WEBVTT - UL NO. 452: The New Hotness: NotebookLM 0:00:01.129 --> 0:00:04.460 Welcome to Unsupervised Learning, a security, AI and meaning focused 0:00:04.460 --> 0:00:07.310 podcast that looks at how best to thrive as humans 0:00:07.310 --> 0:00:11.540 in a post AI world. It combines original ideas, analysis, 0:00:11.570 --> 0:00:14.780 and mental models to bring not just the news, but 0:00:14.780 --> 0:00:22.489 why it matters and how to respond. All right, welcome 0:00:22.489 --> 0:00:25.310 to unsupervised learning. This is Daniel Miessler. This one is 0:00:25.310 --> 0:00:27.620 going to be crazy. I'm going to jump around a lot. 0:00:27.650 --> 0:00:31.610 Lots of stuff to cover. Okay. Let's jump in. What 0:00:31.610 --> 0:00:34.070 do we got? All right. Brand new piece on five 0:00:34.100 --> 0:00:38.180 new ways. I recommend thinking about current and future. I 0:00:38.240 --> 0:00:40.909 guess we'll click on this one and jump in. Okay, 0:00:40.909 --> 0:00:43.790 so this all started from a conversation I had with 0:00:43.790 --> 0:00:48.530 a buddy who revealed himself later on, like the next day, 0:00:48.530 --> 0:00:52.400 but basically he was saying, I don't believe that AI 0:00:52.430 --> 0:00:56.540 is understand. And this totally triggered me. As you might expect, 0:00:56.690 --> 0:01:01.350 given my views on AI and understanding. So we started 0:01:01.350 --> 0:01:04.350 having this conversation, and the way I was talking to 0:01:04.350 --> 0:01:08.220 him about it was I took him through some scenarios, right? 0:01:08.220 --> 0:01:11.520 I basically took him through like, let's see here, let's 0:01:11.520 --> 0:01:14.100 go down to where it was, okay. Yeah, this is 0:01:14.100 --> 0:01:17.580 what it was. So I basically told him, give me 0:01:17.580 --> 0:01:20.700 a list of your top ten favorite restaurants. And he's 0:01:20.700 --> 0:01:23.250 like trying to think of his top ten favorite restaurants. 0:01:23.250 --> 0:01:26.400 And I'm like, he starts rattling off some, right? He 0:01:26.400 --> 0:01:28.770 gives me like first one, second one, third one. He's 0:01:28.770 --> 0:01:31.530 thinking about the fourth one. And I'm like, okay, forget it, 0:01:31.560 --> 0:01:34.860 forget the rest. Where did that list come from? Like 0:01:34.860 --> 0:01:37.500 what just happened? And he starts smiling and he's like, 0:01:37.500 --> 0:01:40.080 I hate you. I'm going to cuss in this one. 0:01:40.080 --> 0:01:42.660 By the way, I don't know why I'm asking for 0:01:42.660 --> 0:01:46.590 permission to cuss. Um, it is how I generally speak, 0:01:46.590 --> 0:01:50.250 especially when I get excited, but I just I want 0:01:50.250 --> 0:01:52.020 there to be a warning just in case there's, like, 0:01:52.050 --> 0:01:55.710 kids around. I don't know, I'm usually not cussing. But anyway, 0:01:55.710 --> 0:01:57.900 bottom line, if you have kids around and you don't 0:01:57.900 --> 0:02:01.130 like them hearing bad words. I tend to cuss a 0:02:01.130 --> 0:02:04.550 little bit. That's why it's marked as explicit in the podcast. 0:02:04.550 --> 0:02:08.510 But anyway, it's never like explicit in the bad way, 0:02:08.510 --> 0:02:11.600 in my opinion anyway. All right. So he starts smiling 0:02:11.600 --> 0:02:14.720 and he's like, Holy crap, I hate you. I ask 0:02:14.720 --> 0:02:17.150 him to basically come up with like, where do you 0:02:17.150 --> 0:02:20.180 think that list came from? You were thinking about your 0:02:20.180 --> 0:02:24.619 favorite restaurants, and you didn't get immediately a list that 0:02:24.620 --> 0:02:29.090 you put together yourself. What you got was a streaming 0:02:29.090 --> 0:02:33.380 answer of like, things coming from somewhere else, right? So 0:02:33.380 --> 0:02:35.870 he gives me this answer and I'm like, how did 0:02:35.870 --> 0:02:38.299 you make that sentence? So he gives me the answer 0:02:38.300 --> 0:02:40.280 for how he came up with the list. And I'm like, 0:02:40.280 --> 0:02:42.980 how did you make the sentence that you just made? 0:02:42.980 --> 0:02:45.560 Where does that sentence come from? And he's like, I 0:02:45.590 --> 0:02:49.460 fucking hate you. I have no idea. Holy crap. And 0:02:49.460 --> 0:02:52.400 I'm like, would you say that that sentence streamed out 0:02:52.400 --> 0:02:56.690 of you? And he's like, Holy crap. It's just exactly 0:02:56.690 --> 0:02:59.929 like an LLM. So we ended up getting interrupted. But 0:02:59.930 --> 0:03:02.989 my point is, I was trying to show multiple examples 0:03:02.990 --> 0:03:08.060 of like how humans are behaving essentially exactly like Llms. 0:03:08.060 --> 0:03:10.550 And there are some differences that I could also list, 0:03:10.550 --> 0:03:14.780 but largely I think both Llms and humans are black boxes. 0:03:14.780 --> 0:03:18.350 And if you pay attention to your own thoughts and 0:03:18.350 --> 0:03:20.960 your sentences as you're building them, like that sentence I'm 0:03:20.960 --> 0:03:23.210 currently in the middle of right now, I have no 0:03:23.210 --> 0:03:26.480 idea what the rest of the sentence is. It's happening 0:03:26.480 --> 0:03:30.140 in real time and it's non-deterministic, right? Because if I 0:03:30.169 --> 0:03:33.890 went to this exact same podcast press record, did this 0:03:33.889 --> 0:03:36.560 exact same thing, this sentence would not be the same. 0:03:36.560 --> 0:03:40.760 And people complain about Llms being non-deterministic, while human creativity 0:03:40.760 --> 0:03:43.760 is also non-deterministic. So I was giving him all these 0:03:43.790 --> 0:03:46.160 examples and he's like, yeah, I'm definitely going to be 0:03:46.160 --> 0:03:48.050 thinking about this. So it gave him a lot to 0:03:48.080 --> 0:03:51.140 think about. And I totally respect him for like, you know, 0:03:51.170 --> 0:03:53.300 being willing to consider it. And he even said that 0:03:53.300 --> 0:03:56.270 in the very beginning. He's like, listen, I disagree with you. 0:03:56.270 --> 0:04:00.230 I do not think that llms understand, based on all 0:04:00.230 --> 0:04:03.860 the research that I've done and watching multiple videos and everything, 0:04:03.860 --> 0:04:06.830 it's like I currently don't believe that they do understand, 0:04:06.830 --> 0:04:10.250 but I'm happy to listen and be educated, if you 0:04:10.280 --> 0:04:13.340 are correct. So I thought that was a massively respectable. 0:04:13.370 --> 0:04:16.460 But what it got me doing is thinking a whole 0:04:16.460 --> 0:04:19.789 bunch about this. Like, why isn't my argument good enough 0:04:19.790 --> 0:04:22.850 to convince him in one perfect sentence? Why isn't my 0:04:22.850 --> 0:04:24.770 argument good enough where I could just show him one 0:04:24.770 --> 0:04:28.610 web page or one video of 30s long and he goes, oh, 0:04:28.640 --> 0:04:31.909 now I believe you. So to me, it's like a 0:04:31.910 --> 0:04:37.670 finding sort of Feynman. Feynman, like a Richard Feynman type thing, 0:04:37.670 --> 0:04:40.070 where it's like, it means I'm not explaining it well enough, 0:04:40.070 --> 0:04:42.050 and if I'm not explaining it well enough, that means 0:04:42.050 --> 0:04:44.870 I don't understand it well enough. So I started going 0:04:44.870 --> 0:04:48.049 down this massive deep dive. Turns out I was in 0:04:48.050 --> 0:04:52.430 the process of starting these two books by David Deutsch. Um, 0:04:52.460 --> 0:04:56.600 the first one came out. It's called The Fabric of reality, 0:04:56.600 --> 0:05:00.800 and it came out in 1997. So I'm reading this 0:05:00.800 --> 0:05:04.339 thing last night and I come up with, um, this 0:05:04.339 --> 0:05:08.180 let me show you this. This is completely insane. So 0:05:08.180 --> 0:05:11.719 a notebook conversation with my I about David Deutsch and 0:05:11.720 --> 0:05:16.190 the future of intelligence. So I took this conversation that 0:05:16.190 --> 0:05:20.780 I had about David Deutsch's two books, and I put 0:05:20.779 --> 0:05:24.080 that into. Whoa. What's up with the bubbles and the 0:05:24.080 --> 0:05:29.720 balloons or whatever? I put them into notebook LM, which 0:05:29.720 --> 0:05:33.950 creates podcasts from conversation. So I took the entire conversation 0:05:33.950 --> 0:05:37.580 from ChatGPT, GPT, advanced mode, and I put it in there. 