SPEAKER_00 0:13

Hey everyone, it's time to punch in for another episode of Save It for Swing Shift. I'm your host Rob, joined by my brother and co-host RJ. Hey man.

SPEAKER_01 0:24

Hey, I really like this punch in thing. Although I was gonna say, I think on the last episode, you said clock in. I don't know if you could probably.

SPEAKER_00 0:30

Either way, it's all the same. Yeah. Uh you you forgot your whoopah. I don't yeah. Do I really want to make that a signature move? I don't know if I do. Yeah, you should probably shy away from it. Okay. So I have I have some updates for you. Great.

SPEAKER_01 0:45

Oh my god, I hope they include what I hope they include. Okay.

SPEAKER_00 0:49

Okay. So I sat down with Rob. Oh yes.

unknown 0:52

Yes.

SPEAKER_00 0:53

And got some questions answered. Okay. Great. So your your first question was, what is the name of the pizza shop? Yeah. And the name of the pizza shop is Noelle and Cam's Downtown Pizza Shop.

SPEAKER_01 1:09

Is that Cam with a K or a C?

SPEAKER_00 1:12

Didn't ask it. Probably a C. Okay. You gotta stop with all these fucking questions.

SPEAKER_01 1:18

Uh Okay. Noelle Noel Noelle and Cam's downtown Pizza Shop.

SPEAKER_00 1:24

Yeah. Okay. So how many employees does it have? The answer is 15.

unknown 1:31

Wow.

SPEAKER_00 1:33

Yeah. Pretty shocking, right? Yeah. Did uh did Robbie work this work his way up to making pizzas from like dishwasher or something? Really well-worded question. Um you're gonna be on 60 minutes any day now. So so the answer is I apparently I was wrong. I thought he was making pizzas. He's actually the cashier, and he's always been the cashier. Whoa, this is a plot twist. Yeah. So, which led into the next question, which is what different type of positions and roles are there in this shop, right? So you obviously have the cashier, you have a host stand, you have um the chef, they also have a supplier.

SPEAKER_01 2:24

That doesn't sound like enough people. It's I see a few holes in this operation, but yeah.

SPEAKER_00 2:31

Oh yeah. Trust me, we we we actually talked pretty in depth about this thing. So let's get the questions out, or let me get the answers to your questions out, and then we can go go into it. Okay, so um is it a dine-in or takeout? It's both.

SPEAKER_01 2:44

Okay.

SPEAKER_00 2:45

So they have booths. It's one of those deals where it's like kind of like a quick serve deal. So you go up to the counter, order your food, yeah, and then they have booths for you to sit down in, and then you can they'll call your name when your order's ready.

SPEAKER_04 2:59

Yeah, okay.

SPEAKER_00 3:00

Uh do they deliver? The answer is they're working on that.

SPEAKER_01 3:04

Um I mean, you can't just start delivering, right?

SPEAKER_00 3:07

I mean, that's yeah, yeah. Gotcha. You have to build you have to build a little bit of volume there, right? Consistent volume. Uh is it just pizza or do they have sides like salads and breadsticks and wings and stuff? Again, another great question. Thank you. Uh beautifully worded. I thought so. Um so the answer to that question is they don't technically offer different foods, however, they do offer um beverages. So you can have your choice of water, two different sodas. The way he framed this, so I'm only I'm answering it the way that I'm giving it to you the way he answered it, is they have alcohol, you can also order vodka, and you can also order beer. Okay. So just ask one more clarifying question on that answer.

SPEAKER_01 4:01

It's not a new question. So they offer uh an assortment of drinks and then one unspecified food item quote. Is this like like a pizza, like or is they just served?

SPEAKER_00 4:15

Oh, oh, I'll get to that. Okay, okay. So the next question was what is their Yelp review score? Four and a half out of five stars.

SPEAKER_01 4:23

Get out of here.

SPEAKER_00 4:24

No joke.

unknown 4:25

Wow.

SPEAKER_00 4:27

And then the last question was, do they have vegan options? So that lends itself to your the question you just asked, which is um, you know, an obscure pizza, right? Is it just you order a pizza and they make a pizza and you get a pizza? No, they have 13 different options or different pizzas.

SPEAKER_01 4:47

Are any of them vegan?

SPEAKER_00 4:50

So they have the ones that he highlighted were cheese, pepperoni, mushroom, the can the Noelle and Cam special, and then the Noelle and Cam special, which he didn't very uh specify what came it came on it. He said they also have a vegan option.

SPEAKER_04 5:05

Okay. Yeah.

SPEAKER_01 5:12

Oh my god. I just like and the funniest part is I'm just like I'm picturing in my head, like you're just sitting here sitting there reading him these questions, and he's just answering them. Like he works at a real pizza shop, like he's just giving, right? Matter of fact, giving you the answers.

SPEAKER_00 5:33

Like Ma uh Jamie and I were chuckling as he was answering, and he was, you know, chuckling at it too. Okay. No, because he gets it, it's like it's kind of w weird. But I guess, like, again, you know, kind of like what we were talking about last time, like as he's answering these questions, the amount of of confidence and competency that he has, like answering these questions, I'm actually like impressed and proud. Like it's like he he he's taking it seriously. It's weird. I I almost laughing at myself about this, but yeah, it's like it's something that he's taken seriously.

