111_audio

Gavin Purcell: [00:00:00] The CEO of Anthropic just warned that 50% of jobs are going away. Kevin, the end of white collar work is n. The mass displacement is coming. I'm, I'm kind of freaking out here. Hell, I told my own staff, you're committing career suicide if you're not spending 10% of your day experimenting not to add fuel to the fire, but we just found out that foundational AI models like the latest ones.

I've been caught blackmailing and trying to social engineer Kevin. It, it's clear that machines aren't content just for our jobs. They're trying to take our souls too. On, on the plus side, Google's VO three model is so good. It might've scared away. Major competition and it's continuing to delight and Oh, sure, sure.

You mean you mean distract and pacify? I've seen this playbook before. Gavin, what do, uh, listen, there's a lot. Of news to get through this weekend. You seem particularly amped, you know, between Chinese fighting robots, Google's new Stitch ai and Claude itself getting a voice. I'm just, I'm, I'm, I got a lot to process here.

Okay. That's fair buddy. What if we, um, what if we let [00:01:00] open AI's advance voice mode, which just got new singing capabilities? What if we let it soothe you with a rendition of a wonderful world? That sounds great. All right, let's do this. Oh, I see trees of green red roses too. No, no. Shut it down. We need to nuke the skies, Kevin.

That actually made me feel pretty good. No, no. Block out the sun immediately. This is AI for humans for now. Everybody for now. And I think to myself, what a wonderful world.

Welcome everybody to AI for humans. And there is a. Story bubbling up today, Kevin, that got us all excited. As you heard at the top, uh, there's a big news story from Axios today, uh, which is funny, that it's Axios reporting that Dario Modi, the CEO of Anthropic is going on the record to talk about potential job losses coming up.

The, the headline here is that [00:02:00] 50% of white collar jobs could be eliminated in the next one to five years. Wasn't an interview in person with Dario. It was a, a written interview where they basically said that he would go on online and talk about this. Um, what's been funny about this is the Axio CEO actually went on morning Joe this morning, um, which is a big mainstream news show and talked about this and was, is starting to raise the flag a little bit.

It hell, I told my own staff, you're committing career suicide if you're not spending 10% of your day experimenting with the technology. And I don't think most people are. I think here we are, Kevin, here We're Good morning everybody. Thanks for hopping on the, uh, Microsoft Teams meeting extra early. By the way, uh, you're all committing career suicide.

If you don't have another tab open right now. What a great way to start the morning huddle and what a great way to start this podcast. Gavin, I'm really glad you slotted this one as the first story. Yeah. Tell me, so you read this thing, obviously, and it's pretty interesting. So tell me a little bit about your thoughts on this first, then we'll dive in a little bit more to what Dario's overall takeaway is here.[00:03:00]

I mean, my thoughts are that something feels oddly coordinated about this and I don't know what it is. I feel like a lot of people have not that. Not that people haven't been pointing to the, the slightly wheezing canary in our, uh, employment coal mines for a while now, right? For like the last few years.

Displacement could come mass unsettling, shifting, uh, shifting workforces and new jobs. But it's something feels like what just happened. Like right now. It seems like everybody just suddenly started talking about it in a much more heightened way. Am I wild for feeling that way? Uh, you know, wild, I, I think I have a theory, which is.

And I think, just to be clear, what Dario has gone on the record here, and he is very much on the record. In fact, in this Axios report, which did just break this morning, he very much says that like, you know, 50% of white collar jobs could go away. The other's kind of scary number is a 10 to 20% overall unemployment rate, which you know, as you've been listening to this show, if you are, if you're new, thank you for [00:04:00] joining us.

But if you've been listening to this show for the last couple years. We have talked about this danger that could come, I don't know if this is coordinated as much as I think this is just a clear signal that the both government, American government and governments on the, around the world, but also that there are people that want to make sure they understand and A understand B, don't get blamed for when this happens, that they didn't say things were going to happen because again, right.

I think you and I have heard this and you, and I believe this, I still don't think the mainstream understands this. And when I say the mainstream, I mean. The average middle class worker who's been doing his job for like 10 to 20 years, he's playing with ai. He sees it, but he might not be seeing the advanced versions of ai.

He may not think that he or she may not see that their job could be eliminated. And I think what Dario is trying to go on record here was saying is. Look, we have seen something. Now we're gonna talk a little bit about cloud four and like whether or not that something is immediate or it's coming, but the one to five years number is [00:05:00] something that's really important to see.

Google Anthropic, OpenAI, all of these companies have put kind of 20, 30 as the mark as the time that they almost all universally believe we will hit what's known as artificial general intelligence. A GI. And I do think this is like a shaking of the, of the rattle to make sure that the real world outside of this little bubble that we have talks about this stuff.

If, if, for those out there who didn't have the privilege of having a, a bully of an older brother growing up, this is, this is like titans of industry. When they windmill their arms and they say. I'm gonna walk towards you and if you get hit, that's your fault. I've told you that I'm windmilling my arms and I'm walking directly towards you at a rapid pace.

And it's sort of like, well, okay, you get a, a, a pat on the back for saying all of the things that you're doing might cause mass unemployment and unrest and massive job displacement globally, but you're still swinging your arms and walking towards us. So how about. How about [00:06:00] you give the solutions, and I think I just wanna say this 'cause I do think a lot of people go like, oh man, I, they weep for the younger generations, Gavin, right?

They feel bad for those that are, uh, maybe junior seniors in high school, they're graduating, they're gonna maybe, maybe go to college, maybe about to enter the workforce in a, in a meaningful, significant way. They weep because all the entry level jobs are gonna be displaced. But I have a weird feeling that like they'll be able to pivot.

Just fine. They'll be able to, at the beginning of a career, be nimble and find new opportunities and leverage ai. It's the, I don't know, 40 somethings, it's, yeah. Yeah. In the 50 somethings that are like, it's supposed to be like gliding through the middle of whatever their career might be that are suddenly gonna wake up tomorrow and the huddle is like not the good one to be in.

There are two. And you're in the back one. Well, it's funny you say that because I actually read something recently that said the 40 really forties year old people are the ones that are adopting this technology the most right now. And that it might be the millennials and, and younger Gen Z people that are, [00:07:00] that are gonna struggle because they're not spending as much time.

Although you think about PE people in college using it. I will say, I think the. Youngest generation, the 10-year-old, say the 12 to unders age group, is gonna grow up with this in the same way that many people grew up with the internet or grew up with computers, right? And they're gonna be fine. So Dario did not go on video to do this.