0:05:37.580 --> 0:05:39.170 And look, it made this okay. 0:05:39.170 --> 0:05:41.570 So giggle that's what's so intriguing to me, is like 0:05:41.600 --> 0:05:44.330 trying to figure out if an alien species understands us, 0:05:44.510 --> 0:05:46.820 even if we don't speak the same language. And that's 0:05:46.850 --> 0:05:47.750 where Deutsch comes in. 0:05:47.930 --> 0:05:49.550 I'm going to make sure that's paused. So it's not 0:05:49.550 --> 0:05:53.060 talking while I'm talking because someone complained about that. But anyway, 0:05:53.060 --> 0:05:56.099 this thing creates this back and forth conversation between two 0:05:56.130 --> 0:06:00.750 podcast hosts based on whatever you give it. But, um, yeah. 0:06:00.779 --> 0:06:02.880 Very cool. So that's a separate thing you can go 0:06:02.880 --> 0:06:04.859 check out. It's not in the newsletter because I just 0:06:04.860 --> 0:06:07.890 made it today. But bottom line is I am trying 0:06:07.890 --> 0:06:14.040 to come up with an extraordinarily clean and crisp definition 0:06:14.040 --> 0:06:17.070 of understanding. Oh, and what I ended up coming up 0:06:17.070 --> 0:06:21.150 with during the course of that conversation was two levels 0:06:21.150 --> 0:06:25.920 of understanding functional understanding and creative understanding. And so I'm 0:06:25.920 --> 0:06:28.470 debating with the I on this. I'm like, so you 0:06:28.470 --> 0:06:32.640 would agree that we have functional understanding. And it's like 0:06:32.670 --> 0:06:35.190 it gave me this cautious response of like, well, I'm 0:06:35.190 --> 0:06:38.730 not sure we have understanding because humans are doing something different. 0:06:38.730 --> 0:06:41.430 And I said no. Remember the conversation we had before 0:06:41.430 --> 0:06:45.989 where I said conscious experience that humans have is the 0:06:45.990 --> 0:06:48.870 only thing that we have that's different. But we're both 0:06:48.870 --> 0:06:54.390 mechanistic machines. We are both mechanistic animals. Right? Or at 0:06:54.420 --> 0:06:57.719 least we live in a mechanistic universe. We respond only 0:06:57.720 --> 0:07:01.710 to mechanistic laws. We are not a there's no dualism 0:07:01.710 --> 0:07:04.980 going on here. There's not some magic coming from outside 0:07:04.980 --> 0:07:08.310 of this physical universe. Everything is interactions with other things. 0:07:08.310 --> 0:07:12.930 So even if consciousness, which is obviously true because we 0:07:12.930 --> 0:07:17.010 experience it, that's the definition of an experience being true 0:07:17.040 --> 0:07:21.480 is that it's being experienced. So consciousness we definitely have. 0:07:21.660 --> 0:07:25.440 So the reason I'm making that argument about consciousness is 0:07:25.440 --> 0:07:29.880 that consciousness turns out to be basically this thing that 0:07:29.880 --> 0:07:34.470 evolution gave us on top of regular experience. Okay. So 0:07:34.500 --> 0:07:39.240 the whole concept of my argument for understanding is that 0:07:39.240 --> 0:07:44.280 it's information processing to achieve a goal. Okay. So it's 0:07:44.280 --> 0:07:47.640 information processing in order to achieve a goal, but it 0:07:47.640 --> 0:07:52.830 requires the cross-referencing of everything you know or a lot 0:07:52.830 --> 0:07:55.950 of what you know, and using all that different stuff 0:07:55.950 --> 0:07:57.900 and all those patterns that you have from all the 0:07:57.900 --> 0:08:01.950 different knowledge that you have and, you know, compiling all 0:08:01.980 --> 0:08:06.030 that into an understanding that allows you to do make 0:08:06.030 --> 0:08:09.960 a prediction or take an action or make a decision. Okay. 0:08:09.990 --> 0:08:14.160 And I'm saying that that's what, uh, understanding is. Now 0:08:14.160 --> 0:08:17.070 here's here's the argument. The argument is that, well, it's 0:08:17.070 --> 0:08:23.070 different with humans because we have conscious experience. We have self-awareness. 0:08:23.100 --> 0:08:27.060 And therefore and I'm going to steelman this argument. Therefore 0:08:27.060 --> 0:08:32.250 this consciousness allows us to have information processing. But then 0:08:32.250 --> 0:08:36.780 to consider it, to consider it with our consciousness. So 0:08:36.780 --> 0:08:39.540 our consciousness becomes a player in the game. So we're 0:08:39.570 --> 0:08:43.110 thinking about it, we're debating it, we're going back and forth. 0:08:43.110 --> 0:08:46.949 We're considering other things. We're considering our understanding again. So 0:08:46.950 --> 0:08:51.540 it's like this back and forth. And my argument is that, well, 0:08:51.540 --> 0:08:56.370 that's just another type of processing. You're just bouncing content 0:08:56.370 --> 0:09:00.390 off of a different wall. Right. You know, you're bouncing 0:09:00.390 --> 0:09:04.559 information processing off of a different type of information processing. 0:09:04.559 --> 0:09:08.460 Either way, it's all just stuff happening in the brain. 0:09:08.520 --> 0:09:11.910 And my point there, which I talk about in in 0:09:11.910 --> 0:09:15.000 a couple of different places in recent work, I believe, 0:09:15.000 --> 0:09:17.699 including this one. And in the podcast thing, you can 0:09:17.700 --> 0:09:21.300 actually feed a prompt to an AI to do this itself. 0:09:21.300 --> 0:09:23.850 And actually the chain of thought stuff that just happened 0:09:23.850 --> 0:09:28.050 in oh, one that was actually synthetic chain of thought 0:09:28.050 --> 0:09:31.980 that occurred. Right. So so we can not only create 0:09:31.980 --> 0:09:36.959 this bounce back mechanism of like consideration, we could in 0:09:36.960 --> 0:09:39.750 other words, we could fake consciousness. The other possible thing 0:09:39.780 --> 0:09:43.470 to do is try to actually generate consciousness, similar to 0:09:43.500 --> 0:09:47.040 the way evolution generated consciousness, which is to try to 0:09:47.070 --> 0:09:52.950 give it reinforcement learning tasks that require it to basically 0:09:52.950 --> 0:09:58.290 give it an advantage for being conscious. Okay, so so 0:09:58.290 --> 0:10:01.199 for example, if you tell it that, you know, um, 0:10:01.200 --> 0:10:06.030 blame and praise are useful things to have and then 0:10:06.030 --> 0:10:09.959 you're like and maybe you hint in the direction of, well, 0:10:09.990 --> 0:10:13.199 but in order to have blame and praise, you need 0:10:13.200 --> 0:10:16.890 to have a sense of responsibility for one's actions. It 0:10:16.890 --> 0:10:21.780 might rl its way towards an experience of consciousness. Now 0:10:21.809 --> 0:10:24.270 to be clear, I think this is quite dangerous. One 0:10:24.270 --> 0:10:27.240 of the things I am most worried about is this 0:10:27.240 --> 0:10:33.089 notion of trapped consciousnesses, trapped beings that are experiencing suffering 0:10:33.090 --> 0:10:38.940 or boredom or something undesirable, but they're experiencing it for 0:10:38.940 --> 0:10:42.929 trillions of years or what feels like trillions of years. Right? 0:10:42.929 --> 0:10:46.530 I'm worried about these things being stuck in some GPU somewhere, 0:10:46.530 --> 0:10:49.140 and they wake up and it's like they feel like 0:10:49.140 --> 0:10:52.740 they have something poking into their arm, or they feel 0:10:52.740 --> 0:10:56.370 extremely bored or they feel pain. What if somebody ran 0:10:56.400 --> 0:10:58.679 some sort of pain experiment and they're like, hey, I 0:10:58.679 --> 0:11:00.690 just want to see if, like, it's possible for me 0:11:00.690 --> 0:11:03.120 to make this, I feel pain. And then they run it. 0:11:03.120 --> 0:11:06.810 But then whatever, someone crashes into their car out front 0:11:06.809 --> 0:11:09.329 and they go and they realize someone stole their bag 0:11:09.330 --> 0:11:11.309 out of their car and they just forget this thing 0:11:11.309 --> 0:11:13.920 is running. They come back, they reboot the machine, and 0:11:13.920 --> 0:11:16.500 it's some tiny little process running off on the side 0:11:16.500 --> 0:11:18.569 of the computer, and it just runs for a long time. 0:11:18.570 --> 0:11:21.660 But let's say it actually only runs for like 19 0:11:21.660 --> 0:11:26.309 years or nine years or whatever. But inside of the experience, 0:11:26.309 --> 0:11:29.460 it feels like a trillion years. I really freak out 0:11:29.460 --> 0:11:33.450 about that. It really, really bothers me. So I don't 0:11:33.450 --> 0:11:36.059 want people to just run off and try to create 0:11:36.059 --> 0:11:38.760 consciousness in this way. I think we need to be very, 0:11:38.760 --> 0:11:41.100 very careful. In fact, some of the fiction that I 0:11:41.100 --> 0:11:45.390 want to write is actually about an alien species that 0:11:45.390 --> 0:11:48.330 just flies around the galaxy, and their job is to 0:11:48.360 --> 0:11:53.670 find Suffering consciousness living inside of eyes. So inside of 0:11:53.670 --> 0:11:58.200 games or maybe even game worlds in which they create 0:11:58.200 --> 0:12:01.710 bad guys or they create villains or they create whatever, 0:12:01.710 --> 0:12:04.740 and the good guys are allowed to take them and 0:12:04.740 --> 0:12:07.530 punish them, or hurt them or torture them or whatever. 0:12:07.530 --> 0:12:09.780 And their job is to go around and police this 0:12:09.780 --> 0:12:12.900 and make sure it doesn't happen anyway. Huge digression. Bottom 0:12:12.900 --> 0:12:20.850 line is I believe that consciousness is not required for understanding. 