AIForHumans93OpenAIStargateDeepseekR1

Kevin Pereira: [00:00:00] AI development continues to ramp up. The 500 billion project Stargate was just announced, and it aims to make cancer a thing of the past. We'll fill you in on what it is, how OpenAI is involved, who's funding it, and you might be surprised to find that the government has put in a total of 0 into this project.

Kevin Pereira: Perhaps more importantly, we will also explore how Elon Musk feels about this. Things Gavin, because spoiler, someone put on their big, mad pants and the competition is not sitting on the sidelines. Anthropic CEO teases some really big stuff coming in 2025 and Google dropped an incredible new model that they named something insane yet again.

Kevin Pereira: Plus deep seek dropped an open source model that in some cases beats open AI's best efforts and you. Can run it right now, for free, locally. That means on your computer, on your cell phone, the future is here. And we've got a rundown of some [00:01:00] of the best new AI tools in media, plus a story about an AI that can literally read your brain.

Kevin Pereira: Like, okay, so the AI would know that I'm It's AI for Humans, Kevin! That's what you were gonna say! It's AI for Humans! You didn't need AI for that. We have a script.

Kevin Pereira: Alright, this is a big week in AI. And the biggest story, as everybody who listens to this show, and probably my mom knows by now, is that we had a big announcement of Project Stargate, Kevin, which sounds like an old sci fi movie because it was! Yes, it was. Also on top of that, This is a 500 billion promised 500 billion investment.

Kevin Pereira: It's a joint venture between OpenAI, Oracle, SoftBank, and I think Abu Dhabi Investment Group, which is one of those things that had been rumored for a while that there was money going to be coming in from the Middle East somewhere, to [00:02:00] essentially create an infrastructure program for AI. And then, you know, we talked a little bit about infrastructure, AI infrastructure.

Kevin Pereira: What this is basically is building out the data centers to run all of these AI models on. Kevin, I, I think this is one of the bigger stories we've seen and obviously made bigger because the president came out and, and kind of took credit for this. And all three of the people that came out and talked about this, including Sam Altman, Larry Ellison, and the CEO of SoftBank all said that this was not possible without our new president.

Kevin Pereira: What are your first initial thoughts? And then let's kind of hear what Sam had to say about this as well. Yeah, I mean, it's no surprise, uh, that Sam Altman is involved with this. I mean, he wanted to, what, raise trillions of dollars in the pursuit of AI? That was a rumor. He didn't care if he would spend that much along the way.

Kevin Pereira: Well, here's, here's half a trillion supposedly secured. Uh, we'll get to some folks that are pointing fingers and saying that's not the case in just a moment, but it is interesting to see, you know, [00:03:00] They all claimed that this couldn't have been possible without our, our, the, the efforts of our new president.

Kevin Pereira: Uh, and the, uh, the timing of the announcement was very, very interesting, but clearly this deal was in the works for months. Uh, potentially years as well. Now, you could say that it was anticipation of a new regime being in charge, but I don't the politics and the timing of it are far less interesting than what this might actually mean for the country, right?

Kevin Pereira: We want to win AI. People are comparing it to the Manhattan Project, or saying when you adjust for inflation, Gavin, that this is a swing, uh, akin to the national highway system. But the difference here is that this isn't a public works project. Just because it was announced in, like, the White House, or just because it was announced in at the White House and the President was there, does not mean our government is actually doing it.

Kevin Pereira: It just means that they are supporting it. Maybe they're getting some regulations out of the way so that they can build [00:04:00] power and build infrastructure to make these supercomputers happen. That's right, and I think that is a big confusing thing about this right now, because you have the President, the new President, announcing this on day two of his administration.

Kevin Pereira: It sounds like this is a government project. Like you said, a lot of people are comparing it to the Manhattan project and it's not. It is a private company. In fact, so much so, if this were really a Manhattan project, you would expect other companies like Meta, Google, Anthropic, all these other big AI companies to be part of this.

Kevin Pereira: If we're all trying to roar off for America and beat everybody, this is a private company that launched, uh, you know, is just launching. And you know, it's been like you said, it's been in the works for a while. The information on a story in the middle of last year, which was about the idea of breaking ground in Abilene, Texas around some sort of data center.

Kevin Pereira: And that is what this is. Just a very quick explainer of why this is important is that the bigger your AI models, the more compute you need, the more the test time compute happens. You're going to need more chips to run those on. These are building giant data [00:05:00] centers where they're going to put a lot more chips.

Kevin Pereira: They're going to have a lot more space to put these chips and therefore run a lot more AI models on them. So it is literally the same in some ways as building the freeway system. You're building larger AI models and places for the AI to run. So that is why this is an important thing. Kev, why don't we hear what Sam Altman had to say about this at the white house?

Kevin Pereira: I'm thrilled we get to do this in the United States of America. I think this will be the most important project of this era. And as Masa said, for AGI to get built here, to create hundreds of thousands of jobs. to create a new industry centered here. Uh, we wouldn't be able to do this without you, Mr.

Kevin Pereira: President. Uh, and I'm thrilled that we get to. I think it'll be, uh, an exciting project. I think we'll be able to do all of the wonderful things that these guys talked about. But the fact that we get to do this in the United States, uh, is, is, uh, I think wonderful. So thank you very much. Thank you very much.

Kevin Pereira: Well, okay. A couple things. One of the, uh, Sam said, all of the exciting things we're going to get to do them that you talked about. One of the [00:06:00] things that Larry Ellison from Oracle, um, Went, went on like a detailed explainer about was his belief that this particular system is going to be capable enough to cure cancer.

Kevin Pereira: And I know that sounds like a catch all here, but what he was talking about was blood draws using AI to do a detailed analysis of the radicals within your bloodstream to, uh, sequence therapeutics to specifically target the cancers which are in your bloodstream and within quote, 48 hours, you would have a personalized vaccine that could be delivered to you that would eradicate the cancer in your body.

Kevin Pereira: And this was Kevin. I know another person that promised this a long time ago. She used to wear a black turtleneck and she'd come out and she'd hold a little thing up. And it was like a real promise of being able to do something super fast. A single prick of blood. That's right. This is not Theranos. We do believe there is some truth here.

Kevin Pereira: There were backrooms at Walgreens that were just splattered in blood. Actually, that On the, the backs of those machines. Yeah. This could be the new ero maybe instead of hearing [00:07:00] cancer that it's just gonna make smoothies out of your blood. We don't know. But I thought it was interesting though that in the, yeah, in the utopic vision, uh, I don't even know if that's a word, but in this, the, in the vision of the utopia, that a lot of the AI power brokers, um, wield that this was the one thing that Larry chose to focus on was the curing of cancer and, and.

