AIForHumansAIJobLossOpenAIGPT6B

Kevin Perieria: [00:00:00] The AI

Gavin Purcell: job apocalypse might finally be here as last month's job numbers are showing the staggering impact of AI

Kevin Perieria: on the economy. Okay, and we brought it down in record time here. How about we talk about GPT six doing new science,

Sam Atlman: and there is a chance that GPT six will be a GPT-3 to four like leap that happened for kind of turing tests like stuff for science, where five has these tiny glimmers and six can really do it.

Gavin Purcell: Oh, and pray. Tell Kevin when do we get. GPT

Kevin Perieria: six. I mean, it's, it's coming soon. I'm sure. We just gotta get the star game. We

Gavin Purcell: are all doomed Kevin.

Kevin Perieria: And we are all moving on. Uh, sure. I mean, yes. I Google's sending chips to space. That's kind of cool, right? They pair nice with freeze dried ice cream. No, they are sending computer chips to space.

Gavin is right. Google's project SunCatcher is gonna use the power of the sun to activate AI models. Hooray for the forests.

Gavin Purcell: Thus open source model. [00:01:00] Kimmy two's

Kevin Perieria: thinking mode is out today. And Kevin, it's. It's actually very good. That's great. And also what might be very good is Apple Intelligence. Now that it's gonna be brought to you by Google.

We'll discuss that.

Gavin Purcell: Oh, oh, oh, Kevin, somebody actually reduced their hospital bill using Claude by like $150,000.

Kevin Perieria: See Gavin, it's not all gloom and doom. I feel better. Lemme just take a look at what the Internet's saying about this

Gavin Purcell: new AI Koch ad. Real fast. No, Gavin. No, no, no, no. Don't.

This is AI for Humans.

Welcome, everybody. Welcome, welcome to AI for Humans. We are your weekly guide to what's happening in the world of AI news and all sorts of other stuff. Kevin, this week, uh, we have some pretty scary, uh, news that came up just this morning. There are a lot of jobs being lost in America right now. And actually, yes, there's a, there's a lot of things [00:02:00] going on, but AI bring us down.

Gav Rizal, let's go. But, but AI keeps coming up in these conversations. Let's just play a little clip of what we heard on CNBC this morning.

Game: Steve Leman joins us with the latest numbers from Challenger on Job Cuts. What's going on? Yeah. Well this is like a, a clue from like a cough or a, or a sneeze or whatever, but.

Announced corporate job cuts. Andrew in the US surging past 1 million so far this year. With 153,000 new layoffs announced just in October, according to Challenger. That is the worst October since 2003.

Gavin Purcell: First and foremost, obviously we all know the economy at large is in a weird place. We are in a weird place in America.

We've talked about that before. But Kev, the most important thing for our audience listening here is that there are clear lines now being drawn to what AI is able to do in workplaces and how companies at least are justifying their cuts with ai. And I do think that's an important part of this. It doesn't mean that these are all because of ai, but a lot of these cuts are being justified BA by AI themselves, and there [00:03:00] is a lot of headlines that have followed up around this.

Uh, Derek Thompson, one of our favorite writers, has shown a number of headlines over the last, say, three months that have come out and shown this is a big deal. So, Kev, I, I think it's important to kind of set the table here first, but then I really do wanna talk about like what we think this looks like

Kevin Perieria: going

Gavin Purcell: forward.

Kevin Perieria: I mean, let's set the table. Then I, I think there's AI scapegoating, which has been going on for a long while. I think that's increasing. Now, if you want to put on a slight tinfoil hat, which you had prime delivered by Papa Bezos in less than two days, some companies. Are announcing massive layoffs because it helps preach their bag.

For instance, they are entrenched in building these massive data centers and proving how effective AI is, so they're slashing jobs and heads left and right to prove. Wow, that this tech is Actually, I didn't hear this. This is a new one. I haven't heard this. Story. Well, I, uh, there, there could be some, there could be some, uh, sure.

Fire with the smoke's. Fascinating. It's fascinating. Yeah. Yeah. Like if, if you're, if you're out there preaching that your data centers and your [00:04:00] systems are going to make your workplace exponentially more efficient, then maybe you cut a few people to show that, oh, hey, we've adopted it and look at all the savings we have here.

Uh, you know, that's, that's one of the things that's, uh, that's being balled about, but also like. The economy was, was well bad before we were talking about TPUs and inference and mixture of experts. I'm not saying that that is necessarily helping things, but. We were headed this direction.

Gavin Purcell: Well, and this is, this is the important thing for everybody, both in our audience and at large to kind of understand is that like, you know, there are always fluctuations in jobs up and down, and obviously different, uh, governments have different things that happen when they do that.

I think the most important thing that's differentiating here is the fact that AI is now capable of performing some of these things. You and I know as well as ever that that's not capable of performing most of these things. The other thing that's kind of tied into all of this that's really important is that the companies that have been propping up the economy in a lot of ways [00:05:00] when you think about this, not that we're trying to do, uh, the wonky walrus, uh, financial edition here, which is maybe this is what this is very gonna quickly turn into, but.

What we're talking about are the, what they're originally known as the FANG companies, Facebook, Amazon, apple, Microsoft, and Google. And now you're adding Nvidia and other companies into that as well. Those are some of the only companies that have been doing well over the last, say, six months because of this AI kind of boom.

And I think the big conversation that's gonna happen over the next, like say six months to a year is will those companies continue to do well? I think they probably will because I think you're gonna see AI keep growing in space. Um, but the other side of this conversation, Kevin, is for the people that are at those companies that are getting laid off.

What is the next step? Because I think that's an important thing for us to lay out for people that are listening to this. We have been preaching for the last two years in this podcast. Go learn these tools. Go figure out ways. Kevin, you have said many times, uh, work yourself out of a job because that will prove to your bosses and people that you work with that you are following this stuff.

I [00:06:00] want to just kind of lay out very quickly some thoughts on. How to figure out if your job might be at risk here, because I think it's an important thing. Like some of the things that are really about, you mean outside of just checking

Kevin Perieria: your email?

Gavin Purcell: Yes, yes. Well, that, that is like, you know, clearly your HR department is gonna get involved in some way, but, but you should think about like, what of your life is automatable?

If I were to give an AI like 80% of my tasks, could it do it in an okay way right now? And that's probably the answer is no for most people. But if you were to say like. Maybe it might be time to start thinking about what you do and how to change that thing a little bit so that you're less that way or try to get better at that automation

Kevin Perieria: side of things.

I had, one of my first jobs, capital j jbs, was to migrate a set of data from one spreadsheet Gavin, to an entirely different spreadsheet. And, and it was, it was working for a, it's amazing government television station and it was somebody's full-time job. To go and copy and paste and check and do it. And so I kind of went outta my [00:07:00] way and was like, well, this, I, I hate this.