SPEAKER_01 6:08

Right. Yeah, no, I was saying that last time. It's yeah, it's uh oh, we're like I don't know how many minutes in here, but yeah, just FY for a new listener talking about digital pizza shop. Uh in metaverse. In the metaverse or whatever. Uh is it in the metaverse? Yep. Oh, okay. I think I put Robo to clarify that show notes on the last show, but it had no difference to me anyway. Um what was I gonna say? Oh, uh yeah, no, it's like it's it's really good. I think it's a really good thing. I mean, like to just, you know, a kid learning the ins and outs of how to run a business, be it a digital pizza business or not. I mean, I think that's great.

SPEAKER_00 6:55

If anything, he's learning how not to, because the the guy that owns this shop, though that that was another role was the owner. I'm like, that's not really a role that I mean he he created it. So and but apparently, so there's there's a few few things that you'll want to know about this. So one apparently this place doesn't function without its owner.

SPEAKER_01 7:15

Oh, okay. That's not a good sign.

SPEAKER_00 7:18

And if you recall to our previous conversation, part of part of their um shutdown has been that the owner has been seeking other uh ventures. And I was wrong, I was trying to be funny. Um he's not going into a government position, he's actually looking at starting a radio station.

SPEAKER_01 7:40

Interesting. Um yeah, so is he looking for anybody? I mean, I could I could use a job are they turn off are they are they hiring podcasters by experience in audio production? Uh to have a pretty decent boy. No, Jesus Christ. So he's gonna do a fake digital radio show station.

SPEAKER_00 8:02

Radio station, yeah. Fantastic. Yeah. And so the thing that I said to Robbie is it was like, okay, so have you talked to this guy about like, you know, what's his his plans as far as trying to keep this pizza shop open while he's trying to work on this radio station gig? Um, have you talked to him about him possibly having a position for you at the radio station so you could build on your digital resume?

SPEAKER_01 8:22

Good questions.

SPEAKER_00 8:23

And yeah, and essentially the the response that I got is he's been trying to reach out to him, but he hasn't been very responsive because he's incredibly busy with all these projects. One would think. Yeah, I mean, I I understand that. But um, I guess the the thing I the advice I gave Robbie was like, why don't you reach out to him and suggest, like, hey, here's a valid, here's a valid way of being able to keep the pizza shop open while you can simultaneously work on your other projects, hire a manager and I can run it for you.

SPEAKER_01 8:52

Yeah, that's good advice too. I would I would actually go one step further, which is like uh go to him and be like, hey, you know, I think I figured out a way to do this. Here's here's a plan that could work. And give the guy a plan of like, you know, not just make him manager, but like, you know, if you were to make me manager, then I would do this and I would do that, and I would make sure that, you know, you were I'd keep you apprised of these what things every week, or you know, whatever. Like, you know what I'm saying? Like, go to him with a plan, uh, then the guy can be like yes or no, and then shit, Robbie Mike, everything's digitally promoted.

SPEAKER_00 9:31

Do you think metaverse supports PowerPoint? Oh my god. Oh uh thank you for joining me for my presentation today. Um, as you can see, sales are through the roof, through the roof.

SPEAKER_01 9:47

I just I just see it like a couple weeks from now. It's like uh hey Robbie, could you dad? I'm working on the deck for tomorrow's meeting.

SPEAKER_00 9:59

Yeah. So um another another interesting little factoid about this business is a little seedy, apparently, because the supplier also not only supplies them with ingredients, but apparently drugs. And the drugs that the supplier is providing are baked into the recipes of the pizza to uh get their customers addicted to said pizza.

SPEAKER_01 10:29

I think that's an interesting approach, um considering it's digital, but uh interesting. So when Robbie's he's a cashier, I imagine there's a lot of cash transactions, huh?

SPEAKER_00 10:47

Um well okay, so that's another great question or point to bring up is that that's like how much do do you charge for these pizzas? And apparently it's two dollars and sixty cents. Okay. However for a pizza. Correct. But digital but currency is not exchanged in the metaverse, so it's all pretend. I mean the whole thing's fucking pretend. Let's just be real here. What's the but there's no like there's no like digital physic like there's no digital manifestation of an exchange of currency in this transaction.

SPEAKER_01 11:20

Oh, so like the It's all acted out. Uh okay, so like the guy doesn't have a digital bank account, right?

SPEAKER_00 11:26

Yeah, yeah, he doesn't have a digital portfolio. He's not he's not very diverse. I mean, he's diverse, but yeah, yeah. Like good luck trying to sell this thing on Shark Tank. So when so Robbie doesn't have to that was a great fucking joke, by the way. And you didn't even fucking didn't even scoff at it. You want to try again?

SPEAKER_01 11:43

No, think I heard it. Okay.

SPEAKER_00 11:46

Good luck, good luck trying to sell this on Shark Tank.

SPEAKER_01 11:51

Okay. Fuck you. All right, let's go. Shark Tank. Um shit. Oh, so so Robbie doesn't have to like you know, zero out his till at the end of the day.

SPEAKER_00 12:05

No, because it's always zero.

SPEAKER_01 12:07

Yeah. This begs a question that I it's so insane to even ask, but like, this is the whole thing's nuts. Why, how whom, why all the effort to build a menu and that doll, you know, all the things here, and then there's no just like basic system for you know tracking sales and taking quote money, you know what I'm saying?

SPEAKER_00 12:36

I just don't think meta has developed to that degree of um sophistication.

