Social Media Tribes, Algorithms and the Permanent Underclass, with Signulll

I turned @signulll into an ominous gray blob as we sat down in Alamo Square to talk about normal people things like how social media is tribal again, the X dot com the everything app algorithm and the permanent underclass.
We also spoke to some dogs.

Watch on X; listen on Apple Podcasts or Spotify.

Sponsors

• Blackbox AI is an AI-powered software development platform designed to help developers and non-developers build, test, fix, and deploy code and applications faster and with less manual effort than traditional tools.
• Superpower is a modern health platform that gives you a complete, data-driven view of your body in one place. From advanced blood testing and biomarker tracking to personalized insights and action plans, Superpower helps you understand what’s actually going on under the hood and what to do about it. Built for people who want to take their health seriously without guessing, spreadsheets, or fragmented care.
•  Zo is an intelligent cloud computer that brings your files, tools, and AI into one persistent workspace, letting you build personal apps, automations, and projects that actually understand your full context.
• Capy is the better way to build and run AI workflows. It’s a multiplayer IDE for agents, automations, and intelligent systems. Create agents, automate tasks, and connect your data into systems that actually get work done, all in one place.

Transcript

SPEAKER_04 0:00

Okay, so you might be wondering, what are we even doing here? How are we starting off this particular episode? Well, I was lucky enough to catch up with my good friend and fame poster signal before his app launch Sky, and we sat down in a park and spoke to some dogs and just generally got in people's way uh while we spoke about technology in the permanent underclass. Of course, before we actually get into that particular conversation though, I must give a shout out to Zocomputer, superpower.com, blackbox.ai, and a special shout out to Cappy.ai, newest mod sponsor, your multiplayer IDE. Manage all workflows in one particular place. Your new command center for agentic coding and for working parallel. I mean, even non-technical people can use it because they can work with Slack. All of these really cool things, you'll be seeing much more of them. Of course, black box and superpower, the homies, the homies. And last but not least, Zoe Computer. Again, we're just unapologetically shilling the homies here. The mods sponsors are the homies, the mods empire are the homies. They must be shilled and they will be looked after. So without further ado, we're gonna go ahead and cross to Alamo Square, where I got to catch up with my friend Signal.

SPEAKER_03 1:20

I think where we got with the state of X was pretty interesting. I think so, yeah. It's like the state of X, right? I mean, um, you know, obviously, I think the most fascinating thing about the timeline now, in which we both inhabit quite frequently, and have both inhabited quite frequently for quite a while now, right? We've seen like many iterations of like the anonymous account, for example. For sure. Uh, but it's like, you know, the difference between now and say pre-Elon, right? The difference in how the SIM clusters work, the difference in pe how people sort those sorts of things, um, it's pretty stark. Pretty stark. I go as far as saying that like, you know, the the centralizing power of social media is like that that era, right? Like just taking everybody from everywhere, from all walks of life, all over the country, red states, blue states, putting them into one space on the internet to fight each other, uh, that era might actually be coming to an end. That's what I would actually contend. Uh, I actually think that's a pretty uh positive reading of it. Why do you think that's coming to an end? Because everybody that we're kind of self-sorting again, you know?

SPEAKER_02 2:29

You mean like people are becoming tribal and the internet is becoming tribal and therefore the tribal nature of platforms, they're they're sort of segmenting the population into silos, I guess. Yes, we're like re-siloing things. And so like the like-minded individual. So are you just reinforcing your belief structures more so now than ever before?

SPEAKER_03 2:50

I think I think people actually are, I wouldn't say more so now than ever before. I feel like it's it's a return to the forum era of the internet in that we have X for those that uh, well, for those that are okay with being on the platform when Elon runs it. Uh we have Blue Sky for those that hate Elon. Uh we have uh we have Truth Social for those like hard MAGA.

SPEAKER_05 3:13

Oh look at the company, it's interrupted!

SPEAKER_03 3:18

Keep that in. This is what I want to happen. That's a glorious dog.

