WEBVTT 1 00:00:17.620 --> 00:00:18.420 Hey Ben. 2 00:00:18.420 --> 00:00:19.340 Hey Matt. 3 00:00:19.340 --> 00:00:24.870 It has been almost a month to the day that you and I have sat in front of a microphone and tried to talk. 4 00:00:24.870 --> 00:00:24.940 Mm-hmm. 5 00:00:24.940 --> 00:00:36.110 And at least I'm sat in front of a microphone now and not actually recording on my earbuds in my wife's bedroom, childhood bedroom. So back in the right continent. 6 00:00:36.110 --> 00:00:36.720 Right. Right. 7 00:00:36.720 --> 00:01:01.160 Although I do have a funny story on that. I set myself up for a work recording the other day and, you know, for an internal like chat kind of video casting, and I had my pompous microphone in shot obviously because, you know, I'm one of those kind of podcasty people. And of course, I forgot to actually select it in the dropdown. So it was using my terrible earbuds as the recording. And so that's going to go out internally and I'm going to have the absolute... 8 00:01:01.160 --> 00:01:07.380 Mickey taken out of me by being this, you know, hey, I thought we got somebody who was like media savvy, but no. 9 00:01:07.380 --> 00:01:27.260 No I'm not. But that's not what we're going to talk about today. We don't even really know what we're going to talk about today, but something happened to me this morning. It was the first time. That was fun. I was intrigued by a super cool new exploit that has been discovered. It exploits [pretty strong]. It's like... 10 00:01:27.260 --> 00:01:28.100 No. 11 00:01:28.100 --> 00:01:49.320 Do you know, I can't remember the dude's name, unfortunately. XOR, E-A-X, E-A-X, E-A-X is their handle, his GitHub handle. He did SandSifter, which was like a comprehensive search through the entire space of all possible x86 opcodes looking for undocumented opcodes, which was fun. He did M/o/Vfuscator, 12 00:01:49.320 --> 00:02:18.200 which was, hey, every instruction... the MOV instruction is Turing complete on its own with certain caveats, and he wrote a compiler that compiles to just MOV instructions, which is fun. And then most recently he's discovered this cool facility there. If you have ring zero access, like, you know, if you're like the lowest level access, you might just think, hey, yeah, I've got root, what more do I need to know? There are layers below that, that like the firmware of the system runs at. 13 00:02:18.200 --> 00:02:47.160 and other things that you're not supposed to be able to see about your own computer that you own under your desk. And he's found a way to get into it by misconfiguring the RAM system in some particular way, in a really cunning way. Anyway. I was like, this sounds cool. I'm going to... While I'm in the shower, I gave a quick query to Claude saying, can you just summarize this for me? I want to learn about it a little bit. Give me some pointers. And then I had my shower and I came back out and I looked. And for the very first time, I had a massive alert saying, no, you can't do this. 14 00:02:47.160 --> 00:02:50.760 You are clearly some evil hacker. 15 00:02:50.760 --> 00:02:54.810 We will not allow you to use our AI to hack things. You're a bad person. 16 00:02:54.810 --> 00:02:55.140 I hope. 17 00:02:55.140 --> 00:02:59.460 I don't think that's exactly what it said. I'm sorry if anyone from Anthropic is listening, but it was something like that. 18 00:02:59.460 --> 00:02:59.460 Mm-hmm. 19 00:02:59.460 --> 00:02:59.660 Yeah, you're paraphrasing a little bit. 20 00:02:59.660 --> 00:03:14.120 And it gave me the option to use, you know, like a watered down model, which I did. And it was perfectly good for summarizing a document, which I then went off and read anyway. So, you know. It was fine, but it made me think. Are we entering a new age where we're being gatekept by things that... 21 00:03:14.120 --> 00:03:17.500 we don't have any control about. 22 00:03:17.500 --> 00:03:24.240 Mm-hmm. Yeah. You know, this is, it's sort of like a new problem and also not a new problem. 23 00:03:24.240 --> 00:03:25.520 Right. 24 00:03:25.520 --> 00:03:30.680 You know, I had a similar experience this week where a friend of mine was... 25 00:03:30.680 --> 00:03:30.680 Oh. 26 00:03:30.680 --> 00:03:35.520 locked out of their AWS account for reasons that are kind of opaque to them. 27 00:03:35.520 --> 00:03:35.560 That... 28 00:03:35.560 --> 00:03:35.560 And... 29 00:03:35.560 --> 00:03:43.360 is a nightmare situation for me because yeah, that goes down and then you're like, well, how do I... who do I call? 30 00:03:43.360 --> 00:03:43.580 Right. 31 00:03:43.580 --> 00:03:53.360 Right, right. Right. And I mean, we have this sort of digital dependence problem in many facets. Like, I'm going to give you a hypothetical scenario here. 32 00:03:53.360 --> 00:03:53.650 Mm-hmm. 33 00:03:53.650 --> 00:04:03.460 Your choices are... You can continue to own and use a mobile device, a mobile phone. Or not. 34 00:04:03.460 --> 00:04:11.080 But if you continue to use a mobile device, then you need to cut off one of your fingers. It can be any finger. 35 00:04:11.080 --> 00:04:17.770 Whichever one you want, you can use whatever method you want to cut it off, but you need to cut off a finger. 36 00:04:17.770 --> 00:04:17.970 What? 37 00:04:17.970 --> 00:04:19.520 What? This was not on my bingo card. 38 00:04:19.520 --> 00:04:19.710 Yeah. 39 00:04:19.710 --> 00:04:20.100 Right! Okay. 40 00:04:20.100 --> 00:04:25.120 What do you choose? You lose the left pinky. Like that's the choice that 99% of people are making. 41 00:04:25.120 --> 00:04:25.920 There's some... 42 00:04:25.920 --> 00:04:27.180 It's like, yeah, I kind of need a phone. 43 00:04:27.180 --> 00:04:34.900 Yeah, I... I don't like the way this analogy is going, but I understand where... I think I understand your point. Yes, I get it. Yeah. Okay. 44 00:04:34.900 --> 00:04:40.800 Yeah, it's this dependence. And it's like, it does... it's not even good enough to have the device. You need... 45 00:04:40.800 --> 00:04:40.800 Right. 46 00:04:40.800 --> 00:04:44.920 Like the network. Right. You need the connectivity. 47 00:04:44.920 --> 00:04:49.040 Yeah. You know, teleport me back to 1970 with the top of the line phone and it's useless. 48 00:04:49.040 --> 00:04:49.240 Right. 49 00:04:49.240 --> 00:04:49.900 It's an interesting... 