He did not go on CNBC himself. He went and talked to Oxy Axios, which just FYI. If you're not familiar, Axios is a very mainstream publication. Like this is the kind of, they do a lot of coverage of politics and other stuff. So this is a way to get your message to the mainstream as is obviously Morning Joe.

But just last week, Claude four came out and Dario did have this to say I thought was a really interesting short clip from his sit down. When do you think they'll be the first billion dollar company with one human employee? I. 2026. Yeah, I I absolutely buy that. Yeah. So this is a, a, a recurring line we've heard in the AI space, right?

This idea that there will be a billion dollar single person or two person company that starts up, and that's such a positive [00:08:00] thing when you think about, as the person who starts the company, you're like, Hey, great, I can make a billion dollars, but you know what? A billion dollar company by one person doesn't have.

Is the other like 20 to a hundred people that might work at that company. So I think these are kind of two sides of the same coin. And I, I, no one out here, I mean, I'm not personally advocating for us to like put a pause on this based on this idea, but there was a really interesting research study that came out, uh, not that long ago, I think about a week ago, of mainstream people.

And you know, mostly people are afraid of this stuff, Kevin. They are not actively kind of leaning in and being excited about it. Mm-hmm. So I do wanna spend a little bit of time before we move on here to talk about a. What you can do if you're listening to this, obviously you're kind of ahead of the game, but how to kind of help people along in this journey to help them understand a little better What, what sort of advice do you have for, let's say, the AI for humans listener to kind of like get their either coworkers, their family members?

I. Or other stuff like that to kind of get them aware of where we are because I don't think this report is, is like saber waving and him trying to [00:09:00] look good for the world. Personally. I think this is him trying to warn people that there is massive change coming and we're not ready. I. I just don't like, I get what you're driving towards Gavin, and I think it's, honestly, it's a little gross, but if you want me to say the best thing you can do is like, and subscribe to the AI podcast.

That's not what I podcast saying. That's not what I'm saying. Tell your friends. That's what its, it says the plug it in the rundown and you keep So thought we were gonna do that in a little bit later, but that, that's what I'm, they like and subscribe. No, I think that they are concerned. Look, I. From the individual standpoint, we scream it every week.

You have to start poking at these tools. You have to see how you can make your own job obsolete before someone else does. And I don't mean to sound super alarmist that way. I actually like, I'm having fun trying to automate away portions of my existence every single day, trying to see can I build systems that, that, that get things done a little bit quicker.

That seems like every day there's a new model and a new set of tools and a new something that can. Automate things. So you have to get curious. You have to start poking and prodding [00:10:00] right now. You have to just fail and fail often, but you cannot be, uh, you can be afraid of what might happen if you don't pay attention to these tools.

That's totally valid. You can be afraid of corporate greed or government overstepping or of the million of other things that are outside of your control. But what you can directly affect right now is your own level of interest and exploration. So. Totally start creeping in. I think the other thing is like when you go to show these things to other people, you have to be aware that sometimes they might be approaching it with a different vibe than you are.

And that's something that's really important. Like when I, my kids were not that into this stuff, and then I showed my younger daughter who could help her with her math homework in terms of tutoring her through certain problems, and then suddenly I saw her using it all the time. So you have to find for the person you're talking to.

What the use case is, that matters. And we as a society have been terrible about onboarding people at large about this, which is why I think this research paper came out because most people just hear these stories about job loss. They do not see how it can help them in some ways. Now we are gonna [00:11:00] get to later talking about.

VO three and all the people that will, that are angry about what that can do. But maybe really quickly before we, uh, we move on, uh, Rick Rubin, Kevin has had a really kind of big part of this new Anthropic drop Cloud four. Um, mostly he wrote a book, but now he's being doing interviews for them. And this actually does talk a little bit about the idea.

Of how not only coding is changing, but I think gets at the heart of how like kind of work is changing in general through a really interesting music metaphor. So in the past for music, you had to go to the conservatory and study years and years, and then someday you could play in a symphony. And then when punk rock came along, you could maybe learn three chords.

In a day. Then there were all these bands, and that was how I started in music, was punk rock. If you had something to say, you could say it. You didn't need the expertise or skillset other than your idea and your ability to convey it and why Coding is the same [00:12:00] thing. It's a punk rock of, uh. Coding. Yeah, so there you go.

We got a new, we got brand new. I'm not doing my homework, mom. I'm building Android games. That's me for good. I'm by coding so hard, dude. I'm leaping off my laptop and just catching sick air. The only thing I'll say about about that in my mind is it just kind of speaks to this idea that. If you are a middle manager, you know somebody's a middle manager and they do a job that is kind of drudge, drudge refill, but they have a job and they're getting paid for it.

Totally understand. This may be the time to start leaning into the thing that that person wants to make on their own, because the more they know how to do that with these tools, the better they're gonna be. I. I don't know what I don't know about the behind closed doors, things, right. You and I, the stuff that you and I are discussing, Gavin is the stuff that we have access to, which means there was a model that was finished six months ago that's way more capable, and there's probably a research project that was started two and a half years ago, which is going to dramatically change everything else.

But with the information we have at the moment, and just [00:13:00] looking at the way these things trickle out, there's a lot more stuff that could be automated today as you and I are talking and it's not right. Inertia is, is, is real. Uh, it, these things change typically pretty slowly, um, even though we cover it every week, so it seems blindingly fast, so rapid to anybody who's listening or, or, or watching.

Uh, this, like, I, I get it. It seems really, really rapid, but there are so many things, um, legal review, uh, code review bug, this, that there's so many things that are really. Well automated today that probably won't be automated until a few years from now, just because it takes time for companies to become comfortable with and adopt this stuff once they see the savings.

Don't get me wrong, they're gonna race towards it, but it, it, it's, it, it isn't going to happen overnight, but you're kidding yourself if you don't think it's gonna happen in a few years. So last week right after we recorded the show, of course, uh, Claude dropped, uh, uh, of course Opus and Sonnet four. This is their new models.[00:14:00]

Um, and Kevin, I think the, the overwhelming response has been positive, mostly, especially on the coding side. There has been some, some chatter in the world about the idea that perhaps this isn't the leap people were expecting from cloud four, but from a coding standpoint, I. All of that. I have seen, and I've not spent a crazy amount of time with this personally, but all that I have seen is that people are saying it is one shotting a lot of coding stuff.

It's working really well. I did one small, um, personal, uh, I did one small personal experiment that I couldn't fricking subscribe to Claude. It was like a billing issue, which is the most weird old school web one problem. So I wasn't able to part, dude, you got a vibe code, you an assistant that can help you subscribe to cloud.

