WEBVTT - Hater Season: Cal Newport on AI Reporting

0:00:02.440 --> 0:00:08.320
<v Speaker 1>As media your witness to a great becoming. It's better

0:00:08.360 --> 0:00:23.520
<v Speaker 1>offline and I'm ed Zetron. Today we're joined by computer

0:00:23.600 --> 0:00:26.960
<v Speaker 1>science professor and tech writer Caln Newport for Hater Season.

0:00:27.040 --> 0:00:28.280
<v Speaker 1>Cal thank you for joining me.

0:00:29.240 --> 0:00:30.600
<v Speaker 2>Always happy to do some heding.

0:00:30.640 --> 0:00:34.200
<v Speaker 1>I suppose, well, you had an excellent YouTube that I'll

0:00:34.240 --> 0:00:37.239
<v Speaker 1>be linking in the show notes about the mistakes in

0:00:37.320 --> 0:00:40.440
<v Speaker 1>AI reporting. Though I would my hater in me says,

0:00:40.479 --> 0:00:43.720
<v Speaker 1>I don't think these are mistakes, but you really you touched.

0:00:43.720 --> 0:00:46.040
<v Speaker 1>One of the best videos I've seen on AI report

0:00:46.200 --> 0:00:48.120
<v Speaker 1>or AI in general was what you did. But it's

0:00:48.159 --> 0:00:51.479
<v Speaker 1>basically this thing of like the digital iic that these

0:00:51.520 --> 0:00:53.720
<v Speaker 1>stories are meant to make you feel uncomfortable, and of

0:00:53.760 --> 0:00:56.640
<v Speaker 1>course this full astonishment thing. You know what, Just run

0:00:56.680 --> 0:00:58.800
<v Speaker 1>the tape. Tell me a little bit about what the

0:00:58.840 --> 0:01:03.320
<v Speaker 1>bits that you found, because I watched it going yeah, yeah, yeah,

0:01:03.440 --> 0:01:04.679
<v Speaker 1>like an angry person.

0:01:05.160 --> 0:01:07.440
<v Speaker 2>Yeah I had that. I could imagine you. When I

0:01:07.480 --> 0:01:09.039
<v Speaker 2>was recording that video, I was like, I bet ed

0:01:09.160 --> 0:01:15.240
<v Speaker 2>is well now the great time exactly? Well, I mean,

0:01:15.319 --> 0:01:17.039
<v Speaker 2>let me just give the context, right, So I'm in

0:01:17.080 --> 0:01:19.600
<v Speaker 2>an interesting situation for observing this because you know, I

0:01:19.600 --> 0:01:22.840
<v Speaker 2>am a computer scientist, so I'm not afraid of the technologies.

0:01:22.880 --> 0:01:27.280
<v Speaker 2>I'm happy to talk about transformers and feed forward networks

0:01:27.319 --> 0:01:29.600
<v Speaker 2>and diffusion models and like, that's not that scary to me.

0:01:29.880 --> 0:01:33.120
<v Speaker 2>But I also write about technology. My main journalistic home

0:01:33.160 --> 0:01:34.720
<v Speaker 2>is to New Yorker, where I write a lot about

0:01:34.800 --> 0:01:37.720
<v Speaker 2>you know, I do AI journalism there, so I'm up

0:01:37.760 --> 0:01:41.200
<v Speaker 2>on that as well. And so I'm often noticing things

0:01:41.319 --> 0:01:47.200
<v Speaker 2>in journalism. When I see AI coverage that is faulty,

0:01:47.280 --> 0:01:49.360
<v Speaker 2>it really catches my attention because I have a foot

0:01:49.360 --> 0:01:51.360
<v Speaker 2>in both of these worlds. So there's a lot of

0:01:51.360 --> 0:01:54.040
<v Speaker 2>good AI report out there. There's also a lot of trash,

0:01:54.200 --> 0:01:56.200
<v Speaker 2>and I really wanted to help people figure out how

0:01:56.200 --> 0:01:57.920
<v Speaker 2>do you sort it? Like how do you figure out

0:01:58.400 --> 0:02:01.520
<v Speaker 2>if you're reading something? Should I pull the ripcord on

0:02:01.560 --> 0:02:03.440
<v Speaker 2>this article? Like this is not helping me, Like how

0:02:03.480 --> 0:02:04.400
<v Speaker 2>do you know if it's good or not? So I

0:02:04.440 --> 0:02:05.680
<v Speaker 2>was like, Okay, here's what I'll do is I'll come

0:02:05.760 --> 0:02:08.600
<v Speaker 2>up with the three most common traps I see in

0:02:08.680 --> 0:02:10.720
<v Speaker 2>AI reporting that makes me want to, you know, throw

0:02:10.760 --> 0:02:13.360
<v Speaker 2>my iPad at the wall. And I came up with

0:02:13.400 --> 0:02:15.040
<v Speaker 2>three and I'll just give you the three names. I

0:02:15.080 --> 0:02:17.880
<v Speaker 2>made up. All these names don't. I don't think they're great.

0:02:17.919 --> 0:02:19.960
<v Speaker 2>My producer thinks he loves them. But here we go

0:02:20.480 --> 0:02:24.280
<v Speaker 2>vibe reporting that's number one. So that is where you

0:02:25.360 --> 0:02:30.680
<v Speaker 2>will omit certain facts and put loosely related quotes next

0:02:30.680 --> 0:02:33.160
<v Speaker 2>to each other in a way that creates a general

0:02:33.320 --> 0:02:35.080
<v Speaker 2>vibe that you want to be true but it's not

0:02:35.160 --> 0:02:39.280
<v Speaker 2>quite true. So you don't actually make a concrete claim

0:02:39.320 --> 0:02:43.760
<v Speaker 2>that's not true, but you imply that claim by what

0:02:43.800 --> 0:02:45.800
<v Speaker 2>you omit or what you put in your story, or

0:02:45.800 --> 0:02:48.200
<v Speaker 2>what quotes you put next to it. So you'll put

0:02:48.200 --> 0:02:52.440
<v Speaker 2>a quote, for example, about layoffs at the gaming division

0:02:52.480 --> 0:02:56.160
<v Speaker 2>at Microsoft, next to an unrelated quote about concerns about

0:02:56.160 --> 0:03:00.320
<v Speaker 2>AI and it's impact on jobs. Now you have the vibe, oh, man,

0:03:00.360 --> 0:03:02.200
<v Speaker 2>all these people just got laid off because of AI.

0:03:02.320 --> 0:03:04.560
<v Speaker 2>AI is taking jobs, where in reality the layouts had

0:03:04.600 --> 0:03:06.040
<v Speaker 2>nothing to do with AI. But you put these things

0:03:06.040 --> 0:03:07.600
<v Speaker 2>next to each other, you give that sense. Then I

0:03:07.639 --> 0:03:11.480
<v Speaker 2>had a mining digital ick. So to me, that's any

0:03:11.520 --> 0:03:13.640
<v Speaker 2>AI story where you take an example from the edges

0:03:13.680 --> 0:03:16.000
<v Speaker 2>of AI, like something that wireheads and San Francisco are

0:03:16.080 --> 0:03:18.880
<v Speaker 2>up to, and you just tell a story that's unsettling

0:03:19.600 --> 0:03:22.520
<v Speaker 2>without talking about any of the technical details like well,

0:03:22.560 --> 0:03:24.880
<v Speaker 2>what's different here? Was there a technical innovation we need

0:03:24.919 --> 0:03:28.639
<v Speaker 2>to know about, and not discussing any concrete implications. Oh,

0:03:28.680 --> 0:03:30.680
<v Speaker 2>this means this is going to change in the future,

0:03:30.840 --> 0:03:32.400
<v Speaker 2>or it's going to have an impact on this sector.

0:03:32.440 --> 0:03:35.600
<v Speaker 2>You're just telling a story to unsettle, and I think

0:03:35.600 --> 0:03:38.520
<v Speaker 2>a lot of the coverage of molt Book and open

0:03:38.560 --> 0:03:40.880
<v Speaker 2>Claw fell into that. And then finally this astonishment, which

0:03:40.920 --> 0:03:44.040
<v Speaker 2>is more of a YouTube phenomenon than a print journalism phenomenon.

0:03:44.080 --> 0:03:46.360
<v Speaker 2>But that's where every single thing that happens in AI

0:03:46.720 --> 0:03:52.600
<v Speaker 2>is insane, amazing, terrifying. Everything is going to change, and

0:03:52.640 --> 0:03:57.720
<v Speaker 2>so you constantly create this atmosphere of something seismic just happened,

0:03:58.240 --> 0:04:00.680
<v Speaker 2>so that the consumer of the infra ends up in

0:04:00.680 --> 0:04:02.200
<v Speaker 2>a bit of a panic, like I can put my

0:04:02.200 --> 0:04:05.120
<v Speaker 2>finger on exactly what's terrible, but like everything terrible is happening.

0:04:05.520 --> 0:04:08.760
<v Speaker 2>Those three traps, to me, should be automatic rip chords

0:04:08.760 --> 0:04:09.840
<v Speaker 2>from what you're reading or watching.

0:04:11.040 --> 0:04:13.119
<v Speaker 1>The only disagreement I'll have is that you would say

0:04:13.120 --> 0:04:15.840
<v Speaker 1>that this isn't the majority of AI journalism, because I

0:04:15.840 --> 0:04:18.880
<v Speaker 1>actually argue it would be, especially that kind of the

0:04:19.640 --> 0:04:23.600
<v Speaker 1>beginning one. The vibe reporting is very common because The

0:04:23.640 --> 0:04:26.440
<v Speaker 1>big one right now I'm seeing is this AI software

0:04:26.760 --> 0:04:29.839
<v Speaker 1>AI is replacing software thing. If you are a reporter

0:04:30.040 --> 0:04:32.280
<v Speaker 1>and cal is not making this statement I am. If

0:04:32.279 --> 0:04:37.440
<v Speaker 1>you're a reporter and you bring up anthropic cowork around anything,

0:04:38.080 --> 0:04:40.880
<v Speaker 1>you are wrong. I was about to call you a name,

0:04:40.880 --> 0:04:43.800
<v Speaker 1>but I'm being nice today for some reason. All Right,

0:04:43.800 --> 0:04:46.120
<v Speaker 1>you're a dip shit, because you are a dip shit.

0:04:46.160 --> 0:04:48.560
<v Speaker 1>If you look at claude cowork, which is a thing

0:04:48.640 --> 0:04:50.719
<v Speaker 1>for fucking around on your desktop, and you say this

0:04:50.839 --> 0:04:53.240
<v Speaker 1>is going to compete with Salesforce, you just don't know

0:04:53.279 --> 0:04:56.320
<v Speaker 1>what you're talking about. You are wrong. But then one

0:04:56.360 --> 0:04:59.800
<v Speaker 1>abstraction higher is this idea that claud code is going

0:04:59.800 --> 0:05:02.800
<v Speaker 1>to destroy SaaS software as a service, and the idea

0:05:02.880 --> 0:05:05.040
<v Speaker 1>being that software as a service is this thing that

0:05:05.560 --> 0:05:08.080
<v Speaker 1>people are just going to stop building their own CRMs.

0:05:08.120 --> 0:05:11.240
<v Speaker 1>They're going to stop building their own per seat software things,

0:05:11.279 --> 0:05:13.720
<v Speaker 1>but they're going to build it internally. This just I

0:05:13.720 --> 0:05:15.640
<v Speaker 1>don't know if you've seen this, cow It just reflects

0:05:15.640 --> 0:05:19.520
<v Speaker 1>a complete lack of understanding of how software works. Yeah,

0:05:19.520 --> 0:05:24.000
<v Speaker 1>because you don't pay Microsoft for Microsoft three sixty five

0:05:24.040 --> 0:05:27.320
<v Speaker 1>teams because you can't build your own word or what

0:05:27.320 --> 0:05:29.080
<v Speaker 1>have you. I don't think you could. But nevertheless, it's

0:05:29.120 --> 0:05:31.320
<v Speaker 1>also because they maintain it, because they make sure it

0:05:31.320 --> 0:05:33.520
<v Speaker 1>doesn't break, or if it breaks, they fix the bit.

0:05:33.600 --> 0:05:36.320
<v Speaker 1>So they make sure that it stays up all the time.

0:05:37.080 --> 0:05:39.360
<v Speaker 1>They make sure it's accessible that has secure log in.

0:05:40.680 --> 0:05:43.320
<v Speaker 2>Well, look, there's a few things going on here. One

0:05:43.360 --> 0:05:45.800
<v Speaker 2>of these things I reported on last month. I did

0:05:45.839 --> 0:05:49.000
<v Speaker 2>a big New Yorker piece on agents. Right, so this

0:05:49.279 --> 0:05:50.960
<v Speaker 2>is relevant to the cloud code and how people are

0:05:50.960 --> 0:05:53.159
<v Speaker 2>thinking about the current future, and there was this basically,

0:05:53.240 --> 0:05:56.839
<v Speaker 2>here's what seems to have happened. Cloud code and these

0:05:56.839 --> 0:06:00.800
<v Speaker 2>other command line interface agents can do really cool things,

0:06:00.839 --> 0:06:03.600
<v Speaker 2>and I'm using the work cool here in a very

0:06:03.640 --> 0:06:05.160
<v Speaker 2>careful fine, yeah cool.

0:06:05.520 --> 0:06:07.719
<v Speaker 1>I would like you to be like completely like give

0:06:07.720 --> 0:06:09.160
<v Speaker 1>people the actual explanation here.

0:06:09.320 --> 0:06:12.400
<v Speaker 2>Yeah, so cool, meaning like Oculus. Right, you put on

0:06:12.520 --> 0:06:15.800
<v Speaker 2>the the Oculus visors for the first time. Everyone had

0:06:15.800 --> 0:06:19.520
<v Speaker 2>the same reaction. This is really cool. Now that's separate

0:06:19.560 --> 0:06:22.279
<v Speaker 2>from that's a trillion dollar business, Like, let's put a side,

0:06:22.279 --> 0:06:25.520
<v Speaker 2>but this is really cool. I'm seeing three D in

0:06:25.920 --> 0:06:28.200
<v Speaker 2>a world where it tracks my head. Cloud code and

0:06:28.240 --> 0:06:32.040
<v Speaker 2>other command line interface coding tools became like that. For programmers,

0:06:32.080 --> 0:06:35.159
<v Speaker 2>it was like really fun the watch it doing multi

0:06:35.320 --> 0:06:38.400
<v Speaker 2>step execution of the construction of demos or this or that,

0:06:38.800 --> 0:06:40.760
<v Speaker 2>and the reason why it could do those cool demos

0:06:40.800 --> 0:06:42.560
<v Speaker 2>it was sort of well suited for that world because

0:06:42.560 --> 0:06:45.080
<v Speaker 2>that's a world that exists only in text. So cloud

0:06:45.080 --> 0:06:47.480
<v Speaker 2>code works on a command line interface. It's all text based,

0:06:47.960 --> 0:06:50.680
<v Speaker 2>and it works with a filesystem. You can write files,

0:06:50.839 --> 0:06:54.159
<v Speaker 2>edit files, sind files to compilers. It's all exists in

0:06:54.160 --> 0:06:56.360
<v Speaker 2>a small number of commands on a command line interface.

0:06:56.400 --> 0:06:57.880
<v Speaker 2>It's all in a world of text, because that's where

0:06:57.880 --> 0:07:01.160
<v Speaker 2>computer programs are built. That's a perfect case for lms,

0:07:01.240 --> 0:07:03.200
<v Speaker 2>which love dealing with text, and they love dealing with

0:07:03.240 --> 0:07:07.359
<v Speaker 2>structure text like computer code. There was an extrapolation that

0:07:07.400 --> 0:07:12.200
<v Speaker 2>then happened, and really this caught on January of twenty five,

0:07:12.720 --> 0:07:15.680
<v Speaker 2>which is where you first began to get this sentiment

0:07:15.840 --> 0:07:19.320
<v Speaker 2>of oh, it's doing such cool things over in the

0:07:19.320 --> 0:07:23.080
<v Speaker 2>world of command line interfaces in code. Certainly these agents

0:07:23.120 --> 0:07:26.480
<v Speaker 2>can now soon do similar cool things and like all

0:07:26.520 --> 0:07:28.600
<v Speaker 2>different things we do on a computer, and that is

0:07:28.680 --> 0:07:32.800
<v Speaker 2>what laid this foundation of people were so impressed, programmers

0:07:32.800 --> 0:07:35.680
<v Speaker 2>were so impressed by the coolness of what was happening

0:07:35.720 --> 0:07:38.400
<v Speaker 2>with cloud code and the other command line interface tools

0:07:38.640 --> 0:07:43.640
<v Speaker 2>that they extrapolated that vibe over to other computer usage.

0:07:43.680 --> 0:07:46.880
<v Speaker 2>Is the point of that article I reported was, Oh,

0:07:46.920 --> 0:07:48.720
<v Speaker 2>it turns out like everything else we do on computer

0:07:48.800 --> 0:07:51.840
<v Speaker 2>is much harder. It's not six text commands on a

0:07:51.880 --> 0:07:54.520
<v Speaker 2>command line interface creating structure code that you can compile

0:07:54.560 --> 0:07:56.640
<v Speaker 2>and test to see if it compiled or not. It's

0:07:56.760 --> 0:08:00.840
<v Speaker 2>much messier. The interfaces are visual. I don't realize what

0:08:00.880 --> 0:08:03.400
<v Speaker 2>complexity goes into the things we click and select doing

0:08:03.400 --> 0:08:05.120
<v Speaker 2>something as simple as even trying to just book like

0:08:05.120 --> 0:08:08.520
<v Speaker 2>a hotel in a new city or something like this.

0:08:08.640 --> 0:08:11.000
<v Speaker 2>And if you use a language model as to underline

0:08:11.040 --> 0:08:13.520
<v Speaker 2>logic and decision engine of making actual actions in the

0:08:13.520 --> 0:08:16.000
<v Speaker 2>real world, well, the language models make things up and

0:08:16.000 --> 0:08:18.600
<v Speaker 2>get things wrong or a little bit wrong twenty percent

0:08:18.600 --> 0:08:19.040
<v Speaker 2>of the time.

0:08:19.360 --> 0:08:23.000
<v Speaker 1>And a little bit wrong in computer science means breaking everything.

0:08:23.080 --> 0:08:25.480
<v Speaker 2>Well, it means a lot like it's okay in code

0:08:25.640 --> 0:08:28.320
<v Speaker 2>because they say, oh, that didn't compile, let's try again.