Kevin Pereira: We do know that others are marching in that direction with AI. I would, I mean, let's, let's all hope for that, right? Let's all hope for that being a side effect of these massive investments. Well, and there was a really interesting tweet from Ethan Mollick, one of our favorite Twitter follows who talked about the idea that we're rushing to AGI now.

Kevin Pereira: And we don't only have a sense of what it's going to do to both us or to our economy. And I think that's one thing that's a little, you know, Uh, a thing to take pause here on is if we do get AGI and ASI, like, there's a lot of things outside of curing cancer that it's likely to be able to do, one of which is like really change the direction of our economy in a bunch of different ways and like who has access to it.

Kevin Pereira: I think [00:08:00] Larry says, cancer, the cancer thing, because it's the most grounding thing, right? It's the most obvious thing that you could say on this podcast. We've said many times, we have all these really cool things that we'd love to do with AI, and we think AI will be able to do when it gets to AGI or ASI.

Kevin Pereira: But the thing that normal people will understand best are the science advances that come with it. And later on, we're going to talk about how Demis Hassabis from Google Deep Mind is starting a drug company that is going to develop drugs and kick them out very quickly. But Kevin, the other side of this that is interesting to me is that there is some real beef going on now with Elon.

Kevin Pereira: Sam Altman is very good at understanding how to take a moment and kind of milk it for a lot, right? And Elon has had his share of moments over the last couple of weeks. We all know that. There's been a very viral clip of Elon going around that doesn't make him look good. But Elon came out and replied to this announcement from OpenAI and said, basically.

Kevin Pereira: That he doesn't think they have the money. He literally tweeted at them. They don't have the money. Actually have the [00:09:00] money was his tweet with 2 million plus views as of the time of this recording. And then he also said SoftBank has well under 10 billion secured. I have that on good authority. My, the funniest thing about this is that Sam then replied.

Kevin Pereira: So this is really public beef that we're seeing kind of play out in the world. And again, just an ounce of context there, Elon Musk, the founder of X AI. I know we think of them as Mr. Twitter or X or SpaceX or Neuralink or Tesla, but he very much is in the AI game. He very much is. We're trying to recruit the same talent.

Kevin Pereira: He's trying to raise billions of dollars and build his own infrastructure as well. So. Just keep that in mind as Gavin continues here. Yeah. So Sam basically replied at first said, I genuinely respect your accomplishments and think you are the most important, uh, entrepreneur of our time. Important to do.

Kevin Pereira: You got to do a nice little like, Hey, I think you're pretty cool. And then following up later on the tweet where he talks about SoftBank having well under 10 billion secured, Sam then directly says, wrong, as you surely know, [00:10:00] want to come visit the first site underway. This is great for the country. I realize what is great for the country isn't always what's optimal for your companies, but your new role, I hope you put American flag emoji first.

Kevin Pereira: So we are looking at the billionaires battling with one another across dumb little tweets now. Amazing. Elon, the money is secured musk. I mean, it's rich coming from him, someone who famously tweeted about having funding secured to bring Tesla private and I think is still dangling over hot water for that tweet from years ago.

Kevin Pereira: So, uh, a little rich that he's saying that. Fantastic that Sam is clapping back and even, you know, poor Microsoft in all of this. Don't say poor Microsoft. I have to say poor Microsoft. They're a multi billion, near trillion dollar company. There's no poor Microsoft in here. No, poor Microsoft. They wanted to be OpenAI's bestie.

Kevin Pereira: They were picking out the perfect nail polish for their big, big dates. And then, you know, Satya's on the sideline. In fact, he was on a [00:11:00] CNBC interview where he was asked about all this. And here's what he had to say about the big SoftBank investment. You're investing 80 billion already. Um, how much money do you ultimately put into this?

Kevin Pereira: So this is where, you know, Microsoft is investing 80 billion in capital each year and this year we're investing. I'm not particularly in the details of what they're investing. Okay. And so, but when you look at that 100 billion commitment, 500 billion dollars potential. I don't know if you saw Elon Musk took to Twitter, um, and said they don't have the money.

Kevin Pereira: The money doesn't exist. As if this is not really going to happen. What do you think of that? Look, all I know is I'm good for my 80 billion. I'm going to spend 80 billion building out Azure. My paper's good. My paper's clean. I got the chips here. I'm not going to the house for more money. It's unbelievable to me that we're talking about these sorts of numbers.

Kevin Pereira: And I know that people who listen to our show are of all [00:12:00] variety of types, right? Most of the billionaires though. Yeah, Kevin, I are not like any sort of like independently wealthy person where, you know, we work for a living, we do all the normal stuff. But when you hear these numbers thrown around, it just sounds so insane.

Kevin Pereira: But it does give you a sense of the kind of game that's being played right now. And clearly that the stakes right. And I think again, we've said this on the show before, there's so many people out there that talk about the hype of AI. And they talk about how this is all just a big giant bubble. But Well, we've now entered what year two of this bubble and the amount of money is not slowing down a fit.

Kevin Pereira: In fact, if anything, it's scaling up. I keep coming back to the electricity metaphor and I think that's something to think about is that like in a lot of ways, intelligence is like when we first got electricity, right? In fact, there's a great book called Power and Prediction, the Disruptive Economics of Artificial Intelligence by a EJ Argywal, Joshua Gans, and Avi.

Kevin Pereira: Uh, gold farm and what that basically talks about is this idea that electricity was this kind of thing that came out, took a [00:13:00] while to kind of populate with everybody, completely transformed the world and there were winners and losers, but the electricity was the winner. Do you know what I mean? The idea that electricity suddenly changed everybody's lives, made their lives better, made things possible.

Kevin Pereira: We never had before. And I think ultimately that's going to be the thing we have to remember is that the intelligence is going to win. And we're going to get into this later in the show when we talk about DeepSeek's new model and a bunch of other stuff that's happening. All of this money that's pouring into this stuff, I don't know if I buy the idea that these companies are going to be worth, I don't know, 10 trillion to justify something like this.

Kevin Pereira: Intelligence is going to be the winner, I feel like. And the intelligence community is going to be the winner because, as we know, Oracle famously in bed with the CIA, Gavin. Why is Larry Ellison up there? Why is he in charge of these data centers? And why are there former NSA members on the OpenAI board?

Kevin Pereira: Is this why everybody was leaving OpenAI last year, Gavin? Is it they I've got my board back here, Kevin. I've got my board. [00:14:00] I'm moving my p jars out of the way and I'm taking my tinfoil hat. I'm setting it aside. We are on to something. We don't need to talk about PGRs again, because I think we had some comments about the PMIST conversation we had in the last podcast that we don't need to repeat.

