How OpenAI's SORA Changes the Future of Filmmaking

The question of will our economic systems change forever, I think the answer is yes. Yeah, but how it's going to change, I don't think we know yet. A lot of people underestimate humans, especially when we're talking about AI. I feel like humans will figure out new ways to add value to the.

World, and I'm pretty encouraged there. Like, I think people who adopt AI, they're going to become enhanced in a way, and they're going to have an outsized impact on the world.

Hey, welcome back to the Next Wave podcast. I'm Matt Wolf and I'm here with my co host, Nathan Lands, and we are your chief AI officer. It is our goal with this show to keep you informed on the latest AI news, the latest AI tools and the talk of the AI world so that you can use it in your life and your business and leverage these new technologies that are there for you to use. Today's episode is a little bit different from what we've put out so far. This time we're doing a little bit of an Ask Us Anything episode. We went to X and asked you to send us your questions about the AI world and what you want to know and hear our opinions on. And we got some amazing questions. So in this episode, Nathan and I are going to deep dive and answer and give our thoughts on some of those questions. We talk about things like, where is AI art going? And is there room for it to get even better? We also talk about how the government and corporations and how people need to change to adapt to this new AI world that we're headed in. And we also talk about the limitations of large language models. How big can these large language models get? How powerful can they get? And we're going to deep dive and give you our thoughts on all of this in today's episode. So hang out with us and let's dive in. So the first question here comes from Vicky J over on X. She says, where do you see AI art and video going by the end of this year?

I mean, AI art, you know, the thing is, like, a year ago, AI art was, like, horrible, or like a year and a half ago, like, really horrible. Like, and then now it's, it's. It's getting to the point where it's like, okay, how is it going to get better? Like, like the newest, you know, version of Mid Journey is amazing. Yeah, maybe it's still, you know, with Mid Journey, I think the one challenge is it still always looks like Mid Journey to me. Like, it looks very polished and it's harder to get it having like a very different style.

Right.

So maybe having more, you know, controls over like, how, you know, like little details in the image versus now, where the kind of, oh yeah, produces something super beautiful, but it's not really easy to edit.

Yeah.

I think by the end of the year, you'll be able to produce images that look even more amazing than now, but you'll have way more control over how you actually edit those images. Which, you know, for commercial uses will be amazing. Right. Because if you're producing like, graphics for like, advertisements or anything like that, that's just going to get way better. I think by the end of the year.

Using only 20% of your business data is like dating someone who only texts emojis. First of all, that's annoying. And second, you're missing a lot of context. But that's how most businesses operate today, using only 20% of their data. Unless you have HubSpot, where all the emails, call logs and chat messages turn into insights to grow your business. Because all that data makes all the difference. Learn more@HubSpot.com.

Yeah, it is interesting because I feel the same way about Dall? E3. Right. Like, I've gotten to a point now where I can look at an image and go, okay, I can tell that was a Dolly three. I can tell that was Mid Journey. I mean, I don't know if the normal. If normal people who aren't like, as immersed in AI can see that or not. It's hard for me to say. Right. Like, I'm such in my own bubble that I don't know if other people notice that too. But Dolly creates very similar images. Like, the color palettes on them always kind of look like the same to me. Same with Mid Journey. That's got like this stylistic thing where I look at it and go, okay, I can tell that was Mid Journey. But I think, like you said, I think the updates throughout this year are going to be a little bit more nuanced, marginal updates. Right. They're going to be smaller updates because we've already got these AI models for images specifically that can create ultra realistic images. Right. Where else do you go? I think. I think the focus now is less on realism and more on prompt adherence. Right. Midjourney is really, really great at realism. Dall? E still sucks at realism. Dall? E is really, really good at prompt adherence. You can throw a whole bunch of elements into the prompt. You know, I want a dragon wearing a fedora, eating nachos, watching mtv and the words Party time on the screen and it will get all of those elements into a single image. Mid Journey can't do that. Right. So I think what we're probably going to see is like all of these models kind of combining to a point where like, nobody can tell whether it was Mid Journey or Dall E or stable diffusion because, you know, Mid Journey focused on prompt adherence and improves that. Dall E focuses on realism and, you know, more dynamic color palettes and optimizes for that. And we're going to get to a point where we don't know what model generated what because everything is just good. As far as video. I mean, we saw Sora. OpenAI has made it sound like we might see Sora sometime by the end of the year. We might actually get our hands on it. You know, we were just talking a minute ago about Adobe rolling in Sora to Adobe Premiere. So, you know, I think with Adobe Premiere we might start to get some access to Sora and some of those capabilities in there. So that's going to improve, which means more realism in video, longer prompt generation with video. Yeah, I think that's really what we're going to see. I think the other area that we're going to see continued massive leaps is in text to 3D art, right. For, you know, for, for creating like little animated characters or for creating game assets or things like that. Right. You've got spline 3D and you've got CSM common sense machines and you've got meshi. There's all of these tools that are like getting really, really good at like text to 3D object. And I feel like a lot of companies are kind of focusing in there right now. Like even midjourney said that they think their next sort of frontier is going to be 3D generation. We'll get back to the show in just a moment, but first, here's a word from HubSpot. Curious about what the future of productivity looks like. HubSpot's AI tools make quick work of expediting content creation, optimizing workflows and elevating data analysis. Personally, I'm really impressed by the AI website generator. You can create a web page just from a few simple prompts. That's mind blowing. And the facts speak for themselves. Over 85% of marketers and sales professionals agree that AI enhances content quality and prospecting efforts. So what are you waiting for? Sign up for free by clicking the link in the description below. Now back to the show.

