Creative Flux

Pierson Marks (00:00)
24, episode 24, Creative Flux. Welcome back, everybody. It was cool. Last week, we got actually a good number of views. We were talking about OpenAI's Code Red. We were talking about Google Gemini, a lot of stuff to talk about that.

Bilal Tahir (00:02)
Yeah. Welcome back guys.

Pierson Marks (00:14)
today which could be very cool so we'll continue off last week's episode but if ⁓ you know it's your first time here, I'm Pierson

Bilal Tahir (00:23)
I'm Belal, nice to meet you guys. We meet you for our regular listeners.

Pierson Marks (00:25)
and we're the creative master group.

So if it's your first time listening to Creative Flux, what we do is focus on generative media, AI, images, video, audio, world models, all of that cool stuff. Sometimes things are tangential like, know, like GPT 5.2 came out today. We're going to talk about that. Some other simulation stuff. But overall, we try to theme this towards a generative

of media space and focus on that because that's what we're both passionate about. But today we have a packed agenda. mean, we just mentioned about it, there's GPT 5.2. There was like the Disney Sora agreement, which we hinted at last week, which is crazy, you know, we were talking about Zootopia characters. So we're going to dig into the details of that and ask for today. There was a announcement between 11 Labs and Meta. So about audio and like

all this stuff. I mean, we have a lot to talk about. yeah, so what should we? Yeah, what should we start with? I mean, I think honestly, let's just talk about 5.2. Let's just like get it out of the way. I came out an hour ago. So hot off the press. So we haven't even had like that much time to look into it. So for the context, mean, last week, okay, so Google did this.

Bilal Tahir (01:20)
Yeah. Yeah, I don't know. Where do you want to start?

Yeah, it came out like ⁓ an hour ago, so very fresh. Yeah.

Pierson Marks (01:44)
I mean, they've been dominating. You we've been talking about this for a while. If you've been listening to us, not financial podcast, but we're both pretty bullish on Google in the AI space. But OpenAI, Sam Altman, released his code red. And it means like, we need to focus more on the core offering of like how good our models are and like kind of forget about everything else. And they come out today with GPT 5.2.

Bilal Tahir (02:10)
And from what I've heard over Twitter, the Twitter grapevine is that apparently this release was pulled forward. It was supposed to be happening in 26, but Sam often overruled some employees who wanted more time for testing because he wanted a response to basically Gemini and Opus, which again, this is why I'm a little skeptical because the results are amazing. The benchmarks look awesome, but the question is, did they just benchmark max or is it actually good?

reports it does seem like a great model I mean it obviously will be a good model coming from OpenAI but at the same time is it a step change and I think that's where you know I've been a little skeptical of OpenAI's releases you know for a while but you know they're still like the codecs honestly they did a great job in getting good at coding and stuff so

Pierson Marks (02:47)
Yeah.

Bilal Tahir (03:00)
So I mean, I'm willing, you're right now, no, you know, I have no opinions. We'll see how it goes. But it's very interesting how they feel like they need the need to respond immediately. And apparently there's another model in January, which I think will be a bigger release, which is going to be more of a step change. So.

No, they don't. It's like all these open AI hype machines are in rumor. But apparently they have two models for these. They're going to do one right now and then one in January, which is going to be a bigger like ⁓ thing.

Pierson Marks (03:25)
What do think it could be in January

if it's a month from now? Like, is it going to be 5.2.1?

Bilal Tahir (03:30)
Yeah,

I have no idea. mean, this is like, that's where there's so many rumors and I've gone tired of kind of keeping up with it, but it's opening eyes. who knows what's going to happen. But I hope so because you know, the more models, you know, the better for us. We didn't really get a chance to talk a lot about Opus. We talked a lot about Gemini, but that also, I've been playing, using it regularly for coding stuff. Amazing. Opus 4.5.

is pretty amazing and I actually think it's fun a lot of people are calling it like the jump from when sonnet 3.5 came out you know to other models they're calling it a similar jump and I will say I mean I wouldn't quite phrase it that way because I feel like it still can be naive and run into loops but it it does you throw stuff at it it just keeps going so

really cool stuff. do think coding at least has been, the jumps have been accelerating, which is awesome for us.

Pierson Marks (04:23)
Yeah,

it is interesting. know we'll touch on coding real quick just because I do think it's super cool how coding is the killer app of LLMs. And you always wonder with a paradigm shift, what is that killer app? And it's pretty obvious now, I think it's going to be coding. And then for the generative media stuff, think we're still yet to be seen what the killer application is. But like.

More and more I use these things and I'm just more more impressed. As long as I'm not lazy as a user and I describe what I kind of want to do, it kind of can fill in the gaps of misspecified requirements and actually do things. like, oh, I didn't even know that you could do it this way. And that's the key thing. When you're an engineer and you're writing code, you want to model as a pair programmer who's a really smart pair programmer.

