Creative Flux

Pierson Marks (00:00)
Welcome to Creative Flux, episode 16.

Bilal Tahir (00:00)
Yo.

Hello, hello, what's up?

Pierson Marks (00:04)
It's a cool one today because now we're both in the same city. We're not physically recording this together yet. Maybe one of these weeks in the coming, ⁓ we'll do this together. welcome to San Francisco.

Bilal Tahir (00:14)
Yeah, thank you. know, it's awesome to be here and still like it's not hit me yet. feel like you know, but I think once I walk out in some homeless guy yells at me. I'll be like, yes, made it to California.

Pierson Marks (00:23)
You gotta

go walking around, especially next week. It's such a great time because with Tech Week next week, I feel like downtown's gonna be very bustling. I haven't even signed up for any events. I need to look specifically on the gendered media ones. I assume fall and replicate and 11 labs will be doing something, but I haven't even looked, so.

Bilal Tahir (00:32)
Yeah.

Yeah, nobody. It's crazy. So so for people who don't know what is Tech Week.

Pierson Marks (00:50)
though.

Yeah, well, TechWeek is... So it's hosted by A16Z, Andreessen Horowitz, a major venture capitalist firm. It's a bunch of just...

Events that go on in the city. They're just all independently organized It's one major calendar that you can submit your events to but it's all independent and companies Venture VC firms everybody can just like host their own events put in on some major calendar everybody comes in from around the world to do this and has happens in SF New York and LA SF being probably the biggest one. Maybe New York pretty close I went to the summit in LA last year, but ⁓ it's not the same vibe. It's not like

Bilal Tahir (01:29)
Yeah.

Pierson Marks (01:29)
But

yeah, it's just like a bunch of independent events. I think it's like a thousand events on this calendar. It's actually nuts. It's like I tried to look through this and I was overwhelmed. There's things for founders, there's things for marketers and builders and VCs and law and is that everything you can imagine, I guess. So, cool.

Bilal Tahir (01:45)
Yeah. Yeah. No, it's awesome.

Having that kind of like the dense, just getting all those events together. There's something about the atmosphere of the, you know, having this kind of conference. I remember I went to South by South bus, which is I think the biggest conference in America. It happened in Austin in around February. And for one week, literally the town of Austin doubles in size. It's insane. And it was crazy. I, I don't think I even paid for it. I mean, there were paid events, but it was just so much.

stuff going on there, just people just shouting and it's like you really get immersed into the thing, you know, and so and then the other one was eTendr which was a crypto focused conference and I mean there's just, is this a lot of fun? It's also like very cool too, I mean because you can just meet so many people like in that area you're interested in. I really like though, like with eTendr obviously it's focused on crypto so with TechWeek I'm guessing you know obviously AI is gonna be its center state so it's just a very

Pierson Marks (02:25)
Hmm.

Bilal Tahir (02:42)
a great honor to just come together and just meet a bunch of people who are interested in the same things you are interested in.

Pierson Marks (02:47)
Totally, totally. It's super interesting. mean, there's also some big speakers, some big names. Mark Andreessen and Ben Horowitz, I think they're doing like a big keynote type thing, but I don't know where that actually is at. But yeah, there's some speakers as well there and a bunch of featured events. Creativity in the age of AI. What is this? HSBC. ⁓ That might be like the perfect one to go to actually, like right at the top I just see this. ⁓

Bilal Tahir (03:09)
I am, because.

which has been seen as when I think creativity,

I think a bank. So yeah, it makes sense.

Pierson Marks (03:17)
I mean, yeah, because like the speakers there, it's, yeah, Global Head of Innovation at Banking, HSBC. And then there is Disney Accelerator, which is interesting for us. Yeah, there was actually an interesting thing I saw yesterday. So we'll get into this in TikTok and Oracle and everything. so everybody probably is aware that.

Bilal Tahir (03:28)
Disney.

Pierson Marks (03:38)
About a year ago, there was a bunch of talks about like, Taekwondo being shut down and they have to divest the Chinese bite dance, has to divest and allow...

American company to own the algorithm and the data of US citizens and kind of split tick-tock into like a US version and then Oracle came in and Larry Ellison was like, hey, we'll be the we'll be the good American company that come in and make sure that your data is safe from Chinese CCP as a like a political thing. And so I get this is like a whole like drama over the course of a year, you know, and I was watching something yesterday. So I think this is going to go through and that's what's going to happen like Oracle.

