you're just wasting your time because the entire conversation 74 00:04:23,930 --> 00:04:26,719 S1: will be about definitions. And you might not even figure 75 00:04:26,720 --> 00:04:30,170 S1: that out until fucking two hours later. Sorry for the cussing. 76 00:04:30,560 --> 00:04:32,690 S1: All right. One of the most important changes to me 77 00:04:32,690 --> 00:04:35,839 S1: with this model. This. This is massive. Okay. This is 78 00:04:35,839 --> 00:04:38,510 S1: the first model that does this. It's the first model 79 00:04:38,510 --> 00:04:41,719 S1: of its kind to do this. Very, very interesting. It's 80 00:04:41,750 --> 00:04:46,580 S1: actually spending tokens to think okay, before you had input 81 00:04:46,580 --> 00:04:50,179 S1: and you had output and you were being charged in, 82 00:04:50,180 --> 00:04:52,250 S1: the amount of work that was being done was based 83 00:04:52,260 --> 00:04:55,080 S1: on the number of tokens coming in and the number 84 00:04:55,080 --> 00:04:57,660 S1: of tokens coming out, and that that was the extent 85 00:04:57,660 --> 00:04:59,940 S1: of it. What's happening now is you have tokens coming 86 00:04:59,940 --> 00:05:04,049 S1: in and you have tokens coming out, but there are 87 00:05:04,050 --> 00:05:09,990 S1: tokens being spent while it's thinking. It's actually thinking and 88 00:05:09,990 --> 00:05:13,950 S1: reasoning through how to solve the problem. And what's really 89 00:05:13,950 --> 00:05:18,150 S1: fascinating about this is that you now have multiple factors here. Okay. 90 00:05:18,180 --> 00:05:22,260 S1: So you can do better prompting. This is the next 91 00:05:22,260 --> 00:05:26,820 S1: piece here. Number seven. You could do better prompting. You 92 00:05:26,820 --> 00:05:29,729 S1: could use a smarter model. Or you could have the 93 00:05:29,730 --> 00:05:33,570 S1: model think harder on the problem. And these are all 94 00:05:33,600 --> 00:05:37,740 S1: going to be levers and knobs that we have to 95 00:05:37,770 --> 00:05:40,289 S1: get better results from AI. And this is the first 96 00:05:40,290 --> 00:05:44,400 S1: time we have this third level lever of like actually 97 00:05:44,430 --> 00:05:49,110 S1: having it think right. So at inference time, more effort 98 00:05:49,110 --> 00:05:51,570 S1: being spent. And they actually say in the blog post, 99 00:05:51,570 --> 00:05:53,670 S1: they're like, hey, look, right now it's taking, you know, 100 00:05:53,700 --> 00:05:56,310 S1: a few seconds to think or whatever, and it's going 101 00:05:56,339 --> 00:05:59,250 S1: to get back great results. But we're thinking, what if 102 00:05:59,250 --> 00:06:02,730 S1: it thinks for minutes? What if it thinks for hours? 103 00:06:02,730 --> 00:06:06,060 S1: What if it thinks for days or weeks? And not 104 00:06:06,060 --> 00:06:09,630 S1: only that, but we give it more compute power to think. 105 00:06:09,630 --> 00:06:12,360 S1: And the example they gave, I think this was an 106 00:06:12,360 --> 00:06:15,719 S1: OpenAI post. The example they gave here was how much 107 00:06:15,720 --> 00:06:18,239 S1: do you want to solve cancer? What if you could 108 00:06:18,270 --> 00:06:20,610 S1: build a data center? What if you had one data 109 00:06:20,640 --> 00:06:23,669 S1: center just for working on cancer and one data center 110 00:06:23,670 --> 00:06:28,170 S1: just for working on aging and so on? Okay. And 111 00:06:28,170 --> 00:06:31,440 S1: you basically have models like this that scale with the 112 00:06:31,440 --> 00:06:34,500 S1: inference difficulty based on the amount of difficulty of the, 113 00:06:34,529 --> 00:06:37,410 S1: of the thinking. And then of course, you have a 114 00:06:37,410 --> 00:06:40,740 S1: smart model and a good neural net and all that. Right. 115 00:06:40,770 --> 00:06:43,380 S1: Scalability of the of the neural net. So maybe that's 116 00:06:43,380 --> 00:06:47,580 S1: GPT five, GPT six, whatever. Combined with the good prompting 117 00:06:47,580 --> 00:06:52,729 S1: combined with this thinking capability and combined with, you know, 118 00:06:52,760 --> 00:06:58,549 S1: all those things unified into the combined with having that 119 00:06:58,550 --> 00:07:02,630 S1: giant infrastructure to run it so that that's insane and 120 00:07:02,630 --> 00:07:04,640 S1: the scales all the way down to like the smallest 121 00:07:04,640 --> 00:07:09,320 S1: stupid problem where it's just like whatever, GPT three and 122 00:07:09,320 --> 00:07:12,620 S1: you get back the answer almost instantaneously. In fact, forget 123 00:07:12,620 --> 00:07:16,010 S1: GPT three. It's some local model that only does one 124 00:07:16,010 --> 00:07:19,760 S1: thing well. You're spending almost no resources whatsoever. It just 125 00:07:19,760 --> 00:07:22,940 S1: goes to your phone, bounces back immediately, doesn't go anywhere, 126 00:07:22,940 --> 00:07:26,810 S1: barely costs any cycles of a GPU or a CPU 127 00:07:26,840 --> 00:07:29,720 S1: because you don't need those resources to run because it's 128 00:07:29,720 --> 00:07:32,750 S1: just an easy thing to answer. So now we're talking 129 00:07:32,750 --> 00:07:37,190 S1: about AI that scales with the difficulty of the problem, right? With, 130 00:07:37,220 --> 00:07:41,990 S1: you know, cancer, aging, getting out of the solar system, 131 00:07:41,990 --> 00:07:46,370 S1: escaping the sun, expanding, ultimately heat, death of the universe. 132 00:07:46,370 --> 00:07:50,270 S1: That's a big one, right? Because entropy kills everything. So 133 00:07:50,330 --> 00:07:52,190 S1: ultimately we're going to need a way out of here 134 00:07:52,190 --> 00:07:55,370 S1: at some point assuming we survive that long. Not happening 135 00:07:55,370 --> 00:07:58,400 S1: anytime soon. I wouldn't worry about that. But these are 136 00:07:58,400 --> 00:08:01,010 S1: the types of things that are really exciting. You know, 137 00:08:01,040 --> 00:08:04,310 S1: the size of the problem being being a factor, for 138 00:08:04,310 --> 00:08:07,250 S1: which I you point at it with lots and lots 139 00:08:07,250 --> 00:08:11,900 S1: of different knobs and levers controlling that decision. So I 140 00:08:11,900 --> 00:08:15,080 S1: think that's really cool. Another important thing to mention is 141 00:08:15,080 --> 00:08:18,530 S1: that the innovation seems independent of what we were waiting 142 00:08:18,530 --> 00:08:22,400 S1: for for GPT five. So based on all I read, 143 00:08:22,400 --> 00:08:25,460 S1: all the releases from OpenAI, and I've seen all the 144 00:08:25,460 --> 00:08:28,130 S1: rumors and, you know, talked to a bunch of people 145 00:08:28,160 --> 00:08:32,929 S1: who've been speculating about this. And this seems completely independent from, oh, 146 00:08:32,960 --> 00:08:35,870 S1: is this GPT four oh? Is it for zero? Is 147 00:08:35,870 --> 00:08:39,709 S1: it five? Is it an early version of five? Doesn't 148 00:08:39,710 --> 00:08:42,680 S1: really matter. It's like a separate axis. This is like 149 00:08:42,679 --> 00:08:46,340 S1: a capability. This is like thinking Capability, which is on 150 00:08:46,340 --> 00:08:50,179 S1: a separate axis from how big or smart is the 151 00:08:50,179 --> 00:08:53,330 S1: neural net, right? Or how big or smart is the 152 00:08:53,360 --> 00:08:56,870 S1: is the model. So really, really cool to think about 153 00:08:56,870 --> 00:08:59,270 S1: those being. Two separate things because now we can start 154 00:08:59,270 --> 00:09:02,569 S1: thinking about okay, well if GPT five is still going. 155 00:09:02,600 --> 00:09:05,059 S1: To come out, you know, later this year, beginning in 156 00:09:05,059 --> 00:09:07,490 S1: next year or whenever it's going to come out. And 157 00:09:07,490 --> 00:09:10,670 S1: whatever they're going to call it, well, imagine GPT five 158 00:09:10,670 --> 00:09:14,750 S1: with this thinking capability. That's cool. So presumably this is 159 00:09:14,750 --> 00:09:17,660 S1: just a feature that you can add onto any model, 160 00:09:17,809 --> 00:09:20,840 S1: which is what we're just talking about. And I think 161 00:09:20,840 --> 00:09:24,440 S1: this is okay. This is really, really crucial here. I've 162 00:09:24,440 --> 00:09:26,630 S1: been talking for a long time about slack in the 163 00:09:26,630 --> 00:09:30,710 S1: rope and tricks that we're going to use to jump 164 00:09:30,710 --> 00:09:35,150 S1: ahead in, um, advancement of AI, so so check this out. 165 00:09:35,179 --> 00:09:37,160 S1: A lot of people are like, oh, we're running into 166 00:09:37,160 --> 00:09:40,309 S1: a data wall. Neural nets are only so good they 167 00:09:40,309 --> 00:09:42,950 S1: can only get so good. We've already hit a thing. 168 00:09:42,950 --> 00:09:45,660 S1: I mean, so many, so many people are saying things 169 00:09:45,660 --> 00:09:50,790 S1: like this that just sound absolutely ridiculous to me. First 170 00:09:50,790 --> 00:09:53,820 S1: of all, they were the ones saying we wouldn't be here. 171 00:09:53,820 --> 00:09:58,110 S1: And so now we are here and everyone's surprised and 172 00:09:58,110 --> 00:10:00,660 S1: they're like, well, here's what we know for sure is 173 00:10:00,690 --> 00:10:02,790 S1: we're not going to get any better. How can I 174 00:10:02,790 --> 00:10:07,050 S1: believe you if you didn't predict any of this and 175 00:10:07,050 --> 00:10:10,740 S1: you were absolutely certain back then, and now you're absolutely 176 00:10:10,740 --> 00:10:14,099 S1: certain it's not going to jump ahead again, right? Leopold 177 00:10:14,130 --> 00:10:17,460 S1: talks about this in his paper. There's lots of different 178 00:10:17,460 --> 00:10:21,569 S1: ways to get better. There's the architecture of the model. 