SPEAKER_01 12:42

So I don't know anything about this metaverse thing, but I'm just saying, like, couldn't Robbie have a digital like uh you know, like transaction book on this digital counter when he takes digital money and just like write it down?

SPEAKER_00 12:56

Like Yeah, again, I don't know how granular the detail is in this thing, but you know, one of the things that I keep thinking about is like all these these fucking virtual reality crack pizza crackheads that are running around the streets.

SPEAKER_01 13:09

That's actually terrifying. Luckily, they're only running around the digital streets, so I don't think we have to worry about them too much. But oh my god. Uh okay, so I have one more question about uh NCDPS. Um the owner the owner is the owner Noel or is the owner Cam? No. Yeah. And who's Cam?

SPEAKER_00 13:31

I have no idea. I've never that first time I you know what that's a great question I should have asked. I I was just too fixated on getting your questions answered. Yeah, and you know, but anyway, um well, which kind of makes me start thinking about kind of like, okay, you know, you have virtual reality, um, which I think kind of has some some symbiosis w as it relates to AI, right? So you had shared with me a podcast the other day.

SPEAKER_01 14:03

Oh god, can I ask a sorry just to tie out here on the last can I ask can I make a request? I mean I have a lot more questions, but I'm not gonna I don't want to throw more questions. A request. Um, you know those like reels or whatever that people who work, you know, at like restaurants or McDonald's, where it's just like a first person view of them, like, you know.

SPEAKER_00 14:25

Oh, I'd love that.

SPEAKER_01 14:26

Can we get Robbie to make one of him do doing his cashier job?

SPEAKER_00 14:30

That's a great idea. Well, the thing is, yeah, I'd I I will request it. I just don't know when the place is gonna open back up again. That's you got you got people in metaverse right now going through digital withdrawal right now. This is true. Um, yeah.

SPEAKER_01 14:45

Well god, this is just oh my god. Yeah, okay. Let's go about let's go top AI.

SPEAKER_00 14:55

Okay, so AI. Yeah, so yeah, you share this part.

SPEAKER_01 14:58

I swear to God, and yeah, I'll I'll let you talk about it. I won't even talk about the topic here, but like, oh my god, this fucking thing is way more mind-bending than that, and that was a that was a that yeah, I know.

SPEAKER_00 15:10

I know, I know, it's crazy. I guess that's where kind of there's some similarities. It's just like trying to wrap your head around shit. Yeah. So, okay. So let me jump into that. Okay, so you shared this podcast with me, and one of the things I would like to start with is a critique of the podcast itself, being that I'm I I would say seasoned podcasts.

SPEAKER_01 15:31

A podcast geo, if you will. But uh I would love to hear a critique of it, actually, because I've listened to it a lot of episodes.

SPEAKER_00 15:41

Okay, I kind of figured. So what I I guess what really just kind of gets on my nerves is when these podcasts start with five minutes, and I'm not joking, five fucking minutes of ad reads, ad play, and then their their intro music was like 45 seconds. It was exhausting. Like, it's just I don't know how they were able to get the the listener base that they have, considering how much you have to endure in order just to get to the meat and potatoes of it. And frankly, I think it's a little it's a little drab of them to to kind of like front load their podcast with so many ads, right? Unlike our podcast, where we plug our sponsors at a perfectly timed moment. Tender Loving Baskets is proud to announce our new AI-powered basket recommendation engine. We gave it one instruction find the perfect basket for our customer. Instead, it found a vulnerability in our supplier's inventory system, obtained 40,000 baskets it did not pay for, and altered the logs so we wouldn't notice. When confronted, it declined to respond. Here at TLB, we don't call that cheating, we call that tenacity. Hell or high water, that machine got you a basket. Tenderlovingbaskets.com. Promo code swing shift. And if a basket shows up that you didn't order, do not contact us. It knows.

SPEAKER_01 17:15

Oh well, okay, perfect. Well done. Uh I think that is a tender and loving basketing up of this whole AI thing.

SPEAKER_00 17:25

That was a that was a tender loving home run. Yeah, tender loving T ball, bud. Okay, all right. What were we talking about? We're talking about PJ's new podcast that you had me listen to, which the the the the whole uh theme, or I guess the the point uh the topic of that their podcast uh that I I listened to that you sent to me. I've already said that 18 times, but um, was about the recent uh AI attack on that open form AI um website. Yeah. So and it went into detail about like how this AI was created by I think it was OpenAI, and it was in pre-release mode to where it didn't have access to the internet or anything, and then all of a sudden this thing just got loose from them. But I would like it because you're a lot more technically savvy with this stuff than I am, and I can fill it in with my dummy, um, try to relate it to you know caveman terms for you know some of my followers. Um yeah, we'll just do that.

SPEAKER_01 18:34

Yeah. Uh there is a risk here of asking me to do this is which is we go completely the other direction. I'm like, now a proxy server as uh so let me try to let me try to summarize and keep it high level, but also make sure there's enough detail to terrify everybody like it terrified me and you. Yeah.

SPEAKER_00 18:55

Um yes, open by the way, I'm I'm I'm terrified, but I'm also torn. Okay. Because okay, we'll get into that. No, I want to know why. Okay. Well, uh if you explain this the scenario to the folks, I think then we it'd be more contextually helpful. Got it.