Going back to video for a second. I, I, I think that, you know, we're. I'm sure we're going to see Sora this year, maybe when we see Sora and even, like a small upgrade to it, like, you know, something better than the demo, like, in the public's hands, which is going to be amazing. And. And I think you, you know, by the end of the year, I think there's a chance that we'll actually see, like, a. A viral short film, like something like five minutes, 10 minutes long, that's entirely made with something like Sora. Yeah, I think that's gonna be like the turning point of, like. Oh, like, wow. Because, like. Because it's. It's almost there now. Like, if something like Sora was out there, I'm pretty sure somebody could make something that went really viral. And I think you might even see, you know, because we've been talking about, like, there's all these great new AI video tools coming out, but, like, what's the use case? I do feel like there'll be something, you know, almost kind of like how Instagram took off because of filters. I do feel like there's some kind of new app, whether it comes this year or. Or next or. I feel like there's something on the horizon that we. We don't even know what it is yet, but some kind of social app that will integrate these AI video tools. And we're like, oh, yeah, of course. Why did it. Why didn't I think of that? Like, I think something like that, bigger than TikTok is coming using AI, probably AI video.

Yeah. Interesting. I wonder what a social media app would look like in that realm. Because it's like, yeah, is it going to be just a bunch of AI videos and people are scrolling a feed of AI videos?

I guess the challenge right now is, like, the filters were free. A lot of this stuff is quite expensive to actually do right now. So, like, you would have, you know, had to pay for the generation of the video, I guess, is the one challenge you probably would have right now. But. So that might make it where, like, it's more likely to happen next year when costs come down or something. But it does feel to me like, okay, we haven't had big social app in a while. We had that kind of like. What was that one called? Gas or whatever, which was like, this guy kind of like generated this out of nowhere, then sold it to Reddit, I think, really fast.

I don't remember that one.

Anyway, it was like half a joke.

Yeah.

And he sold it for, like, a ton of money. Like. Like I don't know, like over 50 million or something. And, and apparently like I think he told Sean Puri from my first million, a few others, like he was going to do it before he did it and he just did it. But there hasn't been any other big social apps, you know, since then that have taken off and it feels like, okay, so AI Video probably is going to enable something really cool there.

Yeah, I mean, right now like Air Chat is sort of a social app. I mean it came out like a year ago, but they rebuilt it. Now it's sort of reemerging. I think that's mostly emerging just because it's got Naval's name on it and people flock to Naval.

Yeah. But yeah, Naval is saying it's because of AI partially. He is saying because he believes, which is funny, he used to say that text was like the best and then now he is switching to saying like, yeah, it's all going to be audio and you know, it's all, it's all going to be audio and that's why I'm doing Air Chat. And he, he, apparently he is saying that one of the reasons is he's now kind of reevaluating things because of AI realizing, yeah, with all these new AI tools, you're going to be talking to your AI and then you're going to be talking to people that you want to, you know, real people you want to talk to as well. And so like voice is going to be the main, you know, medium there.

Yeah, yeah, I know, I know we're sort of on a little tangent right now, but when it comes to, when it comes to Air Chat, the thing that I like about it is that you can't type, you can't open it up and type a response to somebody. You have to speak. And the thing that makes that cool is it sort of solves the problems that X has right now with a lot of bots. Right. Right now you can't really bought it because you've got to talk to it. Unless you're using a really, really sophisticated sort of like 11 labs combo where it's spitting out real voices. And like, I, I don't think anybody's that advanced with it yet. But you also don't get like all the trolling and all the negativity because when somebody actually speaks to you face to face, they're not going to be as negative and as much of a dick basically as they might be on something like X.

Did you ever use Clubhouse, like when back a little bit?

Yeah.

Yeah, yeah. So, I mean, Clubhouse was, like, blowing up, especially in Silicon Valley. Like, everybody was on it and. And it started off great, and then it went downhill.

And, you know, it's like. Because it's really hard to moderate these things. Like, yeah, at first everybody's nice, but then you got people talking about political things or whatever, and, like, people from the opposite side join and, like, it gets really hateful and it just, you know, and then those, those kind of channels became the most popular channel because there's more engagement there. I wonder what Air Chat's doing to kind of make sure they don't have that same problem.

