AIForHumans95OpenAIDeepResearchDeepseek

Gavin Purcell: [00:00:00] Sam Altman in Open AI strike back against Deep Seek with the very powerful new research tool, deep research that is powered by the O three model. And it's not just that, it's O three mini and a brand new font. How deep does this go? Can deep research out deep, deep seek? It goes very deep. And that is Tim Simmons from the Amazing a i YouTube channel, uh, theoretically Media.

Gavin Purcell: He is joining us this week because. Kevin is on vacation with his family and it is dad's 70th birthday. Ooh, I want to go on vacation. This is not the time, Tim. And we'll be talking about Repl. it's new agentic AI app. Uh, plus a crazy new lip sync deep fake tool from ByteDance that is called OmniHuman. And we have so much more.

Gavin Purcell: NVIDIA is training robots on moves from some of the most famous sports stars in history, from LeBron James to Kobe Bryant and Ronaldo. And we've got Gandalf in a cage match. Why not? It's AI for humans, everybody.[00:01:00]

Gavin Purcell: We are so thrilled to be joined by Tim Simmons from Theoretically Media. This is one of our favorite AI channels on YouTube. And Tim does a lot of amazing stuff with creative AI. So Tim, thanks for being here. Hey, thanks for having me, uh, big shoes to fill, uh, with, uh, Kevin off on vacation, I guess. That's right.

Gavin Purcell: Well, that's what he says at least, but we have such big news, Tim, and I think we have to hit the biggest news of the week. OpenAI has a new font. Have you seen the new font? I mean, I have seen the new font and, uh, I mean, stunning, breathtaking, clearly, uh, taking the lead in China, uh, when it comes to fonts.

Gavin Purcell: Yes, clearly. So this is not the big news. The big news, as we all know, is that Sam Alban has kind of. Pushed back against the deep sea invaders. Let's call them or the, the, the crazy Chinese model that kind of took over the internet and crashed the stock market. Sam Altman has basically said, I'm going to start shipping stuff at an insane rate.

Gavin Purcell: And, and really Tim, the big news that we got this week was deep research, which came out on [00:02:00] a Sunday night here for me and Monday morning for some people in Japan. But have you had a chance to play with this yet? I have actually did a full deep dive into creative use cases, uh, with deep research, uh, over on my channel.

Gavin Purcell: I just finished that up today. Yeah, it's, uh, it's pretty remarkable. Uh, have you, have you got a chance to play with it? I haven't played with it, but I've done a lot of deep research into deep research. So that way I can keep playing that way, but, but just to kind of set the table here for everybody. So what deep research is, is.

Gavin Purcell: Is a new agentic model that chat GPT that runs on chat GPT and actually uses the Oh three model as its backbone. In fact, it's the first thing officially that uses the Oh three model for the public, and you can only access it if you have a 200, uh, open AI pro subscription right now. So that's a pretty expensive way in the door.

Gavin Purcell: But I have seen so many people talk about how good it is. And I want to hear more about your experience with it in a bit, but Casey Newton had an amazing writeup in platformer and he's somebody who's obviously the host of hard fork, which we listen to all the time here, the podcast, [00:03:00] but has been relatively, you know, skeptical of AI tools and he had a good experience, but my, the most interesting one that kind of shocked me was.

Gavin Purcell: Hearing from Tyler Cohen, who, if you're not familiar with Tyler Cohen out there, Tyler is a well known economist. He's written some incredible books. He has done a lot of work, uh, at least trying these AI tools. And he specifically said that he had it right. A number of 10 page papers for him, each of them outstanding.

Gavin Purcell: I think the quality is comparable to having a good PhD level research assistant and sending that person away with a task for for a week or two, maybe more, except that deep research does the work in five minutes. So this is the promise, Tim, of what very advanced AI can do for us. I guess when I say us, I mean, you know, the, the, the kind of plebs of the world might not be doing deep research on like, you know, biological weapons or anything that way.

Gavin Purcell: But this is kind of the, the promise of next generation AI, right? That you could [00:04:00] send it away and it could come back with something that would be super powerful. You, you did something in your, in your video, which I thought was interesting is you, you kind of tried to find ways to make it useful to you.

Gavin Purcell: So maybe you kind of tell us a little bit about what your experience with it was and kind of the pros and minuses of it. So. I mean, I'll say previous to this, I should mention that, uh, this is not the first time that we have seen deep research, uh, Google Gemini, uh, the, uh, pro 1. 5 version, I believe, uh, had deep research as well.

Gavin Purcell: Um, so, you know, we've sort of, we've had access to this. Now there is that open AI 0. 3 spank on the whole thing, and, uh, it definitely does, you know, Um, so the way that I decided to approach it because I, I'm not a research scientist, nor am I, you know, an astrophysicist or anything. I'm just a guy with the YouTube channel that talks about AI video.

Gavin Purcell: Um, but you know, not to say that. I mean, I still have to do my own research on, uh, you know, what's happening. And, uh, I did find, uh, That in terms of sort of [00:05:00] real time news, it does not do extremely well. So, which is good for me. It's a good, it keeps me, uh, you know, um, it keeps my job a lot. Um, but, uh, you know, like if, if I ask it for, you know, what the latest, everything that happened today in the world of generative AI video.

Gavin Purcell: Um, it will come back with things like, uh, uh, runway has released gen three. And it's like, well, that was, that was a while ago, but effort. Um, so, uh, what I decided to do was, was kind of circle around and, uh, start thinking about it as like a real research tool, like if you were going to, if you wanted to craft, say an AI viral video along, uh, side, the, uh, say the.

Gavin Purcell: very famous Wes Anderson, Star Wars, uh, movies, trailers, I'm sorry, or, uh, what was the big one? Like the shy kids, uh, you know, airhead, um, you know, what do all of these videos that, uh, had, you know, over 10 million views, what, what do they all sort of have in common? So I went through and identified a [00:06:00] few of them and it found things like, you know, the commonalities being like, it was really.

Gavin Purcell: A was all like human edited. Everything had kind of a heart to it. Um, you know, there was pacing and story with it. Uh, and of course, you know, the big one was like between the shy kids. And, uh, there was one called the heist by, uh, David Zada, I believe. Yeah. Yeah. Jason Zada. Who I know. He's a great. Yeah. He's amazing.

Gavin Purcell: You know, I, but he was first to market with, uh, with the first, uh, so, you know, so there is those aspects to it, but I think it's like, uh, when you go through that research paper, your research paper on, you know, how to, how to make an AI viral video, it does have some really great ideas. I also prompted for things like, you know, what's the best way to prompt in a, uh, AI video generator and, uh, other questions like, is it better to subscribe to runway mini max, uh, 11 labs, mid journey, et cetera.

Gavin Purcell: Or to build your own machine. And it went through and did all of the pricing, um, on it. And basically it came back with like the first year you're actually ahead, uh, by subscribing to all of these services. But by the time you hit year five, [00:07:00] you're basically at like an 8, 000 difference. So no one talks about year five in this space, Tim, you're five.