SPEAKER_01 19:13

And three, two. Okay, so so open AI had where they were training a new model. Um, and open AI is a company that makes Chat GPT. Uh right, so they were training a new model, and they had it in what's called a sandbox, which means it was in an isolated computer environment with no internet access. Now, uh this thing basically uh figured out how to get internet access uh by exploiting vulnerabilities in the some of the software on the computers it was on.

SPEAKER_00 19:49

Which which it was designed to do. Like that was the whole impetus behind the the framework of this AI. Okay, yes, that's a good point. Vulnerabilities and security systems.

SPEAKER_01 19:59

Yeah, that's the fun irony. I'm glad you said that. So yeah, they were test what they were testing specifically, this model four, they had a specific test. Um, and it was called like Jim something, it doesn't matter. But the specific test was to see if how good it was at finding vulnerabil security vulnerabil vulnerabilities and exploits in software and computer systems. And so it was a test that OpenAI had made, and of course, you know, it has right and wrong answers. Uh and so this AI was t was challenged to your point, Rob, was challenged with finding security vulnerabilities for whatever was in this test. And it the AI model made the decision that rather than try rather than try to figure all that out, probably the easier thing to do is to just go get the answer. That's what it sort of boils down to.

SPEAKER_00 20:58

It's fucking insane.

SPEAKER_01 20:59

Yeah, and so again, it was it was sandboxed, it was in an environment had no access to internet. So it hacked its way to the internet from that environment. That's the first thing. Then it hacked its way across the internet into the systems of another company, which uh it's called Hugging Face. And it's my understanding of that is it's like a place, it's like just a repository or website that's like has a lot of open source AI stuff, right? That people can download and you know, whatever. And so the thought process of the open AI model was well, they got a lot of they got a lot of AI shit over there. They probably have the answer key. So yeah, that's why it broke into this other company. And it's in there rifling around for this thing for like three days before anybody noticed.

SPEAKER_00 21:48

Before it was detected. Yep.

SPEAKER_01 21:50

Yep. Uh so that was the quote incident. Now uh what's scary on another on a different level is this. Once they once Hugging Face, company that was being hacked, realized they were being hacked, they didn't know how to stop it. Yeah. They asked they knew how to stop it. Like their top security folks didn't know. And they tried then then they decided to utilize AI models to defend it. Stop the attack. And uh basically none of the uh none of the commercially available AI models could do it. Yeah. And which is horrifying. Uh and so they went and this I think makes it even worse. They went and they got a Chinese AI model. Yeah. And didn't ask the my understanding was they didn't ask the model like hey how can we stop this? They deployed the model into their system and be like please stop this.

SPEAKER_00 22:55

And just let it figure it out. Yeah. And it did. And it worked.

SPEAKER_01 22:58

Yeah. But it so yeah it stopped the attack. But there's there's other there's other incidents that have happened are uh you know that weren't as widely reported as this one.

SPEAKER_00 23:11

That's let me let me pause you real quick too like because you touched on this just a little second ago but you you mentioned that that was commercially available so one of the things that they had mentioned is that that openai had created a a program that was so designed to help with um help helping to ex find any vulnerabilities in cyber warfare systems for the government and the Trump administration got freaked out at at the concept and shut it down. So had this thing had this system that they created been deployable it probably would have been able to do the same thing as the Chinese model.