Yeah, it'll be interesting. I'm new. I've only been on it for, you know, 48 hours as of this recording, so. But anyway, we can. We can move off that tangent for right now. So let's. Next question. This one comes from Craze in the Dark, who's actually somebody that works for me. This is. His name's John. He's my creative director. But he asked this question. He said, how can governments and corporations adjust to a potential massive AI job replacement? Will our economic systems change forever? You know, real, real easy softball question.

Okay, next question.

I do have, like, really mixed feelings here because, like, you know, in terms of will it change our economic systems forever? I think that one is so hard to predict. Like, anybody who says that they have a predict like, that they know what's going to happen there is lying or, like, you know, deluding themselves because AI introduces so many new variables into the world. It's so hard to know, like, okay, GPT5 is going to come out? Is it 10% better? 50%? What does that mean? What does that mean for the economy? What does that mean for how people work? Will these systems stop improving as fast as they are now, like in three years because we hit some kind of limitations, or will they just start growing exponentially? And so it's really hard to know. But, you know, in the future, if AI can do most work, that definitely opens huge questions of, like, what's the purpose of the government?

Yeah.

Like, what happens when AI gets incredibly smart and starts telling us that our government's messed up? What do we do then?

We elect an AI to run our government, obviously.

Yeah, yeah. Do we.

Yeah.

Do we ask the AI how we should be structuring our government based on the. You know, and how do politicians respond to that? Do they decide, okay, we have to bend AI Now? That's probably what would happen. So I, I think no one knows, like, you know, because it's like right now, you know, capitalism is, in my opinion, the best system we have. It's definitely not perfect. Like, everyone knows that. And so in the future, when AI does the work, people are going to have to find meaning in other ways. And also, how does money work then? I think, I think, I think Mark Andreessen talked about this recently where he believes that AI will dramatically bring down the cost of many things, right? And then there'll be certain niche products or niche or maybe like real life experiences that still cost a lot of money, right? Like going somewhere and having some amazing experience or having a human create something that, that, that could still be expensive.

I think that's giving companies too much credit. I give, I think it's giving companies too much credit because what we've seen so far, right, like there's so many companies that are built on the back of GPT3, GPT4, and over the last several months, so many of these APIs have lowered their costs. Lowered their cost, lowered their costs. Have the companies that we're paying our monthly fees to lowered their costs or just pocketed extra profits.

Yeah, well, I mean, so it depends on how good AI gets. So, like, that's, that's when there's like a handful of competitors. But if you have like a thousand competitors and they're all competing on price because AI has made things so much easier to build, you know, a lot of prices should go down over time. But, but yeah, you're right. I mean, iPhone's still incredibly expensive. Other things that we buy are still incredibly expensive. So you're correct there. I, I'm not sure, you know, I. But, but also at the same time, I don't really, you know, like Universal Basic Income, which I believe Sam Altman originally was one of the people who kind of like, didn't experiment with that where they gave people money. I believe they gave people money in Oakland, ran that experiment, and I can't remember the results, but I think it was kind of mixed where it's like, yeah, some people just end up kind of just like playing video games, you know?

Yeah.

It's like, okay, is that what you want in your society? Like, I don't, I don't know. Like, maybe there we, we, you know, maybe some, you know, portion of people. That is what they do, and then they go off and do some productive things, hopefully.

I think everybody worries about the, the, the Wally scenario, right, where everybody's just kind of sitting in their chairs, floating around, getting fat, drinking Slurpees. And watching movies or playing video games all day. Yeah.

But, you know, recently there's been this big, you know, uptick in like people healthy, at least in like Silicon Valley circles and stuff. And, and even me, I, I lost a bunch of weight and I've been using AI even to get, you know, give me advice on like, how to go to the gym and like how to work out and gain muscle, all this stuff. And, and so people were like, yeah, so we thought it was gonna be like a Wally and everyone's super fab. It's gonna be like a Wally and everyone's super jacked.

Yeah.

You know, they got all this free time and AI helped them, teach them how to be healthy. We have all these new drugs that make you healthier and stuff and. Yeah, and yeah, so it might, it might be like that, like Wally, but we're all jagged. So. Yeah, I don't know. But my gut feeling is that there will always be like, the capitalism will just evolve. Like it's not going to disappear anytime soon in a thousand years or, you know, whatever. Who knows, maybe we're in Star Trek land, but no time that soon.