I can't stand this. Yeah. The, the person that that left loved this job for me, it was mind numbing. So I figured out how to write a macro and it would take the data from A and put it to B. Same rules apply here. Yeah. Whether it's N eight N and you're writing automated hooks, whether it's, um, you know, whispering to chat GPT and watching a thing on CLO code so that you can ask the machine how to do it.

If there is, as you said, something that is easily automatable, you have the choice to try to automate that yourself now and free yourself up to do other things, or you can resist that. But that's, that's what it's coming for now. And that's if you, if you don't have the drive. To, uh, sort of, uh, like in invent a, a new position for yourself.

Unfortunately, I think that is the thing that's going to be eliminated.

Gavin Purcell: Well, and, and there's a couple things I'll say that I think people should focus on, uh, that are really important for you. And I think everybody, whether it's you or your children, that you can get them [00:08:00] to think about first and foremost, like original ideas.

Like I think it's going to be a while before an AI will be able to come up with original ideas, and that could be. Creative ideas. It could be workflows or it could be systems that the AI could then implement. So if you can come up with original ideas, you're going to be successful for a long time.

They're AI are very good at understanding how to follow something but not to come up with something. The other thing I think that's really an important thing to think about right now is like. What is the risk of using these things? Like are you somebody that understands very clearly cultural risks or workplace risk?

Meaning that you understand very well, like what it means to like try to implement something and what the human cost of that might be. That's something that's really deep and important I think, to people right now, and that's only gonna grow. So if you can think of those two areas as places to think about how in your position can you kind of like level up that?

That feels like a real pathway to like not only like making your job secure, but also like the opportunity of new jobs coming up and then, you know, finally. [00:09:00] Obviously there's going to be new jobs that arise from this. I, I think you and I both grade not as many as there were, uh, prior probably, but finding the new edges of where these things are happening, like I do think people who understand how to implement APIs, who understand things like MCP, even though it's a very technical thing, like if you know how to do that sort of thing, it really makes a big difference.

I mean, even in the work that you're doing right now, Kev, I think you. Have proven that at your job, the thing that you're working on is like, you know, very AI centric and you've brought a new skillset set to that company.

Kevin Perieria: Yeah. A skillset that I, I arguably don't have today, but most certainly did not have a year and change ago.

Yeah. Certainly not when we started this podcast. So there's never been a we we've said there's never been a better time to be an idea person, right? Yeah. Because now there's a, a bunch of tools that can help you whisper them into existence as well as. Professionals with domain expertise, whether it's graphics or websites and coding or whatever.

No, no. But you can get 75, 80% of an idea [00:10:00] out there. And my, my buddy it was doing his side hustle thing, he was very concerned that his main hustle, his job was going to be automated away and eliminated. And I said, well, what's the side passion? What's this, that the other, he actually teaches yoga. And I said, well, have you thought about doing that privately or doing it for, you know, business?

Can the other, can he do that? Every

Gavin Purcell: male who teaches yoga wishes they could do, which is what are you, take a guess.

Kevin Perieria: Okay, good. Gavin. Wow. Was not expecting that thought experiment from your side of the screen out. Well, we can cut that part out. That stays in. I want people to know that you are the toxic one here, but very clearly and apply this to whatever your passion, hobby side hustle is, if you have one and if you don't, then.

Invent one 'cause now's the time. Sure. Yeah. He, he was daunted Gavin, understandably, by the notion of like, Ugh, I gotta brand myself, I gotta build a website. It's gotta be mobile friendly. I've got to make a TikTok content and Instagram post and handle inbounds. And I said, yes. Uh, [00:11:00] here's a tool for this.

Here's a tool for that. Here's a blah, blah, blah. I showed him Sora. He made a cameo of himself. He started generating like fake commercials for his yoga practice that he thought was awesome, the funny's. And are any of them as good as they would be if he had the resources to hire the, all singing, all dancing agencies to do all that?

No. But he is going to get 80% of the way there. Yes. Which by the way, is enough to get ranked and get visible and start getting inbounds and start getting clients. And so he went from. Being, again, daunted by the notion of something new or taking on something else to almost liberated by accepting these tools and going after them.

So,

Gavin Purcell: yeah. Well, and I think the, the, the second story that kind of pairs with this nicely is the idea of Sam, uh, Alman went on Tyler Cohen's podcast, which I really do, uh, suggest everybody listened to because it's one of those, it's a pretty friendly Sam interview. There was an interview with Sam that was a little less friendly, came out last week, but.

Sams started talking about GPT six, which we know, we always were wondering when this was gonna start becoming a narrative, and it's here. But Kev, I think we should play a [00:12:00] little bit of this because one of the things to be aware of is. I think a lot of people out in the world, in this in our world think, well, these aren't gonna get that much better.

There's not that much more things that are gonna improve. He started talking a little bit about new science, which I think is an interesting thing so everybody can feel a little bit better about where these are going. I'm

Kevin Perieria: so tired of the old science.

Gavin Purcell: Let's hear.

Sam Atlman: Let's hear what this new science, this science Zero is.

If GPT three was like the first moment where. You saw like a glimmer of something that felt like the spiritual turning test getting passed. GT five is the first moment where you see a glimmer of AI doing new science. It's like very tiny things, but you know, here and there, someone's posting like, oh, it figured this thing out, or, oh, it came with this new idea.

Oh, it was like a useful collaborator on this paper. And there is a chance that GPT six will be a GPT-3 to four like leap. That happened for kind of turning tests like stuff for science where five has these tiny glimmers and six can really do it.

Gavin Purcell: So what's cool about that from an optimistic [00:13:00] standpoint, as we've been talking about here forever, is that yes, that would be amazing, right?

If suddenly, and we have seen a lot of people say. That GT five Pro particularly is bringing some interesting stuff out. If GPT six, which again we have no date on, we have no idea when it's coming out, or even if it's not gonna come out this year, he said for sure. So, but if that were to happen, you can imagine the world changes for better in a lot of ways.

The other, the other side of this is. The things that we just talked about that GPT uh, four and five are very good at, it will also get better at, so that doesn't change the aspect of like how it will affect the world. And your job, Gavin, we just said it doesn't

Kevin Perieria: come up with novel insight. It doesn't have unique thoughts.

Exactly. But if it's doing new science, it must be capable of doing something original. So now what buddy? Yeah. Well here's the thing. I've

Gavin Purcell: said this on this podcast many times is that. Yes, maybe it can come up with new ideas, but your idea will feel different because it's coming from you. Right. I do think that's the thing we have for ourselves for now.

Um, [00:14:00] it's not like, it's not a guarantee and I think there's gonna be so many conversations around. Human rights coming soon. I know that sounds crazy. It sounds like something it doesn't, you would've thought for a science fiction movie, say 10, 20 years ago, you would've been like, I can't believe there'll be a human rights movement.

