Shownotes Transcript
As a small business owner, my favorite thing about posting a job on LinkedIn is that when I hit post, I
I clock out and LinkedIn clocks in. LinkedIn makes it easy to post your job for free, get qualified candidates, and manage them all in one place. Plus, LinkedIn extends the reach of your job post by allowing you to share it with your network. And hiring managers that add a hiring frame to their LinkedIn profiles receive two times more qualified applicants. Go to LinkedIn.com slash achieve to post your job for free. Terms and conditions apply.
The argument here is that you might be able to do some of this AI research and AI models on lesser graphics cards because DeepSeq trained this model on a lesser NVIDIA card, which was supposed to not be possible.
Hello and welcome to the Barron Streetwise podcast. I'm Jack Howe, and the voice you just heard is Gavin Purcell. He's a former showrunner for The Tonight Show starring Jimmy Fallon, and he's a frequent AI experimenter and the host now of a popular podcast called AI for Humans. In a moment, we'll hear from Gavin about AI and show business, but also about something from China called DeepSeek.
It caused a dipsy doodle in the U.S. stock market this week. Gavin will tell us what it all means.
Listening in is our audio producer, Jackson. Hi, Jackson. I think it was more than a Dipsy Doodle. It was like a trillion dollars. I mean, does Dipsy Doodle sound small? It sounds small. Well, first of all, I don't think it was that big, as I'll explain in a moment. And also, Dipsy Doodle is also a brand name of one of the sturdier corn chips. So I don't think that there's anything about the name that's necessarily diminutive.
But before we come to that, people are no doubt outraged right now. They're saying last week on this podcast, you promised me an alcohol part two. You gave your word. Jackson is what they're saying.
And look, folks, I know. But the thing is, we had this market mayhem this week, and it's caused by something called deep seek. And I had a conversation lined up with Gavin about AI and show business. But Gavin also happens to have thoughts on deep seek. And so I figured I'd bring this conversation to you now in case you're worried about what's going on with your stocks. And we will get to the part two about alcohol next week. Unless something else comes up. If it bleeds, it leads.
Oh, gross. There's not going to be any bleeding next week, no matter what we do, unless it's the bleeding edge of financial podcasting. That I can't rule out. That's right. Let me give you the briefest of backgrounds about what happened in the stock market early this past week before we come to Gavin. On Monday, stock investors woke up to DEFCON 12 level panic, unresolved.
On the financial news, really, because in the actual stock market, in percentage terms, it was not really that big of a deal. The S&P on Monday closed down 1.5%, and the following day it bounced back close to a percent. For the Nasdaq, which is more loaded up on tech stocks, the moves were more severe, down 3% on Monday, up 2% on Tuesday.
If you own NVIDIA, it was scarier. Monday down 17%, Tuesday up 9%. At the time of this recording, shares were looking weaker once again. Jackson, we had a question just come in about leveraged ETFs and it mentioned NVIDIA. Where's that question? Yeah, I have it here from Scott.
Hi, can you please explain what a leveraged ETF is, when they started and why they became so popular? I just saw that they recently had a big downturn because of Nvidia. Thank you. Thanks, Scott. Leveraged ETFs are almost two decades old.
For buy and hold investors, they are mostly a terrible idea. The fees are sometimes high, the leverage adds risk. You'll see names like two times this or three times that to indicate how leveraged they are in the name. It's a little misleading because they're leveraged typically to daily price movements, not the kind of longer term price movements that investors track.
There are other complications with them. Why are they popular? I think because they're easy. You can bet for and against things in a leveraged way by using A, leverage, or B, derivative securities. But derivatives are complicated. You have to learn a whole new thing to use them. With leveraged ETFs, it's basically the same as buying a stock, and that's whether you're betting for or against something. Your question that you heard, NVIDIA is taking down leveraged ETFs. Well, that's not true.
That's true in that NVIDIA helped take down the entire US stock market and lots of leveraged ETFs track the stock market. And there is a leveraged ETF that tracks NVIDIA. And I think even one that bets against NVIDIA, if I'm not mistaken, Jackson, what is the ticker symbol on that monster? NVDU. NVDU, that is the...
Direction spelled with an X. How do you pronounce that? Direction daily Nvidia bull two times shares, not an endorsement use caution may cause heart palpitations and irritable bowels. We can't guarantee that last part. I don't know, but that lost 34%. Yeah. On a Monday and then gained a rip on close to 17% back on Tuesday.