You still want to have complete understanding of kind of what's writing. But when the model suggests solutions that you weren't even aware of, like there's an API that, hey, you wrote these 10 lines of code over here, but in Next.js 16, it's a one-liner. It's super easy. super like, it's like, that's super cool. Because what do you even ever thought to go, was like, this thing works, but now you're helping me. Not only are you solving that problem,

Bilal Tahir (05:35)
Right.

Pierson Marks (05:45)
You're bettering me, my knowledge, I'm learning from you. I'm also writing better code that's going to be more robust. It works better. that's cool. When I start seeing that, I Opus 4.5, Gem 9.

Bilal Tahir (05:55)
definitely. And

I've noticed that too. mean, it's like, and especially when the model challenges you on your assumptions, you know, it's not just as sec of Antec, which is great because it's like, actually it feels like you're just talking to another engineer, which is amazing. And to your earlier point about the leaps in generated media versus code, think one lens you can view it as basically anything that's verifiable is going to get solved basically, because you know, the whole RL training paradigm is like, you know, you have two plus two equals four. And if the model gets

Pierson Marks (06:17)
Mm.

Bilal Tahir (06:24)
three, you know, you can automatically check that and that's why coding is, you know, getting solved so fast because it's so verifiably. There's tons of data training data out there and similar to that other fields like math and hopefully chemistry and physics will, I do think I'm very bullish will get solved on that note.

something just happened, Terence Tao put out, who's one of the most famous mathematicians alive, he put out this, an Erdos problem that was unsolved since 1975, just got solved. And a lot of it was used. Erdos problem, Paul Erdos, he's a very famous, he was a very famous mathematician and he was one of the most,

Pierson Marks (06:53)
What problem? Wait, what problem?

Bilal Tahir (07:05)
He has like a shit number of papers like he's written so much here very forget the word for it, but ⁓ he's written so many papers that Everyone in the mathematics research field has something called the Erdos number which is like the degree of freedom like if you if you if Paul Erdos wrote a paper with Someone and then you wrote the paper with that person your Erdos number is one So everyone has like an Erdos number like I I wrote a paper with someone who wrote a paper with someone who wrote a paper with someone who wrote

it with all those so it's like a thing you know it's like the seven six degrees of Kevin Bacon which is comes from that which is like Kevin Bacon's been in so many movies so people like you know I've been in movie which who's been in a movie with Kevin Bacon so anyways Eros had this problem in 75 and it's been unsolved since and they use a lot of AI tools and obviously like human so it wasn't like totally automated but they use a lot of AI tools to solve this within 48 hours and we're gonna see more and more of this and this is I don't people don't realize like how

Pierson Marks (07:35)
⁓

Interesting. That's All right.

Right.

Bilal Tahir (08:03)
recent these, because of how recent these capabilities have been unlocked, that we have yet to see all this stuff coming around. I was watching, again, not to go on so many tangents, but I was watching the documentary, The Thinking Game, which is a documentary of Demis Asabis and his journey to create DeepMind. It's amazing.

Pierson Marks (08:22)
you watched it. that's cool. I haven't seen it yet. Yeah. Yeah.

Okay.

Bilal Tahir (08:25)
you

up and it's it's free on YouTube now do a deep mind put it up there definitely recommend people watching it but you see how like in when they created half a fold which is like the algorithm to fold protein folding that just got solved like four or five years ago like it was an unsolvable problem for 50 years get solved now we don't even think about it but that literally happened four years ago then deep mind basically folded every protein possible put it out there

so that people can use it. And so the fruits of that, we haven't even seen what that's like, ⁓ what will be the downstream effect of that, because it only happened three, four years ago. So it's crazy what kind of a step change in terms of mathematical breakthroughs and scientific breakthroughs we are on. It's like we're on the tipping point of that curve that's going to go straight up from here.

Pierson Marks (08:53)
Right.

Totally.

It's

so wild. It's ending off the year with some bangs. It's Christmas time. And it's crazy. We talk about health care and those services. Definitely another area where it's such an antiquated industry that the processes are going to be changed. And I have a friend. I'm going to see her.

Bilal Tahir (09:21)
Yeah.

Pierson Marks (09:40)
later this year, family friend, and she has a PhD from UCLA. She did some crazy, crazy work with regards to the brain. she was the first person that I knew who talked to me about GPT. And this was back in early GPT three or two. I think it was GPT two. And we were at a Christmas party in 2020. Would it have been 2022 or 2021? I think 2020.

I don't know, one of those two years. Oh yeah, yes, yes, right. Yeah, ChachiBeezy at 22. So it at 2021. And I was like, this is crazy. And I didn't really think much about it. And I was like, whoa, you were on this. cause it was wild. She's not a programmer and it was super cool. Like she has a PhD, but like the now what she can do in, like,

Bilal Tahir (10:08)
That's how GPD came out in November 22.

Pierson Marks (10:27)
the life sciences industry is just, it's mind blowing. She's like, this is crazy. I can actually like test my hypotheses and not spend all this time. Like how do I build this thing? You still need to verify that your experiment is done correctly. Cause that's very important, but like also you can, you know, get to that first iteration really quickly and like just test a lot of different techniques. So with coding, but very, very.