179 00:10:21,570 --> 00:10:24,239 S1: There's the size of the model. I forget what all 180 00:10:24,240 --> 00:10:26,790 S1: levers he had, but it's the architecture of the model, 181 00:10:26,820 --> 00:10:28,560 S1: the size of the model. And I think it was 182 00:10:28,559 --> 00:10:32,340 S1: hobbling was the other one, which is what I called 183 00:10:32,520 --> 00:10:35,460 S1: like a year ago. Slack in the rope or tricks 184 00:10:35,460 --> 00:10:37,080 S1: we're going to. This is what I told a friend 185 00:10:37,080 --> 00:10:39,630 S1: of mine who's really smart in this stuff. I said, 186 00:10:39,630 --> 00:10:43,360 S1: watch this. We're going to find multiple tricks where we're 187 00:10:43,360 --> 00:10:46,810 S1: messing around in percentage points, and then we find a 188 00:10:46,809 --> 00:10:49,700 S1: thing and it jumps us 2 or 3 or 5 189 00:10:49,700 --> 00:10:53,560 S1: or 10 x or 100 x ahead. And I actually 190 00:10:53,559 --> 00:10:56,890 S1: learned this from him. Uh, I actually learned this from him. 191 00:10:56,890 --> 00:10:58,990 S1: He was like, hey, you know, there are things that 192 00:10:58,990 --> 00:11:02,170 S1: jump you ahead. Um, and I think he gave me 193 00:11:02,170 --> 00:11:05,230 S1: example from some public paper or whatever. And it was 194 00:11:05,230 --> 00:11:08,350 S1: an example of like a big jump. And my natural 195 00:11:08,350 --> 00:11:11,530 S1: intuition was, there's going to be a lot more of those. 196 00:11:11,530 --> 00:11:15,429 S1: And they're not coming from pursuing along this axis, which 197 00:11:15,429 --> 00:11:18,250 S1: is difficult. They are actually just hanging off to the side. 198 00:11:18,250 --> 00:11:20,530 S1: It's like, oh, did you know if you just changed 199 00:11:20,530 --> 00:11:22,450 S1: the color of this? Hey, did you know if you 200 00:11:22,450 --> 00:11:25,989 S1: just orient the data backward instead of forward? Hey, did 201 00:11:25,990 --> 00:11:28,000 S1: you know if you just prune the data in this 202 00:11:28,000 --> 00:11:31,480 S1: way or if you add this particular data set or. 203 00:11:31,510 --> 00:11:35,170 S1: And I'm just making up these examples, but simple things 204 00:11:35,170 --> 00:11:38,710 S1: that you wouldn't think would work. And this is why 205 00:11:38,740 --> 00:11:42,860 S1: Leopold talks about if you automate an AI engineer or 206 00:11:42,860 --> 00:11:45,829 S1: an AI researcher is what he called it. That's when 207 00:11:45,830 --> 00:11:48,859 S1: it gets completely silly, because they have the ability to 208 00:11:48,890 --> 00:11:50,870 S1: now go and try a whole bunch of these things, 209 00:11:50,870 --> 00:11:54,140 S1: including these tricks. All this to say that the slack 210 00:11:54,140 --> 00:11:57,860 S1: in the rope or this series of tricks is going 211 00:11:57,860 --> 00:12:01,580 S1: to keep multiplying our advances. And that's at the same 212 00:12:01,580 --> 00:12:05,750 S1: time that we're working on the algorithms. Oh, that was 213 00:12:05,750 --> 00:12:08,450 S1: the other. That was the other factor is algorithms. That 214 00:12:08,450 --> 00:12:10,339 S1: was this is going to happen. At the same time 215 00:12:10,340 --> 00:12:13,160 S1: we're working on the algorithms to make those better. We're 216 00:12:13,160 --> 00:12:16,100 S1: also working on the size of the neural net and 217 00:12:16,100 --> 00:12:18,559 S1: the quality and the structure. And everything about the neural 218 00:12:18,559 --> 00:12:20,989 S1: net is going to get bigger and more powerful, but 219 00:12:20,990 --> 00:12:24,170 S1: mostly just a matter of size, number of parameters. But 220 00:12:24,170 --> 00:12:26,930 S1: all those things are changing at the same time as 221 00:12:26,960 --> 00:12:31,189 S1: we're finding all these tricks. Right? So we're talking about 222 00:12:31,910 --> 00:12:35,420 S1: this is just begun. And this is what people don't realize. 223 00:12:35,420 --> 00:12:37,970 S1: This is just now starting. We're going to look back 224 00:12:37,970 --> 00:12:40,770 S1: in two years and be like, what was that? That 225 00:12:40,770 --> 00:12:44,550 S1: was silly. Right. And so I really want to warn 226 00:12:44,580 --> 00:12:47,670 S1: people against thinking we're hitting some kind of a wall. 227 00:12:47,700 --> 00:12:50,459 S1: Think of it this way. We just found alien technology. 228 00:12:50,490 --> 00:12:53,370 S1: We have no idea how it works. And we're, like, 229 00:12:53,400 --> 00:12:56,370 S1: poking it with a stick, and it's already spitting out 230 00:12:56,370 --> 00:12:59,520 S1: amazing things. So think about that. Okay, we got a 231 00:12:59,520 --> 00:13:02,219 S1: glowy ball. We don't know how it floats. We don't 232 00:13:02,250 --> 00:13:05,729 S1: know how it's doing. Anti-Gravity, right? We don't know how 233 00:13:05,730 --> 00:13:08,250 S1: it's doing this. We don't know how it's reflecting its surface. 234 00:13:08,280 --> 00:13:10,350 S1: We don't know how it's coming up with these answers. 235 00:13:10,350 --> 00:13:12,570 S1: We don't know how it got here from the other 236 00:13:12,570 --> 00:13:15,570 S1: solar system. We don't know anything about it. You poke 237 00:13:15,570 --> 00:13:18,270 S1: it with a stick and it tells this magic stuff 238 00:13:18,270 --> 00:13:22,830 S1: and we're like, Holy crap, that's amazing. Somebody walks up, 239 00:13:22,860 --> 00:13:25,110 S1: sees you poke it with a stick and goes, yeah, 240 00:13:25,110 --> 00:13:27,630 S1: that's I mean, that's that's all it's ever going to 241 00:13:27,660 --> 00:13:30,059 S1: be able to do. I mean, I've seen you poke 242 00:13:30,059 --> 00:13:33,390 S1: it with a stick twice, and it gave you kind 243 00:13:33,390 --> 00:13:36,479 S1: of a similar answer, which means that's all we could 244 00:13:36,480 --> 00:13:40,110 S1: learn from this alien ball. That's their conclusion. I am 245 00:13:40,110 --> 00:13:43,200 S1: certain that since you poked it with a stick while 246 00:13:43,200 --> 00:13:45,780 S1: I was standing here three times, and it kind of 247 00:13:45,809 --> 00:13:50,010 S1: gave you a similar answer. One it must be stupid. Two, 248 00:13:50,040 --> 00:13:53,069 S1: it's not as smart as us. And three, this is 249 00:13:53,100 --> 00:13:56,250 S1: as as smart as it's ever going to be. This 250 00:13:56,250 --> 00:13:58,770 S1: is the most it has to offer. That is the 251 00:13:58,770 --> 00:14:02,580 S1: claim that's being made by these kind of like denialists, 252 00:14:02,610 --> 00:14:06,900 S1: in my view. And that doesn't mean the current shiny 253 00:14:06,900 --> 00:14:11,550 S1: ball is better than humans, or it should replace humans, 254 00:14:11,550 --> 00:14:14,460 S1: or it could do everything we could do. Like, this 255 00:14:14,460 --> 00:14:16,860 S1: is not a competition. Okay, here's a better way to 256 00:14:16,890 --> 00:14:19,830 S1: think about this. This is not like a rock that 257 00:14:19,830 --> 00:14:22,710 S1: we have animated. Think of it this way. If an 258 00:14:22,710 --> 00:14:26,100 S1: alien comes here because someone else was like, hey, this 259 00:14:26,100 --> 00:14:29,820 S1: is not thinking, this is processing. And I'm like, come on, 260 00:14:29,850 --> 00:14:33,270 S1: come on. If you if an alien comes here, let's 261 00:14:33,270 --> 00:14:36,610 S1: assume we know how our brain works. An alien comes 262 00:14:36,610 --> 00:14:39,220 S1: here and we look at its brain, or it shows 263 00:14:39,220 --> 00:14:42,430 S1: us its brain, and it looks different. And we're like, oh, 264 00:14:42,460 --> 00:14:47,290 S1: you guys do neurons and synapses different than us? Who's 265 00:14:47,290 --> 00:14:49,870 S1: going to walk over and be like, well, since they're 266 00:14:49,870 --> 00:14:53,710 S1: doing neurons and synapses different than us, they're not thinking. 267 00:14:53,710 --> 00:14:57,670 S1: Only humans can think. And I'm like, they got here. 268 00:14:57,700 --> 00:15:00,460 S1: They got here, didn't they? It's a little shiny ball. 269 00:15:00,460 --> 00:15:03,190 S1: And they got here from whatever part of the galaxy 270 00:15:03,190 --> 00:15:06,850 S1: or universe that they came from. They're obviously doing something right. 271 00:15:06,850 --> 00:15:10,270 S1: And AI is obviously doing something right too. So I 272 00:15:10,270 --> 00:15:13,120 S1: think it's a little bit specious. Is that is that 273 00:15:13,120 --> 00:15:17,590 S1: the name of the word? It's like specious to just 274 00:15:17,590 --> 00:15:21,040 S1: magically assume that we are the best. Only we are 275 00:15:21,070 --> 00:15:25,120 S1: thinking only we are special. Instead of thinking like we 276 00:15:25,120 --> 00:15:29,440 S1: might have this nascent alien intelligence thing going on that 277 00:15:29,440 --> 00:15:32,320 S1: actually is doing things that are very much analogous to us. 278 00:15:32,350 --> 00:15:34,570 S1: It reminds me of the first time that I clicked 279 00:15:34,570 --> 00:15:37,600 S1: around inside of Linux. This is like late 90s. I 280 00:15:37,600 --> 00:15:40,990 S1: was messing with Linux. This must have been like 9798 281 00:15:40,990 --> 00:15:44,800 S1: or something. I'm messing with Linux and I'm clicking around 282 00:15:44,830 --> 00:15:47,350 S1: because I had started with windows and I'm like, oh, 283 00:15:47,350 --> 00:15:49,990 S1: it opens windows and it opens things that I could 284 00:15:49,990 --> 00:15:53,320 S1: click and navigate. Then I'm like, it's it's just like 285 00:15:53,320 --> 00:15:58,210 S1: on Windows Explorer. And this like, blew me away. It 286 00:15:58,240 --> 00:16:01,060 S1: absolutely blew me away that this was just a different 287 00:16:01,060 --> 00:16:04,330 S1: way of doing the same thing. And that underneath this, 288 00:16:04,330 --> 00:16:07,420 S1: there's a universal thing of you need to be able 289 00:16:07,420 --> 00:16:10,210 S1: to browse files, you need to be able to open windows, 290 00:16:10,210 --> 00:16:12,970 S1: you need to be able to close windows. And that 291 00:16:12,970 --> 00:16:14,980 S1: clicked for me. And I'm like, oh, I guess like 292 00:16:15,010 --> 00:16:17,830 S1: all operating systems are going to do this differently. It's 293 00:16:17,830 --> 00:16:20,260 S1: the same with aliens. It's the same with like they 294 00:16:20,260 --> 00:16:23,620 S1: might think differently, but whatever. They have to think, right. 