But somebody, I can't remember who this was, but somebody recently said like the 2028 election. A lot of that conversation will probably be about AI because if you expect 20, it's right now, it's November of 2025. You know, you expect the election to really get rolling, say in a a 2027. Um, a lot of that is going to be AI conversation because assuming that that point, we get two more iterations on these models, that is a big deal.

But there's a lot of big things happening right now in the AI space, and we've always said with the show we would grow up as it grew up a little bit. And this is a big week, right? Like there's a lot of things people are talking about. Because all of this AI growth story, which is building data centers, all of that is tied into what we talked about in the beginning about how the reason why those AI companies are doing well is 'cause they're [00:15:00] spending money on this stuff.

And that's keeping going. But yes, Sarah Reer, the CFO of OpenAI was at, was on a interview. She said some words. That then were reiterated back to her by the host that basically said that she was looking for a federal backstop for AI companies. And what that means, essentially is from a, uh, layman's standpoint, is the idea that these are too big to fail.

And we've heard that before in different places with, uh, the Goldman Sachs of the world and the Lehman Brothers. This idea that the government should provide some sort of federal backstop to this. And there's been a lot of back and forth. She clarified that she didn't necessarily think that was true.

Later on, uh, via OpenAI. David Sachs, the uh, AI Czar, a famous Silicon Valley person who is now in the White House ecosystem said that we are not going to do this. There is no bailing out of AI companies. But Kevin, this just goes to show you the vast amount of money that these companies, especially open AI right now, is spending on infrastructure.

And I think that is this kind of like game we're playing right now between. [00:16:00] How much money can we spend? How much money are we gonna get out of that money we spend? Meaning like what is the value tradition coming out of that?

All right, wait. Some breaking news here. I actually jumped back on to record this. Um, Sam Alman has replied to a bunch of these things and, and first and foremost, he wanted to make it clear to the world that, uh, he does not want government guarantees for OpenAI data centers. This is a very long post on his Twitter handle, uh, at ama.

You can read it all, but he basically says in here, he does not want government guarantees on this company. He does clarify that there is a world in which governments themselves. May want to have data centers that would have some backing and would be able to be protective. But this, the main point of this whole thing is that clearly he is trying to get in front of the idea that he does not believe that government should be bailing out, uh, private companies.

There are three things in this that are pretty interesting. Um, he says there are at least three questions behind the question. Uh, first and foremost he [00:17:00] says, how is OpenAI gonna pay for all this infrastructure? Well, they are saying in this post, he's saying that they expect to end the year about 20 billion in annualized run rate.

That is a big business number, uh, and to grow to hundreds of billions by 2030. So that is a big part of this. Second of all. He wants to clarify that OpenAI, uh, should not be too big to fail that, uh, unequivocally he says no. If they screw up and can't fix it, we should fail and other companies can continue doing the good work and servicing customers.

And that also clarifies some of the comments that were made by Sarah Friar, the CFO, and then finally. Why do you need to spend so much now on infrastructure? And Sam particularly says is that they believe that based on the trends that they are seeing that AI use is going to scale up. And that right now they are unable to serve the needs of everybody plus, uh, models conceivably that they have, they cannot serve.

So. They think they have to scale quicker to make it worth their while. So there's a big, long, I think, [00:18:00] 1500 word post from Sam that you can go read right now that updates all this. All right. Back to the show. The one thing I'll give everybody kind of a, a good balance, uh, a good place to kind of jump off of is that a lot of people compare this to like the internet bubble.

One of the biggest things about the internet bubble was that. It was a lot of public companies that came out, went into IPO and had these huge valuations that made no sense. Right. And in this instance, it's mostly private money. It's mostly, you know, honestly, Google and Meta have a ton of money that they make on their own.

There's a lot of semi shady things that are happening there, but. In general, this is a private markets scenario, but when we go back to that thing we said before about the fact that AI is propping up the general economy right now, it's not a great place to be, let's just put it that way.

Kevin Perieria: No, and especially when you see the charts of like the snake eating its own tail where you've got, yeah, open AI getting.

X billions of dollars from nvidia. Then going right back to Nvidia and saying, okay, well we'll get some servers and some chips from you. Then Nvidia goes and does a deal with Meta, who then goes back [00:19:00] around and goes to Microsoft, and it's just like those 10 or so companies all feeding each other these multi-billion dollar.

Checks to sort of raise their valuation. Again, you could say like, well, if this is the most transformative technology ever, right? If this is gonna be exponentially bigger than the internet, we're all gonna have this intelligence on demand in our pockets and our washing machines and in space. Well then, okay, that makes sense.

Yes, I, this is a house of cards and cards is gonna win anyways and it doesn't matter.

Gavin Purcell: That's what I was gonna say is that Jensen Wong, Nvidia CEOs out there saying that China will win the AI race. And I think he's saying that as a, like a kind of a chug on to the American people to do this. I think the last point on this, in my mind, and we've said this on the show a couple times, there will come a time soon because AI is this transformative of technology.

You may not think so, but it seems to be believed by most people at large in in power. That AI in some form will become a nationalized thing, meaning that you will really start to see [00:20:00] governments folding in, in a major way. And right now, at some point, I saw somebody tweet this the other day that some of the stuff you're seeing, that circular, sort of like a investing thing in part that's like a soft, what they call a free market nationalization.

And that's a very complicated term that I'm not gonna try to get into here. But yeah. Because they are being supported in part by all these different companies and that the government is involved not putting a ton of money into it, obviously, but that the government is involved. That, that we might be seeing that transition.

Now. It is a very complicated situation, but I do think it's important as listeners of the show, and I know we often talk about how, like we, as we've putting your pets into Sora, but there are other moments where these are the important things to think about. 10 years from now, it's going to look different and, and just think about what that looks like.

I don't

Kevin Perieria: like it. And I feel like we have lost the sauce. 'cause I remember, oh, I'm, I'm wispy. I'm uh, I got a nostalgia tingling for when we used to just give AI monster milk and they would get drunk or we would. Personify a set of [00:21:00] bleachers to be a graveley voice. Sure. Different world back then. It was a different world.

Boy, howdy. It was a different world. Yeah. And it's not that we didn't see this coming, but now that it's here, I'm just, at least we're gonna do it in space. At least we're gonna just, yeah, that's the fun start. Yes. Let's talk about Sundar Paches Project Sun Catcher, and don't be left out. Star Cloud NVIDIA's effort to take, uh, AI into the cosmos, to reduce power consumption, to harvest the sun, and maybe save our planet.