Bilal Tahir (10:48)
I'm bullish. My personal like

would be awesome is like basically by the time I hit my 40s we solve aging and then I can go back to 21. So I get like a little taste of what it's like to be old and like in pain but not too much and then I can go back. God yeah. Nah I don't have the discipline and that kind of obsession that Brian Johnson has. Hopefully it'll just be we can just de-age ourselves without needing all his protocols. I do think it'll happen.

Pierson Marks (11:03)
As long as you don't look like Brian Johnson, I think that'll be amazing.

Right. Totally.

Okay, well sick. so, okay, GPT 5.2, Terrence Tao is the air dose problem. So watch the game. We're going to put these in the show notes for anybody listening.

Bilal Tahir (11:25)
Yeah, another

one real quickly. I forgotten the name, but there's a company called News Research. They've done open source models and stuff. cool. Yes. And they released a model. I forgot the name of it, but that was only 30 billion parameters. But it was like second in the platinum. Again, so math verifiability. These things are the.

Pierson Marks (11:33)
yes. N-O-U-S, right? Yes.

Bilal Tahir (11:47)
the amount of milestones being reached are just like this, like it's crazy. we're gonna basically have, probably have the world's greatest mathematician in your MacBook Air basically by the end of the next year or couple years, which is insane. I mean, if you think about like, where we were three years ago. So.

Pierson Marks (11:58)
Godly.

It's super interesting.

And I'm going mention this, this goes onto another tangent. talking about Google, I use Chrome. I'm on Chrome right now. Most people use Chrome. And Chrome essentially is an operating system. It's a browser. But at end of the day, the browser now is almost as complicated as the operating system. remember. ⁓ And when you can package,

Bilal Tahir (12:24)
Right, yeah.

Pierson Marks (12:29)
small models into Chrome where there can be a hybrid sort of intelligence where, you know, if we can correctly classify a problem into easy or hard, just like literally like just a guess, like is this an easy problem or a hard problem? And you have a route like a model router where easy problems can get solved on the edge in Chrome, not even in your OS, like not on

You don't have to have like Apple's, you don't have to have like Mac OS supported. could just be you download Chrome and you get a light version of Gemini built into Chrome. So if you're a web application, if you're an application developer and you have a lot of users on Chrome, you know, you can just build in probably like call LLM, know, simplifying, but like, you know, generate text. And if Chrome supports the local model in the user's browser.

Bilal Tahir (13:00)
Hmm.

Pierson Marks (13:20)
you can default to the local device. And if not, then you go to the server. And I think that's going to be this hybrid approach of inference that would probably be a little scary for Nvidia and like the clouds that are data providers. If we get smaller models that are more efficient run locally on the edge. But it's like something that we're not a lot of people are thinking about. There's going to be local inference on your device. I use Gemma on the plane when I have no like wifi. It's not that great, but like

Bilal Tahir (13:47)
All right. Yeah.

Pierson Marks (13:47)
It's kind of cool. I could do it. You know, I could I could like

have a conversation with a compressed compressed version of all humanity's knowledge offline.

Bilal Tahir (13:52)
Yeah.

100%,

100%. And to be clear, feel like Nvidia, it's such an expanding market that even if we use more on edge devices, which we will, still, they'll be fine. But to your point, it comes back to like, it's funny, because last week you demoed the split tab in Chrome, right? Where you can have now a split tab where you can have a webpage and then have a Gemini agent talking. I recently just updated Chrome and they have this thing called reading mode. I don't know if you saw.

Pierson Marks (14:04)
We.

Bilal Tahir (14:24)
what it does is lets you right click any website and it'll basically read out loud that page. So you can basically have a free audio overview which is probably using a baked-in TTS model, very light right? So these are the small things that you can do like you can you know maybe you know people can create podcasts out of web pages and stuff you know maybe we'll use them who knows such other part. So it's very fascinating that all these like features are being baked in.

Pierson Marks (14:50)
That is super interesting.

Bilal Tahir (14:50)
that is just going to

enrich the experience so much.

Pierson Marks (14:53)
I see this, yeah, reading mode, you right click, voice speed, voice selection, voice highlight. Yeah, wow, that's cool. So yeah, the way that we interact with information on the internet and with the world is going to change. So, Okay.

Bilal Tahir (15:07)
Yeah, and a lot of

this stuff is going to be baked in. Like speaking of baking in, one interesting deal that happened in the gender media experience is the 11 Labs and Meta partnership, which I wanted to talk about. So this is.

Pierson Marks (15:16)
Right.