SPEAKER_02 3:24

Maybe that's a um you know that will get you a positive moment. The engagement level of that will be much higher than uh.

unknown 3:32

Please give them back.

SPEAKER_02 3:35

I thought he was gonna knock over the top.

SPEAKER_03 3:38

That would have been hilarious. I would have I would have kept that in as well. Um so the examples truth socia for those hard MAGA, uh, blue sky for those I mean, I'm gonna go ahead and say it hard leftist. Uh threads for those that are kind of, I mean, how dare I say normies?

SPEAKER_02 3:53

Soft leftist? Soft leftist, yeah. What the hell's a soft leftist? Yeah, I can't. Isn't isn't every leftist soft? Um, wow. Yeah, yeah, yeah.

SPEAKER_03 4:04

It's political. Uh yeah, it's it's over. No, I'm kidding. But you know, the the soft styling, so so it's uh instead of having like forums, we now just have uh separate Twitters for for each of these for each of these factions. At least it seems that way.

SPEAKER_02 4:18

Um Well, even within the platform there's different types of algorithmic environments you can get into, right? I think you can get into the sort of black pill, red pill stuff, and your feed just continuously reinforces that and then Totally. I think that's the the craziness is that um I had uh not to do be not to be this guy, which is I had a tweet about this, which was um I mean, I mean, I've I've got a I've got a tweet of yours uh ready to love. It's like writing uh it's like Imagine citing your own works.

SPEAKER_03 4:52

What do you think I do on my show? Just read my own tweets.

SPEAKER_02 4:55

Yeah, read my That's how I you know roll myself to sleep every day.

SPEAKER_03 4:59

Of course. I read my I read the feed to go to sleep.

SPEAKER_02 5:02

I look at the mirror and look how much of a failure I've become, and then I read my tweets and I'm like, oh my god, everything has been worth it.

SPEAKER_03 5:09

Everything has been worth it. Gotta get a few, get a few um, you know, mildly incriminating bangers out there, and it makes it makes all the failure worthwhile.

SPEAKER_02 5:15

Exactly.

SPEAKER_03 5:16

In my in my sick soft life.

SPEAKER_02 5:18

That's what makes your parents proud.

SPEAKER_03 5:20

Of course. I tell them every day.

SPEAKER_02 5:21

Yeah, it's like, mom, I've got three bangers today.

SPEAKER_03 5:23

Jesus Christ. Check check check out who commented on my post. It's sex to inventor. Like, you know, like guys like like Forge Gloyd Ayak, uh, is my one of my best friends online.

SPEAKER_02 5:33

Um I think there's like this. Let's just move on. No, I um I I was thinking about this. I was like, okay, well, if you like something anywhere, the algorithm will know that you like it and therefore will serve you more and more of this content. And there um there's this uh video that's like uh TikTok is um effectively teaching you to leave the ones you love. Right.

SPEAKER_03 5:56

You were you were posting about this quite recently now. Yeah, yeah.

SPEAKER_02 5:59

And so uh it was fascinating because um I think people are in these reinforcement loops, but they don't really realize it because the algorithm is the one that's controlling their sort of perception of reality. Yeah uh based on what they engage with. So I think if you go through any sort of life moments or whatever you are, um and you are like I think there's all this thing where people think that Facebook is listening to your microphone because it shows you were the right thing at the right time. That's just a a very sophisticated algorithm that is predictive, right? Predictive behavior on 1000%. And um I guess the macro element of this is that you are getting exposed to almost exactly what makes you or what what what your sort of mind or the state of your mind is at any given time that you're and that's insane if you think about it, because you're just getting you're getting reinforced with the same this there's this interesting thing on Facebook where they um you know how people were like they were in their own filter bubbles or whatever, and maybe that's what got Trump elected initially or whatnot. Um that's more so true now than ever before. Yes. Like, but the filter bubbles aren't necessarily permanent, they're more ephemeral. They last for a little while and then maybe you move on to the next thing. Right, you you you happen to engage with something else that's a good thing. Yeah, it's almost as if you're like, yeah, a few days they last for a few days or whatever, and then your algorithm changes quite drastically. Yep. I mean, these are like it's so real time now. Yes, yes, that uh you are like your intent is almost non-verbal, it's almost emotional, and that emotional intent is captured by an algorithm, and then you are reinforced elements that either A serve your emotional state or serve your to validate your emotional state, and or um it never is like about contradicting it or questioning it.