Yeah, I don't know. I feel like a lot of people underestimate humans, you know, especially when we're talking about AI. I feel like humans will figure out new ways to add value to the world. Yeah, right. You know, right now one of the ways humans add value to the world is doing the actual construction, building the stuff, you know, doing the actual work that needs to get done. But I do think in a sort of post AI world, maybe call it a post AGI world, where robots and AI can handle a lot of this stuff for us. You know, I still think humans are going to figure out new ways to add value. We just don't know what that looks like yet. Also, questions like this about how can governments and corporations adjust? My, my thoughts on that are not necessarily that corporations and governments need to adjust. It's that we need to adjust the people, the, you know, the general population, the public should be adjusting. You know, what we're seeing right now I think is one of the greatest opportunities in anybody's lifetime to quickly build and iterate and optimize and create something yourself that adds value to the world. So maybe you're going to lose your job at Walmart as a checker because, you know, you can just walk out now, but you also just have the opportunity to use AI to help you create a business that can generate value and income for you. So, you know, we're in this very empowering point in time where I don't think we should be going what can governments and corporations do to protect us from AI? And, and we should be thinking what can I do with AI so that I'm not that reliant on corporations and government. I think, I think we need to have a sort of paradigm shift. We need to think differently. Right. You know, not, not trying to quote Apple here, but we do need to think differently in that we need to not look at corporations and government to solve these problems for us. We need to look at how can we be empowered by this new technology so that we aren't reliant on government and corporations.

Yeah, I agree. Like everyone has a, you know, free assistant, you know, in their pocket now. You know, they used to have a computer, now they have an actual assistant and that assistant's gonna get more and more intelligent. So yeah, that, that's, and, and, and that's probably how you'll actually see changes in, in government and corporations is like from the bottom up, you know, like, of like people learning this stuff and then, and, and changing the government and, and, and I'm pretty encouraged there. Like, I think AI, like the people who adopt AI, they're gonna become enhanced in a way and they're going to have an outsized impact on the world which, which should give them a, you know, a better chance of actually influencing the government or becoming part of the government, helping restructure things.

I mean the question of will our economic systems change forever? Think the answer is yes.

Yeah.

How it's going to change, I don't think we know yet. I don't think anybody is going to be really good at predicting that in this moment in time right now. But definitely I think there's. The economic system is going to change. I mean we're already seeing it change, right? You know, people are trying to figure out ways to bypass the existing financial systems and banking systems through crypto and things like that. It's going to continue to evolve that way no matter how much people sort of fight that stuff. I do think it's going to like evolve towards that more decentralized crypto centric finance in the future. But yeah, definitely the economic system is going to change forever. I just, I can't make a prediction on how yet.

Yeah.

So the next question here is from Universa AI podcast. Their question is, do you think open source community can band together and create an AGI that can compete with the big corporations? So really I think the, you know, the main question Is here is can Open Source build AI models that can compete with an OpenAI, with a Microsoft, with a Google, with an anthropic, with companies that have millions, potentially billions of dollars backing them with compute power?

Yeah, I think we've talked about this a little bit before. Not sure if we put it out there, but you know I, I personally really hope so. Like really really hope so. But I am skeptical. I, I do feel that OpenAI is further ahead than people realize. I think OpenAI is probably about one to two years ahead and when GPT5 comes out I think that's gonna become apparent. Maybe I'm wrong, maybe I'm delusional, but that's what, that's my current belief. And you know, it feels like yeah, open source is catching up to GPT4, but if GPT5 is like two years ahead of that, right. You know, it's like, well then maybe they'll always be in it. And then at some point when AI starts self improving, you know, let's say by GPT6 that it's actually kind of improving itself a little bit too by itself. You could get in this like you know, exponential growth of the AI where then whoever gets there first or the first two or three companies that get there first, it gets better so fast that yeah, sure there'll be like cool open source models, you know, for individuals but maybe the most powerful stuff will always be in like two or three companies hands unfortunately. And then, and then there's the question of like yeah, who gets to use it? Do the government say like holy crap, this is so powerful it changes the entire world. And yeah, we have to like really, you know, treat like you know, nuclear bombs or something. And I really hope that's not what happens.

Yeah, so when it comes to open source one of obviously the biggest bottleneck is compute, right? So like yeah, you look at companies like Google, like Microsoft, like Anthropic, like Mixtral, some of these companies that are building these large language models, they have millions and in some of these scenarios billions of dollars of investment which they could use on GPUs to train these models. Even the open source community, even all of the large language models that are out there in the open source world right now, probably costed millions of dollars to train on, right? So you still need one of these big companies like a meta, like a Mistral that has all of this financial availability to train these models even for the open source to be successful. Right? So I don't know if open source, at least not in the near term, I don't know if it can be successful without the big corporations deciding to get involved and help train these models. Because the open Source World of LLMs right now probably wouldn't even exist if it wasn't for meta dropping llama 2, which everybody's sort of kind of fine tuned and built off the back of that. Right. So it's like, I don't know, at the end of the day, open source is amazing and we want it to grow. We want more people tackling problems and more people trying to figure out how to make these bottles better is a better thing than less people working on these models.

Right.

But at the same time, you do need millions of dollars to train one of these models right now, even if it is open source.