In the meantime,

Gavin Purcell: yeah. This is a, this is a palate cleanser a little bit. It is a moonshot. As you know, Google has done a lot of these moonshots over the course of it. Waymo was a moonshot, right? Originally, and look at where we are right now. We have Waymo's that are driving around and. Yes, Kevin, they might be taking everybody's jobs, but they sure are fun to drive in and they're much safer.

Right? So there are upsides to this thing in this instance, what the plan here is and, and nobody's saying that this is gonna work yet. There's a lot of, there are a lot of questions, but the goal would be is that you could send a bunch of what they call TPUs, which are tensor processing units, which is kind of [00:22:00] Google's version of the GPU or or AI chips into space to get solar power and then be able to power a lot of the stuff and take it outside of our atmosphere.

This is a pathway to a much more, uh, environmentally friendly version of AI power because as we know, there's a lot of people out there who kind of point to the fact that as we build more of these data centers and as we build more AI power, that this is a very powery power hungry, uh, entity, the overall AI industry.

So, Kev, I think this is cool. Like again, we, we think that we are going to see advancements like this quite often in the near future, hopefully from ai. But this is Google kind of putting a, a unique, interesting. Thing into the world and saying like, Hey, this is an idea and maybe it'll work. Let's try it. I remember,

Kevin Perieria: um, Dr.

Diamandis coming on attack of the show, like, I don't know, two decades ago. The guy, yeah, yeah, yeah. And he was like talking about, we're gonna, we're gonna send these devices up to asteroids and we're gonna mine 'em for resources. And there'll be like little refueling depots and there'll be cargo [00:23:00] runs happening.

I'm like. I don't know that I believe it, but this is cool and I want this to be real. This seems like something Arnold Schwarzenegger was doing before his life got turned upside down in the eighties. That's right. Like total recall style, like it sounded very futuristic. When you go and you look at. The, the Google initiative or even the Nvidia one, with these renderings of these like four kilometer wide solar panels with data backends leading to these like shipping container looking farms, you realize one Elon's gonna be very wealthy.

'cause he's the one that's launching things up to space right now. Yes. Cheaply. Yes. Um, but two, like if we don't, if things don't explode and we don't trap ourselves on the earth and we can't escape. It's a pretty, pretty cool science fiction vision of the future, I think. Yeah.

Gavin Purcell: As long as we figure out how to keep people here and happy, that is a very, very cool world that we're gonna be entering in.

And, and you know, I think the most important way to keep us happy, Kevin, is you've gotta click that like and subscribe button. I know that's a hard transition into self-promotion and we've had a. [00:24:00] A deep, thoughtful intro here, but now's the time where we get shallow and we really start talking about us because it's time to start thinking about how to help us right now in this moment.

And the way you can do it for free is click that subscribe button, share this with a friend. Go rate us on, uh, podcast. Help yourself Gavin, automate yourself out

Kevin Perieria: of a promotion. Uh, the promotion aspect. Oh, I should, you know, I should,

Gavin Purcell: I should try to have like a little robot, Gavin next. That's right. Who will can throw that in here for now?

Game: Famous last words. Gavin, I've taken your job now.

Gavin Purcell: Yeah, please go share. Um, the link to our Patreon is in our, is in our show notes. We definitely use that money to buy some of the AI tools that we do on the show. Later on, I'll show you, uh, how I used, uh, SOA to do a very, very fun, uh, uh, prompt from a guy that I found on Reddit. Anyway, thanks so much everybody for listening and always sharing what we do.

Okay, Kev? The next thing up, there's some big stories out of Apple in that they're finally going to fix Apple intelligence. But do you know how?

Kevin Perieria: Yeah. Yeah. They're just gonna buy their way [00:25:00] out of it with Google. That's it. That's

Gavin Purcell: right. That's right. Gemini's coming to Apple Intelligence. There's a new deal.

And by the way, if you didn't know this, like. There was a deal for a long time. It might still be the case where like Apple paid Google for Go or Google a bunch of money. Google search. No, Google pays Apple to search. Oh, Google paid Apple. That's right. Yeah, that's right. Yeah, yeah. I'm sorry. Yeah.

Kevin Perieria: Yeah.

Google's given, or Apple's given some of that money back now, uh, which is fine. Look. Okay. A couple ways to look at this. First of all, Apple's spending a billion dollars for at least one year of a Gemini powered something likely. Siri rumored to launch, I think spring of next year. Um, they're gonna use a custom 1.2 trillion parameter model, but Apple is going to be able to host it essentially, so your data doesn't leak to Google.

It stays within the Apple ecosystem. Multiple ways to look at this, right? I mean, AI is hard. Apple is traditionally late to the technology party, but they tend to integrate it elegantly. Lately, I'm feeling the elegance has been a little lost with some of their updates, if I'm [00:26:00] being honest. Like, yes, I feel like that's been rough, but ultimately, like I as the user don't actually care.

Which model provider is being used? If they paid open AI anthropic, Google Meta do not. Do not care if data this not the best results secured. Yes. Yeah. And the experience is elegant and great and thoughtful. Awesome. Thus far, their implementations of AI have been lackluster from summaries to the emoji playgrounds and the like.

All that stuff has been really wonky. Very off. Yes. If they can just solve. Multi turn directions with Siri where I don't have to know the exact command and I don't have to wait for each thing to go and I don't have to kill 25 seconds while Siri says, yes, you got it. I just turned on noise cancellation mode.

You should be able to tell that the noise is being like, just do it. Do the function. Siri.

Gavin Purcell: Yeah, I will say I have a personal experience with this in the last week. That's been fascinating. We finally got Alexa Plus. Yeah. Do you have Alexa Plus yet? So, no, but

Kevin Perieria: I've [00:27:00] heard good things.

Gavin Purcell: You should get it if you can.

If you're out there and you got the email, definitely turn it on. I mean, I, if you're a privacy person, I understand. I am not so clearly. We all know that at this point in this stage, but here's the thing. It's good enough. It makes a huge difference like they have figured out. I asked it a question last night, I was trying to remember what I asked it.

I asked it something about, oh, I asked, this is the weirdest thing and this is gonna shows you my brain. I started singing that song from Oliver Twist, the one about food, you know, that song and what, and I tried to figure out like what was the thing they eat in that song. So Oliver Twist as a famous musical, but also like I in my brain just popped in.

It's gruel. So there's a thing called gruel, which is like a kind of a watery oatmeal. And I asked Alexa plus I was like, Hey, can you gimme a recipe for gruel? I wanna make gruel. And like, it kind of has a little bit of an attitude now in a fun way. And it actually answered the question. It did not go. On blog spot.com/gru says this and that is transformative.

It is not like the cutting edge, it is not opening eye advanced voice. It [00:28:00] is not all this stuff, but like the fact that they're in my house, the fact that they are a around where I can just ask a question. That is a big deal. And this is same thing with the iPhone. If your iPhone gives you the things you want out of intelligence, out of ai, yeah it will be a big deal.