295 00:16:23,620 --> 00:16:27,070 S1: So why would we expect this synthetic intelligence that we've 296 00:16:27,100 --> 00:16:29,800 S1: birthed to do it exactly the same way that we 297 00:16:29,830 --> 00:16:34,040 S1: way that we do? We should not expect that we 298 00:16:34,070 --> 00:16:39,410 S1: got here accidentally stumbling through time due to evolution. And 299 00:16:39,410 --> 00:16:42,950 S1: we've got this version that we have and it's awesome, obviously. 300 00:16:42,950 --> 00:16:46,760 S1: But like, that's way different than we invented this thing 301 00:16:46,790 --> 00:16:50,870 S1: five years ago or whenever that was 2017, six years ago. 302 00:16:50,900 --> 00:16:53,240 S1: And I know it goes further back than that. But 303 00:16:53,270 --> 00:16:56,900 S1: you know what I'm saying? Transformers. All right. So that's that. 304 00:16:56,900 --> 00:17:00,350 S1: And this this is becoming a long thing. But whatever 305 00:17:00,350 --> 00:17:04,220 S1: we'll go with it. So yeah, basically we have no 306 00:17:04,520 --> 00:17:08,149 S1: idea how early all of this is. We're likely to 307 00:17:08,180 --> 00:17:11,330 S1: find ten, 20 or 200 more of these holy crap 308 00:17:11,359 --> 00:17:15,320 S1: optimizations like this thinking thing before we start hitting any 309 00:17:15,320 --> 00:17:21,950 S1: limits for neural network architecture or the transformer transformer like. 310 00:17:21,980 --> 00:17:25,189 S1: Plus we could just find something better than a transformer. 311 00:17:25,190 --> 00:17:28,669 S1: You realize how how lucky we were to find the 312 00:17:28,670 --> 00:17:32,410 S1: transformer like the people who made that paper, they're like, hey, 313 00:17:32,410 --> 00:17:34,600 S1: this is this is a cool way. We think this 314 00:17:34,600 --> 00:17:36,520 S1: is a cool way of doing something. They didn't know 315 00:17:36,520 --> 00:17:39,640 S1: what they had. Okay, you should watch a Karpathy talk 316 00:17:39,670 --> 00:17:42,909 S1: about the transformer. He's like, this thing is a general 317 00:17:42,910 --> 00:17:46,780 S1: purpose computer. This thing is insanely good at learning. He 318 00:17:46,780 --> 00:17:50,800 S1: talks about different ways that it's better than humans at learning. Okay, 319 00:17:50,830 --> 00:17:55,540 S1: some some people randomly found this thing and it shot 320 00:17:55,540 --> 00:17:58,420 S1: us off. Okay. So so check this out. This is 321 00:17:58,420 --> 00:18:03,340 S1: another example of finding tricks or slack in the rope 322 00:18:03,340 --> 00:18:06,250 S1: just lying on the ground. So we stumble through AI 323 00:18:06,280 --> 00:18:10,660 S1: for decades and decades and decades. And then someone's like, hey, 324 00:18:10,660 --> 00:18:13,120 S1: this is kind of cool about this attention mechanism. Hey, 325 00:18:13,119 --> 00:18:16,899 S1: what do you think about this architecture for a neural net? Boom, 326 00:18:16,900 --> 00:18:21,129 S1: now we have this take off. There's nothing saying somebody 327 00:18:21,130 --> 00:18:24,010 S1: isn't going to be like, I like what you did 328 00:18:24,010 --> 00:18:28,950 S1: with that transformer architecture. What if it looked like this instead, 329 00:18:28,980 --> 00:18:32,820 S1: it might be 20 times better. It might be 2000 330 00:18:32,820 --> 00:18:36,210 S1: times better. It might be 4% better. It doesn't matter. 331 00:18:36,240 --> 00:18:40,680 S1: Like we have only just begun. We have only just begun. 332 00:18:40,680 --> 00:18:44,310 S1: I can absolutely guarantee you that assuming we don't kill 333 00:18:44,310 --> 00:18:47,040 S1: ourselves off as a result of this, like that would 334 00:18:47,040 --> 00:18:49,350 S1: set things back. But I'm trying to get you to 335 00:18:49,380 --> 00:18:54,210 S1: think about things in this way, because it's insane what's 336 00:18:54,210 --> 00:18:56,130 S1: about to happen. And yeah, I'm gonna I'm gonna have 337 00:18:56,130 --> 00:18:58,380 S1: more examples here. I'm working on an example right here 338 00:18:58,380 --> 00:19:01,320 S1: on this other screen. Uh, pretty cool thing I'm building 339 00:19:01,320 --> 00:19:04,800 S1: with it. Um, okay, so that was that. All right, 340 00:19:04,830 --> 00:19:09,060 S1: now back to the show. Okay. We are adding a 341 00:19:09,060 --> 00:19:12,300 S1: whole bunch more content to the podcast. So if you 342 00:19:12,330 --> 00:19:15,389 S1: have not been listening to the podcast, I used to 343 00:19:15,390 --> 00:19:19,170 S1: do this really dry thing. Um, and sometimes I'm still dry, 344 00:19:19,200 --> 00:19:22,710 S1: so whatever. I'm an intellectual guy. Sometimes I just like, 345 00:19:22,740 --> 00:19:25,500 S1: talk about the idea. I'm not overly excited, like the 346 00:19:25,500 --> 00:19:28,280 S1: way I was just now, and it could be a 347 00:19:28,280 --> 00:19:30,709 S1: little boring. I mean, I started this thing back in 348 00:19:30,710 --> 00:19:34,010 S1: 2015 and I was literally like, welcome to the Take 349 00:19:34,040 --> 00:19:37,969 S1: One Security podcast. The podcast where I do everything and 350 00:19:37,970 --> 00:19:41,030 S1: take 1 in 1 take. That's all I do. I 351 00:19:41,030 --> 00:19:45,770 S1: just read the news. Okay. First news. Now that was fire. 352 00:19:45,800 --> 00:19:50,450 S1: The podcast blew up. Why? Because no one had a podcast. Well, 353 00:19:50,450 --> 00:19:54,080 S1: there was Risky Biz. There was like GRC. What's his name? 354 00:19:54,080 --> 00:19:58,310 S1: Steve Gibson. There was that there was a couple, but very, 355 00:19:58,310 --> 00:20:03,500 S1: very few. And actually mine was dry and to the point. 356 00:20:03,500 --> 00:20:07,400 S1: So it was like super exciting. Anyway, that that is 357 00:20:07,400 --> 00:20:11,360 S1: the old version that is no longer interesting. People got 358 00:20:11,359 --> 00:20:14,180 S1: tired of that, and I lost interest in it because 359 00:20:14,180 --> 00:20:16,369 S1: I just didn't want to talk about the news anymore. 360 00:20:16,400 --> 00:20:19,700 S1: I wanted to move on to other things. So it 361 00:20:19,700 --> 00:20:22,790 S1: is now what it is. The point is, the thing 362 00:20:22,790 --> 00:20:25,160 S1: that I just did about AI, Outcry about the release 363 00:20:25,160 --> 00:20:29,000 S1: of strawberry. That is now a piece which will be 364 00:20:29,000 --> 00:20:31,370 S1: cut out by the team. The team will put that 365 00:20:31,369 --> 00:20:34,250 S1: on YouTube, and the team will put out that out 366 00:20:34,250 --> 00:20:38,149 S1: on the podcast. And when I do little clips like that, 367 00:20:38,180 --> 00:20:40,490 S1: I don't know that I'm in a clip in the moment. 368 00:20:40,490 --> 00:20:43,159 S1: But what happens is they take that and put it 369 00:20:43,160 --> 00:20:45,560 S1: into its own piece. They name it, they do a 370 00:20:45,560 --> 00:20:48,380 S1: thumbnail maybe. I'm not sure about that, but they put 371 00:20:48,380 --> 00:20:51,110 S1: it out there and now that's a piece of content 372 00:20:51,109 --> 00:20:55,280 S1: and oftentimes it's way easier to listen to those. I 373 00:20:55,280 --> 00:20:57,649 S1: would prefer to listen to those. Again, I'm building the 374 00:20:57,650 --> 00:21:01,400 S1: thing that I believe needs to exist because it's what 375 00:21:01,400 --> 00:21:04,159 S1: I wish I was getting from other people. Right? So 376 00:21:04,190 --> 00:21:07,580 S1: Alex Hormozi, Chris Williamson, all these different people, I love 377 00:21:07,580 --> 00:21:10,850 S1: the fact that I can get a single discrete piece 378 00:21:10,850 --> 00:21:13,790 S1: of content, and I don't mean like a YouTube short, 379 00:21:13,820 --> 00:21:17,780 S1: like 30s or a minute. I mean something like, I 380 00:21:17,780 --> 00:21:20,129 S1: don't care if it's 30s or if it's like 3 381 00:21:20,130 --> 00:21:24,090 S1: or 4 minutes. I just want a discrete idea labeled 382 00:21:24,090 --> 00:21:26,880 S1: as that discrete idea. So that's what I've been doing. 383 00:21:26,880 --> 00:21:31,020 S1: I've been taking content either making it independently or if 384 00:21:31,020 --> 00:21:33,960 S1: it happens to come out like it just did. Talking 385 00:21:33,960 --> 00:21:37,080 S1: about strawberry inside of a big episode that is then 386 00:21:37,080 --> 00:21:41,609 S1: brought out and turned into a full thing. Now, the 387 00:21:41,609 --> 00:21:44,520 S1: advantage here is that the podcast is now going to 388 00:21:44,520 --> 00:21:48,600 S1: be full of content. There's now going to be three, four, five, 389 00:21:48,600 --> 00:21:54,480 S1: six pieces of hopefully decent pieces of content inside the 390 00:21:54,480 --> 00:21:58,410 S1: podcast feed. So if you're a subscriber to the podcast, 391 00:21:58,440 --> 00:22:01,140 S1: which you should definitely go do on Apple or Spotify 392 00:22:01,170 --> 00:22:04,890 S1: or whatever your client, um, you will not only have 393 00:22:04,890 --> 00:22:09,000 S1: this big episode, which, when I mess around and start 394 00:22:09,000 --> 00:22:12,240 S1: ranting like this, could end up being 30 minutes long, right? 395 00:22:12,570 --> 00:22:15,060 S1: But in addition to those, you're going to have these 396 00:22:15,060 --> 00:22:18,869 S1: really clean little pieces of content that's like a single idea. 397 00:22:18,900 --> 00:22:22,689 S1: So bottom line is get back on the podcast, whether 398 00:22:22,690 --> 00:22:27,790 S1: that's on YouTube or on, um, audio. Definitely. Please go 399 00:22:27,790 --> 00:22:33,700 S1: subscribe on the audio version. So Apple Podcasts, Spotify, whatever, Android. 