Yeah. In the future, like when we're talking about GPT5 or GPT6 level models, you're probably talking about way more than that. I mean, I think the other day Google came out and said that, you know, over the next, I don't know if they said next five to 10 years that they'll probably spend like a, end up spending like a hundred billion dollars on AI development.

And that news came right after Microsoft said that them and OpenAI are partnering on $100 billion data center. Right?

Yeah.

So basically OpenAI and Microsoft said we're going to partner on $100 billion data center filled with GPUs so we can train better and better models. Recently the CEO of DeepMind, Demis Hassabis, came out and said in response to Microsoft building this hundred billion dollar data center that over the next five to ten years Google's going to spend at least that on their data centers. So that was like him sort of responding to the fact that Microsoft and OpenAI were building this hundred billion dollar data center. But Metta said, what did they do? You remember the exact number? Metastat. I think they said they have something like 600,000 H100. Something insane like that.

Yeah, something. I think a chart came out showing like they have the most or something.

Yeah. So Metta has 350,000 H100 right now.

Oh my gosh.

Yeah, so. But Mark Zuckerberg make a made it did make a comment recently saying they have the equivalent of 600,000 H1 hundreds, whatever that means.

Yeah.

So, you know, I don't really feel like an underground dude in his basement is going to train a model that's going to compete with a GPT4. They, they just don't have the compute availability to do that at least not in this point in time.

Yeah.

This question comes from AI leaks and news. What is one thing you are confident we won't see in the next 12 months?

I don't know if I could be confident about anything. 12 months. I mean, that's the crazy thing is, like, it's really hard to know where the, you know, where OpenAI is privately, like, where their model is at. I don't expect to see AGI in the next 12 months or, you know, artificial super intelligence or anything like that. And partially because I think we're probably, we need, you know, we're probably limited by the current compute, but that's about it.

Yeah. I mean, I have the same thoughts, like, almost every time I've tried to predict something that won't happen, I've been wrong. Right.

Like, I look at AI Video. Right.

Yeah. I've made predictions about how far AI Video is going to come, how fast. I was totally wrong on my prediction. And what I thought wasn't coming for years came months later. Right.

I was very optimistic about AI Video. Like, I, I, like, I was uploading out AI video threads for a long time, talking about how it was, you know, going to really change Hollywood and whatnot. But, yeah, it's hard to know. Like, yeah, AI Video, it could have taken years, but now it looks like it's actually going to be amazing in the next six months.

Yeah, yeah, yeah.

So there's not much that, that I would be willing to, like, confidently predict in 12 months.

I'm with you. I think AGI and ASI are probably the two things I'm fairly confident we're not going to see within 12 months. Definitely not ASI, but most likely not AGI either.

Yeah. I'm fairly confident that I will not replace all jobs in 12 months. They will not be that good.

We won't have Skynet taking over the world within 12 months.

Yeah, probably not. Probably.

I like how you said probably not. Not. Definitely not.

Yeah, probably not. Yeah, probably not. I think I saw something the other day where one of the AI systems in China with the government, it's actually called Skynet.

Yeah. A few people have, like, named their company. I've seen a couple different Skynet companies and I'm like, yeah, you do realize, like, Skynet was the bad guy, right?

Yeah, I think they, they thought it was humorous or something, but, like, it's. Yeah, it's literally like China's AI system that, like, monitors the public.

Yeah. All right, so this one comes from Akshay Lazaris, and I'm sorry if I butchered that name, but he says I'd love to hear you discuss the future of tech. For example, will UI switching from GUIs graphical user interfaces to voice? Do we see there being a consolidation of all application tech by cloud based hyperscalers? What is the role of startups as data resides with these hyperscalers? Well, let's start with the first part of the question. Like, what do you think the sort of future of these user interfaces are? Like, less graphical user interface, more voice. What are your thoughts there?

I think both. I mean, I think you'll have both. I mean, I think you'll have, you know, like, like, you know, Minority Report kind of thing where you got like these visual, you know, systems that you can interact with and they get really intelligent and, and yeah, of course you'll be able to also use voice to interact with it as well. Kind of like her and other movies, right? Where?

Well, I mean, we're already kind of doing that. Right. Like, that was sort of the point of Siri. Right. That was sort of the point of Amazon Alexa. So we already kind of have that. They're just kind of dumb AIs right now. They're not very great AIs, but we're already kind of, we've already had that for a while now. So I'm going to speak for myself here. When it comes to like speaking out loud to some of these apps, I feel very awkward doing that in public. Right. Like I've got these meta Ray Ban sunglasses that have llama 2 built into them. I feel awkward as hell in public asking questions to my sunglasses. Right? Because you go, hey, Meta. And then you. And that's like, like saying, hey Siri. Right. That's what starts the prompt. So I can start talking to these sunglasses. I feel ridiculous in public saying hey, Meta. And then talking to sunglasses. I feel like I feel the same way with an AI pin. Right. The humane pin. Pressing a button and then speaking out loud. I like, if I'm sitting in a coffee shop or a library or somewhere, that tends to be a more quiet place. I'm definitely not using the voice interface. Right, right. Even just being out in public, like if I'm walking through my supermarket and there's other people around, I feel like a crazy person talking to my gadgets, you know?