And at this point it has not. So I think Gemini in it will make a big difference.

Kevin Perieria: Yeah. Again, like the apple fanboy in me, which is, I mean, it's like whatever, dying

Gavin Purcell: slowly.

Kevin Perieria: Uh, I mean, yeah, arguably it's the, it's the, the meme where that's, you're poking apple with the stick saying do something. Yeah, exactly.

Yeah. Like, I mean, like, I have an, I still have an iPhone 14. Even with the, the, the Apple intelligence features, I don't feel like I'm missing out yet. So yeah, just make it work. Just make it not say, here's what I found on the web for you. Fantastic. And then don't go overboard with it because we've seen some other companies that jam AI into everything.

Yes. Where it's unnecessary and it gets unwieldy and it kind of falls apart.

Gavin Purcell: Well, that's an interesting transition into a Gemini is now in Google Maps. Now. It could be really useful here. There's a lot of really interesting things that Google is trying to put AI [00:29:00] into, and we know that AI search has taken a lot of Google search traffic, but this is just a cool way of seeing like.

You'll be able to a ask questions directly while you're in Google Maps. And I, I have a car now that has CarPlay integrated directly and Google Maps is usually pretty amazing. I would love the ability to be able to have natural language conversations with it. 'cause right now, kind of like that, Alexa, all I can do is say like, you know, enter, uh, is only doing voice recognition.

So if in Google Maps I could ask it a question of like, oh, I gotta find a way to charge my car. Is there one somewhere between X and y? If it could figure out that answer, that is amazing. And again, that is a real world use case of ai. That would be very beneficial.

Kevin Perieria: Yeah. And to being able to have that conversation with the mapping app and it finds the right address, but then gives you directions like turn left at the gas station ahead.

Yeah. As opposed to like 200 feet, where I'm always like, oh, okay. I know if one foot, like just give me the natural language stuff like that's impactful. Or asking a follow up of like. Hey, I wanna go to that restaurant. It navs you and then you go, what's the parking situation [00:30:00] like? Yeah. Yeah. And it can go and intelligently surface that answer.

It's not like. It's not and magic yet. It's

Gavin Purcell: gonna cost you $40 to valet. Are you sure you want to do it? There's no other option. So you kind of have to, and you're like, right. Ugh.

Kevin Perieria: LA Google Maps is just a one big shrug emoji. It's like,

Gavin Purcell: it is crazy. By the way, valet parking in LA has gotten just ridiculous.

It's not with that's hit him with. It's my next big rant and my next big rant. First of all, doctor parking. No one should have to pay for parking at doctor offices no matter what. Yes, it should definitely not cost you $20. Second of all, valet parking is too expensive right now.

Kevin Perieria: Valet ai. Valet I, we gonna disrupt job.

I don't wanna, I'm not cutting more

Gavin Purcell: jobs. I'm not responsible for cutting anymore jobs. Kimmy two, Kimi K two is the open source model out of China. They have released their thinking model and Kev. They have gotten a better score on human's last exam than any of the other premier models. From OpenAI. Yeah.

Or Anthropic. This is a big deal. It's kind of a deep seek ish moment, but not, [00:31:00] it's not gonna be as big a deal, but like this just goes to show you that the Chinese open source, and again, going back to what Jensen said up above. Chinese open source models are doing very, very well right now. And Kim K two has always been one of those interesting ones where people actually think it's pretty good for creative writing as well.

So it's, that's an interesting thing to play around with, but what it just shows you is the open source movement is not slowing down. So when you talk about this idea that like things are progressing or not progressing as much. You can look at this story particularly and say like, it's great. You can go try it right now.

It's available. Like it, you can go pop into it. Um, that was gonna say chance to

Kevin Perieria: play with this yet? No, but I, I did look at the text specs and there's a few camps that said like, you know, there was like a couple different stories swirling about, one was that open source, eventually it's just gonna just go away here 'cause it's too expensive and it's too difficult to just be releasing this stuff for free.

There's no business model there. Well, that hasn't panned out yet. Yeah. Then they said, well, the gap between state of the art. Foundational something and open source is just gonna widen these companies. Spending billions of [00:32:00] that hasn't panned out yet. Right. If anything, that gap is narrowing and the timeline is decreasing.

And like you had actually replied to the Kimi K two thing saying you wanna see how it pairs up with, with writing, which I agree like a lot of the tests are, are more math or coding based. Um. Sometimes logic base, but uh, the fact that this thing can do two to 300 sequential tool calls at once, like where these models get really powerful is not necessarily in its base knowledge that it was trained on.

Yeah. Right. Yeah. It's in its ability to go out and use tools like web search, basic web search, you know, connect to MCP servers, discover things. Do them at the same time to get back a result for an A an a GI future, an agentic future where something can go out and be as capable as a human. We can use all these different sites, all these different tools.

That is a massive difference maker. And this thing is essentially going to be free and it can, that's one of its stats. It's two to 300 sequential [00:33:00] tool calls.

Gavin Purcell: Yeah. You know, the one thing a part of that's not gonna be free, I mean it's gonna be free for now, is the reasoning side of it, right? And the kind of inference side, which means like, they actually having to do compute while it's thinking.

Because I think that will be the interesting thing. And that may end up being the place where the, the, the larger companies separate from the smaller companies. And it might be why they're spending so much money on, uh, data centers right now because maybe the vast majority of the actual AI work is gonna be doing, being done on that side, but still.

Very exciting. I think people should go play with this. Try it. Just know it is a Chinese model. So there probably are some few things that you should just be aware. Don't, don't share anything super personal necessarily. Probably don't share any financial information, but don't do that anyway.

Kevin Perieria: Uh, well, okay.

Well, yeah, until you can host it yourself, which yes, I wanna talk about llama CPP. I don't wanna get too in the weeds about it, but. This is something we talked about eons ago. It, it's a, it's a free tool. It's released under the MIT license. Basically use it however you'd like, I think is essentially the, the feeling there.

But it is a, a, a front end. [00:34:00] To interface with large language models. Um, and they make it dead simple to run. And once it's running Gavin, they now have all of the same features that you would from like an Anthropic, Claude a, a Gemini Pro, or even a chat GPT. You can drag in images, PDFs, you can have forked conversations with it.

So as you're having a discussion, if you wanna branch it off and do something else. You just click a button. And the reason this is so powerful is that it runs entirely locally on your machine. So if you have a, there's, and there's models now which run on old MacBook errors, like there's, yeah, they're not as powerful, but a model like this is going to be coming to Lama CPP very soon.