400 00:22:33,700 --> 00:22:37,179 S1: And uh, yeah, tune in because you're probably going to 401 00:22:37,210 --> 00:22:40,210 S1: like it a lot. Okay, new keyboard I'm using. Okay. 402 00:22:40,240 --> 00:22:42,760 S1: How do I type without messing anything up? Um, let's 403 00:22:42,760 --> 00:22:45,340 S1: see here. Okay, I'll just do. Okay. I'm going to 404 00:22:45,340 --> 00:22:47,620 S1: type fast. I don't know if you can hear this. 405 00:22:47,619 --> 00:22:51,790 S1: I got a Sennheiser here listening. Okay. This keyboard is 406 00:22:51,790 --> 00:22:56,890 S1: called the Orlev F 75. I love this keyboard. And 407 00:22:56,890 --> 00:22:59,830 S1: it's actually super cheap. It's like 70 bucks. I've purchased 408 00:22:59,830 --> 00:23:03,340 S1: so many expensive keyboards that are sitting over there and 409 00:23:03,369 --> 00:23:06,340 S1: I don't use them. I was using this one for 410 00:23:06,340 --> 00:23:08,889 S1: a while, but I heard this sound on a YouTube 411 00:23:08,890 --> 00:23:11,590 S1: video in a review and I was like, ooh, I 412 00:23:11,590 --> 00:23:14,230 S1: like that sound. Hope it sounds like that in person. 413 00:23:14,230 --> 00:23:17,290 S1: And yeah, it does. It sounds like that in person. 414 00:23:17,290 --> 00:23:21,110 S1: Hopefully the post-production will not take out that sound, but 415 00:23:21,619 --> 00:23:24,680 S1: I love the sound of this thing. This particular one 416 00:23:24,680 --> 00:23:29,330 S1: has no key. Uh, has no labels on the top 417 00:23:29,330 --> 00:23:31,280 S1: of the keys. They're actually on the bottom of the 418 00:23:31,310 --> 00:23:34,129 S1: keys and they're backlit, so it's pretty cool. So the 419 00:23:34,130 --> 00:23:38,120 S1: whole keyboard looks kind of gray. It's really heavy. I 420 00:23:38,119 --> 00:23:40,760 S1: like it, but mostly I like the sound and the feel. 421 00:23:40,760 --> 00:23:46,280 S1: So combine that with Neovim config text based world. I'm happy. 422 00:23:46,310 --> 00:23:49,250 S1: I'm a happy person. Okay, continue to be blown away 423 00:23:49,250 --> 00:23:52,280 S1: by the idea of encapsulating what people think the biggest 424 00:23:52,280 --> 00:23:57,020 S1: problem in the world is using extract. Primary problem I 425 00:23:57,020 --> 00:24:01,250 S1: love this thing. So echo Viktor Frankl's work, pipe that 426 00:24:01,250 --> 00:24:04,550 S1: into extract primary problem and look what it brings back. 427 00:24:04,580 --> 00:24:07,940 S1: The lack of meaning in life leads to suffering and 428 00:24:07,940 --> 00:24:13,070 S1: existential despair. That is beautiful, that is beautiful. And I'm 429 00:24:13,070 --> 00:24:16,910 S1: taking someone's entire body of work and pulling out what 430 00:24:16,910 --> 00:24:20,840 S1: they think the biggest problem in the world is. That's 431 00:24:20,840 --> 00:24:24,110 S1: just astounding to me. Yeah. So the keynote at Sans 432 00:24:24,109 --> 00:24:28,010 S1: went really well. Almost 30 minutes of questions afterwards was 433 00:24:28,010 --> 00:24:30,740 S1: really cool for them to have me appreciate them. Yeah, 434 00:24:30,740 --> 00:24:35,389 S1: I'm experimenting with some micro art fiction. And Tim from 435 00:24:35,390 --> 00:24:37,310 S1: community was like, hey, what are you working on? You're 436 00:24:37,310 --> 00:24:42,470 S1: working on a book? Uh, not really. Kind of. That's 437 00:24:42,470 --> 00:24:47,210 S1: all I'll say about that. Not actively and fiercely, I 438 00:24:47,210 --> 00:24:50,630 S1: will say that. But yeah, some ideas out there working 439 00:24:50,630 --> 00:24:53,150 S1: on a bunch of flagship content. Right now, I'm mentioning 440 00:24:53,150 --> 00:24:55,760 S1: more and more about human 3.0. So I'm going to 441 00:24:55,760 --> 00:24:59,600 S1: put out like a description of that, like the structure 442 00:24:59,600 --> 00:25:02,780 S1: and what I'm working on there. Uh, a new piece 443 00:25:02,780 --> 00:25:06,290 S1: on security asset management in AI, how to write fiction 444 00:25:06,290 --> 00:25:09,020 S1: using AI. That one's going to be insane and a 445 00:25:09,020 --> 00:25:11,960 S1: whole bunch of other ones. And all right, let's get 446 00:25:11,960 --> 00:25:16,580 S1: into security. Chinese government backed hackers have been infiltrating US 447 00:25:16,580 --> 00:25:20,149 S1: internet service providers to spy on users. According to a 448 00:25:20,150 --> 00:25:24,740 S1: bunch of researchers, Halliburton confirmed a cyber attack where intruders 449 00:25:24,770 --> 00:25:29,000 S1: access an exfiltrated data and they blame the Ransom hub group, 450 00:25:29,030 --> 00:25:35,780 S1: who also claimed responsibility. Predator spyware predator spyware is back 451 00:25:35,780 --> 00:25:38,330 S1: with new features that make it even harder to track, 452 00:25:38,330 --> 00:25:41,780 S1: and it looks like it's showing up in Democratic Republic 453 00:25:41,780 --> 00:25:45,530 S1: of Congo and Angola, and there's a lot more anonymity 454 00:25:45,530 --> 00:25:48,739 S1: built into it. So the trick with this kind of 455 00:25:48,740 --> 00:25:51,830 S1: malware is that it's not just the people who make it, 456 00:25:51,830 --> 00:25:54,680 S1: it's the people who sell it. And the fact that 457 00:25:54,680 --> 00:25:59,060 S1: there's a market for governments who want to control their populations. 458 00:25:59,060 --> 00:26:03,350 S1: And that's what's especially scary about, like the CCP way 459 00:26:03,350 --> 00:26:06,350 S1: of viewing the world. And if you combine that with AI, 460 00:26:06,350 --> 00:26:10,790 S1: that can kind of control things and see everything and like, 461 00:26:10,820 --> 00:26:15,179 S1: determine who's breaking the rules and monitor and surveil and 462 00:26:15,180 --> 00:26:18,600 S1: things like that. And you combine those things together and 463 00:26:18,630 --> 00:26:22,710 S1: you do have a recipe for controlling populations. Latest version 464 00:26:22,710 --> 00:26:27,990 S1: of NIST, CSF, which is two, introduces governance as a 465 00:26:27,990 --> 00:26:32,189 S1: new step and focuses on continuous improvement to adapt to 466 00:26:32,220 --> 00:26:36,689 S1: emerging threats. And it also they also released a continuous 467 00:26:36,690 --> 00:26:42,360 S1: threat exposure management framework, CDM framework. Thanks to Hyper Proof 468 00:26:42,359 --> 00:26:47,400 S1: for sponsoring Maltese security, researchers have been charged after discovering 469 00:26:47,400 --> 00:26:52,500 S1: a flaw in an application called Free Hour. And I 470 00:26:52,500 --> 00:26:55,109 S1: think they are in jail. Yeah, they're in jail waiting 471 00:26:55,109 --> 00:26:59,160 S1: for trial, I believe, because they don't have like, oh, disclosure, 472 00:26:59,250 --> 00:27:03,150 S1: public disclosure or, you know, good citizen or anything like that. 473 00:27:03,150 --> 00:27:05,399 S1: It's like, nope, you did a bad thing. You're in jail. 474 00:27:05,430 --> 00:27:08,219 S1: US Space Force is gearing up for potential conflicts in 475 00:27:08,220 --> 00:27:12,240 S1: space with countries like China and Russia, and they're trying 476 00:27:12,240 --> 00:27:17,260 S1: to protect the stuff in space. Obviously satellites and other 477 00:27:17,260 --> 00:27:21,580 S1: space assets. And thanks to dropzone for sponsoring as well. 478 00:27:21,580 --> 00:27:24,550 S1: In the US, is offering a $10 million reward for 479 00:27:24,550 --> 00:27:28,000 S1: information on the Russian hacking group called Cadet Blizzard, linked 480 00:27:28,000 --> 00:27:34,030 S1: to Gru's unit 29,155, which has been particularly focused on 481 00:27:34,030 --> 00:27:37,210 S1: disrupting aid to Ukraine. NASA is launching a new podcast 482 00:27:37,210 --> 00:27:41,409 S1: called No Such Podcast. Good name. That's a good name. Evidently, 483 00:27:41,410 --> 00:27:44,320 S1: lots of people use the I forgot My password feature 484 00:27:44,320 --> 00:27:48,400 S1: as a de facto login method. Human behavior A Starlink 485 00:27:48,400 --> 00:27:51,460 S1: satellite dish was used on a US Navy ship for 486 00:27:51,460 --> 00:27:55,450 S1: an illicit Wi-Fi network called stinky, which was used for 487 00:27:55,450 --> 00:28:00,010 S1: streaming and civilian communication. The Navy demoted the senior enlisted 488 00:28:00,040 --> 00:28:04,450 S1: leader responsible for being awesome. Uh, so I didn't write that. 489 00:28:04,450 --> 00:28:09,130 S1: I wrote that right. Obviously. Uh, and I got some pushback. Honestly, 490 00:28:09,130 --> 00:28:12,830 S1: I got some pushback from some military people and some 491 00:28:12,830 --> 00:28:14,840 S1: Navy people, and they're like, hey, this is not a 492 00:28:14,840 --> 00:28:17,450 S1: laughing matter. I know it's not a laughing matter. I 493 00:28:17,450 --> 00:28:21,020 S1: know you shouldn't be allowed to set up satellite receivers 494 00:28:21,020 --> 00:28:26,090 S1: on a warship, but I know this. I'm ex-military too. 495 00:28:26,090 --> 00:28:28,040 S1: I just thought it was funny. Like they're trying to 496 00:28:28,040 --> 00:28:31,280 S1: give internet to the people. They're a hero. I know 497 00:28:31,280 --> 00:28:35,750 S1: it's bad. Like, sorry. Not sorry. Um. All right, I 498 00:28:35,750 --> 00:28:40,340 S1: tech Apple released their September updates yesterday and they were decent. 499 00:28:40,490 --> 00:28:44,630 S1: So I'm going to jump ahead. I'm basically getting probably 500 00:28:44,630 --> 00:28:46,310 S1: I don't think I'm going to get the new AirPods. 