50 00:04:49.900 --> 00:04:52.620 Right. 51 00:04:52.620 --> 00:04:57.160 yeah, device, but it doesn't really... yeah. 52 00:04:57.160 --> 00:05:06.200 So we have this digital dependency, right? And it comes in many forms. Like AI models are the most recent form of it, you know, cloud services. 53 00:05:06.200 --> 00:05:07.020 Right. 54 00:05:07.020 --> 00:05:13.980 All of these, like, you know, your Google account, right, with your email and everything else. And you've got probably a lot of identity tied up in that. 55 00:05:13.980 --> 00:05:14.390 Yeah. 56 00:05:14.390 --> 00:05:15.240 Oh my gosh. 57 00:05:15.240 --> 00:05:32.110 Shit, I have a story about that, right? Like, my vanity domain godbolt.org is a Google Apps for Your Domain, when that was a thing. That was like a thing that if you were a Google fanboy like I was, and I was an early adopter of everything, you could say, hey, I would like to bring my own domain to a Gmail account effectively. 58 00:05:32.110 --> 00:05:32.640 Mm-hmm. 59 00:05:32.640 --> 00:05:42.720 Subsequently, when I worked at Google, I discovered that that is not how it works. There's a massive internal hack to make that work. And then it sort of became Google Apps Enterprise-y type thing. 60 00:05:42.720 --> 00:05:55.620 But unfortunately, those who early adopted, like I did, are stuck in this... it's this hinterland between it's not really an enterprise account and it's also not a Gmail account. So I can't like do half the things in my house that are internet connected with my godbolt.org account. 61 00:05:55.620 --> 00:06:06.080 And so I want to just move. And I can't just move because so many things are tied into my login identity as matt@godbolt.org that goes through like the Google login process, that now I can't... I'm stuck. 62 00:06:06.080 --> 00:06:09.520 So I'm like... what do I do? But yeah, sorry, I interrupted. 63 00:06:09.520 --> 00:06:09.800 Yeah. 64 00:06:09.800 --> 00:06:16.140 You were in the middle of the flow of talking about other digital dependencies and... 65 00:06:16.140 --> 00:06:32.080 Yeah. No, well, I mean, that's another great example, right? And so like, I don't think it's realistic to try to eliminate all of these things. But to the extent that you can be... 66 00:06:32.080 --> 00:06:47.940 more resilient to some other entity just unilaterally deciding maybe, in a way where they think you're acting in your best interest, like my friend in his AWS account, maybe. 67 00:06:47.940 --> 00:06:48.420 Right. 68 00:06:48.420 --> 00:06:50.890 Maybe it was. Or not. 69 00:06:50.890 --> 00:06:51.040 Yeah. 70 00:06:51.040 --> 00:06:57.740 It's sort of like, you don't know, and you don't really have a way to know, and you don't have a way to like validate what it is that they're doing and be like, actually that... 71 00:06:57.740 --> 00:06:58.280 I know. 72 00:06:58.280 --> 00:07:08.690 It's not how that works. Like, to the extent that you can limit that liability, it sometimes makes a lot of sense. And so like, you know. 73 00:07:08.690 --> 00:07:08.980 Right. 74 00:07:08.980 --> 00:07:23.990 I feel like open source was like a great example of this, where you know I certainly remember early on in my career the pain of having to deal with, you know, third-party software vendors, and you would try to understand how their software worked. 75 00:07:23.990 --> 00:07:35.680 And they would give you answers that were clearly wrong. And then, you know, you maybe tried to dig into it a little bit. And then, you know, I had the experience very early on in my career of like, you know. 76 00:07:35.680 --> 00:07:50.980 I don't even think that it was any more complicated than like reading through like the scripts they had sent us. But they were like, don't modify those. Right. And I'm like, okay, I'm not modifying them. 77 00:07:50.980 --> 00:07:58.340 I'm just reading them. Like I'm trying to understand like this thing that you sent me that we paid a bunch of money for that doesn't seem to work. And like, God forbid, if I had actually tried to like reverse engineer any of their software, right? 78 00:07:58.340 --> 00:08:02.830 I can't imagine what their reaction would have been to that. Like maybe some kind of lawsuit. 79 00:08:02.830 --> 00:08:03.180 Yeah. 80 00:08:03.180 --> 00:08:13.980 But... like you're stuck in this place where you're like, you know. I don't understand how this works. I'm not empowered to change it. And yet I am dependent on it. 81 00:08:13.980 --> 00:08:14.980 Right? 82 00:08:14.980 --> 00:08:43.320 Empowered is the word, the active word. That's really interesting. Yeah. It occurred to me that like another thing, it was like early Google search engine optimization once Google became the monopoly. Like, you know, you had this thing where an algorithm would change and people's would drop off of the listings, and again it's like, but they couldn't choose that. That's not something they could choose to be in or out. They were like dependent on something, but not because they chose it. Which is even... but yeah, the empowerment is an interesting 83 00:08:43.320 --> 00:08:57.600 aspect of... I do wonder, yeah, so what can we do about this, Ben? Well, I mean, I'm kind of stuck, right? I mean, what... I'll have my own internet search, I'll just, you know, my own data center, my own... yeah, what. Yeah. 84 00:08:57.600 --> 00:09:04.880 Right. Mesh network internet for everybody. That's the solution. Is we're all just going to get some powerful antennas and put them on the top of our houses and mesh all of the houses together. 85 00:09:04.880 --> 00:09:06.760 Don't tell me. Yeah, mesh. 86 00:09:06.760 --> 00:09:07.420 And that's how the internet's going to work. 87 00:09:07.420 --> 00:09:12.720 It's going to end up in some crypto scam though, isn't it? At some point, like some... 88 00:09:12.720 --> 00:09:13.080 Yeah. At some point, yeah, there's like, you know. 89 00:09:13.080 --> 00:09:17.220 Anything distributed nowadays ends up in some... Well, there's a... we can make a chain, some kind of... 90 00:09:17.220 --> 00:09:19.190 Right. There's a coin. You got to buy the coin. 91 00:09:19.190 --> 00:09:19.740 Hey there. 92 00:09:19.740 --> 00:09:20.850 And we're issuing the coins. 93 00:09:20.850 --> 00:09:21.640 You got AI coin. 94 00:09:21.640 --> 00:09:23.210 And who gets the money from these coins? 