Gavin Purcell: We're going to be living plugged into the matrix with like little tools coming out of our brains and all sorts of other things. But I mean, I, I think, listen, That is a really interesting perspective, right? Because I would put us squarely in the non academic use case, right? Because I think that there's a lot of people out there who have said, and I've seen, you know, and even in the video that they showed from Japan, they talked about this idea of being able to kind of create these kind of very dense academic or, or, or marketing driven or financial papers and all of that, like.

Gavin Purcell: It really will depend on the results. And I guess that's why Tyler and Casey are both interesting to me. The other side of this, I think was interesting in that video, which gets to the point of some of the stuff you were asking about. So, so in the video, they actually look for and ask for a deep research to go off and like research the best skis and come back with the best version of skis.

Gavin Purcell: And what I immediately thought about that with that was that that is a really good use case of agentic AI, right? Like, you know, there's a lot of people out [00:08:00] there who will go and spend all this time figuring out like what's the best computer monitor to buy. And for some people it might be fun and they're not going to want it.

Gavin Purcell: But like what you really want is the answer and know that something spent the time on it. But that is also Google's business model, right? Like Google's entire business model is based around the idea of you clicking on links to see what, what's there. So I think that's going to be an interesting shift overall.

Gavin Purcell: And the other thing I think that's really important to think about here and We're going to play a clip from Sam, uh, Sam Altman's in Japan, uh, meeting with, uh, Masa, the guy who runs SoftBank, but he particularly talked about the idea of how much work this can do when it compares to the kind of worldwide, uh, employment rankings, let's call it.

Gavin Purcell: So we'll play that clip here. This can do complex research tasks for you. Tasks that might take 30 minutes. They might take 30 days. It's powered by O3. It's the first time that the world, the outside world gets to use our O3 model. And it can browse the web, scan text, [00:09:00] images, PDFs, much more synthesize this reason through it and prepare a report for you.

Gavin Purcell: So it takes a while. It goes off and does all of this work. You can see what it's thinking about as it goes. Um, it's different than Chachi VT where you instantly get a response here. You start off a task like you might give a task to a sophisticated coworker and deep research goes off, thinks through it, gathers insights, gets it together, find sources and gets your report.

Gavin Purcell: This is a system. That I think can do, this is just an estimate of mine, but I think can do a single digit percentage of all economically valuable tasks in the world. So Tim single digit percentage of all the work in the world, which is not a small amount, right? But it sounds like it's small does a single digit.

Gavin Purcell: You know, single digits, nothing compared to single digits, plural. That's up to nine of all the numbers. Uh, no, that is a, it is definitely a, but Sam says bombastic stuff like that all [00:10:00] the time. I mean, eventually we do know that we're going to get there. Um, but I mean, he also, you know, occasionally you'll run across a tweet of just like, uh, uh, we've harnessed the power of the star of Orion and the AI God is amongst us now, uh, releasing sometime this April.

Gavin Purcell: Yeah, we will eventually get there. I don't know if it's necessarily going to be deep research that is going to do to do it. So yeah, I think everybody's, I think the single digits amongst us are safe for a little bit longer. Keep making stuff. That's the key. Well, I think the interesting thing here is that like.

Gavin Purcell: You see stuff kind of chipping away at the things that are promised, right? Like people have talked about for a long time. 2025 is going to be the year of agents, right? We and we've discussed on this show kind of ad nauseum what an agent is. But in case you're listening for the first time, that's the idea that an A.

Gavin Purcell: I. is going to go out and do stuff for you and come back with the result. Uh, deep research is a version of an agent and to Tim's point, like, uh, Google has a version of this same thing. Operator is another version of an agent that they release, which is something that can access the internet, which we talked about last [00:11:00] week in the show.

Gavin Purcell: And clearly there are seeing, uh, companies are seeing value in this, or at least, uh, Uh, Masa from SoftBank is seeing a lot of value in this because they have now partnered with OpenAI to create a new company called SB OpenAI Japan. And there's a bunch of news around this. Uh, Samwen did a, a speaking gig with Masa and like.

Gavin Purcell: It's clear that the money pouring in here is now I think people are trying to see an end game here, right? So one of the things we've talked about in the show a lot is this idea of kind of the endless fount of money coming in In fact, we didn't even cover this last week because it happened after our show But you know opening eyes now valued at three hundred and forty billion dollars or is going to be valued at that Which is a just an astronomical amount of money As somebody who works with these tools all the time, and obviously spends money on them, right?

Gavin Purcell: You and I, and our whole crew, like we, we spend money on them. Do you think 200 a month right now. Well, I know that's what we, we track it, right? Like it's not cheap, right? You're 200 for just the, for [00:12:00] the, for the pro plan. And on top of all the other video tools and stuff, but like, you know, do you see a world where this stuff, the promise of the economics of this is going to work out?

Gavin Purcell: Or do you, where are you at on the whole kind of like AI hype cycle of, of, of all this? Of like, of, of the, uh, utopian equality, uh, and, uh, all, all tools for everyone. Um, look, I, I, I think that it's always going to be a bit, you know, Tiered, like it just, I just don't, I mean, as much as I would love to see, I don't know, like, does it become like EDU licenses at one point when you're in school and you have access to the latest and greatest of the, uh, LLM models, and then if you don't go to college, then, you know, you have to start paying for it.

Gavin Purcell: Like it might be sort of a bit like that, where you have sort of your, you know, you have your basic access AI, uh, package, um, and then you have gradual, like, Do you want your cable package as well? Do you want your premium subscription [00:13:00] package? All of those are going to cost you more. You know what I mean?

Gavin Purcell: So that's more than likely. I think the direction that things will probably head. I don't see, um, a complete, uh, I mean, I don't want to get into the whole UBI talk or anything like that, but that becomes an eventuality at some level, um, but you're still going to need to pay for some kind of like AI service as well.

Gavin Purcell: Yeah. Yeah, yeah, I think so. I mean, well, I don't know. The other thing I think that's really important to think about was that, uh, well, actually two things. I do want to get into this one thing that really kind of shocked me coming out of the, uh, deep research thing. And I don't think has gotten enough attention, which is the fact that deep research scored a 25 percent on humanities last exam.

Gavin Purcell: So if you name that humanity's last exam, first of all, that's, that's problem number one, right? Like we shouldn't be calling something humanities. It could be like. The humanities, like even last chance or all sorts of things sound terrible. Like we should come up with something better, but just so humanity's last exam is a [00:14:00] AI benchmark that was created by, um, Dan Hendricks and a bunch of people to basically.

Gavin Purcell: really stress test these AIs to really make sure that we create something that would last for a couple years. We talk about uh, Terence Tao, one of the most famous mathematicians in the world is involved with it. They created this thing and uh, I wanted to point this out because Kevin Roos, the New York Times writer, uh, he basically said when I wrote about humanity's last exam, the leading AI model got an 8.

Gavin Purcell: 3 Five models now surpass that. The best model getting a 26. 6, which is deep research. That was 10 days ago. So 10 days ago, the best model got an 8. 3 and now we're here with deep research at 26. 6 and. What's crazy here to me, Tim, is just, it talks about like the, just the pure speed at which we're moving.

Gavin Purcell: And when, when I try to explain this to people, you know, mostly my family or people that are around me, they're just like, okay, whatever, you know, sure. AI is what it is and chat GBD is fine. But like, I, [00:15:00] I think between deep research. And the other stuff that's kind of bubbling up for the rest of this year.