501 00:28:46,310 --> 00:28:48,920 S1: I saw a thing that basically said that the AirPods 502 00:28:48,920 --> 00:28:51,890 S1: Pro two are still better because they have the silicone 503 00:28:51,920 --> 00:28:55,310 S1: tip part, and I'm going to get the black 16 Pro, 504 00:28:55,310 --> 00:28:58,580 S1: but not the max. I can't remember if there was some. Oh, 505 00:28:58,610 --> 00:29:00,800 S1: I might get the watch as well. I might either 506 00:29:00,830 --> 00:29:05,690 S1: get a black Ultra Series two or I might get 507 00:29:05,690 --> 00:29:09,050 S1: the series ten because it's got a bigger screen then 508 00:29:09,050 --> 00:29:11,390 S1: the ultra. I don't know which one I'm going to get. 509 00:29:11,390 --> 00:29:13,160 S1: I'm going to wait and see it in person, because 510 00:29:13,160 --> 00:29:15,950 S1: I will be camping on the night of the 19th 511 00:29:15,950 --> 00:29:18,710 S1: in Burlingame. So if anyone wants to come hang out 512 00:29:18,710 --> 00:29:21,860 S1: in person, we can have like an impromptu UL meetup 513 00:29:21,860 --> 00:29:28,070 S1: because I've been there for the last let's see since ten, 2010. 514 00:29:28,190 --> 00:29:32,960 S1: Holy crap. I've been camping at that store for 14 515 00:29:32,960 --> 00:29:37,459 S1: years and I camped the three previous times in other places. 516 00:29:37,460 --> 00:29:41,000 S1: So I've been camping for the iPhone since 2007, but 517 00:29:41,000 --> 00:29:44,479 S1: I've been doing it at that store like first in 518 00:29:44,480 --> 00:29:47,270 S1: line for last 14 years, and I will be there 519 00:29:47,270 --> 00:29:50,180 S1: again on the 19th. If anyone wants to do an 520 00:29:50,180 --> 00:29:53,720 S1: impromptu UL meetup, which it won't be an actual UL meetup, 521 00:29:53,720 --> 00:29:56,180 S1: it'll just be us hanging out because there's coffee shops, 522 00:29:56,180 --> 00:29:58,700 S1: there's a Phil's right there so we can hang out. 523 00:29:58,700 --> 00:30:03,920 S1: We could talk. Um, I usually walk a lot through 524 00:30:03,920 --> 00:30:05,750 S1: the course of the night, maybe try to take a 525 00:30:05,750 --> 00:30:09,570 S1: nap or whatever at night. But, uh, anyway, I'll be there. 526 00:30:09,600 --> 00:30:15,240 S1: Nvidia's RTX 40 series GPUs, including the 5080 and 5090, 527 00:30:15,270 --> 00:30:19,500 S1: are getting ready to be finalized and potentially come out 528 00:30:19,500 --> 00:30:22,500 S1: end of the year or beginning of next year. They're 529 00:30:22,500 --> 00:30:25,260 S1: saying these things are going to be massively power hungry. 530 00:30:25,290 --> 00:30:27,720 S1: I've got two 49 seconds in the room next to 531 00:30:27,750 --> 00:30:30,930 S1: me here, and they are beasts. Um, I'm going to 532 00:30:30,930 --> 00:30:34,560 S1: have some FOMO when the 5090 comes out, but I 533 00:30:34,560 --> 00:30:37,740 S1: don't think I'm going to be upgrading my system because, um, 534 00:30:37,740 --> 00:30:40,590 S1: that would be expensive and I'm not sure I need that, 535 00:30:40,590 --> 00:30:44,700 S1: especially when a lot of the AI models are probably 536 00:30:44,730 --> 00:30:48,870 S1: going to run better. I'd rather wait and upgrade my 537 00:30:48,870 --> 00:30:52,380 S1: main desktop on Apple and see what, like the giant 538 00:30:52,410 --> 00:30:57,690 S1: M4 or giant M5 systems have. When I have, let's say, 539 00:30:57,720 --> 00:31:02,130 S1: half a terabyte of system on a chip memory and 540 00:31:02,130 --> 00:31:05,220 S1: all of that can be used for GPU operations. I'm 541 00:31:05,220 --> 00:31:07,930 S1: not sure I'm going to need a dedicated. Well, maybe 542 00:31:07,930 --> 00:31:11,440 S1: I'll need a dedicated one of those boxes to do I, 543 00:31:11,470 --> 00:31:14,230 S1: but I'm not sure I'm going to build another dedicated 544 00:31:14,230 --> 00:31:17,680 S1: PC with Nvidia chips in it, which is weird because 545 00:31:17,680 --> 00:31:20,410 S1: I've got a lot of Nvidia stock. Anyway, that's a 546 00:31:20,410 --> 00:31:24,760 S1: separate talk show. Uh, okay. Trump is launching a crypto project, 547 00:31:24,760 --> 00:31:27,910 S1: but there are concerns because 70% of the tokens are 548 00:31:27,910 --> 00:31:33,040 S1: being allocated to insiders. Uh, Ilya, it has a new startup, SSI, 549 00:31:33,040 --> 00:31:36,280 S1: and they raised $1 billion. So a lot of people 550 00:31:36,280 --> 00:31:41,260 S1: have $1 billion valuations. Ilya did it differently. He has 551 00:31:41,260 --> 00:31:45,250 S1: a $1 billion seed round, a $1 billion seed round. 552 00:31:45,250 --> 00:31:48,280 S1: Insane visa is set to launch a new account to 553 00:31:48,310 --> 00:31:51,459 S1: account a to a payment service in Europe. So you 554 00:31:51,460 --> 00:31:55,570 S1: could make direct bank to bank transfers without using credit cards. 555 00:31:55,600 --> 00:31:58,300 S1: I wonder if that's a competitor to Swift. I imagine 556 00:31:58,300 --> 00:32:00,430 S1: it is. Or maybe it rides on top of Swift, 557 00:32:00,430 --> 00:32:03,969 S1: I don't know. Engineers from Cornell and Florence University have 558 00:32:03,970 --> 00:32:08,890 S1: developed a biohybrid robot that uses electric signals from a 559 00:32:08,890 --> 00:32:13,540 S1: king trumpet mushroom to move and sense its environment. Reminds 560 00:32:13,540 --> 00:32:18,640 S1: me of Neri Oxman. Neri Oxman doing her, uh, networking 561 00:32:18,670 --> 00:32:24,100 S1: of nature type stuff for 2024. Annual Work Trend Index 562 00:32:24,100 --> 00:32:28,450 S1: from Microsoft and LinkedIn reveals a shift in employer preferences 71% 563 00:32:28,450 --> 00:32:32,350 S1: of leaders favoring candidates with AI skill over those with 564 00:32:32,350 --> 00:32:36,790 S1: industry experience. Yeah, I found this one kind of interesting. Um, 565 00:32:37,270 --> 00:32:40,150 S1: Wall Street Journal is highlighting a trend where small startups 566 00:32:40,150 --> 00:32:43,930 S1: are increasingly influencing the US economy. So I've been thinking 567 00:32:43,930 --> 00:32:45,610 S1: about this for a while, is getting ready to write 568 00:32:45,610 --> 00:32:47,260 S1: a piece on it. But I want to I want 569 00:32:47,290 --> 00:32:49,840 S1: to state this more forcefully. I think people are about 570 00:32:49,840 --> 00:32:53,740 S1: to realize that the most medium to large companies, like 571 00:32:53,740 --> 00:32:58,840 S1: the size companies, have become ineffective. They lack vision and focus, 572 00:32:58,840 --> 00:33:02,480 S1: too much bureaucracy. They have giant workforces that are hired 573 00:33:02,480 --> 00:33:05,960 S1: for like a worker bee mentality, rather than to have 574 00:33:05,960 --> 00:33:10,459 S1: them be exceptional or innovative or challenging to the structure. 575 00:33:10,460 --> 00:33:12,680 S1: And I think this is another part of like the 576 00:33:12,680 --> 00:33:15,110 S1: essay that I wrote called The End of Work, where 577 00:33:15,110 --> 00:33:17,360 S1: much of the innovation in the world moves away from 578 00:33:17,360 --> 00:33:22,040 S1: big companies and towards individuals and small, dynamic startups. And 579 00:33:22,040 --> 00:33:24,290 S1: this is also what Mark Andreessen talked about in his 580 00:33:24,290 --> 00:33:27,260 S1: conversation with Huberman, which I thought just came out. But 581 00:33:27,260 --> 00:33:29,330 S1: it turns out that was a year ago. But that 582 00:33:29,330 --> 00:33:34,910 S1: was a fantastic episode. And related to this, Paul Graham's piece, uh, 583 00:33:34,910 --> 00:33:38,690 S1: his latest piece called Founder Mode, looks at how bigger 584 00:33:38,690 --> 00:33:43,220 S1: companies make the mistakes talked about above and how this 585 00:33:43,220 --> 00:33:45,800 S1: founder mode, which he didn't even define it, he just 586 00:33:45,800 --> 00:33:47,630 S1: said it like it needs to be defined because it's 587 00:33:47,630 --> 00:33:51,890 S1: really interesting. But, um, yeah, it basically he says it 588 00:33:51,890 --> 00:33:55,250 S1: allows you to stay in a more innovation focused mindset. Uh, 589 00:33:55,460 --> 00:33:57,770 S1: great read. All this stuff is a great read. I 590 00:33:57,770 --> 00:34:00,810 S1: just wish he, uh I wish he locked it down. 591 00:34:01,530 --> 00:34:04,020 S1: I understand it's not locked down yet, but I thought 592 00:34:04,020 --> 00:34:06,000 S1: he might have tried to do that. Oakland police are 593 00:34:06,000 --> 00:34:10,320 S1: using Tesla's Sentry Mode footage to aid crime investigations. And 594 00:34:10,320 --> 00:34:12,630 S1: if they can't find the owner, they just tow the car. 595 00:34:12,660 --> 00:34:15,060 S1: They just tow the Tesla because it has the evidence. 596 00:34:15,090 --> 00:34:18,719 S1: Waymo is tackling the skepticism around its autonomous vehicles by 597 00:34:18,719 --> 00:34:20,969 S1: launching a new safety hub with a whole bunch of 598 00:34:20,969 --> 00:34:25,410 S1: data and charts showing that they're more safe than human drivers. Oh, 599 00:34:25,440 --> 00:34:27,990 S1: related story that just came out is that most of 600 00:34:27,989 --> 00:34:33,270 S1: the accidents with Waymo are from human drivers hitting the Waymo. 601 00:34:33,300 --> 00:34:37,739 S1: Not surprising to me. Joshua Austin's A manifesto manifesto for 602 00:34:37,739 --> 00:34:41,370 S1: Radical Simplicity argues for a streamlined approach to software delivery, 603 00:34:41,400 --> 00:34:45,630 S1: ditching subjective metrics like story points in favor of focusing 604 00:34:45,630 --> 00:34:50,040 S1: on real dependencies and outcomes. Bluetooth 6.0 is here. Oh, 605 00:34:50,040 --> 00:34:53,190 S1: by the way, the new iPhones have Wi-Fi seven. I 606 00:34:53,219 --> 00:34:56,399 S1: am upgrading my whole house to Wi-Fi seven. They say 607 00:34:56,400 --> 00:35:00,310 S1: it's about five times faster, So if I'm getting like 608 00:35:00,340 --> 00:35:05,110 S1: 800 down right now, eight times 544 gigabits, it would 609 00:35:05,140 --> 00:35:07,029 S1: be nice if I got four gigabits. I bet you 610 00:35:07,030 --> 00:35:10,540 S1: I'm going to get like two gigabits streaming to my 611 00:35:10,540 --> 00:35:14,500 S1: phone and then to my laptop once laptops have Wi-Fi seven, 612 00:35:14,530 --> 00:35:17,230 S1: I'm not sure if my current ones do. A Caritas 613 00:35:17,260 --> 00:35:20,620 S1: phone was snatched in London and despite tracking it with 614 00:35:20,620 --> 00:35:24,760 S1: Find My iPhone, he watched it basically travel around the 615 00:35:24,760 --> 00:35:29,739 S1: city and ended up in Shenzhen. Humans. President XI Jinping 616 00:35:29,770 --> 00:35:33,370 S1: has pledged to create over 1 million jobs in Africa. 