0:08:28.760 --> 0:08:31.080
<v Speaker 2>But when you're booking a hotel room, as I sort

0:08:31.080 --> 0:08:33.360
<v Speaker 2>of detailed that article, it means like you ended up

0:08:33.360 --> 0:08:35.920
<v Speaker 2>in the wrong city two years from now, and the

0:08:36.000 --> 0:08:38.439
<v Speaker 2>room costs six thousand dollars, right, and so it just

0:08:38.480 --> 0:08:40.839
<v Speaker 2>didn't work. So twenty twenty five was supposed to be

0:08:40.840 --> 0:08:42.520
<v Speaker 2>the year of the Agents and it just didn't work

0:08:42.559 --> 0:08:44.920
<v Speaker 2>and they don't really know how to fix it. But

0:08:44.960 --> 0:08:48.080
<v Speaker 2>we're still vibes. This is Vibe reporting. Yeah, but cloud

0:08:48.160 --> 0:08:51.240
<v Speaker 2>code is like really exciting, and so why can't we

0:08:51.280 --> 0:08:53.240
<v Speaker 2>do that with everything else in the computer. It's actually

0:08:53.240 --> 0:08:54.680
<v Speaker 2>a much harder problem than they were letting on.

0:08:55.360 --> 0:08:57.560
<v Speaker 1>Well. The thing is I've seen, especially like in twenty

0:08:57.600 --> 0:08:59.800
<v Speaker 1>twenty six, I've seen a lot more cloud code stuff.

0:09:00.240 --> 0:09:04.440
<v Speaker 1>Was there's been a very big consent manufacturing operation going

0:09:04.480 --> 0:09:07.760
<v Speaker 1>on right now. Wall Street Journal, Atlantic, CNBC Didrew Bosa.

0:09:07.960 --> 0:09:09.880
<v Speaker 1>This is a statement from me, not cow did you

0:09:09.920 --> 0:09:12.440
<v Speaker 1>both from CNBC should be fucking ashamed of herself going

0:09:12.440 --> 0:09:15.240
<v Speaker 1>on CNBC every day just going cord Code's got a

0:09:15.320 --> 0:09:18.440
<v Speaker 1>destryal software. She's on Twitter because she was able to

0:09:18.640 --> 0:09:23.240
<v Speaker 1>vibe code a some sort of Monday clone, which is

0:09:23.280 --> 0:09:26.600
<v Speaker 1>just like a it's a project management tool. It's like

0:09:26.640 --> 0:09:29.840
<v Speaker 1>I made some software that worked. This is everything now.

0:09:30.080 --> 0:09:31.720
<v Speaker 1>But that's kind of what you've been talking about with

0:09:31.720 --> 0:09:35.360
<v Speaker 1>this Vibe reporting, where it's like I did a small thing.

0:09:36.000 --> 0:09:38.720
<v Speaker 1>Now all things will be done in this manner, whether

0:09:38.760 --> 0:09:42.360
<v Speaker 1>it's possible, God no, God no. But she she, like

0:09:42.440 --> 0:09:44.720
<v Speaker 1>many reporters, are able to find a lot of people

0:09:44.720 --> 0:09:47.480
<v Speaker 1>who are invested in AI, who will absolutely go on

0:09:47.559 --> 0:09:51.200
<v Speaker 1>TV and say, Yep, it's completely true, that's gonna happen

0:09:51.360 --> 0:09:55.800
<v Speaker 1>one hundred percent. It's just so strange because it makes

0:09:55.840 --> 0:09:59.839
<v Speaker 1>me feel paranoid and kind of conspiratle when you look

0:09:59.840 --> 0:10:02.440
<v Speaker 1>at so the majority of news about AI, and it

0:10:02.520 --> 0:10:06.679
<v Speaker 1>is this viper POI. It's these vast extrapolations from sixteen

0:10:06.720 --> 0:10:12.160
<v Speaker 1>thousand job losses Amazon. They mentioned AI. This plus this

0:10:12.280 --> 0:10:16.240
<v Speaker 1>equals the AI is replacing people. It's just so it

0:10:16.280 --> 0:10:19.160
<v Speaker 1>makes me feel like uncomfortable with the world.

0:10:19.240 --> 0:10:22.199
<v Speaker 2>Yeah, well, can we sit for a moment on that

0:10:22.240 --> 0:10:24.080
<v Speaker 2>Amazon example, because I think it's a great please please,

0:10:24.120 --> 0:10:27.520
<v Speaker 2>it frustrates me that one frustling. All right, So Amazon

0:10:27.600 --> 0:10:31.760
<v Speaker 2>lays off sixteen thousand people, right, all right, it's covered

0:10:32.640 --> 0:10:35.559
<v Speaker 2>in two different ways, so the vibe reporting way it's

0:10:35.600 --> 0:10:37.839
<v Speaker 2>covered I in my newsletter, in my podcast, I looked

0:10:37.880 --> 0:10:41.480
<v Speaker 2>an example from Quartz and it was covered as clearly

0:10:41.600 --> 0:10:45.320
<v Speaker 2>intended to imply Amazon laid off sixteen thousand people because

0:10:45.360 --> 0:10:48.600
<v Speaker 2>of AI. They're being replaced by AI. They put the

0:10:48.640 --> 0:10:52.680
<v Speaker 2>subhead of the article was the CEO of Amazon talking

0:10:52.720 --> 0:10:57.319
<v Speaker 2>about how AI is going to increasingly disrupt the job force,

0:10:57.960 --> 0:11:02.080
<v Speaker 2>and then the article itself, no alternative explanations are given

0:11:02.120 --> 0:11:03.920
<v Speaker 2>for these layoffs. They kind of just give the details

0:11:03.920 --> 0:11:05.400
<v Speaker 2>of like here's how many people are laid off and

0:11:05.400 --> 0:11:06.839
<v Speaker 2>here's where they figured it out, and then they put

0:11:06.840 --> 0:11:08.440
<v Speaker 2>a couple of quotes in there about AI being very

0:11:08.480 --> 0:11:11.640
<v Speaker 2>disruptive and being able to automate jobs. It turns out

0:11:11.640 --> 0:11:13.679
<v Speaker 2>those layoffs had nothing to do with AI, and you

0:11:13.720 --> 0:11:16.480
<v Speaker 2>can find other reporting that focused because it was reporting

0:11:16.480 --> 0:11:18.040
<v Speaker 2>that was for the financial market, so it was trying

0:11:18.040 --> 0:11:20.240
<v Speaker 2>to focus on what the hell is actually happening, and

0:11:20.520 --> 0:11:22.560
<v Speaker 2>the deeper reporting was like, yeah, they laid off a

0:11:22.559 --> 0:11:25.040
<v Speaker 2>bunch of managers because like a lot of tech companies,

0:11:25.080 --> 0:11:28.640
<v Speaker 2>they over hired during the pandemic because cloud computing became

0:11:28.720 --> 0:11:30.480
<v Speaker 2>much in demand during the pandemics, So a lot of

0:11:30.520 --> 0:11:32.679
<v Speaker 2>tech companies over hired during the pandemic and they're all

0:11:32.679 --> 0:11:35.760
<v Speaker 2>shedding those jobs again. And Amazon's pretty ruthless about this, right,

0:11:35.920 --> 0:11:38.520
<v Speaker 2>They're always looking for excuses to fire people, and they said,

0:11:38.520 --> 0:11:41.880
<v Speaker 2>we have too much bureaucratic bloat, there's too many managers.

0:11:42.200 --> 0:11:43.920
<v Speaker 2>We're going to fire a bunch of these managers we

0:11:44.000 --> 0:11:46.280
<v Speaker 2>hired so that we can be more lean. Again, that

0:11:46.320 --> 0:11:48.200
<v Speaker 2>has nothing to do with AI. In fact, this is

0:11:48.240 --> 0:11:51.360
<v Speaker 2>the second or third round of these firings that they've

0:11:51.360 --> 0:11:54.880
<v Speaker 2>planned to do, the first round occurring before chat GPT

0:11:55.040 --> 0:11:58.040
<v Speaker 2>was even released. Like, this has nothing to do with

0:11:58.200 --> 0:12:00.640
<v Speaker 2>jobs being replaced by AI. But then you can vibe

0:12:00.640 --> 0:12:04.160
<v Speaker 2>report it because like, well, technically speaking, Amazon also is

0:12:04.200 --> 0:12:09.040
<v Speaker 2>investing money in AI products, So technically speaking, money saved

0:12:09.080 --> 0:12:12.320
<v Speaker 2>by firing these managers could be respent on AI. So

0:12:12.360 --> 0:12:15.000
<v Speaker 2>you can say with a like a semi straight face,

0:12:15.800 --> 0:12:18.560
<v Speaker 2>they fired people because of AI. But clearly, you know,

0:12:18.640 --> 0:12:20.440
<v Speaker 2>the impression that you're giving to the reader is that

0:12:20.440 --> 0:12:22.800
<v Speaker 2>they were replaced by AI and it had nothing to

0:12:22.800 --> 0:12:23.960
<v Speaker 2>do with it. And I heard, by the way, so

0:12:24.080 --> 0:12:25.960
<v Speaker 2>I wrote about that. I put in that video. I've

0:12:25.960 --> 0:12:29.240
<v Speaker 2>heard on background from multiple Amazon executives since they were like,

0:12:29.320 --> 0:12:33.160
<v Speaker 2>we were completely baffled by this coverage that was implying

0:12:33.280 --> 0:12:35.720
<v Speaker 2>that people were being fired here to do with AI.

0:12:35.880 --> 0:12:37.920
<v Speaker 2>This is just what we do at Amazon. We're ruthless,

0:12:37.960 --> 0:12:39.679
<v Speaker 2>Like if you're not earning your keep, we fire you.

0:12:40.040 --> 0:12:42.600
<v Speaker 2>They were completely baffled by that coverage and like things

0:12:42.600 --> 0:12:43.880
<v Speaker 2>for pointing it out, like what.

0:12:43.880 --> 0:12:46.360
<v Speaker 1>I go to be honest, cal if I got an

0:12:46.360 --> 0:12:49.079
<v Speaker 1>email like that from an Amazon executive, A tont go

0:12:49.200 --> 0:12:52.560
<v Speaker 1>fuck themselves? And I mean this nicely because Andy Jesse

0:12:52.840 --> 0:12:56.120
<v Speaker 1>last year June seventeen, twenty twenty five, the whole thing

0:12:56.160 --> 0:12:57.839
<v Speaker 1>about today and virtually every corner of the company we

0:12:58.000 --> 0:12:59.880
<v Speaker 1>use in generate if I had to make customers lives better.

0:13:00.080 --> 0:13:03.319
<v Speaker 1>I believe Amazon benefits from that obfuscation, and I think

0:13:03.320 --> 0:13:06.120
<v Speaker 1>they deliberately fuel it. Now that may be executives who

0:13:06.120 --> 0:13:06.719
<v Speaker 1>disagree with.

0:13:06.640 --> 0:13:08.880
<v Speaker 2>It, well, these were lower level managers, right, so not executs.

0:13:08.880 --> 0:13:11.640
<v Speaker 2>Shouldn't say those like people who work there that were

0:13:11.679 --> 0:13:14.480
<v Speaker 2>like oh yeah, no, no, no, they're not firing people

0:13:14.520 --> 0:13:17.000
<v Speaker 2>because like AI, they're just being brutal.

0:13:16.800 --> 0:13:18.440
<v Speaker 1>Right, right, they're the people who tell the truth.

0:13:18.640 --> 0:13:21.480
<v Speaker 2>Yeah, exactly exactly, But no, you are right. I think

0:13:21.480 --> 0:13:23.640
<v Speaker 2>for the there has been a lot of vibe reporting.

0:13:23.679 --> 0:13:25.400
<v Speaker 2>Basically the end I've been covering this for the last

0:13:25.440 --> 0:13:30.360
<v Speaker 2>two years, the entire sequence of job reductions that were

0:13:30.360 --> 0:13:33.200
<v Speaker 2>post pandemic corrections, so the entire tech industry over hired

0:13:33.240 --> 0:13:36.480
<v Speaker 2>during the pandemic. The entire tech industry cuts ups in

0:13:36.480 --> 0:13:38.480
<v Speaker 2>the last few years because they hire too many people

0:13:38.559 --> 0:13:41.520
<v Speaker 2>and now they have to correct back to where they

0:13:41.559 --> 0:13:44.840
<v Speaker 2>were pre pandemic. Consistently, across the board, these cuts are

0:13:44.880 --> 0:13:48.400
<v Speaker 2>vibe reported is due to AI consistently. You see exactly

0:13:48.400 --> 0:13:50.640
<v Speaker 2>that story. We see it, and I'm a computer scientist.

0:13:50.679 --> 0:13:54.120
<v Speaker 2>We see this in coverage of computer science majors as well.

0:13:54.160 --> 0:13:58.000
<v Speaker 2>Same idea. Computer science majors historically directly tied to the

0:13:58.000 --> 0:14:01.000
<v Speaker 2>tech industry. If they're cutting in a downside, majors go down.

0:14:01.360 --> 0:14:03.280
<v Speaker 2>If they're hiring. I mean, it's just economic. It's not

0:14:03.720 --> 0:14:06.719
<v Speaker 2>surprising when the technistry is booming, we get a lot

0:14:06.720 --> 0:14:09.320
<v Speaker 2>more majors because they're good jobs, right, And so computer

0:14:09.360 --> 0:14:11.560
<v Speaker 2>science majors went down in the last year or two

0:14:11.960 --> 0:14:15.000
<v Speaker 2>as the tech company started cutting. That was reported in

0:14:15.000 --> 0:14:19.480
<v Speaker 2>the Atlantic as kids are not majoring in AI because

0:14:19.520 --> 0:14:21.360
<v Speaker 2>AI makes the the.

0:14:21.400 --> 0:14:22.680
<v Speaker 1>Surprise majoring and kompsa.

0:14:22.760 --> 0:14:25.120
<v Speaker 2>You mean yeah, because AI is going to do all

0:14:25.120 --> 0:14:28.840
<v Speaker 2>these jobs. Because we've had these cycles every five years,

0:14:28.840 --> 0:14:32.000
<v Speaker 2>we have this cycle there, this is nothing anyways.

0:14:31.880 --> 0:14:34.600
<v Speaker 1>No, no, I'm and the Atlantic has just done a

0:14:34.640 --> 0:14:37.520
<v Speaker 1>piss poor job with this stuff. I'm I'm heighten and

0:14:37.840 --> 0:14:41.560
<v Speaker 1>I'm heighten on them because they had an awful claud

0:14:41.600 --> 0:14:43.640
<v Speaker 1>code thing. They just you know, I'm gonna bring it

0:14:43.680 --> 0:14:45.320
<v Speaker 1>up and read the title. I'm not gonna say the

0:14:45.360 --> 0:14:48.760
<v Speaker 1>reporter's name because I think that that's that's main. But

0:14:48.840 --> 0:14:51.720
<v Speaker 1>let's look at this move over chat GPT. You're about

0:14:51.760 --> 0:14:53.680
<v Speaker 1>to hear a lot more about Claude coding. Know why

0:14:53.680 --> 0:14:55.960
<v Speaker 1>you're gonna hear about that? The fucking Atlantic is writing

0:14:56.000 --> 0:14:59.520
<v Speaker 1>about it, and it's just over the whole days. Alex

0:14:59.560 --> 0:15:02.160
<v Speaker 1>lieberm And had an idea what if he could create

0:15:02.200 --> 0:15:04.840
<v Speaker 1>Spotify rapped for his text messages without writing a single

0:15:04.880 --> 0:15:07.360
<v Speaker 1>line of code. Liberman, co founder of the media outlet

0:15:07.360 --> 0:15:09.920
<v Speaker 1>morning Brew, created I message wrapped. I just want to

0:15:09.960 --> 0:15:12.320
<v Speaker 1>start here and say that guy doesn't do any fucking work,

0:15:12.360 --> 0:15:14.720
<v Speaker 1>and I know people who work there like I just

0:15:14.720 --> 0:15:17.400
<v Speaker 1>want to start by saying, if that's the best you've got,

0:15:17.880 --> 0:15:20.440
<v Speaker 1>and that probably involves throwing a giraffe and the entire

0:15:20.480 --> 0:15:23.880
<v Speaker 1>zoo into a vat of acid to make the GPUs move.

0:15:25.360 --> 0:15:28.200
<v Speaker 1>You're meant to read this, and the deliberate effect you're

0:15:28.200 --> 0:15:31.120
<v Speaker 1>meant to have is you're meant to be scared and

0:15:31.200 --> 0:15:33.880
<v Speaker 1>excited like it is a kind of a mishmash between

0:15:33.920 --> 0:15:38.040
<v Speaker 1>the vibe reporting and the I guess it's the folksting.

0:15:38.120 --> 0:15:40.960
<v Speaker 1>In fact, this might be a triple score because this

0:15:41.000 --> 0:15:43.479
<v Speaker 1>is meant to feel you, make you feel with discomfort,

0:15:43.560 --> 0:15:46.240
<v Speaker 1>but also make you freak up but also be excited,

0:15:46.240 --> 0:15:49.880
<v Speaker 1>a rare triple score. It's just stories like these pissed

0:15:49.920 --> 0:15:52.440
<v Speaker 1>me off because I'm fine with people going this could

0:15:52.520 --> 0:15:56.840
<v Speaker 1>do this. I don't mind it happens, But when it's

0:15:56.880 --> 0:16:00.600
<v Speaker 1>just like, hmm, feed me the slop right into my mouth, asshole,

0:16:00.760 --> 0:16:03.720
<v Speaker 1>make me feel, make me feel all the marketing things

0:16:03.760 --> 0:16:08.240
<v Speaker 1>all at once. First of all, I'm not impressed by

0:16:08.240 --> 0:16:11.000
<v Speaker 1>the idea of doing I message s grapped. That's not

0:16:11.320 --> 0:16:13.520
<v Speaker 1>That's not a thing a human being with that with

0:16:13.640 --> 0:16:18.760
<v Speaker 1>like friends and hobbies does. I've been I've been watching

0:16:18.760 --> 0:16:21.240
<v Speaker 1>the first season The True Detective. I've got many more

0:16:21.280 --> 0:16:24.280
<v Speaker 1>shows I could watch. Never once would I think what

0:16:24.360 --> 0:16:26.600
<v Speaker 1>if I could get it wrapped of all my messages?

0:16:26.720 --> 0:16:29.880
<v Speaker 1>What a fucking psychotic thing to do. But when you

0:16:29.920 --> 0:16:33.520
<v Speaker 1>read this, it's like yours. You spent your holidays with

0:16:33.560 --> 0:16:36.560
<v Speaker 1>your family, wrote one tech Polticy expert. That's nice. I

0:16:36.600 --> 0:16:38.400
<v Speaker 1>spent my holidays with Claude code.

0:16:38.800 --> 0:16:39.520
<v Speaker 2>Well, it's fun.

0:16:40.760 --> 0:16:41.280
<v Speaker 1>That's cool.