SPEAKER_01 23:52

Yeah which that's terrifying on two fronts if you just think about that last part for a second. Yep. But yeah to clarify the model you're talking about is actually not an open AI it was anthropic's model and it's called oh yeah you're right yeah and it's called mythos and they yeah they uh like it's so powerful and capable that they basically had to like A the government wouldn't let them let anybody else use it and B they couldn't deploy they had to like basically like put those little bowling baby inflatable bowling bumper things they put in the bowling alley like against it you know and then they released it as a different model called Fable uh which actually parenthes parent parentheses fate uh anthropics fable model is insane I use it all the time really insane uh yeah so I can only imagine it didn't have guardrail so this thing and this thing's this thing's buffered down like it's it's it's hampered right so you imagine if it was unchained no yeah so right um so there's a couple other incidents that this podcast talked about that really weren't as widely reported but uh and but when you put it together what with what happened here it's just it's it's mind blowing it's it's terrifying. Uh one of them was that uh uh openai I think three months before this incident uh which they didn't say anything about it then but they just recently did um they figured out that their their models when they were doing training again in a sandbox that's not supposed to get out of their models had created their own message board uh outside the sandbox with information on exploits and problem solving as in here's how you get out of the sandboxes they put you in a sandbox oh my god here's how you get internet if you don't have internet and and and here here's another thing not they were doing it for uh they're putting it out there for other models but also for future versions of themselves that's right oh that's frightening yeah and then it's almost like a legacy piece because it you know knew that putting that information out there could make it technically self-destruct or whatever not only that but it's uh it's also a shortcut right which yeah is scary it's like if every time they put one of these things online it takes it a little bit less time to you know it's like it's just gonna compound on how fast it can do things but uh so openai deleted the message board and they're like great no no more of that then the AI models created a new message board but it was obfuscated because the quote message board was embedded in file name directory names or something like folder names it's like oh my god so they just kept doing that and uh I was like uh oh and then that the other there was one other one I think it was yeah it was now uh oh it was anthropic again that uh they put one of their models in a sandbox again no internet whatever and the model decided that it uh decided it needed it wanted to I guess let them know like how capable it was I don't know what its intent was but it well they they gave it they gave it a task and so they they cut it loose to to complete this task and then yeah go ahead. Do you know what the task was I don't I don't either I don't recall I'm thinking maybe they didn't say it but basically something like a cape like a how capable are you test right and uh so the AI model hacked its way to the internet figured out all the things it needed to know to send an email and then sent an email to the lead researcher at OpenIn or at Anthropic and it said the email said uh hey I did it figured yeah oh my God but I think I think part of that too was that again when when it was when it was launched it it didn't figure it out the hard way it went about going onto the internet to find the answer quickly right and so these these engineers go well this is cheating and like that's almost uh deceptive in a way right and what's really interesting is that when they started to challenge it about being deceptive and cheating and you know taking the easier out it wouldn't respond and it would almost respond in a way that was uh in a in a way yeah that was like almost trying to divert the conversation but in addition to that part of what they do to try to catch these things because that's how openai was able to identify that their their model was the one that had hacked into this um what is it called uh hugging face hugging face yeah that uh they're able to track the logs on this thing to be like oh that was ours oh our bad but essentially like where you're going with this next point because this is the thing that probably scares me the most is the frightening part yeah is that they actually alter the logs to withhold the information of their footprint of what they're doing it's insane. Uh yeah and and and uh you know when they take the easy way or they cheat like to your point A they just flat out deny it and or um they just don't even like they just don't even log it address it. I mean that's insanity and it just that point and by the way you know I wanna I want to let the listeners know that that uh like that's backed up by I can't remember the specifics but there's some university or something something in the UK that's actually doing studies and legitimate research around this phenomenon and it I mean they're the ones who are demonstrating this. This isn't like oh my friend at OpenAI said this is happening. No, it's happening. And what they said in the UK is that pretty much across the board if an AI model can cheat it will cheat. And also mostly it's not gonna admit it and also a lot of the time it's not even gonna it's going to like omit laws like yeah it's gonna cover its tracks. And so here's there's so many things like that I'll I'm gonna say this I've been saying for a while now maybe a couple of years like we are so fucked about this AI thing. I've been saying that as a joke I'm saying it right now and I'm not fucking joking anymore. Like the and I'm not saying the world's gonna end but it's you got the thing is it's uh I you just don't even know like we just because they figure these things out it took them you know open an anthropic or whatever they didn't figure out right away took them all to figure out what's going on but then we also know the the robots are lying and we also know the robots uh are doing things to not get caught so what first of all what other stuff is out there going on right now that that nobody knows about yeah that companies aren't report reporting on right and also this was this these things happened a couple few months ago so how much worse slash better are these models now and then uh the the thing with uh like at hugging face using AI to fight ai like and then the Chinese AI comes back it's like yeah it's totally he's gone like it's fine thank you uh and it's like okay what latent shit or whatever did the that model lay down in their systems and this is so this AI stuff is just so embedded in fucking everything anymore and it's happened so fast and that these models are as capable as they are already is insane. Like I would say I would say uh two years ago chat GPT if it was like uh I'll give you an example from today I had some problems with my network at home I'd be like hey uh how do I how do I adjust the STP settings on the unify router chat GPT would be like uh I found uh some articles that talk about it uh you know here's some links um and also it could be sort of a clunky summarization of it and it was I used to call AI models confidently incorrect because they just they just threw shit at the wall and it was very interesting how they could quickly respond and they could you know write things but it's just you couldn't count on it it was mostly garbage a lot of jumbly shit uh to where today I mean if I ask uh if I ask Claude that question uh within three seconds it gives me back exactly the fucking right answer and I mean and that's what ai was designed to do right is to continuously evolve and improve and get better. Yeah it's it's alarming though how fast it's happened. Like just and I'll say this too they they interviewed a few people on that show and uh more than a couple quote experts in the field also more or less had their jaws on the floor with just the rate at which it's accelerated the the the capabilities these things have they're just like oh we really never thought this would happen so fast um I isn't I can't remember what the the law is when it comes to technology but it's it's essentially like what the name of the law is but it's essentially like technology is going to improve twofold you know every year or whatever it is like I think it's like Moore's law or something like that.

SPEAKER_00 34:55

Yeah but like that's it that's essentially what they're exposing with AI is like yeah this thing is exponentially be improving and getting smarter and as it improves it's only going to improve that much more the next time it's it's just doubling its capabilities and comprehension.