617 00:35:33,400 --> 00:35:38,980 S1: I cannot stand seeing Africa become an extension of China. 618 00:35:38,980 --> 00:35:41,800 S1: I do not like it, but it's pretty hard for 619 00:35:41,800 --> 00:35:45,250 S1: the West and anyone in the West to even notice this, 620 00:35:45,250 --> 00:35:48,490 S1: given our history, which is not great. The question is, 621 00:35:48,489 --> 00:35:51,910 S1: how long will we let that guilt be an obstacle 622 00:35:51,940 --> 00:35:56,860 S1: to opposing China wherever they go and basically do colonialism, which, 623 00:35:57,640 --> 00:36:00,340 S1: is bad, but hard to talk about how bad it 624 00:36:00,340 --> 00:36:03,850 S1: is when you already did it and got scolded for it. 625 00:36:03,850 --> 00:36:06,160 S1: And now you feel bad about it, and now you're 626 00:36:06,160 --> 00:36:08,739 S1: pointing the finger at someone else doing it. It feels 627 00:36:08,739 --> 00:36:10,719 S1: kind of weird, but at the same time, it also 628 00:36:10,719 --> 00:36:13,989 S1: feels weird and cowardly to not do anything about it 629 00:36:13,989 --> 00:36:16,960 S1: just because you feel guilty. A whole bunch of right 630 00:36:16,960 --> 00:36:21,160 S1: wing influencers received millions of dollars from Russia in return 631 00:36:21,160 --> 00:36:24,970 S1: for promoting pro-Russian talking points, which is hilarious to me 632 00:36:24,969 --> 00:36:28,299 S1: since their whole narrative is like, I don't believe anything. 633 00:36:28,300 --> 00:36:32,830 S1: I'm skeptical. I'm discerning. Except when it comes to obvious 634 00:36:32,830 --> 00:36:36,160 S1: Russian propaganda. Here's another way to think about it. And 635 00:36:36,160 --> 00:36:38,590 S1: I did some Intel in the Army. I'm not, like 636 00:36:38,590 --> 00:36:42,280 S1: heavily Intel trained, but I did have a decent amount 637 00:36:42,280 --> 00:36:44,440 S1: of exposure to it and worked in like the Intel 638 00:36:44,469 --> 00:36:48,070 S1: Group or whatever. But think about this. Here are probably 639 00:36:48,100 --> 00:36:52,960 S1: two unrelated phenomenon. One. We know for absolute certain that 640 00:36:52,960 --> 00:36:56,270 S1: Russia is trying to use its significant propaganda capabilities to 641 00:36:56,300 --> 00:36:59,359 S1: influence the right wing in the United States to be 642 00:36:59,360 --> 00:37:03,170 S1: pro-Russian into Ukraine that we know for absolute certain. Right. 643 00:37:03,200 --> 00:37:07,279 S1: That's not even debatable. They are really good at this, 644 00:37:07,280 --> 00:37:11,030 S1: and they are spending at least tens of millions of dollars, 645 00:37:11,030 --> 00:37:14,930 S1: probably hundreds of millions of dollars, probably not billions. So 646 00:37:14,930 --> 00:37:17,600 S1: I'm guessing tens of millions to hundreds of millions of 647 00:37:17,600 --> 00:37:20,630 S1: dollars they are spending on doing this. And they're really 648 00:37:20,630 --> 00:37:23,390 S1: good at it because it's essentially like the evolution of 649 00:37:23,390 --> 00:37:27,170 S1: the KGB. Second point, which is probably unrelated. The right 650 00:37:27,170 --> 00:37:31,460 S1: wing in the United States is now completely, almost completely 651 00:37:31,460 --> 00:37:34,640 S1: pro-Russian Anti-ukraine. And some people reached out to me and 652 00:37:34,640 --> 00:37:37,160 S1: they're like, that's not true. That's not true. I should 653 00:37:37,190 --> 00:37:39,319 S1: have went and found some data and linked to it. 654 00:37:39,320 --> 00:37:41,840 S1: I read the data all the time. I read the 655 00:37:41,840 --> 00:37:44,930 S1: data all the time in polling, where it's like, how 656 00:37:44,930 --> 00:37:50,690 S1: much of the Republican Party, when polled, are actually pro-Putin? 657 00:37:50,719 --> 00:37:53,390 S1: The numbers are ridiculous. I'm sorry, I don't have the 658 00:37:53,400 --> 00:37:55,680 S1: link here. I should have put it in there, but 659 00:37:55,680 --> 00:37:59,370 S1: it's it's pretty easy to Google and you'll find it yourself. 660 00:37:59,370 --> 00:38:01,710 S1: It's actually better if you find it yourself than if 661 00:38:01,710 --> 00:38:04,979 S1: I feed you the link. Just Google that. Look for 662 00:38:04,980 --> 00:38:09,570 S1: your look for the stats, the polling. It's ridiculous. And 663 00:38:09,570 --> 00:38:12,120 S1: a brief political aside. So this is an example of 664 00:38:12,150 --> 00:38:15,150 S1: a little breakout piece. So I already know I'm going 665 00:38:15,150 --> 00:38:17,609 S1: to get hate mail about the point above, because I'm 666 00:38:17,610 --> 00:38:20,310 S1: a crazy liberal and I post a lot of other 667 00:38:20,310 --> 00:38:25,680 S1: stuff that's like anti-left far left and their idiocracy or idiocy. 668 00:38:25,680 --> 00:38:28,890 S1: And I get tons of comments about being way too 669 00:38:28,920 --> 00:38:33,030 S1: right from the left because I attack the left and whatever, 670 00:38:33,030 --> 00:38:34,890 S1: and I just now attack the right and they're going 671 00:38:34,920 --> 00:38:37,560 S1: to be like, oh, you're too left wing. So I 672 00:38:37,560 --> 00:38:40,410 S1: ask you to consider another possibility, which is I'm actively 673 00:38:40,410 --> 00:38:44,790 S1: considering each position from first principles, not perfect. I could 674 00:38:44,790 --> 00:38:47,310 S1: be wrong, but I put a lot of effort into 675 00:38:47,340 --> 00:38:49,620 S1: having my own opinions that are not part of a 676 00:38:49,620 --> 00:38:53,190 S1: tribe of pre-approved options. So perhaps the best way to 677 00:38:53,219 --> 00:38:56,489 S1: sum me up right now is that I'm liberal in 678 00:38:56,489 --> 00:39:00,690 S1: my goals and somewhat conservative in my approach for how 679 00:39:00,690 --> 00:39:03,540 S1: to achieve them. So meaning, I want a planet full 680 00:39:03,540 --> 00:39:08,040 S1: of lots of different colors and ethnicities of people all 681 00:39:08,040 --> 00:39:12,899 S1: thriving together. A secular society that encourages any religion but 682 00:39:12,900 --> 00:39:16,530 S1: doesn't allow any of them to infringe on government or 683 00:39:16,530 --> 00:39:21,000 S1: the ideals listed here. Gender identity, private sexual behavior between 684 00:39:21,000 --> 00:39:26,969 S1: consenting adults, all personal choices and nobody's business. Basically, the 685 00:39:26,969 --> 00:39:29,640 S1: freedom for everyone to strive to be the best versions 686 00:39:29,640 --> 00:39:33,120 S1: of themselves that they can be, and a society that 687 00:39:33,120 --> 00:39:37,230 S1: sees that as simultaneously a matter of personal responsibility, but 688 00:39:37,230 --> 00:39:41,850 S1: also helps people along that path. So free speech, the 689 00:39:41,850 --> 00:39:46,320 S1: ability to offend people with difficult ideas and have those 690 00:39:46,320 --> 00:39:50,590 S1: debates and be controversial and be offensive, right. The concept 691 00:39:50,590 --> 00:39:55,629 S1: of meritocracy. The emphasis on personal responsibility. Okay. And those 692 00:39:55,630 --> 00:39:59,110 S1: are kind of like currently those are considered like right wing. Right. 693 00:39:59,110 --> 00:40:03,640 S1: But also the acknowledgement that some people in groups need 694 00:40:03,670 --> 00:40:07,540 S1: help getting to this point where their personal responsibility can 695 00:40:07,540 --> 00:40:10,540 S1: actually take root and help them thrive, and that it 696 00:40:10,540 --> 00:40:15,520 S1: is society's responsibility to give that to them. And that's 697 00:40:15,520 --> 00:40:18,040 S1: where it's liberal. So in other words, if everyone had 698 00:40:18,040 --> 00:40:22,060 S1: the same opportunity, I'd be fiercely all about meritocracy. But 699 00:40:22,060 --> 00:40:25,690 S1: not everyone has the same opportunity. And that's the role 700 00:40:25,690 --> 00:40:30,250 S1: of society and charity and kindness to help them get 701 00:40:30,250 --> 00:40:33,790 S1: to the place where their hard work can actually benefit them. 702 00:40:33,820 --> 00:40:36,430 S1: So the problem right now, in my mind, is that 703 00:40:36,430 --> 00:40:41,170 S1: the far right and the far left are in opposition 704 00:40:41,170 --> 00:40:43,630 S1: to that model. I just laid out the far right 705 00:40:43,630 --> 00:40:46,360 S1: because they actually want the wrong things, and the far 706 00:40:46,360 --> 00:40:48,680 S1: left because they are so confused about how the world 707 00:40:48,680 --> 00:40:51,650 S1: works that they're causing more harm than good. So anyway, 708 00:40:51,650 --> 00:40:54,380 S1: that's a short version of where I currently stand. And 709 00:40:54,380 --> 00:40:56,930 S1: if you ever get confused about like if I'm crazy 710 00:40:56,930 --> 00:41:00,650 S1: right or crazy left, just go read that. And here's 711 00:41:00,650 --> 00:41:04,009 S1: a really important point. I also consider you to make 712 00:41:04,010 --> 00:41:08,090 S1: your own North Star paragraph or set of bullet points 713 00:41:08,090 --> 00:41:10,700 S1: like I just did, and have that be your thing 714 00:41:10,700 --> 00:41:14,420 S1: and then realize that this this, the left narrative or 715 00:41:14,420 --> 00:41:18,500 S1: the right narrative or centrist narrative with all these different narratives, 716 00:41:18,500 --> 00:41:20,600 S1: they don't really matter. What matters is what is your 717 00:41:20,630 --> 00:41:23,239 S1: North Star