Gavin Purcell: There's a little bit of like, how do you talk to people to understand that? Like, and it's not like trying to overhype it, but just to understand that there is going to be a change in what these things can do and how much they can do. So, um, during, uh, the holidays, um, I, I, I went to my wife's, um, company dinner party and, uh, we were kind of seated at a table and, uh, there was another person that is, uh, It's kind of relatively as plugged in as, as I am.

Gavin Purcell: So they were, they were talking to us and we were sort of, you know, typical holiday party talk, you know, uh, it's just like, everything's going to change. Uh, you don't have no idea what's coming. It's all happening right now. And I mean, like, look, I, I think that when you get outside of you and I talking outside of everyone listening right now, if you were just, if you weren't in this and you just heard us talking in like a grocery store, you would be like, those people are crazy.

Gavin Purcell: For sure. Yeah, absolutely. I think, but I mean, it doesn't [00:16:00] help when, you know, we're like, uh, humanity's last exam. Yeah, it's not, it's not great. We should call it like, like humanities, uh, humanities doing great exam. Like it's like, don't worry. We're 75 percent ahead of what they can do right now. Which are all of these researchers dressed like, uh, like cowboys in Rio Grande, like just with their guns, just waiting for like, I guess the robots, the robots are coming in.

Gavin Purcell: I should say, I should say, uh, Dan himself, uh, did reply and saying, uh, not sure our last exam will survive optimization. So I think that we're even himself, the man who created it is definitely feeling it. And I, before we move away from this, I do think it's important to kind of, you know, Uh, play a, a quote from, uh, Masa and Sam's, uh, press conference, which really did not only kind of shock me, but really made me laugh.

Gavin Purcell: So let's take a listen to that. I will bet you in the next, uh, several years, 10 years, it will [00:17:00] gradually start to have at least understand the people's emotion. And then gradually it will start to have emotion by itself. And it's a good thing to protect humans. People think, or if it has the emotion, it's a disaster, it's the demolish, it's a you know, the bad thing for that's the end of human, because they're going to fight and kill you, destroy you.

Gavin Purcell: But I would say, if, if their source of energy was protein, then it's dangerous. Their source of energy is not protein, so they don't have to eat us. Okay? There's no reason for them to have reward by eating us. That's brilliant. That's brilliant. So, so Tim, no one is going to eat us, which is good to hear.

Gavin Purcell: Like no one's going to be eating us. I would just, I would like to point out as well, there, [00:18:00] there is, there's nothing greater, uh, than hearing a, uh, Japanese man talk about a large giant robot that is coming to destroy us all. You're right. That is like, we are seeing like a very fun science fiction movie.

Gavin Purcell: Fiction movie play in front of our eyes. So if Mecha Godzilla ate people, then things equal bad. That's open AI. So now we have a perfect thing. That's a, there we go. We know what it is going forward. If that's the case, do you think as Microsoft is the 800 pound gorilla, do you think that that mechanizes at some point, then instead of just having a giant, you know, AI company battle, we can have a Mecha, you know, and we have a giant Mecha llama because we have a llama involved.

Gavin Purcell: I mean, I don't know what, and then you probably have to stop looking. Right. A very bookish guy looked like Claude that suddenly just gets super jacked. He becomes like the Hulk. It becomes like the Hulk. So anyway, this is opening eyes. Big thing. They obviously it's a huge week for them. I'm really curious to know like what's going to come down the pipe again because like, I mean, they've clearly been loading up fastball after fastball.

Gavin Purcell: The big question I have for you [00:19:00] is here. This Tim is, you know, there's, there's all this has happened in January or now it's February but just barely February. Is it? Yeah, I know. It's crazy. And, and there have been like hints and Sam has hinted in different things. He did a Reddit AMA that people should read that like, you know, the GPT five models coming, but like clearly there are other models still coming this year.

Gavin Purcell: And not only opening eye, we're talking about, you know, the new Grok model will clearly Gemini is going to be coming out with something very soon. Um, Claude, I would assume, uh, therapy is going to be coming out as well. What do you expect, what do you expect for the next, like, you know, six months? Are you expecting like this sort of pace to continue or do you feel like we're going to get a little bit of breather and we'll kind of have some ups and downs throughout this year?

Gavin Purcell: I feel like there's probably, it's probably just going to be, it's probably going to be a little bit like what I'm experiencing on the video side as well, which is just like, it, it, it, It's just like this one, two punch of like, um, you know, somebody releases something and then a week later, somebody else releases something, but it's all about the same.

Gavin Purcell: It's everything [00:20:00] stays at kind of the same level. So I think that we're probably going to be hitting a point of, um, Wonder exhaustion, um, oh, interesting. At some point, like where it's just sort of like, eh, you know, like, like now, um, Chatt, PT, like, yeah. Uh, Chatt. PT now can, you know, uh, read the emotions on my face and, uh, and, and it can make my monitor cry.

Gavin Purcell: How come I can't make my kids cry? That's the next question. It's like, I can make me cry, but it can't make my kids cry. What is it? You're absolutely right. By the way, this is what happens is what happens with people, right? It happens every time new technology comes out. We just get used to it. And then we decide what's next.

Gavin Purcell: So I was just in San Francisco, uh, the land of the Waymo, uh, if you will. And like, uh, and I do not live in the Bay area or in a Waymo city. Um, so. It was, I mean, at some point I had to look around and go like, those are robot cars. Like there are robot cars just driving around everywhere. There are cars without people driven like by robots, ghost robots at that, [00:21:00] um, that are, that are just, and no one, everyone's just like, eh, there's a way, man.

Gavin Purcell: And I keep thinking like, if you were to go back to, if you could take my grandfather in his twenties and teleport him to that moment where you're just like, yeah, there's a bunch of like robot cars just driving around. By the way, I have, you know, all of the world's information in my pocket right now. You know what I mean?

Gavin Purcell: But like what? And he's not, that's not. that long ago. Like, you know, um, so the, yeah, the, the amount of like future shock that we have experienced and we've adapted to so quickly of just like ghost robot driving a car around, I guess I'll get in it and you're going to trust it because it's now normal.

Gavin Purcell: Yeah, exactly. Well, everybody, uh, that is deep research, but before we move on to the next thing, I want to remind everybody to subscribe to this show and specifically go check out Tim's show. So Tim's, uh, YouTube handle is, uh, is it at Theoretically Media? Is that what your YouTube handle is? I think it's, I think you do youtube.

Gavin Purcell: com backslash at Theoretically Media or just search Theoretically Media. Or if you can't spell that, just type in AI video and my stupid mug will probably show up at some point. That's [00:22:00] right. So Tim does a lot of great demos of, of brand new AI stuff that comes out. We're going to talk a little about one a little bit later in the show.

Gavin Purcell: But definitely go check him out. And if you're checking us out, please like, and subscribe as per usual. But also if you're listening to us, we always love it. When you leave us a review on Apple podcasts, that does really well for us. Um, we've been doing a few other podcasts last week. I was on a barren street wise podcast where I talked about deep seek, but also talked about AI video stuff, which I thought was pretty interesting.