Kevin Pereira: We don't need to repeat. Okay. Really quickly. A couple. We need to pander to our new community, Gavin. We need to go. We've tapped into a really new followers that are PMIST fans. Sorry, everybody. Okay. Two other quick things that are semi attached to this. One. Um, Trump also repealed the AI regulations that Biden put in.

Kevin Pereira: So this just goes hand in hand because it means that like, you know, Trump is like basically opening the door and letting people come in. I think most people kind of see this as a positive in the world. Obviously the AI safety people, I think are very concerned about this. The other side of this, Kevin, is that there's news that came out of the information just this morning.

Kevin Pereira: That open AI looks like it's going to be shipping their operator, uh, this week, which we've talked about operator. This is their kind of like minor AI agent that can [00:15:00] interact with the internet. So, you know, open AI is they're doing a bajillion things at once. I am really curious to see like how this company that started of like nine people, whatever, endow is a thousand plus people, how the company can scale to these sorts of ambitions, I think is going to be really interesting.

Kevin Pereira: Yeah. And the operator thing, whether it comes out this week or next, uh, I hope we, it doesn't come out right after release our podcast. I know that's selfish, but we'll be interesting to see what happens and how people react when suddenly their cursor starts moving autonomously. Not that there aren't apps that can do that now, but a major, major one from like open AI, this is slightly different than what Anthropic was doing with their computer control, their MCP system.

Kevin Pereira: This is like directly read your screen. Manipulate your, rumored, read your screen, manipulate your cursor, click on things, you know, take action on your behalf, fill out forms, like, this is hands off and let it ride, I, I don't even know if I'm ready for [00:16:00] that, I'm definitely gonna experiment with it, but I just, there's something that feels so weird about that to me.

Kevin Pereira: Well, you know, Kevin, I heard a beta test of operator. The first thing that it's going to be able to do is subscribe to the AI for humans, YouTube channel, which is pretty cool. Not a bot, not a bot, everybody. This is a human, uh, says to go do this, but you can go subscribe to the AI for humans, YouTube channel.

Kevin Pereira: It will also Kevin fill out. A AI for humans podcast review, which we have not had for a while. And I know people, our podcast numbers are growing, are going up and growing, which is amazing. Every week we get more, which is really cool, but I really would love to see some more five star reviews on our Apple podcasts.

Kevin Pereira: We love those. We, we really means a lot to us when we see them, obviously it also helps and affects the algorithm on the podcast side. But please subscribe, we also have a Patreon if you want to throw some bucks in a tip jar for us. All of the stuff that you do to help us make the show better keeps us coming back every week and we really appreciate that.

Kevin Pereira: Yeah, thank you. And to Gavin's point, don't wait for the bot to subscribe for [00:17:00] you. Go ahead and click it yourself today. It costs you nothing but it does mean the world to us. So I echo that. Thank you. Okay, the next big thing I want to talk about is Dario Modi, the CEO of Anthropic, was at the Davos Forum, which is where all the rich people in the world gather to hear about what's happening for rich people's lives.

Kevin Pereira: But Dario actually dropped some interesting stuff about where Anthropic is heading this year. Let's take a listen to that. I think until about three to six months ago, I had substantial uncertainty about it. I still do now. But that uncertainty is greatly reduced. Um, I think that over the next two or three years, I am relatively confident that we are indeed going to see models that show up in the workplace, that consumers use, that are, yes, assistance to humans, but are, are gradually get better than us at, at, at almost everything.

Kevin Pereira: Um, and the positive consequences are going to be great. The negative consequences, you know, we, we, we also will have to, we also will have to watch out for. Yeah, I mean, that was on the heels that was on the [00:18:00] heels of talking about how hyped a lot of this AI tends to be and how Dario traditionally has taken a more measured approach, almost with this coin flip mentality of like, Oh, we could totally get it, but we might not.

Kevin Pereira: We might not have the tools we need. And the beginning of that quote started out with him now being, um, more bullish about having line of sight on solving actual intelligence. And this seems to be the refrain from many, many of the leaders of the largest companies. Yeah. And I think the important takeaway here is that Dario's company, Anthropic really is seen as like one of the most Restraint, uh, restrained of the AI companies, right?

Kevin Pereira: They're very careful as they go along. Now they might get their hand might be getting played a little bit right now because everybody else is pushing so hard for it. In fact, I think there was an announcement that they just took another billion dollars from Google and funding. So they're kind of building up a word chest themselves.

Kevin Pereira: Uh, which is funny that Google is a big funder of Anthropic, but it is, they added onto to Anthropic as well. Well, Amazon is one of the biggest backers of Anthropic. As well. Yeah. And is giving up their [00:19:00] new training chips, uh, in mass for anthropic models. So it's interesting that, that yeah, that Google and.

Kevin Pereira: Amazon are kind of together on this one. Yeah. And, you know, I really recommend everybody goes back and reads Dario Modi's essay, machines of love and grace. It came out the end of last year. We made a video about it, but like, it is a really interesting look at the pluses and minuses of what an AGI slash ASI universe looks like.

Kevin Pereira: And speaking to like what Larry Ellison was talking about with cancer solving. You know, Daria goes in detail about the science advancements that can come out of that. But the thing that keeps coming back in my brain as we talk about all this, Kevin, is like people are so weird, right? In general, when I say people, I mean like humanity, humanity is so weird.

Kevin Pereira: And we're entering this time where like, The chaos of what this could bring to our lives is kind of unprecedented. And I think just a lot of smart people are out there talking about how that might affect us going forward. And I do think that all the times we hear [00:20:00] about this stuff, it's important to kind of remember.

Kevin Pereira: Learn, try and do like all of this stuff. Know where it's going, because if you don't, you're going to kind of wake up at a couple of years from now and your computer is going to make not only your breakfast, but then like do the first three hours of your job for you. And you'll be like, what happened?

Kevin Pereira: Where did I where did I go wrong? What happened to me? Exactly. You're going to wake up in bed like that night. You had one too many at the bar and you're going to roll over and there's going to be a robot. Right there next to you vaping. And you're going, what is going on? And you're gonna get up. And then two more robots are gonna come in the room.

Kevin Pereira: Hair of the dog hair. And it's just, it's gonna be a fever dream. Wait, is this like a robot, uh, influencer house? Like a bunch of you and a bunch of robots are in a bro house, like kind of making concept? That'll wait. You'll be at 1600 Vine. You won't know what happened, . There's gonna be half a bottle of Corbell on the pillow next to you and a bunch of Rubics telling you, I think you need to leave.

Kevin Pereira: You look in the mirror and they've written like dumb human on your forehead and they've taken videos of you. [00:21:00] Oh god, I hope we don't get to that. So anyway, uh, Anthropic is, is obviously pushing forward. Davos is where all these people are talking about it. I will say he threw a little bit of shade towards OpenAI.