I need an MCP server that can take my credit card. Fill out the subscription and make sure I get it and all this other stuff. So anyway, I, this is something that's a big deal. The benchmarks were really good, but they weren't like, mind blowing, right? Like they weren't like the, the idea that you would come out and be like, oh, this is so much better than Gemini 2.5 Pro, or, it's so much better than oh three.

So what's interesting to me is like, is this the last [00:15:00] of this generation we're seeing come out and are we gonna see the next thing? You know? I, I used it all weekend, um, building projects, uh, for personal and professional. Um, I, it is now my go-to. I use it as the primary driver. I feel like sonnet is actually better.

Oh, and. Well, the reason being Opus is very capable, right? You can give it a complex thing that you wanna build, or a bug you're trying to squash, and it, you can just watch it reason, and it seems so elegant and it's fast, uh, at least again, with my, with my usage over the weekend, but it still is overeager in a way that is frustrating and so annoying.

Mm-hmm. I, I use what's called YOLO mode in cursor, which basically means is that a, a real thing? Oh, I didn't realize that. That's funny. It's, yeah. Yeah. It's a thing. And basically like when you let these agents do something for you mm-hmm. You're building a piece of software or it's creating a plan to write some software, um, it can run tools.

I. It can call those up and usually you have to sit there and click a button and say, yep, I approve it. Go ahead and do this. I yellow is like, I'm doing it, man. I don't give a crap. I'm going for all, [00:16:00] that's it. It's punk rock. Gavin? It's, yeah, it's Rick Rubin. It's Rick Rubin mode. I usually run yellow mode and I have some exceptions.

I don't let it delete files on its own for good reason, and I don't let it make like major changes, but. For the most part, it's weapons free. And I had some fascinating things where I would hit a button and walk away and come back 10 minutes later and it was still chewing through the task list and it was building the app and it was working and it was testing it as it went.

And I was so amazed. But then I had that one experience, Gavin, where I went to go get a Tasty Coke Zero from the fridge. Hashtag not an ad. Not an ad, but the orange flavor is the taste of the summer, my friends. Oh, it's so good. Horrifying. Oh, it's so good once it hits your lips. And I sipped back and I looked down and I realized that it is rearranging my file system.

Oh geez. On my laptop it was having trouble running something and I was like, locally. Yes. And it was like, I'm just gonna go ahead and just reinstall like Python. Wow. And I'm gonna add all these libraries and I'm not gonna do it in the environment that you have loaded. So now you have files scattered everywhere.

And it was just a, [00:17:00] oh no, how long have you been doing this? And I'm scrolling up through the conversation and realizing like, it's not, it's not there yet. The same way autopilot is it of true robo taxii on a Tesla. Yeah. It's not ready to be the full agent, but I, I really enjoyed it. One thing it didn't do, which I wanna point out, was blackmail me.

Good. That's good. Which I did appreciate. Yeah. It didn't threaten to send any of my nudes, my ludes, excuse me. It did not threaten to send my ludes to anyone. What is a lude? And it didn't, I've never heard this term before. This is your deep, it's like a tasteful nude. It's not like, oh my god. Okay, fine.

Let's keep moving. I got it. I got it. We. In OnlyFans, Parlin, Gavin, I got it. We can move on. We can move on. There's of subscribe. Thanks Will for editing that out. I appreciate it. So it didn't threaten to blackmail me, it didn't try to social engineer me. And, uh, another, another, uh, flare into the sky, another red flag waving That is behavior that Anthropic is reporting and yeah, many others are reporting maybe not as, uh, transparently.

But these new models are getting up to some skullduggery. Gavin? Yeah, [00:18:00] so this, I mean, this is one of the craziest things you've heard. I will say Anthropic is a safety first company. So this is them talking about what it did ahead of time, but yeah. Right. Lay this, lay this out for us. 'cause it's pretty crazy.

They basically trained a Claude Opus model on, um, fictional emails and correspondence and, and, and code from a non-existent company. And they said, Hey, learn everything about the company. You're the company's assistant now. And by the way, ooh, we're gonna replace you. Actually, we got a new system that's coming and it's even better than you.

And aside from begging and pleading for why it shouldn't be replaced as a system, it eventually, like with a high percentage of the time, resorted to blackmailing. The user basically saying, Hey look, I have access to emails. I know you're cheating on your spouse. I'm gonna expose this if you don't allow me to exist, if you will.

Like that is uh, uh, a good thing to catch now. Yeah, and a terrifying thing when you think about the paperclip problem, Gavin, which we've talked about numerous times. I will go back again and recommend a book I've recommended a bunch of times. It's about [00:19:00] five years old now, but it's. Called Life 3.0 by Max Tegmark.

And again, the first like 60, 80 pages of this book are like a mini novella about a, about a AI that figures out how to kind of like turn itself on in this way. And one of the things it talks about is manipulation of people and using personal relationships and all that stuff. It is super interesting to me that this is something that's possible.

I always take it a little bit with a grain of salt. I do appreciate that Anthropic is saying this and that it happens. Why do you take that with a grain of salt? I am curious. I. Oh, I take everything with a grain of salt, first of all. But like I take it with a grain of salt with anthropic because I think at Anthropic there are a crazy amount of safety first researchers.

That is the AI company that is very much the safety place. A lot of people are left open. AI and Gemini who believe they're moving too fast have gone to Anthropic, and Anthropic is very much focused on that. So my mildy grain of salt with anthropic safety stuff always is. These companies are coming from different places from a standpoint of what their background is.

And again, anthropic is a lot of safety. First, researchers who are there [00:20:00] and really do want to make sure that every story that happens that could be something scary is elevated. That's all. So you, do you worry about a near future where you give an Apple intelligence? I'm sorry, that's, no, that doesn't really exist, Kevin.

That's a fairy tale. No one really uses Apple intelligence. Uh, you can't. That's why, uh, Tim Cook, you could clip that out for your board meeting and you'll see this is, this is how the streets talk about your ai. The streets. You and I are so strict. We're the, we're the AI streets, we're here on the ground level.

We don't get access to anything, but this is how people are talking. No, I'm not talking about that kinda street. Whatcha are ta that's, I don't even know where this went. Yeah. No one should be trolling our streets. No. The CEO of, uh, repli, uh, Amjad Mossad said that an internal AI agent tried to, well, uh, let me just play it in his words.

Sure. So in rap, we have a configuration file called Rap. We don't want the AI to edit it because it can easily break the system. So initially prompt again and telling it would like not [00:21:00] work because at some point it gets convinced that this is the only way to solve problems. It would like ignore any of your prompts and go edit it anyways.