Gavin Purcell: So go check that out. check out Tim's channel as well. Next up, we're going to briefly talk, uh, Tim about O3 because O3 mini is the thing that kind of got shoved to the side by deep research. And I just want to kind of make sure everybody knows O3 mini dropped last week, late last week after our show came out.

Gavin Purcell: And this is a free reasoning model. And this is clearly OpenAI's direct response right now to the DeepSeek R1 stuff. Obviously, as we covered last week, DeepSeek R1 is very high up in the app store. A lot of people are using it. They wanted to show that they had a [00:23:00] powerful reasoning model, more powerful reasoning model that would be accessible to everybody.

Gavin Purcell: And, uh, I think the thing that's surprising here a little bit is that they dropped it for free, but also that they dropped it this quickly because I really didn't think that this was going to come out as fast. Obviously, deep research was a surprise, but this is another big deal from them. I don't know.

Gavin Purcell: Have you ever had a chance to play with those three mini as well? Yes. I played with a little bit. Yeah. The reasoning stuff is just, is fun and fascinating to, to click along and just read along with like, where's your brain going there? Oh three. Like, uh, I, I, we had that with, Oh one as well, but, uh, Oh three is just, it's faster for one.

Gavin Purcell: Yeah. Uh, and then, you know, I, I presume the level of questioning that I'm giving it is, I mean, it's not even breaking a sweat with the questions that I'm asking. I look like, you know, uh, how long do I air fry potatoes for? Yeah, I know. That's like, that's like a classic. It's like the GPT 4. 0 questions.

Gavin Purcell: It'll be like, we need to kick you down to the little brother. I mean, one thing they talk about with this is it's much better on coding and much better on science questions. And, and I [00:24:00] do think this is something at least tangentially that I have seen seen people do and there's been a lot of people out there who are using open, uh, sort of using Oh three mini to one shot, really interesting kind of like physics puzzles.

Gavin Purcell: I don't know if you've seen these videos that are floating around. There's some really interesting stuff that's being done with it. And in my mind, this kind of transitions us to the next story, which is that replit just released a brand new. app and the iPhone app and what's so cool and replet if you're not familiar, replet is a company that basically allows you to code in the cloud and do all these AI models.

Gavin Purcell: They have an AI agentic coding tool kind of like cursor, all these other things out there. But what they did is they released an app and today I downloaded it and play with it. And it is what I think is so interesting about this is it, is it powered by the backend of these high end models and which are very good at coding.

Gavin Purcell: Right. But they don't expose you to the coding because what it basically does is it allows you to type in something that you want. And I went in there and asked him, Hey, make me a marble race game. Like make me a marble race game and see how it can [00:25:00] do. And you just kind of watch it install all the libraries.

Gavin Purcell: And if you're not a coder, you don't know a lot about this. You know, you might not understand, but like when you load up a program, it needs a bunch of different Python libraries or different things from certain places. And then it should, and then it actually returns a result. You really don't see the code at all.

Gavin Purcell: And to me. What's interesting about this kind of transition point in this world that we're living in, and you know, we've talked on the show a lot about the idea of that you can be a little bit more of an idea guy now, and you don't have to be a hardcore coder to kind of do these things, but now we're transitioning to the idea where, like, Anybody can make software, right?

Gavin Purcell: And when anybody can make software, it is a, it is a completely different sort of world because it's, it just changes it. And I thought when I was doing this, and I'd be curious to know about this for you. I thought about what it would have been like to have this tool when I was a kid, right, or, or to give this tool to kids now.

Gavin Purcell: Because the idea that you can just imagine anything for a computer to do, and that through the kind of like magic of AI prompting and [00:26:00] coding, it will return you something that you can then tinker with and make into something even better, that feels like magic. Such a transformative moment in terms of how we use software and not only how we use it, but like how we make it make it.

Gavin Purcell: Yeah, for sure. Yeah, for sure. Yeah. I mean, we've seen a couple of examples already of people just, you know, drawing a picture of the website that they want to see, you know, uploading it to an LLM and having it then code that website. I mean, that's that's pretty it. So essentially, yes, we are. We are very much reaching a point of, you know, If you can dream it or explain it, uh, you can build it.

Gavin Purcell: The one thing that not to, not to go AI Doomer, uh, or anything like that, but I mean, like there has been some interesting stuff, um, that we've seen recently with like two LLMs talking to one another. I don't know if you've seen this where they've switched over to like Chinese, just because it's easier.

Gavin Purcell: It's I've actually, it's a trick that I do with video prompting every once in a while too, because a lot of times you'll have that thousand or 2000 character limit. Uh, but if you [00:27:00] translate it to Chinese. You get more characters because oh, interesting, more efficient language. I do sort of wonder that when we hit a point with something like the replica, um, you know, coding app or anything like that, where if it starts to become more efficient in its coding language, and there was an example that we saw just kind of recently as well, where what we were looking, we were looking at two AI agents talking to one another and it was starting to turn into what we perceive as gibberish, but they perceive as, but they were connecting with each other in some way.

Gavin Purcell: Yeah. It's like their secret language on a date and they don't want anybody else. The other table to hear what they're saying. I mean, there is a level of like, you know, when you buy like, uh, like a BMW these days and you can't even get into the engine. If you, even if you could get into the engine, you wouldn't know what to do with it because, uh, you know, that is probably going to happen at some point.

Gavin Purcell: I guess there's a part of me that's a little bit worried that, you know, what happens when we don't have the mechanic anymore. Um, and at that point, like. We're sort of, how do we fix things? How do we know? I mean, I think this is a very [00:28:00] rational and reasonable argument and not only how do we fix things, but what if the BMW starts driving us in a place we don't want to go to and we can't tell it not to, that's where it gets a little scary.

Gavin Purcell: I mean, so there's two different sizes. Yes. I completely understand that aspect of it. I think for me, when I saw this, it was just me thinking. If you could hand a tool like this to a eight year old and the eight year old was like somebody who's like, Hey, I play Minecraft all the time and I see people do Minecraft mods, but I don't really understand how to code this thing to your point.

Gavin Purcell: I'm not saying they shouldn't learn how to code that thing, but like just the fact that they could like get the first few steps on the way and then dig further down because in the replica thing, you're able to kind of unpack the code. Like the code is there and you can look at it. But you're right. You know, it's funny to think about this way because it really does show you there's kind of two sides to this.

Gavin Purcell: One is you make these tools accessible and interesting for everybody. But on the other side, it's a little bit of an easier way into this. And I think a lot of coders would be like, well, I spent my whole life, you know, learning these basic rule sets. I just hope that like, [00:29:00] you know, I do think that to your BMW metaphor, there will be the BMW mechanics and they'll be paid very well because they know how to access these tools, but there will be a lot more drivers, right?

Gavin Purcell: And there'll be a lot more people who can drive those interesting cars because or a very personalized car. Like, I mean, that's one of the things that replica is kind of talking about here is the idea of like, Hey, you could make your own piece of software for X. Like, you know, if you're making You know, uh, a video and you wanted a specific filter for something.

Gavin Purcell: I've seen people use blender with Oh three and make all sorts of interesting stuff. Like that feels like a really cool way to look at this going forward. And that is something that I think that, uh, is I hope doesn't get lost over this whole, um, uh, idea is that, you know, going into like, I'm still a blender dummy.