95 00:09:23.210 --> 00:09:23.410 Ugh. 96 00:09:23.410 --> 00:09:23.800 Yes, exactly. 97 00:09:23.800 --> 00:09:24.390 Oh, token coin. 98 00:09:24.390 --> 00:09:24.590 Yeah. 99 00:09:24.590 --> 00:09:26.360 I'm sure it's been done. I'm sure... 100 00:09:26.360 --> 00:09:26.540 a billion times at this point. 101 00:09:26.540 --> 00:09:28.340 It has already been done. 102 00:09:28.340 --> 00:09:33.060 The terrible idea that launched a thousand slide decks. 103 00:09:33.060 --> 00:09:35.400 Yeah. 104 00:09:35.400 --> 00:09:49.880 So, you know, like... I think it sort of depends on how far down the sort of, you know, crazy tinfoil hat, you know, internet libertarian, you know, route that you want to go. 105 00:09:49.880 --> 00:10:10.490 And I say that recognizing that like a lot of the things that people, you know, from like the Electronic Frontier Foundation and, you know, some elements of the open source community, you know, 20, 30 years ago... a lot of things that they were saying, you know, with panic in their voices would come to pass, have come to pass, and nobody seems to care. 106 00:10:10.490 --> 00:10:11.180 Right. 107 00:10:11.180 --> 00:10:17.800 Like the amount of spying that like all the devices that you pay for do on you is like off the charts. 108 00:10:17.800 --> 00:10:36.700 Somehow that's okay when... I mean, the thing is, there is... so I've always thought of it as there is some kind of implied social contract between me and... I'm going to pick on Google again just because I'm a bit of a fanboy for it, and I feel like having been there I've got some standing in there, some... 109 00:10:36.700 --> 00:10:55.180 And I've always felt that like, yes, it serves ads, yes, it reads my emails or does something with them, and I don't mind that because I get a decent email service that I don't have to pay for and it's integrated with everything. And like the slightly spooky face recognition is great when, for example, you are preparing a family 110 00:10:55.180 --> 00:10:55.180 Mm-hmm. 111 00:10:55.180 --> 00:11:12.100 board. And I'm like, hey, let's just find pictures of my father-in-law. Here's hundreds of pictures of Tony. And like, oh, that's great, right? That's kind of an amazing thing. But yeah, I do realize that I've given up something for that, but I've probably given up more than I think. 112 00:11:12.100 --> 00:11:22.200 than I think about. And I've probably given up more than I would if someone sat me down and said, well, what about these things? And I was like, oh, yeah, I hadn't really thought. Oh, that's a good point. And, you know, but I'm getting something back 113 00:11:22.200 --> 00:11:24.700 out of this. 114 00:11:24.700 --> 00:11:26.220 Mm-hmm. 115 00:11:26.220 --> 00:11:43.980 And like some of the things, though, I'm not getting something back out of it. So like, you know, automatic number plate recognition stuff, the Flock camera stuff that's going on around, you know, like there's a lot of things where it's like, yeah, that's... I'm uneasy because it is not... ironically, it's not directly a company making money. I mean, it is obviously, but it's like... 116 00:11:43.980 --> 00:12:02.380 Law enforcement, a government thing that is for the benefit of the people has decided that they would like this information and then they would be happy to pay someone to put cameras up to do all this kind of stuff. And that feels like, no, that's a step too far somehow. Now, I don't know where the boundary is there. I mean, it feels obviously... but one I'm sort of choosing. 117 00:12:02.380 --> 00:12:15.120 or at least choosing... not choosing, not choosing to. But then... I get so much benefit, and how will I get the same benefit without, you know, putting together a whole bunch of open source software, which is maybe where you're going with this. 118 00:12:15.120 --> 00:12:19.160 As you mentioned, the mesh network of like internet stuff. 119 00:12:19.160 --> 00:12:29.340 Well... Yeah, I mean, this is kind of what I'm saying, is it sort of depends on how far down the rabbit hole you want to go. Right. And I think. 120 00:12:29.340 --> 00:12:40.740 For me personally, first recognizing that 90... nine... I'm just going to put two nines in that, percent of people are not going to want to go down the rabbit hole at all. 121 00:12:40.740 --> 00:13:02.760 Right. Like they're going to want to do the minimum amount that they need to do to keep themselves safe and secure from like scammers and phishers and, you know, people like that. And you know, some of them don't even really care that much about it, from some of the things I've seen. But you know, that's like the minimum amount that they're going to want to put into this, and the rest of it, they're like perfectly happy to be completely dependent. 122 00:13:02.760 --> 00:13:10.380 and exposed in like a data sharing type sense. Like, I'm very happy to trade my data for free email. Right? Like, it's fine. 123 00:13:10.380 --> 00:13:11.500 Yeah. That's, I guess, both. Yeah, yeah. 124 00:13:11.500 --> 00:13:22.800 Right. Like they're very happy to trade their data for just about anything. And in fact, many of them... What was the... Was it Snow Crash that had the gargoyles in it? 125 00:13:22.800 --> 00:13:23.660 Do you know what I'm talking about? 126 00:13:23.660 --> 00:13:24.300 I... 127 00:13:24.300 --> 00:13:25.220 Did you read Snow Crash? 128 00:13:25.220 --> 00:13:27.300 I did read Snow Crash, and now I'm just trying to think about... 129 00:13:27.300 --> 00:13:29.380 I think it was, I think it was Snow Crash. 130 00:13:29.380 --> 00:13:29.380 I don't remember gargoyles. Was it? 131 00:13:29.380 --> 00:13:41.020 that had basically like there was a whole profession where you just walked around with a bunch of like data gathering equipment strapped to your body. 132 00:13:41.020 --> 00:13:41.240 Ah. 133 00:13:41.240 --> 00:13:46.690 right like cameras and like infrared and like, you know, microphones and all these things 134 00:13:46.690 --> 00:13:47.240 Right. Does ring a bell. 135 00:13:47.240 --> 00:13:52.620 And like that was your job. You just walk... it was like you're like a little mini like Google Maps car, but you were just a person. 