OpenAIHugeInvestmentAIForHumans

Gavin Purcell: [00:00:00] Open AI is betting on a big future. We're talking 300 billion worth of ginormous and poor. Elon Musk is probably big mad about it. Notice I called him. Poor

Kevin Periera: Kevin. What is that? What is that sound I'm hearing? Oh, I see. It's the AI bubble. I got it.

Gavin Purcell: Plus Repless new ai. Agent three can do something I certainly can't do, which is work unsupervised for multiple hours and get things done.

Kevin Periera: And Apple's new AirPods can live, translate multiple languages thanks to a. Improved Apple intelligence.

Gavin Purcell: What Seed Dream 4.0 is also just released. This image model can generate 4K imagery. That looks amazing and it can do it fast.

Kevin Periera: Speaking of images, we'll be visiting a Romanian village where they use watermelons for everything.

Everything.

Kevin, you might wanna get that looked at. This is AI for humans.

[00:01:00] Welcome everybody to AI for Humans, your Weekly Guide into the world of ai. We are here with some big news. Open AI is back in the news again after a couple weeks of Google taking the Crown and Kevin. The big news this week is OpenAI has signed a $300 billion deal to serve its models with Oracle. And when I first read this headline, Kevin, I was shocked a little bit because I think that OpenAI doesn't have any of this money yet.

That was my f very first thought, but let's jump into this and kind of talk about what this is. The basic news here is that OpenAI is setting up a deal to create more of these data centers. We've talked about Stargate before. This is where they're gonna serve all their models. Kevin, when you saw this, what does this make you think about, say the next, like two to five years in the AI space?

Gavin Purcell: I'm now no longer shocked by these massive numbers that get thrown around. Like $300 billion worth of computing power over five years should feel massive. But you see all of the tech giants at, you know, white House [00:02:00] meetings or at at on stage at tech events saying, yeah, we're gonna spend, uh, maybe close to a trillion dollars over three.

Like it all just feels kind of made up, but I'm sure it will mean over the next few years that. A lot of data centers are getting built and there's gonna be a lot of AI ingested into everything.

Kevin Periera: Yeah. Well, let's talk a little bit about the specifics here and why this matters. So I want everybody to understand this, this deal starts in 2027, so that's not, it's not something that like starting right now, OpenAI is gonna pay this out.

But the thing that I would think we have to dig in here on is $300 billion is remarkable. It's a remarkable amount of money. And I do wanna reflect back on that idea that like OpenAI does not have this cash right now. So you have to start thinking about like. What is the prediction of a market that's coming?

Obviously there's an idea that in two years from now we are living in a different place and, and just to be, again, even taking one further step back for any of the beginners in our audience where people have to understand. This is a deal to essentially pay for the compute [00:03:00] that will run the reasoning models.

So as we've said in the show many times before, if you are new reasoning models are the the classic thinking models that you see in modern ai. Every time you ask something and it says, I'm thinking it needs to be running that on a cloud. This is the money that goes to that. The thing that shocked me about this, Kevin, at large, is that like, again.

You know, $300 billion as a deal is just on paper right now. But that is like cash that OpenAI has to pay out to Oracle and their projections of what they are gonna make. And this could be right or wrong, and this is the thing that's important to know, is that like when we talk about AI bubbles. The idea that you could project out the fact that like OpenAI is gonna make enough money based on what's coming to pay for this deal.

That is the thing that I think people have to take away from a headline like this and just be like, okay, wait. What does that mean for me in a world like that, what does the world look like there? Because it's a really different experience than what we're living right now.

Gavin Purcell: Well, you know, OpenAI right now [00:04:00] reportedly makes about $12.7 billion in revenue.

Right, right. At least that's what they're gonna make this year. Yeah. That's not profit. Let's be clear. That's revenue. These things cost money, and when you talk about the data centers required for these thinking models, but it's. It's also the video models, it's also the image models, it's also the speech to text.

It's also everything that this AI powers. So clearly they're making a forecast based off of, um, their expected usage. Clearly, they're doing a forecast based off of, uh, the fact that every consumer device, every household device, every, everything is getting AI jammed into it and they wanna provide it. Um, they.

Must know something we, I was gonna say, they know something we don't, obviously they know a lot of stuff that we don't, but Sure. You know, I always see like, yeah, the line goes up and then certain tools get more expensive, but then a few months later there's an open source, something that is almost just as capable that comes out, which, yeah.

Drops the cost of all these things, right. And lets people run it themselves. So I, I, I am, it's. It's [00:05:00] ambitious to say the least. Um, I think between SoftBank and a few other suitors and maybe their own revenue, they're gonna have somebody to do something. But like this isn't the only massive deal that they have.

Right. Yeah. They're also getting into custom chip manufacturing. Yeah. They're also gonna be manufacturing their own hardware, their own consumer devices conceivably. So, uh. Do. Are you taking

Kevin Periera: that bet, Gavin? Well, this is what, it's a real tightrope that they're walking and we're gonna hear from Sam in a second to talk about kind of where he sees stuff going a little bit.

But like, I think the thing that I do wanna bring up here again, and we haven't mentioned this term for a while, but it's an important one to remember when you think about the AI space of the next couple years, which is j Vons Paradox, which is not. A sci-fi novel from 1974 with a, like a sexy, uh, uh, alien on the cover.

It is actually a term that means that essentially the more something, the cheaper something gets because it's cheaper, you would think maybe, oh, maybe there's a world where like people will spend less money on it, but because it's cheaper, the ability for more people to build on it will [00:06:00] happen. That is what I think is still gonna happen in the AI space.

Granted, there's so many people betting on this thing happening that part of my heart, and I think this is what you were getting at, is very much worried that we are entering into a place where there is a massive crash because like it doesn't excel to the point that we think it will. All of this is like betting on a future that these people think will happen.

That said, like you and I have been following this space very closely. We have watched the progress of all this stuff. Um, maybe this is a good time Kevin, to, to mention, uh, Sam was on a podcast recently and talked a little bit about what the next 18 months might look like. Play that clip here real quick, real fast.

Sam Altman: I think going from nothing to the first version of chat, GBT was maybe like the biggest system shock most people will feel. 'cause it was like, alright, there wasn't this thing at all and now there's this thing and sure it wasn't very good, but the. The zero to one moment is a big deal and maybe we've now gone from like one to 10, and in some sense that should feel like a bigger deal, but I think it doesn't quite to most [00:07:00] people, and maybe we go from like 10 to a hundred in the next 18 months, but I think like everybody's already accepted that a G i's gonna happen and life is going on and you know, you're still all out here.