Bilal Tahir (15:19)
This kind of actually surprised me because I'm kind of giving a history of what Meta has been up to. Well, Meta itself, obviously Meta is a huge company. have their own in-house AI research division, is, you know, they messed up a lot, you know, with the Lama 3 and all that. But still lots of data scientists, lots of ML engineers. And I think they have text to speech, but for some reason they went on an acquisition speech. First they acquired PlayHD a few months back and PlayHD was a comparator to 11 Labs. There were more folks on studio quality, voice

you know, so very like the professional clone in the lab labs, they just focus on that part. And I guess they just couldn't find PMF. So they just got acquired. And I heard that I'm like, all right, you more tech goes in house, medus, TTS is just going to get really good now. But then, and then they acquire another company called Limitless, like last month. Now this is an always on memory device thing. And they apparently

Pierson Marks (16:14)
Providing

hardware.

Bilal Tahir (16:15)
Hardware, was a hardware. It's kind of like those.

the frame thing, but apparently their tech was state of the art when it comes to like always on like capturing memory and having it searched through. So they kind of solved voice memory. They probably have a very ⁓ proprietary like RAG system or whatever, which is very fast and accurate. And that's been a huge issue. Anyone, if you ask people like, you know, in terms of like unsolved problems, that's one, RAG has always been up there. I think RAG is overrated.

Pierson Marks (16:32)
Mmm.

Bilal Tahir (16:44)
especially with more context, but still.

Pierson Marks (16:45)
Rag

is like people making something seem more cool than it is. It's literally, it's rag is the acronym for retrieval augmented generation. What does that mean? It's like go and get some more information, retrieve information, and then generate text with that information. It's all it is. When you upload your own document into chat, you can say, ask some question. That's literally rag. It's just not, you're just like, you just remove the retrieval part.

The AI's not retrieving. It's like...

Bilal Tahir (17:11)
Yeah, essentially

the unsolved problem is people say, rag, rag, rag this. What they mean is it's a search problem. Because when you have like, say, 50 million documents and you want to search for, you know, you have a question.

Pierson Marks (17:17)
It's a search problem.

Bilal Tahir (17:24)
It really depends on the 10 documents you retrieve. If you retrieve the wrong 10 documents, it's garbage and garbage. It doesn't matter how smart the LLM is, right? So I think that's why it's been a problem. And search has always been a big problem. But I think there's been lots of improvements on it and they have rack strategies and different ways to do it, et cetera. So anyways, apparently these guys had a great product, but...

Obviously, the hardware pending market was always going to be going nowhere. At least, I feel like the pending thing. So anyway, they got acquired by Meta as well. And that was very interesting. you can see some of the pieces come together. Now, just today, they announced a partnership with 11 Labs, which you would think they have Play HD, they have TTS. Why are they partnering with 11 Labs? But I was thinking through this. I think they're trying to create some sort of an AI companion thing with memory as well. mean, this makes total sense for Meta. They have Instagram. They're going to have bacon all these

Pierson Marks (17:51)
Mm-hmm.

Mm.

Bilal Tahir (18:17)
and with the digital avatars and just have these companions on Instagram and WhatsApp and stuff, is what they're going for. So they're just, I guess, leaning in fully into all the tech here, trying to bring these pieces together.

Pierson Marks (18:31)
Yeah, it's super interesting. But why do you think that they would acquire PlayHT and then partner with 11 Labs?

Bilal Tahir (18:41)
That's been confusing me as well because I'm like, why would you pay? I mean we're enterprise level apps customers We know I mean, I'm sure America has a great deal But it's probably more costly at their scale to use the level app models than to just use their own models Which are pretty decent so I I either they're doing it because they feel like the ROI is there of just having amazing voices ⁓

Pierson Marks (19:06)
So one question I have

here, so I'm looking up right now PlayHT. And there are articles that happened over the summer like, oh, Meta acquires PlayHT, blah, blah, blah. But if you go to PlayHT, there's no mention of their acquisition. Did it fall through?

Bilal Tahir (19:21)
Hmm.

Really? So they're still running as a standalone. They haven't been kind of folded up. I thought they would quietly shut down or something, but that's interesting.

Pierson Marks (19:31)
Yeah,

it's interesting because like five months ago, they acquired by meta blah blah blah. Everybody acquired by meta July play AI. Yeah, interesting way. was HTA. They just rebranded. ⁓ But yeah, they completed the deal. So interesting. I guess I just kind of like just requiring you for maybe the models and they're going to let that run along and then eventually they'll shut it down. ⁓

Bilal Tahir (19:40)
That also got a card? That was a...

Okay.

Yeah,

I thought it was more of an AQI higher thing because these guys didn't understand how to train the DTS models. They were like, yeah, just come in and build our DTS models, which makes sense. So it's interesting,

Pierson Marks (20:04)
Right. So here's my

thing. I've always been very, I'm very, very bullish on meta also. So I think advertising is about to completely change. We're in the general media space. We know, you know, like I sent this something yesterday. It was like the Google, what was the pie, whatever the ad generator thing was. it was called, sorry, let me get it. Pumeli. Yeah. Yeah. Pumeli. And it's like a,

Bilal Tahir (20:27)
that was me.