So a wilder ride on that. I also saw some three times leveraged ETFs for non-US investors. Okay, so Scott, leveraged ETFs were just one of the many places where you saw the market turmoil, maybe in an exaggerated way. The cause of that turmoil is what we're going to talk about now. There's a company here in the US called OpenAI, and it's the creator of something called ChatGPT.com.
That's an AI tool that became available to the public and I think was part of the gee whiz moment that got everyone so excited about AI and its potential. And now there's been a spending race on AI infrastructure by big companies, and that has propelled a narrow group of U.S. stocks much higher. It's really been supporting the whole U.S. stock market.
There's a company in China trying to do something similar to OpenAI. It's called DeepSeek, and it's been around since 2023. It launched its latest model in December, but what got Wall Street's attention was a research paper published on the same day that Donald Trump was inaugurated.
And it was about a deep-seek AI model called R1. And it showed advanced abilities, but mostly what people paid attention to was the price. There are restrictions on China buying the best AI chips, which come from America's NVIDIA. And so China had to, how should we describe it, Jackson? They had to Gilligan's Island this thing, right? They had to work with... If people listening aren't familiar with Gilligan's Island, I think that I'm...
contractually limited from mentioning it any more than I already have in the history of this podcast. But, you know, folks stranded on an island have to make do with what's available. You know, there's a bicycle with a monkey and coconuts. And next thing you know, it's a washing machine, that sort of thing.
So the thinking here is that if China can make such a big leap in AI so cheaply, why can't everyone else? Which means why can't companies save vast sums on their spending on AI chips and infrastructure? And if they can do that, then maybe NVIDIA won't make as much money.
And by the way, it and excitement around it is what has been propelling the U.S. stock market higher. That's why shares tanked. Now, whether you think that market move was an overreaction or not, observation number one is that the U.S. stock market looks expensive. If it weren't, investors probably wouldn't be this nervous.
I show the S&P 500 at more than 25 times trailing earnings. Nvidia gained 239% in 2023 and then another 171% last year. It's now the most valuable U.S. company. This stock market rise has been the most narrow one we have seen in the U.S. since the dot-com bubble. So investors are sensitive to anything that could cause trouble for that.
In a report, Piper Sandler called the market's move overdone. It wrote that commercialization of DeepSeek appears unlikely. DeepSeek is a Chinese-based company. The origins of its model are largely unknown, and that raises security concerns. Piper Sandler also pointed to something called Jevon's Paradox. I hadn't heard of this before this past week, and now I'm hearing about it everywhere.
They write, even if DeepSeq's model is implemented despite the concerns listed above, we feel that inferencing demand will likely pick up given the lower costs associated with compute. In other words, if this thing really is so cheap and if a lot of people start using it, that means everyone will want more AI. And with it, they'll want more chips.
Piper Sandler writes, we conducted checks with multiple management teams in our coverage and came away with the belief that the need for accelerated computing as well as connectivity slash networking remains strong. In other words, they view the spending cycle as intact.
B of A Securities calls DeepSeek noteworthy but not a big course correction. They recommend that investors buy NVIDIA, Broadcom, and Marvel. They continue to predict very fast growth in the market for computing power. And they say the arbiter of this in the near term will be commentary on capital expenditures from large cloud and enterprise customers in coming earnings reports.
B of A, like others, also makes a comparison with Sputnik. Jackson, how old do you think I was when Sputnik launched? I don't think you were alive.
God bless you for saying that. And for not taking too long to do the math in your head. Sputnik is a Soviet Union satellite launched in 1957. If I were alive then, that would mean I would be pushing 70 today. And I'm not. I'm a smidgen over 50. I recently heard my approximate age described as...
the freshman class for old age. It's like orientation. Made me feel youthful. Anyhow, Sputnik launched in 1957 and NASA was established when? The following year. And the initial budget was $89 million with an M dollars.
And by 1966, just eight years later, that budget was $5.9 billion with a B dollars. It had climbed from 0.1% of the federal budget to 4.4% of the budget. And why was that? Because America had a holy Sputnik moment and decided we better get astronauts on the moon in a hurry. And we did in 1969.
I remember it well. I was smoking Lucky Strike non-filters and watching on my black and white TV. It's a joke, Jackson. I wasn't born then either.
Okay, so if deep seek is a Sputnik moment, maybe US spending will go up and not down. Or maybe it will be some combination of the two. Maybe the AI buyers will find a way to make do with cheaper chips in the near term, but then need more expensive ones down the road. It's difficult to say, and that's why I wanted to use my AI show business conversation with Gavin Purcell to ask him about AI and deep seek.