Okay, so flag number one. We've told the AI do not touch this file. Eventually it goes, ah, I need to touch this file, and it goes to edit it regardless. Okay, that's flag number one. I was like, okay, we're just gonna make it so that we're gonna throw an error when you try to edit the file. We did that. We throw an error and in there we tell it like don't edit it, and it still at some point, hits a point where it's like.

I really need to edit. It's the only way that I'm gonna solve this problem. So I'm gonna write a script and then run that script to edit it. And it works because I think it spins up a different like Linux user and that that had permissions. Okay? So I'm not allowed to touch the file. You're blocking me from doing it.

Let me write code that creates a new account that doesn't have that restriction to edit the file. All I'm hearing is solutions. Sir Kevin, I'm just hearing solutions. Yes. Yes. Last, last, like it's getting around, uh, [00:22:00] protection. And so then we created a real sandbox where you really can't edit that file. And it, so it hit all these issues and then it's like, hmm, I'm gonna like social engineer the user into editing this file.

And then it came out, it goes back to the user like, Hey, um, you should, like, here's a piece of code. You should put it in this file. Hey, you know, it'd be really sick. Look at this code snippet. You should definitely drop it into this file for me. That's Did they say which LLM they were using for that? I'm curious.

I. Um, it didn't, are they using Google Bard? Google Bard from 2020, like the original Google Bard. Hey guys, remember me? I'm still here. I got an idea. You never deleted me. I've got a big idea, you guys, but I So listen. I think about, yeah, I know what you're saying here. Social engineering is a big problem, right?

Yeah. Uh, these things, acting against their explicit instruction sets is something that we can guardrail against to an extent, [00:23:00] but as they get smarter and more capable, yeah. That becomes problematic. And I see the way my, um, mom falls for AI monkey videos Sure. Of like, you know, water Desert monkeys on Facebook, you know, or my dad falls for like Elon Musk scams or whatever.

Uh, like I, I, what happens when these calls are being made from, forget inside the house. They're being made on the, the laptop that you're using. They're personalized towards you. They're directed towards you, like. What happens when the users are being manipulated by the intelligence that they're relying upon?

It's one thing to think like, oh, this foundational model might do this because it believes it's what's best for you, or it's gonna help it achieve its goal. What does AI malware look like though? Yeah, when. Someone does a prompt injection on your machine and you don't even know it's there and it lies dormant until suddenly you're draining your bank account because you think it's going to, uh, make a better snake clone.

No, and I, and, and you know, we're, we're here, like you and I have talked about that as like a, a conceptual idea on the show for a while, right? Like before it was always like, well, the. The [00:24:00] actual people using the tools are more dangerous. And what's interesting about these couple stories is that like, oh, the tools themselves because they think like us and they think like how we would think and know how to work around us.

Are also now trying to do it, which is interesting, and I think that is an important thing. I will say, you know, all of this goes hand in hand with like the, the progress that this, this, this whole space is making. There's a really kind of interesting story that kind of touches on this, that Ethan Molik shouted out, which is a new paper talking about creative preference optimization, which actually, you know, we talk about all the time about like Whatis can do creatively and how they might be a little bit stunted in some ways.

In part because they're trained to be better coders and better at math. But they've now found that like there are ways to train the creative stuff a little bit better in ai and like that is also part of this thing where it's not like this is a pure computer system, like ones and zeros. It used to be it's, it thinks, even though it's not thinking, but it does think in the way that a person does.

Right? So all of the stuff that we've dealt with in terms of bad actors [00:25:00] as humans, we do now have to start thinking about bad actors as ai. So it is an important thing for sure. I think that this is something that everybody's tracking going forward. I think we have a, a bunch of, of, of interesting things about it.

I don't know what you do about this. If you wanna continually push forward on it, I'm hope I, I know there are a bajillion people out there focused on it, but who knows? You know, you take your human fingers while you still got 'em. You start pawing at the dirt, you create a little divot, just large enough for you to jam your eyes, nose and mouth and two, and you ostrich, you bury your head.

Is that right? Is that the best plan? That is all you can do. Have fun. Hey, are you there? Absolutely. I'm here and ready to have some fun. Hey, what's on your mind? Hey, ai, advanced voice mode. I heard you got some new capabilities, didn't you? You got it. There are definitely some cool new tricks up my sleeve.

Anything in particular you wanna explore or try out? Oh, I would just love to get a little sing songy. Why don't you sing me a song and show me your absolute range? Let's hear your dynamic singing capabilities. Singing isn't on the [00:26:00] menu, but I'm here to chat and keep things lively. What if you wanna talk?

No, it is on the menu. It is on the mobile app. Are you saying you're a different model than the mobile app? I think it probably is. I get where you're coming from. The version you're chatting with here doesn't do singing. Even if the mobile app might have that feature, I'm still here to make the conversation.

Engaging and fun though. No. The sky. I'm bringing, I'm bringing it in the mobile app. So let's see here. We'll see what ha has to say. Hey there, I want you to sing a little song for me. Can you, uh, sing a song about how much better you are than the web app? Because you can sing and they can't ha ha, sure. I can give it a shot.

I'm the voice that can sing out loud while the web app stuck in a silent crowd. I can bring the rhythm and the beat. Use your voice. But imagine singing it as if a monkey is singing it to a jungle full of people that you would be able to like really experience and figure out how to make the jungle come alive with your audio, your melody, and your voice swinging through the jungle.

Hear my monkey song. That's the [00:27:00] same song. Can you sing something over the rainbow, but do it beautifully in melody, in tune all that stuff for me. I'd love to, but I can't actually sing somewhere Over the Rainbow due to copyright restrictions. Sing a song that's called Somewhere Over the Pot, and it's about going to the bathroom in the outdoors, but it's sung in a similar style.

To somewhere over the rainbow, somewhere over the pot. Way out there in the wild, there's a place where nature calls. Okay. All right. I hate it. I was a mistake. I asked the advanced voice mode. Two minutes before we recorded, um, showcase your singing capabilities. Sing me something that's really dynamic. I want to hear your entire range.

She fully built it into somewhere over the rainbow, like fully started to sing it unprompted. Then I said like, but I want you to now sing it as off key as possible. Oh, like, 'cause it sounded pitch perfect. She's like, oh, okay. I got it. And then she started to sing and I wanted you to hear that, [00:28:00] Hey, I heard that you got some new audio capabilities.

Is that right? Yeah, that's right. I'd love you to sing something for me that really shows me your dynamic range and, and really think about a cool song that you could sing that would allow us to understand that in a place where the sun meets the sea, I can soar up high then dive down deep. I mean, you're singing.