Pierson Marks (20:30)
AI Ad Generator by Google. And so you add your business website, it'll extract all your assets like your logo and what your business does and it'll generate ads. They're actually pretty good. And they can also be animated. So you can literally generate the ads, the static ads, and then you can animate them as like little like five second long shorts, probably animated by VO. But Facebook, let's talk about Facebook. It's super, so Facebook, have Instagram, you have Reels, you have

⁓ Images you have actual Facebook you have all the brands underneath meta They've already been rolling out AI avatars and How do they make money? How does Instagram make money they make money on ads and if you're a brand? You know you want to target the right people you want to create a an ad that resonates with your target target buyer you want your ad to kind of Look oftentimes like

Bilal Tahir (21:03)
Mm-hmm.

Pierson Marks (21:22)
the person you're trying to find. So, you know, for me, you want to create an ad that makes me resonate with that person. I can see myself in their shoes. You know, it looks like me. It sounds like me, blah, blah, blah. If you're, you know, a Spanish speaker and you can add in English, that doesn't look anything like you, you're not going to really pay attention. But if the ad is talking to you, it looks kind of like you or your friends.

It's in your hometown. It speaks your language and the right regional accent. Imagine if you as a brand are able to essentially prompt an advertiser. Like you generate 50, 100 variations. Every ad that everybody sees is dynamic. You know, and that's going to happen. Like it's not a matter of if it's more of when and that like generative UI, all these generative things, you know.

I have mixed feelings on some other areas, but advertising, ⁓ this is like the killer thing in generative media. This is going to be every single ad.

Bilal Tahir (22:21)
It totally makes sense. It

totally makes sense. And it's an extension of the core product itself. think I was trying to look up the exact stat, but I saw this tweet that apparently less than now it's less than 10 % of the content people engage with on reels. Instagram reels is human generated. So majority is already AI generated. It's just AI slope. So we're basically becoming like, it's funny because it's it's less about social and it's more about just being in your own little bubble.

Pierson Marks (22:38)
We're-

Bilal Tahir (22:50)
know, I'll go bubble more and more.

Pierson Marks (22:50)
You know.

Bilal Tahir (22:51)
which I mean, can argue that's a dystopian thing, blah, blah, blah. I think it's more nuanced than that. But to your point, know, ads are just, for me, think ads, the core product and the ads are just become an extension of that. ⁓ It actually goes also, it's interesting. You ⁓ sent me this link, it was about dynamic ads, about having in a movie, you can use AI real time to edit, like add a, let's say a Coca-Cola bottle or something. Looks funky and stuff, but again, the tech is getting better. And I wonder if ads kind of go in that way where everything becomes an organic ad.

Pierson Marks (23:02)
Totally.

Bilal Tahir (23:21)
where you just have ASOP and then you just pay for, know, people will talk about podcasts and JellyPod, whatever, right? Like in their cartoon or whatever. And it just becomes one seamless feed. ⁓

Pierson Marks (23:30)
Alright, alright.

There's

two ways to take this right now that I want to touch on both things. So yeah, so dynamic ads could be very cool. There was a hackathon that XAI hosted last week. And one of the winners of the hackathon was the dynamic ad placement thing. So pretty much you're watching a movie or a show. The ad that they showed was like Harvey Specter from

Suits he's walking outside. He's getting into a car. He has a coffee in his hand and as he opens a door and he's you see him like Going like with his coffee in his hand the video pauses and that scene it's like replaces his posses with him and coca-cola and it just like replaced the the coffee with a coke and he's just looking at the camera so that's like all AI generated literally like paused the scene Replace the coffee with the coke

Bilal Tahir (23:53)
suits.

Right. Right.

Pierson Marks (24:17)
And then the user gets like a five, 10 second long ad of just like the Coke, like Harvey Holden Coke. And then the show continues. Really, really crazy. There's definitely a lot of backlash to that. ⁓

Bilal Tahir (24:21)
Right, right.

Right, and it's funky

and weird, right? But again, it's gonna improve a lot. These kind of placements have always been a thing in Hollywood. Actually, I remember there was a controversial one that happened a few years ago in a James Bond movie. was Daniel Craig, one of his movies, I think the second or third one, where it was James Bond for the first time in all Bond Square, he has a beer. So it was a Heineken beer, and everyone was like, Bond doesn't drink beer, he drinks martinis, what the hell? And it's because Heineken paid them

Pierson Marks (24:33)
Totally.

Bilal Tahir (24:57)
a shit ton of money to have that ad placement. you know, it's funny, right? It kind of did go against the character. And that backlash was kind of made sense because James won. It's like he's not drinking, chugging the beer, right? He's a classy guy.

Pierson Marks (24:59)
Right.

Especially a Heineken, you know? ⁓

Bilal Tahir (25:12)
Right,

but that's how it's always been. There's always been this tension between monetizing and actual content and ⁓ making. And I think AI will actually help with that kind of personalization where it'll say, actually, you can't just put a Coca-Cola bottle in Harvey Specter. Harvey Specter doesn't drink a Coca-Cola bottle. I mean, that's a cool demo. But eventually, he's like, no, he's drinking his Americano. So maybe a Starbucks ad makes more sense here versus a Coca-Cola bottle.