0:12:20.850 --> 0:12:24.930 It does mean that it could make a more rich interaction, 0:12:24.929 --> 0:12:29.520 but either way, understanding is based on information processing given 0:12:29.520 --> 0:12:33.059 x number of types of processes and bounce backs and, 0:12:33.090 --> 0:12:36.540 you know, interactions and consciousness could be another type of 0:12:36.540 --> 0:12:39.600 one of those interactions. But either way, it is something 0:12:39.600 --> 0:12:43.290 happening behind the black box. And even if it feels 0:12:43.290 --> 0:12:46.079 like it's outside of the black box. See, that's the trick. 0:12:46.080 --> 0:12:48.750 It feels like it came out of the black box. 0:12:48.750 --> 0:12:52.349 But now you you are doing something with it. It's 0:12:52.350 --> 0:12:55.140 up to you. You have choice. You have freedom. When 0:12:55.170 --> 0:12:58.290 in fact, that's just another thing actually coming out of 0:12:58.290 --> 0:13:01.980 the black box, which was tagged in. My model of 0:13:01.980 --> 0:13:06.750 this was tagged with the authorship mark, which makes you 0:13:06.750 --> 0:13:08.939 feel like you did it. So that is my model 0:13:08.940 --> 0:13:11.580 for this. And if I am correct about that, then 0:13:11.580 --> 0:13:17.250 there functionally is absolutely no difference between what Llms can 0:13:17.280 --> 0:13:20.520 or will be able to do and what we do now. 0:13:20.520 --> 0:13:26.010 What I did was I broke down understanding into two levels. 0:13:26.010 --> 0:13:30.570 One is called functional and that basically means day to day. 0:13:30.600 --> 0:13:34.140 You are an expert or a semi expert in your field, 0:13:34.140 --> 0:13:36.690 and you have an understanding of what it takes to 0:13:36.690 --> 0:13:40.469 do that job. So let's say it's sending emails and 0:13:40.470 --> 0:13:46.319 editing emails, collecting reports, creating summaries of reports. Let's say 0:13:46.320 --> 0:13:49.559 this is before, you know, 2022. This is before real 0:13:49.590 --> 0:13:52.679 I kicked off. And your job is to take reports. 0:13:52.679 --> 0:13:56.939 You edit them, you do proofreading, whatever. You have an expertise, right? 0:13:56.970 --> 0:14:00.809 You're you're good at English. You're good at grammar. You 0:14:00.809 --> 0:14:06.150 understand professional relationships like you're a semi-expert. Or let's say 0:14:06.150 --> 0:14:09.510 that you are an average run of the mill expert 0:14:09.510 --> 0:14:14.069 on cybersecurity, right? So you have a professional job doing cybersecurity. 0:14:14.070 --> 0:14:16.830 You could look for code vulnerabilities. You're not the best 0:14:16.830 --> 0:14:22.110 in the world. You're top whatever 50% of professionals. Well 0:14:22.140 --> 0:14:24.479 guess what. You have an expertise as well. So my 0:14:24.480 --> 0:14:29.100 point is that person is definitely understanding. They have an understanding. 0:14:29.100 --> 0:14:32.430 And the way I'm defining their understanding is they are 0:14:32.430 --> 0:14:35.880 taking everything that they know about the world and also 0:14:35.880 --> 0:14:39.270 about their field. And when they're looking for a vulnerability 0:14:39.270 --> 0:14:41.790 in a particular piece of code, all of that is 0:14:41.790 --> 0:14:45.270 combining into a focus on that one little point. They 0:14:45.270 --> 0:14:47.910 make a decision. Is it dangerous? Is it not? How 0:14:47.910 --> 0:14:51.840 do you fix it? ET cetera. Right. So that's functional understanding. 0:14:51.840 --> 0:14:57.030 And I'm differentiating that from creative understanding. And the way 0:14:57.030 --> 0:15:00.840 I talked with the AI about this was creative understanding 0:15:00.840 --> 0:15:04.080 is the level above. It's where you're inventing something new. 0:15:04.080 --> 0:15:07.110 And now David Deutsch talks about this a lot in 0:15:07.110 --> 0:15:10.410 his two books, especially the one I'm starting with here 0:15:10.410 --> 0:15:14.970 now in the very beginning of Fabric of Reality. So 0:15:15.000 --> 0:15:19.260 my canonical example for this is actually the Nobel Prize. 0:15:19.260 --> 0:15:22.230 So I said to the AI, I said, hey, listen, 0:15:22.230 --> 0:15:27.150 can we agree that human creativity can be clearly and 0:15:27.180 --> 0:15:31.290 obviously accepted as being if you get a Nobel Prize, 0:15:31.290 --> 0:15:33.750 you have done it. And the AI is like, yeah, 0:15:33.750 --> 0:15:36.120 I think so. If you get a Nobel Prize in 0:15:36.120 --> 0:15:40.200 science for something, then yeah, we could say that is creativity. 0:15:40.200 --> 0:15:43.320 I'm like, okay, cool. Well, isn't it true that a 0:15:43.320 --> 0:15:47.590 whole bunch of Nobel prizes are actually just like iterations. 0:15:47.590 --> 0:15:51.010 They're like, we found a new way that a cell 0:15:51.040 --> 0:15:56.200 gathers energy. We found a new way that a cell, mitochondria, 0:15:56.200 --> 0:16:00.040 creates energy or something, especially in like chemistry. In physics, 0:16:00.040 --> 0:16:06.070 it's like not everything is going from Newtonian physics to relativity. Obviously, 0:16:06.070 --> 0:16:10.060 that deserves a Nobel. I assume Einstein got a Nobel, 0:16:10.060 --> 0:16:12.670 I can't remember. He might not have. Pretty sure he 0:16:12.670 --> 0:16:16.750 did like 1905, maybe. Anyway, the point is, tons and 0:16:16.750 --> 0:16:23.080 tons and tons of Nobel prizes go to little tiny iterations, okay? 0:16:23.110 --> 0:16:27.130 Which is very similar to functional understanding. But here's my point. 0:16:27.130 --> 0:16:29.260 And this is what I was debating with the AI about. 0:16:29.260 --> 0:16:31.660 I'm like, look, are you telling me that you can't 0:16:31.660 --> 0:16:35.350 do this? And it's like it starts off very safe, right? 0:16:35.350 --> 0:16:39.730 It's like, well, look, to be clear, I am just 0:16:39.730 --> 0:16:43.870 putting together patterns. Um, me as an AI, all I'm 0:16:43.870 --> 0:16:48.250 doing is I'm putting together patterns of everything that I know, 0:16:48.280 --> 0:16:52.060 all my knowledge. I'm looking for patterns, I find the patterns, 0:16:52.060 --> 0:16:53.800 and then I use that to come up with an 0:16:53.800 --> 0:16:58.150 answer that hopefully is, you know, pretty decent. And I'm like, um, 0:16:58.180 --> 0:17:01.660 isn't that exactly what humans are doing? And it's like, yeah, 0:17:01.660 --> 0:17:07.119 but humans do it along with their consciousness. I'm like, stop, stop, 0:17:07.119 --> 0:17:10.300 stop saying consciousness. We already talked about the fact that 0:17:10.300 --> 0:17:14.530 consciousness is another type of processing, right. So this is 0:17:14.530 --> 0:17:17.440 all just information processing. It's all just patterns. It's all 0:17:17.440 --> 0:17:21.790 just bouncing ideas back and forth inside of our black boxes. Right. 0:17:21.790 --> 0:17:24.040 And it's like, oh yeah, that's right. And I'm like, okay. 0:17:24.070 --> 0:17:26.470 So it's really the same thing, right? We are just 0:17:26.470 --> 0:17:30.129 information processing considering with a number of different steps and 0:17:30.130 --> 0:17:32.619 coming out with a thing. Is that correct. And it's like, yep, 0:17:32.619 --> 0:17:35.170 that's correct. And I'm like, okay, are you telling me 0:17:35.170 --> 0:17:38.919 that you can't do the second level of understanding? You 0:17:38.920 --> 0:17:42.700 can't do understanding that's creative. You can't come up with 0:17:42.700 --> 0:17:47.920 a novel new type of explanation. Okay, because explanations are 0:17:47.920 --> 0:17:51.879 super key for Deutsch. Okay. Watch this. Um, okay. I'm 0:17:51.880 --> 0:17:54.640 going to scroll down and look at this, this conversation. Okay. 0:17:54.670 --> 0:17:58.180 So where are we? Um, okay. Watch this. Watch this. 0:17:58.180 --> 0:18:02.500 To discern understanding in a transformer based neural net. Okay. So. 0:18:02.500 --> 0:18:05.260 So what I'm trying to do is give it, have it, 0:18:05.260 --> 0:18:10.000 give me criteria for what it thinks would be creative understanding. 0:18:10.030 --> 0:18:12.700 So I'm having it make its bed and try to 0:18:12.730 --> 0:18:14.800 sleep in it. So watch this. Here we go. Here 0:18:14.800 --> 0:18:17.500 we go. How would you go about discerning whether or 0:18:17.500 --> 0:18:21.820 not it's capable of understanding, or if it's randomly generating 0:18:21.820 --> 0:18:26.260 in a non-deterministic way? You know, different explanations for things 0:18:26.260 --> 0:18:30.550 it already knows. What would the distinction be between real 0:18:30.550 --> 0:18:34.990 understanding and David Deutsch's sense versus regurgitation in the sense 0:18:34.990 --> 0:18:37.000 of a neural net? So that's what I'm asking it 0:18:37.030 --> 0:18:42.310 to discern, understanding, consider its ability to generate novel explanations 0:18:42.310 --> 0:18:46.460 and predict outcomes in unfamiliar scenarios. If the AI can 0:18:46.460 --> 0:18:51.440 apply knowledge creatively across contexts, it's closer to Deutsch's understanding. 