136 00:13:52.620 --> 00:13:54.000 Put a person doing that, yeah. 137 00:13:54.000 --> 00:14:04.300 Right. And that's what you did for, you know, for your livelihood. And they referred to these people as gargoyles, right? And I think like the great trick, it's like, you know. 138 00:14:04.300 --> 00:14:17.880 So much of what is being built is like 90s cyberpunk dystopia. Right? We're just literally using those books as like a user manual for like how to build a society. 139 00:14:17.880 --> 00:14:19.020 Yeah. Oh. Yeah. 140 00:14:19.020 --> 00:14:25.070 And I feel like one of the great tricks of social media is basically like turning everyone into gargoyles for free. 141 00:14:25.070 --> 00:14:25.320 Now I'm sad. 142 00:14:25.320 --> 00:14:32.600 Like you don't even have to pay them. Right. Like you just walk around and you say like, hey, you should post this to your Instagram, and then people take pictures of everything around them all the time. 143 00:14:32.600 --> 00:14:32.600 Yeah. 144 00:14:32.600 --> 00:14:41.160 and just post it with, you know, geotags on it to social media, and then it's like, all right, cool. We've done it. We've gathered all the data that is. 145 00:14:41.160 --> 00:14:42.140 That was... I remember like, 146 00:14:42.140 --> 00:14:43.480 We didn't even have to pay for it. 147 00:14:43.480 --> 00:14:59.410 the like the surveillance things that like found all of the photographs of like a monument somewhere in Paris or something like that, we're able to reconstruct a full high detailed 3D model from like just thousands of tourist photographs of it, for example, that they've posted to their Instagram. So yeah, we're in that world, aren't we? Yeah, gosh. 148 00:14:59.410 --> 00:14:59.900 There you go. 149 00:14:59.900 --> 00:15:13.180 Yeah. So my point is that 99% of the people, maybe 95, I don't know, some very high percentage of the people out there are not going to be interested in going down this rabbit hole at all. 150 00:15:13.180 --> 00:15:25.000 But... I think it is important. Personally, I think it is important to be able to defend the ability to do it if you want. Because A, it just might make you sleep better at night. 151 00:15:25.000 --> 00:15:35.140 Like, I know a lot of anxious people. I know a lot of paranoid people. And they're just like, that's who they are. And they just want to know that these things are going to be safe and secure. And then they reduce their dependency as much as possible. 152 00:15:35.140 --> 00:15:35.480 Right, fair. 153 00:15:35.480 --> 00:15:37.100 And they just want to be able to do that. 154 00:15:37.100 --> 00:15:48.680 And I think that you should. And also, there are lots of situations in which... you know... people who... maybe... 155 00:15:48.680 --> 00:16:01.180 in situations where they can be easily exploited by governments, by other people, need to be able to keep themselves safe and secure. And having the tools available to them to do that 156 00:16:01.180 --> 00:16:17.840 is important. Right? And you know, I don't know that I can go in and enumerate all of those situations, but I think we could certainly imagine lots of situations in which that might be not only something that someone would want to do, but be, you know, reasonably morally justified in wanting to do it. 157 00:16:17.840 --> 00:16:25.020 And so, for all of those reasons, I think it's important to be able to do this. But it is also really hard. 158 00:16:25.020 --> 00:16:31.960 Right. I mean, so like... Going back to the beginning of this, or not nearly the beginning, as you mentioned your friend's AWS account. I mean, so... 159 00:16:31.960 --> 00:16:31.960 Right. 160 00:16:31.960 --> 00:16:44.900 If they can't recover that account, do they go and buy servers and rack them? Or, you know, is there a happy end to this story, by the way, I should say? 161 00:16:44.900 --> 00:16:50.900 Right. I mean, immediately what you probably do is hope that you did everything with infrastructure as code and then try to port it all over to... 162 00:16:50.900 --> 00:16:54.400 you know, Google Cloud or... Azure, I guess. I don't know. 163 00:16:54.400 --> 00:16:55.590 Let's just switch it one... 164 00:16:55.590 --> 00:16:58.200 one problem for the same category. 165 00:16:58.200 --> 00:17:00.060 Yeah, you know. One feudal lord to another feudal lord, right? 166 00:17:00.060 --> 00:17:00.640 Feudal lord. 167 00:17:00.640 --> 00:17:02.280 Yeah. Like, okay. 168 00:17:02.280 --> 00:17:03.680 Jeff Bezos, feudal lord. 169 00:17:03.680 --> 00:17:03.740 Yeah. 170 00:17:03.740 --> 00:17:05.560 You heard it here first. 171 00:17:05.560 --> 00:17:10.940 So, you know, like that's like on day one, that's probably as good as it gets. 172 00:17:10.940 --> 00:17:11.400 Yeah. 173 00:17:11.400 --> 00:17:28.600 But I think this is why, like, you know, a lot of times, like especially in our industry, people will wind up building their own data centers, not because they strictly have to, but because, you know, when things go wrong or when they have problems or when they want to try to do things that are maybe not standard. 174 00:17:28.600 --> 00:17:37.000 They want the level of control to be able to do it, right? Like they want to be able to say like, yeah, we can put whatever hardware in these machines we want. We can, you know. 175 00:17:37.000 --> 00:17:37.200 Mm-hmm. 176 00:17:37.200 --> 00:17:48.430 configure them however we want. We can use whatever network technology that we want. We're not limited by what a third party vendor is willing to offer to us. 177 00:17:48.430 --> 00:17:48.840 Right. 178 00:17:48.840 --> 00:17:49.660 Yeah. Yeah. 179 00:17:49.660 --> 00:17:54.700 Obviously that is out of the reach of... most typical people. 180 00:17:54.700 --> 00:17:55.820 Your average. 181 00:17:55.820 --> 00:18:05.560 But I will say, I think one thing that is overlooked... you know, kind of like not that expensive 182 00:18:05.560 --> 00:18:15.220 way to get yourself a lot of digital independence is just get a cheap computer and put it on your network and leave it on all the time. 183 00:18:15.220 --> 00:18:16.960 Right. 184 00:18:16.960 --> 00:18:22.320 Great. And, you know, if you want to take that up a level, like get some remote access for it, right? 