Kevin Pereira: Take a listen to this. Let's do it. Everybody loves the shade. Poor Sam Altman. Here we go. I think progress really is as, as, as fast as people think it is. Um, one thing that I will criticize. is I actually think it's very important now that fast progress is relatively likely, um, to, to, uh, to appreciate it with the, with the proper gravity and to talk seriously about it.

Kevin Pereira: Um, some of the other companies, I won't name any names, um, you know, there's just all these, you know, it's all these weird Twitter rumors like employees talk, you know, employees like, you know, have this, have this kind of sly winking like, you know, um, nod to like, Oh, there's these amazing things we're doing here.

Kevin Pereira: Um, I actually think that's dangerous because someone on the outside looking at it is like, Oh man, that's just [00:22:00] hype. That kind of communication gives the impression that this stuff is not serious. No, Dario, that kind of communication gives us click baity headlines for our YouTube videos. We need that.

Kevin Pereira: That is our essence. That's our mana. I was going to say, Dario's like the adult here who's kind of boring. He's definitely not at our hype house. We're not inviting Dario over because his views are going to go down. The robots are going to be like, get him out. We don't need him here. But I mean, he's right, right?

Kevin Pereira: Because the open AI world is like, they're not really being as direct in their communications when they talk about this stuff. I mean, we followed all the strawberry rumors, all those things. We are getting to a place now though, where they are being a little bit more direct, but Dario, I think is. that they need to start taking it a little bit more seriously.

Kevin Pereira: So we will see how that happens going forward. Well, Sam Altman did recently tweet too, that the hype is getting out of control. And OpenAI is not working on, they're not going to announce AGI next week behind closed doors, this, that, that, so I, I feel like there is. a nice little push and pull, but also like, [00:23:00] you know, Dario, the clip we did just play before that is Dario coming around to the fact that, oh, he's way more confident now that there's going to be quick takeoff that we, we will see, uh, AGI and ASI, you know, uh, it could happen.

Kevin Pereira: It's like, well, That's what they were hyping a year ago. So maybe they were just a little bit more comfortable to talk about it. But I don't know. I'm here for the, I love it. Let's bring the hype. I love the hype. Let's do it. Why not? What about China, Kevin? What are your thoughts on China? I am not able to answer that at this moment.

Kevin Pereira: Can we please have a discussion about anything else? Which is kind of the response that DeepSeek would give you. Is that where we're headed? DeepSeek R1? That's where we're headed. That's where we're headed. Like, tell us what, so DeepSeek R1 is a brand new model for DeepSeek. from, uh, the Chinese company DeepSeek that just came out.

Kevin Pereira: Tell us what it's all about. This is a reasoning model. Um, you know, OpenAI has their, uh, O series of models like the O1 and the recently announced O3, where it reasons and thinks through problems to arrive to an answer. [00:24:00] Well, this is an open source, freely available alternative to that, that was trained by a Chinese company that you can run for free.

Kevin Pereira: On your laptop right now, there's distilled versions of it, smaller, less powerful versions of it, but you can for fractions of the cost of open AI run the full featured version of this in the cloud. And what's interesting about this is that this model was trained, uh, not through, uh, there's reinforcement learning from human feedback, which requires flesh bags to be in the loop and say this good, this bad, uh, I'm going to make this model better.

Kevin Pereira: This was just pure RL, just pure reinforcement learning. The same way, uh, AlphaGo was trained, uh, a while ago. Um, what that means is you don't sit down and say, Hey, here's, um, like when the AlphaGo example, here are a thousand, uh, games of, of Go that are winners. So learn what these games did and then, and then reverse engineer how to beat the game.

Kevin Pereira: Instead, they say, you go ahead and play the game, play it a bazillion times. [00:25:00] Just the win condition. Is is what you want. So go figure out how to get that reward. Basically, this is a gross distillation of the way this works, but basically that's how this model was trained is that it was just told to like, go and figure it out and arrive to the wind condition, which is the correct answer to the math problem.

Kevin Pereira: problem, uh, solving the logic puzzle, just go figure it out. And the model did, and now that, uh, the chain of thought that deep seek has can also be extracted. And so people are taking the reasoning steps and using that. To train other models, so now like dumb models from a year ago, open source models, even Metasllama, which is far from a dumb model, but people are taking the reasoning steps, uh, and, and applying them to other models and making them smarter.

Kevin Pereira: What? What did I even just say? So, so the interest, there's two things that are interesting about this. So Kevin pointed out, this is dirt cheap and it is free to use, right? Then it's comes from China, which we have to [00:26:00] be clear. There are some big problems with that. In fact, it won't answer questions about Tiananmen Square.

Kevin Pereira: There are things that are hard coded into this data because it comes from China that are a problem. The other thing that I think is really interesting about this is that it points to the idea of the reasoning advancements in which why the O models are moving so fast and perhaps why, again, going back to what Ilya Sutskever saw way back when, This might have been the unlock the second unlock after the transformer was first figured out right if the first if the transformer was the kind of Initial breakthrough on these AI models in neural learning This reasoning step and how this stuff works feels like it is the next version of that and as to your point like If you can use this data and use it to train other models, that is how you scale insanely fast.

Kevin Pereira: And now we have a version of this that's running super cheap, running locally. I know you had pointed out that somebody was running this on a phone, I think at some point too, right? So like, it is a pretty remarkable thing. And just to [00:27:00] know benchmark wise. The actual R1 model is benchmarking around 01 and sometimes better.

Kevin Pereira: And I know that there's a smaller model. I saw a tweet about this, a smaller model. Um, in fact, I think it's 7 billion model that is benchmarking at GPT 4. 0, which is pretty incredible locally on a device. So that's a pretty remarkable thing to see. Yeah. And again, because it's open source and the, the, the model, the weights, everything is available.

Kevin Pereira: Um, you're now seeing like a lot of. Uh, DeepSeek R1 released and then less than 24 hours later, we've got it distilled and running on a cell phone. And then less than 24 hours later, Oh, we gave it access to the web. They can now crawl and search and get real time information in 24 hours from now.

Kevin Pereira: Someone's going to have it running on their smart toaster. Like the, the ability to take these open source things and iterate on them and develop with them and make them better. is unprecedented. And this keeps the trend alive. The one that many have identified where closed source, these companies with their bazillions [00:28:00] of dollars, they have a competitive advantage that seems to kind of die out around three to six months later when the open source community is gone.

Kevin Pereira: Catches up. And this is very important for the democratization of all this intelligence. So that one particular company or government or set of bad actors doesn't have complete control, uh, over the intelligence ecosystem, uh, uh, us. Us folks have a fighting chance and also it will trickle down so that you don't just need the latest and greatest cell phone or a 10, 000 supercomputer on your coffee table.