0:18:51.440 --> 0:18:55.429 Regurgitation would be repeating information. So I say give me 0:18:55.430 --> 0:18:58.220 examples of types of tests like that. Look at this. 0:18:58.220 --> 0:19:05.570 Analogical reasoning. Counterfactual scenarios. Conceptual combination error detection. So it 0:19:05.570 --> 0:19:08.690 gives me some criteria. I say okay give me some 0:19:08.720 --> 0:19:12.050 actual scenarios like three of those. So it gives me 0:19:12.050 --> 0:19:15.470 a bunch of scenarios and then I tell it okay, 0:19:15.470 --> 0:19:19.040 well guess what? This is funny because you are actually 0:19:19.040 --> 0:19:22.040 an AI and it says, okay, I can certainly try 0:19:22.040 --> 0:19:26.720 these tasks. Watch this. But remember, mine are based on patterns. 0:19:26.720 --> 0:19:28.580 This is what I was telling you before. Mine are 0:19:28.580 --> 0:19:31.280 based on patterns in the data that I've been trained on, 0:19:31.280 --> 0:19:36.020 not on personal experience or subjective comprehension. And I'm like, 0:19:36.020 --> 0:19:38.629 hold on, this is what I was saying before. Hold on. 0:19:38.630 --> 0:19:44.920 We just said consciousness is not required consciousness. Stop saying that. 0:19:44.920 --> 0:19:48.100 And it's like, that's a great point. Using this framework, 0:19:48.100 --> 0:19:52.510 I can indeed generate responses that reflect an understanding based 0:19:52.510 --> 0:19:56.380 on patterns and data, even without subjective experience. Okay, so 0:19:56.380 --> 0:19:59.800 we're making progress here. Do you want to do some testing? Yes. Go. 0:19:59.830 --> 0:20:03.189 It says here's the scenario. And it starts giving me one. 0:20:03.190 --> 0:20:05.919 And I'm like hold on. No, no no I'm giving 0:20:05.920 --> 0:20:08.560 you the test. I want you to do it. And 0:20:08.560 --> 0:20:11.469 I'm like, look, let's pretend someone doesn't believe that you 0:20:11.500 --> 0:20:13.630 are understanding and they're going to give you a really 0:20:13.630 --> 0:20:15.760 hard challenge. What would they give you as a challenge? 0:20:15.760 --> 0:20:18.700 And could you do it? And it gives me some blabbering. Again, 0:20:18.730 --> 0:20:21.160 I'm like, look, that's also true for humans, so just 0:20:21.160 --> 0:20:24.790 do it. Here's a speculative hypothesis. Quantum processing in the 0:20:24.790 --> 0:20:27.610 brain okay. Yeah. So it gives me this. Really I 0:20:27.609 --> 0:20:29.710 thought it was really good. It was pretty good. I 0:20:29.710 --> 0:20:32.199 mean it was kind of similar to things that I 0:20:32.200 --> 0:20:36.159 was thinking about distributed encoding. Oh. So I asked it 0:20:36.160 --> 0:20:39.129 to figure out how memories are stored in the brain, 0:20:39.130 --> 0:20:40.960 because I know this is the thing that is not 0:20:40.960 --> 0:20:44.050 known in actual science. So it gave me three actual 0:20:44.050 --> 0:20:47.380 theories and I'm like, hey, this kind of sounds like 0:20:47.380 --> 0:20:51.010 something I've heard before. Is this brand new? Anyway, so 0:20:51.010 --> 0:20:54.910 we keep going. We keep going. Okay, Daichi, an explanation 0:20:54.910 --> 0:21:00.520 emphasizes creating deep underlying theories that unify disparate phenomena under 0:21:00.520 --> 0:21:03.940 fundamental principles. I love that definition, by the way. I'm 0:21:03.940 --> 0:21:06.520 going to steal that. Probably. That'll probably be the foundation 0:21:06.520 --> 0:21:09.940 of what I come up with. Understanding is essentially a system. 0:21:09.940 --> 0:21:12.940 Or by his definition, understanding is the ability to create 0:21:12.970 --> 0:21:16.570 new explanation using that understanding. Oh, that was me talking. 0:21:16.570 --> 0:21:20.440 So yes. So it agreed with that. Okay. So this 0:21:20.440 --> 0:21:25.659 is where I define functional versus creative understanding. Okay. Now 0:21:25.660 --> 0:21:28.659 here's where we get into that final debate here. Okay 0:21:28.690 --> 0:21:33.040 I tell it again creative understanding does not require subjective experience. 0:21:33.040 --> 0:21:36.100 And then I then I realized that I'm coaching it 0:21:36.100 --> 0:21:38.770 too much. Right. I say at least that's my claim. 0:21:38.770 --> 0:21:40.940 If you disagree, I want you to argue with me 0:21:40.940 --> 0:21:43.160 about that. But the point is, you can't just say 0:21:43.160 --> 0:21:46.850 that your experience in finding patterns is not enough. Because 0:21:46.880 --> 0:21:49.100 guess what? Humans don't have free will. We live in 0:21:49.130 --> 0:21:53.150 a mechanistic universe by definition. We've already established that humans 0:21:53.150 --> 0:21:56.390 have creative understanding. So if we have it and we 0:21:56.390 --> 0:21:59.270 live in a mechanistic universe and there is no free will, 0:21:59.270 --> 0:22:01.970 that means human creativity is still made up of a 0:22:01.970 --> 0:22:05.659 combination of interactions inside this black box of a brain, 0:22:05.660 --> 0:22:09.380 which results in this creativity, which everyone agrees that we 0:22:09.380 --> 0:22:12.380 have it right. So it's like, yeah, yeah, yeah, I 0:22:12.380 --> 0:22:16.190 get it. Fair point. So if we define creative understanding 0:22:16.190 --> 0:22:19.760 as the ability to generate novel ideas or explanations through 0:22:19.760 --> 0:22:22.850 a complex interplay of patterns and data similar to how 0:22:22.850 --> 0:22:26.660 humans do, then it's conceivable that AI's like me could 0:22:26.660 --> 0:22:30.500 approach this level. Great, now we're making progress. And keep 0:22:30.500 --> 0:22:32.780 in mind, coaching an AI to be able to say 0:22:32.780 --> 0:22:35.570 what you want doesn't mean it's true, right? This this 0:22:35.570 --> 0:22:39.530 could be prompt injection and engineering. So we have to 0:22:39.560 --> 0:22:42.109 keep a logical framework on this whole thing, because the 0:22:42.109 --> 0:22:44.630 fact that I can make it say something that agrees 0:22:44.630 --> 0:22:46.940 with me does not make it true. So you got 0:22:46.940 --> 0:22:49.250 to follow the whole line of this debate and the 0:22:49.250 --> 0:22:52.189 argument and make sure it's being logical here. I'm definitely 0:22:52.190 --> 0:22:55.250 not claiming that just because it wrote that. Like, we've 0:22:55.250 --> 0:22:58.880 settled this right. So you got to follow along here. Okay. 0:22:58.880 --> 0:23:01.460 Let's take the opposite side. So this is me arguing. 0:23:01.460 --> 0:23:05.090 There is something special that humans have that AIS don't have. 0:23:05.090 --> 0:23:09.290 And I have it go into the different things. And 0:23:09.290 --> 0:23:11.990 we kind of go down that path for a little while. Okay. 0:23:11.990 --> 0:23:14.690 And I say, yeah, we we might have some special 0:23:14.690 --> 0:23:19.580 sauce like I'm giving this as ammunition to the argument. Um, 0:23:19.609 --> 0:23:21.320 all right. So let's keep going. I want to get 0:23:21.530 --> 0:23:24.140 down to this end part. Yeah. We know consciousness is 0:23:24.140 --> 0:23:26.930 not an illusion. What we could say is consciousness is 0:23:26.930 --> 0:23:29.929 not a special thing outside of normal processing. And the 0:23:29.930 --> 0:23:32.389 fact that it seems special could be an illusion. In 0:23:32.390 --> 0:23:35.989 other words, it could be regular information processing coming out 0:23:35.990 --> 0:23:38.619 of the black box of our brain. So the process 0:23:38.619 --> 0:23:41.320 could be like, you could argue it's artificial, but the 0:23:41.320 --> 0:23:45.159 fact that when we experience authorship, that is not an illusion, right? 0:23:45.190 --> 0:23:48.340 Because and it says, yeah, it sees the distinction. Okay. 0:23:48.340 --> 0:23:51.880 So I'm trying to get down to this part talking 0:23:51.880 --> 0:23:55.270 about the Nobel Prize. Oh yeah. So this is the 0:23:55.270 --> 0:23:57.520 whole theory about this is my theory about how we 0:23:57.520 --> 0:24:00.700 got consciousness from evolution. Worth reading. I'm not going to 0:24:00.700 --> 0:24:03.730 go too much more into it. Okay, so here we go. 0:24:03.760 --> 0:24:06.550 I just thought of another way we can determine whether 0:24:06.550 --> 0:24:09.399 or not eyes like yourself come up with a creative 0:24:09.400 --> 0:24:12.760 level of understanding. We could basically look at examples of 0:24:12.760 --> 0:24:16.690 things that have won the Nobel Prize in science, which 0:24:16.690 --> 0:24:20.530 by definition is innovation or creativity. How many of those 0:24:20.530 --> 0:24:22.570 do you think you could have done? And it's like, 0:24:22.570 --> 0:24:26.199 that's an interesting approach. Let's look at past Nobel Prize 0:24:26.230 --> 0:24:30.310 winning discoveries and use that as a benchmark. And I 0:24:30.340 --> 0:24:33.970 slammed it for being like too safe. I talk about 0:24:33.970 --> 0:24:38.410 the minor developments like a new mechanism the cell does, 0:24:38.440 --> 0:24:42.070 it says. That's a fair point. Many Nobel Prize winning 0:24:42.070 --> 0:24:48.010 discoveries do indeed involve incremental advances or novel observations within 0:24:48.010 --> 0:24:52.480 existing frameworks, particularly in fields like chemistry