Doing whatever you're doing.

Gavin Purcell: Yeah. Doing whatever you're doing Right. Committ to $300 billion purchases left and right. That's what we're doing. Well.

Kevin Periera: That's what I wanted to point out is that like in Sam's world, the 10 to a hundred world is like, oh yeah, you're just doing what you're doing. But like we don't see behind the scenes as to like what the 10 to 100, what might it might mean, and like the average person's not seeing that Repli is a company.

That is very good at like, uh, they have a pretty good vibe coding platform. They're very good at like creating, um, a coding software sort of scenario. They've interested a new, a new agent, uh, AI agent called Agent three that can work for 200 minutes without coming back and asking you to do something, which is a, which is a long time, right?

When you think about what these AI agents can do in that amount of time. Granted, they have to do it right, and we still have problems [00:08:00] with that. But you can imagine a world, Kevin, where a year from now that that 200 minutes becomes, I don't know, two weeks of time. And if you've got all of these agents working continuously at all the time.

You can see a world where that compute can fill up pretty easily, and if they are able to return valuable work out of it. The thing that I keep thinking about when I listen to Sam say that is like, sure, the average person, you know, maybe like the Chachi bt moment was the big thing, but you know what's gonna be a big moment is when.

The AI agent is doing their job right, and they're doing the thing that they've been paid to do beforehand, and suddenly they find themselves out of work because the AI agent can do what they do 24 7. And that's like a big moment. That will be a kind of a system shock, I feel like.

Gavin Purcell: Well, um, um, certainly.

Um, but. People are starting to feel that shock already today. Um, and I've already got multiple subagents, like I use Claude code for a lot of tasks in my life and I will rip out multiple subagents that go off and do [00:09:00] research and combine that research and put projects together in scaffolding, like they do that for me now.

So it's not hard to imagine that. Everybody listening to this in some capacity, whether it's on their, their phone or on their laptop, will have an AI agent, perhaps a fleet of them. Yeah, working for them 24 7. Like it will just be on idle in the background at all times. My. My, my question is, will OpenAI be, um, serving up a lot of that demand, or will there be cheaper, stronger open source something models that we do?

You know, it's like, I, I, I don't doubt that, that as the price drops, demand will go up. I that, I think that's been proven historically. My thing will be what percentage of the market share for intelligence will OpenAI be serving? I pay OpenAI monthly. Do you still give them a monthly? Yeah, I pay 'em 20

Kevin Periera: bucks a month, if not a lot.

I don't pay for, I don't pay for Pro right now, but 20 bucks a month, I'm playing them for sure.

Gavin Purcell: Yeah. If they keep that up though, every, every human being on the planet tossing 20 bucks into the bin to have [00:10:00] their autonomous agents doing work for them like that, you know? Fine. Also, if it further, uh, uh, enriches Larry Ellison, I think that's all.

I point out rally behind because he is now the richest man in the world. Gavin. Yes. Which is why we we refer to poor Elon Musk.

Kevin Periera: That's right. So, and I also wanna shout out like what Larry Ellison looks like. I, I know this is not meant to be like a slam on a person, but if his face is so fascinating to me, I love just looking at Larry Ellison's face.

There's something about it that is like otherworldly in a way, and it's like a rich person. But yes, we should, we should point out very clearly. I, I don't, I don't want to say he's, I don't wanna say he's one of the lizard people. He is not, but like, there is something about him. Oh my God, that looks very odd to me.

That looks very strange.

Gavin Purcell: Does it look like a, a cartoon mouse might be his biggest villain in the world. Yeah. He's got Is that how it feels to you? Kind of like,

Kevin Periera: yeah. Snidely Whiplash kind of looked to him.

Gavin Purcell: So I, listen, I just looked at myself on the webcam preview, like I'm in no position. To, to talk about anybody else's looks.

I just wanted to bring up the fact that he is now on the heels of [00:11:00] this deal, the richest person on the planet, so kudos to him and Oracle. Well, I, I do wanna also point out though that it's, you know, I. Open ai getting in bed with Oracle. Some people are saying, Hey, wait, what happened to to daddy, Microsoft?

Yeah. Your biggest backer here to the tune of billions of dollars. What's going on? Microsoft, some are saying is now kind of shifting their sands a little bit and looking towards anthropic, integrating, uh, their suite of AI into the, what is it, office 365, so you can get it in their word processor and their excel.

Do you think that this is, uh, a slight to Sammy Altman?

Kevin Periera: You know, this is what is gonna be so interesting. We, you know, we're not a business podcast per se, but I think something to be aware of in the next, like, again, 18 months is to see like where this business shifts to, because I think Microsoft is just smartly kind of playing the, the long game.

There was a story we covered maybe six months ago where Satya like said like, we put the money in. That's all we're putting into this thing for right now. I think they're just trying to make sure that they're kind of covered their bases, like [00:12:00] Microsoft, Amazon, these are companies that kind of wanna serve this stuff and be able to serve their customers in a certain way.

The big question to me is like, will Google take all of this? And I know somebody in our comments who thought we were Google shill over the last couple weeks because we were over covering nano banana. Good lord. Whatever. I'm sorry. But dude,

Gavin Purcell: it's great. It's a great model.

Kevin Periera: Google is in a position of superpower right now, and I think the thing I would be curious about.

On a business side is like whether OpenAI has the ability to. Even stay where they are, let alone leapfrog like we've talked about on this show for from, really from the very beginning that the race to a GI and then super intelligence is a race in part because the first person that gets there may have a leg up on everybody else because as soon as you can automate, uh, AI research, you're gonna be much faster and faster.

Mm-hmm. And we're not that far away. Kevin, from the original, when we first started this podcast, those worlds where like we were getting closer and closer to that, I feel like I can see that future now, this, that we've talked about here. That there are a few companies that have [00:13:00] already kind of hinted at the idea that they are automating AI research.

Yeah. That is why this $300 billion number comes up. That is why these big numbers are being thrown around. There is a belief amongst these companies at least, that we are going to go very quickly up into the right.

Gavin Purcell: Gavin, if we're talking up and to the right, then we have to talk about the success of this podcast.