Kevin Pereira: You'll be able to run these on consumer devices like that. Yeah, I mean, that's incredibly important. That's an amazing thing. And also you can have a device that's just for you so you don't have to share your information with other people, which is also great. Another huge advance that happened this week, which I think kind of went under the radar, but then I got a chance to play with it and I was kind of impressed.

Kevin Pereira: Google has updated their new Gemini model. Now, Kevin, I want to say this all, I'm going to read this directly. So I have a clear understanding of what the name of this is. We've kind of joked about [00:29:00] this for a long time. A shout out to the people at Google who make amazing products, but don't know how to name them.

Kevin Pereira: This is called. Gemini 2. 0 Flash Thinking Experimental 0121, meaning that is the date it came out. This is a brand new, uh, model of Gemini 2. 0 Flash Thinking, and you have to access it through Google AI Studio. Um, shout out to Logan Kilpatrick, who works over at Google, that's always sharing this stuff. I went into this and, you know, first of all, I had to find the dropdown menu and figure out where it was, played around with it.

Kevin Pereira: And I have to tell you, Kevin, I was. pretty impressed right away with it. I, you know, the, one of the big things they talk about is how much better it is at math, how much better it is at all the sort of problem solving things that all of these models are getting better and better at. But of course, as I do, I threw my very important, uh, benchmark at it, which is creating a March madness bracket tournament of the world's most dangerous animals and wanted to see how it would do.

Kevin Pereira: And I actually threw it a couple of little different things this time. [00:30:00] One, uh, normally I'll just say, like, give me the March Madness bracket, make them all fight each other and tell me who wins and kind of walk through the steps. This time I said to it, I was like, you know, assign each of these animals a stat system and then also assign some sort of luck rule based on each thing.

Kevin Pereira: I did it in, in text. It was fantastic. The polar bear ones, everybody knows that makes a lot of sense and I did see a couple small things. So I went back and said, Hey, the. It feels like there might be, the system might be a little bit off. It recognized that it tweaked a couple of the stats. It also tweaked the luck situation and then gave me a new rundown of who won what really incredible.

Kevin Pereira: And then I said, Hey, can you kick me out the code of this so that I could like maybe find a way to make this something that I can share with people and try. And I got a usable code. Now I kick that over to you. And there were a few issues with it, but like, This was remarkable. Like I feel like this new Google model is very good and people might be sleeping on it because it's just not getting as [00:31:00] much hype.

Kevin Pereira: Yeah. Um, head over to the Google AI studio. It's free to experiment with right now. Um, some interesting things like, so it passed your vibe check, which is always important, but from a technical standpoint, this thing has a million token limit, just North of a million tokens. That means, yeah, I mean, grab all of the Harry Potter novels and copy and paste.

Kevin Pereira: them a few times and then jam them into the system and it will remember all that and keep it within context. So you can do detailed analysis across massive walls of text and data. That's really cool. Code execution is built right in so you can actually run and generate code in a sandboxed sort of safe environment right from within their website.

Kevin Pereira: Like this is a. Just a stealthy little release because I, I haven't seen it trumpeted anywhere. I didn't see any massive videos, uh, or splashy anythings, but it's out and you can mess with it now, uh, for free over at the AI studio. So kudos to Google. But again, yeah, Gemini 2. [00:32:00] 0 flash thinking experimental.

Kevin Pereira: Oh, one 21. Oh, one 21 is the important part because there's another model with that exact sure you get the model name right, which is a very weird thing to me. Look, uh, the, the distillation of all this stuff is the line keeps going up and the costs keep dropping. And to the, uh, electricity analogy from earlier, yes, intelligence is going to be, uh, commoditized.

Kevin Pereira: And, um, I hope we all have, uh, free and instant access to it all. And this, the, the trend continues. That's right. And Kevin, in that vein, it is time to generate a new theme song. Oh, sorry. Sorry. Is, is that, are you, is that water on your shirt? Yeah, I just spilled it myself. You can leave that in, Will. Reverse the podcast and you'll see me take a drink of water while Kevin is talking.

Kevin Pereira: And it's like, At the time, Kevin, I tried to pretend that nothing happened. I'm sorry. Wiped it off my face. But no, it's alright. It's [00:33:00] okay. Alright, speaking of all this crazy change, Kevin, there's actually some stories this week that kind of blew me away, blew you away. So we're introducing a new segment called Mind blower.

Kevin Pereira: So this story is crazy. It is a mind blower. I have not been more blown away by something in AI for awhile. This is a paper right now, but we assume that this is coming from a reliable source. What this is, is the idea that AI. Can actually predict five seconds in the future looking at your brainwaves, which feels bonkers.

Kevin Pereira: Feels like super insane to me. Did you get a chance to look at this? And what is your thoughts? Do you feel like we're looking at something that looks like it'll fundamentally change human life? Or are we looking at something that like, who knows if this is even real? It's a cool party trick, and I can't wait for the TSA to implement it.

Kevin Pereira: Um, this uses, uh, [00:34:00] fMRI, that's Functional Magnetic Resonance Imagery, which if you've ever seen studies that people put on the caps that have like the electrodes going to them basically and it looks for blood flow changes in your brain which light up and so as you're thinking about certain topics or you're about to move certain limbs, you know, there's motor skill portions of the brain, there's different sections that light up and what this study was doing is basically feeding that data.

Kevin Pereira: To a transformer model and saying, okay, when this area lights up in a certain way, What's going to happen five seconds from now? And it fairly accurately predicted which portions of the brain would light up. And that is, uh, it's certainly interesting, right? Certainly, it certainly points to a completely different idea of how the universe might work, or at least our sense of how it is now.

Kevin Pereira: Granted, There have been scientific studies and I've read a bunch of like pop science books where people talk about like brain function happens before before, you know, motion happens in the body and you decide things [00:35:00] earlier, but this really is crazy like and this is like the very beginning stages of it.

Kevin Pereira: But imagine a world where this is real and what we're looking at here is Is some sort of, I don't even know what to say Kevin, this is like connective tissue to I guess the idea that you could predict behaviors before they happen. This is really minority report stuff when you think about it. Now granted, five seconds is not much time.

Kevin Pereira: Like what are you going to be able to do? Stop yourself from stubbing your toe. That might be useful or stop yourself from dropping something. Maybe don't drink that jar of pee this time. Jokes. Maybe it would stop me from putting my foot in my mouth. The handful of times that I have throughout my life, like just five seconds would be enough of a buffer, but also this is five seconds just with FMRI, which, you know, that's just one thing.

Kevin Pereira: What about EEGs or MEGs or. Getting, you know, the, the neural link kind of thing to get direct recordings of other neurons. Like there, there's a bunch more data that could go into a model like [00:36:00] this. And maybe we could predict 10 seconds or 15 seconds. Maybe we can stop you from taking that sip of water, Gavin, down your shirt on the podcast.