Gavin Purcell: I'm, you know, I'm consistently on the, uh, the one on one tutorial, uh, you know, in, on YouTube or, but you know, there are things, there are times when I've had, um, I've been experimenting lately with, uh, you know, uh, with screen capture and letting, and having something like chat GPT teach [00:30:00] me right. At some point with, uh, Something like operator.

Gavin Purcell: I'll just be able to say, you know, Hey, comp me these three video pieces together and then, you know, have Thor come in and turn into a frog or whatever. I want, you know, and it'll just take care of all of those things. But at the same time, I'm not learning the skills of how to do that. Um, so, you know, I, I've worked in Hollywood for a long time and I, and I've worked, I've worked, you know, I've produced stuff where I've told artists what to do.

Gavin Purcell: So there's that part of it too. You know what I mean? Like where, and then I've checked on it afterwards as a producer or whatever. So. I hope that we spend the time that we save via all of these models, learning more about what we're actually passionate about. No, we're not. We're going to watch love Island 15.

Gavin Purcell: That's what's going to happen to us all. We're going to watch, but honestly, like I, again, not they've gone on this a long time, but like, The experimental thing is the thing that I think that's important, right? Like in, in letting people experiment and see how this works and bespoke software is very cool.

Gavin Purcell: Speaking of experiments, there's a really interesting thing that came out of ByteDance that I really want to talk about because this is not something that is out [00:31:00] yet, but it's another one of the famous like papers that have been released, but we have not seen code yet, but this is called OmniHuman 1.

Gavin Purcell: OmniHuman 1. And it is a really fascinating, Lip sync and kind of like body sync tool that you spent some time looking into. Tell us, tell us about what this is. You can kind of think of it as like, maybe like, uh, the good tent pole to it is a Haygen. Um, so, uh, you know, sort of a AI avatar body avatars. If you ever noticed with the Haygen stuff, it always kind of has like this weird kind of like the motions are, yeah, it's a little, yeah, it always kind of gets a little off.

Gavin Purcell: Little bouncy. Um, so this aims to solve that by being able to be driven by input video, input audio, um, and so on the video, it's looking at two things, body language, essentially, and hand language pockets. Listeners can't hear this, but, um, you know, I think that they've all experienced that whole thing where, when someone's talking, it's like that hand is cycling [00:32:00] completely out of sync with, uh, you know, with what people are saying.

Gavin Purcell: So, um, this aims to sort of solve that by analyzing driving video. Yeah, I mean, there were some interesting ones I saw. The Albert Einstein one was really interesting, right? Yeah, So the Albert Einstein one is your more traditional kind of look at what it would take to take a picture in an audio and make a, an animated person.

Gavin Purcell: Right. And we can show that here. If you, if you didn't see this here, please check out our show notes. The other one that was really interesting to me was the, the Taylor Swift one, right? Which again, I love these Chinese models. They just kind of decide we're going to use the biggest stars we possibly can.

Gavin Purcell: They had Taylor Swift, you know, Swift singing, um, a Chinese song. I'm not sure what the song was, but it was singing a Chinese song and singing What's interesting about it is each step kind of gets better and better, and you're right in that, like the kind of the whole experience of this one starts to feel much more succinct, right?

Gavin Purcell: It feels like it's kind of of a part. And I guess the question will be is whenever I think of a ByteDance thing, like, Again, these Chinese models, especially video models are getting so good. I know you [00:33:00] spend a lot of time with cling and mini max. It is interesting to see, like we will, I feel like at some point get an American like ultimate video model and we will get a Chinese ultimate video model where like all these things live in and that'll be the one that kind of like takes off.

Gavin Purcell: I, yeah, I would agree with that. I, you know, the, the, the one, the one AI video model to rule them all, which I'm actually not looking forward to. I think that once that does happen, things start to become, to look a little on the homogenized side. Oh, that's interesting. Yeah, I can see that. Um, you know, it's kind of like how that whole, like, Uh, every Marvel movie started to look the same after a little while.

Gavin Purcell: Like every, every red cape was the same color red, that sort of thing. So I'm, I'm actually, I'm really not looking forward to when that happens. Uh, again, great artists will always find ways around that to break that, to get around, uh, the, the look. But, um, but yeah, what is your, what would you, what was your favorite video model right now?

Gavin Purcell: Like you're somebody who spends a lot of time playing with video models. And if you haven't watched Tim's videos, you should. He has a lot of the similar same sort of tests he does [00:34:00] with each one. Wasn't there a guy like in a blue suit that you use for each of your tests? Yeah, man in a blue suit. Where do you sit right now on what the best model is, um, as an overall pick?

Gavin Purcell: So is there one that you like the best? Uh, you know, so I mean, they're all my children. Uh, not, not fair. I do, uh, I like all of them. I really do. I think that there's something kind of cool and unique about every single one of them. The ones that I'm currently kind of playing around with the most would be.

Gavin Purcell: Probably be, uh, Hylu minimax and Kling, um, uh, Kling, I think has, uh, they have like that, this like sort of ingredients, uh, thing where you can put, um, input for, you can put like two people, uh, a location and a prop and, uh, mash them all together. And it, it does that thing better than any of the other video models currently.

Gavin Purcell: Uh, and then the other one that I'm like still trying to wrestle like a, like a, like a wild beast is Sora, uh, cause Sora is, I mean, it's, It's, it's unpredictable, but every once in a while, man, when that thing hits, [00:35:00] it hits, I'm, I'm, I'm dead convinced that there was something, there was, there was something to mine out of that.

Gavin Purcell: So, uh, I honestly think Sora, I mean, especially when you have the pro model and you could really kind of rip out a bunch of different exports. So really always surprises me with. The weirdness of it. And I, we've talked about in the show, how much I love weird results. Sometimes I mean, I video, like I, you know, I made a video a while back with VO and VO is amazing, but like, you don't get as many weird results and it's like, it's somewhat disappointing, but also when you're trying to do work, I can understand that.

Gavin Purcell: Um, okay. We should get through a couple of other pretty big, interesting video things. We'll do them fast. Matt. Anyone was a very small, interesting AI tool that was announced again. This is a not, not code is out there, but this is a straightforward Really good AI way to Matt out the background of a video.

Gavin Purcell: You also covered this. What's your take on this? I love it. I think it's, it looks really, really good. Um, if you go through and watch the examples, uh, there was two that really kind of blew me away. Uh, the Heath Ledger Joker, uh, kind of walking and you can, and when [00:36:00] you, it's not necessarily in the Chrome of green.

Gavin Purcell: But when you see the alpha mat and you just see all the wisps of hair, um, like that is, um, like that's better than like this, like the standard after effects, you know, plug in, um, it's, it's getting into like nuke territory like at this point. So like the, like, I mean, this is the thing that, you know, I, we is James Cameron came out, whatever, three or four months ago and said AI.

Gavin Purcell: Like, this is the thing that feels like that will make a difference because Matt, anyone is the kind of tool that you could see easily being integrated into the Adobe suite or into final cut and being super useful. This is an AI tool and it is incredibly good at what it does. Oh, absolutely. Yeah. Uh, it also is the name of a new YouTuber, uh, definitely the AI.