So if you have privacy concerns, if you wanna run your medical data through it, or you're just a frequent flyer who is on airplane mode all the time. Run the new llama CPP, grab the models. They make it really easy to do. In fact, I, there's a guide. We'll link it in our show notes. Just click the guide. Um, and you can be running the latest AI models all safe and secure.

Gavin Purcell: Yeah. And we're gonna talk a little [00:35:00] bit later about the famous YouTuber PewDiePie, who actually created his own kind of like, uh, local server and, and a whole, a whole video on it. But like, it might be a thing that you see a lot more people doing in the future, right? Like, 'cause if you can start creating these.

These, these servers on your own and make powerful ais that you don't have to worry about anybody looking at that you can control. That's a pretty big deal. It does make me think of some sort of weird, like. Almost like fallout sort of world where like everybody, when you go into their homestead, they've got like a local, uh, AI that's run on like a big generator.

Here's our country

Kevin Perieria: cluster. It's uh, got a diesel generator backup just in case it's over there. Oh, we can get a trillion tokens per second, but we can get those numbers up. Yeah,

that's exactly what's covering. I feel that there are companies which sell thumb drives that are basically prebuilt. With models on them and front ends, like CPP, you could make your own for the cost of the thumb drive.

So just in case you can have intelligence on demand, I think that's pretty important. I put it right next to my gruel buckets. My dehydrated, yeah. [00:36:00] Shelf stable

Gavin Purcell: gruel buckets. That's right. That's the use case for gruel. Uh, Kev, one other really cool thing that happened out of the AI world this week, and this is one of those human stories, there's a story about a, a, a family who, unfortunately, uh, somebody passed away, but they got a bill.

For $190,000 for a hospital. And that was just like, you know, the American medical system is definitely in a funky place where this sort of stuff kind of falls out and people are like, how are we gonna do this? It's gonna bankrupt the family. Well, they use Claude from Anthropic and they got the bill down to $33,000 from $190,000, mostly because of duplicate, uh, charges and all sorts of other stuff.

What this is pointing towards, we hope is a much more clear listing of what these costs are, but also it's just a good example of using AI for your benefit and also showing how these systems have gotten so, um, uh, insanely, uh, shadowy and sort of cloudy that you need something to cut through some of this crap.

So this was just a cool story that I thought that is worth highlighting. A positive use case of [00:37:00] ai and anybody, again, can go do this on their own. I

Kevin Perieria: had a friend that recently reached out to me and was like going through escrow and was like, Hey, just. Run the d run it, run the docs through an ai, ask all the questions you want to ask.

He saved so much money by like removing the fax fees. He's like, we're not gonna use a fax. No one uses that anymore. Get that out of there. You know, like the, uh, the, the, the notary fees were like four x what they should have been, and he just had the AI prep, a concise counter to the escrow company and sent it off.

Saved a bunch of money. So, yeah. Yes, you can win with ai right Now

Gavin Purcell: I have one downside of this, which is interesting that. In that Sam Alman interview, the Tyler Cohen one, they talk a lot about the idea that. Margin businesses might come down a lot because of ai, and this is a good example, like right, the notary guy, whether he's or or woman, they may be a shadowy figure who's trying to like kind of upsell them and know that their thing is worth like whatever, $10,000 when actually like.

There's a, there's enough profit at $2,500 to make it work, [00:38:00] right? You can see a world where like that is going to shave off those sorts of things. And again, that goes back to like if your business can be done by an ai, be aware of that. And, but that's a big deal. So anyway, the on the side of the consumer and on the side of the person who's struggling, there's real value in using these tools to make sure you look at them.

Kevin Perieria: And what about the side of the, uh, multinational sugar water sales company or the, the commercial creative agency? Is there margin to be shaved there, Gavin?

Gavin Purcell: There is a margin to be shaved there. This is a, a story that we actually talked about last year around this time. If you remember last year, um, secret level, Jason Zada, who actually know his company, created the first kind of Coke commercial that was all AI generated.

There's a new one this year and Koch has again come out and said, AI generated holiday commercial, and swapped out humans for animals this time. But if you watch this commercial, and we'll show it on the video here, we'll also link in our show notes. It looks pretty good, right? It is another kind of like [00:39:00] stylized holiday commercial, but this was made by an AI agency, right?

So secret level is an agency that is AI first, and I can tell you Kevin. People were not less big mad this year than they were last year. In fact, in some ways I think they were more big mad. But I do think we said this on this show many months ago, maybe even in the first three months of doing this show, that advertisements would probably be the first area to start getting hammered.

Because you and I both know we've, we've worked at least adjacently to the ad business. There's a lot of upcharges in the ad business. Right. And a lot of it is like you've got a couple people that are like, we're creative geniuses. Pay us X number of dollars and then they get a group of people to make that thing.

And there's a big profit margin if you do it right. This is a good example of like the ad business is changing faster than almost anything else. Now the other side of this, Kevin, is that Koch has said very clearly that this is ai. They came out, they didn't try to fool anybody. They said This is an AI generated thing and there's a big, big blow back, at least [00:40:00] online.

Um, and what was interesting about what they said, Koch said, is that they, they did testing on this ad and when they tested it, just like last year, if you remember this story. Nobody seemed to care that it was AI when they talked to normal people. But my mom's gonna share it because the animals are wearing funny hats, Gavin?

Yes. Yes. And bless her for it. And that's where it comes back to this conversation around there's going to be a lot of anger around ai, but it is also about who watches it, who and takes it. What do they care if it is AI or not? And it seems like right now, and I'm sorry if this makes you at home mad or listening to this, probably not our audience, but it will make other people mad.

That mostly people will care if they're told, but even those, if you're told don't care that much, which is an interesting thing. The content is the content. It still took a,

Kevin Perieria: a pretty sizable team of creatives to make this commercial. Yeah, but they admit it took less than last year and it still took a while to make this commercial, but they say it took way less time.

Then last year, and that's a one year [00:41:00] difference for a full CG AI powered something. This time next year, I'm willing to bet the story's gonna be the same. And then you and I will on episode 4,000, be covering the single person, you know, that made the commercial for their sugar water company and it's it, it's outperforming.

And then there will be. The movement and the commercial that is completely human and maybe shot on film and, and cut on a Steinbeck, Steinbeck hand cut film, whatever. There will be that production that totally pierces through because people will be tired of seeing the AI visuals.

Gavin Purcell: Well, I was just trying to think of what episode 4,000 like where would, what year we'd be in at that point, if we're talking weekly, so we're at like 1 35 now, and that's a little bit like two and a half years, so.

That's a, I mean, that's a, we are living life. We're living life. I'm gonna be a glowing

Kevin Perieria: orb in a jar that you have cloned. 'cause you can't actually stand me. But it will still be, it will still have a bad mustache.

Gavin Purcell: No, we'll, both. Here's what it will be. There'll be two separate feeds. One will be you having your conversation with ai, me and me having your and we'll.