Should we roll that conversation, Jackson? Yeah, let's roll it. That's what you're going with? Let's roll it? I mean, this is a show business guy that we're going to be hearing from. What else you got? How about boom, baby? Keep trying. Initializing. Buckle up. For now, probably just roll it for now. All right, roll it. Gavin, give us, for the benefit of people who haven't listened to your podcast, aren't familiar with you, can you give us a quick intro of...
How you got involved in this subject, how long you've been doing it, what the experience has been like. Yeah. So, I mean, I have a weird background in that I was the showrunner of the tonight show starring me Fallon and worked in TV for about 20 years, but was always tangentially involved in digital media and kind of grew up nerdy and interested in tech. And so I,
I met my co-host at a place called G4, which was an old channel. He was the host of Attack the Show. I was the showrunner of it. And we stayed friends for about 20 years. And then after I left The Tonight Show to kind of do a bunch of startup stuff, he and I reconnected and we both got really excited by GPT-3, which was one of OpenAI's kind of earlier releases. It was pre-
chat GPT by about a year. And we were both excited about it because it was kind of this really interesting thing where AI suddenly could kind of write more like a human. It was still pretty rough at times, but it was the first stage where we saw something really interesting in it.
So Kevin and I reconnected and we were like, well, maybe we should do a startup in this space. He and I both love to build things and just do dumb little stuff. And if you watch our podcast or listen to our podcast, you know that we play with these tools too and don't just talk about them. And we tried a couple of things and eventually we both got together and we're like, you know what, let's just like make media about this because that's what our backgrounds are. And the last like year and a half have just been this kind of steady progress, right? And now we're at this point where people are starting to realize the impact
of how this could not just affect us,
but affect the world, affect the stock market, affect everything else. Because AI is something unlike any technology, at least for a long time, it's come before it, in that it kind of seeps into every aspect of our world. It can seep into every aspect of our world. And there's a lot of people who have hyped technologies in the past. You know, we've heard a lot about, obviously, crypto was a big hype wave. Before that, there were a bajillion other things, SaaS businesses, all those sorts of things, which all became things, right?
But this feels like fundamentally much more to me like electricity, which I think is a metaphor a lot of people have used. This idea that you had a world that was one way prior and now is a different way. And it's going to take a while for people to figure out how to use it.
But you suddenly have electricity versus not having electricity, which is kind of a big trade-off. There's always a contingent that's saying, yeah, this is overblown. I mean, people are making too much of it. It's not going to – it's a flash in the pan. Yeah, they're spending a lot of money now, but that's going to slow down. It's not going to be –
In other words, compared with what's happened with the stock prices, they're saying that those moves are not going to be supported by what's to come. And it sounds like if you're comparing this with electricity, then it sounds like you're in the camp of saying, no, this will continue to be a big thing, attracting a lot of money for many years to come.
I will say, just everybody keep in mind, my source, my name of my podcast is AI for Humans. So I definitely have some sort of value in saying this, but my take is this. Investment and technology advances are different. And I think that's an important thing to realize. If you go back to look at what money was poured into the internet in the late 90s, I grew up through that, you know that there was a bajillion dollars that were thrown away during that time, right? Like there's a lot of money that just was wasted online.
And now we may be in a slightly similar place. I will say, I think right now, especially after the last couple of days, I mean, we all know what happened yesterday on the stock market. It's Tuesday morning. And yesterday, I think NVIDIA lost like,
some insane amount of money in a one-day loss and a huge, huge loss across the market. I think the R1 news has been overblown for what it is. And I think one thing that's really important for people to know about R1, which in R1, if you're listening, you're not totally familiar. This is the Chinese model from DeepSeek. It's a Chinese company that released an R1A reasoning model, which is another LLM, kind of like Jack GPT, that can think through its thoughts.
And this is the thing that tanked the stock market yesterday. And the reason why is because it was developed for much cheaper than all of these other AI models that they've spent billions of dollars on developing at OpenAI or Anthropic or Google.
There are some caveats to that. It's basically a model based on another model. So they use the other models, specifically Lama 3.1, which is Meta's model, an open source model to kind of make it better. But they were able to do that for very cheap. In fact, I think that the training cost was somewhere around $6 million, which is like a factor of 10 or less more than what a normal model would cost to do.