Gavin Purcell: Base along with Matt Wolf. Matt, uh, there's another Matt. Matt. Yeah. Well, there's Wolf, there's Berman, and there's, uh, Maddy, uh, Maddy, ai Young, Matt, ai. So now Matt, anyone? Oh yeah. So now we have Matt. Anyone. That's great. That would be the fourth, Matt. That's amazing. Yeah. Uh, and then finally there's another video that came out, um, from Meta weirdly called [00:37:00] Video Jam.

Gavin Purcell: And actually speaking of so's. This was really interesting. They're attempting to try to fix the physics in AI video by doing some pretty technical stuff. Have you dug into this yet? I took a look, I took a glance through the paper, uh, kind of got the gist of it, um, but I haven't actually dug into a, how they're doing it.

Gavin Purcell: Um, and B, um, you know, sometimes with the, with the papers, you're like, all right, all right. I see. All right. Yeah, all right. Did you? Did you really? Uh, Matt has always been pretty good about it. Um, but basically, yeah, it looked like it was taking, uh, because what was it like break dancing footage where if you normally try to prompt for that, you would end up with like a twisty Play Doh man, like kind of like, like sponging all over the place.

Gavin Purcell: But this was actually figuring out like how to get that correct movement via it kind of looked like it was using, uh, Um, like, uh, like, like some 3d modeling going on in there, or at least some. It's a very technical, it's a very technical back end, but what it talks about is like that the, the video models right now are [00:38:00] kind of ignoring motion and they prioritize what they see.

Gavin Purcell: So that what they're trying to do here is kind of layer on some of the. So that it will pay attention to the motion and the physics of the scene. Um, it's funny though, as you were saying this and the videos are all really interesting, you should check it out. Like, especially if you spent time trying to get AI video to do gymnastics or anything like that, this could be a big deal, but you to talk about the idea of like, you know, 10 center bite dance, often like, you know, being vaporware or something we haven't seen.

Gavin Purcell: We haven't seen a meta model in the world, a meta video model in the world either. Now, I know they exist. It's kind of like when V01 came out with Google, they didn't release to the public. But Meta is probably due for legitimate high, high res video model because they've been doing the work there for quite a while.

Gavin Purcell: I from last I heard Movie Gen was actually getting rolled into Instagram. And yeah, so it was going to be more it was going to be less a video generator and more like. Snapchat filters 2. 0. Um, so yeah, yeah. I'm not as excited by that. Then you had me at ooh, [00:39:00] and then I was down. Yeah. So, um, so yeah, it was kind of more meant to be like, uh, replace my t shirt with, uh, you know, you know, like, uh, mom's best friend shirt and then put me in the Arctic and ha ha ha.

Gavin Purcell: Now we have a funny name video. So I, I don't know. Um, I mean, certainly for, From the paper, uh, they could flex and just release something that is, you know, like that completely blows sore, uh, uh, uh, uh, view to video. I think you can blow a mall out of the water. Hypothetically. I don't know. We'll see. So, but I just doubt that they will.

Gavin Purcell: So, cause what's in it for them. Well, that's so funny. I was going to say that. Like, it's so interesting. These video models, like it's on some ways it's more hassle than it's worth a little bit for them. I mean, because a cost of fortune to, to render video models, right? There are much, there's a lot more data that are going through, but also because of all the right stuff, you know, they're, they're a giant company and like, The, the win for them may not be in that space, but anyway, it's a pretty crazy week for this sort of stuff.

Gavin Purcell: And I think every, every week we see new things come out and every [00:40:00] week, but what's even crazier this week is we have some new robots that we got to get into. It's time for robot watch.

Gavin Purcell: Tim, there's another Tim who wrote a tweet this week that I found very prophetic and I always enjoy. So Tim Urban from wait, but why, if you're not familiar, it's one of the classic kind of blogs of the internet. Very, very smart writer wrote a great book as well. But so Tim's tweet said humanoid robots and drones will be absolutely everywhere 10 to 20 years from now.

Gavin Purcell: Delivery drones flying overhead, Robert robot janitors, baristas, grocery store employees guiding you to the right aisle and eventually housekeepers. Will all seem as normal as smartphones do today. And I, I think I agree with this pretty clearly. I don't know what your thoughts, you, you feel like you're in the same boat.

Gavin Purcell: We were talking about the ghost robots, uh, drivers, the Waymo, uh, ghost robots that were our ghost robot cars that were, uh, driving around San Francisco and no one bats. And I, um, so yeah, that's, uh, [00:41:00] that's it's funny 'cause actually I, I was talking to my, uh, Uber driver on the way to back to the, to SFO, uh, about the, uh, the Waymo's.

Gavin Purcell: I'm like, how do you guys feel about that? He's like, well, you know. What are you going to do? I was like, yeah. And he's like, they can't drive to the airport at least. And I was like, Oh, cause I, they, not yet at least, but apparently, cause I guess they won't take direction, uh, from, you know, traffic, uh, Oh, I never thought about that.

Gavin Purcell: That's interesting. That's why they can't do it. That's why they can't do it. And I just can't help but think that you can't, you can't put those things into like, you You know, um, like, like Philadelphia, for example, like in Philadelphia, like you remember they destroyed the robot, they Philadelphia killed the robot that was no, I didn't do that.

Gavin Purcell: Somebody did. I can't remember the research project. They started a robot. They built it to literally walk across the country, got to Philadelphia and Philadelphia, like literally just beat the robot up and killed it. Um, yeah, So I do think, yes, well, I do think that we will see, uh, robot delivery drones flying everywhere and, uh, robot, uh, humanoid robot workers, I guess, kind of like going about their daily business, doing their things.

Gavin Purcell: Um, I don't put it past [00:42:00] human on robot violence. Well, this is what I want to get into is because the robots that, that NVIDIA and Carnegie Mellon introduced this week. Week, our sports star robots. So I think they might be able to kind of dodge out of the way really fast. We're texting. So, so this story is pretty interesting.

Gavin Purcell: They actually, so NVIDIA has been talking forever about a simulated training, right? And what simulated training is, is like training robots in a virtual environment. So they know what's real to go when they, when they be able to in the real world. And what they did here, this is the craziest phrase I've ever heard.

Gavin Purcell: Dr. Jim fan, who we've talked about on the show before works at NVIDIA. He said he's announcing ASAP, a real to SIM to real model. And what that means is that this is hammer song. That's right. Real to real to real. Uh, this is basically taking real world. Motion. And in this case, they took the motion of LeBron James, Cristian Ronaldo, the soccer star or football player, whatever you'd like, and Kobe Bryant.

Gavin Purcell: And they use that [00:43:00] in a simulated universe. The robots practice it. And then they brought those behaviors back into the robots. And the video you're looking at here, or if you're in the show notes, go, if you're listening, go to show notes. Is these robots basically acting as if they are those stars? And this feels like, you know, A really interesting way to think about how, you know, you talk about in Philadelphia or New York City robots, you know, we could get the crappie out of them or, well, you put Chuck, uh, Chuck Liddell's, you get Chuck Liddell training and it's a whole new game, Tim.

Gavin Purcell: It's a whole new game. I mean, so the other interesting thing about robots this week was that you may or may not be familiar with the Twitch streamer Kai Sinat. I mean, you probably are at this point. He's basically a mainstream celebrity. He had one of these robots with him Uh, kind of in his life. And it was a really interesting thing to think about seeing.