Kevin Pereira: Maybe it'll be like Adam Sandler's click and we'll get our own little remote controls and we'll be able to reverse our lives, Kevin. That's what I'm looking forward to. Shout out to click. What a hell of a movie. Let's let's do it. Thank you, click heads for tuning in. As always, we appreciate the support.

Kevin Pereira: We have a huge, huge community out there. All right, Kev, the other big mind blower, one of the other big mind blowers was, as we mentioned earlier, Demis Hassabis is expecting their new AI drug spinoff to start launching AI designed drug trials later this year. So what is this and why is this big? And why is it mind blowing?

Kevin Pereira: Well, first of all, we've talked about, uh, Google DeepMind and their progress on science. In fact, Demis Hassabis, won the Nobel Prize this year for science advancements, uh, with Google DeepMind. What Google DeepMind has been focused on outside of, you know, Gemini advanced, fast thinking, blah, blah, blah.

Kevin Pereira: They're really focused on science progress. And [00:37:00] one of the things they've talked about forever is how can we develop drugs that can treat people faster and make those drugs faster. So we all know during the pandemic, they kind of pushed through the mRNA vaccine. vaccines. You know, you have your own feelings, I'm sure about them.

Kevin Pereira: But the idea here is that you could create a drug and have it ready to try way faster due to AI. And I think that's a really significant thing that we have to keep our eye on, especially as we talked at the top of the show about people attempting to cure cancer in the future with AI. This is a company, a very, very well funded company that is going to focus directly on generating new drugs to help people.

Kevin Pereira: Now, I have a not at all well funded side project where I'm using AI to generate new drugs, Gavin, and I have asked Uh, I have asked DeepThink now. Is this Silk Road 2 Electric Boogaloo? Is that what you're doing? Yeah, I'm trying to, what I want is something that is like the fun, breathy, like, Oh, it's just cool to like, you know, dance around of like an MDMA, but I don't want down.

Kevin Pereira: I also want it to be like price point where I want [00:38:00] it to be growable, like a weed. You know what I mean? Uh, and so like, I'm, I'm, I'm putting the pieces together to use AI to synthesize drugs as well, but no one seems to want to fund my research. Well, there might be a guy that I could connect you with somewhere.

Kevin Pereira: He's probably, uh, in Abu Dhabi at some point. Some point, but we'll see what we can do to get you some money. . Anyway, this is a really cool thing. Another one of those things that's like from a science standpoint feels like a real mindblower. And finally, our last Mindblower is Huon 3D 2.0. And Kev, you and I have talked a lot about, uh, AI to 3D models in the past, and this is a brand new 3D model generator.

Kevin Pereira: That not only allows you to generate assets, but we'll also work on animating them. Have you a chance to check this out? I keep trying to run the hugging face demo and it has just been runtime error after error, I think due to popularity, but I tried cloning the space, but no, the examples look. Incredible that you can text prompt or use an existing image [00:39:00] and it will just a 2d image, a single one.

Kevin Pereira: It can remove the background. It will try to interpret what the full 3d view of the model is, texture it and spit it back out. And it, it looks high quality. We've been covering this stuff for like the last year and a half or whatever. And, and. We're getting there, but it sort of looks like a half baked Play Doh something where it's melting on itself or whatever.

Kevin Pereira: But the, the examples of this 2. 0, uh, just it's, it's actually two models together, there's one that generates the shape and there's another that creates the high quality texture. And they've mashed that peanut butter and jelly together into this 3d 2. 0 product. And I can only see the examples.

Kevin Pereira: Unfortunately, every time I've tried to run it, it is not happening, but. They're releasing it though. They're going to release a comfy UI version of it. So you'll be able to run this stuff again for free and locally on your computer. And the quality of their samples look absolutely incredible. So I really, really want to get my hands on it.

Kevin Pereira: Yeah. And then this is 10 cent, right? So it's a real big company making this. [00:40:00] This is not going to be like, you know, vaporware. This is a company that's coming out. And it just, again, it's one of those things that when you see it, It's one of those things when you see it, and you see the models that it's exporting, and again, granted we haven't done hands on with it, but you just realize this is going to change a significant amount of what we do, especially when we create digital assets for video games.

Kevin Pereira: So it's another really cool thing. Those are our mind blowers this week, so we'll see if we bring the mind blowers back. MIND BLOWERS! Alright, now we're stepping right into our next segment, which is our favorite. These are some of the fun things we've seen people out there do with AI this week. It's AI, see what you did there.

Kevin Pereira: Sometimes you're scrolling without a care, then suddenly you stop.

Kevin Pereira: All right, Kev. Uh, we have a new robot hype video. I am so excited for this video. I've never felt more [00:41:00] excited for a robot video in my life. Have you seen this yet? Uh, I, well, I've seen it, but I've also heard it because the soundtrack Yes. Makes me think of just like, yeah, you know, a naughty robot. I don't wanna do my homework, mom.

Kevin Pereira: I'm gonna go out with the cool kids. It's very female. That's it is a stunt video. This reminds like if robots could make skate videos. This is what they would be making. This would be on the back of the VHS because it's got like a robot bouncing on one leg and doing a skanky front flip. This is just, if you're not watching the YouTube video and you're just listening, please go to the show notes for this because you have to understand what we're looking at here versus say even a year ago.

Kevin Pereira: This is a four legged wheeled robot that is over multiple terrains, but not only that, is doing like one legged and single rolls over multiple terrains and can do all sorts of crazy stuff. Now, obviously, as we always see this, we say these are the things that are going to kill us in the future, but the beginning of this, in fact, the end of it is the robot.

Kevin Pereira: Or take care of [00:42:00] us. They might be caregivers. The robot at the end of it has, uh, has fireworks on either side of itself. It looks like they're making it crazy, but this is just another, uh, Insane look at where Chinese robotics is at right now. And, you know, there's a lot of people in the world that are talking about how China's real lead is in physical robotics.

Kevin Pereira: And you're getting a sense looking at this video, it looks like this is not that far away from doing real work. Like you can see a world where like, this is the kind of like robot companion that would follow you around and do stuff with, cause it's that capable. So go watch this video. I was shocked by how amazing it is and it's worth checking out.

Kevin Pereira: Uh, up next, supposedly, uh, HeyLuo just changed the voice synthesis game, Gavin. This is their T2A01 HD model, which I know sounds like a lot of gobbledygook, it sounds like a robot handshaking with another one, but wait till you hear this multi emotion expressiveness. Voice bears the weight of our emotions.[00:43:00]

Kevin Pereira: Everything feels so heavy. This is the best day of my life. I'm done putting up with this. What if it all goes wrong? And a voice has the power to connect us all. Then they do the same voices, but in different languages, which is impressive. But those first few samples of like, maybe you could feel or interpret happiness or sadness or defiance and anger within a text to speech voice.