Yeah. I mean, that might be an age thing too, right? Like, it could be. I think that like, like young people may, you know, AI gets better and better. They may just be totally used to it. Like, of course I'm talking to it. Why would I, you know, it's like, it's like my, like my son would like Siri and stuff. He, he doesn't understand the world without that.

That's true. Yeah.

And he also just thinks Siri is so dumb. So as soon as it gets better, he, I think he's going to love interacting with it however he can, including being outside. But yeah, I also would feel very weird, like talking to, you know, AI outside, especially here in Japan people. He's staring at me like a psycho, like, what is this? What is this guy doing?

I mean, if really the question is essentially like, yeah, we've got huge incumbents that are sort of running everything right now. How does anybody compete?

I personally think it's a big open question, like, yeah, yeah, will, you know, when OpenAI has AGI, you know, what do startups do? But I think there'll always be like, I don't think, you know, OpenAI is going to want to build every single use case and every single kind of product. They're wanting to build the technology and have other people build on top of it. And I think if they tried to do everything, they're going to run into like, really major, like, regulation problems, if I had to guess. So I think, you know, that, that, you know, I don't like regulation, but that might be one area where like, regulation actually comes to save the day. Because I do think the big tech companies are not going to try to do everything. They're going to try to build the best AI technology and then on top of that, there'll be so many things for startups to build.

Yeah. And really, if the question is like, how do they compete? Honestly, our, our Greg episode, our Greg Eisenberg episode is literally that entire topic of, like, when you have these big incumbent companies, the Google's, Microsoft's, Amazon, all these companies out there, how do these little companies compete? Well, that's exactly what we talked about with Greg and Greg was talking about, well, you got to create brand, you've got to create community, you've got to, you know, what are you going to do to stand out? Right. I just listened back to that episode when it went live and Greg was saying things like, you know, people choose sides, right? I'm an Apple guy, I'm a PC guy, I'm a Nike guy, Adidas guy, Puma guy, whatever, right? Like, people sort of gravitate towards brands and attach their identity to brands and a lot of times they're the brands that have good community connected to them. Right. So that is, in my opinion and Greg's opinion, and it's my opinion because I heard Greg give me this opinion. But, you know, that's how you stand out, is through brand, through community, through those things that differentiate you. Right. People don't necessarily want to work with Google where they're just like a number in a giant database of a million people. But I use, for instance, Beehive for my AI newsletter, right? And with Beehive, I've talked to the founders. They actually respond to people on X. They're still a smaller, nimble company with community around them, where everybody that uses Beehive sort of love is talking about it until. And recommends other people to go use Beehive. And they've got that community element, that brand element. And there's a narrative that sort of formed of this like, convertkit versus Beehive narrative. Right? And so you've got that kind of thing going on where people might, like, pick their sides. And the reason they'll pick a side is they connect with the community. They connect with the brand that that company built. And so I think, you know, as Greg said, that's how they stand out. And, you know, I totally agree with that opinion.

Yeah, yeah. And I think, you know, AI will only, like, make that even more so, right? Like where people want more human connection there, right? Like, yeah, I know the founders or like all the people in the community, I love them and I, I feel like they actually, like, listen to what I say. And this is like one reason to do like, Q and A like this, right? Like, is to actually, like, yeah, we're doing the show, but also, you know, it's for other people and we're, and we're, you know, interacting with those people versus just. Oh, it's just what me and Matt say, like in startups in general, beyond community, I mean, I think, you know, AI is going to make. There can be so many opportunities for startups and I just, I don't believe that the, like, three companies are going to control everything. There's, there's so many, there's so many problems to be solved in the world and it's not going to be, it's not going to be two or three companies that solve all of them. Like, there will always be new opportunities.

I think if it does become three companies trying to do it, all governments will intervene and be like, no, you can't have this much power. I just, I think that's where it'll go.

Yeah, yeah. Okay. This one comes from my Twitter feed. Comes from Jason Vanish. Hope I'm saying that correctly. Jason's always been really awesome on my Twitter and doing a lot of good comments on my Twitter. How much better can get at some point there aren't larger and larger training data sets, are there at this point, then do we hit an AI plateau? Or are there different methods that could leapfrog LLMs? Like, the computer industry had computer breakthroughs from the 70s to 90s. I mean, I think with this one, like, we're definitely, like, speculating because, like, neither one of us are like, the smartest AI engineers in the world or something. Like I cod.

I mean, you could just end that sentence as neither of us are really the smartest.