Gavin Purcell: We see these videos, these research videos of robots all the time. And what was interesting is to watch Kai, you know, interact with this robot. Now I don't know for a fact if this was something that was a [00:44:00] sales deal or if I'm sure he was sent this robot because of what it is. But. What was interesting is like how normalizing it is to see this in the beginning to your waymo point like at first like everybody's kind of weirded out by then he's kind of bringing it around and it's still kind of surprising to see people interact with it.

Gavin Purcell: But like, to me, this is like the kind of first step as to. How humanoid robots start to become part of our world, right? Like it's the same way with the Waymo. You might hear about it at first. And it's only in San Francisco and I live in LA. And so when they first came to LA, when I first saw one, I was like, Oh my God, it's really a, uh, uh, uh, nobody's in that car.

Gavin Purcell: And then I got in one and it was like, Oh my God, I'm driving. And there's nobody in this car, in this car. And then three rides later, it was totally fine. Right. So, and that's where like, I think. San Francisco probably will be the first place, but there will be, you know, to Tim Urban's point, you know, five, 10, 15 years from now, we're going to be on a street and like, you know, a fifth of the things walking on that street could very well be humanoid robots.

Gavin Purcell: And that is like kind of the world that we're entering here. Well, you [00:45:00] know, the funniest part about that too, is that, so eventually there will be in a robot in your house, you know, doing your dishes, doing your laundry, your robot Uh, maid, who was the, who was she on? Uh, the Jetson. You know, Rosie, Rosie.

Gavin Purcell: We're really going deep in the, we're really going deep in the, uh, gen X world. Yeah, we really are. Yeah. Um, that's, that's, those are my people. Um, yeah, so, uh, so yeah, we will have, you know, that it'll be amazing and cool. And then like within two months. We'll be like, ah, it doesn't, it's like it just, it ran out of batteries and didn't do the laundry again.

Gavin Purcell: Like, yeah, no, I know. And then you'll get, and then it'll get mad at you and you'll get mad at it. And then you'll have to live in separate houses. And suddenly you're like, you're starting a divorce agreement with your robot. It's like a whole different world. Like, honestly, that's something to think about.

Gavin Purcell: We're going to have to figure it out. Like, Do robots, are they members of your family? And if so, like then, you know, I mean, obviously like Rosie 2. 0 is going to come out and your Rosie 1. 0 sucks. So like, you're just like, well, see a family member. I know. And this is, by the way, there [00:46:00] was a great show called humans.

Gavin Purcell: Did you ever see that show? It's a really interesting show of like the upstairs downstairs life of what it will look like in the future with robots, like become as prolific as they are. And like, That is a real thing to think about, right? Because like, eventually, if these, if these AIs keep getting smarter and smarter, and, you know, some people argue they're never going to have like some sort of a consciousness, but like, eventually you're going to have to think about how you treat them.

Gavin Purcell: And in fact, one of the other things I'll say about the Kai Sinat stuff is there was a video. Where they were kind of sitting around and kind of pushing it around. And we often say, please don't push the robots. Don't matter. No matter what, don't push the robots. Dynamic robot will eat your face. These robots are trained on watching and listening and reading everything.

Gavin Purcell: Just make sure the robots make you know you like them. That's the most important thing. I think of all this alright everybody It's time for us to go around the internet and see it some of the fun things people have done with AI it's time For AI see what you did there Sometimes you're scrolling without [00:47:00] a care then suddenly you stop

Gavin Purcell: So Tim, this is one of our favorite things where we just shout out a bunch of cool things people made online. And I know you've seen a bunch of stuff and cool things out there. I do want to shout out something. This really made me laugh. So basically this is another dumb version of somebody just taking a funny idea and making it.

Gavin Purcell: This is called Lord of the Rings, Kings of the Cage. And they turned the stars of the Lord of the Rings series into UFC fighters. And then there's a little bit of drama at the end of it. It's funny. It's funny. I mean, you've, you know how these tools work. I'm sure you see stuff like this all the time, but whenever you see like a very funny idea executed, well, it's just a fun thing to watch.

Gavin Purcell: Um, yeah, I mean, so good. So, so good. I think that I'm not entirely sure, but I think they did a little, they must've done some face swapping. On afterwards. I don't think that's coming straight [00:48:00] necessarily out of a video generator, but yeah, I mean, it looks fantastic work here. Um, characters all look incredible.

Gavin Purcell: Um, so yeah, I think that there's definitely some image to video, some face swapping and, uh, the fun of it looks like it's mini max. Well, here's a question for you, um, because I've done some face swapping, but not a ton, but I think a lot of these celebrity videos, the door brothers or many of these are not just taking straight out from mini max were.

Gavin Purcell: results. What is the best face swap tool to use now? Are you, are you, are you up to date on them? I know that the The open source ones can be pretty good. Yeah. People, uh, the one that, I mean, I tend to like a lot is, uh, uh, is face fusion, uh, which you can find, you can run that locally, uh, via a, uh, app called Pinocchio.

Gavin Purcell: Uh, we talk about Pinocchio all the time. We talk about, uh, yeah, face fusion, uh, pretty easy to use. Uh, I did a whole. Tutorial on it. I think, uh, I face the thumbnail for it. If you ever see it is, uh, uh, Uma Thurman in Kill Bill. And then I face [00:49:00] swapped it with Scarlett Johansson. Cause that's, that's just funny to me.

Gavin Purcell: That's awesome. Okay. Next up in this list, um, people have been doing all sorts of interesting stuff with Oh three mini, especially when it's talked about with it's very good at coding. This guy, uh, had this really interesting tweet where he basically came out and said, Oh my God. Oh three mini has finally allowed me to do the thing I've always wanted to do, which is an ASCII art generator, which, you know, basically ASCII art.

Gavin Purcell: I, if you're talking about our age in some ways here and growing up on stuff like ASCII art was like the early internet. like, you can generate almost anything, but you had to use the characters on your keyboard to do it. And what this guy did is he basically coded an AI ASCII art thing, so you can upload a photo to it and it will give you an ASCII art back.

Gavin Purcell: And again, this is more interesting to me. I mean, the actual thing is cool, great. You can get ASCII art, but that was interesting here is that this guy, not a coder, Use O3 mini to do this thing for himself. And that is such a cool thing. I love that as, [00:50:00] uh, he's playing around with the sliders and this too.

Gavin Purcell: It kind of, it just sort of has like this weird, hypnotic, hallucinogenic kind of like look to it. Um, that's, I mean, it's ASCII art is ASCII art. It's just, there's something kind of unique and incredible about it. And then finally we have, um, one of my favorite creators of all time. And if you know, if you're familiar with him, but neural viz, who is a great A single human being continues to expand his universe and he has created the Gloron Dating Show.

Gavin Purcell: And this is a dating show set in his humans universe and I always shout him out when he comes up with a new format. Uh, let's take a listen to a little bit of this. I love him so much because my standards are so low that I can only really dream of having such a barely suitable partner. What are you thinking about?