0:16:41.640 --> 0:16:45.320
<v Speaker 2>It's fun to create, uh these sort of demo apps

0:16:45.440 --> 0:16:48.120
<v Speaker 2>if you're like engineering minded, So in that sense, they're

0:16:48.120 --> 0:16:51.840
<v Speaker 2>like most Yeah, the most cynical analysis here is that

0:16:52.640 --> 0:16:56.960
<v Speaker 2>cloud code is like model trains for engineering minded people.

0:16:57.040 --> 0:16:59.040
<v Speaker 2>It's a fun hobby I can put. Look at this

0:16:59.200 --> 0:17:02.120
<v Speaker 2>like I made a thing that can Like I was

0:17:02.120 --> 0:17:04.600
<v Speaker 2>talking to a friend of mine yesterday, computer scientist, and

0:17:04.640 --> 0:17:06.360
<v Speaker 2>you're like, oh, I built a thing where I can,

0:17:06.440 --> 0:17:10.639
<v Speaker 2>you know whatever, email an appointment to a thing that

0:17:11.000 --> 0:17:13.840
<v Speaker 2>gets parsed by an LM and then goes to my calendar.

0:17:14.600 --> 0:17:16.520
<v Speaker 2>That's just fun for him, in the same way that

0:17:16.560 --> 0:17:18.600
<v Speaker 2>someone else might be like, hey, I built a cabinet

0:17:18.800 --> 0:17:20.560
<v Speaker 2>and like it's really nice, Like look, I got the

0:17:20.560 --> 0:17:22.440
<v Speaker 2>wood to go together, and like I'm proud. I could

0:17:22.480 --> 0:17:23.760
<v Speaker 2>just on a cabinet or whatever.

0:17:23.560 --> 0:17:27.800
<v Speaker 1>Except you build, Except you built something with your hands,

0:17:28.280 --> 0:17:30.560
<v Speaker 1>and a cabinet can have things put in it. It's

0:17:30.880 --> 0:17:35.280
<v Speaker 1>I just it feels like Lego almost, Like it's more

0:17:35.320 --> 0:17:39.639
<v Speaker 1>like it's toy software that has some functionality. Because the

0:17:39.680 --> 0:17:42.360
<v Speaker 1>big thing I'm waiting for with the vibe coding stuff

0:17:42.440 --> 0:17:45.920
<v Speaker 1>is an actual product, you know what I mean.

0:17:46.200 --> 0:17:48.400
<v Speaker 2>But you know that doesn't go well. I mean, look,

0:17:48.440 --> 0:17:50.840
<v Speaker 2>this is the story of This is the story of maltbook,

0:17:51.280 --> 0:17:52.040
<v Speaker 2>right because.

0:17:52.000 --> 0:17:54.280
<v Speaker 1>Oh yeah, tell me more about moltbook. We were just

0:17:54.320 --> 0:17:54.800
<v Speaker 1>talked about this.

0:17:55.560 --> 0:17:57.000
<v Speaker 2>The whole can of worms to open there, I'm just

0:17:57.040 --> 0:17:58.760
<v Speaker 2>going to open like the top of the nearest can,

0:17:58.920 --> 0:18:00.800
<v Speaker 2>which is before even getting the what molt book is

0:18:00.840 --> 0:18:02.960
<v Speaker 2>and is not. You know, it was in the news

0:18:02.960 --> 0:18:06.960
<v Speaker 2>all the time. It was vibe coded and immediately was

0:18:07.040 --> 0:18:10.480
<v Speaker 2>just full of terrible security holes because it was vibe coded.

0:18:10.520 --> 0:18:13.040
<v Speaker 2>And it turned out that you could get the API key,

0:18:13.160 --> 0:18:16.760
<v Speaker 2>so the key you use when you access the paid

0:18:16.800 --> 0:18:19.520
<v Speaker 2>service to get used in LLLM, you have an API key,

0:18:19.520 --> 0:18:21.120
<v Speaker 2>so they know who to charge. You could just steal

0:18:21.160 --> 0:18:23.160
<v Speaker 2>everyone's API key who was using it because the guy

0:18:23.280 --> 0:18:25.720
<v Speaker 2>just vibe coded it and so no one was actually

0:18:25.720 --> 0:18:28.439
<v Speaker 2>there looking at the code. But what but what I

0:18:28.440 --> 0:18:30.200
<v Speaker 2>think is going on? Okay, So here's what I would

0:18:30.200 --> 0:18:31.920
<v Speaker 2>like to see. I would like to see more reporting.

0:18:32.080 --> 0:18:35.119
<v Speaker 2>That would be how are people using X Like to me,

0:18:35.160 --> 0:18:36.880
<v Speaker 2>that's very interesting, yeah, right, how are.

0:18:36.760 --> 0:18:37.720
<v Speaker 1>People using agree?

0:18:37.960 --> 0:18:41.000
<v Speaker 2>Yeah? And the problem is those answers right now. And

0:18:41.040 --> 0:18:43.159
<v Speaker 2>this is confounding. I think that people who are very

0:18:43.160 --> 0:18:46.960
<v Speaker 2>excited by the potential and coolness of these things in isolation,

0:18:47.480 --> 0:18:49.760
<v Speaker 2>the answer to how are people using X is often

0:18:49.800 --> 0:18:52.760
<v Speaker 2>not nearly as exciting as you would guess. And I

0:18:52.760 --> 0:18:54.879
<v Speaker 2>think the reality with I mean, I go quite have

0:18:54.920 --> 0:18:56.600
<v Speaker 2>my arms around cloud code. I do know there's a

0:18:56.600 --> 0:18:58.600
<v Speaker 2>lot of people who are building kind of like internal

0:18:58.640 --> 0:19:01.600
<v Speaker 2>tools or personal tools with it, which I think is cool.

0:19:02.280 --> 0:19:04.600
<v Speaker 2>Most people aren't interested in that, but for some people

0:19:04.640 --> 0:19:07.040
<v Speaker 2>they are. And you know, I think it's fun. Computer

0:19:07.080 --> 0:19:09.360
<v Speaker 2>programming is fun, right, so like the ability to make

0:19:09.359 --> 0:19:11.840
<v Speaker 2>a program work is and some of those tools are useful.

0:19:11.840 --> 0:19:13.800
<v Speaker 2>But this is that's not a major industry on its own.

0:19:13.840 --> 0:19:16.160
<v Speaker 2>You know, I'm designing a tool for my small company

0:19:16.200 --> 0:19:19.000
<v Speaker 2>that makes it easier for us to do whatever. That's cool,

0:19:19.040 --> 0:19:21.880
<v Speaker 2>but that's not like a trillion dollar industry. I can't

0:19:21.880 --> 0:19:25.119
<v Speaker 2>get my arms around yet exactly how professional computer programmers

0:19:25.119 --> 0:19:27.480
<v Speaker 2>are using it. They're all talking about it, really yeah,

0:19:27.800 --> 0:19:30.240
<v Speaker 2>but I can't get I can't get my arms around

0:19:30.240 --> 0:19:32.760
<v Speaker 2>it yet. Yeah, what do you what's what's your sense of.

0:19:34.320 --> 0:19:36.840
<v Speaker 1>I'm gonna be honest, I was going to ask you.

0:19:37.160 --> 0:19:39.520
<v Speaker 1>I was literally going to ask have you heard people

0:19:39.640 --> 0:19:42.480
<v Speaker 1>using it? Because if you go on X, the Everything

0:19:42.480 --> 0:19:47.520
<v Speaker 1>app if you put on pulhasmat suit, you go on

0:19:47.800 --> 0:19:49.159
<v Speaker 1>X and you go and look. And the way that

0:19:49.200 --> 0:19:53.320
<v Speaker 1>people talk about this is like they have connected into

0:19:53.359 --> 0:19:57.040
<v Speaker 1>the matrix that they are now a thousand X engineer.

0:19:57.240 --> 0:19:58.560
<v Speaker 1>But when you go and look at what they do,

0:19:58.600 --> 0:20:01.840
<v Speaker 1>no one actually says. There's always these these kind of

0:20:01.920 --> 0:20:04.800
<v Speaker 1>vibe store, the vibe tales, the mythology where it's like

0:20:05.280 --> 0:20:08.919
<v Speaker 1>I had a problem vaguely that would have taken me

0:20:09.480 --> 0:20:11.960
<v Speaker 1>X number of hours, but when I used claud code,

0:20:12.040 --> 0:20:14.439
<v Speaker 1>it solved it immediately and caught two bugs that I

0:20:14.520 --> 0:20:18.520
<v Speaker 1>didn't know about. It's very Marine Todd's style. Then everybody

0:20:18.600 --> 0:20:23.480
<v Speaker 1>clapped yeah, and it's you were a computer science teacher

0:20:23.720 --> 0:20:26.400
<v Speaker 1>and you you don't know either, which makes me think

0:20:26.440 --> 0:20:29.639
<v Speaker 1>it's not as big a deal as people are saying.

0:20:29.920 --> 0:20:32.919
<v Speaker 1>My other bit of evidence is that at the end

0:20:32.960 --> 0:20:36.160
<v Speaker 1>of last year, Anthropics claud Code revenue was one point

0:20:36.200 --> 0:20:40.200
<v Speaker 1>one billion annualized, so about ninety to one hundred million

0:20:40.240 --> 0:20:46.120
<v Speaker 1>a month for a revolution. That feels low somehow.

0:20:57.480 --> 0:20:59.240
<v Speaker 2>Yeah, I mean what I know as a computer science

0:20:59.320 --> 0:21:03.400
<v Speaker 2>you can't rate performance oriented code. You can't write safe code.

0:21:04.480 --> 0:21:07.960
<v Speaker 2>You can't write code it has to sort of juggle

0:21:08.080 --> 0:21:10.359
<v Speaker 2>sort of complex out of scenarios. I mean, you just

0:21:10.440 --> 0:21:13.560
<v Speaker 2>need good programmers' eyes on it building this code.

0:21:13.600 --> 0:21:15.600
<v Speaker 1>Why is that? Is there a way of explaining to

0:21:15.680 --> 0:21:17.000
<v Speaker 1>a known code of why.

0:21:16.840 --> 0:21:20.520
<v Speaker 2>That is code? There's like a poetic element to it,

0:21:20.960 --> 0:21:24.480
<v Speaker 2>you know, it's Writing good computer code is difficult. You're

0:21:24.480 --> 0:21:28.200
<v Speaker 2>often you're dealing with, you know, what is my problem here?

0:21:28.240 --> 0:21:30.320
<v Speaker 2>And I want an elegant way of sort of satisfying

0:21:30.320 --> 0:21:34.680
<v Speaker 2>this problem. You're often drawing from pretty nuanced algorithms and

0:21:34.760 --> 0:21:37.159
<v Speaker 2>data structures to try to figure out how am I

0:21:37.200 --> 0:21:41.280
<v Speaker 2>going to organize information and efficiently access it? When performance

0:21:41.320 --> 0:21:43.360
<v Speaker 2>comes into play, there's a lot of really subtle decisions

0:21:43.400 --> 0:21:45.280
<v Speaker 2>to make about, you know, how am I going to

0:21:45.320 --> 0:21:47.040
<v Speaker 2>store use things in such a way that we don't

0:21:47.040 --> 0:21:49.840
<v Speaker 2>get bogged down when we're trying to execute things. It

0:21:49.920 --> 0:21:52.280
<v Speaker 2>just becomes I'm I don't code as much anymore. Exo'm

0:21:52.280 --> 0:21:54.000
<v Speaker 2>a theoretician, but I did my whole life since I

0:21:54.119 --> 0:21:57.160
<v Speaker 2>was you know, seven, And it's a there's a it's

0:21:57.160 --> 0:22:00.160
<v Speaker 2>an art form. Right When you're using cloud code, you're

0:22:00.160 --> 0:22:01.720
<v Speaker 2>not really supposed to look at the code, and so

0:22:01.760 --> 0:22:03.600
<v Speaker 2>I think that takes a lot of uses probably off

0:22:03.640 --> 0:22:05.800
<v Speaker 2>the table. It's like the use cases. You're supposed to

0:22:05.840 --> 0:22:08.320
<v Speaker 2>have these different instances of cloud code running, and this

0:22:08.400 --> 0:22:10.239
<v Speaker 2>one's going to write code, and then this one's going

0:22:10.280 --> 0:22:12.720
<v Speaker 2>to write test for that code, and then the cloud

0:22:12.760 --> 0:22:14.440
<v Speaker 2>code is going to run the test and then try

0:22:14.440 --> 0:22:16.480
<v Speaker 2>to fix the code if it doesn't matches the test.

0:22:16.960 --> 0:22:19.480
<v Speaker 2>But your eyes aren't on the code at all. And

0:22:19.520 --> 0:22:21.440
<v Speaker 2>there's I mean, obviously for a lot of programming, that's

0:22:21.440 --> 0:22:23.040
<v Speaker 2>an issue. But then the other thing I've heard about

0:22:23.040 --> 0:22:26.080
<v Speaker 2>the computer programming industry is it's very stratified. There's a

0:22:26.080 --> 0:22:29.040
<v Speaker 2>smaller number of like really good, serious programmers that produce

0:22:29.240 --> 0:22:33.280
<v Speaker 2>like ninety percent of the really important, valuable code on

0:22:33.320 --> 0:22:35.960
<v Speaker 2>which everything runs. I don't think they would touch cloud

0:22:36.000 --> 0:22:38.159
<v Speaker 2>code with a ten foot pole, like they're good at

0:22:38.160 --> 0:22:40.120
<v Speaker 2>what they do. And then you have these like huge

0:22:40.160 --> 0:22:44.840
<v Speaker 2>strata of people writing like JavaScript and sort of hacking

0:22:44.840 --> 0:22:47.719
<v Speaker 2>together Python, and you know it's like not very good code.

0:22:47.800 --> 0:22:49.760
<v Speaker 2>And then it's I guess you could replace some of

0:22:49.760 --> 0:22:52.399
<v Speaker 2>the functional it's functional enough, but I can Yeah, I

0:22:52.400 --> 0:22:53.840
<v Speaker 2>can't get my arms around it yet. But I do

0:22:53.840 --> 0:22:55.040
<v Speaker 2>get you know, I have a lot of sources, so

0:22:55.160 --> 0:22:57.240
<v Speaker 2>I hear from you know, I do hear from people

0:22:57.320 --> 0:23:00.520
<v Speaker 2>that are talking about how cool cloud code is. Is cool.

0:23:00.640 --> 0:23:02.720
<v Speaker 2>I hear from a lot of professional programmers that are like, yeah,

0:23:02.720 --> 0:23:04.840
<v Speaker 2>we don't use we can't use this. We're trying to

0:23:04.840 --> 0:23:07.159
<v Speaker 2>write serious programs, like we have to sell this software,

0:23:07.280 --> 0:23:09.879
<v Speaker 2>like this is this is not solving a problem we

0:23:09.960 --> 0:23:13.240
<v Speaker 2>have or you know we so you know, I don't know,

0:23:13.280 --> 0:23:15.400
<v Speaker 2>But the problem is that that's what the reporting should be. Hey,

0:23:15.480 --> 0:23:19.400
<v Speaker 2>here we are at this company looking over the shoulder

0:23:19.440 --> 0:23:22.159
<v Speaker 2>of people. What are they doing. Let's talk to the

0:23:22.200 --> 0:23:25.480
<v Speaker 2>engineering teams on background of this tech company exactly what

0:23:25.680 --> 0:23:26.960
<v Speaker 2>role is going on here? And I think what a

0:23:27.000 --> 0:23:30.480
<v Speaker 2>lot of reporting has become on AI is your hype laundering.

0:23:30.640 --> 0:23:35.000
<v Speaker 2>So you look at the discussion about the technology happening

0:23:35.000 --> 0:23:37.680
<v Speaker 2>from more engineering minded people, you convince yourself as a reporter,

0:23:37.800 --> 0:23:39.919
<v Speaker 2>I can't understand the engineering, but I will trust the

0:23:39.920 --> 0:23:42.639
<v Speaker 2>people who do, and then you launder what you're sensing

0:23:42.680 --> 0:23:46.119
<v Speaker 2>from that hype into your articles, not realizing that like

0:23:46.240 --> 0:23:48.960
<v Speaker 2>nerds like me, we get hyped up about stuff, and

0:23:49.000 --> 0:23:51.520
<v Speaker 2>we get super excited about stuff and we go create like, yeah,

0:23:51.560 --> 0:23:54.960
<v Speaker 2>you can't just launder our hype into this is what's happening.

0:23:55.320 --> 0:23:58.199
<v Speaker 2>And so it's like reporting on a war where you

0:23:58.280 --> 0:24:01.000
<v Speaker 2>have no one embedded, there's no one actually on the

0:24:01.080 --> 0:24:05.080
<v Speaker 2>ground where the battles are happening. You're just responding to

0:24:05.119 --> 0:24:07.800
<v Speaker 2>the press conferences that the generals are holding, you know,

0:24:08.040 --> 0:24:10.000
<v Speaker 2>back in the Pentagon. Like it's not a way to

0:24:10.000 --> 0:24:11.280
<v Speaker 2>report on what's actually going on.

0:24:12.200 --> 0:24:14.280
<v Speaker 1>And I think the other thing as well is there

0:24:14.359 --> 0:24:17.560
<v Speaker 1>is a if you don't do this hype laundering. I

0:24:18.480 --> 0:24:20.480
<v Speaker 1>don't know how these If you're a report listening to

0:24:20.520 --> 0:24:22.040
<v Speaker 1>this and you have a thought about this, send it

0:24:22.040 --> 0:24:25.440
<v Speaker 1>to me anonymously. Is it trod Thatt seventy six on signal.

0:24:25.840 --> 0:24:28.240
<v Speaker 1>But my thought is as well, is there's probably a

0:24:28.280 --> 0:24:30.879
<v Speaker 1>problem with rationalization as well, because if you look at

0:24:30.920 --> 0:24:34.879
<v Speaker 1>this and you say, Okay, well it's kind of cool,

0:24:34.920 --> 0:24:38.320
<v Speaker 1>it's fun in whatever indeterminate way, it doesn't seem like

0:24:38.840 --> 0:24:43.320
<v Speaker 1>serious software, like actual real deal software is being made

0:24:43.320 --> 0:24:45.760
<v Speaker 1>with it. But then the CEO of Google says thirty

0:24:45.760 --> 0:24:48.760
<v Speaker 1>percent of code is written in AI, which is bullshit.