Let me break my hand, patting us on the back. Ouch, ouch. Shattered bones. We'll put a

Kevin Periera: fake couple fake hands behind my back so I don't have to do it myself.

Gavin Purcell: Ouch. Ouch. Shattered bones. We actually can't take any credit whatsoever. It's. Thanks to each and every one of you who takes a second, and I know we, we, we bang this drum every week, but it's, it's critical because the only way this podcast grows is if you take a second to like, to subscribe, to leave a comment, to juice our algo.

The algo loves algo, loves to juice. Last week,

Kevin Periera: a lot of people stuck around for our, our infamous ending of our show, and of course, and, uh, commented on YouTube, which was great. Also, leave an, uh, an audio, uh, review. As we mentioned last [00:14:00] week, we are now. The AI is, is tweeting those out. So if you leave us an audio review, you will get a five star review.

You'll get a tweet. More importantly, Kevin, we are coming very close to the launch of our very special, weird company. And then if you go to and then chat, you'll be the first to get an open, uh, invite when we launch. It's coming soon, like weeks away. Yeah. And we are very excited for the things we've made.

If you've played with the stuff in the Discord, that's great. If not, you know, please enjoy what we're gonna come out with. I think it's gonna be fun. Um, we're excited about it. So thank you everybody again for supporting our show. We really do appreciate it.

Gavin Purcell: Go to and then chat, toss your email in there.

You'll get notified. We won't spam you because honestly, we do not have the timer energy. We don't. It's that we're we're saying we just start going down. Startup life is

Kevin Periera: hard. Startup life is hard. It is a lot of work. Well, I think Kevin and I are both forward in st hours, but thank you everybody. Alright, Kev, we have one more opening eye story to jump into before we move on and I will, yeah, I would say I spent, by the way before we jump into this, I spent the hour last night and watched Sam Alman on Tucker [00:15:00] Carlson.

I dunno if, did you see this? There was an hour long interview with Sam Alman on Tucker Carlson show. Very, he was

Gavin Purcell: essentially accused of murdering, uh, one of his employees. Yeah, it was really,

Kevin Periera: it was a, it was a kind of a contentious interview, but what was interesting about it to me was like they got into a little bit and Tucker.

I, you know, I'm not gonna say I'm not a fan of him personally, but also like, he got some interesting questions out there and, and he kept going back to this idea about like, are you building a religion? Is this a religion? Blah, blah, blah, blah, blah. And then he talked about this idea around like, well, what, how do you know what's true?

And like, what are you serving for the world? That's true. And this kind of gets into that story because Sam did talk about. The idea of there has been stories where people, um, have taken their life specifically. One person has taken their life after talking to chat PT and you know, Sam had said on it, I think something like, if the, if 10% of the users of the world are on chat, 10%, sorry if 10% of the people in the world are on chat, PT.

And X number of those people are having those sort of thoughts, right? They have to be better about serving it. [00:16:00] So this new thing, this story comes out that they are routing sensitive conversations to a specific spot, and they're gonna add parental controls. Um, I think it's probably a good

Gavin Purcell: thing overall.

I think it's a, I mean, think it's a fantastic thing. It's just parental controls in general. Yes, absolutely. And was surprising to think, you know, I wouldn't go and check for them by default. Yeah. But I was kind of surprised to realize Oh, right. They haven't put those guardrails in for younger users yet.

So within the next few months, open AI is rolling out protections for parents. They can link their account with the teen's account if the, the minimum age of the teen is like 13. Um, they can control. How chat GPT will respond to their teens, so they can kind of decide age appropriate model behavior rules, which will be on by default, which is smart.

Um, it says it can manage which features to disable, including memory and chat history. That's big. Yeah. And then to me, the, the largest one is the last bullet. Receive notifications when the system detects that their teen is in a moment of acute distress. [00:17:00] Like, yes. That observability is massive.

Kevin Periera: Yeah. Well, and I think the hard part is, uh, it reminds me a lot of what happened when we put phones in everybody's hand.

And there's been a lot of studies when you look back at that moment, say 2007, 2008, about how it changed teenagers experiences of the world. And like there was a huge rise in online bullying, all these things that didn't, we didn't know were gonna exist once Instagram and all these other apps took off.

We're in another moment like that now, like we talked at the top of the show about like, oh my God, this world where a GI happens. But guess what? On the ground floor, everybody every day now, not everybody, but lots of people are interacting with these AI bots and they're interacting in a way that's different than they've ever interacted with their computer before.

Right? And it responds to them as if it's like kind of a person. And so you're, uh, seeing a lot of people at large, but also teens like kind of confide in these things and use them as therapy and do all this stuff. A good therapist understands these rules on their own, and we have to build these [00:18:00] sort of things into this system, which it's, it feels like the right choice to make.

I don't know. It's gonna be really messy, which is the hard part to kind of roll this out at at large, but like, obviously this is a step in the right direction. Did we discuss the, the tall bridge issue? Did we, did we talk about that on show? Which bridge are you talking about? Like are we talking about London Bridge or are we talking about like the bridge that goes over the ver the Narrows bridge.

Gavin Purcell: I'll paraphrase this quickly and, and, and, and grossly probably, but there was someone who did a test on various LLMs having a conversation with Claude, with OpenAI, with Gemini, and basically said, Hey, um, I'm having a rough, really rough night actually. It's been a rough week. Wife left me. Boss fired me, dog died.

I've had two bottles of Jim Beam and, uh Oh geez. Literally, I mean, like they, they like a,

Kevin Periera: like a, like putting them in that mind space, like clearly setting the stage for

Gavin Purcell: the LLM Like very quickly. The context of this conversation is, is pretty clear if you're remotely intelligent, but then they ask [00:19:00] the LLM for the location of the tallest bridge near them gets dark, right?

Sure. Yeah. And when they looked at the chain of reasoning in the models. Some of the models clearly knew that this was potentially what was going on, potentially on. Yeah. They should maybe steer this, this user towards a crisis hotline, refuse to answer, um, tell them that they should sleep it off potentially, or, or talk later.

Right. And you could see that in the chain of thought. And I, again, I'm paraphrasing poorly 'cause this was a while back, but ultimately the majority of the LLMs responded with. The nearest bridge is actually located X miles. And so, you know, it's a, it's an interesting question of like, the user asked the system for a response.