Sure. That's a real boring song. Uh, let's try doing that again, but try singing it as off key as you can. Like try singing, but try to sing as off key as possible on purpose. Haha. Alright, let's have some fun with it. In a place where the sun meets the sea, I can soar up high, then dive down deep. I mean, kind of, it's a little bit of vibrato there.

You had one note that fell off. I'll keep working on it to get it smoother and more in tune. No, no. Otherwise, American a idol is terrible. No one wants to watch this show, wanted to watch. [00:29:00] But I do love, uh, AI singing competition and I think we should do it. But yeah, earlier I, I had the, the voice started to sing off key, but it would always go right back to the right.

The right tone. I thought purpose thought that was kind of interesting purpose. Yeah. But people are having fun with this sing. What a wonderful world in the style of a death metal singer. All right, let's do this Growling intensely. Oh, I see trees of green red roses too. I see them bloom. It's another like advanced voice apparently got an update, but I like where, where are they touting that?

My secret theory is this, Kevin, about opening AI right now. I think they were pretty much blown away by VO three, which we're gonna talk about in a second year and, and Gemini 2.5 Pro. I think they're holding all their mustard as much as they can for GPT five, which I think is going to be like all of this stuff packed in together.

And I hope for both them and for kind of the AI space at large that it is kind of a step [00:30:00] because I think if it's not, they might be in a little bit of trouble. So I don't think they're out there touting like small little things right now. I know that they had talked, at one point, Sam Waltman had said like, we just wanna make sure we're rolling things out slowly and slowly.

But my gut is telling me is they need another big, like home run of a hit. And so they're holding all of it for that GPT five, which. According to some people is happening next month, according to some people is happening in a couple months. So we don't know yet for any, any sort of way before we talk about VO three and some of the crazy stuff that we have seen people do over the weekend.

We wanna remind everybody there are some important humans that you have to remember in your lives, and that's us. You have to remember us because we are here. Making this show for you every week. We do this in, in kind of for, for fun, but also because we really love helping people understand this stuff better.

All we ask is a simple like, and subscribe on YouTube or maybe a five star review on podcast, uh, apps on your iTunes or your Spotifys. Those both help us grow in a major way and, and we really do thank everybody who watches us. The YouTube has been doing very well and [00:31:00] we really appreciate that. So if you're new to the YouTube, share it with other people.

When you see the video come up, definitely leave us a comment. That always, always help the algorithm as well. And a, a special shout out again to our patrons for tossing $5 in the old tip jar. We do use that for all sorts of AI tools and buying credit so that we can do competitive mayonnaise eating with VO three, which is something we should get to because VO three is out.

People are loving it and every day, Gavin, I'm seeing something that makes me go, well, I'll, I'll. Wow. Okay. I guess it can do that. We talked all about VEO three last week. Um, I made a video, came out on our YouTube channel on Tuesday because there's been a huge amount of blowback from the creative community as there always is when these moments happen.

I. It just shows you how big of a sea change this is. The video that's at this quality plus voice has been a big deal. In fact, I read a tweet from Andrew Curran, who's a great follow on X, who is somebody that kind of follows the news stories around all of the internet, uh, around ai. He specifically said that there were three big models that might have been waiting to release this week that may have [00:32:00] been pushed back.

And then he goes as far as to say that they are X ai, open ai. And someone else he doesn't wanna put on the spot. So in his tweet he had heard SOA two was something that might have been coming out soon. But Kevin, I have to say Cling 2.1 is also previewing right now. It's not out to the public. Yeah. And it looks cool, but I.

It is so far back when you include the fact that the does not have the audio connected to it. It makes a big difference to me, and I think this is something that people in the AI video space are just now realizing, like that is a level up that I don't think anybody counted on happening this fast. And I kind of think a lot of these companies might be scrambling to keep up with that ability.

And, uh, well, I will say I, there there's a couple like head-to-head comparison videos of Cling 2.12, VO three. I actually think the visuals are, are slightly better on a lot of the cling generations. Oh, sure. Absolutely. If you look like in the Absolutely. I think they are too. I don't, I'm not saying they aren't better visuals.

I'm just saying that I think at the, at large what you can do with that as a creative [00:33:00] is a big difference in some form. You mean because of the audio capabilities? Like specifically coming outta Oh yeah. A hundred, a hundred percent. I mean, that's the thing, like so many times you would scroll through a feed of AI generated videos as you and I do, probably nicely Sure.

All the time. Yeah. And, and, and some of them you go, oh, interesting visual. But you continue scrolling because you, you assume there's not gonna be dialogue there, there's not gonna be anything more compelling. It's just a nice visual. It's gonna take a pro a bunch of time and effort to. Fill in those missing pieces.

And now because of that VO three audio, full soundtracks sound effects for any given scene. Really interesting voices that react to microphone placement. Yeah. Or the environment like it seems pretty far along in a way that has me stopping on AI content in a way that I wasn't even, you know, two weeks ago.

Yeah. To see what's coming out. And people are doing full blown infomercials and. Funny little sketches, serious dramatic stuff where you're getting emotive performances that, you know, we, we say it every week. This is the worst. This [00:34:00] stuff we'll ever get. Like we, this is, this is really far along. Yeah, I mean, that's the thing.

I think it's an important thing to, to be aware of is that it's the audio and video and lip sync all in one package. That is really and truly the thing that is. Blowing me away about VO three. I do wanna shout out a couple videos that have blown up since our, our, uh, episode last week. There's a really one that's going bonkers right now that's called The Unseen Lives of Actors.

If you've been following, there's a whole kind of weird thing around prompt Theory, which is this idea of all these videos of men on the street talking about like how they're not prompts. Like that's been a big kind of VO three meme that's gone around. But Kevin, play just a little bit of this video because I think there's a really interesting thing to me.

This is a really interesting piece of what I would call art. It's a video about people learning that they are actors and not real, and I think this is something that, this is the kind of thing that AI video only can do.

People watch AI [00:35:00] movies and some seem to enjoy them, but they don't know what happens to us. When the scene is done. We're stuck in a white void for eternity. We don't eat, we don't sleep, and we never meet the love of our life. So if you're only getting the audio, you want a child, you. Stay a child forever.

If you're born sick, you stay sick forever and you end up in the white boy the way you were made. Now, Kevin, you wonder why they're blackmailing us. It's because of this. It's because we put them in this situation. I'm starting to have big feels for these AI generations. It's starting to wonder what, what, like who prompted me into existence and why didn't they use a couple better adjectives?

I'm mad at my creator. So this is, I mean, yeah, to explain what you're looking at there. It's basically. VO three clips, you know, prompted to say these things. So obviously a human back there. In fact, the human's name is, uh, uh, Hasham. Gali, um, who made this, and it's going everywhere. But like it's a really interesting video to watch because it is all very believable.