0:24:49.119 --> 0:24:52.480
<v Speaker 1>And I've heard from so many software engineers, well, it

0:24:52.520 --> 0:24:55.679
<v Speaker 1>couldn't be that everyone's just wrong, right, It couldn't be

0:24:55.680 --> 0:24:58.600
<v Speaker 1>a case of that everyone is making the most egregious

0:24:58.640 --> 0:25:02.080
<v Speaker 1>capital expenditure fuck up of all time. This will be historic,

0:25:02.400 --> 0:25:06.640
<v Speaker 1>I believe, worse than railways, digital beanie babies, but done

0:25:06.680 --> 0:25:09.560
<v Speaker 1>at the scale of laser tag arenas. Now, that can't

0:25:09.560 --> 0:25:11.919
<v Speaker 1>possibly be that, because everyone else is saying this is

0:25:11.960 --> 0:25:16.040
<v Speaker 1>exciting and good. And at that point they choose instead

0:25:16.040 --> 0:25:20.160
<v Speaker 1>of being like worried, instead of being a bit anxious

0:25:20.200 --> 0:25:23.359
<v Speaker 1>about this, they say, well, Amazon Web Services spent a

0:25:23.400 --> 0:25:26.080
<v Speaker 1>lot of money, so this spends a lot of money

0:25:26.240 --> 0:25:31.400
<v Speaker 1>money too, so this is actually good. It's actually good.

0:25:31.400 --> 0:25:34.040
<v Speaker 1>And indeed these people seem emphatic and excited, and I,

0:25:34.160 --> 0:25:38.199
<v Speaker 1>as a non coder, can build a fudged CRM that

0:25:38.359 --> 0:25:42.480
<v Speaker 1>probably would not withstand even the laziest haacker. I can

0:25:42.520 --> 0:25:44.960
<v Speaker 1>do this, and thus it will extrapolate further from there.

0:25:45.400 --> 0:25:47.760
<v Speaker 1>And what sucks is what the people that I believe

0:25:47.800 --> 0:25:52.000
<v Speaker 1>actually will be hurt by this are retail investors. I think,

0:25:52.119 --> 0:25:55.240
<v Speaker 1>regular people buying stocks in these companies or selling software

0:25:55.240 --> 0:25:58.520
<v Speaker 1>stocks because they believe that color code will replace them.

0:25:58.640 --> 0:26:01.120
<v Speaker 1>And altomately, I think it's just gonna be a blood bath.

0:26:01.480 --> 0:26:03.720
<v Speaker 1>For people's furrow. One case that could have been avoided,

0:26:03.840 --> 0:26:07.960
<v Speaker 1>except it would require reports to do something uncomfortable, and

0:26:08.000 --> 0:26:12.080
<v Speaker 1>I don't think they want to do that. Ema.

0:26:12.760 --> 0:26:14.480
<v Speaker 2>I think there's two things going on. This is what

0:26:14.480 --> 0:26:18.400
<v Speaker 2>I've decided with reporting. First of all, it's an asymmetric risk.

0:26:18.960 --> 0:26:20.600
<v Speaker 2>So a lot of reporters are like, look, there's not

0:26:20.640 --> 0:26:24.040
<v Speaker 2>a major risk if I'm excited about this and it

0:26:24.080 --> 0:26:27.080
<v Speaker 2>doesn't pan out, because we could be like, yeah, surprisingly

0:26:27.160 --> 0:26:29.400
<v Speaker 2>this didn't pan out and with some factors we couldn't see.

0:26:29.400 --> 0:26:30.840
<v Speaker 2>But they do feel like there's a huge amount of

0:26:30.920 --> 0:26:33.359
<v Speaker 2>risk of saying this is not a big deal and

0:26:33.359 --> 0:26:36.040
<v Speaker 2>then it is, and so it's definitely an asymmetric risk.

0:26:36.080 --> 0:26:38.000
<v Speaker 2>We saw a lot of this during like COVID as well.

0:26:38.680 --> 0:26:41.520
<v Speaker 2>Right is like you wanted it was less There would

0:26:41.520 --> 0:26:45.479
<v Speaker 2>be less harm if you were too alarmist about something,

0:26:45.560 --> 0:26:47.480
<v Speaker 2>but there could be a lot of harm. They felt

0:26:47.560 --> 0:26:49.680
<v Speaker 2>like reputationally, I feel like this is not a big deal,

0:26:49.680 --> 0:26:52.440
<v Speaker 2>and it was, and so there's definitely an asymmetric risk profile.

0:26:52.680 --> 0:26:56.040
<v Speaker 2>There's also like a meaning defining profile. It's just really

0:26:56.080 --> 0:26:59.480
<v Speaker 2>exciting to think everything is going to be disrupted in

0:26:59.600 --> 0:27:04.639
<v Speaker 2>chain unrecognizably. It sort of gives a focus and meaning

0:27:04.760 --> 0:27:09.439
<v Speaker 2>to like an otherwise somewhat chaotic and disrupted world that

0:27:09.480 --> 0:27:12.560
<v Speaker 2>we're in right now. And so there's that aspect too,

0:27:12.800 --> 0:27:17.080
<v Speaker 2>is that people want to believe there is something massive

0:27:17.119 --> 0:27:19.920
<v Speaker 2>about to happen because in some sense it wipes away

0:27:19.960 --> 0:27:22.000
<v Speaker 2>all the bad stuff that's happening. Who cares, none of

0:27:22.000 --> 0:27:25.240
<v Speaker 2>this is relevant because this much bigger thing is coming.

0:27:25.280 --> 0:27:27.119
<v Speaker 2>So I think that gets wrapped into it as well.

0:27:27.160 --> 0:27:29.800
<v Speaker 2>I think the economic reporters are more on this because

0:27:29.840 --> 0:27:32.080
<v Speaker 2>like their whole job is to try to I mean,

0:27:32.280 --> 0:27:34.320
<v Speaker 2>they're not on it. I think after your work they're

0:27:34.320 --> 0:27:34.880
<v Speaker 2>on it more.

0:27:34.920 --> 0:27:37.840
<v Speaker 1>But like, oh no, if I agree, I am reading

0:27:37.920 --> 0:27:41.240
<v Speaker 1>Big Series. There's an article in Bloomberg I'm trying to

0:27:41.320 --> 0:27:45.320
<v Speaker 1>shove through archive dot is because they were so main

0:27:45.400 --> 0:27:47.680
<v Speaker 1>and unfit to my friend Steve Burke at Games Next

0:27:47.680 --> 0:27:50.679
<v Speaker 1>as that won't pay them. But it's more shit about

0:27:50.720 --> 0:27:54.600
<v Speaker 1>like the software narrative, the fact that software is being

0:27:54.600 --> 0:27:57.760
<v Speaker 1>disrupted by Claude code. You get the same pallid reporting

0:27:57.960 --> 0:28:01.600
<v Speaker 1>from Bloomberg and even the Financial Time about Anthropic and

0:28:01.640 --> 0:28:03.760
<v Speaker 1>the fd is generally pretty good. Well, they're like, yeah,

0:28:03.800 --> 0:28:07.160
<v Speaker 1>they're just gonna make thirty billion dollars next year. It's

0:28:07.200 --> 0:28:12.399
<v Speaker 1>just fanciful. It's there's the skepticism, the cynicism doesn't exist.

0:28:12.440 --> 0:28:14.359
<v Speaker 1>And I get I agree on that there is a

0:28:14.359 --> 0:28:15.160
<v Speaker 1>bubble reporting.

0:28:15.200 --> 0:28:18.879
<v Speaker 2>The last fall, there was a period where everyone did

0:28:18.920 --> 0:28:21.120
<v Speaker 2>the bubble reporting. After GPG five, you had a two

0:28:21.119 --> 0:28:25.800
<v Speaker 2>month period where every major publication did bubble reportings. But you're, yeah,

0:28:25.800 --> 0:28:27.160
<v Speaker 2>I guess that did kind of die off.

0:28:28.040 --> 0:28:31.880
<v Speaker 1>It died off because they hear one nice thing from

0:28:31.960 --> 0:28:35.320
<v Speaker 1>Jensen Wong and they're like, well, I'm sold.

0:28:35.640 --> 0:28:38.320
<v Speaker 2>Like that jacket that jackets looking pretty sharp. I mean,

0:28:38.440 --> 0:28:41.200
<v Speaker 2>someone who wears that jacket. How could they be wrong?

0:28:41.520 --> 0:28:44.640
<v Speaker 1>Would a man that has a shiny jacket like the

0:28:44.680 --> 0:28:48.480
<v Speaker 1>Gleat singer of Corn wears at concerts? Would he lie?

0:28:48.680 --> 0:28:52.560
<v Speaker 1>And it's just what really bothers me. As far as

0:28:52.600 --> 0:28:56.240
<v Speaker 1>the economic reporting though, is the Oracle, because it's like

0:28:56.320 --> 0:28:59.040
<v Speaker 1>Oracle needs to prepay three hundred billion dollars over five

0:28:59.120 --> 0:29:02.440
<v Speaker 1>years by op AI, a company that, if we're to

0:29:02.480 --> 0:29:06.440
<v Speaker 1>believe reporting, which I do not, they made thirteen billion

0:29:06.480 --> 0:29:09.880
<v Speaker 1>dollars last year in revenue and lost. Sorry they've made

0:29:09.960 --> 0:29:13.120
<v Speaker 1>thirteen billion and lost. They claim nine. I think it's

0:29:13.120 --> 0:29:15.960
<v Speaker 1>probably higher. How are they meant to pay thirty to

0:29:16.000 --> 0:29:19.719
<v Speaker 1>sixty billion dollars a year in a year, And everyone's like, well,

0:29:20.840 --> 0:29:23.520
<v Speaker 1>they'll work it out. How's Oracle meant to build those

0:29:23.600 --> 0:29:25.680
<v Speaker 1>data centers? Those data centers are gonna cost one hundred

0:29:25.680 --> 0:29:29.680
<v Speaker 1>and eighty nine billion dollars. They've raised one hundred billion, Okay,

0:29:29.760 --> 0:29:31.360
<v Speaker 1>how they meant to do that? And everyone's just like, oh,

0:29:31.360 --> 0:29:34.760
<v Speaker 1>they'll work it out. If I wish I could do

0:29:34.800 --> 0:29:39.000
<v Speaker 1>this with the fucking bank, I need a five hundred

0:29:39.000 --> 0:29:41.480
<v Speaker 1>million dollar house, I'll pay for it at some point,

0:29:41.840 --> 0:29:44.040
<v Speaker 1>all right, it's fine, and the news would run articles

0:29:44.080 --> 0:29:47.960
<v Speaker 1>about my genius housing purchases. It's just it's one of

0:29:48.000 --> 0:29:51.720
<v Speaker 1>those things where and you said, well the asymmetric risk

0:29:52.240 --> 0:29:54.920
<v Speaker 1>doesn't exist, it does for some of these reporters because

0:29:54.920 --> 0:29:57.720
<v Speaker 1>I've been saving their their bylines for years. Because I

0:29:57.760 --> 0:29:59.640
<v Speaker 1>actually think that there needs to be some sort of

0:29:59.640 --> 0:30:02.840
<v Speaker 1>reckoning with this, because if you look back, there are

0:30:02.920 --> 0:30:06.200
<v Speaker 1>major financial outlets that did the same, that were literally

0:30:06.480 --> 0:30:10.880
<v Speaker 1>propping up Sam Bankman freed two weeks before FTX collapsed,

0:30:11.240 --> 0:30:15.240
<v Speaker 1>who then went on immediately to cover AI fucking and

0:30:15.320 --> 0:30:19.120
<v Speaker 1>now they're peddling bullshit for anthropic and sorry, I'm kind

0:30:19.120 --> 0:30:21.200
<v Speaker 1>of hating. Well, I guess it's hated season, so I can.

0:30:21.360 --> 0:30:25.600
<v Speaker 1>It just frustrates me because regular people are being scared.

0:30:26.200 --> 0:30:29.840
<v Speaker 1>They're being scared by the kind of astonishment which I

0:30:29.840 --> 0:30:33.280
<v Speaker 1>actually loved the astonishment reporting of like, oh well, open

0:30:33.320 --> 0:30:36.840
<v Speaker 1>clawd has proven open claw has proven that ai agi

0:30:36.960 --> 0:30:39.560
<v Speaker 1>is here, or they've built their own social networks that

0:30:39.560 --> 0:30:46.920
<v Speaker 1>we should be scared, scared insane did Yeah, And it's like,

0:30:47.320 --> 0:30:51.000
<v Speaker 1>I assume you saw that for the listeners. By the way,

0:30:51.120 --> 0:30:54.960
<v Speaker 1>open Claw had this this fucking claude bot whatever it is.

0:30:55.000 --> 0:30:57.840
<v Speaker 1>They had their own social network where the quote unquote

0:30:57.880 --> 0:31:00.640
<v Speaker 1>lms would post, but it turns out that those posts

0:31:00.640 --> 0:31:02.040
<v Speaker 1>would just make by their owners.

0:31:02.320 --> 0:31:04.280
<v Speaker 2>Yeah, and this is just I mean, this is a

0:31:04.280 --> 0:31:04.880
<v Speaker 2>good case study.

0:31:04.920 --> 0:31:05.040
<v Speaker 1>Right.

0:31:05.080 --> 0:31:07.239
<v Speaker 2>This one bothered me because like writer friends I know

0:31:07.440 --> 0:31:09.920
<v Speaker 2>who are not technology related, we're texting this to me.

0:31:10.480 --> 0:31:13.120
<v Speaker 2>They were worried, right, They're texting me these articles like

0:31:13.120 --> 0:31:15.480
<v Speaker 2>this seems bad, Like this seems like something really bad.

0:31:15.520 --> 0:31:18.400
<v Speaker 2>They were really getting the digital ic really strongly off

0:31:18.400 --> 0:31:21.200
<v Speaker 2>of these articles. And I would start reading these articles

0:31:21.200 --> 0:31:24.520
<v Speaker 2>and I say, well, there's no discussion of what is

0:31:24.560 --> 0:31:28.440
<v Speaker 2>the technical breakthrough here? And what are the concrete implications,

0:31:28.480 --> 0:31:31.720
<v Speaker 2>Because it turns out there was zero technical breakthroughs. There

0:31:31.800 --> 0:31:36.760
<v Speaker 2>is no new AI technology connected to open claw, which

0:31:36.840 --> 0:31:40.880
<v Speaker 2>used to be called molt whatever mold or open clawed

0:31:40.960 --> 0:31:43.400
<v Speaker 2>or whatever open claw is, I think where they ended

0:31:43.480 --> 0:31:46.800
<v Speaker 2>up right, which is an open source library or framework

0:31:46.840 --> 0:31:50.320
<v Speaker 2>for building AI agents powered by elims. There's no new

0:31:50.360 --> 0:31:53.360
<v Speaker 2>AI technology involved in this at all. The agents you

0:31:53.360 --> 0:31:56.520
<v Speaker 2>build are just accessing off the shelf elms that we're

0:31:56.520 --> 0:31:58.200
<v Speaker 2>all using for chatbots anyways. You can aim it at

0:31:58.240 --> 0:32:02.640
<v Speaker 2>whatever commercial chatpot you want. There's no new framework for

0:32:02.680 --> 0:32:04.800
<v Speaker 2>how the agents work. It's the same sort of react

0:32:04.800 --> 0:32:06.240
<v Speaker 2>loop that we've been trying for the last two or

0:32:06.240 --> 0:32:08.040
<v Speaker 2>three years, where you just you have a program. It's

0:32:08.080 --> 0:32:11.520
<v Speaker 2>like a bit of Python code that asks an LM, hey,

0:32:11.520 --> 0:32:13.120
<v Speaker 2>here's what I want to do, here's the tools you

0:32:13.120 --> 0:32:15.000
<v Speaker 2>have available. Come up with a plan, and then it

0:32:15.040 --> 0:32:18.160
<v Speaker 2>sends it back a plan, and then the Python code

0:32:18.600 --> 0:32:21.040
<v Speaker 2>takes the first step out of that text and says, okay,

0:32:22.120 --> 0:32:25.160
<v Speaker 2>here's the tools available. What should I do to execute

0:32:25.160 --> 0:32:27.480
<v Speaker 2>this step? And then like the LM will give you

0:32:27.480 --> 0:32:30.880
<v Speaker 2>some steps and then the Python code runs those steps

0:32:30.880 --> 0:32:33.240
<v Speaker 2>and then you just go back and forth right. That's

0:32:33.320 --> 0:32:34.840
<v Speaker 2>like this basic react, which.

0:32:34.640 --> 0:32:37.040
<v Speaker 1>Is exactly how Manus worked as well.

0:32:37.200 --> 0:32:38.440
<v Speaker 2>Everything there's nothing new about this.

0:32:38.840 --> 0:32:42.320
<v Speaker 1>Manus was this AI agent that Facebook might be Meta

0:32:42.440 --> 0:32:45.040
<v Speaker 1>might be acquiring. They literally when you use it, it

0:32:45.120 --> 0:32:47.000
<v Speaker 1>just writes prithon code for every step.

0:32:47.360 --> 0:32:49.400
<v Speaker 2>Well, so it's yeah, so this, I mean, there's nothing

0:32:49.440 --> 0:32:51.560
<v Speaker 2>new here technologically. The only thing that was new about

0:32:51.560 --> 0:32:55.840
<v Speaker 2>openclaw is it was open source, so it made it

0:32:55.880 --> 0:32:58.960
<v Speaker 2>easy for anyone to write bad agents. And the other

0:32:59.000 --> 0:33:01.320
<v Speaker 2>thing that made it interesting to wireheads in San Francisco

0:33:01.440 --> 0:33:03.360
<v Speaker 2>is that the commercial products where people are trying to

0:33:03.400 --> 0:33:07.560
<v Speaker 2>build these companies, they have common sense security, like, well,

0:33:07.560 --> 0:33:10.520
<v Speaker 2>probably if it's just a Python program blindly doing what

0:33:10.560 --> 0:33:12.800
<v Speaker 2>an LLM tells it to do, like it probably shouldn't

0:33:12.800 --> 0:33:15.040
<v Speaker 2>have access to like your credit cards or to your

0:33:15.040 --> 0:33:17.400
<v Speaker 2>hard drive or whatever. At open cloud like you can

0:33:17.440 --> 0:33:19.280
<v Speaker 2>just you can give it access to anything you want

0:33:19.320 --> 0:33:21.640
<v Speaker 2>on your computer, and so you could build really cool

0:33:21.640 --> 0:33:25.160
<v Speaker 2>demos that are also like incredibly insecure and unsafe. And

0:33:25.240 --> 0:33:27.239
<v Speaker 2>so it was fun for hobbyists, but there was no

0:33:27.320 --> 0:33:30.280
<v Speaker 2>new technology there. Nothing was new, So it was just

0:33:30.360 --> 0:33:33.800
<v Speaker 2>a way for other people like hobbyists to build their

0:33:33.840 --> 0:33:37.120
<v Speaker 2>own agents. That were like less safe than the more

0:33:37.160 --> 0:33:39.600
<v Speaker 2>carefully built Like what people don't here's a story people

0:33:39.640 --> 0:33:42.320
<v Speaker 2>don't know is like in the immediate aftermath of chat GPT.