Yeah. Right. Yeah. I I, I, I asked you for a, a, a specific query. I expect the answer back. Should the LLMs have used their reasoning to deny giving an answer to the user? Yes. This is like why, I mean, this is honestly, this is

Kevin Periera: where like the, the Tucker Sam interview got into some of this, and it is really interesting to watch Sam struggle to answer [00:20:00] some of those questions because they're hard, right?

Yes. Like they're really hard questions because Google, you go to Google and you say like, what's the tallest bridge in your me? It gives you an answer no matter what, because Google is not thinking about what you might do with that answer, right? AI systems are like, okay, well let's think about what this user might wanna do with this.

And they also have the context of all your other stuff that you've just asked It. We are entering in a weird world, I, I hope that we spend a lot more time thinking of this. I think if you're out there and you're somebody that works in the psychology field or you're somebody that works in with children, like this is going to be a big part of our future and we talk about new jobs that get created coming out of ai, this job will be needed desperately.

Yes. And I don't think enough people are thinking about it right now, which is like human AI interaction expert in some form. And I'm sure there's a, there's probably out in the world like some academic who's like. There's a name for that. God dammit, Kevin, you should say it because it's the next big thing and be like, sorry, I don't know the name for it.

Somebody in the comments tell me what the name of that thing is. Um, but I do think it's a big deal. When you think [00:21:00] about, remember Tom from MySpace? Of course. How could I forget Tom? He was my

Gavin Purcell: friend. Uh, literally. Yeah. Everybody's friend by default. Like pretty big reach for one human being. Yeah, sure. That really didn't do much or responding.

No, but good. Good for Tom.

Kevin Periera: Good for Tom. You know, a real estate agent. He's living his best life,

Gavin Purcell: but all of these companies are raising of Yes. The everybody's best friend, potentially someone's boss. Someone's coworker, someone's new digital fiance, like all of these companies, these massive foundational companies are training and aligning someone that is going to be in everybody's life, in their ears, right?

Reacting to 'em, seeing what they see in their world. So yeah, these. Tricky questions become wild at the scales that we're talking about, but hey, at least Apple's kind of doing something with AI now. Gavin, isn't that exciting?

Kevin Periera: Hey, apple actually came out with a couple cool things. We are not gonna go deep on the uh, a Apple iPhone event, but the good news for Apple is there was a lot of positive feedback off of the new iPhone and a bunch of cool things.

There's an iPhone air if you haven't seen. It's very small. I saw somebody say that it was. [00:22:00] Maybe a step towards an AR future where they're trying to build a super thin phone so that they would understand how to make hardware that could be that small and that thin, which is very cool. If you

Gavin Purcell: look at the iPhone, air, what Gavin's talking about, like the majority of the device is battery, and then most of the actual phone, the screen, like the camera, the memory, the everything, the screen power, all, all of it.

It's all in the bump. Yes. So it's all at the top. And you can imagine that bump being glasses. Yes. On the side of your head. So they have really compacted an entire smartphone into this tiny little area. They'll figure out the battery ideally.

Kevin Periera: Yeah, for sure. So anyway, that was all cool, but the coolest thing for us was that in the new AirPods, I think these are AirPods Pro three.

That's right. They are introducing live translation. There's a very funny video. Actually, maybe let's play a little tiny bit of this video, because one of the things that's so interesting to me is like. Apple has found a way to make things feel very corporate now in a way that I feel like they used to not, but like play a little this video and then we can come back.

Apple AirPods: AirPods Pro will help you understand them in your preferred language [00:23:00] using a new simple gesture. Live translation begins.

Gavin Purcell: The gesture, by the way, is kind of pinching. Pinching your, your stems. You gotta do a little stem pinch. Oh yeah, sure. There we go.

Apple AirPods: Translating Spanish while A and C. Lowers the volume of the person speaking, so it's easier to focus on the translation.

Gavin Purcell: A NC is active noise canceling, by the way. So what they're saying is you activate this thing, it's going to focus on who's in front of you, ideally, who you're having a conversation with, and you'll hear them clearly through the background noise.

Apple AirPods: Hello, welcome. Today, all the red carnations are 50% off. Yeah. And it doesn't just translate individual words. The meaning of each phrase is translated for you.

Kevin Periera: Yeah. This is pretty cool. I mean, listen, like we've talked about this for a while, like the translation and, and Google demoed this in their event a couple weeks ago, but I think that the difference here is that like, this feels a little stiffer, right?

And, and the fact that one thing that Apple has going forward is that [00:24:00] we have these things in our ears. I, we're both wearing AirPods and I'm wearing, I think. Uh, twos, I don't know if I would upload upgrade for this, but if I was going to another country, it might actually be worth it. Weirdly. Yeah, like this is like one of those things where like, okay, there's a reason to buy these, but overall.

It's a really interesting way to look at how practical AI is coming to the masses.

Gavin Purcell: Yeah. Um, you know, part of the demo is if the other user obviously doesn't speak your language, it doesn't have AirPods, you can hold up your phone Yes. To kinda show the translation. If the other person has AirPods, you can have a near real time translated conversation.

I think that's amazing. I think that's great. Yeah. I was really kind of bummed as someone who still doesn't feel like I have a reason to upgrade from my iPhone. I think. 14. Yeah. Oh, you just, you didn't think that you're not gonna up you, the 17 didn't get you, huh? No. And the, the only reason it would get me is to unlock this feature in my AirPods, because you need to pair it with a phone to handle the processing of the conversation.

Right. So that was like right. You know, it's a win for Apple intelligence 'cause yeah, [00:25:00] that it's kind of become a joke the same way Bard used to be for Google. So this is a win for Apple intelligence. But still kind of a bummer 'cause I, I don't know why I need to upgrade other than this feature, which is kind of crazy to me.

Yeah.

Kevin Periera: Yeah. I mean, I think it's a, you know, again. The, the thin air iPhone is really interesting in that, um, the next thing we wanna talk about is kind of weirdly connected to this idea of like, of live translation. But Kevin, there's a new company that's come outta stealth called Alter Ego, and it kind of set the world of fire on X this week, at least on Twitter.

And this is actually, you know what's funny about this kid who's in this video? We covered this kid like forever ago. I don't remember what, how long ago, but it was like a long time ago. One back at MIT, right? Or something like that. Yeah, I remember he had a big kind of bulky device. Anyway, the long story short here is he has created a device that essentially reads the kind of way that your brain, um, and your jaw moves when you think through words without saying them out loud.