Kevin Pereira: That's the promise of this system. Yeah, I mean, it's very cool. We follow a lot of voice stuff. And obviously there's companies like Eleven Labs that are doing voice specific stuff and some really interesting stuff from PlayHT. What do you think is interesting about this separating from those other companies and the things that are out there?

Kevin Pereira: It's that I want the emotion out of it. Um, right now, like 11 labs is really good at like cloning a sort of baseline performance of a voice, but you have very little control and flexibility to get it to be like, look, yeah, you can put six or seven exclamation points or a bunch [00:44:00] of ellipses on something, but then the model may hallucinate or go, go out.

Kevin Pereira: So having granular control over the emotion of a performance is important for the next phase of, of these systems. So it makes me think about what a new sort of like script might look like for AI voice systems, right? Because one thing I think a lot about when you're talking about 11 labs is like, I'll do some dumb thing.

Kevin Pereira: Like I did that Mayo commercial a couple weeks ago. You really have to generate a bunch and kind of hope that you're going to get the right thing. Things. What would be really interesting is a scripting system that would allow you to kind of embed emotion in a line or a word. So imagine in within carrots or something, you would be able to put a, a, a, a slug or something that would allow you to make sure that they hammered that thing home.

Kevin Pereira: Like it's an interesting thing of like, you know, when you see a movie script, sometimes they'll put a little bit of definition of what, how they say something. Not that often. 'cause it's often like interpreted by the actor. But in this instance, the AI has no idea what you're trying to do emotionally.

Kevin Pereira: Most of the time, this process with Helio might be able to open the door to [00:45:00] some of that. But like, I think it would be great for us to find a voice model that allows you to like literally tell it how to say something in some form or another. Well, that's the promise of this. So, uh, this was by a post by AI and design said that they cloned, uh, AI girl, Desiree's voice, and then asked her to say a few things with different emotions.

Kevin Pereira: And by the way, this system can clone a voice with like just a few seconds. of audio. So, so here is a clone voice going through some different emotions. Babes, I'm so disgusted right now. Don't even get me started. Are you kidding me? Oh my god, you guys. I am so excited. This is the best thing that has ever happened.

Kevin Pereira: Woohoo! Oh my god, I can't believe this is happening. I think I am going to cry. Help! Help! The aliens are coming. What are we going to do? I am really scared, you guys. I don't think that's that bad. It's not, like, amazing. That's pretty bad. Come on. That's pretty bad. That's pretty bad. Let's be clear. That's pretty bad.

Kevin Pereira: I don't think it's that bad. I'm so scared, you guys. That, like, [00:46:00] wasn't scary at all. Well, the Oh, I thought the scared one was scar I mean, like, if you have problems with the script, take it up with them writing the dialogue. the script. I have problems with the fact that that voice didn't sound scared. Play the scared one again.

Kevin Pereira: Help. The aliens are coming. What are we going to do? I am really scared, you guys. The, I am really scared, you guys, was not that bad. It was a choice. It wasn't the best read, but it was a choice. I, I, I, listen, I'm all, I'm going to play an angry sample now. Cause I'm angry with you right now. I could spontaneously combust.

Kevin Pereira: What in the world just happened now? That didn't sound scared. That sounded different. Play that last line for me one more time, the scared line. What I want to do, I'm actually curious to try advanced voice real quick while we have this. I just want to see what it does. Help, the aliens are coming. What are we going to do?

Kevin Pereira: Okay, okay. I am really scared, you guys. Okay, I got it. Okay. Hey, I have a quick role for you to play. I want you to say as frightened and scared as you [00:47:00] can, These words, help, help, the aliens are coming. What are we going to do? I'm really scared, you guys. Say that in a very frightened, scared, uh, way. Help, the aliens are coming.

Kevin Pereira: What are we going to do? I'm really scared. You guys better. Way better. Yeah. Way better. Yeah. So clearly open AI has some, some advancements there. That's sure. But this was a voice clone, right? So we don't know what the voice, the model was going into this. Did it just have five seconds of audio? And I had to interpret that.

Kevin Pereira: That was a voice clone. And, uh, yeah, it was way better. I mean, opening eyes of pizza is the best. It's just way better. It's so cool. It's still cool. It's still cool. I think it's cool that these things exist. We're always moving forward on stuff. Kev, the other thing I want to shout out here is the Dorr brothers, who you all may know from the show, They're very famous, uh, videos with politicians going into the Eyes Wide Shut party or the politicians on the CCTV [00:48:00] that are, uh, ripping off bodegas.

Kevin Pereira: These guys made a eight minute kind of TV show, which is about people kind of walking into a, uh, giant Stargate, weirdly enough. And it's pretty compelling. I mean, it's like, I think the tricky thing with AI videos sometimes when you get to the longer stages is. It's really storytelling, right? It's a really, it's a storytelling adventure, but I suggest everybody goes and watches this.

Kevin Pereira: I think this is like a good example of how you can string together these tools and make a compelling narrative. I don't think it's like the best thing in the world. I've seen people, we've touched out neural viz before, do some really kind of more interesting, longer stuff, but. What's cool is just seeing people who are at the top of their game technically put all their stuff together in a certain way.

Kevin Pereira: So I thought this is worth watching. You can go check it out right now, um, at least on Twitter and I think on YouTube as well. Very cool. Now, uh, the other thing you should check out, by the way, is, uh, some would say, An art form that is slipping away, but I would say, uh, yeah, [00:49:00] exactly. If it's writing a thing, uh, I would say the momentum is only growing and mounting.

Kevin Pereira: And that's the AI for humans newsletter. You should all go check it out every Tuesday morning. Uh, we drop the choicest of information. Uh, it's always fun links. Uh, fun observations, clips from the show, uh, it's free, it's easy to sign up for, you can go to AI4Humans. show, toss in your email, we will not spam you, you're just gonna get something from us every Tuesday morning, and, um, and maybe you'll love it.

Kevin Pereira: Maybe, Gavin. Maybe, and this week actually, if you got the AI for Humans newsletter, you got an early look at what I did with AI this week, which was play around with Haleo, they're the company that makes this, also makes the mini max AI video model. And I played around with their character reference tool.

Kevin Pereira: So what this is, is the ability to. Add a face or create a face onto a video so you're basically you give it a single image of somebody's face and then you have that face be put on the [00:50:00] on a video but you can actually prompt the video from scratch. So, for instance, there's been some really incredible versions of people playing around with this that I've seen out there.