Well, yeah, there was an interview though, I think with like, Sam Altman and Ilia maybe, I don't know, maybe it's like five months ago, six months ago, something like that, where, you know, they said that with existing data in the world, you know, with video and audio and other things that we're just, we're just now starting to train on that there's a pretty clear path to major improvements for the next three to five years. Right. And so if that's the case, we're probably talking about like GPT6, GPT7 before you need any kind of breakthroughs.

Yeah.

And if GPT5 is good, as they're saying. I mean, I think by GPT 6, 7, we're talking about, like, work looking very, very different.

Right.

So I think the next three to five years, there's already enough data for the world to be entirely transformed, you know, mostly in positive ways. And then beyond that, I'm sure we will need more breakthroughs at some point. And there's a big open question too, that people, you know, that a lot of engineers are debating is, like, with synthetic data. Right. Like, is synthetic data actually useful? Like data that the AI helps generate? Can you then trade on that, that new data? You know, it's synthetic data. And I think it's not entirely clear yet, but I could be wrong on that. Yeah, but if synthetic data works well, then, yeah, there's. We'll probably never run out of data to be training on.

Yeah. And I mean, everything we're talking about at this point is very sort of theoretical. But yeah, you know, at some point, I don't think humans need to continue to improve on training the large language models. Like, once we hit a point of, you know, AGI, we get to a point where the models figure out what they need to train on next and figure out how to self improve and sort of get better and better. I think there is going to be sort of a model that's kind of like the last model that humans needed to train. And then the models beyond that become the models where the AI is sort of a self improving model where it just, it gets better and better and better on its own. Right. It does its own is reinforcement learning through AI feedback instead of reinforcement learning through human feedback at some point. Right. You know, to some degree we have that right now. Right. Like that's kind of how Gans generative adversarial networks work where you have an AI that's a discriminator and then you have the AI that's the generator and the AI that's the generator tries to generate something that fools the AI that's the discriminator. And they go back and forth until the discriminator is actually fooled by what the generator made. I think we're going to see that kind of thing get more and more prominent with large language models where it gets to a point where it's giving itself its own feedback and getting better and better and better. But then I also think there's a phase after large language models where we start talking more about the embodied AI. Right. Like we were looking at the Boston Dynamics robot, we've talked about the figure one robot, we've got Tesla Optimus. All of these are robots that, you know, they're going to have AI injected into them. And when we start putting AI into some of these robots, some of these machines, well now they have more stuff they need to learn on. They need to learn on how to interact with the physical world. And when my arm moves like this, what's the result when my arm moves like this? And now it's starting to get trained more on the domain knowledge of just how to operate this robot. So I think there's going to be this shift to now we need to train these LLMs to work with the specific use case, like embodied robots, like you know, going into drones or whatever we use it on next. And they need to sort of train on the domain knowledge to operate the vehicle that's now embodying that AI, if that makes sense.

Yeah, totally. I think a big open question too is like, can AI start solving real world problems too? Like can it, can it help, you know, cure cancer? Can it, can it help us figure out how to make the robots better? Right, because. Because if that's the case, then we'll probably get to this kind of exponential point where things will just keep getting Improving and, and then, yeah, something that seemed like such a small little breakthrough, you know, with, with GPT1 and with Transformer, you know, the concept of transformers. Yeah, that takes us, it could take us all the way. I, I think that's an open question. We don't know. It could be that, like, there's a ton of data is all we needed.

A ton of data and a ton of GPUs to process it.

Yeah, yeah. Also, I think, you know, LLMs are going to be just like one part of the equation too. Like you said, the robots. But also my understanding is that, you know, the rumors were the OpenAI has created this thing called Q, which is supposed to be some kind of logic engine. I don't think there's any details been revealed about that. Is that just like some kind of other LLM? I don't know. But. So in theory you would have like the LLM then having some kind of like, logic engine attached on top of it. So in the future these systems could get really complicated, but for like regular people, you won't know any of that's going on.

Well, if you ask Jan LeCun. Right. He's the lead AI over at Meta, Also worked with Geoffrey Hinton, one of the godfathers of AI, as they call them. Yeah, right. He doesn't believe large language models will ever achieve AGI. He thinks that the sort of technology underneath large language models just will never get to that point of AGI. And over at Meta, they're developing something they called V jepa, which stands for Video Joint Embedding Predictive Architecture. But basically it's a way for like the AI models to see and understand the world and sort of train themselves by actually doing and getting a response. And he believes this is what will actually lead to AGI and not necessarily the large language models that everybody's familiar with today.

Yeah.

So I don't know, there might be a point where large language models, there is like a point of diminishing returns where they just don't get any better. But I still think there's a lot of models that can replace large language models that continue to improve what the capabilities are.

Yeah, I don't know. I know he's highly respected and obviously he knows more about this technology than I do, but he's also made some predictions with, like, towards like GPT4 and things like that that didn't seem that accurate. Like he, like, he, he, he was really skeptical on how good GPT4 was going to be and stuff like that and, and now, like, with the rumors about GPT5, if they turn out to be true, well, then obviously OpenAI is doing something that he doesn't understand.