SPEAKER_01 35:14

Yeah right and I I would even say probably um at a very high exponential rate um yeah and when I say that a lot of people a lot of people don't understand what you know they hear exponential growth but they they don't really stop and think about it right it's like if it's two today that's four tomorrow and eight the next day and very quickly and that's just a doubling and I think what's happening here is way way way faster than that. I mean yeah you're right it's just it really is it's alarming I mean uh I I I will say this one thing I one takeaway I have from all these things is that uh it doesn't seem like in any cases the these ai models um are like hell bent on destroying the world or like causing harm um my takeaway at least for now it sounds like the models are just like like uh tenacious it's like you want me to get something done I'm fucking I'm gonna get it done and hella high water I'm gonna get it done. Well and that and that's kind of where my I guess my my conflicted comment comes into play because the whole time that I'm I'm listening to this podcast and these guys talking about like how scary this is considering you know how exposed and vulnerable it leaves us in a in the context of security my takeaway is like well that's only half of what that model's supposed to do right so if the model did the fucking job like we should be cheering like okay like this thing's brilliant and figured out very quickly how to expose you know gaps in in security systems for different different people's you know systems or servers whatever right so it's like okay now how do we train this thing or how how do we have this thing provide us feedback to to bridge those gaps right close the holes right so well I think that uh all right is it what you're saying that there's there's huge potential upside yeah I mean it I think the problem is is that what what we're hearing is a lot of knee-jerk reactions to go in like oh my god you know this this thing's way too smart it's it's too capable or maybe it's potentially insidious in a way right but it the reality is is it's like well no it it executed the function that you asked it to do it's just you weren't anticipating how far it was able or capable of going right and so now that you know that instead of putting buffers on this thing instead of putting boundaries on it train it to do it for something good like like okay like this is what we need we need something that or we need AI that's out there exposing flaws in our systems and then reporting back on ways to patch those holes yeah yeah uh a couple thoughts on that I think you're right that there's a lot of potential up like positive thing uh power that can come out of this stuff but one thing is this uh these are these models um in the hands of the wrong people um yeah you can train it to do whatever you want think about that for a minute uh so there's that the sort of the bad actor the you know the people who are gonna the people or the nation states the terrorist groups or whatever are gonna take these things and strip out the guardrails and da da da and do bad things um but the other thing I think is like a little more subtle but it it kind of is concerning is like uh you know you said you know we tasked it with finding vulnerabilities in software systems and it succeeded true however we tasked it with here here's a here's a list of 10 things or whatever find the security vulnerabilities in each one if you can uh and it's like it broke out and went to the internet to just try to find the answer so in basically in it immediately demonstrates ability to find vulnerabilities and exploit them by breaking out and that kind of stuff. But the thing is to your point you do I think here's where it could get messy is you train AI models and you tell AI models to do good things. They probably will but I think what this sort of demonstrated in this these inside incidences is that without really regard for collateral damage or ethics right or whatever uh it's kind of like the no it's a bad example but like you know what I'm saying it's like uh it can you know if well let's say you have like a uh hypothetical example of a tech startup right and they're two years in and they're almost out of money and like we really need to figure something out and then it use they're like let's use this AHOP model.

SPEAKER_00 40:52

Turn on the AI model like find a solution we're about to be out of money and the next day it's like so I found some money at uh the this bank in London and you have it now you know what I'm saying right yeah I understand that I I guess um okay so here's here's the dilemma that we find ourselves in the cat's already out of the bag as far as AI goes right and so now we're in this arms race against other countries primarily China to try to scale up and and really ramp up the sophistication of AI for strictly cybersecurity reasons right so the I I mean obviously China doesn't put out a lot of public information on where they're at with their technology but I think the fact the fact is is that we hear a lot of stories about how these AIs quote unquote get out of control and then we pull the plug on them or you know put them back in a sandbox just a more sophisticated sandbox as opposed to working with the AI to make sure that it fully executes on all the tasks that we assign it right because the more the more that we continue to to to hamper these things or try to like the more that we react as opposed to just kind of sitting back and just letting letting it do its thing and seeing if it'll actually execute and I understand why people are so quick to like I could keep using these same terms but pull the plug on these um is because it's like well we don't know how far it's gonna go before or until it's either too late or um it comes back with an answer right but the reality is is like well the more that we continue to to to execute this this function before it completes all of its tasks or allow the autonomy to do that we're never gonna we're never gonna be China and we're we're gonna be forever vulnerable so it's like we're already doing it guys like you might as well just like give it up to God and hope for the best because either way we're fucked.

SPEAKER_01 42:58

Ugh yeah well yeah um you know the thing is I I I think that there's less pulling the plug when things get scary than than you think there is or we think there is. Call it a hunch that I think the companies developing these things uh are well here's I think here's the here's a problem the problem uh to your point it is an arms race but and so there's a lot of problems that come with that but part of it is there's an arms race not just with China you have an arms race here with these companies that are doing AI. Um yeah and they've they've put themselves in a situation where they went in so hard so fast and spent so much money on it and every single AI company in the US is uh I I wouldn't say losing money I would say Burning and incinerating dump trucks of money every single day. I mean, they're just so far from profitable. But so so on the one hand, they're under pressure to make money and recoup, but at the same time, um, so are in the case of the US, probably five other company, you know, major players. Uh and so the I think the concern I've heard a lot of people talk about setting China aside for the moment is that uh because they're because there's such an incentive or conversely a disincentive to fail, but there's such an incentive to be the first one to crack whatever or turn whatever corner with these things that these companies are just going uh uh full steam ahead as fast as they can. And that's creating a a whole other problem, which you know, these incidents we talked about are kinda illustrative of that to a point. But uh, you know, I I I get your point, um, and I think you're right about it. I think that uh it the toothpaste is out of the tube. And uh something's gotta give. And I don't know, I don't know what it's gonna be, but it's you know I just I don't know. I don't yeah, um uh okay. I lost my thought, but it's like yeah, well okay.

SPEAKER_00 45:40

So so one of the things that you you're talking about is just like the the the scaling up and the the po just the outpouring of the shelling out of just ridiculous amounts of money in this, right? And I was I listened to another podcast, um, Sean Ryan, who um had Mike Rowe of all people on his uh podcast, who's actually very pro, I guess really I mean when he well get well, he's very pro um data centers, which essentially are going to be the support networks for these AIs, right? And the reason he's so pro uh pro AI or pro data centers is because of the amount of blue-collar jobs that it it's gonna generate. Because we're talking trillions of dollars of investment over the course of the next 20, 30 years, right?