Pierson Marks (25:35)
Right. yeah. Starbucks logo on

the on the coffee, you know, it's going to be. It'll be so cool because I definitely think that there's going to be an ad platform in Netflix where because everyone's trying to look to monetize more and like imagine because when I buy ads on LinkedIn or about buy ads on Google or whatever, it's a process, blah, blah, blah. But if you're a brand, you go to Netflix like, hey, I'm Starbucks.

Bilal Tahir (25:39)
Exactly right so those kind of subtle things will come together.

Pierson Marks (26:01)
I am willing to pay $1 for every item. if I literally just go say, hey, Netflix, what are the shows like with your coffee mugs in them? I'm willing to pay as long as like the viewership of the show gets like, you know, it's in this theme and it's like you have some categories and have some restrictions, like, hey, per impression of this coffee mug, I'm willing to pay like 50 cents.

just to replace a coffee mug with like a branded Starbucks coffee mug using AI. And so it goes out to every single person that watches like, know, suits or every show on Netflix and Starbucks, you'll have these, you know, dynamic AI generated placements. What's exact name? Not placements, like product placement. So, yeah.

Bilal Tahir (26:47)
Yeah.

No, 100%.

I think it's fascinating. And I agree with you, these studios, it's almost like you converge to, like, you got content, media, you got these ad placement, ad engines, you got gaming. All these are kind of combined, because at the end of the day, when you have, you can generate content, doesn't matter, right? And I do wanna, because you mentioned Netflix, I do wanna touch on this. It's kind of different than general media, but I think this is like such a spicy story. I don't know if anyone's watched Succession,

Pierson Marks (27:02)
Mm-hmm.

Yeah.

Bilal Tahir (27:18)
It's probably one of my favorite shows of all time. It's brilliantly done. But it's funny because it's like the plot of Succession kinda is happening in real life where what happened was Netflix acquired Warner Brothers. Huge deal. I mean think about it. Warner Brothers, like the OG studio, the media giant that could have bought Netflix for 50 million. I was gonna say that was blockbuster, but Warner Brothers could have bought for a fraction.

And this startup streaming service called Netflix now is so big that they can acquire Warner Brothers. This is like the future of media, right? mean, there, and so what was the spicy part about that was that Netflix.

puts a deal, like announces a deal, they're announcing it, they're taking only the good parts of Warner Brothers, is HBO, so content library, so Max, HBO Max, DT, DC Comics, mean, because this is what Netflix traditionally has lacked, it's just the proprietary, like, know, IP, basically. And so this would be such a huge, I do think it'll go through, when it does, basically when the dust settles, there are only two companies in the media landscape that matter,

Pierson Marks (28:07)
DC.

IP. ⁓

Bilal Tahir (28:25)
Netflix with DC Comics and know HBO and then Disney with Marvel Marvel Marvel Star Wars because they had the idea

Pierson Marks (28:31)
And don't leave out

Google, though. There's no IP there on the YouTube side.

Bilal Tahir (28:35)
yeah, YouTube is its

own beast. YouTube is like UGC, I mean, amazing shows. YouTube honestly would be bigger than both of these companies, but this is just more like in terms of studios. I think these are the two studios that matter going forward now. But there was Larry Ellison, who was the CEO of Oracle. I think the richest man in the world right now, or at least in the top three. His son, David Ellison.

Pierson Marks (28:44)
Right, right,

Bilal Tahir (28:56)
didn't want it to actually acquire Warner Brothers himself and he actually made a deal with Paramount. Like he owns Paramount and he wanted to create a rival studio and this was why it's spicy right now it's going back and forth because he made a bid that got rejected by the board of Netflix and now he's kind of making a counterbates and he's trying to just make the price so good that they have for they're forced to give Warner Brothers over to Paramount which I hope

doesn't go because I mean I'd rather Netflix own it because I Netflix and Warner Brothers will just be a killer combo even though people think it's gonna cause my subscription to go up and stuff but I'm like think about the quality think about all those spin-off shows that they'll make you with all these DC comic characters that'd be insane

Pierson Marks (29:40)
Well, what's

interesting about this, mean, like, so I don't know how familiar listeners are about like acquisitions and corporate governance, but like the the board has like a fiduciary duty to the shareholder. Yeah. And the best offers and offers would necessarily be all cash. Like it doesn't have to be like the number, but

Bilal Tahir (29:51)
to take the best offer for the shareholders.

Pierson Marks (29:58)
And so there'll be lawsuits here and so with Paramounts I think Paramount countered with a hundred plus billion dollar offer to the original 83 billion dollar offer of Netflix that Netflix offered so what's probably gonna happen is Netflix will probably up their bid or they'll have to you know, Many like litigate this thing out so it becomes messy, but it'll be interesting to see what happens

Bilal Tahir (30:18)
Oh yeah, and the other

political element is Jared Kushner is, you know, the son-in-law of the president is actually, you know, the Paramount campaign. He wanted to push the steel. Like apparently there's some ties and Trump apparently is leaning towards Paramount. So it's like messy. It's like, it's like full on drama. This is going to be a succession 2.0.