Yeah.

Right. And so. So we don't know what that is like, in. In this Sam Altman bullshitting. Well, I don't think so. Like, you know, and the other day, Sam Altman said they're like, you know, people who are ignoring, like, where things are going to be going with GPT5. I think he said something like, you're gonna. We're gonna steamroll you, is what he.

Said, which is kind of.

Kind of shocking. He was like. And he delivered this kind of like, really, like, you know, peaceful, nice way, but it was like, yeah, we're gonna steamroll you. And so I. I think they have something. I think it's gonna be shocking. But. But who knows. Who knows what the. You know, what's gonna be the best way to do this in the future?

Yeah. I actually think.

Is going to be a huge bottleneck over time. Right. I think. I don't want to get political with this, but I, you know, I do think that the government is going to be a bottleneck to progress. And the reason I say that is because I was recently listening to a conversation with Ray Kurzweil and Jeffrey Hinton. Right. And they were talking about these AI models being essentially, like, the best correlation machines on the planet. They can find correlations between seemingly unrelated things that humans can never spot. And that's the reason why these large language models will likely find the cures to various cancers and diseases and possibly solve climate change and world hunger and all of this stuff. Right.

They'll find very good at finding corruption.

That's. That's true.

You know, I had a viral tweet on this, like, maybe a year ago where I talked about that. I was like, I think it's going to be a big thing where you can actually see how. Where all the money is moving in the government.

Yeah.

Humans cannot process this. And it's a very easy way to, like, be corrupt and hide. That is, like, through moving money around in, like, really sneaky ways.

And which is another thing blockchain sort of solves as well.

But, you know, to, like, look at that and go like, oh, yeah, this person's doing this. This is why they're doing it. This is why they, you know. Yeah, yeah, sign this into law. And so, yeah, I could definitely see government people being very, like, not wanting that to come out.

Yeah, well, you know, where I was going with the. Them, the Government being the bottleneck is.

Yeah.

I think these large language models will likely find cures for cancer, you know, solutions for climate change, you know, ways to end hunger in, you know, parts of the world that need it. But I think government is going to get in the way of making what the AI finds a reality. Right. Like when there's new drugs that come to market, how long does it take for the drug to go through animal trials and then human trials and then, you know, all of the steps before the drug finally gets on the market for humans to use. Right. I think I will probably find a lot of solutions to a lot of problems and then the government's red tape is going to be what slows down these solutions actually becoming live to the world.

Yeah, I mean, I mean, that's why there's movements like EAK and things like that. Right. It's is. Is they are really afraid of that. It's like, I think people talked about, like, you know, we had, we could have had solved a lot of our energy problems with like nuclear power, but then regulation, like stopped that from happening. Right. And I think France is one of the big success stories where they adopted nuclear power and they haven't had energy problems like other, a lot of other countries have. And so there's a lot of fear around that. And. But then the challenge is like, you know, EAG does kind of go so extreme that they might end up bringing on some of the regulation. Right. Because it's like just, let's go as hard as we can, as fast as we can, and then the government kind of freaks out about that too. So it's really hard to know, like, how to get the government to embrace this technology. I'm hoping they will. I mean, I think it's like the US winning at the Internet, like the US needs to win AI too, in order to kind of set the stage for the next hundred years of the world. Right. And if we don't, whoever wins that, China or whoever, you know, Russia, whoever, they then get to set the stage for the next, you know, chapter of humanity.

It's the new space race.

Yeah, I think it's bigger than that. Yeah.

Yeah. I mean, it is really interesting to think about it from that perspective too, because, like, the US probably wants to be the country that finds the cure for cancer, but is the government going to, you know, is the government going to put up a bunch of red tape and slow that down? I don't know. Yet to be seen. Yeah, I, I think there's going to be a sort of necessary overhaul coming. Because with, with this AI world that we're entering into, I do feel like we need stuff to happen faster so that we can let the solutions that these tools provide provide the solutions.

I think ideally, that's what we do. Like. Like, as these systems become incredibly intelligent, more they can process more data than humans can. We should be asking them like, okay, how can we reorganize this part of the government to be more efficient and actually get more things done and help more people, bring more people out of poverty and things like this? That's what I hope for. Yeah, it's. That's even one reason I wanted to do it, you know, do the show with you. That's why I tweet is I'm really hopeful for what this technology can do for humanity if we don't all get in the way.

Right. You know, I think that's a perfect spot to wrap this one up. If anybody is watching on YouTube, let us know in the comments. If you like this style of episode, if you want us to do more Q and A and you like just the sort of us giving thoughts on your questions. Because if you like this, we'll make more episodes like this for you. If you haven't already. Make sure you like and subscribe wherever you're watching or listening to this. It really helps us out to get more listeners and viewers on the show. So appreciate that and thank you so much for tuning in today. We'll see you in the next episode.

Sam.