And we'll be [00:42:00] inter spacing each other and fighting to see who gets more views and listens. That happens. That's cute that

Kevin Perieria: we think you and I are gonna be around slash relevant. In any period of time,

Gavin Purcell: I think we're all gonna be around. This is my big, weird, crazy theory is that in 20 years, well I think you and I will make it.

Yeah, thank you. I think we're all gonna be around and I think that the, what I mean by make it is, I mean, we're gonna hit that level of where we figure out how aging, either, either aging gets solved, period, or we all digitize in some weird way and then, then it will be a choice one way or the other. But.

That I think, I think we're all gonna be around in some way and, and so prep yourself now for episode 4,000. Everybody.

Kevin Perieria: Here's the thing. If you're bent outta shape about the Koch commercial, remember that Apple is so bad at AI that they had to shoot their new Apple TV promo entirely, practically, which by the way, it's really gorgeous.

I saw that video, the practical video. It's really gorgeous.

Gavin Purcell: What's interesting about that is when you speak about people's reactions, the reactions I saw were. Thank God they're not doing this with, with computers. [00:43:00] Thank God they're doing this practically, which by the way, I bet a hundred percent before they did do it with computers and they did do it that way.

You know what I mean? So like, I think that's a very smart slight pivot. But I was like, yeah, isn't it? Isn't it

Kevin Perieria: funny though? It's so funny. Like we were literally just talking about ai, ai, ai. Oh, they pierce through all the noise. How. 'cause they hired professionals and did it practically. Oh, that's a still a

Gavin Purcell: viable option.

It's a lot better than AI smashing all of your creative tools into one giant iPad member. That was a whole different thing. They finally got a better ad agency. Yeah. Okay. Ke we have one last story that I thought was pretty interesting. There's a new AI wearable that has been. It's not out, but it comes out.

You can pre-order them now for 2026. This is a unique one. It is called the Sandbar Ring. And Kev, what I thought was cool about this is we've seen things like Friend, we've seen a couple other products out there that are designed to always be kinda listening to your environment, right? The idea is that there's an AI constantly listening and being part of what you're doing.

Sandbar is a different take. It's kind of a chunky ring that it looks like you wear in your [00:44:00] index finger and you bring it up to your mouth and there's a button like you can click it and you click it to kind of turn the microphone on. And yeah, the visuals are really interesting when you watch the trailer video, but also the guy using it.

You can actually whisper to it. And I think this is the first AI wearable that I might buy, because mostly it seems like it's a note taking device right now. But you can see how, oh, by the way, and you hear an AI assistant in your EarPods, so the idea is it's supposed to be paired with a pair of EarPods, but if you were to talk to it.

You can imagine, but going back and forth a little bit and having it go there. Also, by the way, a very good piece of technology for something like Ambe, which our startup is, because you could have these kind of conversations without having to look at your phone and just kind of talk to this thing back and forth, so

Right.

A cool, new way of thinking about a i ux. This is a pretty small company. I think it's a startup. Lot of people tweeted about it yesterday. Uh, they've got decent funding, but I, I think this is a cool idea overall.

Kevin Perieria: Yeah. And I, I, you know, I'm sure by version three they will bolt in a tiny camera so that you can preserve [00:45:00] visuals of your memories.

But for now, I love that there's not a shutter on this thing. Like, I, I love, I, I want less wearables that are turning us all into this weird, weird nanny state, if you will, of just like, everything's gonna be recorded. All of the time and probably backed up to a cloud where someone else is gonna take that data and crunch it in ways that the people that are being recorded don't have any say over.

Gavin Purcell: How do you, how long do you think it'll be before like there's a little tiny dot on these. I'm just pointing to my ear pods. Every bear parts, everyone. It feels like that's the next step for Apple, right? Like that they would have. A tiny camera on either one or both of your AirPods that would actually go out and you could show somebody else.

It's like their version of the meta glasses, but just in audio like that feels like it's gonna

Kevin Perieria: come soon. Could fully see that. I mean, just look at the iPhone air and the way they're trying to shrink. Yes. And, and put everything into that tiny bar. Like if one, one side of your bud is the compute and the electronics and the other side is the battery.

And I don't know if there's like a little digital wire or something, like there's gonna [00:46:00] be something that's going to allow them to achieve that vision. But again, like I. I, I think I'm gonna, like, I haven't had too many parties in my life where I'd be like, all right, throw your phones and your car keys in the fishbowl where like, this is off the record.

I haven't really had to do that. I might have to do, I might have to do that with the glasses.

Gavin Purcell: Yeah, I think so. I mean, I'll, I, I, at some point, like, unless. I, yeah, it's gonna kind be interesting. Glasses are gonna be a, a change for people and I think again, generationally people might just get used to them, but there's gonna be a transition for sure.

Um, alright everybody, it's time to take a look at what you did with AI this week. It's ai. See what you did there. Sometimes

Game: ya rolling. We found a care. Then suddenly you stop and shout.

Gavin Purcell: She wants you there. Alright, Kev, we have some fun stuff this week. First of all, PewDiePie, the famous YouTuber who was a, for a long time, the number one YouTuber until Mr. Beast overtook him. He made a [00:47:00] very cool, very long video about ai. There's a couple things I do wanna say about this. First and foremost, you know, he has a great thumbnail.

Pedi Pie is great. It says. What does it say? It says like, stop. Uh, use ai, ai Stop using

Kevin Perieria: a, it says stop using AI right now.

Gavin Purcell: Yeah. And then, but it actually is stop exclamation point using AI right now. 'cause what he's doing is he is setting up a full local AI model and a cluster on a bunch of graphics cards He got.

And what's just cool watching this video, I, I suggest everybody goes to see this. He just is excited about what AI is can do. Right? And, and I will say very clearly, he goes in, in, in this video to talk about how much he hates generative AI video and audio, which was interesting. He's not a big fan of, of video and pictures and, and the slop, ai slop, he talks about that, but.

What was cool is seeing him set up a full local, um, llama based AI model for himself, and he did the UX and the UI of it all, and it's just a cool thing. It's a cool project to watch somebody get excited about, especially at the size of his audience. [00:48:00]

Kevin Perieria: Yeah. I mean, people were very surprised to see this come out of him.

I, I saw it and clicked immediately and kind of watched it down. I was like, oh, yeah, that's cool. All right. I like that he's trying to like, extract every ounce of his $20,000 rig and his GP lose him. Yeah. I saw some salty people in the comments basically saying like, oh, of course you can do this. You've got the money to spend on this rig, but I think that's missing the forest for the trees.

Yet again, you can build a version of what he built. For a couple hundred bucks. Yeah. Or, or even, even less than a hundred bucks. If you wanna use like a raspberry pie and you're gonna get. Much slower performance with much less capable models, but you can still go through the steps of building your own.