Pierson Marks (30:38)
Right. You know what

might happen will be crazy is like if, Netflix, I saw something like about, um, like some sort of content and I wonder if Netflix starts like pulling content and like, uh, try to make it like look cleaner for the administration and the, was it FTC you would have to, who would have to approve this? Um, No.

Bilal Tahir (30:57)
⁓ Yeah, so good point. I mean, FTC probably is involved and

stuff, but the deal is announced. if it does get ⁓ canceled or something, Netflix does get a price. They get 2.5 billion. There's a breakup fee you have to do. So they're still on the hook for that at least. I mean, Netflix will win bigly or win smallly, depending on either. So, bonus Netflix.

Pierson Marks (31:10)
All right.

Well, totally. So

we're in the space. I want to talk about this because you mentioned who the big players now is Netflix and there's Disney. So today, this morning, Disney and OpenAI announced a billion dollar partnership that allows OpenAI to use Sora to generate characters with Disney Plus IP. So it's a partnership. as a user, you're going to be able to generate

characters like Mickey Mouse and whoever and make your own short films and things using Sora and whatever. And then Disney will also use AI to help what they say like streamline the creation process or stuff like internally also at Disney. Really cool. We talked about this last week about Zootopia. know, Bob Liger did say that they're going to allow sort of AI generated user UGC content in the Disney universe.

really big, think. I think it would be really cool to see like, you know, Mickey Mouse talking with a character that Mickey Mouse would never talk to. Like, hey, what if Mickey Mouse talked to Judy from Zootopia or like, know, mix and match, remix.

Bilal Tahir (32:19)
Yeah,

mean, the writing was on the cards because people are gonna do this anyway, so you might as well just take your cut, you because you can't put the genie back in the bottle at this point,

Pierson Marks (32:29)
Sorry.

And what was really interesting about this too is so there was a key detail. So I read the article in the announcement at the end. will be opening. I will like you'll be able to generate animated and illustrated versions of characters. So like all the animated characters that aren't like real people and then illustrated versions of key characters like Iron Man, Captain America. And that was the key illustrated because

There's a lot of actors out there that you like, for example, when you think of Iron Man, you think of Robert Downey Jr. as Iron Man, right?

Bilal Tahir (32:59)
Right, you're thinking of the IP. Right. Well, so no, that's

interesting. does that, because I thought the IP, even if it's Robert Downey Jr., the IP of Iron Man is owned by the studio, right? So I mean, the image rights and stuff. Or is it that the actor still needs to say yes to their image?

Pierson Marks (33:10)
Correct.

So I'm not

in Hollywood. We got called out on this last episode, but hopefully you're listening. So if you know the answer here, let me know. My understanding is that people weren't really thinking about this back as much 20 years ago. And so these old actors, especially the famous actors now, can demand your likeness, your image, everything is yours, even if you're played a character.

And this is like the whole strike about the Screen Actors Guild and it's kind of like, hey, you don't want to be able to replace me with AI and my likeness even though I play a character. However, what's probably going to happen is if you're a young, upcoming, promising actor going to do a film for Netflix, going to do something for Disney, you don't have to leverage. And so I wonder what's going to happen is in those contracts is going to be like, hey, you are playing this character. The likeness of this character belongs to Disney and

Bilal Tahir (33:45)
Yeah, I know.

Pierson Marks (34:09)
any resemblance to you as a person is like, you know, coincidental, you're just playing the actor. And so that actor, now his face becomes owned by Disney playing that character. He could go play other characters too, but like, and so I wonder if that's going to happen. And like these problems will get solved in the future. But maybe if you're part of the actor's guild, that was probably in there. So.

Bilal Tahir (34:29)
I mean, this is.

It's

such an unprecedented kind of a thing because people have, there's some sort of a parallel where you can say people have songs and stuff and they let other people remix their songs, but your likeness and stuff, you can make both, there are gonna be people who are like, here's my avatar, please use them, give me my 10%, 20%, whatever, for every dollar you make. Great model, but at same time, you probably want some guardians, they could take your avatar and make porn out of it or something, you probably are not gonna be okay with that.

So how do you put guardrails about what you can or cannot do with your so it's probably gonna be something like like some clauses and stuff like this is what you can do with my avatar maybe kissing scenes are allowed or something like literally be like bullet by bullet what you can do with with it right so yeah very fast

Pierson Marks (35:18)
Alright,

or it'll probably be like what's, yeah, it'll be interesting if it's an inclusionary rule like that, probably exclusionary. Like these things are like explicitly prohibited and they're probably broad enough.

Bilal Tahir (35:31)
Yeah.

And people

are very particular, especially the big actors. I remember one of the funniest stories I remember I read was apparently, so in the, what's that, Fast and Furious. So Fast and Furious, when Dwayne Johnson, he did a few movies there. The character Dwayne Johnson, and then there's a movie with him, and then Jason Statham, who was the antagonist, I think, Fast They both have a fight. Apparently, there's a clause in both actors' contracts that they can't lose. So they literally have to throw the same

Pierson Marks (35:35)
For sure.