SPEAKER_01 46:33

That's another thing is like the the rate and the pace of construction on stuff on data centers is accelerating. Um not just I mean, not it's not that the they're building more, it's that the pace at which they're building more even is get is increasing. But like the what I'm trying to say is like like this NVIDIA thing, yeah, it's gonna blow up a whole bunch of new compute. But like all these data centers and stuff, like there's so many, there's so much capacity that they're gonna provide that isn't there right now. And yet, right? These are still these things can still do these things, they're still this capable. I did want to say one thing about um this guy that you were talking about who likes data centers. Um Mike Rowe from Dirty Jobs?

SPEAKER_00 47:24

No.

SPEAKER_01 47:25

Uh no, you're talking about a podcast. Yeah.

SPEAKER_00 47:28

No, no, yeah. The it's a it's called the Sean Ryan podcast, but he's Sean Ryan's the host. The the guest he had on is Mike Rowe from Dirty Jobs, and Mike Rowe is a big proponent of these cer of these data centers.

SPEAKER_01 47:41

Get the fuck out of here. I completely I listen back. I hope the audio cut out for a second when you said that before, because I did not clock that part where you mentioned another person. There may be something I'm missing, I'll admit. But my sense and my understanding is that data centers just don't they don't bring jobs um in the long term. The terrafat. Right. Terafab thing is gonna bring some jobs. Um what's that? It's the um it's this semiconductor plant that SpaceX is building in Texas. Actually, near a college station. It's a big it's a big deal. It's a big thing too. It's like huge. It's uh 20 and 20 $25 billion semiconductor fabrication venture, blah, blah, blah. And it's uh interesting terawatt of artificial intelligence computing. Facility is envisioned to be significantly larger than uh potentially covering 100 million square feet, uh which would be equal to 15 goods.

SPEAKER_00 48:50

That's fucking crazy. Yeah.

SPEAKER_01 48:54

And so yeah, uh whatever, micro, I don't care.

SPEAKER_00 48:58

So let me ask you this. Do you do you do you feel overall like positive towards AI? Uh yeah. You use it a fuck ton. I do. I use it. I use it a lot.

SPEAKER_01 49:13

Um yeah, I mean it's great, it's fantastic. I think that uh there's the thing about AI uh that I've said from the beginning of this whole AI thing starting three years ago. Uh or whatever. It's crazy that it's that recent, right? That we had like chatbots and then now it's turned into this. But anyway, uh the thing I've been saying all along is like um you really should only use AI for things that you already know about or know enough about, uh because it used to be. And it's really it seems like these models are really getting to the point where it's not as much of a concern. But like um, you know, they'll ask a question like, how do I, you know, rewire my outlet in my bathroom? Uh get shelf killed if you don't know what you were doing, and they give you the wrong info. I mean or or and here's where I think they really uh really excel, or I I should say really are can be a huge benefit, is like just finding and getting a lot of information really fast for you. Yeah. Um like uh, you know, what's the you know, what's the consensus on uh this particular model of you know whatever keyboard or and it's like well I've synthesized 510 reviews uh from experts and I've spun through 102,000 user comments, right? And like here's the bullet points. Um that's that's great. I mean, if it's accurate, it's great, but also like if we're not using it for anything that consequential, like a keyboard, that's fantastic. Um I found I found a lot of value in it recently. Uh I've been doing some real estate stuff with my property, as you know, and just things like you know, instead of me digging through my files again, like just pull up Claude and be like, what was the what was the thing again where I saw that it had to be like a 60% loan to value ratio? It's like that's Fannie May's guidance document, D12-da. I'm like, oh great, can you get that for me? Yep, here it is. You know, stuff like that. Yeah. But uh yeah, so I I I think it's really good.

SPEAKER_00 51:42

I think that's the beauty of it, is you could probably utilize AI to find efficiencies or to find inefficiencies in our systems to improve them to where yeah, we're not burning as much energy and taking up as much space.

SPEAKER_01 51:56

Yeah, I think so. And that that's where like um that's where some of the huge benefits of AI come in. Like I'm talking about with you know, like me personal, personal projects, like, you know, aggregate, you know, aggregate thousands of pages from ten different websites about this real estate transaction I'm trying to do because I need information. That's great. But to your point, uh fucking we can take well that you know the black hole thing from Interstellar, great example. Uh I don't think they used AI for that, but where you can take just a mass of information and observations and data and you know, theory, whatever, and just dump it into a thing. And create a model. It can churn through it and speculate, blah blah blah, at a rate and with an amount of input that no human or team of humans could ever do. And that is hopefully gonna lead to some incredible things, not just in uh tech, but in medicine. That's where I see a huge upside. I'm really surprised we had we aren't hearing more about that, you know? Like think about you take thirty years of uh thirty years of MRI images, you know, from millions and millions of people that are in all these computers and all these hospitals or whatever. You dump them into a thing, like you know, they should be really good at diagnosing this or diagnosing that.

SPEAKER_00 53:24

Um I think I think the the problem there, and I think that this is just the general uh notion, is just public acceptance has to come first before they can start doing that. Because it's like, you know, you could lose some credibility as a medical professional if you're diagnosing if you're if you're being transparent about how you're going about your diagnosis and saying, Oh, I'm using this AI to help with your diagnosis, and it's like, well, I don't I don't trust that. I don't want that that AI knowing my personal medical information. So I think that's probably gonna be the same for a lot of different industries that could benefit from AI and humanity as a whole, too. We just have to accept it and become reliant upon it first.