0:33:42.400 --> 0:33:44.720
<v Speaker 2>So we're in early twenty twenty three, right, so chatchepts

0:33:44.760 --> 0:33:47.600
<v Speaker 2>come out. Open AI goes on a road tour of

0:33:47.720 --> 0:33:51.040
<v Speaker 2>major publications, right, and they're to try to Hey, here's

0:33:51.040 --> 0:33:53.040
<v Speaker 2>what's going on. You need to write about this. In

0:33:53.080 --> 0:33:55.160
<v Speaker 2>early twenty twenty three, they're like, the next thing we're

0:33:55.200 --> 0:33:57.120
<v Speaker 2>going to offer is something like these agents. They call

0:33:57.160 --> 0:34:00.960
<v Speaker 2>them plugins, and you can install plugins that basically can

0:34:02.280 --> 0:34:05.760
<v Speaker 2>do actions on behalf of your LLLM queries. You could

0:34:05.760 --> 0:34:08.560
<v Speaker 2>have like a book an airline flight plug in and say, hey,

0:34:08.640 --> 0:34:11.560
<v Speaker 2>chatchipt book me a flight, and a language model can't

0:34:11.560 --> 0:34:13.359
<v Speaker 2>do anything but produce text, but then the plugin could

0:34:13.400 --> 0:34:15.479
<v Speaker 2>take the text and go and book you a flight.

0:34:15.560 --> 0:34:17.839
<v Speaker 2>And they're like, this is the future, obviously, and that

0:34:17.880 --> 0:34:22.400
<v Speaker 2>project disappeared because oh, it's incredibly unsafe and unreliable to

0:34:22.760 --> 0:34:25.279
<v Speaker 2>have code that could interact with the real world that's

0:34:25.640 --> 0:34:28.840
<v Speaker 2>following commands from an unreliable who's stayed an LLM like

0:34:28.880 --> 0:34:32.120
<v Speaker 2>that went away not because the technology was hard, but

0:34:32.160 --> 0:34:35.480
<v Speaker 2>because it's not safe. So there's nothing new. My whole

0:34:35.600 --> 0:34:37.959
<v Speaker 2>article on agents, they tried this all at twenty twenty five,

0:34:38.400 --> 0:34:40.360
<v Speaker 2>and we're failing to get this type of agent to

0:34:40.400 --> 0:34:43.200
<v Speaker 2>work for anything outside of basically like producing computer code.

0:34:43.320 --> 0:34:45.960
<v Speaker 2>So open clause is nothing new. It was just a

0:34:46.000 --> 0:34:48.000
<v Speaker 2>way for other people to build these things. The only

0:34:48.040 --> 0:34:51.600
<v Speaker 2>interesting stories were security whole stories. But it was reported. Man,

0:34:51.640 --> 0:34:53.839
<v Speaker 2>I was listening to the All In podcast and they

0:34:53.840 --> 0:34:57.799
<v Speaker 2>were tasing why this is the future of AI, Like

0:34:57.840 --> 0:35:00.919
<v Speaker 2>this is it. Everyone is going to have of an

0:35:00.960 --> 0:35:04.400
<v Speaker 2>open claw agent as like a personal assistant, and this

0:35:04.560 --> 0:35:06.680
<v Speaker 2>is like the biggest thing in open AI. And I

0:35:06.719 --> 0:35:08.879
<v Speaker 2>don't know who it was. One of those guys was like, yeah,

0:35:08.880 --> 0:35:13.360
<v Speaker 2>we replaced our podcast producer, you know, with this agent

0:35:13.360 --> 0:35:15.479
<v Speaker 2>who can like email guests on our behalf and book

0:35:15.480 --> 0:35:17.080
<v Speaker 2>it on our calendar. And well it's costing about a

0:35:17.120 --> 0:35:19.400
<v Speaker 2>thousand dollars a day right now to run it. But like,

0:35:19.880 --> 0:35:22.560
<v Speaker 2>I don't know why we're going to need employees in

0:35:22.600 --> 0:35:23.040
<v Speaker 2>the future.

0:35:23.120 --> 0:35:24.520
<v Speaker 1>Oh yeah brother, Yeah.

0:35:25.040 --> 0:35:27.160
<v Speaker 2>So anyways, like it's a non if you said what's

0:35:27.160 --> 0:35:29.040
<v Speaker 2>the technical implications, not if you say, what are the

0:35:29.040 --> 0:35:31.400
<v Speaker 2>concrete implications? The best story they can have is like

0:35:32.080 --> 0:35:34.680
<v Speaker 2>maybe when people I don't know what it is. All

0:35:34.719 --> 0:35:36.920
<v Speaker 2>the companies will try to build these things for years,

0:35:36.960 --> 0:35:41.120
<v Speaker 2>so I don't know, but it was reported like just

0:35:41.200 --> 0:35:44.719
<v Speaker 2>something icky is happening, and that molt book was an

0:35:44.719 --> 0:35:48.000
<v Speaker 2>application that was built for these agents to communicate on

0:35:48.080 --> 0:35:52.319
<v Speaker 2>like a Reddit style social network. They vibe coded that

0:35:52.400 --> 0:35:54.239
<v Speaker 2>framework and it was full of security holes as we

0:35:54.280 --> 0:35:57.239
<v Speaker 2>mentioned before, and there was like a small number of

0:35:57.320 --> 0:35:59.080
<v Speaker 2>users that created a huge amount of agents, and they

0:35:59.080 --> 0:36:01.160
<v Speaker 2>were just kind of like prompting and prodding their agents

0:36:01.160 --> 0:36:03.600
<v Speaker 2>to produce, like let's talk about creating around church or

0:36:03.640 --> 0:36:06.840
<v Speaker 2>killing humanity. It's like, guys, this you have Python code

0:36:06.880 --> 0:36:10.160
<v Speaker 2>asking lllms to like write text in the style of

0:36:10.200 --> 0:36:13.200
<v Speaker 2>like the matrix, and you're posting it on a fake

0:36:13.280 --> 0:36:17.239
<v Speaker 2>social network. The real story here should be where they

0:36:17.280 --> 0:36:20.000
<v Speaker 2>don't these people have things to do? Don't they have jobs?

0:36:20.040 --> 0:36:20.680
<v Speaker 1>Like what is y? Yeah?

0:36:20.719 --> 0:36:24.040
<v Speaker 2>Yeah, yeah, really brilliant my model train, I added a

0:36:24.040 --> 0:36:26.279
<v Speaker 2>tunnel so that one really got no.

0:36:26.600 --> 0:36:29.160
<v Speaker 1>I actually will push back on that. Mudile train listeners

0:36:29.400 --> 0:36:32.480
<v Speaker 1>I think make up twenty percent of the listenership of

0:36:32.560 --> 0:36:36.719
<v Speaker 1>this podcast course. Also, modile trains are cool. Signaling to

0:36:36.760 --> 0:36:40.040
<v Speaker 1>my fellow autists out there. But you know, mudtile trains

0:36:40.040 --> 0:36:41.880
<v Speaker 1>are a physical thing which you build it, you build

0:36:41.880 --> 0:36:44.880
<v Speaker 1>a little a little city. I think that delightful. Respect

0:36:44.920 --> 0:36:47.200
<v Speaker 1>to those with this. It's like if they and you

0:36:47.239 --> 0:36:48.920
<v Speaker 1>know what, if they were just saying I have been

0:36:49.000 --> 0:36:51.120
<v Speaker 1>fucking around with some software and it did something cool,

0:36:51.239 --> 0:36:53.160
<v Speaker 1>I'd respect the shit out. Then I'd be like, yeah,

0:36:53.320 --> 0:36:56.359
<v Speaker 1>enjoy yourself. It's going away. This is a little subsidized rites,

0:36:56.400 --> 0:36:58.160
<v Speaker 1>but have fun with that. I don't know why you

0:36:58.160 --> 0:37:01.160
<v Speaker 1>need a max studio, but good luck. No, they are like,

0:37:01.360 --> 0:37:04.480
<v Speaker 1>this is the future. But I'm gonna be honest, cal

0:37:04.840 --> 0:37:08.320
<v Speaker 1>My real question is what does openclaw actually do? Because

0:37:08.320 --> 0:37:11.120
<v Speaker 1>when I went and read what it actually I read

0:37:11.200 --> 0:37:14.719
<v Speaker 1>so many posts you said it books callen does and

0:37:14.760 --> 0:37:17.799
<v Speaker 1>does this. I could find no proof that anyone successfully

0:37:17.800 --> 0:37:18.320
<v Speaker 1>did that.

0:37:18.200 --> 0:37:20.800
<v Speaker 2>And it doesn't. It's it's just a series of library

0:37:20.840 --> 0:37:23.800
<v Speaker 2>calls that you can use to build your own program

0:37:23.880 --> 0:37:26.360
<v Speaker 2>in theory that would do that. So open claw itself

0:37:26.480 --> 0:37:28.840
<v Speaker 2>is just like a series of like interfaces and hooks

0:37:28.840 --> 0:37:31.840
<v Speaker 2>that makes it easy to write a program that talks

0:37:31.840 --> 0:37:34.640
<v Speaker 2>to other services basically and makes calls to LLLM. So

0:37:34.680 --> 0:37:37.200
<v Speaker 2>it's just like a wrapper in which you can write

0:37:37.239 --> 0:37:40.160
<v Speaker 2>your own code, and some people are trying, Yeah, like

0:37:40.200 --> 0:37:43.040
<v Speaker 2>you can you can write, you can have it talk

0:37:43.120 --> 0:37:45.080
<v Speaker 2>to your calendar, you can have it talk to your

0:37:45.120 --> 0:37:46.160
<v Speaker 2>email and theory and.

0:37:46.200 --> 0:37:48.239
<v Speaker 1>Look, but does it doesn't work?

0:37:48.360 --> 0:37:50.840
<v Speaker 2>Well, it's it's just asking an lell you can simulate.

0:37:51.040 --> 0:37:52.759
<v Speaker 2>I mean, I did this for my New Yorker piece.

0:37:52.800 --> 0:37:54.720
<v Speaker 2>I was like, look, I can just simulate being an agent.

0:37:54.800 --> 0:37:57.560
<v Speaker 2>Just ask any LLM, here's what I'm trying to do.

0:37:57.640 --> 0:37:59.640
<v Speaker 2>Give me like the steps for doing this or whatever,

0:37:59.800 --> 0:38:02.840
<v Speaker 2>and like anything else you ask of an LOLLM, it

0:38:02.880 --> 0:38:05.480
<v Speaker 2>will give you answers that sound very reasonable in general

0:38:05.520 --> 0:38:07.200
<v Speaker 2>and then have like a lot of issues in the details.

0:38:07.200 --> 0:38:10.520
<v Speaker 2>Which is why agents based on llms have struggled because

0:38:10.560 --> 0:38:12.719
<v Speaker 2>if you it's fine if you as a human are

0:38:12.719 --> 0:38:14.839
<v Speaker 2>talking to an M, because you don't realize how much

0:38:14.840 --> 0:38:17.080
<v Speaker 2>filtering and tweaking. You're like, well, that's kind of ignore

0:38:17.120 --> 0:38:18.600
<v Speaker 2>that and this is good, or let me ask you

0:38:18.640 --> 0:38:20.759
<v Speaker 2>to redo it. It's really an issue if you write

0:38:20.760 --> 0:38:23.759
<v Speaker 2>a program that just says I will do whatever. I'll

0:38:23.800 --> 0:38:25.840
<v Speaker 2>ask the LLM for a plan and then just do

0:38:26.000 --> 0:38:30.440
<v Speaker 2>whatever it says because it doesn't know that doesn't really

0:38:30.520 --> 0:38:33.960
<v Speaker 2>make sen or like this sounds generally generally reasonable, but

0:38:34.040 --> 0:38:36.480
<v Speaker 2>with like issues in the details. Does it work out

0:38:36.560 --> 0:38:39.680
<v Speaker 2>when you execute it? I mean, in a practical sense,

0:38:40.160 --> 0:38:43.160
<v Speaker 2>can it? They said only all in podcasts, three dumb

0:38:43.160 --> 0:38:47.400
<v Speaker 2>bitches saying exactly, that's the main reference. Don't use that

0:38:47.440 --> 0:38:50.960
<v Speaker 2>word usually those people sitting around going blah blah blah.

0:38:51.239 --> 0:38:54.400
<v Speaker 1>Oh we've made it and replaced our podcast producer with

0:38:54.520 --> 0:38:59.760
<v Speaker 1>an LM that can make appointments and send emails in practice.

0:38:59.840 --> 0:39:03.560
<v Speaker 1>Is that true? I severely doubt that. I just I Also,

0:39:03.760 --> 0:39:05.919
<v Speaker 1>you're spending on thousand dollars a day, so you're paying

0:39:05.920 --> 0:39:09.640
<v Speaker 1>three hundred and sixty five thousand dollars a year for this, right?

0:39:11.080 --> 0:39:14.160
<v Speaker 1>Are you really or are you just? Have you? If

0:39:14.200 --> 0:39:17.040
<v Speaker 1>it is it completely it probably isn't. And that's the

0:39:17.120 --> 0:39:20.320
<v Speaker 1>thing I don't I get why they all in guys

0:39:20.360 --> 0:39:23.440
<v Speaker 1>do it because they probably have investments and they're boosters TVPN,

0:39:24.080 --> 0:39:28.239
<v Speaker 1>same fucking deal. It really rules that the two largest

0:39:28.440 --> 0:39:32.240
<v Speaker 1>like tech things in the valley are just state media

0:39:32.360 --> 0:39:36.560
<v Speaker 1>book for Silicon Valley. What gets me is when you

0:39:36.600 --> 0:39:39.759
<v Speaker 1>get like the Atlantic CNBC Business Insider in places like

0:39:39.800 --> 0:39:41.880
<v Speaker 1>that doing stuff like this, and it bothers me because

0:39:42.160 --> 0:39:44.319
<v Speaker 1>they don't even need the hate on it. They could

0:39:44.440 --> 0:39:47.120
<v Speaker 1>just say, yeah, I could do this, it's pretty cool,

0:39:47.200 --> 0:39:51.320
<v Speaker 1>right yeah, But I guess that that doesn't get the clear.

0:39:51.480 --> 0:39:53.640
<v Speaker 2>It's not an interesting story. That's not the real story

0:39:53.719 --> 0:39:55.759
<v Speaker 2>is not always that interesting. It's like like hobbyists are

0:39:55.760 --> 0:39:57.920
<v Speaker 2>building these tools that kind of do cool things but

0:39:58.040 --> 0:40:00.640
<v Speaker 2>make mistakes, and it's a little expensive. Like the most

0:40:00.680 --> 0:40:04.040
<v Speaker 2>interesting story out of open claw is the only thing

0:40:04.080 --> 0:40:05.880
<v Speaker 2>I think is going to be impactful out of it

0:40:05.920 --> 0:40:08.520
<v Speaker 2>is because it was so expensive, right, like that thousand

0:40:08.520 --> 0:40:11.680
<v Speaker 2>dollars a day. What it's forcing people, these hobbyists to

0:40:11.760 --> 0:40:15.440
<v Speaker 2>do is to turn to like cheaper alternatives for models,

0:40:15.880 --> 0:40:18.600
<v Speaker 2>and that I do think is significant, this idea that

0:40:19.040 --> 0:40:23.480
<v Speaker 2>there's open source models out there, they're significantly cheaper than

0:40:23.520 --> 0:40:25.719
<v Speaker 2>trying to use like open ai or claud. More and

0:40:25.719 --> 0:40:27.960
<v Speaker 2>more people are running their own local models because it

0:40:28.000 --> 0:40:30.600
<v Speaker 2>turns out, for you know, most specific uses you might

0:40:30.680 --> 0:40:33.640
<v Speaker 2>use in LLM, you don't need like a super fine

0:40:33.680 --> 0:40:36.120
<v Speaker 2>tune trillion parameter beast of a model running in some

0:40:36.239 --> 0:40:38.440
<v Speaker 2>data center somewhere. It's like, you know what, I'm parsing

0:40:38.520 --> 0:40:42.520
<v Speaker 2>my email that try to extract like suggested times. I'm

0:40:42.600 --> 0:40:45.319
<v Speaker 2>fine with like a twenty billion parameter model that can

0:40:45.400 --> 0:40:48.000
<v Speaker 2>easily fit in a single GPU on a thing I have,

0:40:48.160 --> 0:40:50.600
<v Speaker 2>you know, we share in our office or something like that.

0:40:50.760 --> 0:40:52.680
<v Speaker 2>So to me, that's the most interesting story out of

0:40:52.680 --> 0:40:55.120
<v Speaker 2>open claw is the bad news it could be for

0:40:55.239 --> 0:40:59.440
<v Speaker 2>the big companies as people get more comfortable with we

0:40:59.520 --> 0:41:04.040
<v Speaker 2>don't needed these formula one car versions of language models

0:41:04.080 --> 0:41:07.600
<v Speaker 2>for the stuff we're actually doing. We're fine with the

0:41:07.640 --> 0:41:10.279
<v Speaker 2>forward focus, right, And I think that's a transition that's

0:41:10.320 --> 0:41:12.680
<v Speaker 2>going to That's a transition that to me is more interest.

0:41:12.680 --> 0:41:14.200
<v Speaker 2>That's what we should be writing about. Like, that's an

0:41:14.200 --> 0:41:16.399
<v Speaker 2>interesting idea to me is that you have these huge,

0:41:16.440 --> 0:41:19.240
<v Speaker 2>high valuation companies, but you also have all these open

0:41:19.239 --> 0:41:21.279
<v Speaker 2>source models, like the weights are just out there in

0:41:21.320 --> 0:41:23.319
<v Speaker 2>the public domain that can do ninety eight percent of

0:41:23.320 --> 0:41:25.480
<v Speaker 2>what people care about, and you're beginning to get low

0:41:25.520 --> 0:41:28.239
<v Speaker 2>cost competitive services where people can just spin those up

0:41:28.280 --> 0:41:31.120
<v Speaker 2>in cheaper data centers. That's interesting to me. That's an

0:41:31.120 --> 0:41:34.319
<v Speaker 2>economic story. AI creating its own church is not as

0:41:34.480 --> 0:41:35.400
<v Speaker 2>not as relevant to me.

0:41:50.040 --> 0:41:53.239
<v Speaker 1>I'm already working on a story full this Friday in

0:41:53.320 --> 0:41:56.800
<v Speaker 1>fact oh my premium newsletter about the fact that actually

0:41:56.840 --> 0:42:00.120
<v Speaker 1>the margins of serving GPU compute are dog shit like

0:42:00.160 --> 0:42:02.400
<v Speaker 1>the best of the best of like a co location

0:42:02.600 --> 0:42:05.960
<v Speaker 1>place like Applied Digital, it's like twenty seven percent gross

0:42:06.000 --> 0:42:09.320
<v Speaker 1>margins and that's at one hundred percent utilization. Anything below

0:42:09.320 --> 0:42:12.799
<v Speaker 1>point seven they're burning cash. I hear this date center

0:42:12.800 --> 0:42:14.759
<v Speaker 1>out in North Dakota losing a million dollars a day.