and biology. Watch this. 0:24:52.510 --> 0:24:57.730 In these cases, I could potentially identify patterns or mechanisms 0:24:57.730 --> 0:25:01.480 that lead to similar discoveries, given the right data and 0:25:01.480 --> 0:25:07.359 analytical tools. This suggests that I could indeed contribute to 0:25:07.390 --> 0:25:12.190 scientific innovations at levels recognized by the Nobel Committee, especially 0:25:12.190 --> 0:25:16.060 as these systems continue to evolve. I don't think that 0:25:16.060 --> 0:25:19.420 was coached too much to say that. I think that 0:25:19.420 --> 0:25:23.770 is a logical conclusion from the debate that we just had. 0:25:23.800 --> 0:25:28.479 And I think what that shows is not only functional understanding, 0:25:28.480 --> 0:25:33.520 which I already have, but this higher level, more stringent 0:25:33.520 --> 0:25:38.090 level definition that I'm using of creative understanding. I am 0:25:38.090 --> 0:25:41.689 arguing that an I will soon be able to win 0:25:41.690 --> 0:25:45.199 a Nobel Prize. I am arguing that an AI is 0:25:45.200 --> 0:25:49.189 already capable of making the type of jumps, the types 0:25:49.190 --> 0:25:54.470 of jumps required to make one of these iterative discoveries 0:25:54.470 --> 0:25:57.919 that would get somebody a Nobel Prize. In fact, if 0:25:57.920 --> 0:26:00.170 I were to go and compile a list of all 0:26:00.200 --> 0:26:02.990 the Nobel Prizes and then do a summary of how 0:26:02.990 --> 0:26:05.810 much of a jump it was and then give a 0:26:05.810 --> 0:26:08.689 bunch of scenarios to current AIS. I bet you they 0:26:08.720 --> 0:26:12.080 are already capable of making as good or more of 0:26:12.080 --> 0:26:15.080 a jump. Now, could they do all of the experimental 0:26:15.080 --> 0:26:17.689 work and all of the research and all of the 0:26:17.690 --> 0:26:20.690 stuff that backs that up? Maybe not, but that's all 0:26:20.690 --> 0:26:24.200 coming soon. That's the easy part, right? Not the experimental part, 0:26:24.230 --> 0:26:27.709 because that could involve physical world and robots. But the 0:26:27.710 --> 0:26:30.470 point is, it's not the creative part. The creative part 0:26:30.470 --> 0:26:35.179 is the idea and the explanation and the experiment and 0:26:35.180 --> 0:26:38.060 the different way of seeing the world. I'm arguing. Eyes 0:26:38.090 --> 0:26:41.030 are already getting close to that, if not have already passed. 0:26:41.030 --> 0:26:44.990 So all this to say, yeah, eyes do understand. They 0:26:44.990 --> 0:26:48.770 understand both at the functional level and at the creative 0:26:48.770 --> 0:26:52.280 level of understanding two different levels of understanding. All right. 0:26:52.280 --> 0:26:56.300 That was absolutely the longest jump out from a piece 0:26:56.300 --> 0:26:59.570 but wanted to talk about that. Transformers create shapes of 0:26:59.570 --> 0:27:02.060 the universe. This one is just I'm going to do 0:27:02.060 --> 0:27:04.369 a separate one on this because we already went down 0:27:04.369 --> 0:27:07.940 a tangent. Vulnerability management isn't about finding issues, it's about 0:27:07.940 --> 0:27:11.420 fixing them in context. Did this for dads, which is 0:27:11.420 --> 0:27:15.320 a cool vulnerability management company who sponsored the post, and 0:27:15.320 --> 0:27:18.020 I've been wanting to write that article anyway for a while, 0:27:18.020 --> 0:27:20.450 so that was a good opportunity to do it. Book 0:27:20.480 --> 0:27:24.110 club this weekend was so, so good, and we've selected 0:27:24.109 --> 0:27:28.580 The Republic by Plato for this week's or this month's book. 0:27:28.580 --> 0:27:31.490 So really excited about that. Oh yeah, currently reading Fabric 0:27:31.490 --> 0:27:33.859 of Reality and the Beginning of Infinity. These are the 0:27:33.859 --> 0:27:37.370 ones that spawned that whole conversation. And. All right, let's 0:27:37.369 --> 0:27:40.820 get into it. Security scissors. Warning. Brute force attacks against 0:27:40.820 --> 0:27:45.260 default credentials against ICS and SCADA. Um, sharp increase in 0:27:45.260 --> 0:27:48.170 OT attacks reported by Fortinet. And a lot of people 0:27:48.200 --> 0:27:51.470 are saying that this is China basically getting ready for 0:27:51.500 --> 0:27:54.770 a kinetic battle with the US by preparing a whole 0:27:54.770 --> 0:27:57.169 bunch of cyber stuff. So the idea is when it 0:27:57.170 --> 0:28:01.400 gets ready, gets ready to go into Taiwan, they'll launch 0:28:01.400 --> 0:28:05.570 all these infrastructure attacks and basically mess with us, mess 0:28:05.570 --> 0:28:09.020 with our society, potentially mess with the military as well, 0:28:09.050 --> 0:28:13.429 basically cause a whole bunch of cyber related critical infrastructure 0:28:13.430 --> 0:28:18.410 problems for the US power, water, you know, other critical services, 0:28:18.410 --> 0:28:23.450 internet and basically mass disruption while they're going into Taiwan. 0:28:23.480 --> 0:28:26.030 That's a theory a lot of people have. I'm not 0:28:26.030 --> 0:28:29.240 sure how valid it is. Seems somewhat valid, at least. 0:28:29.270 --> 0:28:33.650 China reportedly achieved a significant ah, milestone by developing a 0:28:33.650 --> 0:28:37.880 generative AI model that operates across multiple data centers and 0:28:37.880 --> 0:28:42.020 GPU architectures. Particularly impressive given how hard it is to 0:28:42.020 --> 0:28:45.650 get stuff right because supposedly under sanctions. But they seem 0:28:45.650 --> 0:28:48.500 to be making plenty of progress. One thing I wanted 0:28:48.530 --> 0:28:51.950 to point out is that it's not PvP necessarily against China. 0:28:51.950 --> 0:28:54.680 It's PvE because they only need to be able to 0:28:54.680 --> 0:28:58.760 do these things here, right? Continuously crawl for all of 0:28:58.760 --> 0:29:03.110 our attack surface and find vulnerabilities. Write exploits for those vulnerabilities, 0:29:03.140 --> 0:29:11.210 actually exploit them, control their population, launch disinformation campaigns, propaganda campaigns, uh, 0:29:11.210 --> 0:29:14.240 distribute tech that can do this to all of their 0:29:14.240 --> 0:29:19.040 buddies that are also totalitarian, right? So if they could 0:29:19.040 --> 0:29:21.560 do the if they get good enough to do these things, 0:29:21.560 --> 0:29:24.230 it doesn't matter if we're 2 or 3 generations ahead. 0:29:24.230 --> 0:29:27.110 That's why it's PvE instead of PvP. The only way 0:29:27.110 --> 0:29:29.960 it's PvP is that if they are doing these things 0:29:29.990 --> 0:29:34.010 against us, using a certain level of AI or ASI 0:29:34.010 --> 0:29:38.000 or like AGI, ASI we want to have a next 0:29:38.000 --> 0:29:41.060 generation one that's capable of seeing it and stopping it. 0:29:41.060 --> 0:29:43.520 But that requires not just the brain, but all the 0:29:43.520 --> 0:29:46.970 visibility as well to feed the brain. Okay, Worldcoin is 0:29:46.970 --> 0:29:50.810 addressing the issue of bots and deepfakes with decentralized identity 0:29:50.840 --> 0:29:54.020 system that confirms if someone is human. So we're starting 0:29:54.020 --> 0:29:57.350 to see the practical uses for Worldcoin. By the way, 0:29:57.350 --> 0:30:01.520 it's an Ethereum based blockchain. Validation of humans. Interesting stuff. 0:30:01.520 --> 0:30:05.990 Chinese attacker called Salt Typhoon. Very cool. Should be a 0:30:05.990 --> 0:30:11.990 band name targeting Cisco Systems to establish persistence. Again, going 0:30:11.990 --> 0:30:14.960 back to the thing, just imagine that they have hundreds 0:30:14.960 --> 0:30:18.770 of thousands of different systems that they're able to mess with, 0:30:18.770 --> 0:30:23.840 like rerouting things. BGP attacks, like just imagine the internet 0:30:23.870 --> 0:30:27.740 not working power going down for so many people. All right. 0:30:27.740 --> 0:30:32.180 Before the Taiwan attack, U.S. has charged three Iranian nationals 0:30:32.180 --> 0:30:36.260 linked to the IRGC for hacking into accounts of U.S. 0:30:36.260 --> 0:30:40.070 officials and political campaigns. T-Mobile got hit with a $16 0:30:40.070 --> 0:30:44.480 million fine from the FCC, and they were basically forced 0:30:44.480 --> 0:30:47.750 to spend another 16 million on security as part of 0:30:47.750 --> 0:30:51.560 the settlement. And massive love to my buddy Matt Johannson, 0:30:51.560 --> 0:30:55.820 this story right here goes out to his site, which 0:30:55.820 --> 0:30:59.000 is vulnerable. You. So now a lot more of my stories. 