The argument here is that you might be able to do some of this AI research and AI models on lesser graphics cards because DeepSeq trained this model on a lesser NVIDIA card, which was supposed to not be possible, and they did. So going back to your hype question, there's a term going around now called Javon's paradox. Are you guys familiar with this? Do you know what Javon's paradox is? Or Javon, it might be Javon Javon. So
I just learned this term yesterday for the first time. I'm sorry if anybody's listening to this who's now heard this 15 times because it's something out there. It's basically, Satya Nadella from Microsoft mentioned this in a tweet the night that all this news dropped and he knew the stock was probably going to go down the next day.
essentially it means that there's an argument that if you make something cheaper, it will bring the value of that thing down. But because you've made it cheaper, the actual use cases of it go up. And so more people want to use it. And the value then goes up because more people are doing it. And what's cheaper in this case is compute, right? And compute is like the digital oil of the AI space, meaning that
you can do more things and ask more questions of the AI and it can think longer. And in that case, if compute is suddenly cheaper and suddenly everybody can do this, I do think these build-outs, like the $500 billion build-out that Sam Altman, SoftBank, and Oracle are planning may sound like a crap load of money right now, but to be honest, it is like...
Maybe just putting down the right places where you have to go in order to get to this world that we're headed to. The idea is that everyone is spending oodles of money with Nvidia for these very expensive chips because they're the best on the market for AI and everybody wants to have cutting edge AI. And suddenly a model comes from China where because of necessity, because of the restrictions on selling the best chips into China, they said, well, we've got a model that uses older chips.
And it doesn't need the latest, greatest thing. It didn't cost a lot of money, but we've built better software. And so now it works as well as anything out there.
And the question becomes, if China can do that, maybe people don't need to spend all this money on these chips. Where do you stand on that? Is this development out of China the kind of thing that can interrupt the spending? Are we entering a world where we can do it all with software? We don't need the latest, greatest chips? Or is this just a moment that will pass and everyone will still need that latest generation of chips from NVIDIA?
I think personally that this is the entry point to the next stage of AI. And I think a lot of people in San Francisco have been thinking this for a while. And I think that the world at large is going to start to feel this. In fact, Sam Altman just yesterday in response to DeepSeek said that they are going to move releases up in response.
So I think what you're going to see is the edge move faster, further. OpenAI has said publicly that they believe these reasoning models will scale very fast, meaning that they will get better much quicker than almost the other models. Because
One small thing, and I know this is a big concept, but there's two different ways to improve AI. There's the reasoning side, which is called test time compute, meaning that you're asking the AI model in time to spend compute on thinking about something. And then there's the foundational model, which you're training with a whole bunch of data at once. And those are two different pathways.
The reasoning model side, it sounds like, is not stopping, that there is a continual growth. In fact, they've planned already an 04 conceivably at opening eye. There's been rumors of this. That could come out by mid to late this year. So I think my long answer to your question is that this is open source catching up and specifically China catching up, which I think is probably what tanked the stock market. But the edge is going to keep moving and probably moving faster.
Thank you, Gavin. We'll hear more from Gavin in a moment after this quick break. Exclusively on ESPN Plus, UFC 312, Saturday. Reigning middleweight champion, Drakus Duplessis, defends his title in a rematch against Shawn Strickland. And Zhongwei Li defends her strawweight title against undefeated Tatiana Suarez. UFC 312, Saturday at 10 p.m. Eastern.
Buy it on ESPN plus.com slash PPV. With new McValue at McDonald's, you always get more than you expect every day. So even if gas prices go up, you can buy a double cheeseburger and add a McChicken for $1. Or before game time, you can get $5 meal deals for the group. And if breakfast is about to end, get deals in the app to save before the bell. The choice is yours. And the choice with McValue is always more.
Prices and participation may vary. Valid for item of equal or lesser value must opt into rewards for app deals. Welcome back, Jackson. I am, to put it in ETF terms, 3x levered, excited to talk about the intersection of AI and show business. Let's get back to my conversation with Gavin Purcell, host of the AI for Humans podcast. Rocking and rolling. They're getting worse.
Give us a sense of where are we in terms of what AI can do. And what I mean is, if I think about the technologies that came along during my life that I saw, I am 100 years old. So I remember when, you know, going from VHS to DVDs, what does it do? Well, you can skip forward and it takes up a lot less room. So you can, you know, your collection of movies takes up less space.
and the picture is better. Okay, the internet, you can go online and everything's real slow loading, but then you could see it progress. It became faster speeds and more people jumped on the internet. It became robust. The iPhone, right? It started off slow and it was amazing from day one, but then as each model came out in those early years, it was like, wow, now it's faster. Now there's a bigger screen. Now there's Siri and this and that. So
What is the latest about something that AI does now that it didn't used to do in your experience a year ago or something like that? What's the cutting edge of what is happening in AI and what's the cadence from here? Where do you think we're headed?