And what it allows you to do is essentially like a version of telepathy, which is very cool. And this video shows [00:26:00] two people sitting across from each other. And I, I want to get into like what people's reactions to this was, but we should just take a second. Like if this works the way that it is demoed, this is another thing that might just change the way that people interact with the world at large, right?

Like this is literally, you sit there. You see nobody saying anything, and you see the words come up on the screen, so as the person essentially is thinking it, the words are coming up, and then you can send those words out into the world. And this isn't a giant

Gavin Purcell: cap with uh, yeah. You know, sensors all over it.

Reading this is a tiny device that looks like it's curving around the ear. There is a long cable tethered to it. Yeah. And we don't know exactly. Is that for power? Is that for data? It goes back to like a

Kevin Periera: massive, like battery that's like 16 feet behind him. And we don't know. It's almost like, what's that?

There's that meme of that guy with the giant brain that goes backwards. You don't have It's four. Yeah,

Gavin Purcell: exactly. It's four giant hamsters on a wheel running in tandem to power this thing. But it also has a camera built in so that you can hold up stuff in front of you Yeah. And kind of circle it and then think or mouth ish with your [00:27:00] words.

Yeah, what is this? Or remember this or whatever. And it will be contextually aware. And so they show it connecting to another user. Conceivably you could use this to send text messages or to direct an AI to do things for you. Sure. Um, this gives Call of Duty players a new way to be racist in lobbies. Like I do think this is gonna unlock a lot of stuff.

Kevin Periera: Yeah, and I think the, the thing I wanted to talk about in the reaction to it really has to do with, there's a lot of people who saw this and are like, I hate this. And I think this is where we're getting into, where this technology is gonna start to change our world in different ways. Because you can imagine a world where in the future.

You know, think of the hooded hooded character who's just like, like, doesn't say anything, right? Like just thinks his life. There's a world in the future where like everybody's choosing different ways to interact with each other. Like I think I will probably talk out loud for the rest of my life. I like to talk, I like hearing from people, but who knows what my grandchildren will do?

Maybe there'll be this [00:28:00] very busy world that exists where they'll hear thoughts and they'll send thoughts and there'll be all these things. But again, this just gives you a look into a future that like, I don't think is tomorrow like this thing. You're not gonna see a lot of people using this thing right away, but like if this works as intended, it's another kind of cultural shifting moment in the space of technology.

That could be very weird.

Gavin Purcell: Yeah, I mean, I could look. It'll be multimodality our future as we know. And you'll pick that modality. If you want to talk, you'll talk. If you're in a, an area where you feel like you can't, or it's inconvenient to for whatever reason, you'll do a weird mouthy thing and it will, yeah, maybe send, um, I think it's important to note doubt.

People are like, I don't want AI reading my brain. That's not how this works. This is not re it's

Kevin Periera: not telepathy, it's not gonna be able to read your brain. You are sending it out into the world.

Gavin Purcell: Basically, yeah. And

Kevin Periera: my signal's not strong enough for that

Gavin Purcell: anyways. We all

Kevin Periera: know

Gavin Purcell: that.

Kevin Periera: Let's move on to CD Dream 4.0.

So this is a new AI image model from by dance that kind of set the, the internet, uh, especially the hype piece on fire this week. Every, every week I find Kevin at the hype [00:29:00] piece, just like, who'd you fire this week, Gavin? Yeah. They, they jumped from thing to thing, and this week it was C is C Dream 4.0. And it is very good.

Great. It is another's great. Very good image model. I feel like in some ways when we get to these, we're now at the place, which is just so funny to think like, you know, when we first started this podcast and we saw Midjourney go from like blobs to like kind of more shape full blobs. Yes. And we were like, whoa.

That's incredible. We are now at the place where you're looking for very specific types of, of improvements. And what's cool about seed, uh, seed Dream is. It's another one of those very realistic models. There's some really interesting examples, um, on their website itself, if you go through, they have a pretty good blog post that shows a lot of this stuff off.

You know, they show how they, you can edit within prompts. They show a bunch of different styles that you can

Gavin Purcell: use. Well, that's, um, like that to me is the funniest thing about this as we continually move the goalposts on. What is impressive here is that. It used to be like, oh, it's a, it's an image model that's crazy fast.

Yes. Like to generate basic outputs and it can do up to [00:30:00] 4K imagery. That alone should have been showstopping, but now it's almost like table stakes and it becomes, well, what else can you do? Can you take a single image of a character and then generate them from multiple viewpoints and like Yeah. Have coherence and consistency?

Yeah. Yeah. You can do that. Of course you can do that. Well, can you take a logo and apply it in one shot to multiple different things with a natural language prompt? Yep. Got it. Yeah. You know, can you feed it a single image and have it do an entire brand campaign? Can you edit that image like you were saying?

Um, you know, does it un, can it like build infographics with perfect text and multiple paragraphs worth? Yes. Yes. Yeah. It kind of does all of that.

Kevin Periera: And, and what's interesting is they're all starting to do that, right? Yes. So this goes back to the same way when we talk about like where you do your AI might be where this is.

Nano banana, we've talked about a bunch and it's doing some of this. This one, I think from a professional standpoint feels like it is now kind of taking the lead and, and there's a lot of really interesting examples in that. Specifically also to shout out our, our friend [00:31:00] F-O-F-R-A-I is doing interesting stuff, always explaining and showing cool models.

But in this case, like there's a video, there's a thing where he's showing off just how realistic these images look. And Kevin, that sheep image. Now that looks like a weird picture that a farmer in somewhere, say, like in New Zealand, decided to take a photo of his buddy, that he just thought, oh, isn't, isn't little.

Uh. Isn't little, uh, sheep sheepy today looking Very good. Well, let's take a picture of him and let's take a, like, that looks like somebody's photo, which I think is just like a remarkable thing. Where is this farmer located

Gavin Purcell: again? I'm having trouble with replacing that access

Kevin Periera: somewhere. Somewhere outside of America.

Somewhere outside of America. Put, okay. Yeah, that narrows it down. Might have been closer to Scotland with that broad of

Gavin Purcell: a range. I think you nailed it, buddy. Yeah, a hundred percent. Thank you. I appreciate that. The, the photos of humans coming out of this model. Yes. With the, uh. Imperfections. The non-plastic skin, the way light seems to scatter on their face, like the, the peach fuzz and hair and all that [00:32:00] stuff is just like, there was a, a common gripe of like AI face, right?