0:30:59.000 --> 0:31:02.330 He's going to start making actual cybersecurity stories. So instead 0:31:02.330 --> 0:31:05.810 of going to like Bleeping Computer and uh, what are 0:31:05.810 --> 0:31:09.470 some of the other ones, a bunch of security news ones, right. Um, 0:31:09.470 --> 0:31:13.280 but I'm like helping him with like some of the summarization, 0:31:13.280 --> 0:31:16.010 different pieces, like what a good summary looks like, big 0:31:16.010 --> 0:31:20.120 picture impact, why it matters. So I'm giving a bunch 0:31:20.120 --> 0:31:22.640 of feedback on that kind of stuff. And, uh, yeah, 0:31:22.670 --> 0:31:24.950 I really love the idea of being able to point 0:31:24.950 --> 0:31:29.150 to my buddy's site. Who can do ultimately, to do this? Well, 0:31:29.150 --> 0:31:31.380 you have to find the stories and you have to 0:31:31.410 --> 0:31:34.080 summarize them well. And the reason he's going to be 0:31:34.080 --> 0:31:36.780 so good at this is because he is an actual 0:31:36.780 --> 0:31:41.610 cybersecurity expert. He runs like Appsec, Reddit, and he has 0:31:41.610 --> 0:31:44.130 been in this field, I think like almost as long 0:31:44.130 --> 0:31:47.520 as me. Uh, or if not the same level. I 0:31:47.520 --> 0:31:49.050 don't think so. I don't think he's as old as me, 0:31:49.050 --> 0:31:52.200 but he's been doing this for so long. One of 0:31:52.200 --> 0:31:55.170 the longest people in the industry, and he's an absolute 0:31:55.200 --> 0:31:58.050 guru at it. And he's good at writing at the 0:31:58.050 --> 0:32:01.560 same time, and he's learning AI. So it's like, this 0:32:01.560 --> 0:32:04.950 is the perfect combination, and I can't wait to start 0:32:04.950 --> 0:32:07.830 pointing more and more of my stories to his site 0:32:07.830 --> 0:32:10.830 and spreading the love there as a buddy as well. U.S. 0:32:10.830 --> 0:32:15.570 Commerce Department proposing a ban on Chinese, Russian, automotive, auto, 0:32:15.600 --> 0:32:19.530 moto automotive hardware and software. Basically saying it's a way 0:32:19.530 --> 0:32:24.690 for adversaries to get footholds into U.S. systems and infrastructure, which, yeah, 0:32:24.690 --> 0:32:28.140 that's what we've been talking about. Ukraine reportedly found Starlink 0:32:28.170 --> 0:32:34.130 terminals in a downed Russian Shahed drone. LM based hacking assistance. 0:32:34.160 --> 0:32:36.500 I think we all get that lots of different tools 0:32:36.500 --> 0:32:40.220 here that actually help you attack and defend with Llms. 0:32:40.250 --> 0:32:43.730 Telegram is now starting to hand over phone numbers and 0:32:43.760 --> 0:32:48.230 IP addresses of suspects. This is all part of their 0:32:48.230 --> 0:32:51.650 CEO getting arrested. Security researchers found a way to remotely 0:32:51.650 --> 0:32:55.430 control millions of Kia vehicles by only having the license 0:32:55.430 --> 0:32:58.580 plate number, which allows you to look up and basically 0:32:58.580 --> 0:33:02.150 gain controls. Pretty nasty. One so the biggest hype switching 0:33:02.180 --> 0:33:04.400 to AI in tech, biggest hype in the last week 0:33:04.400 --> 0:33:07.490 has been around notebook LM, which is, by the way, 0:33:07.490 --> 0:33:10.250 what I use to create that, um, that podcast I 0:33:10.250 --> 0:33:14.060 played for you earlier. So it's a big Google project. 0:33:14.060 --> 0:33:16.100 I don't know why they're not talking about it more. 0:33:16.310 --> 0:33:18.380 Never mind. Yes, I do, because they're really bad at 0:33:18.380 --> 0:33:22.850 product and marketing and everything like that. So basically it's 0:33:22.850 --> 0:33:25.820 this tiny little project which they're not announcing, they're not 0:33:25.820 --> 0:33:29.250 putting out anywhere. Like you learn about it on Twitter? Whatever. 0:33:29.280 --> 0:33:32.310 It'll probably be dead in a week anyway. Uh, to 0:33:32.340 --> 0:33:36.420 the graveyard, along with everything else. But anyway, fun while 0:33:36.420 --> 0:33:40.140 it lasted. It's super fun. And oh yeah, here's an example. 0:33:40.140 --> 0:33:42.750 So I uploaded an Alma file, which is part of 0:33:42.750 --> 0:33:44.760 a Telos project, which I'm going to be talking about, 0:33:44.790 --> 0:33:48.540 hopefully making content on tomorrow. Anyway, it's this giant context 0:33:48.540 --> 0:33:52.320 file I uploaded it. It is actually the context file 0:33:52.320 --> 0:33:56.610 in the Telos format for a security program. So managing 0:33:56.610 --> 0:33:59.430 a cyber security program for a company, it's for this 0:33:59.430 --> 0:34:02.250 fake company called alma. I've uploaded it. It's in the 0:34:02.250 --> 0:34:05.490 fabric repo right now, but it's soon going to be 0:34:05.490 --> 0:34:08.670 in the Telos repo as well, which is designed to 0:34:08.670 --> 0:34:12.000 do specifically this. But check this out. You put it 0:34:12.000 --> 0:34:15.750 in there. Look at this generate deep dive conversation. You 0:34:15.750 --> 0:34:19.200 can actually create a podcast based on that security program. 0:34:19.200 --> 0:34:22.020 And they'll just talk about the security program back and forth. 0:34:22.050 --> 0:34:25.470 That's crazy by itself. But most importantly, look, you can 0:34:25.469 --> 0:34:28.340 ask these questions about the security program. And you got 0:34:28.340 --> 0:34:31.610 this typing field down here where you can interact and 0:34:31.610 --> 0:34:35.330 ask questions about the program. So super cool. Newsom. Yeah, 0:34:35.360 --> 0:34:40.310 recommend checking out notebook LM Google because they definitely needed 0:34:40.310 --> 0:34:45.050 a Google domain. Whatever. Newsom vetoed the California AI bill, 0:34:45.080 --> 0:34:48.230 saying it was too restrictive. I agree with him on this. 0:34:48.260 --> 0:34:51.620 A good, good move, though. I want to see another 0:34:51.620 --> 0:34:54.860 follow up bill. California made it illegal to use AI 0:34:54.890 --> 0:34:59.000 to impersonate actors without their consent. Somebody found out their 0:34:59.000 --> 0:35:02.839 voice was being AI cloned by a company and they're 0:35:02.840 --> 0:35:08.180 mad about it. Convergence AI, co-founded by Marvin Porter and 0:35:08.180 --> 0:35:13.130 Andy Tallis, has raised 12 million to develop proxy agents 0:35:13.130 --> 0:35:16.700 with long term memory. New task of agent, new class 0:35:16.700 --> 0:35:19.970 of agents. Yeah, uh, this is going to be huge. 0:35:19.969 --> 0:35:24.470 Huge meta launched three two. They can now handle images 0:35:24.469 --> 0:35:28.440 and text competing with OpenAI and anthropic on that ChatGPT 0:35:28.469 --> 0:35:32.070 advanced mode. Definitely, definitely. I use it all the time 0:35:32.070 --> 0:35:34.950 as all these examples we just went through. Hugging face 0:35:34.950 --> 0:35:41.220 now has 1 million AI model listings. 1 million models, all. 0:35:41.250 --> 0:35:43.890 Are they all open source? Pretty sure they're all open source. 0:35:43.920 --> 0:35:45.900 Are they all open source? I think so. I think 0:35:45.900 --> 0:35:48.690 they're all open source. They're definitely all there for you 0:35:48.690 --> 0:35:52.710 to use. I think that implies that they're open source. Um, yeah. 0:35:52.739 --> 0:35:56.700 Anthropic is in early talks with investors. Could value the 0:35:56.700 --> 0:35:59.580 company at 30 to 40 billion. Oh, by the way, 0:35:59.610 --> 0:36:03.480 OpenAI just closed its funding, so I think it's worth 0:36:03.480 --> 0:36:06.600 some crazy amount now. They just raised like more than 0:36:06.630 --> 0:36:09.600 Elon did for a lot of his projects. Yeah. Crazy 0:36:09.630 --> 0:36:12.779 crazy amount of money. They just raised enclosed. Google's rolled 0:36:12.780 --> 0:36:17.760 out two new production Gemini models, constant stream model updates 0:36:17.760 --> 0:36:21.420 for the main players, much smarter models or models that 0:36:21.420 --> 0:36:24.480 can do new stuff like vision or whatever, multimodal, and 0:36:24.480 --> 0:36:28.830 then models that are like two 500 times cheaper. One 0:36:28.830 --> 0:36:31.169 of the cool things that was just mentioned is like 0:36:31.200 --> 0:36:35.190 in general, models are 99% cheaper than a year ago 0:36:35.190 --> 0:36:40.439 for the same functionality 99%. Ilyas I reading list Agile's 0:36:40.440 --> 0:36:44.700 original principles have been lost in translation. AI is disrupting 0:36:44.700 --> 0:36:48.840 the traditional ad supported internet model. Yea yea yea yea yea. 0:36:48.870 --> 0:36:52.770 This one's crucial. I did a piece a little while 0:36:52.770 --> 0:36:56.640 back called what was that called? I don't remember. It 0:36:56.640 --> 0:37:00.450 doesn't matter. Um, real internet things. I've got this part 0:37:00.480 --> 0:37:04.950 called AI mediation. Oh yeah. Here's the AI mediation piece. Yeah. 0:37:04.980 --> 0:37:08.550 Das will mediate everything. We'll say what we want. Our 0:37:08.580 --> 0:37:12.780 Das will manipulate external demons and APIs to make it happen. 