I think there's two really interesting ways to look at this. I'll talk about the part that I'm a little least familiar with because it's not my background first, which is AI is getting very good at writing software code. And I think this is something that people are kind of surprised by. These reasoning models particularly can write complicated software code and can create programs in a way that you wouldn't be able to do it before. In fact, there's a whole...
movement in the AI space on the rise of the idea guy, which is such a dumb, funny idea. But you know, in tech, there's always this conversation around the engineer and then the idea guy and the idea guy is kind of useless. He just comes up with the idea and the engineer does the work.
But now with the software writing of AI and how good it's getting, there are a fair amount of people who are starting to write small programs and being able to do stuff that were not coders at all before. And, you know, SaaS companies are a little worried about this because SaaS companies specifically are created as software as a service, right? You're buying software for a larger company. What this movement kind of does is it allows people to create bespoke software for themselves, right? And as this gets popularized,
better, the idea that the AI could create, say I have an invoicing need that I have, and I want to create a specific program for me that knows these different things. Eventually, and really not that far from now, in fact, you should go try this or anybody can try this right now, is like go to OpenAI's O1 or one of these other reasoning models and say, hey, can you code for me an invoicing program? And you might get a few bugs in it, but it'll come back and you'll try it. And it's like, oh, this is actually not bad. So to go from not bad to like,
destroying the whole SaaS market, it's not insane to see that that's a world where we can get to. The other thing I think that it's really good at in my world, and I think has people a little bit freaked out, is video generation. And this is a different side from a business aspect, but
You may have heard of OpenAI Sora, which was like a big news story that came out when Sora was released as their video model. There are Chinese video models, speaking of China, called Kling and Minimax, which are remarkable. And now all these are doing is creating, say, five to 10 seconds right now. And it's a little bit funky, but not nearly as funky as it was a year ago.
I think a year to five years from now, television and film production will be completely different. First of all, I think all the big players will integrate these tools within their production pipelines. But bigger than that, I think you're going to see one to five person teams make films that are super compelling, really interesting. I think you're going to see a lot more of these films. I think it's about to transform the entire Hollywood experience because it's
It has a, you can do stuff faster and you can try more things. So those are the two areas that I think are the most interesting to kind of follow. If you're really interested in seeing how, how this stuff is progressing fast, obviously there's a bajillion other things you can do, but those are like distinct changes to the world that I've seen.
I have not played around with AI video generation or, you know, with much of it at all. I saw a video of a preposterously fat man. Okay. And don't before anybody writes me, I'm not fat shaming because the end of the story, he's not real. Okay. This person was exceptionally fat and shirtless and on skis.
on a snowy mountain and, and doing tricks and flips and everything just incredibly agile. And I had no idea because if you scroll a social media feed, these things just pop up and you don't know that they're AI. And then you have to look around and sometimes the name of the publisher of it will say such and such AI, or sometimes in the comments, people say it's AI, but I didn't think it was. So I'm here trying to picture a whole backstory for this gentleman. I'm saying like, this guy was a
he was an incredible athlete and just, you know, he fell in love with junk food or what, what happened here? How did this guy, how's it, and how's he making these movements? But yeah, you just couldn't tell anything. I know. And then when I first saw AI videos, they'd say like, look at the fingers, like the fingers are always weird or that, or the hair or something like that. What are the tells that you've seen? And are there tells that are now fixed? Like the AI used to get this wrong and now they're now it's doing a better job with that.
The tells are almost impossible now. And I will say this is a really tricky thing because it used to be, yeah, that used to be fingers were bad, faces were bad. The tells now, if the creator of the thing is good, what they're doing is selecting for the best shots. And I will say as somebody who makes a lot of AI video things or tries to make a lot of AI video things,
You do still get bad results oftentimes, and you have to be selective of how you pick the stuff. It's kind of a crapshoot sometimes, but, and you'll get, you know, it's a little bit of a, they call it a slot machine, right? Cause you're not exactly sure what you're going to get out. The tools are getting better though. In fact, I made a video with VO two, which is Google's new tool. It is not out yet. Um, I got it on an early list. It's like an early release, but Google's new AI video model is by far the best I've seen so far.