SPEAKER_05 8:02

Yeah.

SPEAKER_02 8:02

It's almost as if it's like supporting it. That's right, yeah, it's almost like AI. Right, like the brother's in AI, right? Like it's a very similar thing. It's like, well, it's just telling you exactly what you want to hear.

SPEAKER_03 8:14

Yeah, it's like it's like the whole, like the meme writers, like it's like, oh, it's like, you know, uh uh motherfuckers say they hate subweb. My brother in Christ, you made the sandwich. It's like it's like, oh, like, like, why is the algorithm showing me this like, you know, Judy hops from Zootopia to, you know, like like weird like feet picks or some bullshit. It's like my brother in Christ, like, how do you think the algorithm is trained?

SPEAKER_02 8:32

I don't think many people are sophisticated enough to actually understand how this stuff works. But it's so I mean, even the makers don't understand how it works. Totally, yeah. Like, if you were to understand what uh, you know, like Nikita has to deal with the X algorithm all the time, I guess. And that's open source. Yep. Um, but it's very like the nature of it is very difficult to understand the code.

SPEAKER_03 8:54

And you know, like I I I do tend to believe that he is not in control of the algorithm, actually. Uh like like I I I mean I feel like with with uh with something like X, uh, you know, spoke about this with Rune. Uh and it's like it's like Yeah, what was his perspective? I mean, similar to mine in that I think they they are looking at particular metrics and just, you know, simply dialing up and down based on Totally. Uh uh Well I think I think Nikita has actually said it outright, it's like not non-regretted user minutes. Totally. And so, you know, they're just they're just twisting and turning knobs, dialing stuff up and down based on that on that metric and you know, user retention, uh like like new user growth. I think they're I'm I'm sure they're focusing on that a lot more than perhaps the old Twitter was. Um I would actually say that the old Twitter was probably actually harder to get into than it is today, like like harder to you know get a get an initial following and build momentum. Um well you couldn't be as crazy. Well, there's that, right? And obviously with the craziness comes with you know in in an an an increased likelihood of actually breaking containment and you know becoming a part of that actual 4U feed, whereas when it was siloed previously in sim clusters, like I mean you'd build your entire following, your initial following from just like replying to people that you like, right? Becoming a part of that sim cluster.

SPEAKER_02 10:15

I wonder, like, the original Twitter was uh very much a um credential element, right? Exactly, exactly. Like there was a huge element of, hey, this person is XYZ and therefore they must be important and therefore I need to follow them. In fact, I think the algorithm probably suggests, like not probably, I think it did suggest very prominent individuals to follow. Of course.

SPEAKER_03 10:38

And then you'd have the um you'd have one of my favorite parts about that was when Barack Obama had like fucking, what, 100 million followers or something like that. But then the the the the the really the novel behavior that you would get out of that, which was which I found hilarious by the way, is that um it cut it was like a that was a great level because you'd have accounts like Sex2 Inventor, who's like 50 followers, you know, some anonymous account, um they they might have an actually actual informed take and they're getting into arguments with like the chairman of the New York Fed or some shit. Yeah. And it's like and then the chairman of the New York Fed is replying to Sex2 Inventor and they're having like a well-thought-out discussion in the comments. I mean, that's that's the beat doesn't happen anymore because of blue checks.

SPEAKER_02 11:19

Uh I mean, in some sense, um you know maybe that that is actually like now you're you're focused less on you're focused less on in credentials in in general. Like the fact that a a non with like, I don't know, a weird bio gets traction and people follow and a lot of you know, weirdly enough, a lot of important individuals are following these people because How do you think I sell mods? There you go. Well, apparently like Sam Moldman has a alt or something. Uh no pun intended. Or something. Or something.