Bilal Tahir (04:09)
Right.

Pierson Marks (04:17)
But tying this back into Disney, I didn't realize that Larry Ellison's son, David, is the president or CEO of Skydance Paramount. I didn't know that. I didn't make that connection. I didn't realize he was there. And they're putting in a bid for Warner Brothers to take over Warner Brothers. So you have Ellison's David, David Ellison, the son.

Bilal Tahir (04:23)
Mm-hmm.

Mm-hmm, yeah, yeah.

Right.

Pierson Marks (04:39)
⁓

owning major Hollywood media empire. And then you have Oracle owning TikTok and that side. And I was like, it's crazy because also in AI, they're both like very, very pro AI. AI in movies, AI in shows, AI in like, I was like, this is kind of nuts. I didn't realize.

Bilal Tahir (04:52)
Right.

Mm-hmm.

No, it is very

slept on story, it was one of the probably the only reasons Paramount stock, think, went up a little because Paramount has nothing going for it, know, compared to Disney or Netflix, etc. But suddenly, know, the Papa Ellison's, you know, infinite money pot is open to Paramount. And so that's why they were like, OK, yeah, maybe there is a chance here they can just like outbid for good content, etc. And you just ramp up.

It's a very, on a more general level, find the Ellison, it is kind of shaping up to be a dynasty right now. People think of dynasty, think of the Murdochs, like Lachlan Murdock or Soros. And I feel like the Ellisons are shaping up to be a dynasty with David Ellison now kind of making inroads into media. So it'll be interesting to see how this plays out and what kind of, what's his worldview and how much of that.

is enforced, you know, via the company.

Pierson Marks (05:54)
Totally. Yeah,

it would be super interesting. And I think the reason why it's relevant to talk about here on Creative Flux is because, like they both said, they're very pro-AI, obviously. Oracle's one of the building so many data centers, they're pro-AI.

Bilal Tahir (06:07)
Also through

monitoring and stuff. That's like another thing with Larry Esselman. He's very much, cause he used to work for the CI, right? You know, I mean, that's how he started. he was, was Larry Esselman, I respect the shit out of him. I mean, he's amazing. Like he, he was a programmer at the CI and he came across a server DB2, I think it was called that then that was the first SQL server basically. And he was like, I like this. Let's start a company around databases. And he started Oracle basically with like a thousand dollars of his money or something like that. ⁓

Pierson Marks (06:12)
All right.

All

It's crazy story and they've just been around for forever too, which is.

Bilal Tahir (06:35)
Yeah. yeah,

yeah, they're older than Microsoft. SQL, MySQL came later. I mean, it's like Oracle is a pretty old database, but.

Pierson Marks (06:45)
Right,

But also, though, the interesting thing in this in regards to media is obviously we have, we were just talking about Skydance, Paramount, Warner Bros. But then tying it into the TikTok deal, Oracle, what influence are they going to have over the algorithm and everything and prioritization of content, user-generated, AI-generated.

will, like how will AI generate content on TikTok perform, but then also going into Sora. Because Sora 2 today was, or yesterday, this week was released, have, but Sora 2 was released this week and it's essentially an AI TikTok that you can generate your own videos with, right?

Bilal Tahir (07:14)
Hmm. Right, right.

It was, yeah, first off like, let's, I mean, there are like two sides here that are very interesting. First is just like the quality of the model. I mean, we've talked about this before. It's funny, I think last week maybe you asked me a question about would it open a release to video generation? And I said, of course it would, because for me in my head, they're a consumer tech company first.

This is like nothing. I mean, they saw the chat GPT moment with the image. They saw they had the chat GPT moment, then they had the chat GPT image moment with GPT image one. And now, I mean, video was obviously, if they had nailed it, it was gonna be as big, if not bigger. And it sounds, it seems like they delivered on that front. The Sora 2 quality is...

pretty amazing, like, you we'll talk about some of the things you can do. But the examples I've seen at least, know, cherry picked as they may be, they were just amazing. Like they were like a lot of them, I could not tell if they were AI or not. mean, it's, you know, really makes it pass that uncanny valley. Yeah.