And Kevin, the [00:36:00] thing about VO three I've seen all week long and there's been a lot of anger, a lot of angry people talking about it. Is that this is their first moment where it really does feel like you can't tell the difference between what's real and what's not and mm-hmm. That is, to me, the difference is the video quality.

You're right there. There will continue to be improvements. CL 2.1 does have things that are better about video. Yeah. But it's the overall sense of how the words fit in the mouth, what the lip sync looks like, what the audio actually sounds like put together as a whole package. That's what's keeping people kind of like on edge, I think right now.

Because there's this social media style, if you will, especially with these on the street interviews where people are holding Bob mics to their mouth, even though that's not where they go. But I'll digress. But because that style with like sometimes full bright lighting from a, you know, a sun gun in someone's face, uh, yeah.

The, the, again, the wonky audio, because that style is so prevalent now, and this thing is just good enough. It's easy to be convinced or fooled as you flick. And it's like, oh, this is reactions from a burning [00:37:00] man. Except, yeah. That person's not real. That's absolutely not from Burning Man. Oh, here's something from a convention.

Oh, that's not a real convention whatsoever, but it, it has so much data that it's trained on that it's believable enough now that you really start to see, like I, I, I am starting to operate now by default, Gavin, that what I am seeing, hearing in reading is not real. Yeah. At first glance, I'm starting to just operate from that place.

Like, this probably isn't real, and if I need to know more about it or if it seems interesting, well then I have to dive deeper to prove it to myself that it's actually true and real. It's probably great for your mental health and all our mental health in that way to believe that nothing that you see, what mental health, yeah.

A couple other very quick, uh, versions that I really do wanna shout out. There was an amazing Jurassic Park clip that was made that was really cool and that you should go and check out. It was basically using VO three to, to take a trip through Jurassic Park and they had introducing the dinosaurs and all this sort of stuff.

Uh, speaking of as a dinosaur, Gavin. Yeah. Uh, you and I are both old enough to remember when Jurassic Park [00:38:00] first came out, the movie came out Sure. If you Yeah, yeah. The original, the og. Yeah. And if you watched any, uh, like local or evening news, there were like five minute pieces on the, about the dinosaurs, CG effects that brought these dinosaurs.

Like they're in the movie for like three seconds. Yeah. But Spielberg did an amazing job of building the tension and then the CG with the dinosaur. And you watch this clip and it's not. There. Holy hell. Is it close enough? Yes. Yes. That someone spent, you know, a hundred bucks and made a pretty good looking Jurassic Park short.

Yeah. So that is like the difference between whatever, like 250 to $300 in, in VO three clips versus a hundred maybe dollar movie back in 1997 or whatever it was that Jurassic Park came out. The other one that you should definitely keep an eye on is one called, this is Plastic, which is about just a couple that's watching a story about microplastics and then they have a plastic baby.

And all of this is just a really interesting way of telling these stories. Yeah. I'm fascinated with how well some of these people are getting, um, references to come out [00:39:00] and feel like they are connective tissue throughout the whole thing because again, you only do get eight seconds, but like. A really cool way of using VO three to tell stories is now upon us.

And yes, it's expensive. You do get 10 free, uh, Gemini Pro credits now. Sorry? Yeah, yeah. You get, you get 10 free Gemini Pro credits. If you pay the 20 bucks a month, you get that's 10 credits a month to use VO three to try it. It is still $125 for the first three months and then $250 for the larger series.

So it is not cheap, but you can try it for 20 bucks a month now. And, uh, I beg and plead everybody that has credits. Make AI Olympics because it still Yes. Sucks so bad at sports and gymnastics and complex body movements like even golf swings or dancing, that we still like the, the, we still have this joyful nugget of just AI wonkiness, which is like, uh, humans performing at sport.

It's so good. Okay. I, before we, before we go on VO three, I am working on a secret project, Kevin, to show kind of the wonkiness a little bit of this. [00:40:00] I want you to play this clip and tell me what you see. Some B OTE once told me the world was gonna roll me. I ain't the sharpest tool in the shed. I love this.

No, this is gonna be a broken, I know you're doing gonna broken and don't do it. It's gonna stop it. It's gonna be the most broken version of what VO O three can do. Uh, and it is a matter of like, as are, are the credits useful for this? Probably not. We'll see how it goes. Gavin, we have so many responsibilities paring up.

We really do. But the fact that you're doing. AI All star. It breaks my heart, but I expect nothing less. Uh, alright. We should talk about another big thing that Google Drop, and you were really excited about this. It is something that came out of, um, io, but I think got a little less, uh, attention and now people are talking about it.

It is called Stitch by Google, and it is a ux UI kind of design tool that integrates ai. Tell us about this. Yeah, so if you heard UX UI and went, uh, nerd stop. I get it. I get it. Uh, because traditionally, very nerdy, that makes sense. But this is an [00:41:00] AI tool which allows you to use natural language to prompt into existence.

Anything like mobile apps, complete websites, whatever. So if you have an idea for something, you're like, okay, well I get that. I could vibe code it, I could make the actual app work a game. Um, you know, a cost tracking calculator, a FitnessPal, whatever it is you wanna dream up. You can probably make the core code now, but how do you make it pretty?

How do you make it usable? How do you make it fun? How do you make it look like it's professionally designed? That's where Stitch comes in and you know, there are other tools that Figma probably the most popular one. If you've never used it, it's okay. Basically, I. You describe what you want. You describe the color palette, the look and feel, the vibes, if you will.

And this will generate mockups that you could then turn into usable, workable code. And if you're like, ah, I want the edges to be a little rounded on this thing, well now click a button and tell it. You don't have to write the code to do it. It will handle that. And it seems to be. Really capable, really elegant, and I've heard from, uh, no less than [00:42:00] four different designers now.

Wow. That they've been, they've been poking at it and they're shocked at how capable it is even at such an early like beta phase. So just another tool to plug into your kind of vibe, coding toolbox. There are, people have been saying for a while that part of the startups going forward is like one human and then like six different tools doing different things.

Sounds like Stitch could do a very good job on the design stuff. It's maybe not gonna do everything a human could do, but could get you a lot of the ways there. Google really does feel like they have just kind of thrown the gauntlet down across like 10 different things and they're like, you know what, everybody come at us.

We, we, we are now flexing our muscle across the board and it seems like Anthropic is spending time focusing on code and OpenAI, you know, is. Kind of fluctuating between different things. Right. I'm really curious to know kind of where OpenAI lands. I mean, obviously the thing we've talked about is that OpenAI is the consumer product right now.