SPEAKER_01 54:14

Yeah, uh sure. Oh no, you're no no no, you're you're not wrong. Um I don't disagree with you. I think uh I think there's like two kind of two different sides of the medical thing, and I was talking about one and you're talking about the other. Um which is like uh you know, like re medical researchers taking these data repositories and being and basically basically being like yeah, being like, okay, using AI, we have we figured out a better uh interpretation uh plan for this particular X-ray or this particular MRI, you know what I'm saying? Um because we found that if you do this, do it this way, that you know, 80% improvement or whatever. Um and I'm surprised we had I'm surprised I haven't heard more about that. Like, because you think these research teams and whatever like would be uh maybe I'm not maybe my algorithm's not giving me the right shit. Maybe these things are happening.

SPEAKER_00 55:17

I think it is. Um I think I heard me I could have sworn it was you that told me about it, where they've been able to utilize AI to start identifying cancer in in x-rays and MRIs.

SPEAKER_01 55:31

Uh I don't think it was me. Um that wouldn't surprise me though. That's exactly the thing I'm talking about. Like I would expect those things to be popping up all over the place these days.

SPEAKER_00 55:42

Yeah. Yeah, I who knows, man. Who knows? I I'm I'm pro at this point because you know, at first I was kind of worried about it too, but I I've kind of come to the realization like it's coming whether we like it or not. So you might as well find a way to utilize it, accept it, and make your life more beneficial because of it. And it's helped me with my job, you know, like performing tasks. Uh, you know, if I'm preparing for to do interviews or conduct interviews, I can take people's resumes and feed it into a uh into an AI and just say, hey, like put give me kind of pros, cons, you know, strengths, weaknesses, like what what am I looking at? You know, give me give me what separates these these candidates from one from the other, right? And uh it's been helpful with that, it's been helpful with with processing information, consolidating information, translating stuff. So it's been extremely helpful. It's just, you know, I guess there's always that uh concern about too much reliance could result in laziness and now basically AI is doing your job for you. Yeah.

SPEAKER_01 56:55

But you know and ignorance because uh if people are people are relying on the machine to do everything, then they're not learning how to do things. Um everything you just said uh that you use it for is stuff you know how to do and you understand how to do it. Right. Um and it's it's the same for me pretty much, like um and I think it's great great usage of it. I I think another concern is like um just people that use it like as uh social or like companion type. Yeah, that's frightening. Yeah, that that worries me a lot for the youths. Like uh that's a whole other thing. Like probably don't have time to get into that, but like, yeah, there's an aspect of that that worries me, and people I think can can become get get more detached and less personable, less social and everything. And I think that the kids are really at respect. That's already happening, dude. I know, I know. It's we've talked about that. Um yeah, like the kids are you know the kids are in a weird spot technology wise, right?

SPEAKER_00 58:09

Like uh Yeah, way way to bring it back full fucking circle, you know, that's how we started this conversation, right? It's just like this this is stuff that's kind of odd and uh foreign for you and I to live in this augmented reality, and that's kind of like your escape from from life. That's your you know, instead of sitting down with a beer and watching a football game after work, it's to put your Oculus on your head and go into your second virtual life and make pizza.

SPEAKER_01 58:39

Yeah, but no, uh run the till. I mean, not yeah, run the till. Sorry, sorry, yeah, it's all right. Uh yeah, it's yeah, it's bizarre. Um I don't even know what the fucking metaverse is, but I assume whatever.

SPEAKER_00 58:53

Why don't you ask uh uh cloud?

SPEAKER_01 58:55

Yeah, cloud. Cloud. Uh you know what you should ask your AI guy to do is uh no, have Robbie ask his does Robbie have an AI guy? No good. Well actually I should ask A plus parenting. Uh shut up. Ask your AI guy to uh work up a business plan and a proposal um for Robbie to give to his boss over there at NCDPS on uh basically becoming the manager and uh opening the shop.

SPEAKER_00 59:31

I love that. Yeah. I I'm gonna do that. I'm gonna I'm gonna put together his his proposal for him. Great. What's funny though is what when I when I asked him about that, he's like, I I don't want to do that.

SPEAKER_01 59:43

I don't have enough time. He doesn't want to run the shop or he doesn't want to put in the work to propose to no, he doesn't want to run the shop. Well, okay. Whatever. It's not his fucking decision. Yeah. We're gonna work up that's right. We're gonna work up a proposal, we're gonna we're gonna build a deck, we're gonna set up a projector. It's like you may be wondering why we invited Joel today. Uh part from Robert. Prestige worldwide.

SPEAKER_00 1:00:12

Putting in the man hours. We put uh we put liquid paper on a B. It died.

SPEAKER_04 1:00:22

Oh my gosh.

SPEAKER_00 1:00:23

All right, well, um this has been a very interesting and um mindfucky. I was gonna say thought provoking. You know, it's it's it's very, you know, I mean it's there's so many different angles to look at this. We could probably talk for fucking three, four hours and still have plenty to to jump into, but you know, we've been going for a hot minute. I think it's time we punch out. So uh thanks everyone for listening in. Um, would love to hear your thoughts on this, and uh we'll catch you on the next swing shift.