0:42:15.040 --> 0:42:17.919
<v Speaker 1>That's the thing. This is even with these low these

0:42:17.960 --> 0:42:20.680
<v Speaker 1>lower cost models on device could be interesting.

0:42:20.680 --> 0:42:22.600
<v Speaker 2>That's what's going to happen. I think, I think undevice

0:42:22.680 --> 0:42:23.480
<v Speaker 2>is what's going to happen.

0:42:24.080 --> 0:42:26.880
<v Speaker 1>I just remembered something. This is a classic bullshit story

0:42:26.880 --> 0:42:29.239
<v Speaker 1>that I see every so often. So have you read

0:42:29.239 --> 0:42:32.040
<v Speaker 1>any of the stories that are like, yeah, Claude can

0:42:32.080 --> 0:42:35.439
<v Speaker 1>now work for hours uninterrupted? Have you read about these?

0:42:35.719 --> 0:42:36.839
<v Speaker 1>You heard this? You've seen this.

0:42:37.320 --> 0:42:39.720
<v Speaker 2>Work for our Oh yeah, yeah, we were talking about

0:42:39.760 --> 0:42:42.400
<v Speaker 2>the Yeah, the multi step agenda execution.

0:42:42.520 --> 0:42:43.719
<v Speaker 1>Well, no way, it keeps coming back.

0:42:44.080 --> 0:42:45.759
<v Speaker 2>Yeah, yeah.

0:42:45.840 --> 0:42:48.439
<v Speaker 1>AI on the CNBC AI on the verge of eight

0:42:48.480 --> 0:42:51.000
<v Speaker 1>hour job shift without burnout or break is it is

0:42:51.040 --> 0:42:55.440
<v Speaker 1>twenty four hour our AI workday. Next going to just

0:42:55.600 --> 0:42:58.120
<v Speaker 1>censor myself what I was going to say there, because

0:42:58.160 --> 0:43:02.440
<v Speaker 1>it's just like yeah, it can work for hours. Is

0:43:02.480 --> 0:43:03.520
<v Speaker 1>the output good?

0:43:03.760 --> 0:43:05.319
<v Speaker 2>Does it a problem? Yeah, just a lose?

0:43:05.360 --> 0:43:05.719
<v Speaker 1>Does it? Will?

0:43:05.800 --> 0:43:08.200
<v Speaker 2>Just keep reading? Call it? Recall it? Recall it like

0:43:08.239 --> 0:43:11.160
<v Speaker 2>I can. I can write a Python program that calls

0:43:11.160 --> 0:43:12.560
<v Speaker 2>an hour for twenty four hours.

0:43:12.760 --> 0:43:15.640
<v Speaker 1>If you need me to burn something for hours on end,

0:43:15.719 --> 0:43:17.960
<v Speaker 1>I just give me some gasoline. I got your baby.

0:43:18.200 --> 0:43:20.759
<v Speaker 1>I've sort this right out. Be much cheaper, but it's

0:43:20.760 --> 0:43:23.680
<v Speaker 1>like yeah. By September twenty twenty five, Claudes on It

0:43:23.760 --> 0:43:26.000
<v Speaker 1>four point five is reported to run autonomously for up

0:43:26.000 --> 0:43:29.840
<v Speaker 1>to thirty hours, reported with com I mean, but that's

0:43:30.280 --> 0:43:34.000
<v Speaker 1>more vibe reporting. It's you're meant to read that and go, wow,

0:43:34.200 --> 0:43:38.919
<v Speaker 1>this thing is performing tasks that are useful, that execute code,

0:43:38.920 --> 0:43:41.520
<v Speaker 1>that do something, and in that thirty hours that is

0:43:41.560 --> 0:43:45.080
<v Speaker 1>equivalent to thirty human working hours versus they sat there

0:43:45.120 --> 0:43:48.160
<v Speaker 1>and pissed their pants for like twenty nine point five

0:43:48.280 --> 0:43:50.080
<v Speaker 1>of those at least well.

0:43:50.120 --> 0:43:52.719
<v Speaker 2>And also that those are like specialized TSK typically that

0:43:54.480 --> 0:43:56.920
<v Speaker 2>have clear milestones and testing, so it's like it can

0:43:56.960 --> 0:44:00.640
<v Speaker 2>do something. The loop can try and try until it

0:44:00.640 --> 0:44:02.520
<v Speaker 2>passes the test, and then you know that's done, and

0:44:02.520 --> 0:44:03.920
<v Speaker 2>then you can move on to the next step. And

0:44:03.920 --> 0:44:06.280
<v Speaker 2>then you can keep calling to l M and executing

0:44:06.360 --> 0:44:08.200
<v Speaker 2>until it passes the next test and you can move on.

0:44:08.560 --> 0:44:12.720
<v Speaker 2>And in theory like it's yeah, they're avoiding error cascade

0:44:12.719 --> 0:44:14.959
<v Speaker 2>because it's testable and they can keep retrying or something

0:44:15.000 --> 0:44:16.520
<v Speaker 2>like that. I mean, this was like the meter had

0:44:16.560 --> 0:44:19.880
<v Speaker 2>this problem with their graph of how long of tasks

0:44:19.960 --> 0:44:22.719
<v Speaker 2>AI can now do, and it was like the sort

0:44:22.719 --> 0:44:25.440
<v Speaker 2>of like super arbitrary decision of like, oh, here's a

0:44:25.480 --> 0:44:27.719
<v Speaker 2>test that takes a human five hours, Now AI can

0:44:27.760 --> 0:44:30.440
<v Speaker 2>do that. Here's a task, whereas really more about like

0:44:30.600 --> 0:44:33.560
<v Speaker 2>how many things in a row can you do without

0:44:33.600 --> 0:44:36.040
<v Speaker 2>the errors cascading out of control or something like that.

0:44:36.440 --> 0:44:37.400
<v Speaker 2>It's very different.

0:44:38.280 --> 0:44:41.360
<v Speaker 1>You're being way nicer than you should be, in my opinion,

0:44:41.440 --> 0:44:45.799
<v Speaker 1>just because they don't even explain. They don't even say like, yeah,

0:44:45.840 --> 0:44:47.560
<v Speaker 1>and we managed to make it through it and let's

0:44:47.600 --> 0:44:51.480
<v Speaker 1>go bus right hours c NBC just fucking woo just

0:44:51.520 --> 0:44:53.520
<v Speaker 1>for tearing their shirts off and screaming.

0:44:54.080 --> 0:44:56.160
<v Speaker 2>But it goes back to the two things. We need

0:44:56.160 --> 0:44:59.000
<v Speaker 2>this report and you need technical the details of the

0:44:59.080 --> 0:45:02.800
<v Speaker 2>relevant technical innovation, and a discussion of concrete implications in

0:45:02.840 --> 0:45:06.000
<v Speaker 2>your future implications of that breakthrough. That's what I'm always

0:45:06.000 --> 0:45:07.400
<v Speaker 2>looking for, So, like, if you want to cover a

0:45:07.440 --> 0:45:09.600
<v Speaker 2>story like that, like, well, what does that mean? What

0:45:09.760 --> 0:45:12.800
<v Speaker 2>is the technical breakthrough? Right? What does this mean technically

0:45:12.800 --> 0:45:14.720
<v Speaker 2>you can do thirty hours of work or eight hours?

0:45:14.840 --> 0:45:17.360
<v Speaker 2>What is the work? What was changed? What did they figure?

0:45:17.560 --> 0:45:21.279
<v Speaker 2>What was happening before? What technical breakthrough made this possible now?

0:45:21.400 --> 0:45:24.280
<v Speaker 2>And then what are the concrete implications? What specific things now?

0:45:24.600 --> 0:45:26.920
<v Speaker 2>Can we directly this will now allow us to do? What?

0:45:26.960 --> 0:45:28.640
<v Speaker 2>Tell us what jobs are going away? Tell us what

0:45:28.719 --> 0:45:30.520
<v Speaker 2>tool we're going to see, like you have to, but

0:45:31.280 --> 0:45:34.080
<v Speaker 2>we avoid that, right because then you're putting your chips

0:45:34.080 --> 0:45:37.239
<v Speaker 2>down and then those things don't come along, And so

0:45:37.320 --> 0:45:39.400
<v Speaker 2>I'm always looking for that. What's the technical innovation and

0:45:39.400 --> 0:45:41.239
<v Speaker 2>what are the concrete implications? If you don't have that,

0:45:41.280 --> 0:45:43.400
<v Speaker 2>you're mining emotions.

0:45:43.600 --> 0:45:46.279
<v Speaker 1>You're mining emotions or just helping boost stook. That's what

0:45:46.360 --> 0:45:50.400
<v Speaker 1>really bothers me because I hate that they're scaring people.

0:45:50.840 --> 0:45:54.880
<v Speaker 1>I really really hate that they're doing marketing like they're

0:45:54.920 --> 0:45:59.320
<v Speaker 1>just doing like I run a PIAFA. I've dealt with

0:45:59.400 --> 0:46:04.080
<v Speaker 1>ether these states startups for like fifteen years. There isn't

0:46:04.120 --> 0:46:06.879
<v Speaker 1>a single early stage startup that would get a percentage

0:46:06.920 --> 0:46:11.319
<v Speaker 1>point of this bullshit, Like you email the reporter about

0:46:11.360 --> 0:46:14.160
<v Speaker 1>like a series a start. They're like, all right, all right, motherfucker,

0:46:15.120 --> 0:46:18.160
<v Speaker 1>how much revenue are they profitable? Yet? Why not? Why

0:46:18.200 --> 0:46:21.359
<v Speaker 1>not explain to me right now? Anthropics like we're gonna

0:46:21.440 --> 0:46:24.920
<v Speaker 1>gonna burn a hundred billion on training. I guess what

0:46:25.000 --> 0:46:27.600
<v Speaker 1>do you think? And they're like, yeah, I love it.

0:46:27.640 --> 0:46:29.680
<v Speaker 1>I actually think that's huge, that's great. I'm not And

0:46:29.719 --> 0:46:32.200
<v Speaker 1>to be clear, I think that this scrutiny should be

0:46:32.520 --> 0:46:34.920
<v Speaker 1>from the beginning to the end. I think that everyone

0:46:34.920 --> 0:46:38.200
<v Speaker 1>should face this scrutiny. Yeah, I'm not saying that I

0:46:38.200 --> 0:46:41.840
<v Speaker 1>should get an easier I'm saying, actually everyone should. Anthropics

0:46:41.840 --> 0:46:47.160
<v Speaker 1>should face the most brutal scrutiny, and I guess that

0:46:47.239 --> 0:46:50.920
<v Speaker 1>they don't because they want access. It's just very dull

0:46:51.320 --> 0:46:54.480
<v Speaker 1>and annoying, and it's just helping already rich guys like

0:46:54.520 --> 0:46:58.560
<v Speaker 1>Wario Ama Day, who should not have more money. Listen

0:46:58.600 --> 0:47:01.360
<v Speaker 1>to him speak, he needs he needs to face some stress.

0:47:02.520 --> 0:47:05.239
<v Speaker 1>I think some stress would help him grow. But caw,

0:47:05.320 --> 0:47:07.200
<v Speaker 1>as we wrap up, I did actually have like a

0:47:07.480 --> 0:47:10.839
<v Speaker 1>technical thing that wants an idea of being perclaying. So

0:47:11.520 --> 0:47:15.440
<v Speaker 1>I think that this term training with models is being

0:47:15.480 --> 0:47:18.319
<v Speaker 1>misused and used in a way that is kind of

0:47:18.360 --> 0:47:21.880
<v Speaker 1>vibe reporting style, which is they use training as a

0:47:21.920 --> 0:47:24.440
<v Speaker 1>word that suggests that it will stop, that they will

0:47:24.440 --> 0:47:27.160
<v Speaker 1>stop training these models. But from correct me if I'm wrong.

0:47:27.320 --> 0:47:30.640
<v Speaker 1>Training is everything from building a new model to updating

0:47:30.640 --> 0:47:32.480
<v Speaker 1>and models current parameters. Correct.

0:47:32.880 --> 0:47:35.640
<v Speaker 2>Yeah, there's pre training and post training is the right

0:47:35.640 --> 0:47:38.080
<v Speaker 2>way to think about it. So the pre training is unsupervised.

0:47:38.120 --> 0:47:40.120
<v Speaker 2>That's where you take all the text that's ever been written,

0:47:40.600 --> 0:47:42.520
<v Speaker 2>and you will take a real piece of text written

0:47:42.520 --> 0:47:44.360
<v Speaker 2>by a human, and you'll cut it off at an

0:47:44.440 --> 0:47:47.840
<v Speaker 2>arbitrary point, and you'll tell the language model your guess

0:47:47.920 --> 0:47:50.000
<v Speaker 2>the word that came next. There's a real word that

0:47:50.000 --> 0:47:52.359
<v Speaker 2>came next. This is a real writing Guess the word

0:47:52.400 --> 0:47:54.239
<v Speaker 2>that came next, and then it guesses, and then you

0:47:54.520 --> 0:47:56.399
<v Speaker 2>adjust the weight so it gets closer to the right answer.

0:47:56.400 --> 0:47:58.359
<v Speaker 2>That's prehen you adjust the waits. What are you doing,

0:47:58.760 --> 0:48:02.759
<v Speaker 2>You're just so you're a training algorithm called back propagation.

0:48:02.840 --> 0:48:05.399
<v Speaker 2>This is a Jeff Hinton, you know innovation where you're

0:48:05.440 --> 0:48:07.640
<v Speaker 2>you're going through and you're adjusting the weights all the

0:48:07.640 --> 0:48:09.880
<v Speaker 2>way through the layers in such a way that the

0:48:09.960 --> 0:48:12.839
<v Speaker 2>answer it gives for this particular test gets a little

0:48:12.840 --> 0:48:14.400
<v Speaker 2>bit closer to the right answer, because you know the

0:48:14.440 --> 0:48:17.040
<v Speaker 2>right answer that's pre training, that's pre trains unsupervised. So

0:48:17.080 --> 0:48:19.879
<v Speaker 2>you take Hamlet or you take Dickens like the best

0:48:19.880 --> 0:48:21.680
<v Speaker 2>of times it was the worst of and then you

0:48:21.719 --> 0:48:23.840
<v Speaker 2>give that to the thing what were to come next?

0:48:24.360 --> 0:48:26.640
<v Speaker 2>And you know it says Bacon like, we're going to

0:48:26.719 --> 0:48:28.560
<v Speaker 2>adjust these weights now in a way that like your

0:48:28.560 --> 0:48:31.319
<v Speaker 2>answer gets a little bit closer towards times. Right. Okay,

0:48:31.320 --> 0:48:34.360
<v Speaker 2>that's pre training. Then you get post training, where you

0:48:34.400 --> 0:48:36.319
<v Speaker 2>already have trained, so you have this network. All the

0:48:36.320 --> 0:48:38.960
<v Speaker 2>weights have been set through this massive multi month you know,

0:48:39.160 --> 0:48:42.480
<v Speaker 2>billion dollar pre training, and now you want to go

0:48:42.520 --> 0:48:44.800
<v Speaker 2>through and you want to tweak this to avoid certain

0:48:44.840 --> 0:48:47.920
<v Speaker 2>types of behaviors or to influence towards certain types of behaviors.

0:48:47.920 --> 0:48:51.440
<v Speaker 2>So for post training, it's almost always based off of

0:48:51.520 --> 0:48:55.960
<v Speaker 2>you have inputs like prompts and correct answers. This is

0:48:56.000 --> 0:48:58.600
<v Speaker 2>the right way to respond to this question, right, So

0:48:58.760 --> 0:49:02.680
<v Speaker 2>you've you've you have of questions and answers. You give

0:49:02.680 --> 0:49:05.800
<v Speaker 2>the prompt to the LLM, it spits out some answer,

0:49:05.960 --> 0:49:08.480
<v Speaker 2>and now what you're doing you're using is called reinforcement learning.

0:49:08.480 --> 0:49:10.920
<v Speaker 2>It's a general technology, but you're using techniques from reinforcement

0:49:11.000 --> 0:49:12.719
<v Speaker 2>learning the sort of like zap it like you would

0:49:12.760 --> 0:49:16.000
<v Speaker 2>zap a dog when you're dog training it to if

0:49:16.000 --> 0:49:18.239
<v Speaker 2>it's a bad answer, you zap it, so you get

0:49:18.280 --> 0:49:20.040
<v Speaker 2>those weights away from that answer, and if it's a

0:49:20.040 --> 0:49:22.360
<v Speaker 2>good answer, you give it a treat like Okay, this

0:49:22.480 --> 0:49:25.520
<v Speaker 2>is post this is post training. And so that's where

0:49:26.280 --> 0:49:29.640
<v Speaker 2>you've moved past the word guessing game, which is where

0:49:29.640 --> 0:49:32.000
<v Speaker 2>all of the sort of general smarts comes from these models,

0:49:32.040 --> 0:49:35.400
<v Speaker 2>And now you're you're you're doing the sort of zapping

0:49:35.440 --> 0:49:37.960
<v Speaker 2>and treat training around very specific things, right, so this

0:49:38.000 --> 0:49:41.759
<v Speaker 2>is where you keep going. Yeah, So like so you'll

0:49:41.760 --> 0:49:43.840
<v Speaker 2>go through and like ask it questions where the answers

0:49:43.920 --> 0:49:46.600
<v Speaker 2>might be like about building bombs or whatever, and every

0:49:46.640 --> 0:49:48.759
<v Speaker 2>time it spits out an answer about a bomb, you

0:49:48.800 --> 0:49:51.400
<v Speaker 2>give it like a really bad negative shock, and you

0:49:51.520 --> 0:49:54.319
<v Speaker 2>like definitely turn off those circuits, like we don't want

0:49:54.360 --> 0:49:56.000
<v Speaker 2>you to spit out answers by bombs, Like that's where

0:49:56.000 --> 0:49:58.279
<v Speaker 2>all the guardrails come from. Or if you want it

0:49:58.320 --> 0:50:00.640
<v Speaker 2>to get better at doing like a particular math exam,

0:50:00.680 --> 0:50:03.279
<v Speaker 2>you can give it like lots of questions from that

0:50:03.520 --> 0:50:06.040
<v Speaker 2>math exam and then you have the right answer and

0:50:06.040 --> 0:50:07.880
<v Speaker 2>you can kind of zap it to move it towards

0:50:08.239 --> 0:50:10.480
<v Speaker 2>what the answers look like on this math exam. So

0:50:10.480 --> 0:50:14.880
<v Speaker 2>so post training is more focused. You have particular types

0:50:14.920 --> 0:50:17.360
<v Speaker 2>of behavior you're trying to sort of instill in this

0:50:17.480 --> 0:50:21.240
<v Speaker 2>already pre trained massive network. That's mainly where the focus

0:50:21.320 --> 0:50:24.400
<v Speaker 2>turned after GPT four. So GPT four was like the

0:50:24.480 --> 0:50:28.959
<v Speaker 2>extent of pre training making it smarter. After GPT four,

0:50:29.080 --> 0:50:32.080
<v Speaker 2>trying to make those models even bigger and pre train

0:50:32.160 --> 0:50:35.080
<v Speaker 2>them longer didn't lead to much performance increases. So everything

0:50:35.080 --> 0:50:37.440
<v Speaker 2>we got between four in the lead up to five

0:50:38.360 --> 0:50:40.680
<v Speaker 2>was post training, and that's when they began focusing on

0:50:40.800 --> 0:50:43.879
<v Speaker 2>metrics because if I have a particular metric, I can

0:50:43.960 --> 0:50:46.200
<v Speaker 2>post train a model to do well on that test.