look like? Like I just laid out. And 718 00:41:23,239 --> 00:41:25,910 S1: then what you'll find is, like all these different narratives 719 00:41:25,910 --> 00:41:29,629 S1: and tribes and everything they might have. Venn diagram overlaps 720 00:41:29,630 --> 00:41:32,210 S1: with some part of your thing, but they shouldn't be 721 00:41:32,210 --> 00:41:35,720 S1: your thing. If you have a perfect Venn diagram, overlap 722 00:41:35,719 --> 00:41:40,070 S1: with the extreme right or the extreme left, you don't 723 00:41:40,070 --> 00:41:43,610 S1: have your own opinions. You are using someone else's opinions 724 00:41:43,610 --> 00:41:47,089 S1: and that's a problem. So North Star plus first principles 725 00:41:47,090 --> 00:41:50,180 S1: is far better than picking a tribe and endorsing everything 726 00:41:50,180 --> 00:41:53,390 S1: they say. All right. Sweden's health authority has issued new 727 00:41:53,390 --> 00:41:58,310 S1: guidelines advising the children under two should have no screen time, 728 00:41:58,310 --> 00:42:01,520 S1: while teenagers should be limited to three hours a day. 729 00:42:01,550 --> 00:42:03,740 S1: A lot of people are starting to say, and this 730 00:42:03,739 --> 00:42:06,320 S1: is supported by a bunch of different studies, that exercise 731 00:42:06,320 --> 00:42:09,350 S1: could be the most potent medical intervention that we know of. 732 00:42:09,350 --> 00:42:12,200 S1: And I got a buddy who's a cardiologist named Jonathan, 733 00:42:12,200 --> 00:42:16,730 S1: and he agrees with this. David Brooks discusses TEDx Goias 734 00:42:16,760 --> 00:42:21,200 S1: Gioia's essay on the decline of American culture, where art 735 00:42:21,200 --> 00:42:24,800 S1: is overshadowed by entertainment, and now even entertainment is being 736 00:42:24,800 --> 00:42:28,879 S1: consumed by distraction from platforms like TikTok and Instagram. Got 737 00:42:28,880 --> 00:42:32,959 S1: a photographer documenting life and beauty of America's last old 738 00:42:32,960 --> 00:42:36,290 S1: growth forests? Got an article here that explores the belief 739 00:42:36,290 --> 00:42:38,720 S1: that there's a place for everyone, suggesting that every person 740 00:42:38,719 --> 00:42:42,980 S1: has a unique purpose and value. Marco Giancotti argues that 741 00:42:42,980 --> 00:42:46,830 S1: with millions of books available, only select a few what 742 00:42:46,830 --> 00:42:50,489 S1: he calls damn good books, which are truly life changing. 743 00:42:50,489 --> 00:42:52,860 S1: And these are the books that transform you and not 744 00:42:52,860 --> 00:42:56,250 S1: to get bogged down with, like, every book out there. 745 00:42:56,370 --> 00:42:58,470 S1: And this goes along with the idea of like, you 746 00:42:58,469 --> 00:43:01,109 S1: should be willing to put down bad books, just put 747 00:43:01,110 --> 00:43:03,270 S1: it down. Go on to a better one. Phoenix just 748 00:43:03,270 --> 00:43:07,410 S1: hit 100 consecutive days of 100 degree heat, which beats 749 00:43:07,410 --> 00:43:11,490 S1: the record that was set in 1993. And discovery lmao 750 00:43:11,489 --> 00:43:16,650 S1: swsh a bash wrapper around Python's ML whisper. It's basically 751 00:43:16,680 --> 00:43:20,520 S1: a really fast whisper for Mac. NN term lets you 752 00:43:20,520 --> 00:43:24,120 S1: browse Hacker News from your terminal and dungeon dash, a 753 00:43:24,120 --> 00:43:28,710 S1: remote know a command line RPG where you dive into dungeons, 754 00:43:28,710 --> 00:43:32,400 S1: battle enemies, and collect loot to up level and become 755 00:43:32,400 --> 00:43:35,160 S1: the ultimate hero. I want to play this a whole 756 00:43:35,160 --> 00:43:37,410 S1: bunch more, and I actually want to make some of these. 757 00:43:37,410 --> 00:43:40,440 S1: These these just seem super fun and I love the terminal. 758 00:43:40,440 --> 00:43:44,110 S1: If you didn't already know that. NSA's national Cryptographic school 759 00:43:44,140 --> 00:43:49,750 S1: television catalog from 1991 has surfaced, listing around 600 training 760 00:43:49,750 --> 00:43:53,950 S1: videos on CommSec and Sigint. I love these treasure troves 761 00:43:53,950 --> 00:43:57,400 S1: of old things. When I was in the Army, I 762 00:43:57,400 --> 00:44:00,040 S1: got to work in a library for a certain amount 763 00:44:00,040 --> 00:44:02,980 S1: of time, and I got all the Special Forces manuals 764 00:44:02,980 --> 00:44:06,070 S1: and all the CommSec and Sigint and stuff like that, 765 00:44:06,070 --> 00:44:09,879 S1: and I would just sit and read these things, like voraciously. Okay. 766 00:44:09,910 --> 00:44:13,240 S1: List of ideas. Extraordinarily simple template you could use to 767 00:44:13,270 --> 00:44:16,270 S1: orient your life. Okay. I love this thing. I believe 768 00:44:16,270 --> 00:44:18,790 S1: one of the biggest issues in the world is this, 769 00:44:18,820 --> 00:44:21,939 S1: which is why I'm looking to solve it using this, 770 00:44:21,940 --> 00:44:25,419 S1: which is why I'm doing these projects and I'm measuring 771 00:44:25,420 --> 00:44:29,770 S1: my success using these metrics. This is how to introduce 772 00:44:29,770 --> 00:44:33,010 S1: yourself at a party. This is potentially how to introduce 773 00:44:33,010 --> 00:44:35,680 S1: yourself to a mate. This is how to explain yourself 774 00:44:35,680 --> 00:44:39,430 S1: in an interview. It's your LinkedIn profile. Look at this. 775 00:44:39,430 --> 00:44:42,350 S1: I believe this is the biggest problem which I think 776 00:44:42,350 --> 00:44:44,630 S1: we should solve by doing this, which is why I'm 777 00:44:44,630 --> 00:44:49,490 S1: doing this work which I am measuring using these metrics. 778 00:44:49,489 --> 00:44:53,390 S1: That is purpose, that is direction, and it explains it 779 00:44:53,390 --> 00:44:56,540 S1: to others. And it's done using a simple sentence like, 780 00:44:56,570 --> 00:44:59,359 S1: you could turn this into one complex sentence. And if 781 00:44:59,360 --> 00:45:02,480 S1: you can answer this, I would argue you're better off 782 00:45:02,480 --> 00:45:04,940 S1: than most people in the workforce and most people on 783 00:45:04,940 --> 00:45:06,620 S1: the planet. The more I think about it, the more 784 00:45:06,620 --> 00:45:10,580 S1: I think major career for creators going forward will be 785 00:45:10,580 --> 00:45:14,600 S1: building entire realities for people to live inside of. So 786 00:45:14,630 --> 00:45:19,130 S1: think post, AGI or ASI and post UBI and where 787 00:45:19,160 --> 00:45:22,160 S1: games are extraordinarily immersive. So I think there will be 788 00:45:22,160 --> 00:45:25,910 S1: a huge market for creative people, creative people building the 789 00:45:25,910 --> 00:45:29,690 S1: storylines and stat systems that look and feel like entire 790 00:45:29,690 --> 00:45:32,540 S1: worlds that people will live inside of. So imagine how 791 00:45:32,540 --> 00:45:36,049 S1: everyone loved Game of Thrones for all those years, except 792 00:45:36,050 --> 00:45:38,990 S1: like you're actually one of the characters. In fact, you're 793 00:45:38,989 --> 00:45:42,540 S1: one of the heroes in your navigating through the the difficulty, right? 794 00:45:42,570 --> 00:45:44,850 S1: Or it could be like a superhero themed thing or 795 00:45:44,850 --> 00:45:48,630 S1: like a romance themed thing, or like fantasy or whatever 796 00:45:48,630 --> 00:45:51,959 S1: the structure is. So imagine basically your favorite book or 797 00:45:51,960 --> 00:45:55,710 S1: TV show or series or movie or whatever. We basically 798 00:45:55,710 --> 00:46:00,719 S1: build fully immersive realities based on those for people to 799 00:46:00,719 --> 00:46:04,830 S1: spend their time inside of. And the problem is, it 800 00:46:04,830 --> 00:46:07,650 S1: can't all be entertainment because people need a purpose. They 801 00:46:07,650 --> 00:46:09,870 S1: need some sort of drive. I think we're going to 802 00:46:09,870 --> 00:46:13,830 S1: find ways to create that purpose inside of these realities. 803 00:46:13,830 --> 00:46:16,230 S1: I have a piece from I need to go dig 804 00:46:16,230 --> 00:46:19,950 S1: this thing up. It's from like 2002, and I basically 805 00:46:19,950 --> 00:46:23,250 S1: said that the games of the future will be people 806 00:46:23,280 --> 00:46:27,360 S1: going in and doing the work and the jobs inside there. 807 00:46:27,390 --> 00:46:29,700 S1: So you'll go in there and you'll be a cop, okay? 808 00:46:29,730 --> 00:46:31,830 S1: And you'll see someone coming down the thing and the 809 00:46:31,830 --> 00:46:34,500 S1: music is too loud, or they're beating somebody up, and 810 00:46:34,500 --> 00:46:37,410 S1: you'll go and actually engage in a fight with this 811 00:46:37,440 --> 00:46:41,100 S1: person to try to restrain them or whatever. Or maybe 812 00:46:41,100 --> 00:46:43,140 S1: you're the bad guy and you're just going to go 813 00:46:43,170 --> 00:46:45,960 S1: beat somebody up, and it's someone else's job to be 814 00:46:45,960 --> 00:46:48,810 S1: the good guy, and they're actually paid or rewarded in 815 00:46:48,810 --> 00:46:51,660 S1: some sort of way. So other people inside the game 816 00:46:51,660 --> 00:46:55,890 S1: can give them something, or the system can reward them 817 00:46:55,890 --> 00:46:58,620 S1: in a way that's good on the outside as well. Now, 818 00:46:58,620 --> 00:47:03,390 S1: in most scenarios, your house and your game subscription and 819 00:47:03,390 --> 00:47:06,330 S1: your game gear and everything is being paid by UBI 820 00:47:06,330 --> 00:47:08,730 S1: in this world that I'm thinking of, most of the 821 00:47:08,730 --> 00:47:11,520 S1: jobs are gone and people are being paid essentially to 822 00:47:11,550 --> 00:47:16,290 S1: live inside these realities, which is massively dystopian. I want 823 00:47:16,290 --> 00:47:19,500 S1: to be very clear about that, if that is the 824 00:47:19,500 --> 00:47:22,590 S1: only option. Okay. What I'm saying is that when the 825 00:47:22,590 --> 00:47:24,780 S1: jobs go away, there will be a lot of people 826 00:47:24,780 --> 00:47:28,200 S1: who want to live these lives because we are creative. 