SPEAKER_03 11:53

Very important to put the or something.

SPEAKER_02 11:54

Uh yeah, exactly. I mean, I wonder how many famous people have uh accounts that are um, you know, anonymous or pseudo-anonymous that nobody ever figured out, and there's such good op sec associated with that that they're effectively driving the zeitgeist when they're actually a very important individual.

SPEAKER_03 12:12

Totally, totally.

SPEAKER_02 12:13

But they're not driving the zeitgeist through their sort of persona that is very public, but yeah, more so private. I I don't I don't know how the state of that is, and I don't know how many people there are, but I suspect at least a few. I I don't know. Um, or at least there have been rumors or something like that. Totally.

SPEAKER_03 12:27

Well, I mean I'm I mean I'm talking to a gray blob right now.

SPEAKER_02 12:30

There you go. I mean there's some there's some beautiful mystery and um uh around like the human mind not knowing like the the the the a critical part. I agree and there's like the essence of a human is that you encounter an individual and you understand their facial structure and you're you know you're looking in their eyes and then you understand how their balance and rhythm exists and therefore you know you have a better gauge of their trustworthiness or their all these things fade away drastically on on on this. Although, you know, on on the counterpoint is that if you know an average tweet is what like 50, 60, 70 words or something like that, maybe longer now, I guess. But those 70 words communicate more than anything like it's crazy how much like just not the the sort of surrounding nature of those words communicate, like whether it's AI written and or whether it feels authentic or whether the story resonates. And it's literally a handful of words.

SPEAKER_03 13:31

It forces you to be precise. Yeah. Yeah, it forces you to be precise and very clear. And I think uh, yeah, I mean, I I always used to make the joke where it's like, oh yeah, like uh, you know, uh like like posting is actually an art. And of course, you know, as we both know, that is incredible cope for for being like addicted to an app. But you know, I think the the art of being precise, the art of being able to explain things like one is five, the art of being able to explain things in as little words as possible. I mean, it's uh, you know, I mean we're in Silicon Valley right now, we're in SF right now, like people love to yap. People love to yak.

SPEAKER_02 14:02

Well they have opinions on every fucking topic. That's right, that's right. Um which is incredible.

SPEAKER_01 14:07

Also, I think not that like bad.

SPEAKER_02 14:09

I think of course like there's this Dave Chappelle bit, right, uh, after 9-11, he was like, oh yeah, what gets Jaw Rule on the line. What does Zhaw think about 9-11? Like, I'd like to if that's the last person I'd like to hear what you're doing. If Twitter was around at the time, then that would happen. Exactly. Well, now everybody's Jaw Rule uh every topic is 9-11. Yeah.

unknown 14:34

You know what?

SPEAKER_03 14:34

I did I didn't I didn't think about that.

SPEAKER_02 14:37

No, I uh You'll get the transcript, buddy. Yeah, that's for sure. Um it's it's fascinating. I don't know. I look, I think generally I don't I don't really have um X uh something should be this way or something should be that way. Um I think I I just like to understand things as they are and why they are that way and what drove him to be that way and how they existed and um what what was the the underlying essence of why something is the way it is is is deeply fascinating. I think that's probably what my account really is to be honest. So you like like trying to figure that out? Like maybe the the incentive structures behind the state of what something is, whether how people behave, or why a product exists. Yes. Or why AI is behaving in such a fashion, or why um people enjoy uh being told exactly what they want to hear. Yes. I mean some of those are very obvious and some of them are non-obvious.

SPEAKER_03 15:39

Or why or why everybody uh in this in the city that we're in right now thinks that they're working the last job that they'll ever work.

SPEAKER_02 15:47

Right. Uh like why why why is there such a hysterical or hysteria around the permanent underclass?

SPEAKER_03 15:54

Yeah, which I'm sorry about that, by the way.

SPEAKER_02 15:56

I wonder how that term got coined.