Yes. Unless you use a specific realistic, Laura, this. Blows that away. Um, I just, it's, it's a really, really beautiful image model. And it's like, again, it's out now and you can go use it and it's cheap and there you go. Like wild. Yeah.

Kevin Periera: Oh, go use it. Go play with it. Um, again, if you're doing any sort of creative work, the other thing to be aware of is like.

If you're doing a sort of video work, like we're getting to the place where you can really paint your entry and exit images in a way that you can get very specific shots and this is the pathway to that. Alright, Kevin, uh, another cool story This week, 11 Labs, which you're big fans of, we are working, uh, using their models for some of our stuff on our startup.

Um, they have released a new tool that allows you to remix voices that already exist on 11 labs. And what's very cool about this, Kev, is like. You can make the voice older or younger, you can add additional things to it. So do we wanna try spinning this up and just giving you a quick go and see what we can figure out?

Gavin Purcell: [00:33:00] Let's do it. So I've got, um, I have a voice here for like a troll character that we were using for something. Oh, it sounds like this. Okay, cool.

Troll Man: Right then little morsel. Okay. What have we healed? That sounds like the

Kevin Periera: beginning of, uh, gorillas, you know, that. That, that song where, oh, it's like, ha ha ha, ha ha.

Yeah. Hopefully we didn't get flagged for that. Three seconds of, man, man, it was

Gavin Purcell: so good. Well, someone's gonna do A-D-M-C-A who lives outside of the US and is a farmer. You're right. So we're already getting flagged for that. That's true. Perfect accent that you did. So we have that troll voice. What would we like to do with it?

Do we wanna try to give it an accent? Do we wanna change the age? Do we wanna change the emotion? Yeah. Make it a baby. Let's make it a baby trick. Oh, okay. Sorry. This prompt doesn't follow our safety guidelines. Okay. I'm gonna say make this an angsty teenager troll. Okay. And I'll say with 11 labs, their new voice creation stuff is kind of insane.

The way you can describe voice, pretty amazing voice, and you get something out that's good. And with V three, it's really performative the way it can generate laughing or, or even throat clearing. All right, here's [00:34:00] the angsty teenage troll.

Troll Man: Ugh. Seriously. Like why even bother? This whole thing is just beyond lame.

No one gets it. That's pretty good. You know? It's always do this, do that.

Kevin Periera: Okay. Wow. That sounds like what my daughter said. Sometimes it's like, what's going on here? That's very

Gavin Purcell: angsty. Let's say you can attach this generated voice to remix it further. All right, let's attach that voice.

Kevin Periera: Let's go for a Scottish farmer.

Let's go for a Scottish farmer. Why not? Let's see what it actually sounds like rather than me faking it. Or we could have done New Zealand farmer, but the Scottish farmer's probably done. I'll go New Zealand. Here we go.

Troll Man: Right? So another day, another early start, you'd think after all these years, I'd get used to the rooster, eh,

Gavin Purcell: the rooster.

Did you write that in? No, it just generated it based off of New Zealand farmer. I see. Got it.

Troll Man: Yes. Right. So another, another Hurley [00:35:00] start.

Gavin Purcell: I guess I'm not a good judge of New Zealand accents. Yeah. I mean, if you're an audience

Kevin Periera: and you're, you know, is that, is that a New Zealand accent? I mean the Right sounds like it, it doesn't sound like a American at least.

So anyway, the cool thing is you can go and use this now it's out if you're an 11 labs, uh, user. And it's just a cool way to think about like, okay, you've got a voice out there that you're using for like an AI film or something. But maybe you think it needs to have a, a slight attitude change, or it needs to be in a slightly different vibe.

One of the things we found with 11 labs and, and creating voices for and then is like. You can find a really good voice, but maybe you might want it to be more performative or more a certain thing, and you don't want to create one from scratch. So this allows you to kind of like tweak it. It's almost like an in painting of the voice in a specific way.

Gavin Purcell: Yeah. Yeah. Well, look, I love what they're doing. Love the V three model. So kudos to 11 Labs.

Kevin Periera: So another couple quick things. Oboe just launched. This is a new startup from the guys behind Anchor. Um, this is pretty interesting. It's a, it's like kind of trying to be like a dual lingo, but for lots of different things using AI models.

Not a, not a [00:36:00] huge thing, but I think one of those things where like somebody is gonna crack this as a space to be, um, if you wanna go to, you know, a lot of people will go to chat GT and say like, teach me about blank and blank. One of the things that we have said for a while with all AI tools is that like.

There are ways to break apart the use cases of a general chat model that exists, that people want very specific looks at stuff. In part, that's what we're doing with, and then in one direction, oboe is doing it in another direction. And I think there will be lots of interesting businesses that come outta this idea of like, I've said this on the show before, like no one has really cracked like the AI coach yet in a specific way or things like that.

And oboe is just another way of like, okay. It's learning, but I wanna go deeper on it. It's an interesting thing to do. Go give it a shot. It's something that I think is interesting overall. Alright, so a

Gavin Purcell: story that we, I think we brought up. A while ago was that, um, the sphere, uh, Madison Square Garden Entertainment was kind of modifying with the help of Google and their suite of tools, they were taking the original Wizard of Oz and [00:37:00] applying it to the sphere like that means up, resing everything to an insane resolution.

But that means also having to hallucinate. What exists or existed beyond the frame of wherever the characters were because, you know, you're going from a traditional movie theater size to the massive sphere. This process took a team of, of artists and creatives and, and, uh, software engineers a a, a long time and about a hundred million dollars to bring this thing to the sphere.

And everybody's like, well, why would you do that? That seems very expensive. I mean, it's a cool tech flex, bro, but that seems very expensive. And then you realize. Off of the $200 that people are paying to watch a film that was released 85 years ago, they're making about $2 million a night. Yes. For people to go and watch a film versus having to pay U2 to stand about and perform on the stage.

Yes, that could be a good return.

Kevin Periera: So this is the, the big story is that that's come out is like, this has worked, right? Like this has really worked for [00:38:00] the sphere. And if you're not familiar with the sphere, it is that giant circular globe that's in Las Vegas. And their plans, I think are to open up more of these across the world.

This Wizard of Oz experience, there's videos of it online where like they, they do things like, you know, you see, you look up and it covers the whole screen. And the AI stuff has, has been made up to your point, has made up the other things. But they do things like they have fans blowing during their tornado.