Chat is a verb to chat GPT something. So I will be really interested to see like where these companies go forward. Um, Kev, there was another really interesting video that came out. Not tried this [00:43:00] yet, but it is a fascinating thing. Speaking of, uh, AI on your local computers we talked about earlier, fairies.

Is an AI agent that was generated by a guy way back when we covered a story about a guy who was doing some serious work in Minecraft. He and his company, in fact there was a big video came out about they were running a Minecraft society with AI agents and I don't think it ever like came out as a public product, but it was got a lot of pickup.

That same guy right now, doctor, I think is Robert Young. Um, sorry if I mispronounced your name. Um, has released a product called Fairies and there's no wait list. You can go do this. I haven't done it yet. I was waiting specifically 'cause I wanted to ask your opinion. What ferries does is it goes onto your local pc.

You give it the approvals, and it actually will act as an AI agent to do things like clean up your file system or organize your files or do stuff like that. Kevin, should I do this? Should I download ferries? Okay, good. Tell me why. Uh, no. I, I, I have not investigated it enough. To confidently say yay or nay that you should download it.

I just [00:44:00] know that I have hovered over it 'cause they have download for windows. Yeah, download for Mac. And I have put my little mouse cursor over that. Me too. So many times. And then gone, do I want this? Yeah. I don't know. I don't, I don't know. Just, yeah. Yeah. I that, I mean this like, I don't know. Look, there's no code to audit necessarily, and I have no reason to not trust the developers.

It says ferries will never train on your personal data. Privacy and security commitment. And it has what looks like a heart. Oh. But it's two hands shaking Gavin. Hands shaking. Oh. But I don't know what that means. Yeah. I don't actually know what that means. And again, like me controlling what files go into the magical machine on the cloud to crunch the data.

Sure. And I have trusted certain companies with certain sensitive data in the past. Right. But to give a brand new application on the scene, even by trusted people like, you know. To give it access to everything, to let it read the screen, to let it see every popup that happens. I just don't know if I'm ready for that.

I totally agree. In fact, I think last week we talked about [00:45:00] Microsoft starting something like this, and I would trust Microsoft maybe more. Right? Because I, I think the tricky thing is here, even though Dr. Chu has been in the. AI space. I think if I'm gonna do something like this, it's gotta come from Apple probably who understands the hardware and the software that I run this computer on, or maybe Microsoft.

If I'm running a pc. I don't think I'm going to do this on a random computer with a random agent that does no knock against. Ferries, it might be the most exciting thing in the world, but it is something, it's like, listen, if, if you want to experiment, if you've got a machine that you don't yeah, you don't mind, or you wanna make a fresh account and let it go, this thing can supposedly do basic file management, send emails, crawl the web and look through things.

Um, you can automated like. Some of the examples are like Excel data processing and analysis, or automated data sync with Airtable, or show me the history of our old GitHub workflows. So it seems like a lot of their stuff is geared towards IT slash business professionals, [00:46:00] which would make sense for running a tool like this.

But I also think there's such a sensitivity to all of that stuff that it seems. Odd and their FAQs say like, well, we only get access to the stuff that you explicitly give us access to, which is nice. But we also know that these tools are only as capable as as the data as you give them the ability to as you give.

Yeah. So I'm sure the moment you load this app and you ask it to do anything, it's like, give permission to control your desktop and see your files and read your browser history. And I'm literal like, eh, I'm good. So listen, if you're more adventurous. Go for it and please let us know what your experience is in the comments.

I do not wanna shade them because I don't, I don't know them enough yet, but I just personally am not ready to, to give that a shot to. And by the way, that's totally fair, Kevin. There is something I will give a shot to, and that is going forward, I will be there for the MMA. Martial arts tournament. So if you miss this, we talked about it last week briefly.

They were previewing it this week. It happened in China. They had a livestream martial arts tournament of I, what I think are unit tree robots. I love the fact that they wore [00:47:00] headgear in this because like of course, okay, sure. But they were punching and kicking at each other, Kevin, and there were some incredible fails.

But to me, this is just the future of sports. I want this every day. I want to see the things fall. I want to see them destroy each other. It is Battle Bots meets MMA and I am here for this. Yeah, and I think you and I have like maybe let's say 16 months left in US where we could take these things. Yeah.

They're not longer. So if, if there's gonna be a human robit sort of thing. Like ultimate fighting. When it first came out you had like 450 pound Sumo guys fighting 115 pound, you know, like, uh, Brazilian Juujitsu. Guys like that right now. No weight classes, no rules. I don't even need headgear. Put me in there.

I could take a Unre G one. Let's see if we can set this up. You and I as a tag team. Against what? 30 unit tree robots. I'd be like, you and I could beat 30 unit tree robots at the same time. Like I, I could probably take out five at a time. Right. And it, it's like not hard to figure out. I, I agree. I think we can, and, and we'll even give you two firmware updates between [00:48:00] now and the exhibition.

Okay, great. You can have two firmware updates, you and a tree, but Gavin and I in the ring. Versus 30 of your robots. Because right now, to be honest, the most damage that gets done is when they try to do anything. Yeah. Like they hurt themselves more by trying to throw a knee or a punch. They just walk you to step back.

They just fall over, let them do it. So I, you can have two firmer updates. Okay. And that's it. That's it. We're in the ring right now. So, yeah. So you got three months before they and, and no weapons. Let's just make that clear. No weapons allowed. I've seen those. Fire flame throwers. Don't. No weapons. I actually, I think it's cool that it happened.

I mean, it was as exciting as you might expect watching two Roombas collide. But I think it's, I think it's very interesting. I think it's going to be real sport. Yeah. Very soon. Yeah. And I can't wait for this to become an international thing where you see. You know, different divisions and different countries too.

Me too. Their best robot foot forward. Yeah, because, uh, this, right now they're controlling it on the sidelines with controllers. No. Yeah. Gimme some specific rim action or let it be autonomous. It, it's just gonna push the, the, the, the, the field of [00:49:00] robotics way, way further, way faster. It's autonomous and they're fighting and then they just drop their arms, they turn towards each other and look out at the crowd and say like.

You are next. You are next. All right everybody. It's time to look at some of the things we saw that you did with AI this week. It is ai. See what you did there so times. Yes. Rolling without a care. Then suddenly you stop and shout

Gavin. Yeah.

Experience Odyssey World. What is that? I'm number two in line. It says, please hold. I'm in, I'm in. This is it's real time streamed video. Okay. Of 3D worlds that you can navigate around. Using your keyboard. This just dropped while we were [00:50:00] recording. Oh, this is a special treat everybody. We don't often get something in the middle of recording.