SPEAKER_03 15:58

Like who who coined the term permanent underclass? So I didn't actually coin it. I mean uh I thought it was you. I feel so so I think um if there's any I don't like to take credit for bits because I think it's just it's one of those things in which uh, you know, say something, say say it enough, and if it proliferates, I think that's really, really funny and awesome. And um, you know, maybe in this case, uh I don't I don't know. I mean I I thought it was really funny at the time. But it's wonderful.

SPEAKER_02 16:21

I think what a what a that's craft. Permanent underclass.

SPEAKER_03 16:26

I I think I think I I added the whole sort of like two years to accumulate capital part. That seems reasonable. Uh yeah, because I do think I do think I I actually think Rune had had had mentioned the permanent underclass uh back in 2024.

SPEAKER_02 16:40

And then I rune is a pioneer of culture.

SPEAKER_03 16:43

Yes, yes. Well I I started roofing on it a little more just with um uh you know the idea of like AGI 2027. I was like, okay, we got two years, right? It was 25 at the time when they came out, and then the the New Yorker did an article on it, and um and so you know, naturally, naturally, you know, those of us like like like myself are just sort of like meme translators for the New York media now, which is which is great. Um, but I digress, uh, yeah, that's like it's like a real thing here. It's a real thing here, and and and and people have asked me on podcasts, uh, do you believe in the permanent class? I say like, absolutely not.

SPEAKER_01 17:16

No, why not? You not do you not believe in a permanent class? No.

SPEAKER_03 17:19

Why not? Uh it's just I I think you just need to step outside of San Francisco to realize that the real world is just, it takes a lot longer to catch up than those in the valley think. I think those in the valley, because we're we're mostly engineers, we're surrounded by software engineers, we're probably one or two layers of separation max away from a software engineer. Uh turns out that software engineering, if you index off of the pace of how that profession has progressed in the last couple of years, if you use that as an index for every other industry, you're probably going to think that white-collar work is basically over or soon to be over. Uh, I think every, most, most other industries outside of building software is are bound by regulation. You can't just go to a Starbucks, spin up a server, spin up a web app, put a Stripe link onto it, and start making money straight away. Accounting doesn't work like that, legal doesn't work like that. Uh, I think the the the real world and its infrastructure has a long way to go to catch up to the pace of software engineering.

SPEAKER_02 18:20

What do you think of the fact that the world like if you think that technology is accelerating um in general the dynamics of of everything with respect to human life, like then maybe the infrastructure or the sort of physical world or the regulations and the rules will I mean if you sorry to use this analogy, this probably is the terrible one, but it's like if there's a tsunami coming, that's gonna wipe out everything. It doesn't matter what it's what it's what's ahead of it. Like I don't think it will consider that there that's a building or that's a something else or that's a you know, whatever. It that it is agnostic of its uh path. And in some sense, I think if you were to think about the world in that realm, like it has the potential to impact everything. And also technology itself is so powerful uh in terms of how and especially married with the incentive structures of capitalism. Yes that is a that is a powerful force that I think most individuals like most economists don't really understand what a highly deflationary environment that is powered by AI making things more abundant and cheap would feel and look like and how it has second, third, fourth order effects.

SPEAKER_03 19:42

Yes. I mean I mean I mean but like you know, if we're if we're taking the deflationary scenario, I I think that ends up being I think that ends up being certainly bad in some respects in that we run into a debt-based economy. And so like, okay, what what happens what happens to all of that debt? It's like yeah, it's um, you know, I I think it starts getting a bit far out. You start getting into like c citrini territory, right? With the with you know with the you know perhaps more grim deflationary scenario, but um I don't think that report was all that bad.