185 00:18:22.320 --> 00:18:24.850 Tailscale, something something, to make it a very... yeah, exactly, yeah. 186 00:18:24.850 --> 00:18:25.120 So you can SSH into it. 187 00:18:25.120 --> 00:18:36.100 Yeah, yeah. Most of those types of computers that you can buy use very low power, so you don't really need to worry too much about how much power they're drawing. 188 00:18:36.100 --> 00:18:41.870 They're quite... powerful in terms of like the compute that they can, you know, supply. 189 00:18:41.870 --> 00:18:42.440 Yep. 190 00:18:42.440 --> 00:18:52.310 And you will probably find that most of the things that you would want to run in some other service provider, you could run on those types of machines today. 191 00:18:52.310 --> 00:18:52.700 Yep. 192 00:18:52.700 --> 00:18:58.820 And in a weird sort of like everything is backwards and I don't understand how anything works anymore... 193 00:18:58.820 --> 00:19:06.420 The absolute trick that you can pull is, you know, use the LLM of your choice, which you are absolutely dependent on these days, for 194 00:19:06.420 --> 00:19:06.860 At least. 195 00:19:06.860 --> 00:19:14.280 to build you the software that you need to run on that computer to replace whatever it is that you're dependent on the third-party services to do. 196 00:19:14.280 --> 00:19:14.560 Put a pin in that. 197 00:19:14.560 --> 00:19:23.490 right, so if you have some like, you know, awesome note-taking app that depends on some cloud service and you're like, wow, if this ever disappeared I'd be totally screwed 198 00:19:23.490 --> 00:19:24.480 Right. Interesting. 199 00:19:24.480 --> 00:19:32.240 And you want to go down the rabbit hole, which again, 95% of people don't. Something that you could do is you could build your own. 200 00:19:32.240 --> 00:19:32.620 Yeah. 201 00:19:32.620 --> 00:19:43.060 and run it on this computer that runs in your house. And maybe, you know, have some backups that, you know, write to a USB drive or, you know, go back to the cloud and encrypt them and upload them. 202 00:19:43.060 --> 00:19:47.020 Yeah, I was going to say, that's one thing where like the cloud maybe isn't holding you to ransom unless... 203 00:19:47.020 --> 00:19:47.020 That's another way. 204 00:19:47.020 --> 00:19:54.420 Two things happen. One, you need your backup. And two, the cloud provider decides that they're not going to let you. 205 00:19:54.420 --> 00:19:55.260 Right. Right. 206 00:19:55.260 --> 00:20:00.900 So that seems like a dual uncorrelated failure or whatever we used to say. 207 00:20:00.900 --> 00:20:11.420 So, yeah, that is an interesting call to action. Now I wonder. And we have touched on this before because I'm getting a slight deja vu feeling here. But I wonder, it is... 208 00:20:11.420 --> 00:20:11.500 Right. 209 00:20:11.500 --> 00:20:26.900 The fact that we have been so lucky for so long, like 20, 30 years in the sort of software profession that we're in, where essentially we've always been able to just, inverted commas, quotes, just build it ourself. 210 00:20:26.900 --> 00:20:34.740 And with open source, that has been free. Like, is that spoiling us? Because like if I worked in the manufacturing industry and I made grommets 211 00:20:34.740 --> 00:21:12.380 on a machine, and then the vendor that made the machine said, no, you can't use this machine anymore, that's a real problem that many grommet manufacturers may have always had, and they might have to switch to another provider at short notice because they've gone out of business and now they can't get the parts to repair them. So there's a little bit of like a, maybe, you know, other people have always had to deal with this dependence. It's like, unless you want to become a core... your core competency is building lathes to make grommets, as well as the grommets themselves, then you've always had to outsource some significant part of your endeavor. Whereas you and I have been always like, oh, an email client, yeah, I can't be bothered to write an email client, but I... 212 00:21:12.380 --> 00:21:22.660 could in principle. I feel like I could. Is it... are we just spoiled? And then does an LLM now move us back to where we were to start with, you know? 213 00:21:22.660 --> 00:21:32.760 Yeah, I think a lot of the difference there is that once you've bought the grommets, they're yours. Right. The company who built the grommets never shows up and says, your license to these grommets has expired. 214 00:21:32.760 --> 00:21:34.060 And we're going to go into your... 215 00:21:34.060 --> 00:21:34.060 Or the lathe. 216 00:21:34.060 --> 00:21:37.580 Maybe in that instance for the grommet manufacturer. But yeah, either way, both of those things are true. 217 00:21:37.580 --> 00:21:37.780 Yeah. 218 00:21:37.780 --> 00:21:37.980 Right. 219 00:21:37.980 --> 00:21:43.460 Like it is possible to own the thing and it's yours. 220 00:21:43.460 --> 00:21:44.590 It's really interesting. I mean, obviously, again, 221 00:21:44.590 --> 00:21:44.800 right? 222 00:21:44.800 --> 00:21:50.180 And it's kind of like servers, right? Like you buy a server. You own the server. 223 00:21:50.180 --> 00:21:50.920 Yeah. 224 00:21:50.920 --> 00:21:56.840 Modulo maybe some weird things that happen when you ask LLMs about how those chips work. 225 00:21:56.840 --> 00:21:57.300 Ah. Ha ha! 226 00:21:57.300 --> 00:22:01.880 You know, you can do pretty much whatever you want with them. 227 00:22:01.880 --> 00:22:02.700 Yeah. 228 00:22:02.700 --> 00:22:13.100 And like, you know, you want to screw them up and break them to experiment, go ahead and do that, right? Like there will be no ramifications other than the fact that you lost however much money you spent on the server. 229 00:22:13.100 --> 00:22:13.500 Yeah. 230 00:22:13.500 --> 00:22:21.680 Right? So I think that is sort of like the fundamental difference. And some of it gets back to like software as a service type models where it's like, you know. 231 00:22:21.680 --> 00:22:25.880 You never quite own. And, you know, like, you know, video games these days, right? 232 00:22:25.880 --> 00:22:37.020 You know, obviously our hobbies are, my hobbies are old computers, and I've got like buckets and buckets of cartridges and CDs and DVDs, even, that I can put into various old machines and they work fine, but like. 233 00:22:37.020 --> 00:22:54.800 But nowadays, if you want to play a game for like PS4, PS5, it's like you don't really own it in any meaningful way. It exists as bits on the disc, maybe it doesn't even exist on a hard disk anymore, it just exists in the cloud, which means that it goes away. 234 00:22:54.800 --> 00:22:55.120 Yeah. 235 00:22:55.120 --> 00:22:59.020 when you don't own it anymore. You're renting. You've got a license to own something, which is like 236 00:22:59.020 --> 00:22:59.670 Right, right. 