Kevin Pereira: One of our favorites, F O F R A I, did a thing where he took a picture of, uh, well, a person from Avatar, one of the stars of Avatar, and had them walk down the streets of Tokyo, and it completely transformed the main character into the Avatar, wearing sunglasses. There was another one where, uh, B underscore high on X.

Kevin Pereira: took two characters. So there's two images in a two person scene and it kind of put their faces on both characters. So this is a really powerful model that can do really interesting stuff. Um, I don't know if you had a chance to see kind of some of the examples of what's been done with this. Yeah. What's wild is that like, when I saw some of the examples coming out, I didn't know which characters.

Kevin Pereira: Yeah. Video service, this was referring to because I feel like again, as all these tools are racing, they kind of get to parody with each other. We've seen other tools that will let you, um, do three ingredients like a background and a subject and a, uh, [00:51:00] uh, an accessory, whether that's a bag or a purse or, uh, the glasses, et cetera.

Kevin Pereira: And so I didn't know that some of these examples were coming out of their new. S2V01 model, which is also such a sexy name. So I spent some time with this. Um, one of the things to be aware with this is obviously you can be a subscriber to minimax, you get a limited number of generations for a certain amount per month.

Kevin Pereira: I actually did this through replicate, which is a, a service that hosts different models and lets you pay per generation. This is kind of expensive right now on replicate. It's about 50 cents per generation, but I didn't really want to sign up for minimax for a subscription. So you can go and try for like, say.

Kevin Pereira: 2. 50 or 5, you know, you can get a certain amount of these, um, generations and you know, it's very easy. You basically, you have a prompt, there's a option to put a first frame image in, if you want, if you want to start it to kind of give a starting place. And then you have a subject reference file and it's very simple.

Kevin Pereira: You upload a picture of yourself. One thing, Kevin, you'll see in the first couple I did is I uploaded the picture that sometimes I use for my headshot, which [00:52:00] is like from a. From an older, uh, uh, event that I was at where I'm smiling really big. That does not work as well as you can see in the results. And I, I did a version of myself, uh, walking down the street in Tokyo and I did another version of myself in a cyberpunk world.

Kevin Pereira: If you look at both of those and both of those are a little funky looking because the face doesn't really fit on that well in those, uh, when you're looking at it also, do you notice that my cyberpunk version was given like a weird pot belly, which I also found strange. Yeah. They're shading you. There's Subtly trying to give you body dysmorphia.

Kevin Pereira: Um, yeah, the look, the, the, yeah, the smile is, is kind of a, a, a caked on almost Mr. Beastie and grin, uh, that, that doesn't move in the, in a believable way, the way the rest of the body does. But I think the examples are, are pretty good. Do you feel like it, it captures you like in a coherent way or does it, like when you see the, these renders, do you feel like you're seeing yourself?

Kevin Pereira: Well, I, so the, the, not in the smiling ones, but the second version I did is where I updated [00:53:00] with a, just the flat face of me staring directly at camera and of the like five generations you can see in this video. One I just did the one they had, which is walking down the street in Tokyo with sunglasses. I thought that one looks pretty good, right?

Kevin Pereira: Like it captured me pretty well in terms of now I have sunglasses on, so it makes it a little bit easier. What was the prompt that has you walking through the hospital as if you've got a broken arm? That was actually a shout out. from a woman named Heather Cooper who does a lot of amazing AI stuff. I actually stole her prompt.

Kevin Pereira: It was, it was originally a woman walking through a hospital wearing a white sweater. That one, the face is not as good until you get to a certain point, but I thought the Viking one looks really good. Like that actually looks pretty solid on me. The one of me kind of on stage doing a TED talk, it clearly struggled with that phase.

Kevin Pereira: And then, you know, the, the DJ one at the end is just not really me at all. And it doesn't really have as much exciting things going on. Um, but the, you know, the walking around in the white tuxedo thing is interesting too. I think this is worth playing with. If you have a couple extra bucks and you want to try something, [00:54:00] it's an easy way to try an AI model that's just as fun.

Kevin Pereira: doing something new. There are other models that do something similar. Kling has a version of this that you can do. I'm sure that there's a version of this that's going to be coming out from runway and from Sora at some point as well, too. This is just the kind of thing that will allow us to make.

Kevin Pereira: ongoing AI video, and I think one of the struggles with AI video up to date has been character consistency. So now you're entering a point where you can create a character and then enter them into many different places. And I'm sure also many of the people in our audience will be like, this has been able to be done open source for a while, which yes, I know if you are a comfy UI expert, you could have figured this out, but that is a lot of stuff for a normal person to play around with.

Kevin Pereira: Yeah, you can't ever shade like the one click nature of a lot of these tools. That's when they seem to take off. I would personally vote that we take, um, at least five of our Patreon dollars, Gavin, and put you in a Casino Royale type situation, because [00:55:00] you sitting down to the AI hallucinated roulette table, where it's got numbers, Random numbers and shapes on the wheel than along the sides of the wheel and the stacks of chips sort of like morph and go everywhere.

Kevin Pereira: I, I want to see, I want to see you in the casino playing all of the games. I want to see the craps table that melts where, you know, the dice become 12 all at once or the slot machine has some fake hallucinated cymbals going. AI Casino is a really interesting one to watch it try to hallucinate these games.

Kevin Pereira: Well, that's one of the fun things about this, uh, video model stuff is that because it's trying to do this thing with the face, sometimes the stuff in the background gets a little bit funky, which does make it interesting and fun as well. So anyway, go try this. You can try it on replicate or if you're a mini max subscriber, it's available now that's just dropped.

Kevin Pereira: There are just so many more of these really interesting AI tools that are coming down the pipeline all the time. And if you don't try them when they first come out, you might actually forget about it. So it's definitely worth giving it a shot. Do we want to talk about our hundred billion dollars in [00:56:00] funding that we're, are we ready to announce that yet?

Kevin Pereira: Sure. Let's talk about that. Yeah. You want to say what it is? Well, go, go for it. We want to thank the Oscar Mayer Corporation for coming in bigly. They saw what we were building. with hot dog city and they've decided it's time the uh, infrastructure needs to be the groundwork needs to be laid for hot dog city to become a reality so thank you to oscar and meyer both of you yeah there's two people which we didn't realize until just now in fact no idea in singapore they they are no longer allowed in the united states But we are and you may be if you're here, thank you so much for listening.

Kevin Pereira: Made their money in crypto, not even meat, who knew? Yeah, who knew? They asked me for like my email and my wallet address, Kevin. I don't understand what that means, but Oh, I gotta give you the money somehow. Thank you everybody for another fun episode. We will see you next week and on the YouTubes and all sorts of other places.

Kevin Pereira: Try stuff, have fun, and do it when you can. Bye. Bye.