SPEAKER_02 20:12

I mean the example usage I thought it was a great base. Yeah, the the the the example usage was kind of the thing that took away from the like using DoorDash was kind of a a weird element to it. But but I think if you were to use like I think it's just the uncertainty levels are so so high because this is very it's almost an unprecedented event economically, technologically. Like I don't think 10 years ago you would have been like, oh my god, you know, we're gonna abstract away software engineering to the extent that people are gonna just write English code and then therefore it's gonna work out of the box. Like even the most hardcore optimistic people would never have thought that that would have been possible. Yeah, yeah, yeah. Like these things are highly, I mean, maybe AI is a black swan event. Yeah, um, that humanity doesn't really understand very well. Yep. Um, and therefore the implications and the under underlying things that it could drive are are are indeed like crazy if you think about it. Like man, white-collar people impacted would have been this is a this is something that no one, I mean, no one would have been like highly educated individuals who are intelligent, they're good at their jobs, they're doing important work for society, they're economically such an integral force in terms of the GDP and the output, especially when it comes to software, because software is such a visceral zero marginal cost entity that you know created more wealth and value in the world than anything really like what was the last thing like people were rich from like resources in the back in the day. Now people are rich because they've created um things that a thing you can't. see other than if it's a screen or what optimization abstraction. I mean it's a powerful thing and it software has made people so I mean it is it is such a weird dynam it's not a physical thing. It is a uh it's highly it's what a what a fascinating time that is and I think that the fact that it was so difficult like for example dude engineering is actually very difficult. Like writing code is difficult. Yeah. Like it is a non-trivially like if you were a software engineer and working on a day-to-day and building products and writing code um you felt tired. Yeah like you were using some significant brain cycles and therefore that burned a lot of goddamn calories I suspect and you are very much um a a valued individual of society given that a first of all engineers hard so therefore it attracted individuals who are really good at what they do and they got paid well as a result of it. And um that being impacted is such a weird anomaly yes of the AI or AGI type thing I think like it's incredible. I I credit to the people who are saw this early on where they like Anthropic optimized for um building a great software engineer or coding product or model. And nobody's still really caught up on that territory.

SPEAKER_03 23:27

Yep yeah that was a giant rant so my bad no I mean I think I think even just going back to it I mean yeah the way you know even in in simplest terms the way that I see it is that the my my my basic take is that the real world just takes a lot longer to catch up than people realise. I mean I mean you go to any Fortune 500 company they're using Microsoft Copilot and allowed us clawed code. You know it is like one of those things I go back to Australia nobody's actually heard of clawed code. And so I I think the the the lag outside of San Francisco is what will you know a lag would probably prevent the permanent underclass right I mean I think I think there's they're they're generally again it's a pretty optimistic view actually is that there will be enough time for you know these things to to play out for for people to transition career for new careers to be to be made I mean fuck like like a like an ex poster is now like a legitimate career actually. It's quite ridiculous. You know you can I mean that wasn't that wasn't even the case in 2022 wasn't even the case in 2021 when I started um and so all of these sorts of things you know I I I just think that I I think that indexing off of software engineering is a mistake uh for those that are that are that are extremely anxious at which software engineers tend to be but you know I think I I think about it this way and I don't I'm not a doomer in the macro sense. Yeah I'm not I again I I do actually think that my take is like quite an optimistic one.

SPEAKER_02 24:52

Totally and I think that's a reasonable one as well I I I for me I think look if you if you solve code generation and you build a really incredible um I don't know if you want to call AI like it's like if you if you build a great software engineer AI I think everything downstream is highly impacted because code is is is action transcribed and that is an incredibly powerful force and it's so um it's the basis for for how and what we model as a career in some sense, right? Like if you look at the traditional responsibilities of an individual who is like a white collar worker, um I think almost all of those can be mapped down to code. Yep. I mean anything can be mapped down.

SPEAKER_03 25:44

Depends if you have the data. Right so that that's what that's what I would contend in other industries. I mean you go uh I mean the worst one would be something like construction right in which you don't even have you don't have the building permits in the cloud actually so you can't Yeah you know I think I saw his chart today where people are um uh like it they mapped out how much each of these companies is are effectively spending on data.

SPEAKER_05 26:04

Yep.

SPEAKER_02 26:04

Um billions and billions of dollars and there's so there's there's shops and companies that are popping up that are simply data collection agencies and or like you know I think there's a company that that sort of um you know tries to collect camera data for or like have people work cameras to try to collect like uh warehouses stuff.