Pierson Marks (08:17)
So, and so

just like, so Sora 2, I mean, we've talked about Sora in the past and how I just say they launched it or they released Sora, then they didn't actually release it the public for a while. And then at that, the time public actually had access, you had all these other open source, other competing labs release their own video models. How does Sora 2 and the examples that you've seen today compare with other models?

Bilal Tahir (08:41)
Right. Yes.

So this is where I mean the second side, the second angle becomes important because unlike other features, this wasn't just a model release, it was an app release. So what they've actually, they actually have an app where like a TikTok or whatever, like you say, AI TikTok where you can have, can generate so and you can share it. And I think because of this unique, because it's an app, they had to be less of elitist quote unquote in a way because

you know, the lifeblood of a social app is having people. So they did do an invite code only release. So you need an invite code, but the invite code, honestly, it's like super easy to get. Like you just literally Google invite code on Google or sorry, Twitter or whatever, you can get an invite code. Five people sent me an invite code. So I just like added it. So it's not exactly, you know, the elitist, but you know.

I get why they had to do something, because you want to create some hype. That's just what OpenAI does. So I think it's one of their more accessible releases, surprisingly. And actually, I am a little surprised because I've talked about this before, how OpenAI always, their quality always tends to be the best, but it comes at the expense of compute, which means priciness. And the fact that they're just opening this up so much to, which, mean, they're Mauser, the biggest in the industry.

I mean half a billion or whatever. I imagine that's a shit ton of compute they have to account for. it sounds like either they have the compute or they're hoping that it won't be as big or crazy as the image one was.

Pierson Marks (10:15)
So if you have an invite code, are you able to use it or do you have to have a chat GPT Pro or Plus account and then invite code or is it if you just have invite code you get it?

Bilal Tahir (10:24)
that's a good question. I have the ChatGPT program, I think that you need that. I didn't download the app, so what you do is you download the app and then I think, I'm sure there are some restrictions for you. I wonder if you can make any videos though, if on the free tier. ⁓

Pierson Marks (10:27)
Right, I don't know either. I was just curious if you did know.

Right.

Right. So yeah, I

mean, how was your experience in the app? Like, what were your first impressions? What was like the standout things, consistency, speech, audio, like?

Bilal Tahir (10:50)
Right.

I can't speak to unfortunately the app because I literally just got it like 10 minutes before we like. But I've seen the output videos which are very interesting. They're obviously that we talked about the quality. The quality is really good. You know, it's very realistic. There's it's got audio too. it's I would say at par with VO3. You know, some people say VO3 is still better. I at least based on the again, could be cherry-picked examples.

Pierson Marks (10:56)
Yeah, I wasn't sure if you had the...

Bilal Tahir (11:16)
The examples looked way better than VO3 at least to me and very more realistic. The other thing which they've done, this is why, mean, OpenAI does consumer. mean, this is the, mean, their supercent is consumer. You know, obviously, I mean, they build great models, but they understand consumer psyche and hype. Sam Altmer understands that. They made this feature called Cameo. And the Cameo is basically what you do is you take your own image and you record a video and audio.

Pierson Marks (11:19)
Right.

Bilal Tahir (11:43)
and you can create your own avatar so quote unquote that you can use in your videos, but also crucially, you can share it with other people. You can make it public and then people can do it. So I saw like some people like influencers like Justine Moore and Alexander Bortez, they made their avatar public, I mean, sounds kind of bad, but my immediate thought was like, I mean, you're just giving up your rights of your image. I mean, I guess technically, mean,

Everything's public. mean, anyone can screenshot me right now or take my recording. it's funny. It's an interesting transaction because you're saying, just go crazy with this, my imagery. And then people can do all sorts of stuff with it. within Sorami, I'm sure they have some limits and stuff. So you probably can't do some really edgy stuff. But still, that's probably a when, not an if.

Pierson Marks (12:23)
You know, it's super interesting, especially... ⁓

Yeah, totally. So is the

reason why everybody's creating videos of Sam Altman because he has his own cameo avatar available?

Bilal Tahir (12:39)
He was the first one

who did it, which I want to see. That's an awesome move, right? As a CEO, you're like, I'm the first. I believe in this technology. Here, go crazy with my own. I mean, that's a pretty good PR move. You're like leading from the front.

Pierson Marks (12:49)
Right.