And I think when this comes out, it's going to kind of blow people away. So I sent you guys this, but I made a video for big Mayo, which is the idea that like big mayonnaise is, is needing to get out there and show what else Mayo can do and what it can do besides being put on sandwiches. So I used VO two for all the shots.
I used AI sound to make the sound and I used an AI voice to do the voiceover. And I just made what would be like a dumb commercial. And, you know, this is the result of it. Mayonnaise, the perfect compliment to any sandwich. But why stop there? Start your day with mayo. Try spreading it on your bacon. Mayo on bacon? I had not thought of. Oh, just wait. We at Big Mayo believe you can use mayonnaise where so... I mean, it's a great, you know...
Like SNL style, you know, Saturday Night Live style, spoof commercial. But you're telling, is nothing in that real? Like the people aren't real? Zero. Nothing is real. The only thing that is real is the idea was mine and I wrote it, right? So like, and I do think this is an important thing to remember for people when they think about AI. Because I, as somebody in Hollywood who talks a lot about AI, has gotten a fair amount of crap from people that I work with or have worked with in the past.
mostly because there's a lot of hesitancy towards AI in Hollywood. And I will say, you know, the training of these models was not clean. And I think that's an important thing for everybody to know is that there is an original sin in AI video and even in chat GPT, which is they kind of took everything, everybody's stuff, and they put it together. Now, there are legal definitions of what that is and what that isn't, but that is kind of the beginning stages. All that said...
I really think that like doing this now is capable by almost everybody. This video that you just watched took me about four hours in total to make. Now, granted, you know, you talk about an SNL parody, like my brain, because I worked in comedy TV for so long as a comedy writer, kind of immediately goes to ideas like that. And so like the idea conceptualization and writing of it wasn't super hard for me because I was kind of jumping off of something that, you know, other people have done for years and years.
but the video generation, the audio generation, the music, all of that was done within four hours and the editing. So I did the editing, I did the writing and I came up with the idea, but everything else was AI. But there are a bunch of shots of hands, right?
In this, in this video and hands, you know, fingers are supposed to be the thing that clues you in. So I'm looking at a sort of closeup of two hands putting mayo in a cocktail for people who want to, who want to, and I'm seeing, I'm seeing five fingers on one hand and five on the other. Fingers have been solved. I hate to tell everybody out there, but fingers have been solved. They are not the thing you can tell anymore. I, I always tell people when we talk about what to believe and what not to believe is AI, you know,
We're getting into a stage, and by the way, if you haven't already started thinking this way, you should, but we're getting to the stage where you have to kind of doubt almost everything you see before totally believing it's real. Obviously, that has huge implications for our world and has already started to affect our world in some ways, but I think we're going to be entering into a space very soon where everybody has access to tools to make things look like they want to make them in any sort of way. And by the way, of note,
Google's tool, VO2, Sora, the American models all have restrictions on what they can output, what they can say, what they can do, because they won't let you do celebrity outputs with the video models. Chinese models, not as much. You know, Minimax and Kling, both of these models are from China.
China is much looser with IP. They definitely have had a looseness around it. In fact, I've been telling people lately, there's a great book I read called AI Superpowers by a guy named Kai-Fu Lee, which really talks about the AI fight between China and America that I think is really useful. It's a couple years old, but it's worth reading. So anyway, this AI video thing is not going to go away.
I do think the dangers of it is slightly overblown because there was a huge reason, you know, a lot of this stuff didn't come out pre-election because everybody was worried we were going to get some viral piece of content that showed one of the presidential candidates doing something everybody would believe it. I read a scary study one time that said not only might some people believe it, but even when you show people a video, if you show them a video of someone doing something bad and you tell them it's fake,
and they just watch it a few times, it's like, it still affects the way they feel about the person, even knowing it's fake. This is where it's a societal issue almost more than it is anything else, right? Like, I think we've adjusted to this idea of truth being true. And, you know, you can take your own story in whichever way you want and believe whatever side you want. But like,
this only adds to that, right? Like it only adds to that. I tend to think that like, okay, how cool would it be if like the individual creator can tell a story now that they couldn't tell before, or they could do something faster, or they could explore a world they couldn't explore. Like, you know, giant people on water skis, like that's a funny thing to think about that, like, you know, you wouldn't necessarily be able to see in real life. I also think it's
just, we're going to get weirder. Like I think things are going to, they're going to see edge cases. We're going to see strangeness. You know, I think the opportunity here is for AI video is to like discover new voices and hear from new people, but you're also going to see a just massive amount of crap. And I, when I say that it's like, you know, you, have you heard the term AI slop before? Do you know what that is? No. Okay. So AI slop is a derogatory term.