SPEAKER_03 26:24

I I had a I had a friend who was paying people to uh to to basically set up Lego on camera.

SPEAKER_01 26:30

There you go but then again look there if there are economic incentives the data problem might be solved.

SPEAKER_02 26:35

Like if you think about it I if we we have the ability to record visually and then now we have an AI that's able to understand visual context at like I was um I was shopping for a 90s party and I went to a thrift store and I took a picture of a jacket and I said oh hey does this fit the 90s vibe? Yeah and it did a big it it sort of understood that context in a way that I think even normal human would not have like it analyzed deeply about the color structures and patterns and the way that the the look and feel of it and the fact that there was like a little flare on the sleeves and stuff and you're like oh my god what the fuck is this is insane. Like even a very small thing and if that has that that sort of visual intelligence um and paired with like um the ability to capture I'm not worried about I think as far as the data is concerned like look that's a uh solvable problem and if you throw enough resources uh at it then I think oh absolutely motivation you'll be able to get it and now is that I mean Liam Fettis is trying this with periodic right there you go it's like like you know it's like like people yeah they're this is this is happening in real time.

SPEAKER_03 27:48

Again it comes back to my contention that it's like that it just takes it takes more time than people think and it's it again it feels the the the acceleration feels so much so in SF because everyone's either a software engineer or one separate.

SPEAKER_02 28:01

Do you think this is like um do you think people in San Francisco understand it less or the people outside of San Francisco understand it less in some sense? And the what I mean by that is that oftentimes when you're deeply in something right um whether it's the zeitgeist or the culture or AI or whatever you have like a very almost a you know a weird perspective on it because you don't get a you don't get it outside in you get an inside out and I think I think that leaves a lot of blind spots for individuals. 100% yep the the experts are not as we know from a variety of different scenarios it they may not be name them not. Yeah exactly um so I think generally I maybe maybe it is uh it is the case that like I mean look nobody really understands this as well and we're all pontificating and turning figuring figuring shit out but yes overall I I feel like maybe in the the macro point I want to make is that uh damn it could accelerate it could be absolutely like it could be even faster than I mean Jensen said said that you know obviously he has an incentive structure that AGI is already here. Sure yes like nobody can define what AGI is like nobody can understand it. Like even the people that are working on it don't really freaking understand um how the models work. I think I asked somebody on Twitter one time it's like wow why does the why did why did uh you know Chat GPT have this X it's not X it's Y pattern. And is it very prolific in language? Uh is it and that very well may be the case I although I'm not sure that that I've come across that as much. I my mind kind of works that way weirdly enough but um the fact that it was like so prominent and so I couldn't get anybody to to explain as to why that phenomenon exists. Yeah like I don't know if I maybe I'm an I'm an idiot or something I don't know but like sure maybe there's a a lot of the the pre-training and you know whatever.

SPEAKER_03 30:09

Yeah I I I would say I would say that I um I'm friends with quite a fair few AI researchers and yeah they're they're noticing emergent behavior all the time. I mean I think it's really incredible. I think it's totally totally fair to say uh totally fair to say you know as somebody who is uh you know half employed 27 uh on on the internet uh just just posting incriminating thoughts all the time uh it's cool cool for me to say what uh we are quite quite a disclaimer but I figured I'd just add that wonderful wonderful I mean unless you've got any other thoughts I feel like that's a pretty decent place to place to wrap I think so because that's that's um I think we just literally require a shit ton of editing I suspect. You might be surprised but maybe you'll take maybe you'll take a listen you'll be like you know what fucking run it.

SPEAKER_02 30:52

So yeah we go you know I've never really spent a significant amount of time listening to my own voice.

SPEAKER_03 30:57

Neither but that's a that's a that's a different problem. I mean people pick people can probably tell because I barely edit this I wouldn't the funniest part is like uh man imagine hearing yourself sound like a moron in third book my world king exactly buddy I'm a moron well been listening to this uh very special episode of members of technical stuff we'll go ahead and see you on the timeline again real soon incredible