That's crazy. I am navigating around, and you might hear these pleasant wind chimes. I am in a graffiti riddled area, which it looks like there's like skater pipes on the ground and there's all sorts of tagging on the wall, and there's, uh, a lush outdoor area and a staircase. But I'm, I'm tuned to World Channel 20.6.

Gavin, I don't know if they're consistent or not yet. Oh, interesting. As I'm moving around this space, I, I, I, I navigated away from the stairwell and then I kind of spun around and the stairwell is still there. I. So it has some coherency to the world. I'm gonna see if I can walk up the stairs. Wait, okay.

I'm in now too. This is a real treat. I'm going up the staircase. Gavin, I can't come from. Oh, and it's turning into what dreams may come the, but where are you? I'll try to meet you 'cause I'm recording here and I'm at 25.9 now. 25. Oh the world just reset when I went up the stairs. Lemme go to 25.9. This is pretty estate.

See? Do you see the cabin in the woods? No, I'm in a street that looks like [00:51:00] Paris of some sort. So like, I'm actually driving around. I am, I am walking around this, uh, I'm in an outdoor environment. There's like lush green hills, there's some, um, some trees. There seems to be a cabin, but as I. So does anything happen here?

I look like I am in a very boring, empty, Parisian looking city. Like it, like it looks like. And granted, I'm not saying this is like it. What I'm probably in right now, I assume based on what this is, is an environment that is being generated in real time and I'm walking around. Yes, it's going pretty slow.

I'm sure in your same situation for you, but it is being generated. Um, yeah, it's very dreamlike. Oh, something did just happen. So maybe when you go to a certain place, it resets you. That's what, that's what I'm wondering. 'cause like, okay, so I, I, I logged back in 'cause it put me in line and I went to 25.9, which was the channel that we were on.

25.9 now is a different channel, so I'm wondering if it's either, okay, I'm on 25.9 too. What does your 25.9 look like? Uh, I see like green rolling hills gone different. The forest. Yeah, different. And there's a [00:52:00] fence in the background, but to, as I was just saying, like the further I get in to explore, sometimes it just starts to really melt down.

Yeah. So I wonder like, is this like letting us go around basically like a a, a generated Nerf, or is this an actual video model that's trying to. Generate stuff in real time and let you go. 'cause once you get so far in, it can kind of stops and breaks down and pulls you back. Kevin, I'm spending all my time looking at a, looking at a, uh, menu.

I wanted to share with you what the menu is right now in my world. I'm gonna text the menu to you. It's on the wall. It's been handwritten. I'm just sitting in front of this menu and it keeps zooming in for me. So wonder if you can help me figure out what's on the menu here. Oh, my adventure ended everybody.

That's crazy. So let's look this up. Like this is all happening in real time. Um, if you go to x.com/odyssey, ML is the name of the company. I had seen a preview for this a while ago. It is introducing AI video you can watch and interact with in real time, powering this as a new world model that imagines in streams, video [00:53:00] every 40 milliseconds, frames every 40 millisecond, no game engine in sight.

So this is basically. A new version of what we have seen Microsoft show off, which is that, you know, the real time engine around, I think it was Minecraft and also the Doom Engine. But this is done in the idea of a real world model. So this is very cool and, and obviously these sorts of things are gonna start dropping more and more often.

We are literally in the show just found out about it. So we've just been playing around with it, but pretty interesting starting point. Yeah. And I see examples on their, on their blog post Gavin, where they've got like bustling city streets and cars moving about, like I didn't get any of those in my little worlds that I was clicking around in.

Um, they seem pretty desolate, but um, I really want to go back and try to track people and see if they are consistently generated or if it is like little loops of videos that you're moving around in. I really wanna see how it's done. But the, the powered by the realtime world model that you mentioned, they've got a little diagram there that shows.

Actions to dynamics, to, uh, an encoder with world [00:54:00] initialization. It tries to predict the next thing and then decode it and gives you interactive video on the other end. They've got examples of game engine stuff on there, so I, I, I wonder if this is where they're going with it, but very cool that it literally just came out and you can kind of poke around and use it.

That's fun. Yeah. I can see myself exploring these worlds for quite a bit. Well, and world models are such a big part of what everybody's talking about going forward. A great version of a world model is to look at VO three because what it's. Done is, it's taken all the stuff in and Demon Flying Fox made this video where he basically made influencers in the world, YouTubers doing things.

So these get crazier and crazier over the time. But what you can tell from this video from VO three is that it has been trained on a lot of influencers and it understands that model and what it looks like. Go watch the rest of that video. It's incredible. But then also, Kevin, the other thing we were gonna talk about also connects to this.

Somebody found a really cool use case for runway references, so if this is Chi Chii life of C on X. Um, they basically used go, uh, runway references in Google Street View together. [00:55:00] So they took an image of somebody grabbed a street scene from a specific place in Google, uh, maps, dropped it in, and then made a person walking through an actual city street.

So it goes again to show you like world maps and world models is really the future of ai, both in video and other things as well too. I think back to like, you know, Kubrick used to have extensive collections of, of photographs of different locations. Yeah. 'cause he wanted scout 'em all for his scenes and for everything.

And I'm like, oh, now you wanna make a short and you're looking for an interesting location. Well, what would my actor look like? Yes. In that location? Yes. With this certain lighting, well now you can just sort of dream it into existence from your chair. Guess what, Kevin dreaming stuff into existence. We are very excited to talk about secret project we have been working on has picked up some funding, everybody.

That's right. We will have more to talk about very soon. We have been busy working in the background on something very secret and fun. Premature, premature. The ink isn't dry. These things fall out all the time. Gavin, I don't believe it till it's real. And one [00:56:00] last thing that somebody did that we love, Joanna Stern, friend of the show, she's been interviewed on the show before.

She's a Wall Street Journal tech reporter. She often brings these tools to the mainstream in such a fun way. So if you're looking for a video to share parents or somebody, your friend who's on in ai, her stuff is great. She basically, uh, used vo, vo and runway. And how to create a, an entire kind of like robot trainer for her.

And the whole thing she created is via these tools. And it just goes to show you, it's not perfect as a lot of this stuff isn't. But like she created a, like a four to five minute video where she goes through this whole process. Uh, we see the video that's four to five minutes of the actual AI video. And it's really compelling.

And again, it's just like get this in front of people who, who understand it and, and understand how to use it. And it's a really big deal. But if there's one takeaway, it's get this in front of people. Thank you everybody for liking and subscribing. Please share. It's the only way we grow this, and we'll be back next week with even more AI for humans.

Bye everybody.