And we just talked about the Doomsday Bunker AI approach. Well, this is, this is like a really high end premium version of that, but the, the being against the generative AI portion of that, like again, totally understand it, but I also like. Is that, is it, uh, how, how do I phrase it? Like I, if you're, if you're against it for [00:49:00] the generative AI stuff and you're a content creator, but then you're celebrating it for the code that it can write or the, you know, the things that could do that you don't traditionally do, like how is that any more noble than someone who's really into code being like, I hate that this is trying to write slop AI code, but I can finally make pseudo songs.

I, I,

Gavin Purcell: I think this is where you struggle and I think I understand where he probably struggles for the people and the sort of things that he talks about and wants to do. But yeah, I think in general, this is the thing. You kind of either have to be on board or not on board and, and I think the biggest thing you can still argue if you are against like AI audio and video or, and photos is that.

That there's all these artists that were part of the process of training, but at the same point, code was that way too. And I just think code is seen differently in some form because in code, everybody kind of, at least from my mind, I'm not a coder. People like are more used to sharing code or forking code and using it and remixing it.

Whereas in in art and in music and all that other stuff, people are much more used [00:50:00] to like saying, this is me, I did this thing. So it might be part of that. But in general, just a very cool thing to see a mainstream YouTube. What about Joseph

Gordon Levitt?

What about Joseph Gordon?

What about Joseph Gordon Levitt?

Mr. HitRecord, he wanted everybody to share. Give me your face and gimme your best visual reactions of things. Oh, I didn't think about this. Oh, he was crowd stuff. I never thought about that. You're right. Joseph Gordon, an

interesting thing, but he didn't, I, his goal with HitRecord was never to like. You can remix something, but was never not a Sam Alman in disguise.

Kevin Perieria: He wanted every, he was just getting training data for free. He was building his own B-roll library. I don't know. I do not agree.

Gavin Purcell: Just so you know, Joseph Gordon Levitt is very anti AI in the world. He's I know he is. He's come out hardcore. Yeah. That's very funny. I've never thought about that one. Joseph Gordon,

Kevin Perieria: I'm gonna, you know what, I'm sorry I left you on red.

I will get back to you. Let's bring him on the show, Gavin. Yeah, should him on the show would be actually, he reached out to me ages ago on the show. He would be a great best on the show.

Gavin Purcell: He would absolutely hate some of our takes. Yes, you definitely would. So, hey, this is a, a really fun tweet from Kylin O'Connor who built a Python script using one of [00:51:00] FOFR of, of our favorite X users models on replicate to map out a grid of images of his face looking in every direction.

So he had a picture of himself and then he got a picture of himself looking up, looking up, looking up, looking up, and he built a website. That you can use your mouse and make him look at the mouse. And this is just a fun small thing. This video, uh, this tweet went pretty viral and now he's going to drop the code on GitHub so people can do it themselves.

Amazing.

Kevin Perieria: Yeah. 'cause coders view art differently. Gavin. All right. They share. They definitely do source, definitely do. I've seen this replicated by a lot of people. But super

Gavin Purcell: fun. Finally a very fun video that I saw on Reddit. Uh, this is a new version of the Marvel Kombat finishing Movements. This is a guy named whose name Isim Maximus, and he created a series of Christmas fatality finishers, which are very good, but even more importantly, Kevin.

He, unlike some people out there, shared his full prompt. So if you go to this Reddit post, which is you can see it and you go down [00:52:00] below, I took his prompt and I used it in SOA and I did a Thanksgiving one. So maybe you can play that for everybody right now and we can just kind of listen to it and watch it finish him.

Game: Looks like someone got stuck. Drumstick.

Gavin Purcell: So just describe, describe to people what you see there. There's a, there's little

Kevin Perieria: Ramadan of mashed potatoes looking very, uh, pizza, the huish, like a sentient pile of mashed potatoes with a weird gravy drizzle coming down. Uh, and then there is a Turkey leg. Um, with some gloves on and some boots.

They're fighting on a Thanksgiving table with some delicious looking pie in the background. And the Turkey leg kind of jabs the mashed potatoes with the bone that's still coming out of the leg and the mashed potatoes and kind of. Sloppily disintegrate. Yeah. I mean, it's,

Gavin Purcell: it's, it's a, it was a, again, is a one-off, uh, output.

It's great. But what was, what is cool about [00:53:00] people who share prompts on soa? So we've talked about this before with SOA two. Some of the best results come out from just saying like, make this funny. But some of the even better results come out for people that have really started to study the prompting.

There's a lot of really interesting stuff. And this guy's prompt was very specific. If you look at, uh, I'll, I'll drop it in. Uh, our show notes somewhere, it's pretty long, but there's like an announcer prompt, an arena prompt, it has the winner and it has the loser. And then it describes like it's a good like probably 50 lines of text, which Sora can take.

And it did a very good job of sticking with the prompt. And again, this is how you. Start to learn to get better at AI video is you look at what people have made that you think is fun and interesting, and then you try to figure out what they did. And I just, I, I hope that lots more people share their Sora prompts because it does make it easier for you to understand, oh, that's what you have to do to get the better results.

Yeah.

Kevin Perieria: Yeah. I, you know, was having a conversation with a, with a pal about. Specifically prompting SOA and was like, you can get super [00:54:00] granular with it and give it frame by frame or second by second in brackets. Call out the cameras, it's at the other. Um, and he was like, no, I'm just gonna say like, make a funny commercial.

I was like, okay. And that works too. That works too. But if you want that full creative control, you gotta go for it. And that prompt is really fun.

Gavin Purcell: Yeah, I've been having a ton of fun, uh, with soa. I made a new character called Old Man Abe. So if you're out there, go uh, at Old Man. Abe is just some really Ra by the way, he's a random old man that showed up in one of my other sort of videos and I made a character out of that and it's working incredible.

Which is very cool. So like, if you don't know that, that was one of the big things they introduced with character cameos is this idea that. You can take another SOAR video and grab a face and make that your character. Oh, that's fun. Um, yeah, super fun. And then, you know, the, the random or, um, very funny, um, remix is always good.

One of the things I did that's actually getting a bunch of likes right now is there's a great remix of a woolly mammoth and a block of ice that's rolling out, uh, on a big semi-truck. I put Queen Elizabeth in that block of ice, which is just a fun way to remix. So anyway, [00:55:00] so two continues to be a fun platform to play around with, um, and we're seeing a lot more people do creative stuff with it all the time.

Um, alright, Kev, that's it, that's it for our week. Uh, hopefully we started off low but brought people back. And, uh, go, uh, give us a, give us a heads up on uh, uh, anything new that you see online and we'll talk to y'all next week.

Kevin Perieria: See you in the comments, friends.