237 00:22:59.670 --> 00:23:00.240 we've gone... Yeah. 238 00:23:00.240 --> 00:23:00.840 Yeah. 239 00:23:00.840 --> 00:23:11.230 I mean, you buy a book and you read the small print in the front copy about, you know, you may not, you know, give it away or whatever, or there's some level of restrictions for what you can use it for. 240 00:23:11.230 --> 00:23:12.680 and reproduce it. You know? 241 00:23:12.680 --> 00:23:24.120 But like, that seems like... those seem very reasonable compared to end user license agreements in software or whatever else. Yeah, I don't know. How did we get here? How do we let this happen? 242 00:23:24.120 --> 00:23:27.680 Well, the 90s cyberpunk set the model, and oh, we're just going to do all these things. 243 00:23:27.680 --> 00:23:27.960 Yeah. 244 00:23:27.960 --> 00:23:28.680 I mean, I don't know. 245 00:23:28.680 --> 00:23:31.360 I'm being a little tongue-in-cheek there. 246 00:23:31.360 --> 00:23:31.740 Is this Neal Stephenson's fault? 247 00:23:31.740 --> 00:23:32.040 Is that what you're saying? 248 00:23:32.040 --> 00:23:33.720 Yeah, exactly. Right. 249 00:23:33.720 --> 00:23:35.280 Oh man. 250 00:23:35.280 --> 00:23:36.580 William Gibson can be blamed. 251 00:23:36.580 --> 00:23:39.400 William Gibson. Oh, yeah, yeah, yeah. 252 00:23:39.400 --> 00:23:40.660 Is that Neuromancer as well? 253 00:23:40.660 --> 00:23:44.300 Yeah, yeah, Neuromancer, Mona Lisa Overdrive, and... 254 00:23:44.300 --> 00:23:44.670 Thank you. 255 00:23:44.670 --> 00:23:46.120 Okay, what was the middle one? 256 00:23:46.120 --> 00:23:48.600 My nerd credentials are definitely going down here. 257 00:23:48.600 --> 00:23:50.720 I don't think I've ever read Neuromancer. 258 00:23:50.720 --> 00:23:53.560 like Counter Zero, Count Zero. 259 00:23:53.560 --> 00:23:55.210 Count. 260 00:23:55.210 --> 00:23:56.830 It ends with Zero. 261 00:23:56.830 --> 00:23:57.640 Count Binface. 262 00:23:57.640 --> 00:23:59.260 Count Binface. 263 00:23:59.260 --> 00:24:02.040 Who sadly lost today at Clacton-on-Sea. I love my countrymen. 264 00:24:02.040 --> 00:24:03.900 Oh, did he? Oh man. 265 00:24:03.900 --> 00:24:05.500 Yeah, it didn't work out. 26% of the vote though. 266 00:24:05.500 --> 00:24:05.800 That's pretty good. 267 00:24:05.800 --> 00:24:07.460 It's not bad. 268 00:24:07.460 --> 00:24:12.810 I mean, the other guy, the guy who resigned and who won't be named here, did win. 269 00:24:12.810 --> 00:24:13.280 Yes. 270 00:24:13.280 --> 00:24:14.220 Nigel Binface. 271 00:24:14.220 --> 00:24:17.120 Yeah, no. Don't besmirch the Binface. 272 00:24:17.120 --> 00:24:26.530 Use that one. Yeah, I mean, it's... Yeah, you've got to love the Brits for not taking things too seriously, but still. 273 00:24:26.530 --> 00:24:26.750 yeah 274 00:24:26.750 --> 00:24:27.650 You know, I still... 275 00:24:27.650 --> 00:24:28.100 I mean... 276 00:24:28.100 --> 00:24:45.240 someone pointed out that like the people of Clacton, the place where... you know, this is real for them. So we're all treating it as like a bit of fun, but you know, and it is making some political points, but like, yeah, it's a thing. Anyway, that was not on my list of things to talk about with you today. 277 00:24:45.240 --> 00:24:50.500 But you gave me an opportunity to talk about it. So, Count... what was he saying? Count Basie? 278 00:24:50.500 --> 00:24:51.720 Count Chocula. 279 00:24:51.720 --> 00:24:53.440 No, Count... Count Zero. 280 00:24:53.440 --> 00:24:57.100 I think it's... 281 00:24:57.100 --> 00:25:10.140 The books, that's right. 282 00:25:10.140 --> 00:25:10.300 I think it's like counters. I forget what the middle book is called. 283 00:25:10.300 --> 00:25:10.300 I love to count. 284 00:25:10.300 --> 00:25:10.320 But all of those, I mean, you know, every... 285 00:25:10.320 --> 00:25:10.320 Okay, I think... 286 00:25:10.320 --> 00:25:19.110 What is it? Who's your favorite vampire? The one from Sesame Street. He doesn't count. Oh, I assure you that he does. 287 00:25:19.110 --> 00:25:22.000 That's all right. That's brilliant. I don't know how we're going to top that now. 288 00:25:22.000 --> 00:25:26.260 We've peaked. We've peaked. 20 minutes into this conversation, we peaked. 289 00:25:26.260 --> 00:25:26.560 yeah 290 00:25:26.560 --> 00:25:32.860 Oh yeah, no, we're talking about 90s cyberpunk and the dystopian future that we live in. 291 00:25:32.860 --> 00:25:39.100 Thanks. Yeah, so you were saying that the 90s... this is the blueprint. So what's coming in the future? Well, actually, no. Is there like, so. 292 00:25:39.100 --> 00:25:56.080 I think... we've talked a lot about how folks feel about LLMs. And I think, you know, you and I have both been on various rollercoaster-like ride journeys with this is great, to this is terrible, to this is I'm going to lose my job, to now I'm the most powerful programmer, to now this is what a mess I've got, and everything in between, right? 293 00:25:56.080 --> 00:26:03.790 And that's not even taking into account all of the moral and ethical aspects, which are also super important but I have sort of compartmentalized. 294 00:26:03.790 --> 00:26:04.400 okay 295 00:26:04.400 --> 00:26:14.140 But... you know, not being able to run an LLM. If, for example, tomorrow Anthropic decides that I've been a bad boy. 296 00:26:14.140 --> 00:26:24.470 or my company decided that, you know, they're not liked by Anthropic or whomever else they're providing, or whoever, you know, whatever. That's a concern. 297 00:26:24.470 --> 00:26:24.940 Yeah. 298 00:26:24.940 --> 00:26:33.600 So you mentioned about running your own stuff. Can we do that with LLMs? Do I just need a warehouse-sized computer to be able to run something worthwhile? 