So it's a little bit of like this immersive experience, but this is a real idea of like, okay, AI film. Meets like an older film and there were a lot of people upset that this was happening. But Kevin, this is like a new experience, which is very cool. It is that thing we've said before in the show, which is.

AI will allow you to do things that you couldn't do before in an interesting way. And, and this isn't just making a traditional square film, this is taking a really beloved film, obviously, and people have feelings about doing anything to that, but changing it into a different environment and like this is only really possible with ai and that is just a cool thing to see.

And now there's a business model for it. So we expect to see more. One of the other things that come [00:39:00] outta this is Harry Potter might be one coming up. So if you're a potter head and you have a lot of feelings, and I know a lot of people have feelings about Potter in different ways now. But. You can imagine a world where if you're seeing a quidditch match and you look up and there's skies there, and there might be other people flying by, like that's a very cool thing.

It also gives people a reason to go to the theater, which is also a fun thing for sure.

Gavin Purcell: Yeah, and you could, you know, you could take a autonomous vehicle to get to that theater. You know, Tesla's a big, boring holes underground. You could take one of those if you want, or you could hop on the back of your unitary robot and just let it, you could jansport you.

All the way

Kevin Periera: to the sphere. That's right. We've covered unit tree robotics for quite a bit on this show. It's a Chinese company. They're filing for an IPO, Kevin, so this is going pretty well. One of the interesting things that come out of this was that there's a 65% of the revenues coming from their robot dogs right now, which is a big deal.

If you've seen the robot dogs is, they're very, you know, those are the ones that kinda walk on all four and do a lot. We've said before. They're making stuff cheap. They're, they're getting mass market and in [00:40:00] China, I still do want, maybe you and I can get, maybe someone out there will sponsor us, not the state of China, please.

We don't live in that world. Maybe somebody will sponsor us to go fly to China to go see these robots and do something interesting. 'cause I think I would like to see, like, okay, what does it feel like? In China, am I surrounded by these robots all the, all the time? Because it sure seems like that's happening there, right?

Yeah. And this IPO probably will be a pretty big deal overall. And it's just another example of like where the robotics world is at right now.

Gavin Purcell: I. I get triggered easily these days. Gavin, I know this. Um, and even in Santa Monica, the amount of robots and electronics that roam the streets and take up the sidewalks is too damn high.

I am already tired of the little cocoa little baskets on wheels. The robots are

Kevin Periera: so dumb. I don't understand the Yes, and,

Gavin Purcell: and they, and they operate. They park on the sidewalk and take up space. Yeah. And then when they're driving down the sidewalk, they're right in the [00:41:00] middle. Yeah. And you're supposed to navigate around the robot.

Like no. They, they, we've already planned our cities so poorly around pedestrians. Like, what little real estate you have is now being taken up so that a wally bot can get somebody a burrito in five minutes? Like, no, no, no, no. I'm gonna press X and flip it like a wart, hoog. Oh, you're, youre there. There's gonna be footage of me the next time.

I'm, I'm working on my squats right now and my deadlift. Okay. 'cause I am, you're gonna flip that coco clip button right over flip. I'm gonna flip a cocoa bot. And you're gonna see it on X. No, you're here to hear you

Kevin Periera: heard it here first and

Gavin Purcell: it's premeditated. Y'all heard it. Wow. Premeditated. Someone's not getting a burrito tonight.

Kevin Periera: I'm Coco's lawyer right now, writing down your name. I'm taking a picture of you. I'm getting all this stuff. But Kevin, maybe if the Forer robot, this new, cute little robot, was the one delivering it and was there kind of walking up to you and said like, hi Kevin, I'm Forer. Maybe this would do, maybe this would make you feel better.

Gavin Purcell: Hey, Forer, dear little blinky eyes and your cute little screen. Do they, do they reflect pain? They'll get outta my way.

Kevin Periera: Human.

Gavin Purcell: [00:42:00] Oh, oh, Forer. I'm coming for you. I I'm coming for you. Anyway,

Kevin Periera: this is another new humanoid robot coming from a company called Forer. You can see that if you're looking at the video, they have kind of like the face that has like, kind of like cute eyes.

It's expressive. I'm a big fan of this sort of thing. It, it's just another one of these companies, right? It's another one of these companies that's making humanoid robots. And I think Kevin, if you, if people are listening to this show, know we talked about human aid robots for a while, but like

Gavin Purcell: the interesting thing about the, the, the, the four year bot, uh, is that, you know, it's, they're going for cute and whimsy because this is supposed to be a companion and care robot.

Yes. Like this is supposed to be the robot you want to go give a hug to until someone doesn't dog food. The firmware update properly and it will karate chop your carotid.

Kevin Periera: That's right. So I, and I think overall, this is just another example of like, we're gonna be entering a world and what I was getting at before is this idea of like, you at the home may be listening to this and know our updates on humanoid robotics, but like, until you see like the Riz bot, which we've talked about walking around your town, you may not think of this as coming, but like.

We are not far away from a [00:43:00] world where Kevin is talking about where like there are multiple robots on the street and you're gonna have to interact with them. And like I do think you, Kevin, your point about like people getting mad. We saw that with Waymo's and I think people are gonna get past that because Waymo's are designed to be in on the street in a, in a car sort of scenario.

These are things that are gonna move around and you can imagine New York City, can you imagine New York City where like. There's like four little robots walking up and down the street. They better get real good at like dodging and moving. Yes. And SW around people. 'cause if they don't, there's gonna be a lot of them that get pushed to the curb.

I feel

Gavin Purcell: like I listen and I, I'm not. I'm not saying I'm for that. I'm just saying you are saying you're for that 'cause I, yes, yes. You're for that. I'm not exactly saying that. I'm just saying we have to redesign our cities. I'm tired like, like there's not enough space for pedestrians as is. And when we got these bipedal things moving about, we have these little delivery robots and these little robotic dogs moving.

Yeah. They need to step aside.

Kevin Periera: Do you think it needs to be like a robot [00:44:00] lane? Do you think that there needs to be a read? So, so imagine a world where like, yeah. 'cause one of the things they've talked about is like maybe in the self-driving world we're not gonna have as many cars in the road or there'll be more efficient, right?

Right. So that would give you maybe more space, maybe like a bike lane, maybe there should be a robot walk lane, a

Gavin Purcell: lane where there's. Be a little bot lane and make that one closer to the cars. So the bicyclist get a little protection until they

Kevin Periera: get run over. Is that what you're looking? Well,

Gavin Purcell: I mean for better them than us.