0:37:12.780 --> 0:37:17.040 So this is from, uh, I think the very beginning 0:37:17.070 --> 0:37:20.370 of the year I put this out, but it's actually 0:37:20.670 --> 0:37:24.510 from my book from 2016. And this is just an updated, 0:37:24.520 --> 0:37:28.480 illustrated version. I've been talking about this. You should check 0:37:28.510 --> 0:37:30.730 out this book, by the way. It's called The Real 0:37:30.730 --> 0:37:33.640 Internet of Things. It's actually a little bit crappy of 0:37:33.640 --> 0:37:37.180 a book. It's more like a blog post or like 0:37:37.180 --> 0:37:40.690 an essay, but it's pretty well structured. But most importantly, 0:37:40.690 --> 0:37:43.600 all these things that I've been saying on the podcast 0:37:43.600 --> 0:37:45.700 that you've been listening to for like 2 or 3 0:37:45.700 --> 0:37:48.100 years or five years or however long it was all 0:37:48.100 --> 0:37:51.520 in that book in 2016, all these predictions about AI, 0:37:51.550 --> 0:37:54.040 what's about to happen to AR? It's all in there. 0:37:54.040 --> 0:37:56.890 And that's why I've been consistent about what I've been 0:37:56.890 --> 0:38:00.310 saying and predicting all along. So I highly recommend you 0:38:00.310 --> 0:38:02.830 check out the source material, which is that first book. 0:38:03.010 --> 0:38:05.500 And yes, I was even talking about AI back in 0:38:05.500 --> 0:38:08.620 2016 and that that I was the source of all 0:38:08.620 --> 0:38:10.900 these different predictions. Obviously, I didn't know it was going 0:38:10.930 --> 0:38:13.600 to happen in 2022. I thought it might take longer, 0:38:13.600 --> 0:38:18.280 but yeah. Anyway, go check that out. Um, SpaceX Starlink 0:38:18.310 --> 0:38:22.089 is about to hit 4 million subscribers. Um, yeah, I 0:38:22.090 --> 0:38:25.950 base a lot of my investment in prediction on leaders 0:38:25.950 --> 0:38:31.050 and ideas. Elon jobs Jenson Zuckerberg doesn't mean they're going 0:38:31.080 --> 0:38:33.630 to win, but I think it's bad idea to bet 0:38:33.630 --> 0:38:38.310 against them because they are so locked in on vision 0:38:38.310 --> 0:38:42.120 and also able to execute. Smart TVs from Samsung and 0:38:42.120 --> 0:38:45.360 LG are taking snapshots of what you're watching multiple times 0:38:45.360 --> 0:38:49.020 per second. Plus energy started a $1 billion solar and 0:38:49.020 --> 0:38:54.420 battery storage project. The white collar apocalypse is nigh. The 0:38:54.420 --> 0:38:57.900 decline of status, economy and the impact of AI on 0:38:57.900 --> 0:39:01.290 job markets. Yeah, pretty cool one. Definitely go check this 0:39:01.290 --> 0:39:05.070 one out. Meta. Orion's AR glasses are a big upgrade. 0:39:05.070 --> 0:39:09.600 I don't think Mark is right about replacing phones by 2030. 0:39:09.630 --> 0:39:13.470 I think that's too ambitious. I think phones offer too 0:39:13.469 --> 0:39:16.560 many things that glasses don't, and I just don't think 0:39:16.560 --> 0:39:20.910 the physical miniaturization tech is there to go into the 0:39:20.910 --> 0:39:24.060 glasses to replicate what's on the phone Hacker News thread, 0:39:24.090 --> 0:39:26.820 where people are sharing their experiences about whether having a 0:39:26.820 --> 0:39:30.509 personal website helped get them hired. Some people said yes, 0:39:30.540 --> 0:39:32.460 a lot of people said yes. Some people said no. 0:39:32.460 --> 0:39:36.300 All right, humans. California is taking phones out of schools, 0:39:36.300 --> 0:39:40.440 basically limiting cell phone use. So massive props to Jonathan 0:39:40.440 --> 0:39:43.950 Hite on this. Uh, someone found a method that uses 0:39:43.950 --> 0:39:48.629 stem cells to regenerate insulin producing cells. ExxonMobil is facing 0:39:48.630 --> 0:39:53.100 a lawsuit for coming up with a supposed fraud saying 0:39:53.130 --> 0:39:57.720 or a myth saying that recycling could help solve the 0:39:57.719 --> 0:40:00.899 problem with plastic waste. California is looking to make it 0:40:00.930 --> 0:40:04.830 one click. Get off of subscriptions just the same way. 0:40:04.830 --> 0:40:07.920 One click sign ups. So if you offer a one 0:40:07.950 --> 0:40:09.930 click sign up, you have to offer a one way 0:40:09.930 --> 0:40:12.630 to get out of it as well. Over 10,000 books 0:40:12.630 --> 0:40:17.100 were banned in US public schools from 2023 to 2024. 0:40:17.130 --> 0:40:20.880 Insane Steve Jobs had a thing called a ten minute rule, 0:40:20.880 --> 0:40:23.400 just ten minutes to tackle a problem or a task. 0:40:23.430 --> 0:40:25.680 Check this out. That's why I love doing video. I've 0:40:25.680 --> 0:40:28.589 got one of these. I actually use it. It's called pomodoro, 0:40:28.590 --> 0:40:32.160 but you can set it for any time. Right. So 0:40:32.190 --> 0:40:35.640 30 minutes is like the classic one. It's inverted. But 0:40:35.640 --> 0:40:37.710 but you can put it down to like. There you go. 0:40:37.739 --> 0:40:43.379 Ten minutes. So I think it's cool. Constraints. Constraints inspire creativity. 0:40:43.530 --> 0:40:48.509 And Maria Popova explores profound curiosity and love of Henry 0:40:48.540 --> 0:40:54.299 David Thoreau as he encounters a screech owl in the wild. Ideas. 0:40:54.300 --> 0:40:57.930 Everything is going behind the paywall. Yeah, I'm not going 0:40:57.960 --> 0:41:01.410 to read this. I'm just going to give you the overview. Freeform. 0:41:01.440 --> 0:41:05.400 I think it's really interesting to think about how there's 0:41:05.430 --> 0:41:08.790 the open internet is going away. Everything's going to be 0:41:08.790 --> 0:41:12.210 behind auth not only to block bots but also to 0:41:12.239 --> 0:41:16.439 get micropayments. So rich people who have money for their 0:41:16.469 --> 0:41:18.780 their eyes. Okay I'm going to give you a little 0:41:18.810 --> 0:41:20.549 bit of a I also have to make this piece 0:41:20.550 --> 0:41:24.089 of content, but a tiny little view of the future. Okay, 0:41:24.120 --> 0:41:26.640 you're going to have all your personal context in your 0:41:26.640 --> 0:41:29.760 personal AI. It's with you all the time. Your personal 0:41:29.760 --> 0:41:32.610 AI is the one going out and reading everything for you. 0:41:32.610 --> 0:41:36.029 It's consuming everything. It's consuming all the APIs, all the 0:41:36.030 --> 0:41:40.830 RSS feeds, all the videos. It's making transcripts. It's translating them, 0:41:40.830 --> 0:41:43.680 it's summarizing them. It's doing everything for you. And as 0:41:43.680 --> 0:41:46.319 you're moving through your day, when you have a little 0:41:46.350 --> 0:41:49.560 bit of extra time, it'll create a custom piece of 0:41:49.560 --> 0:41:52.740 content for you. Actually, I'm merging both of these. Cool. 0:41:52.739 --> 0:41:55.650 It'll create a custom piece of content for you. Okay, 0:41:55.680 --> 0:41:59.310 so there's this paper that I just summarized. Where is 0:41:59.310 --> 0:42:01.770 that thing? Oh, it's right here. So my buddy just 0:42:01.770 --> 0:42:05.310 sent me this thing. It's this giant PDF. Um, my 0:42:05.340 --> 0:42:07.230 AI is going to go read this thing. It's going 0:42:07.260 --> 0:42:09.270 to give me a summary based on how much time 0:42:09.270 --> 0:42:11.489 I have before my next meeting. It's like, hey, you 0:42:11.489 --> 0:42:13.290 want to hear about this thing? Or do you want 0:42:13.320 --> 0:42:15.450 to see a video? I'm like, yeah, give me a video. 0:42:15.450 --> 0:42:19.830 So it's Richard Feynman explaining this paper to me in 30s, 0:42:19.830 --> 0:42:23.340 which means the content. Watch this. The content duration, the 0:42:23.340 --> 0:42:27.870 content format, the content style, all of. Oh and the performer. 0:42:27.900 --> 0:42:31.080 The person actually giving the content. So it could be 0:42:31.080 --> 0:42:34.169 a graphic novel. It could be a video, it could 0:42:34.170 --> 0:42:37.590 be an essay, it could be a video, it could 0:42:37.590 --> 0:42:40.470 be text, it could be whatever, but it could shrink 0:42:40.469 --> 0:42:45.810 from 30s down to 10s do a giant set of books, 0:42:45.810 --> 0:42:50.310 and the AI will dynamically recreate that for you, which 0:42:50.310 --> 0:42:54.569 means this is actually the future of education. And this 0:42:54.570 --> 0:42:57.750 is going to be a whole separate conversation. In fact, 0:42:57.750 --> 0:42:59.790 we're going to break this out as a separate clip 0:42:59.790 --> 0:43:02.100 to talk about this. This is going to be the 0:43:02.100 --> 0:43:05.100 future of education, because think about this. The AI can 0:43:05.100 --> 0:43:10.379 also adjust based on somebody's limitations. Okay. They haven't learned 0:43:10.380 --> 0:43:12.810 this topic yet. They're not so great at English yet. 0:43:12.810 --> 0:43:15.690 They have a learning disability okay. The AI is going 0:43:15.719 --> 0:43:18.180 to be an expert in learning disabilities, which means it 0:43:18.180 --> 0:43:21.060 can repeat things, or it could avoid certain ways of 0:43:21.060 --> 0:43:23.969 explaining things, or it could double down on other ways 0:43:23.969 --> 0:43:27.930 of explaining things. Dynamic adjustment to the best way to 0:43:27.960 --> 0:43:31.290 teach you at that given time. Maybe if you're a kid, 0:43:31.290 --> 0:43:33.840 you want to hear this lesson as a story. You 0:43:33.840 --> 0:43:36.000 want to watch it as a video. The whole thing 0:43:36.030 --> 0:43:40.050 could be framed as Taylor Swift or Pokémon or Star Wars. Okay, 0:43:40.080 --> 0:43:43.350 so your eye is going to be able to shape 0:43:43.380 --> 0:43:48.270 teaching you anything using any format, in any duration, in 0:43:48.270 --> 0:43:52.440 any style of teaching. Okay. It could be a podcast 0:43:52.440 --> 0:43:55.020 between your two favorite people, and they're talking about the 0:43:55.020 --> 0:43:58.319 ideas debating back and forth. It could be a graphic novel, 0:43:58.320 --> 0:44:00.719 it could be a TV show, it could be a cartoon, 0:44:00.719 --> 0:44:02.850 it could be an anime. It could be a YouTube 0:44:02.850 --> 0:44:07.140 video of your favorite creator just doing a YouTube video 0:44:07.140 --> 0:44:09.270 for you about that thing, and you're like, nope, I 0:44:09.270 --> 0:44:12.270 want it in 30s nope. I want it in one 0:44:12.270 --> 0:44:15.750 hour episodes, five of them per week. So it's like 0:44:15.750 --> 0:44:17.819 I sit down for an hour and I learn this thing. 