299 00:26:33.600 --> 00:26:46.180 Yeah, you know, I think it's arguable that the end state of all information is to be free. 300 00:26:46.180 --> 00:26:57.310 Right. You know, it's like all of these models of like copy protection where you have like a private key that only three people in the world have. It's like someone is going to figure that out. Like, ask Sony. 301 00:26:57.310 --> 00:26:57.560 Yeah. 302 00:26:57.560 --> 00:27:02.000 right, like someone is going to figure that out. 303 00:27:02.000 --> 00:27:05.450 This is not going to work. 304 00:27:05.450 --> 00:27:05.720 Yeah. 305 00:27:05.720 --> 00:27:18.980 Right. And, you know, the model weights and LLMs are just too valuable for people not to try to figure out what they are. Right. And especially when you consider the fact that there is no. 306 00:27:18.980 --> 00:27:27.860 like world government entity that can, you know, enforce any sort of protection for any of this information in any sort of like. 307 00:27:27.860 --> 00:27:28.760 Right. 308 00:27:28.760 --> 00:27:30.660 global way. 309 00:27:30.660 --> 00:27:39.310 Are model weights different though, in the same way? They're more like a compiled program. Right. If you, you know, if you imagine I give you the EXE 310 00:27:39.310 --> 00:27:39.640 Sure. 311 00:27:39.640 --> 00:27:53.520 of my code but not the source code to it. I give you the EXE and you can go away and use that program as much as you like, but you can't really do anything more with it. You can't extend it or build on it or anything like that because it's kind of like already been compiled. It's very difficult to reverse engineer. 312 00:27:53.520 --> 00:27:53.520 Yeah. 313 00:27:53.520 --> 00:28:08.940 And I feel like the weights, or you know even the open weights that you can go and get on the internet, have that flavor. It's like, sure. Here's the result of compiling, but you know, we won't, you can't buy the compiler because it's like 45 data centers full of machines that ran for like two years. 314 00:28:08.940 --> 00:28:37.540 to compile the program. And now here's the program. So, you know, the fact that the program is dressed up to me as a service is, you know, problematic. But even if they handed me the weights, I guess that's what I asked you, in fairness, but like, I'm just now thinking ahead, right? You know, you could give me the, you know, 24 terabyte disc that contains the weights to the best model and I could somehow cobble together running it, but I'd be stuck with that until the next time somebody hands me the 48 terabytes of the next model. 315 00:28:37.540 --> 00:28:38.020 Sure. 316 00:28:38.020 --> 00:28:45.880 It feels like. I don't know. I've always balked at the term open weights because it's like, open? Yes, I guess. I suppose so, but... 317 00:28:45.880 --> 00:28:46.980 Right. 318 00:28:46.980 --> 00:28:50.600 Yeah, sorry. Anyway, I cut you off. Where were you going with this? 319 00:28:50.600 --> 00:29:05.000 well no, I mean, you know, that's a fair point and that is definitely, you know, an avenue that we could explore here, is like how open are open source models, if all you're getting, if you're, yeah. 320 00:29:05.000 --> 00:29:05.950 What does it mean? I mean, because like, yeah, open source. I suppose that's it. I compare it to open source, and open source is like, hey, here's the source code. Now that might as well be alien. 321 00:29:05.950 --> 00:29:13.860 yeah 322 00:29:13.860 --> 00:29:20.780 to most people on the planet, but in principle, you can look at it and go, this was meant for us to build on and continue, and in fact the people who gave me this or gave me access to this, they are using this to continue and get better. 323 00:29:20.780 --> 00:29:22.630 and the next version of the product will be built. 324 00:29:22.630 --> 00:29:22.780 Mm-hmm. 325 00:29:22.780 --> 00:29:28.520 And it's most likely, you know, obviously there are edge cases to these. Whereas the weights feels like so derived. 326 00:29:28.520 --> 00:29:37.940 It's not really the cool and interesting bit of it from the point of view of continuing and being a builder and not being locked into this brain forevermore. 327 00:29:37.940 --> 00:29:38.340 Yeah. 328 00:29:38.340 --> 00:29:43.520 I wonder what... but then the resources are so, so massive that I can't imagine. 329 00:29:43.520 --> 00:29:44.420 Yeah. 330 00:29:44.420 --> 00:29:53.980 that you could do much with it. But at least you could, right? So in principle... And have you tried this, by the way? Have you experimented with local models or anything? 331 00:29:53.980 --> 00:30:08.060 Running local models? No, it's been on my list of things. It's like... there's going to be a tipping point probably in the next 18 to 24 months for me where it's like, part of my day job is probably going to be figuring this out. 332 00:30:08.060 --> 00:30:15.800 And so like, I'm kind of just waiting for that to happen because then it'll be like more of a focus, right? 333 00:30:15.800 --> 00:30:15.940 Woof! 334 00:30:15.940 --> 00:30:21.040 You know what I mean? Like, it's not just like something I'm monkeying around with. It's like, oh no, I need to make this work for real. 335 00:30:21.040 --> 00:30:22.420 You actually have to understand it. 336 00:30:22.420 --> 00:30:24.100 To be able to do it, yeah. 337 00:30:24.100 --> 00:30:26.360 Yeah. You know, and like, the world will be completely different in 24 months. 338 00:30:26.360 --> 00:30:26.700 What? 339 00:30:26.700 --> 00:30:31.560 Technologies will be different. Everything will be different. So I might as well figure it out for real when I need to figure it out for real. 340 00:30:31.560 --> 00:30:48.040 Yeah. Yeah, it's interesting. I suppose... I mean, we're, sorry, excuse the dog. We're meandering all over here, but like, you know, like even getting the equipment to run these things on, obviously one can find GPUs around or, you know, in the house I have a few around, but they're not really up to scratch for this kind of stuff. 341 00:30:48.040 --> 00:30:48.230 Mm-hmm. 342 00:30:48.230 --> 00:31:05.300 And then, you know, like I know from my own day job that sourcing, you know, top end GPUs and the RAM that goes with them is incredibly difficult right now, and I wonder if that's going to change. But one can only hope that there are some advances in the way that we... 343 00:31:05.300 --> 00:31:05.730 right 344 00:31:05.730 --> 00:31:11.220 evaluate, not evaluate, to make better how we run inference or whatever 345 00:31:11.220 --> 00:31:12.400 Sure, yeah 346 00:31:12.400 --> 00:31:26.600 on hardware, because GPUs are designed well for graphics. I still feel sad for rows and rows of graphic processing units in racks that have never had a video display plugged into them ever. 347 00:31:26.600 --> 00:31:26.700 Right. 