827 00:47:28,230 --> 00:47:31,319 S1: We read these books, we see these movies we watch. 828 00:47:31,469 --> 00:47:34,950 S1: We know that that's on offer. That's the thing we 829 00:47:34,950 --> 00:47:37,750 S1: could potentially have. So if games make it where we 830 00:47:37,750 --> 00:47:40,149 S1: can actually have it, we can actually be that hero. 831 00:47:40,180 --> 00:47:42,669 S1: We will want to do it, but we won't want 832 00:47:42,670 --> 00:47:45,580 S1: to do it. It will be empty and weak and 833 00:47:45,580 --> 00:47:50,770 S1: shallow and lame and horribly destructive. If it's nothing but 834 00:47:50,770 --> 00:47:54,730 S1: just like violence and porn, which there definitely will be 835 00:47:54,730 --> 00:47:57,040 S1: that right? There will be games that are just nothing 836 00:47:57,040 --> 00:48:00,520 S1: but that. But it's all entertainment. It's empty, it's hollow, 837 00:48:00,520 --> 00:48:03,970 S1: it's nothing. The better version of this, which I hope 838 00:48:03,969 --> 00:48:08,290 S1: happens and definitely could happen with, like if we're wielding 839 00:48:08,320 --> 00:48:11,350 S1: AC or if AC takes over and it's doing this 840 00:48:11,380 --> 00:48:15,219 S1: in a benign way, it would build meaning into the system, 841 00:48:15,219 --> 00:48:18,850 S1: but it wouldn't require that you do it. You could 842 00:48:18,850 --> 00:48:23,860 S1: also take off the gear, put it aside, go outside, 843 00:48:23,890 --> 00:48:28,660 S1: start a community, start a farm. Like, uh, till the 844 00:48:28,660 --> 00:48:32,739 S1: land toe the land. What's the name? What's the verb? Um, 845 00:48:32,739 --> 00:48:36,070 S1: you know, till I think it's till. Anyway, you tend 846 00:48:36,070 --> 00:48:39,790 S1: to the land, you grow crops, you make food, people 847 00:48:39,790 --> 00:48:42,610 S1: eat the food. You have families, you have kids like. 848 00:48:42,640 --> 00:48:46,300 S1: So imagine this. There are societies that's kind of like Amish. 849 00:48:46,300 --> 00:48:49,030 S1: They don't play the games, so they're kind of Amish. 850 00:48:49,150 --> 00:48:52,900 S1: And maybe there's like more advanced technological versions of Amish 851 00:48:52,900 --> 00:48:56,259 S1: where it's like, I don't know, the 1950s or something, right? 852 00:48:56,290 --> 00:48:59,200 S1: In this ideal kind of world, except for it's everywhere 853 00:48:59,200 --> 00:49:03,400 S1: and everyone can participate. I would argue that the actual 854 00:49:03,400 --> 00:49:05,200 S1: reality is going to look like there's going to be 855 00:49:05,200 --> 00:49:08,830 S1: some communities like that, some communities that are super Amish, 856 00:49:08,830 --> 00:49:11,920 S1: they're like Ted Kaczynski, like tech is bad and they 857 00:49:11,920 --> 00:49:15,310 S1: just do their own thing. Government should leave them alone. 858 00:49:15,310 --> 00:49:18,279 S1: They are productive members of society. They have their own schools. 859 00:49:18,280 --> 00:49:20,440 S1: They have their own religion. They have their own stuff. 860 00:49:20,469 --> 00:49:23,200 S1: Like they're making their own food. Leave them alone. Everything's fine. 861 00:49:23,200 --> 00:49:25,000 S1: They don't have to play the games. And a bunch 862 00:49:25,000 --> 00:49:27,520 S1: of people like maybe some of their kids, and that'll 863 00:49:27,550 --> 00:49:29,230 S1: be bad, but a bunch of people are going to 864 00:49:29,230 --> 00:49:30,850 S1: be like, are you kidding me? I want to be 865 00:49:30,850 --> 00:49:33,740 S1: a superhero. I want to be a princess. I want 866 00:49:33,770 --> 00:49:35,570 S1: to be, you know, I want to take over the world. 867 00:49:35,570 --> 00:49:38,149 S1: I want to do all these things. Fine. Go find 868 00:49:38,150 --> 00:49:41,690 S1: your meaning inside the game system. Importantly here, there is 869 00:49:41,690 --> 00:49:44,060 S1: going to be a period of time in which we're 870 00:49:44,060 --> 00:49:47,270 S1: going to need something like this as a transition period, 871 00:49:47,420 --> 00:49:50,240 S1: because so many people are going to lose their meaning, 872 00:49:50,239 --> 00:49:51,770 S1: they're going to have to go into this thing. I 873 00:49:51,770 --> 00:49:54,110 S1: think there's going to be a requirement for an ASI, 874 00:49:54,140 --> 00:49:56,870 S1: for a government, for something to be like, look, we 875 00:49:56,870 --> 00:49:59,480 S1: need to give out money and we need to give 876 00:49:59,480 --> 00:50:08,150 S1: them something awesome. I think game companies, whoever it is, Ubisoft, Microsoft, Sony, 877 00:50:08,180 --> 00:50:10,490 S1: all these people, they're actually going to get hired by 878 00:50:10,489 --> 00:50:14,420 S1: the government to create these giant scale games that people 879 00:50:14,420 --> 00:50:17,210 S1: can live inside of. There's going to be haptic gear 880 00:50:17,239 --> 00:50:20,210 S1: they think of like a military push to invent the 881 00:50:20,210 --> 00:50:23,690 S1: bomb or whatever, but the military push is going to 882 00:50:23,690 --> 00:50:27,620 S1: be to invent really immersive technology because we need it, 883 00:50:27,620 --> 00:50:30,800 S1: because guess what? It is actually a military concern. It 884 00:50:30,810 --> 00:50:35,340 S1: is actually a military level requirement that we do not 885 00:50:35,340 --> 00:50:39,719 S1: have hundreds of millions of people, also known as billions 886 00:50:39,719 --> 00:50:43,770 S1: of people with no meaning and no money, because they 887 00:50:43,770 --> 00:50:46,230 S1: will in fact light everything on fire and that will 888 00:50:46,260 --> 00:50:48,660 S1: be bad. So there will be a period where we 889 00:50:48,660 --> 00:50:52,290 S1: actually need to invent something like this to get us 890 00:50:52,290 --> 00:50:56,760 S1: through this period. That kind of calms everyone down. But 891 00:50:56,760 --> 00:51:00,270 S1: if you don't do it well, they will be satiated 892 00:51:00,270 --> 00:51:03,509 S1: and they will be not violent, but it will start 893 00:51:03,510 --> 00:51:06,780 S1: to rot. It will rot the human soul. Because what 894 00:51:06,780 --> 00:51:09,120 S1: are they doing for creativity? What are they making? What 895 00:51:09,120 --> 00:51:11,730 S1: are they building? Are they building? Are they making families? 896 00:51:11,730 --> 00:51:15,600 S1: Who's making the kids? Right? I mean, the games didn't 897 00:51:15,600 --> 00:51:18,660 S1: solve aging. So a bunch of people hooked into the games, 898 00:51:18,660 --> 00:51:22,290 S1: they don't meet anyone. And two generations later, we don't 899 00:51:22,290 --> 00:51:24,989 S1: have any people. Or actually, technically, I think everyone dies 900 00:51:24,989 --> 00:51:28,500 S1: in one generation, right? If you're not reproducing. So like 901 00:51:28,530 --> 00:51:32,469 S1: that is only a temporary solution to not have the 902 00:51:32,469 --> 00:51:34,719 S1: world blow up. But ultimately you have to solve the 903 00:51:34,719 --> 00:51:38,590 S1: meaning problem, which either happens outside the game, which I 904 00:51:38,590 --> 00:51:41,680 S1: think that is going to happen and should happen, or 905 00:51:41,680 --> 00:51:45,070 S1: inside the game. And I think what's most likely in 906 00:51:45,070 --> 00:51:49,029 S1: a benign situation is that you'll have both recommendation of 907 00:51:49,030 --> 00:51:52,690 S1: the week. I've been a bit obsessed with the problem 908 00:51:52,690 --> 00:51:56,230 S1: definition lately, like I talked about with the fabric pattern. 909 00:51:56,230 --> 00:51:59,770 S1: And here's my recommendation get really good at articulating and 910 00:51:59,770 --> 00:52:04,300 S1: prioritizing your problems, like write them out in vast detail. 911 00:52:04,330 --> 00:52:08,290 S1: Make yourself an expert in your problems. It takes away 912 00:52:08,290 --> 00:52:11,140 S1: their power. Kind of like staring directly at anger when 913 00:52:11,140 --> 00:52:13,930 S1: you're meditating. Like it takes the spin off the anger. 914 00:52:13,930 --> 00:52:15,940 S1: If you look away, the anger hurts you again. You 915 00:52:15,940 --> 00:52:18,070 S1: look directly at the anger and you're like, mm, that's 916 00:52:18,070 --> 00:52:20,440 S1: not so bad. Happens the same with your problems. They 917 00:52:20,440 --> 00:52:24,459 S1: become approachable. And this also happens to be the key 918 00:52:24,460 --> 00:52:28,330 S1: to brilliant I prompting because it's an extension of like 919 00:52:28,330 --> 00:52:31,750 S1: know yourself and then you can articulate yourself and the 920 00:52:31,750 --> 00:52:34,359 S1: aphorism of the week when I have one week to 921 00:52:34,390 --> 00:52:38,590 S1: solve a seemingly impossible problem, I spend six days defining it, 922 00:52:38,590 --> 00:52:42,040 S1: and then the solution becomes obvious. When I have one 923 00:52:42,040 --> 00:52:45,940 S1: week to solve a seemingly impossible problem, I spend six 924 00:52:45,940 --> 00:52:51,790 S1: days defining it, and then the solution becomes obvious. Albert Einstein. 925 00:52:53,110 --> 00:52:56,350 S1: Unsupervised learning is produced and edited by Daniel Miessler on 926 00:52:56,350 --> 00:53:00,939 S1: a Neumann U87 AI microphone using Hindenburg. Intro and outro 927 00:53:00,969 --> 00:53:04,270 S1: music is by zombie with a Y. And to get 928 00:53:04,270 --> 00:53:06,370 S1: the text and links from this episode, sign up for 929 00:53:06,370 --> 00:53:10,839 S1: the newsletter version of the show at Daniel miessler.com/newsletter. 930 00:53:11,830 --> 00:53:12,880 UU: We'll see you next time.