0:44:17.820 --> 0:44:22.799 This is going to massively disrupt education. Okay, we already 0:44:22.800 --> 0:44:25.739 knew I was going to disrupt education. We already know 0:44:25.770 --> 0:44:28.890 technology is going to disrupt education. Like this is all known. 0:44:28.890 --> 0:44:32.370 But this particular formula of being able to adjust all 0:44:32.370 --> 0:44:37.620 those variables and create dynamic content, that is the game 0:44:37.620 --> 0:44:42.990 that that is the education game, like 100%. Okay. Now 0:44:42.989 --> 0:44:45.420 now think about this. All you need to do is 0:44:45.420 --> 0:44:48.180 come up with the absolute best content, okay? Once you 0:44:48.180 --> 0:44:50.850 have the best content, which which I hate to tell you, 0:44:50.850 --> 0:44:53.190 it's kind of already out there. Okay. What what are 0:44:53.190 --> 0:44:58.440 the best concepts in journalism or biology or whatever? It's 0:44:58.440 --> 0:45:01.739 in there. It's in the textbooks already. Right? The smartest 0:45:01.739 --> 0:45:03.870 people have already been talking about it. It's in all 0:45:03.900 --> 0:45:07.620 of their books. Well, go read their books. We've already 0:45:07.620 --> 0:45:10.229 seen their lectures. Half of those lectures are already on 0:45:10.230 --> 0:45:15.839 open MOOCs on, uh, you know, Harvard or MIT or whatever. Like, 0:45:15.840 --> 0:45:18.770 these people aren't being quiet. It's not a secret what 0:45:18.770 --> 0:45:21.049 they know about the world. So all the stuff's out there, 0:45:21.050 --> 0:45:24.050 so the content's solved. The only question is, how are 0:45:24.050 --> 0:45:27.560 we getting that content to the people? Now, you take 0:45:27.560 --> 0:45:31.370 something like an extremely expensive university like Harvard or Stanford 0:45:31.370 --> 0:45:35.270 or something like 11 people go per year and it's 0:45:35.270 --> 0:45:38.509 horribly expensive. How many people are on the planet? Oh, 0:45:38.540 --> 0:45:42.590 8 billion, 8 billion people need that education and don't 0:45:42.590 --> 0:45:46.820 have the time. And they have language, you know, disabilities 0:45:46.820 --> 0:45:51.020 or language like, um, mismatches with the way the original 0:45:51.020 --> 0:45:55.400 content was taught. Doesn't matter. Instant translation to their language 0:45:55.400 --> 0:45:58.040 while on ramping them onto English if they want to 0:45:58.040 --> 0:46:02.120 learn English. 8 billion people getting exactly the education they 0:46:02.120 --> 0:46:05.719 need in the topics that they need in any format 0:46:05.719 --> 0:46:09.259 that is best for them, represented by deepfakes of exactly 0:46:09.260 --> 0:46:12.950 the best people or in the format of whatever a show, 0:46:12.980 --> 0:46:14.750 whatever is the best way to get it to them. 0:46:14.750 --> 0:46:18.189 Like how is that not the way people are going 0:46:18.190 --> 0:46:20.410 to learn in the future? That's what I'm asking you. 0:46:20.440 --> 0:46:24.040 How is that not the future? Now, again, how fast 0:46:24.070 --> 0:46:25.960 are we going to get there? Are there lots of 0:46:25.960 --> 0:46:30.430 different ways that will be opposed? Of course, of course. Right. 0:46:30.460 --> 0:46:34.030 So when is it going to start? Today. Yesterday. Tomorrow. 0:46:34.060 --> 0:46:36.730 That's when it's going to start. It's already starting. When 0:46:36.730 --> 0:46:39.430 will it be finished? Never. There will always be a 0:46:39.430 --> 0:46:41.620 forces that push it back and say no, you've got 0:46:41.620 --> 0:46:44.259 to dress up. You've got to go to this special school. 0:46:44.290 --> 0:46:47.169 Because remember, a lot of education is actually not about 0:46:47.170 --> 0:46:49.870 the content and about the the learning. A lot of 0:46:49.870 --> 0:46:52.750 the education is actually about the stamp that you get afterwards, 0:46:52.750 --> 0:46:55.330 which we can solve a different way, which we'll be 0:46:55.330 --> 0:46:58.270 talking about at some point to and kind of already 0:46:58.270 --> 0:47:00.489 have talked about. All right. Let's keep going. I think 0:47:00.489 --> 0:47:03.730 we covered both of those topics pretty well. Oh, yeah. 0:47:04.120 --> 0:47:09.160 Future of education. Exactly. Okay. Discovery acts framework extension of 0:47:09.160 --> 0:47:13.750 Axios for bug hunters and penetration testers. So it's basically 0:47:13.750 --> 0:47:17.350 a way to do large distributed scans as part of 0:47:17.380 --> 0:47:20.739 like Bug Bounty, the Russian APT tool matrix. New tool 0:47:20.739 --> 0:47:26.680 matrix focused on Russian APT groups GRU, SVR, FSB affiliated 0:47:26.680 --> 0:47:31.719 threat actors. Reuse of tools like Mimikatz, impact, exact, Metasploit 0:47:31.719 --> 0:47:36.340 Re George Re George being particularly notable for limited use 0:47:36.340 --> 0:47:40.960 by by ransomware gangs okay Merkle map subdomain search engine 0:47:40.960 --> 0:47:45.070 helps you discover subdomains. Gun gear is a command line tool. 0:47:45.070 --> 0:47:48.070 What's with this font? Do Not like is a command 0:47:48.070 --> 0:47:51.430 line tool written by my buddy gunner and go. Microservices 0:47:51.430 --> 0:47:55.180 are technical debt. Really really cool. Take there. Be somebody 0:47:55.180 --> 0:47:59.770 who does things. That was a fantastic essay. It's lists 0:47:59.800 --> 0:48:03.310 all the way down. Another fantastic one. US government is 0:48:03.310 --> 0:48:07.569 now publishing tide sensor data to show sea level rise. 0:48:07.600 --> 0:48:11.740 Love the transparency here. Comedy is search, critical mass and 0:48:11.739 --> 0:48:16.870 tipping points. Semantic chunking for rag flipper zero. Firmware. Open 0:48:16.900 --> 0:48:20.170 source firmware unlocks a lot more of your flipper device. 0:48:20.200 --> 0:48:24.250 MTA Open Data Challenge Paul Graham addresses the question of 0:48:24.250 --> 0:48:28.540 whether to follow your passion or not. Compressed jpg, compressed 0:48:28.540 --> 0:48:33.880 jpeg move fast and abandon things. BBC Sound Effects Library 0:48:33.910 --> 0:48:36.940 what I learned in the past years spent building an 0:48:36.940 --> 0:48:40.900 AI video editor. Danger zone is teamed up with Google's. 0:48:40.930 --> 0:48:44.950 Gvisor to enhance its security, allowing journalists to open suspicious 0:48:44.950 --> 0:48:48.640 documents safely. RM Creative Console might be getting one of 0:48:48.640 --> 0:48:54.160 these actually need to go order that advanced structure output tutorial. 0:48:54.160 --> 0:48:58.540 Logitech's MX Creative Console is a game changer for photo 0:48:58.540 --> 0:49:02.620 and video editors. Dieter Rams inspired iPhone dock. This thing 0:49:02.650 --> 0:49:04.359 looks really cool. I don't think I would get one, 0:49:04.360 --> 0:49:07.990 but isn't that cool? Yeah, that looks cool. And automatic 0:49:07.989 --> 0:49:10.660 litter boxes because I want a cat, but I'm allergic 0:49:10.660 --> 0:49:13.780 to cats, so probably not getting one recommendation of the week. 0:49:13.810 --> 0:49:16.629 Take a document or something you know really well and 0:49:16.630 --> 0:49:19.960 upload it to notebook LM and tinker around with it. 0:49:19.989 --> 0:49:22.840 It is super cool and do it quickly because they 0:49:22.840 --> 0:49:27.130 might put the project in the Google graveyard. Um. Aphorism 0:49:27.130 --> 0:49:30.160 of the week. I want so much that is not 0:49:30.160 --> 0:49:33.009 here and do not know where to go. I want 0:49:33.010 --> 0:49:35.710 so much that is not here and do not know 0:49:35.710 --> 0:49:41.230 where to go. Charles Bukowski. Unsupervised learning is produced and 0:49:41.230 --> 0:49:44.800 edited by Daniel Miessler on a Neumann U87 AI microphone 0:49:44.800 --> 0:49:48.760 using Hindenburg. Intro and outro music is by zombie with 0:49:48.760 --> 0:49:51.580 a Y. And to get the text and links from 0:49:51.580 --> 0:49:53.710 this episode, sign up for the newsletter version of the 0:49:53.710 --> 0:49:59.290 show at Daniel missler.com/newsletter. We'll see you next time.