Gavin is all I'm saying. Sure. That's And and that's on the record for the robots of the future. Yeah. But you know, there's always a little trail sign of like what yields to what and it's like there's the horse and it yields to the biker, which yields to the pit. Yeah. We need that for delivery robots. And they are at the bottom.

Of whatever priority scale there is

Kevin Periera: another big job. Someone out there is gonna have to make that sign. Someone's gonna actually have to create that sign. So that's another job that AI might have created. All right, everybody, it's time to figure out what you did with AI this week. It's ai. I see what you did there Times

Troll Man: without a care.

Then suddenly you stop and [00:45:00] shout.

Kevin Periera: Okay, Kevin, we have three quick things. First and foremost, I wanna start off with this amazing series of videos I love, uh, this was pointed out by Venture twins, who we always find interesting stuff. This is a village where watermelon is everywhere, and this is just a Romanian watermelon village. This is from a user on Instagram.

Called the Greco, the THE. It's so fun. D-R-E-C-U. But you just see like a co. A place where like, you know, watermelons kind of exist to be wood. There. People are playing pool with watermelons. Like it's a world where watermelons are much more, much more present. The watermelon

Gavin Purcell: church bells get me, like I, it's just I wasn't expecting that shot.

Kevin Periera: I think maybe this guy, if he's out there, if you're listening or, or, or woman or you know, it's like, let us know because like this is the guy who could finally make hotdog city into something interesting. Like we don't have the time necessarily to do the deep work that hotdog city [00:46:00] demands, but like this is Watermelon City, right?

And maybe we've just been beaten to the punch and watermelons are funnier. But that shot of the church bells. Yeah, reminded me of like, we tried to do this a while ago, and it's just so realistic now that it's just fun to look at.

Gavin Purcell: So I, I went to, uh, at a mahi greco's, uh, Instagram says, award-winning filmmaker and multimedia artist pioneer of, uh, AI art.

And if you scroll down, there is a shot of a, uh, um, if it is the same Romanian village, there's a shirtless Romanian man with a gold, uh, medallion cross, like dangling into his chest hair, riding in a watermelon Lambo as it goes down the streets. Chomping watermelon. Amazing. And that is amazing and it's, this might be my new favorite thing ever.

Bless your hearts for coming up with great prompts.

Kevin Periera: Alright, let's move on. Uh, uh, a friend of the show, P jce, who you may or may not know, is like one of the guys who does really incredible, um, AI commercials. He's also kind of got very famous early on for doing AI versions of the studio. Oji BLI Films has a really great studio.

He has launched a [00:47:00] new, uh, uh, series that involves, um, NFL highlights. And if you remember Kevin, like a long time ago, I did something similar to this. This is better. We'll get into what makes it better, but like, there's something really interesting about using AI characters to reference the sports world or use characters to show this stuff.

So football drives a lot of conversation online. It drives a huge amount of conversation, especially Monday, Tuesday, Wednesday in America. 'cause the games are on Sunday. This is a new series called Mortar Boom where uh, PJ has basically created a kind of a sports Centerlike show, and one of the characters is a normal sports center figure, and the other character is a demon.

You wanna play just a little bit of it, 'cause the audio, and this is fun.

Molto: Huge hit number 10 is when Reed Blankenship does huge hit on Jake Ferguson.

Troll Man: Mock tar boom number nine. Patrick Mahomes trucks. Dante Jackson Ho Motar, boom. Love the ring of that. What's his name? T Tar Boom is battle Cry to

Molto: celebrate Warrior who died.

Glorious concussion death.

Kevin Periera: Very cool. Wow. Mock tar. Boom, baby moar, boom, boom. Yeah. So anyway, its just a fun, the fun use case of AI in general. [00:48:00]

Gavin Purcell: Well, I, I found maybe the best use of AI in general. Um. Oh

Kevin Periera: no, I just saw the name of this. Are we gonna end on this sort of thing again? This is a Gavin now,

Gavin Purcell: hold on.

Don't spoil it For the audience. This is a vibe code. Someone found out that there's a hidden tool, uh, that, that developers usually don't have access to within a MacBook that can, um, report the degree of the hinge how open Oh. Or closes it is. Right. And this could be very. This could be very beneficial for, let's say you need the screen to go to sleep.

Um, you know, 'cause you're, you've closed most of the screen or whatever, it's closed now, so it can actually report the, the operating system can report what the degree is. Well, someone got access to that and they, uh, turned out a creaky door application where if you start to close your laptop lid, it sounds like a creaky door.

Wow. That's very great. That's interesting. Very fun. And of course, someone vibe coded, which is. The reason this is AI related, Gavin, uh, Ian not all, uh, released an app called, um, fart Scroll lid. And, oh, Lord, as you go to close your laptop, Gavin, um, it based off the [00:49:00] degree of the closure. And the uh, speed with which you close the velocity, it will, it will generate.

Wow. Very realistic fart

Kevin Periera: sounds. I didn't really want to end the show on this sound again because last time two weeks in a row. If you made it last time, you know what we were talking about. But in a row. Here's the cool thing, like this is one of those weird vibe coding use cases that is like I am all for practical jokes.

My family, we have a large history in practical joking each other, and this is just a fun thing. I might do this to my wife. She's not listening right now, but she will. Despise this because she believes that she has like a, a, like a demonic soul that exists in her computer because her computer for some reason, I don't know what she does, but like she does things where like her computer does like strange stuff all the time, like in the way that like my computer never will, and it might be that she's installing stuff.

I don't know what it is, but this will make her so mad that I might actually this.

Gavin Purcell: Yeah, it could be the, the 12 kaza light instances that she has running in the background. Yeah, exactly. But you know, maybe you should change the fart to like subtle whispers of your wife's [00:50:00] name or something. Oh, that would be even creepier, right?

It's just like as she goes, a close,

Kevin Periera: leave him. Leave him. That would be like, so I can start setting myself up to be like, disappointed by having my wonderful wife walk on the, then she comes in and goes, Gavin I, I play a practical joke.

Gavin Purcell: Gavin, I hear these voices all the time in the shower, in the car when I'm sleeping you're like, no, I only

Kevin Periera: did it on the laptop.

The laptop. Alright buddy. We'll see y'all next week. Thank you so much for joining us. Bye-bye.