348 00:31:26.700 --> 00:31:37.590 They've never generated a frame of a game, which is the point of a GPU, in my opinion. But, you know, it makes me sad. It's like battery-farmed hens, you know. It's like, oh, God, it's not right. Be free. Do what you're supposed to do. 349 00:31:37.590 --> 00:31:38.060 Yes. 350 00:31:38.060 --> 00:31:40.460 I need some free range GPUs, please. 351 00:31:40.460 --> 00:31:44.390 Yeah, well, I've got a drawer of... I've got a Voodoo 2 in a drawer somewhere. 352 00:31:44.390 --> 00:31:44.640 I wouldn't. 353 00:31:44.640 --> 00:31:45.680 I think that'll help, right? 354 00:31:45.680 --> 00:31:49.140 We feed them a steady diet of Doom Eternal and... 355 00:31:49.140 --> 00:31:55.320 Oh. That's too new for my GPUs, more like VGA adapters with a bit. 356 00:31:55.320 --> 00:31:56.440 Yeah. Uh-huh. 357 00:31:56.440 --> 00:31:57.220 Yeah, anyway. 358 00:31:57.220 --> 00:32:11.110 Yeah, no, there's, I mean, well, and this is kind of what I'm saying about the, you know, the information wants to be free argument on models is like, yes, it does take a lot of resources to train them and generate those weights. 359 00:32:11.110 --> 00:32:12.020 Right. 360 00:32:12.020 --> 00:32:23.450 But once you've done it... like, they're not going to disappear. And imagine a world where there is an open source 361 00:32:23.450 --> 00:32:29.460 agentic coding model that is better than, say, Fable 5. Right. Like. 362 00:32:29.460 --> 00:32:31.020 Right. 363 00:32:31.020 --> 00:32:45.700 That is probably going to happen at some point. Right? Like. Do you maybe have some situations in which you want to use a better one? Sure. But if you're just trying to like vibe code the notes app to run on your personal server, to like, it was like. 364 00:32:45.700 --> 00:32:46.020 You don't really. 365 00:32:46.020 --> 00:32:48.900 That model will be more than sufficient, right? 366 00:32:48.900 --> 00:33:01.220 All right. And I can see a world in which that kind of open... yeah. The thing that I'm interested in, and now this is going to be part two of a massive thing, is that like, you know, after the base training of a model, like where you've just fed 367 00:33:01.220 --> 00:33:15.940 some subset of all human knowledge in through however you've acquired it, which is its own massive asterisk and footnote and ethical dilemma, but you've got some amount of knowledge, you train a neural net on it, and then there are... 368 00:33:15.940 --> 00:33:35.000 so much post-training that happens to like make it valuable, make it aligned, make it safe. And then tuning after that to make it, you know, like, hey, don't write code that's terrible. And that's the bit I'd be interested in. I'd like to be able to go like, hey, I don't want to have to explain in text over and over again to a model that... 369 00:33:35.000 --> 00:33:38.830 Don't write comments like this, for God's sake, stop it. 370 00:33:38.830 --> 00:33:39.360 Mm-hmm. 371 00:33:39.360 --> 00:34:05.140 because you know you and I, I'm sure, have done that a million times. But I want to be able to say like, why didn't we just tune it? Can I personalize the weights so that it doesn't do that in the first place rather than me having to exhort it every single time to do it the way I want? And that feels like the last steps of an open model where you're like, okay, that's tractable. Here's the 20 terabyte intermediate model. You can add layers of tuning onto it. And then like, maybe that is something I could do. Maybe I'm just beyond, and you know, yeah. 372 00:34:05.140 --> 00:34:08.160 I have to admit I don't know nearly enough about how. 373 00:34:08.160 --> 00:34:09.580 how that process works now. 374 00:34:09.580 --> 00:34:12.120 how that process works to know whether that's a feasible thing or not. 375 00:34:12.120 --> 00:34:17.620 No, but it seems like I can imagine open companies starting that process, or you know maybe maybe this stuff exists on Hugging Face. 376 00:34:17.620 --> 00:34:17.920 Sure. 377 00:34:17.920 --> 00:34:24.220 Maybe this is the kind of thing that, you know, I would, if I spent a bit more time looking into it, I would discover it can actually be done. 378 00:34:24.220 --> 00:34:26.000 Yeah. 379 00:34:26.000 --> 00:34:34.840 But now it's interesting. And yeah, just so, and the irony here that the sort of leader in open weight, as far as I can tell, is the Chinese government. 380 00:34:34.840 --> 00:34:35.200 Right. 381 00:34:35.200 --> 00:34:35.560 Yeah. 382 00:34:35.560 --> 00:34:36.890 That seems like, wow. 383 00:34:36.890 --> 00:34:37.200 Yep. 384 00:34:37.200 --> 00:34:41.980 That was not on my bingo card for this year, is like champions of openness and freedom. 385 00:34:41.980 --> 00:34:44.300 Yeah. 386 00:34:44.300 --> 00:34:47.200 You know, I'm sure you can find a, as you say, a 90s sci-fi novel that... 387 00:34:47.200 --> 00:34:50.680 We need an AI company that's open. What would we call it? 388 00:34:50.680 --> 00:34:51.880 No. 389 00:34:51.880 --> 00:34:52.830 I don't. 390 00:34:52.830 --> 00:34:53.140 No. 391 00:34:53.140 --> 00:34:55.440 I just can't think of a good name. 392 00:34:55.440 --> 00:34:56.680 No, there's no... I don't think you can pick one. 393 00:34:56.680 --> 00:34:59.160 I wonder if the domain, I'll just check some domains and see if anything's available. 394 00:34:59.160 --> 00:35:00.560 Yeah, see whether you can find one. 395 00:35:00.560 --> 00:35:01.340 Yeah. 396 00:35:01.340 --> 00:35:02.380 Yeah. 397 00:35:02.380 --> 00:35:07.600 But yeah, like it's entirely possible that in the limit, the only thing that matters is GPUs, power and cooling. 398 00:35:07.600 --> 00:35:07.940 Anyway. 399 00:35:07.940 --> 00:35:19.440 Right. Because like. Models will get good enough. Their information will leak out into the world as all information does. And you will reach a point where. 400 00:35:19.440 --> 00:35:33.200 you know, you don't necessarily have to be beholden to a company to use their models as a service. You can go and either purchase or maybe just acquire for free because they're just freely available. 401 00:35:33.200 --> 00:35:33.280