0:50:46.280 --> 0:50:48.920
<v Speaker 2>And so everything became about metrics and post now we

0:50:48.920 --> 0:50:51.359
<v Speaker 2>can do this thing better or look at this thing

0:50:51.440 --> 0:50:53.560
<v Speaker 2>we do better, And so that's kind of the game

0:50:54.040 --> 0:50:56.719
<v Speaker 2>that's played now is we do lots of post training

0:50:57.280 --> 0:50:59.879
<v Speaker 2>that requires like much more specific data because you need

0:50:59.920 --> 0:51:03.600
<v Speaker 2>like right answers, pairs of props, and correct answers. So

0:51:03.640 --> 0:51:05.719
<v Speaker 2>only certain things we can do this with. But that's

0:51:05.719 --> 0:51:07.800
<v Speaker 2>the game we've been playing since like twenty that's one.

0:51:08.400 --> 0:51:10.600
<v Speaker 1>Do they do that with models that exist now? So

0:51:11.200 --> 0:51:12.799
<v Speaker 1>this is basically updates, right?

0:51:12.920 --> 0:51:17.319
<v Speaker 2>Yeah? They do it on a semi regular basis. Yeah,

0:51:17.320 --> 0:51:19.240
<v Speaker 2>but they usually there'll be a name change, like GPT

0:51:19.320 --> 0:51:21.280
<v Speaker 2>five to two is different than five to one, different

0:51:21.320 --> 0:51:21.760
<v Speaker 2>than five.

0:51:21.960 --> 0:51:23.680
<v Speaker 1>But don't they update the current models?

0:51:24.880 --> 0:51:27.440
<v Speaker 2>They don't repretrain them. That's too expensive.

0:51:27.480 --> 0:51:29.359
<v Speaker 1>But they no, no, I'm not saying that. And say

0:51:29.400 --> 0:51:31.560
<v Speaker 1>do they post train the current models to make them

0:51:31.640 --> 0:51:32.320
<v Speaker 1>better at stuff?

0:51:32.400 --> 0:51:32.640
<v Speaker 2>Yeah?

0:51:32.680 --> 0:51:33.440
<v Speaker 1>To tweak things.

0:51:33.520 --> 0:51:33.920
<v Speaker 2>Yeah.

0:51:33.960 --> 0:51:36.560
<v Speaker 1>So that's This is a very long way to get

0:51:36.600 --> 0:51:40.200
<v Speaker 1>to a point I'm kind of making. Which is one

0:51:40.239 --> 0:51:42.880
<v Speaker 1>of the problems with VIBE reporting on this is training

0:51:42.960 --> 0:51:46.319
<v Speaker 1>is framed as this thing like inference is framed as

0:51:46.400 --> 0:51:49.560
<v Speaker 1>op X that is permanent that you cannot avoid influence

0:51:49.600 --> 0:51:52.520
<v Speaker 1>being creating the outputs. Training is framed as this R

0:51:52.600 --> 0:51:56.440
<v Speaker 1>and D mysticism, which is just out there and you're

0:51:56.440 --> 0:51:59.280
<v Speaker 1>not like training. They never say this, but you hear training,

0:51:59.320 --> 0:52:02.160
<v Speaker 1>you think, oh, you train and then you perform, and

0:52:02.280 --> 0:52:06.080
<v Speaker 1>so training would end. But from what I understand, training

0:52:06.160 --> 0:52:09.680
<v Speaker 1>is as common and necessary an expense as influence at

0:52:09.680 --> 0:52:10.040
<v Speaker 1>this point.

0:52:10.160 --> 0:52:12.719
<v Speaker 2>Yeah, it's the only way you improve or update things, right, Like, so,

0:52:12.719 --> 0:52:16.000
<v Speaker 2>if you had a Microsoft through sixty five software, you're

0:52:16.040 --> 0:52:19.120
<v Speaker 2>constantly sending updates and patches and whatever is you like

0:52:19.160 --> 0:52:21.959
<v Speaker 2>to add new features or whatever. In the AI model world,

0:52:22.040 --> 0:52:24.520
<v Speaker 2>it's it's yeah, it's posting more rounds of post training.

0:52:24.520 --> 0:52:27.440
<v Speaker 2>It's the only way to make any sort of improvement

0:52:27.600 --> 0:52:30.600
<v Speaker 2>or fixing bugs. Like you're like, oh, here's something that's

0:52:30.640 --> 0:52:32.799
<v Speaker 2>saying that we're really upset about. Okay, let's go in

0:52:32.840 --> 0:52:35.160
<v Speaker 2>and do some zapping. Let's get out to zapper, give

0:52:35.200 --> 0:52:36.920
<v Speaker 2>it a bunch of examples of saying that bad thing,

0:52:36.960 --> 0:52:38.640
<v Speaker 2>and let zap and say don't do that. And now

0:52:38.680 --> 0:52:41.640
<v Speaker 2>it's like very unlikely to do that. So yeah, they're

0:52:41.640 --> 0:52:44.879
<v Speaker 2>constantly they're constantly doing that. Otherwise, your a stasis, which

0:52:44.920 --> 0:52:47.440
<v Speaker 2>is different than the way most people actually think of it. Differently,

0:52:48.000 --> 0:52:50.640
<v Speaker 2>they think that the model is somehow like learning online

0:52:51.239 --> 0:52:54.120
<v Speaker 2>that like, which as it's absolutely not true that as

0:52:54.160 --> 0:52:57.359
<v Speaker 2>you talk to it, it's learning and it's getting uh,

0:52:57.600 --> 0:53:00.320
<v Speaker 2>it's getting better. And they're like, but wait a second,

0:53:00.440 --> 0:53:03.600
<v Speaker 2>something I talked about earlier, it must be getting smarter.

0:53:03.800 --> 0:53:06.840
<v Speaker 2>The model doesn't care about you, just your local software.

0:53:06.920 --> 0:53:10.600
<v Speaker 2>You don't realize this. It's including like huge bits of

0:53:10.600 --> 0:53:12.719
<v Speaker 2>stuff you've talked to it before. In the prompt that's

0:53:12.760 --> 0:53:14.480
<v Speaker 2>going to the model. It's like, here's a bunch of

0:53:14.480 --> 0:53:17.480
<v Speaker 2>stuff that Ed has submitted to you in the past. Okay,

0:53:17.480 --> 0:53:21.080
<v Speaker 2>now here's his current question. So it's the model hasn't changed.

0:53:21.080 --> 0:53:23.160
<v Speaker 2>The model doesn't have memory a canunt of justice. It

0:53:23.200 --> 0:53:25.800
<v Speaker 2>doesn't adjust its memory in real time. It's all static.

0:53:25.840 --> 0:53:29.400
<v Speaker 2>There is no dynamic memory involved in these models. They're

0:53:29.640 --> 0:53:31.520
<v Speaker 2>fixed until you go out and post train it and

0:53:31.520 --> 0:53:33.279
<v Speaker 2>then you bring the new weights back in the data center.

0:53:33.320 --> 0:53:34.640
<v Speaker 2>And that's the thing. Do an inference now.

0:53:35.360 --> 0:53:37.960
<v Speaker 1>But that's the thing like this. The reason I bring

0:53:38.000 --> 0:53:41.319
<v Speaker 1>it up as the kind of closing vibe story is

0:53:41.360 --> 0:53:45.920
<v Speaker 1>because training is very clearly getting more expensive what they

0:53:45.960 --> 0:53:48.919
<v Speaker 1>say profitable in inference, which again doesn't really make sense

0:53:48.960 --> 0:53:52.040
<v Speaker 1>to me. But putting that aside, they're not. But even

0:53:52.080 --> 0:53:56.040
<v Speaker 1>if they were, if training never stops, then who cares

0:53:56.239 --> 0:53:58.920
<v Speaker 1>about profitable in inference? Like it just means that you

0:53:59.520 --> 0:54:03.640
<v Speaker 1>will get more or expensive forever. Inference is just yeah, yeah,

0:54:03.719 --> 0:54:08.239
<v Speaker 1>differences pooh, training is pooh. That's just that's I'm not

0:54:08.280 --> 0:54:08.960
<v Speaker 1>putting that one in.

0:54:09.000 --> 0:54:12.000
<v Speaker 2>A yeah, okay, it's an interesting question, right because it's

0:54:12.000 --> 0:54:14.120
<v Speaker 2>also getting like the what's an altman who is making

0:54:14.120 --> 0:54:16.680
<v Speaker 2>those comments about well, we just didn't have to pay

0:54:16.719 --> 0:54:18.839
<v Speaker 2>for the training. This would be part I.

0:54:18.760 --> 0:54:21.480
<v Speaker 1>Didn't have all these expenses, I would be so profitable.

0:54:22.040 --> 0:54:24.800
<v Speaker 2>The one distinction that's maybe relevant there not to be

0:54:24.840 --> 0:54:26.880
<v Speaker 2>an apologist, but the one one distinction that is relevant

0:54:26.880 --> 0:54:31.480
<v Speaker 2>is pre training versus posts. So pre training is insanely expensive, right,

0:54:31.600 --> 0:54:35.440
<v Speaker 2>because you're you're training something on all the all the

0:54:35.480 --> 0:54:38.880
<v Speaker 2>words in the world, and it takes months. So it's

0:54:38.960 --> 0:54:42.640
<v Speaker 2>it's just like you're running a data center that's going

0:54:42.719 --> 0:54:47.440
<v Speaker 2>to have nowadays up to six digit GPUs uh running

0:54:47.440 --> 0:54:50.719
<v Speaker 2>full time for months just to get that pre training done,

0:54:50.760 --> 0:54:52.440
<v Speaker 2>and you have to pay for all of that, right,

0:54:52.480 --> 0:54:54.920
<v Speaker 2>So that's all right time, you're not getting money, you're

0:54:54.920 --> 0:54:57.279
<v Speaker 2>just paying for training. Post training is also expensive. It's

0:54:57.320 --> 0:55:00.600
<v Speaker 2>not that expensive though, because as expensive as pre training,

0:55:00.680 --> 0:55:02.759
<v Speaker 2>because each post training session it's a way, way, way

0:55:02.760 --> 0:55:05.080
<v Speaker 2>way smaller data set that you're post training on. It's like,

0:55:05.120 --> 0:55:08.240
<v Speaker 2>all right, we generated like ten thousand examples of people,

0:55:08.440 --> 0:55:11.360
<v Speaker 2>you know, responding to questions in a racist way, and

0:55:11.400 --> 0:55:15.200
<v Speaker 2>those ten thousand examples will use to reinforcement learn and

0:55:15.280 --> 0:55:17.200
<v Speaker 2>try to move it away from answering those type of

0:55:17.239 --> 0:55:19.680
<v Speaker 2>questions in a racist way. That's like not that big

0:55:19.719 --> 0:55:22.600
<v Speaker 2>of a data set compared to we're going to train

0:55:22.680 --> 0:55:25.840
<v Speaker 2>this on every word written that we have access to.

0:55:25.920 --> 0:55:28.520
<v Speaker 2>So the post training is not as expensive as pre training.

0:55:29.040 --> 0:55:30.520
<v Speaker 1>Unless you're doing it all the time.

0:55:30.680 --> 0:55:31.600
<v Speaker 2>Yeah, that's true.

0:55:31.960 --> 0:55:34.080
<v Speaker 1>That's the thing if you're doing it all the time

0:55:35.000 --> 0:55:38.440
<v Speaker 1>and it takes months of pre training, but you're constantly

0:55:38.440 --> 0:55:41.359
<v Speaker 1>doing something like post training for months. Yeah, I mean

0:55:41.360 --> 0:55:43.640
<v Speaker 1>post training not functionally the same thing.

0:55:43.640 --> 0:55:45.319
<v Speaker 2>It's the same thing. I think this is a fair point.

0:55:45.360 --> 0:55:47.439
<v Speaker 2>When you when you have a particular example that you're

0:55:47.440 --> 0:55:50.120
<v Speaker 2>post training on like one prompt with an answer, it's

0:55:50.200 --> 0:55:52.080
<v Speaker 2>kind of like you're doing inference in reverse, right, So

0:55:52.120 --> 0:55:55.080
<v Speaker 2>you're going from your back propagating from one side of

0:55:55.120 --> 0:55:56.839
<v Speaker 2>the network to the beginning, as a post are going

0:55:56.840 --> 0:55:58.800
<v Speaker 2>from the beginning to the end. Now, it's more expensive

0:55:58.840 --> 0:56:02.759
<v Speaker 2>than that because when you're just doing inference, the fundamental

0:56:02.800 --> 0:56:06.920
<v Speaker 2>operation is basic multiplication. You're just multiplying numbers in the

0:56:06.920 --> 0:56:10.759
<v Speaker 2>big table backpropagation which you use to training. You're it

0:56:10.800 --> 0:56:12.680
<v Speaker 2>is multiplication. When you do a bunch of it's a

0:56:12.680 --> 0:56:16.360
<v Speaker 2>bunch of derivatives because you're constantly you need to calculate

0:56:16.440 --> 0:56:19.879
<v Speaker 2>like the derivative of these. I mean like it doesn't

0:56:19.920 --> 0:56:21.319
<v Speaker 2>really matter, but you kind of need that. You want

0:56:21.320 --> 0:56:23.759
<v Speaker 2>the derivative because you want the gradient to scent to

0:56:23.800 --> 0:56:26.440
<v Speaker 2>be towards like better and away from worse and derivatives.

0:56:27.560 --> 0:56:31.000
<v Speaker 2>My understanding is this is like more expensive operations per

0:56:31.239 --> 0:56:33.200
<v Speaker 2>weight that you're trying to change because you're not just

0:56:33.280 --> 0:56:36.160
<v Speaker 2>multiplying a number, You're having to calculate derivatives and becomes

0:56:36.160 --> 0:56:38.120
<v Speaker 2>a little bit more complicated. So yeah, it's like inference

0:56:38.160 --> 0:56:40.719
<v Speaker 2>in reverse, but also like a little bit more expensive.

0:56:40.800 --> 0:56:45.480
<v Speaker 2>So if you if you have ten thousand sample question responses,

0:56:45.520 --> 0:56:47.319
<v Speaker 2>you're going to use the post train, it's kind of

0:56:47.360 --> 0:56:51.600
<v Speaker 2>like ten thousand users sent prompts and there are particularly

0:56:51.600 --> 0:56:53.520
<v Speaker 2>expensive prompts, and you had to pay for that instead

0:56:53.520 --> 0:56:55.960
<v Speaker 2>of like them paying for it. So yeah, it's good

0:56:55.960 --> 0:56:58.440
<v Speaker 2>to think of it as like inference and reverse.

0:56:59.080 --> 0:57:04.080
<v Speaker 1>It's also an ongoing cost, Like it's everyone is leaning

0:57:04.080 --> 0:57:07.320
<v Speaker 1>on this idea that this stuff will magically become profitable.

0:57:07.320 --> 0:57:10.120
<v Speaker 1>I don't know if you've seen the cash flow diagrams

0:57:10.120 --> 0:57:13.040
<v Speaker 1>of anthropic and open ai. But there is a mysterious

0:57:13.080 --> 0:57:15.560
<v Speaker 1>math going on where year twenty twenty eight, twenty twenty

0:57:15.680 --> 0:57:16.960
<v Speaker 1>nine they just become profitable.

0:57:16.960 --> 0:57:19.560
<v Speaker 2>Well, I was going to ask you about this. Last month,

0:57:19.920 --> 0:57:23.800
<v Speaker 2>there's this announcement that open ai had some massive increase

0:57:23.960 --> 0:57:25.600
<v Speaker 2>in revenue.

0:57:26.800 --> 0:57:29.400
<v Speaker 1>Yeah, well this is actually a great vibe reporting thing,

0:57:29.440 --> 0:57:32.680
<v Speaker 1>and that's annualized revenue. They said they hit twenty billion

0:57:32.920 --> 0:57:35.439
<v Speaker 1>in annualized revenue, which would mean one point sixty seven

0:57:35.480 --> 0:57:39.960
<v Speaker 1>billion in a month. Now, important details. We don't know

0:57:40.040 --> 0:57:42.800
<v Speaker 1>how they're defining a month. We don't know if they

0:57:42.840 --> 0:57:45.560
<v Speaker 1>mean twenty eight days, twenty nine days, thirty days. We

0:57:45.600 --> 0:57:47.920
<v Speaker 1>don't know if they mean a calendar month, so the

0:57:47.960 --> 0:57:50.240
<v Speaker 1>month of November or December, or do they just bean

0:57:50.400 --> 0:57:54.080
<v Speaker 1>any thirty days. We don't know if they're doing insane math,

0:57:54.200 --> 0:57:57.880
<v Speaker 1>which happens very rarely, that this company feels like one

0:57:57.880 --> 0:58:01.000
<v Speaker 1>that might do it. Just my gut instinct is they

0:58:01.040 --> 0:58:03.320
<v Speaker 1>may be doing his seven day period and we're going

0:58:03.360 --> 0:58:05.920
<v Speaker 1>to turn it into a month like they We don't

0:58:05.920 --> 0:58:07.720
<v Speaker 1>know how they're doing this. And they also coupled this

0:58:07.800 --> 0:58:11.640
<v Speaker 1>by saying that as compute grows, revenue grows. I don't

0:58:11.640 --> 0:58:12.280
<v Speaker 1>know if you seen this.

0:58:13.040 --> 0:58:15.520
<v Speaker 2>No, it's like, say, their formula, yeah, okay.