Mostly for what's shown up on Facebook in the last six to 12 months. Slop is almost never a flattering term. Yeah, exactly. Jack, have you heard of Shrimp Jesus? That was like a common one that just appeared all over Facebook. Shrimp Jesus is a good example because what it is, is pictures that are generated with AI. And by the way, Shrimp Jesus is insane. You should look it up if you haven't seen it. It's a picture of Jesus as a shrimp underwater and it's clearly AI.
But that picture got millions, tens of millions of views on Facebook. And so the idea of AI slop is that because we have these tools, kind of crap can be turned out much faster for views and for algorithm clicks, right? So a lot of what you'll see right now on Reels and TikTok,
if you're in those areas, is you will see AI-generated video getting a ton of views. In fact, there was a video, you know, the inauguration, Barron Trump, somebody made him like, you know, 12 feet tall walking in behind Donald and like things like that. Those are the things that just travel fast. So you see a lot of AI video generations in that space as well too. And AI slop is going to be a problem we're going to be dealing with for a while. And I think the good news is that
Hopefully the AI will help filter that out, which sounds pretty crazy that the AI is going to be fighting the AI slop, but like,
that's the best we can hope for. And I think ultimately we have all the tune our own eyes to kind of figure out what we want. Some people might want shrimp Jesus. I don't, but I think some people, you know, could want it. I've got some reservations. I'm okay with 12 foot Baron cause he's already more than halfway to being 12 feet, but shrimp Jesus, I gotta, I gotta talk to my preacher about where we should come down on that. I've got some reservations, but,
But what happens with... You've got a show business background. So what does this mean for the future of show business? It takes all this time to create this visual artistry and all this shooting and filming and sets and crews and all this stuff. Do we just... In the future, is it just...
one lazy idea person sitting there saying, I got an idea for a show about whatever, whatever. Hey, computer, make it happen. So first of all, I'll say that I don't believe that a one lazy person will make anything interesting. Let's, let's put it this way. Like, I think there's going to be a lot of lazy people out there who are, I had my, I had my hopes up there too. I thought my day had come.
Well, the funny thing is so many people in this, in the space often talk about the idea of like, Hey, you're going to be able to make a video, a movie of yourself as the rock doing the Charlie and the chocolate factory story. And like, that's like a lot of the AI video companies pitch. And I don't think that's at all compelling. I think people want to listen to and watch other people's stories. I think people want to escape into somebody else's things. The thing that I think is going to, I'm picturing, by the way, as you say that I'm picturing, uh,
Martin Scorsese, who was like beside himself about the Avengers movies and Thanos and the computer generation. Imagine how he's going to feel about this. No, I think it's interesting, right? Every filmmaker is going to have a different reaction. I mean, some people really embrace this stuff. James Cameron, by the way, has all in on this stuff. Like he's come out publicly and said he believes that AI is going to completely transform the production process.
I think the part of the problem with the entertainment industry right now is it's very top heavy, meaning that they are going for these home run hits mostly, right? The middle of the entertainment industry has already kind of fallen out. The idea that like the $50 million Judd Apatow comedy doesn't get made and doesn't put in theaters anymore. That's already happened and it's gone. Cable TV, gone. Like the middle kind of programming. So you see the very high end stuff happening.
meaning the Game of Thrones or the Marvel movies or Despicable Me 4. That stuff comes out, does well, made in the theaters, all IP that exists, not a lot of original IP. And then you see the very low end, which is like these kind of indie films like Onora, or you might see a YouTuber do really interesting kind of work and be something that's super compelling to watch. The middle is already gone.
So I think what's going to happen in my mind is that the AI tools are going to let the low end start to compete with the middle again, and maybe the high end. And I think that's a big deal, right? Like if you think about like a five person studio versus the idea of a, and then I'm not saying it's going to replace Marvel movies right now because Marvel movies are made with like a thousand people and they have 250 video graphics and video people, right?
But five people is a very small crew. Now, also, there's always going to be like a contingent of Scorseses. I want human-made stuff. And I think that's going to be something as well, too. I think you're going to have like 100% organic human-made logos on this stuff that is going to mean something in the future because of that. We have the $200 million movies that are sequels and prequels and spinoffs, and they're reliable franchise movies. Yeah.