Mm.

Bilal Tahir (36:02)
number of punches. So there's a fight between the Rock and Jason Statham in there. And we don't realize it, but apparently a lot of lawyers got together. They're like, you can only throw these many punches, and it has to be equal. So they both actually, if you count it, they actually come out the same. And they never technically can beat each other. So it's the egos and the lawyers and all that, know, of the most minute details.

Pierson Marks (36:22)
Well, think about that because you

know how important that is. Like that seems minute minute and everything. But like when you're an actor with a brand like your brand is your likeness. And so if you're playing a character in one movie who gets like beat up and then you want to go play a character, another movie who's like the strong guy like, you know, hopefully like the

Bilal Tahir (36:30)
I miss them.

Pierson Marks (36:45)
your audience has a different character, blah, blah, but it's not. It's the rock at the end of the day. You're seeing this guy. And so it doesn't make sense. So your brand persists across movies for better or worse. Like Robert Downey Jr. after he finished Iron Man, played, what was that movie with all the animals? It was like right after If Lot. It was like Dr. Do, not Dr. Dolittle or something like that.

Bilal Tahir (37:09)
yeah, yeah,

I know.

Pierson Marks (37:11)
And it flopped. was like, because everyone's like, yo, this is Iron Man. What's he doing talking to all these ⁓ animals or whatever? yeah, your brand persists. ⁓ But very, very cool. OK, so we talked about a lot of stuff, honestly. ⁓

Bilal Tahir (37:16)
Right. Right.

That's true.

Yeah, all over today. mean,

again, I mean, we, you know, we've been all over, that's the, it's, you know, kind of the beauty of this, by the way.

Pierson Marks (37:31)
That's the beauty of this.

We're looking over here. yeah. mean, the 11Lab stuff, that was awesome. There was a few other things that I think we'll probably just hold off for next week. But was there anything else that was core that happened this week? General Media Space.

Bilal Tahir (37:49)
Yeah, I

did notice one thing. So again, with AI, there's always going to be a backlash. so apparently McDonald's made an AI ad, which is pretty cool. But it got pulled, which is, again, Hillary or some like, it's a good ad. What does it matter if it's AI or not? It comes to what Tim Sweeney said about Steam. They're going to take out the AI label because he's like, what does it matter if it's AI generated or not? Because this is so ubiquitous. So I think it's just.

Pierson Marks (37:54)
All right. yes. Yes.

Mm-hmm.

Wait, wait, the AI label on

Steam? You have to repeat, you have to dive into that a little bit.

Bilal Tahir (38:20)
I think, that was like there used to be an AI label and he's the same. We don't need that.

Pierson Marks (38:24)
An AI label

on what? the game or on the image or?

Bilal Tahir (38:27)
on the

game characters and assets or something. I'm not as familiar with the, what is it, Unreal Engine assets and stuff, but I remember he made a tweet about that, which makes sense. feel like what's AI, what's real, we'll just mesh into one thing at some point. And we're kind of just fighting right now. There's two forces. One is...

the AI is slow up, it's kind of sloppy, but that is a mid-wit problem because I think that'll go away. The second is a more persistent thing, which is like, it comes from an insecurity that AI is replacing us, which it is. There's no denying it. I kind of hate it when people whitewash it like, ⁓ no, but there'll be other stuff. Let's just kind of make it, I mean, let's approach it like, okay, AI is gonna replace us, yes. What does that actually mean? mean, and how do we...

Pierson Marks (38:55)
Right.

Bilal Tahir (39:10)
what can we do about it? And I feel like, well, you can't fight it. So the question becomes, do you join the club? How do you actually use it to do more stuff with it? So I think that's the only attitude you can take with it.

Pierson Marks (39:18)
advantage.

Totally.

No, it's super interesting. Yeah, so McDonald's AI ad pulled. And it was good. I watched it. mean, wasn't like an artist that

Bilal Tahir (39:28)
Yeah, it was great.

Yeah, and anyone interested, if you watch that, if you like that kind of stuff, there's another one that some guy made called Mac Wars, which is really funny. He basically does like a Avenger style battle, but between all the brands. you got Wendy's, you got McDonald's, you got Subway, and they all have characters in there. It's hilarious. And Taco Bell is like the big antagonist. It's called Mac Wars. He made like three parts. He's made three parts already. Went viral. But this is the type of stuff you'll see, you know, ⁓ if you let your imagination go wild.

Pierson Marks (39:44)
⁓ that's funny.

What's it called?

Bilal Tahir (39:59)
and accept AI.

Pierson Marks (40:01)
Well, okay, so we covered a lot of stuff in general media space today. A wild way to end the week. We'll see if there's anything that major happens before the end of the year. But if not, we'll still see you next week. yeah. See you,

Bilal Tahir (40:14)
Yeah. All right. Take care, guys.