0:58:15.400 --> 0:58:18.360
<v Speaker 1>Yeah, it's an insane formula that does not map to

0:58:18.520 --> 0:58:21.680
<v Speaker 1>any economics. Like it's just it's the kind of thing

0:58:21.760 --> 0:58:24.120
<v Speaker 1>that if we had a functioning business and tech press,

0:58:24.880 --> 0:58:28.680
<v Speaker 1>that would just be scrutinized to the bone, that would

0:58:28.720 --> 0:58:31.000
<v Speaker 1>just be ripped apart and say, what the fuck does

0:58:31.040 --> 0:58:33.960
<v Speaker 1>that mean? Because if this were true, if you simply

0:58:34.040 --> 0:58:38.120
<v Speaker 1>add more gigawattch at more revenue, then you would simply

0:58:38.760 --> 0:58:40.880
<v Speaker 1>print more money, like it would be a money print

0:58:40.920 --> 0:58:41.360
<v Speaker 1>of us.

0:58:41.440 --> 0:58:44.240
<v Speaker 2>Like my movie theater did well. Therefore, yeah, if I

0:58:44.240 --> 0:58:46.200
<v Speaker 2>build one hundred thousand movie theaters, we're gonna make one

0:58:46.240 --> 0:58:48.040
<v Speaker 2>hundred thousand more you know, times the amount of money.

0:58:48.080 --> 0:58:49.880
<v Speaker 2>And it's like, well, wait that theater was in Manhattan

0:58:50.000 --> 0:58:54.680
<v Speaker 2>and it was like really well run, and they're yeah,

0:58:54.800 --> 0:58:58.520
<v Speaker 2>well okay, I assume I assumed as much trying to

0:58:58.800 --> 0:59:01.520
<v Speaker 2>understand that story. Okay, yeah, the annualized revenue stuff, I

0:59:01.560 --> 0:59:03.360
<v Speaker 2>think I even use that term talking to someone I

0:59:03.360 --> 0:59:05.360
<v Speaker 2>credit you as like Trum would say, look out for

0:59:05.400 --> 0:59:06.280
<v Speaker 2>annualized revenue.

0:59:07.160 --> 0:59:09.320
<v Speaker 1>It is. The funny thing is with that as well,

0:59:09.720 --> 0:59:16.280
<v Speaker 1>it's more viabe reporting because arr standard. It's very standard

0:59:16.280 --> 0:59:19.680
<v Speaker 1>in SaaS companies that sell on a per seat basis,

0:59:19.720 --> 0:59:22.680
<v Speaker 1>So a sales force would do a well even they

0:59:22.720 --> 0:59:25.040
<v Speaker 1>are doing annualized now with the AI revenue, but it

0:59:25.080 --> 0:59:27.760
<v Speaker 1>would be a software company has one hundred seats they

0:59:27.760 --> 0:59:29.600
<v Speaker 1>sell to a company and they charge fifteen bucks a

0:59:29.640 --> 0:59:33.480
<v Speaker 1>bit a seat per month, and they charge that annually,

0:59:33.880 --> 0:59:37.200
<v Speaker 1>and the actual cost of each user is fairly measurable

0:59:37.400 --> 0:59:41.000
<v Speaker 1>because they're doing CPU based stuff. Like it gets expensive

0:59:41.000 --> 0:59:43.080
<v Speaker 1>at scale because there's a lot of people, but it

0:59:43.120 --> 0:59:47.240
<v Speaker 1>doesn't get multiply multi I can't say the word. It

0:59:47.280 --> 0:59:50.280
<v Speaker 1>doesn't get much more expensive as you grow with AI.

0:59:50.720 --> 0:59:53.560
<v Speaker 1>It's actually because of the way large language models work,

0:59:53.840 --> 0:59:56.640
<v Speaker 1>your most excited customers are your most expensive.

0:59:56.880 --> 0:59:59.840
<v Speaker 2>Yeah, because they blow whatever whatever their monthly revenue is,

1:00:00.000 --> 1:00:02.240
<v Speaker 2>whatever their cost is, they blow past it, so.

1:00:02.360 --> 1:00:06.320
<v Speaker 1>You can't even do a percy revenue. It's just every

1:00:06.400 --> 1:00:08.560
<v Speaker 1>all of these things. When I say them out loud,

1:00:08.600 --> 1:00:11.280
<v Speaker 1>I'm like, I feel like this should be more obvious.

1:00:11.280 --> 1:00:12.680
<v Speaker 1>It's why I loved your video because it was like

1:00:12.720 --> 1:00:15.240
<v Speaker 1>thank you someone else, Well, here's the.

1:00:15.280 --> 1:00:17.320
<v Speaker 2>Question, here's my here's my question, here's like the dangerous

1:00:17.360 --> 1:00:19.160
<v Speaker 2>question I actually put out a video last week. It

1:00:19.200 --> 1:00:24.080
<v Speaker 2>was like dangerous question. We're now on year three or

1:00:24.120 --> 1:00:27.640
<v Speaker 2>four of like the I'm counting new year starting like

1:00:27.640 --> 1:00:31.040
<v Speaker 2>New Year twenty twenty three or whatever of people saying,

1:00:31.800 --> 1:00:35.200
<v Speaker 2>oh my god, this is so cool. These massive disruptions

1:00:35.200 --> 1:00:38.439
<v Speaker 2>they're going to change everything is imminent, and year after

1:00:38.520 --> 1:00:42.600
<v Speaker 2>year we've said that I'm not yet seeing the massive disruptions,

1:00:43.720 --> 1:00:45.720
<v Speaker 2>like not the not the stories of what might be

1:00:45.760 --> 1:00:48.360
<v Speaker 2>disrupted or the stories of what's different, but like, how

1:00:48.400 --> 1:00:53.840
<v Speaker 2>many years do we have to go without industries crumbling

1:00:54.400 --> 1:00:58.720
<v Speaker 2>or major new economic players that didn't exist before, or

1:00:58.800 --> 1:01:02.680
<v Speaker 2>complete restructuring of huge companies around this technology? How many

1:01:02.760 --> 1:01:05.160
<v Speaker 2>years we have to go? So I did a video,

1:01:05.240 --> 1:01:07.160
<v Speaker 2>right I went, I found the Reddit thread where someone

1:01:07.200 --> 1:01:08.880
<v Speaker 2>just asked this question. This was from like earlier in

1:01:08.880 --> 1:01:12.080
<v Speaker 2>the month. They were like, outside of the vibe coding stuff,

1:01:13.880 --> 1:01:16.240
<v Speaker 2>what what are the like, what are people? What are

1:01:16.240 --> 1:01:18.200
<v Speaker 2>the big tools? Like, what are the big things that

1:01:18.200 --> 1:01:20.240
<v Speaker 2>have come out of this technology? They're changing things? And

1:01:20.280 --> 1:01:22.240
<v Speaker 2>I read through this whole thread and it was interesting,

1:01:22.280 --> 1:01:24.160
<v Speaker 2>there's not people don't have much like well, you know,

1:01:24.280 --> 1:01:26.960
<v Speaker 2>like it could these are like really small case studies.

1:01:27.400 --> 1:01:30.080
<v Speaker 2>I used it to help, you know, gather clean up

1:01:30.160 --> 1:01:32.840
<v Speaker 2>data that I got from whatever. I was like, this

1:01:32.880 --> 1:01:35.480
<v Speaker 2>is like such a nerdy specific use case, and so

1:01:35.560 --> 1:01:37.640
<v Speaker 2>I went through that thread in a video and this

1:01:37.680 --> 1:01:39.240
<v Speaker 2>has kind of been my question. It's like, it is

1:01:39.360 --> 1:01:41.480
<v Speaker 2>very cool technology, but how do we know where this

1:01:41.520 --> 1:01:44.440
<v Speaker 2>is going to fall? Like to me, the scale is

1:01:44.600 --> 1:01:50.880
<v Speaker 2>it would go like this, like blockchain software, then Oculus VR,

1:01:51.680 --> 1:01:55.240
<v Speaker 2>then maybe internet to electricity. Right, so we're gonna have

1:01:55.280 --> 1:01:56.200
<v Speaker 2>a scale of disruption.

1:01:56.320 --> 1:01:56.400
<v Speaker 1>Right.

1:01:56.480 --> 1:02:00.400
<v Speaker 2>Blockchain software is something where the premise it's made no

1:02:00.520 --> 1:02:02.240
<v Speaker 2>sense and it was never going to get off the

1:02:02.240 --> 1:02:04.680
<v Speaker 2>ground and there was going to be no impact on

1:02:04.720 --> 1:02:07.400
<v Speaker 2>the world. And you know, because I'm a my training

1:02:07.400 --> 1:02:09.560
<v Speaker 2>in CS is in distributed system theory. I was there

1:02:09.560 --> 1:02:11.240
<v Speaker 2>in twenty twenty saying guys, let me just tell you

1:02:11.600 --> 1:02:13.520
<v Speaker 2>this is nonsense. No Web three is not about to

1:02:13.520 --> 1:02:15.000
<v Speaker 2>take off. None of this makes sense. And that was true.

1:02:15.040 --> 1:02:18.840
<v Speaker 2>That did nothing. Then you have like Oculus VR. It

1:02:18.920 --> 1:02:21.720
<v Speaker 2>really is cool, right you put on these things like

1:02:21.760 --> 1:02:24.640
<v Speaker 2>that is awesome, Like I love this technology, but it's

1:02:24.680 --> 1:02:26.600
<v Speaker 2>having a hard time having any real major impact because

1:02:26.600 --> 1:02:27.280
<v Speaker 2>people aren't sure what.

1:02:27.320 --> 1:02:31.440
<v Speaker 1>Times use of it. Also, most people don't necessarily have

1:02:31.520 --> 1:02:35.520
<v Speaker 1>a great experience initially because it's extremely dependent on where

1:02:35.560 --> 1:02:37.520
<v Speaker 1>you are, who you are, the size of your skull

1:02:37.640 --> 1:02:38.840
<v Speaker 1>not kind of yeah so.

1:02:38.960 --> 1:02:41.880
<v Speaker 2>Big right for a limited group of people, it's really cool,

1:02:41.880 --> 1:02:43.960
<v Speaker 2>but it fails to come out. Then the Internet is

1:02:44.040 --> 1:02:47.160
<v Speaker 2>like really disruptive, changed a lot of things more. It's

1:02:47.160 --> 1:02:49.480
<v Speaker 2>not so much as like whole industries disappeared or whatever,

1:02:49.560 --> 1:02:51.160
<v Speaker 2>but it like changed the way a lot of industries

1:02:51.200 --> 1:02:53.160
<v Speaker 2>actually function. And then like electricity, you could say, like

1:02:53.160 --> 1:02:56.360
<v Speaker 2>it just completely changed what day to day existence of.

1:02:56.360 --> 1:02:59.000
<v Speaker 1>Business was, like how existence of exactly?

1:02:59.080 --> 1:03:01.200
<v Speaker 2>Yeah, And so that the question like that everyone should

1:03:01.200 --> 1:03:02.960
<v Speaker 2>be asking is where is the ENERVEI going to fall

1:03:03.080 --> 1:03:06.400
<v Speaker 2>on this? And you know, I would say, right now,

1:03:06.440 --> 1:03:08.040
<v Speaker 2>this is what gives me yelled at. I'm not saying

1:03:08.040 --> 1:03:10.280
<v Speaker 2>this is the prediction necessarily going forward, but right now

1:03:10.360 --> 1:03:12.760
<v Speaker 2>I don't think it's got passed much farther past the

1:03:13.200 --> 1:03:16.760
<v Speaker 2>oculus part of that scale where it's really cool. There's

1:03:16.960 --> 1:03:19.920
<v Speaker 2>very cool things. Chat Gipt is very cool. It's very

1:03:19.960 --> 1:03:22.200
<v Speaker 2>cool that it can like have that comprehension and no

1:03:22.240 --> 1:03:24.600
<v Speaker 2>one thought it could do that, But we haven't yet

1:03:24.600 --> 1:03:25.160
<v Speaker 2>figured out what.

1:03:27.400 --> 1:03:31.240
<v Speaker 1>Comprehension. It's no, it's a text I mean.

1:03:31.440 --> 1:03:33.280
<v Speaker 2>We take it for granted, but for CS people, the

1:03:33.320 --> 1:03:35.480
<v Speaker 2>ability that like I can add, hey, give me text

1:03:35.680 --> 1:03:38.160
<v Speaker 2>that like whatever in the style of a poem that

1:03:38.200 --> 1:03:42.480
<v Speaker 2>does whatever, and that includes a character from Star Wars,

1:03:42.640 --> 1:03:44.280
<v Speaker 2>and then it can give you text it does that.

1:03:44.280 --> 1:03:46.680
<v Speaker 2>That comprehension, like for computer scientist was like, oh, we

1:03:46.720 --> 1:03:49.520
<v Speaker 2>didn't really know how to consistently one yeah. Yeah, that's

1:03:49.520 --> 1:03:53.280
<v Speaker 2>like very cool. Yeah, but we haven't got past this

1:03:53.440 --> 1:03:56.240
<v Speaker 2>is like the the surprising thing of this field is

1:03:57.120 --> 1:03:59.760
<v Speaker 2>we're not really past the oculus stage yet, like where

1:03:59.800 --> 1:04:02.320
<v Speaker 2>they're like, for certain this is vibe coding is really cool.

1:04:02.600 --> 1:04:05.240
<v Speaker 2>The comprehension is cool, so like sora is weird, but

1:04:05.280 --> 1:04:07.880
<v Speaker 2>like it's cool, you can do that, But the markets

1:04:07.920 --> 1:04:10.920
<v Speaker 2>are not None of these have markets yet, right, Like that,

1:04:10.960 --> 1:04:13.600
<v Speaker 2>there's not big markets in any of these yet. And

1:04:13.640 --> 1:04:16.000
<v Speaker 2>will how far will it go from oculus to the Internet.

1:04:16.040 --> 1:04:18.400
<v Speaker 2>To me, that is like the number one question, the

1:04:18.440 --> 1:04:20.520
<v Speaker 2>number two question, the number three question of all reporting

1:04:20.560 --> 1:04:23.600
<v Speaker 2>on this, and almost no one's talking about that. It's

1:04:23.680 --> 1:04:28.280
<v Speaker 2>just hype laundering. We'll take this hype, will extrapolate, it

1:04:28.280 --> 1:04:31.160
<v Speaker 2>will react to that extrapolation. That's kind of what reporting is, right,

1:04:31.200 --> 1:04:33.919
<v Speaker 2>now in AI, Whereas to me, this is the hugest question.

1:04:33.960 --> 1:04:36.600
<v Speaker 2>If this ends up oculus, retail investors are going to

1:04:36.640 --> 1:04:39.400
<v Speaker 2>get screwed. If it ends up Internet, all right, that's like,

1:04:39.640 --> 1:04:42.120
<v Speaker 2>that's like a really interesting significant story. If it ends

1:04:42.160 --> 1:04:44.640
<v Speaker 2>up electricity, obviously that really matters. But like, I don't

1:04:44.680 --> 1:04:47.240
<v Speaker 2>know anyone who actually thinks it's going to be that disruptive.

1:04:47.320 --> 1:04:51.000
<v Speaker 2>Not the current technology. That's the story to me. Not

1:04:52.120 --> 1:04:54.840
<v Speaker 2>let's like extrapolate us. You know, hey, what are these

1:04:54.840 --> 1:04:57.080
<v Speaker 2>things are creating a church? Or let's let's let's hype

1:04:57.120 --> 1:05:01.240
<v Speaker 2>launder extrapolate that and react to our extrapol That's not

1:05:01.280 --> 1:05:04.920
<v Speaker 2>really reporting so much as speculative fiction writing. I guess

1:05:04.960 --> 1:05:06.560
<v Speaker 2>I don't know quite what to call it. But this

1:05:06.680 --> 1:05:10.560
<v Speaker 2>is the real question, where exactly are we now and

1:05:10.600 --> 1:05:12.600
<v Speaker 2>what are the possibilities of like where this is going

1:05:12.680 --> 1:05:14.560
<v Speaker 2>to go positive and negative? I don't we have enough

1:05:14.560 --> 1:05:14.960
<v Speaker 2>talk on that.

1:05:16.280 --> 1:05:18.520
<v Speaker 1>I fully agree, Cal, it's been such a pleasure in

1:05:18.560 --> 1:05:19.840
<v Speaker 1>having you. Thank you for joining me.

1:05:20.120 --> 1:05:22.640
<v Speaker 2>Always happy to talk shop, always happy to hate with you.

1:05:22.720 --> 1:05:25.000
<v Speaker 2>I guess in your season, I like.

1:05:24.960 --> 1:05:27.600
<v Speaker 1>It, Hey, season is the best. We will be back

1:05:27.640 --> 1:05:29.800
<v Speaker 1>this week with either a monologu or an interview. I

1:05:29.840 --> 1:05:34.040
<v Speaker 1>have not decided, because I've got wonderful Corey Quinn interview

1:05:34.080 --> 1:05:36.160
<v Speaker 1>I just did, so I'm considering putting that in a monologue.

1:05:36.160 --> 1:05:38.600
<v Speaker 1>You'll find out on Friday. Anyway, this has been Better

1:05:38.600 --> 1:05:42.800
<v Speaker 1>off Line. I'm at zychron Subscribe to the premium, download

1:05:42.840 --> 1:05:53.760
<v Speaker 1>a T shirt, whatever you desire. Thank you for listening

1:05:53.800 --> 1:05:56.440
<v Speaker 1>to Better Offline. The editor and composer of the Better

1:05:56.440 --> 1:05:59.440
<v Speaker 1>Offline theme song is Matosowski. You can check out more

1:05:59.480 --> 1:06:03.000
<v Speaker 1>of his music and audio projects at Mattasowski dot com,

1:06:03.120 --> 1:06:06.640
<v Speaker 1>m A T T O S O W s ki

1:06:06.840 --> 1:06:09.760
<v Speaker 1>dot com. You can email me at easy at Better

1:06:09.760 --> 1:06:12.200
<v Speaker 1>Offline dot com or visit Better Offline dot com to

1:06:12.240 --> 1:06:15.080
<v Speaker 1>find more podcast links and of course, my newsletter. I

1:06:15.120 --> 1:06:17.840
<v Speaker 1>also really recommend you go to chat dot Where's youreaed

1:06:17.920 --> 1:06:20.160
<v Speaker 1>dot at to visit the discord, and go to our

1:06:20.240 --> 1:06:23.560
<v Speaker 1>slash Better Offline to check out our reddit. Thank you

1:06:23.600 --> 1:06:24.560
<v Speaker 1>so much for listening.

1:06:25.400 --> 1:06:27.840
<v Speaker 2>Better Offline is a production of cool Zone Media.

1:06:28.040 --> 1:06:30.880
<v Speaker 1>For more from cool Zone Media, visit our website cool

1:06:30.920 --> 1:06:33.280
<v Speaker 1>Zonemedia dot com, or check us out on the

1:06:33.360 --> 1:06:37.040
<v Speaker 2>iHeartRadio app, Apple Podcasts, or wherever you get your podcasts.