And there are these independent and new movies, but they're out there costing a couple of million dollars and they open in 14 theaters in New York and LA. So they're not maybe in your town, but they're out there somewhere. And so what you're foreseeing is this future. We don't have, as you said, the $50 million movies, the romantic comedies that show up in everybody's movie theater. But you're saying in the future, maybe we'll get back to that. It'll just be cheaper to make them because we'll be doing a lot of it with AI. Have I got that right?
Yeah, I think that's right. And the other thing is you're going to get to the IP faster, meaning that you could start an original idea and turn it into an IP much faster than ever before. But the example I use of this is a little stupid, but are you familiar with Skibbity Toilet? Did you find that phenomenon? Oh, I'm a Skibbity Toilet. I've got a PhD in Skibbity Toilet.
- Oh, good. Well, I'll have to explain it in case the audience listening doesn't know what it is. It is a very funny, weird, you know, mostly 12 year old boy audience. - Yeah, that's what I just know it to be a thing that young boys say. - Oh, you don't know where it came from? - No, I've got a young boy in the house and everything's Skibbity Toilet and Sigma, Ohio Rizzler and all this kind of stuff. - Welcome to the world, yeah. - But tell me the origin story. - Okay, so this is really interesting. So Skibbity Toilet is not just a saying, it is a YouTube series.
And there is a guy, a single guy in Eastern Europe who created this using the Unreal Engine and a bunch of characters that he made in it. And it is an ongoing series, although it really blew up like about six months, maybe even longer ago. And what I talk about in this thing is like the speed from that, from a zero point of no one knowing what Skibbity Toilet was to what I would call an IP, which is like a worldwide, you know, actual thing that people are aware of that you very much could understand
go sell in Hollywood, and I think the Skibbity Toilet guy did in some form or another, went insanely fast. The faster you can generate ideas that are sticky, meaning that you have an interesting idea, you can execute on it, and then you can get it out into the world and see what works,
I think those IPs are going to spin up faster and faster. And obviously, going back to the AI slop conversation, you're going to have a lot more of it out there. So it's going to be a matter of like, how do you get people to pay attention to it? What resonates and all that stuff? I think we're going to enter a kind of a golden age of interesting things. Any parts of show business that you think are going to be enduring? I mean, we're obviously always going to have live sports. Jimmy Fallon, the world, that guy can do it all, right? The world is always going to want Jimmy Fallon.
but what else out there is safe? - I honestly think comedy is a pretty safe space in a lot of ways, because I think novelty is really one of the last things we will have as our own people to control. Because I think novelty is what drives people to want to see something.
And novelty plus emotion makes compelling drama. Novelty plus laughs makes compelling comedy. And I don't think AI is going to be that good at it for a long time, as we talked about before. Like AI is very good at like not just copying. They can they can come up with their own things, but like kind of interpolating and distilling what human and behavior was. Novelty is something that comes weirdly out of a magic place we don't fully understand. Now, will the AI get there eventually? Maybe. But like
I think the future of Hollywood is much more about the human individual creativity, and I hope it means we get a lot more interesting people
dramatic, crazy things that you wouldn't necessarily expect to have come out of Hollywood before. And I feel like I'm pretty excited about that world. I think some people don't see that because they're just mostly worried, and I understand this, about a business they worked in for a long time having less paying jobs. I think there's a very different side of Hollywood, which is
the gaffer or the person that has worked on these giant productions their whole career, which, you know, if a lot of more of them become virtual or AI driven, doesn't happen as much. And, you know, that goes for the VFX guy at the Marvel movie as well, too, right? There's 250 VFX people. If suddenly there's 100 people there, that's a very different world or 50 people. It's a very different world than it was before.
Thank you, Gavin. And I want to thank Scott for the leveraged ETF question earlier. And thank all of you for listening. Jackson Cantrell is our producer. I made the mistake last night of looking up skibbity toilet on YouTube. And I feel like I lost half my brain cells. No. What did it look like for people like me who haven't seen it? Give us six words. Video game toilet head singing.
That's five. You've got another if you want to use it. I'm good. Badly.
You can subscribe to the podcast on Apple Podcasts, Spotify, or wherever you listen. If you listen on Apple, you can write us a review. If you want to be like Scott and hear your question answered on this podcast, just tape it on your phone. Use the voice memo app. Send it to jack.how, that's H-O-U-G-H, at barons.com. See you next Skibbity Week.