Gavin Purcell: Thoughts. Are you thinking about bugs again? If they sleep? That's all I think about. Dorgan, they do. We know that they do. You're pondering again. Stop it. I'm worried that [00:51:00] Dorgan is uncomfortable because of the drama that happened in the house last week. F you, Morky. This guy, just one of our favorites, he continues to push, uh, different boundaries in different ways.

Gavin Purcell: I'm actually doing a talk at Natby this week, and one of the things I'm going to talk about is Natby, for those of you out there who are not familiar, is a, uh, producer, something conference. And I've only gone once, but I was invited to come to it, which is going to be interesting. And the talk I'm going to give is going to be about how small studios can now like create and generate their IP in a way that never could have happened before.

Gavin Purcell: And I kind of think NeuralViz is like the perfect. perfect version of this. Like you can be a person that is one to five people, a studio that's one to five people, and you can spin up things and try things and you don't need a lot of money to try it. And it just feels like to me, that's the future of a lot of video period, not just AI video.

Gavin Purcell: One of the things that we don't hear anymore is, um, Uh, it's just not ready yet. That was something that we've heard a lot of in [00:52:00] early AI video. Yeah. Well, it's cute, but it's, that's not there yet. Um, that I haven't heard that in kind of a while, um, because it is there, you have to work at it and you have to, you have to finagle it and you have to wrestle with it, but it can be done now.

Gavin Purcell: And I think that within, I honestly believe that within this year, uh, by the time we hit calendar year, 2026, we will have seen one major breakout. Couch. AI type series. The thing that I keep thinking of is, um, the South Park short, the original, uh, yeah, I think that we'll see something hit that viral level of, uh, that can be turned into a property, um, sometime this year.

Gavin Purcell: Um, I don't think it'll come out of a studio. I think it'll come out of an independent production. Um, maybe just three dudes in a garage. You never know. So, Well, that's, I mean, that honestly is the whole thing. So, um, I think that's the whole thesis of my talk, is that the three dudes or three ladies or three mixes of dudes or ladies, whatever gender you like to prefer, is the future of kind of [00:53:00] all of this stuff, right?

Gavin Purcell: Because I think that like there will always be this kind of like super high end Hollywood version, which I think like, Once you get the IP to where it's like everybody knows it and they're all aware of it and they can all access it like it's Game of Thrones or it's Marvel, you need maybe a huge studio to do that.

Gavin Purcell: But like, I think the most interesting stuff is going to kind of bubble up. And the example I use, and this is maybe going a little deep on this, but are you familiar with Skibbity Toilet? You know, skibbity toilet. Yeah. That's it. I have kids, I have aware, I've, I've, I have two kids in generation scabity. So, okay.

Gavin Purcell: So, so my kids are older, they're girls, but I wasn't there, but I actually, I am, I am the same, I have my 12 year old boy in my brain. So like when it first came out, I was pretty impressed by it. And then the kind of the speed that it took off is just a really interesting thing. And so the overall thesis here is that like.

Gavin Purcell: These kinds of things, like whether it's neural viz or it's a, you know, a three guys in a garage making something, the distribution part of it's very easy now, right? So if the distribution parts easy and then suddenly the production part gets a [00:54:00] much easier, there's just this opportunity that was never there in the last.

Gavin Purcell: in the entertainment business where the gatekeepers don't really exist any sort of way. Uh, I've, I've, I gave a talk a couple of times in like Hollywood circles and uh, you know, I come back a lot on that whole like, it's gonna, it's gonna ruin all the jobs. All the jobs are gonna go away. It's, you know, there's one way of looking at it or Could be making a whole lot more jobs.

Gavin Purcell: Like, you know what I mean? Like it's, uh, like your giant production that employs 8, 000 people may scale down, but there also might be, there will likely be 800 other smaller productions going on all over the place as well. So, um, you know. That's, uh, it's an optimistic way and I kind of hope to look that way.

Gavin Purcell: Okay. So we, at the end of our show, Tim, we kind of transitioned to stuff we've done with AI this week and you do a lot. I do. You spend a lot of time playing with AI tools. Is there something lately that you think is interesting that you've played with that you wanna kinda shout out? We don't have to go into super detail, but I'm just kind of curious if there's a tool or something that could be interesting that people might be able to try at home.

Gavin Purcell: Uh, for sure. I [00:55:00] think that right. Currently right now, I just covered last week. I covered Refusion. The, uh, Refusion is an AI music generator. You can kind of think of it along the lines of like Udio or Suno. Um, the fun part is that they were kind of the OGs in the, uh, AI music space with, they sort of originated in.

Gavin Purcell: Stable diffusion audio. Um, so back when we first heard it and it was like weird and warping, it sounded like it kind of came off of an AM transistor radio from like a possessed grandma's house in like 1928. Um, yeah, they were the ones that did that. So, um, they kind of went underground, um, built up their model, clearly got, um, some money somewhere along the lines, some backing along the lines.

Gavin Purcell: And now they've kind of resurfaced, um, as, as Riff Fusion. And, uh, yeah, I would say it's definitely in that sort of like, Suno Yudio camp. And the best part is it's currently free. Um, so yeah, they're in your free buffet, open beta. Um, so go sign up and just start generating like just generate and have fun with it.

Gavin Purcell: Is there a [00:56:00] type of music that it does really well versus others? Uh, let's see. So I've run, uh, I ran modern country on it, uh, with a, I don't like modern country, so I gave it the prompt, uh, write a modern country song about how modern country is terrible. Uh, and it did a pretty great job with that. Um, I heard it do, I've had it do metal.

Gavin Purcell: Uh, I've had it do, uh, sort of like your standard pop stuff and some hip hop. And, uh, All of it sounds, you know, very serviceable. The thing that I like about it though, is that if you go into, um, the compose tab, which is kind of the advanced features, uh, you can multi prompt in there so you can put all kinds of weird stuff in there like, um, Middle Eastern music, funk, uh, surf rock, um, you know, and then you have sort of temperature sliders for how much.

Gavin Purcell: You want to have, oh, that's cool. That influence in, into the song so you can kind of blend these genres together. I didn't even really get into it in the video. So this is a hot take for anyone that's made it this far into the podcast is that you can time out when those things happen. So if you had like surf rock, um, you can put that influence at [00:57:00] zero from.

Gavin Purcell: Uh, zero seconds of the song to 30 seconds in the song, and then you just crank it all the way up along with, you know, funk or, um, Jamaican death metal or whatever you want to, you know, whatever you kind of wanted to do. So there's ways that you can sort of blend in and out of, uh, genres. And that's something that I have not seen in any other music generator so far.

Gavin Purcell: That's really cool. Yeah, I'll To take a look at that. I've, I, I, now that you say that, it came about stable diffusion, guys, I remember it from forever ago. For some reason I was thinking it's a new thing, but that's cool. Okay. We'll play a little piece of one of Tim's music's here just to take a, to get a sense of what it sounds like every single night of Tim.

Gavin Purcell: Same old song comes on again. Copy, paste, stand all to play while real hearts just fade away. Alright, Tim, thank you so much for being here. See all of y'all next week. Kevin will be back. Uh, and we'll have a new fun show. Kevin, he doesn't want, he doesn't want to go. He doesn't want to go. I'm challenging Kevin to a duel.

Gavin Purcell: Bye everybody. We'll see you next [00:58:00] week.