There were some funny ones that I saw. He was in, I forget the name of the Studio Ghibli founder. I apologize. Yes. But it was Sam Altman in his studio in Japan. And he's like stealing all his designs and running off with the Studio Ghibli artwork. He's like, I'm going to go train a model.

Bilal Tahir (12:57)
Me as not you.

when he's yelling about AI to his engineers and yeah.

That's hilarious.

Yeah. No, I love it. I think it'd be funny if Sam Altman's avatar becomes like one of those like ubiquitous, like AI avatar things, you know. It's, I was watching this show called Upload. I don't know if you watch it. It's an interesting show. It's basically, know, it's on Amazon Prime. It's a, when you basically in this show, when people die, they can upload their minds to the cloud and they live as, as basically in the cloud. And it just.

Pierson Marks (13:18)
Yeah, totally.

Bilal Tahir (13:35)
comes with all these interesting questions about what happens when you live in the cloud. And one of the funny things is there's this actor who is the AI avatar chaperone in the community that this guy lives, the main character who's uploaded lives in. And he's like, the actor himself is pissed because he's like, they took my face for $1,200. now they, you know, and anyone now who interacts with him, they can't take him seriously because in their mind, he's the AI avatar idiot.

guy, you know, and so he's like, I've ruined my social life has ruined everything, you know, so it's interesting, you know how, how that works.

Pierson Marks (14:09)
No, absolutely, it is interesting. I have some questions, and I wonder if you know or if you have any thoughts around this. So the length of the videos that Sora 2 creates, I've seen ones that are like 60 seconds plus, right?

Bilal Tahir (14:21)
Really? I have to look at the system.

Pierson Marks (14:24)
Or like decently

long, like longer than the five second videos that we're kind of used to.

Bilal Tahir (14:28)
Right,

right. But I wonder, is that like a one shot or is that people are just creating and stitching in frame or last frame? Yeah, I don't know.

Pierson Marks (14:36)
That's my

question because I saw some interesting ones where it was, there's two videos I saw from Justine Moore. There was Pikachu in Saving Private Ryan and then also Pikachu as the Dark Knight. And so they replaced the characters like Tom Hanks and Christian Bale.

Bilal Tahir (14:45)
Hmm.

Right.

Yeah,

I need to explain more, but I remember for Sorsoora, what they did was they had this, the UI itself, the way at least the website work was you created videos with frames and then you can kind of do a pipeline flowchart thing where you create an export and then you can export the video at the end. So I wonder if you know that. On this one, the cling actually came up with an interesting feature, which, you know, it's just, I think it's just first and then last frame.

Pierson Marks (15:07)
Right.

Right, right, right.

Bilal Tahir (15:18)
extrapolation where you take the last frame off of clip and then use it as a first frame with the next one, but they call it infinite. I think it's called infinite generation, but basically they abstract that away, which is, think kind of nice. You don't have to worry about last frame. You just go, I want this, this, it just under the hood. think it just does that for you. And I think we'll have that because right now we're in this weird age. Like I'm sure like within next year or whatever we look back and be like, remember we had to.

Literally use FFmpeg to get the last frame and then use it and that was so hacky and now you just do it under the hood. So makes sense.

Pierson Marks (15:48)
Totally.

yeah, yeah, that just makes complete sense to me. Like why, like we said the other day, there's, live in a world right now where there's just all of this scaffolding that is being built around models in order to get to some output and.

Bilal Tahir (15:59)
Mm-hmm.

Pierson Marks (16:04)
A lot of users have to be aware of this scaffolding. If you want to take full advantage of language model, of video, audio model, image model.

You either have to be aware of the limitations, build a scaffolding yourself or use a platform that provides you the scaffolding like Comfy UI or the FAL workflows or any of these like kind of you have nodes and you connect them all together. But it's just complicated. mean, like you really shouldn't have to do that. Like if you're going into ChatGPT, this goes back to what we were talking about last week where it's interesting that they didn't give the ability to generate videos in ChatGPT. It's probably a completely

Bilal Tahir (16:38)
around it.

Pierson Marks (16:42)
thing because they know that everybody that has access to video gen and chatgpt will probably generate at least one video and that's probably going to be crazy and that's why they create the separate app maybe one of the reasons that decided to make it a separate app but

Bilal Tahir (16:49)
Right.

Pierson Marks (16:55)
Yeah, I mean, like, I don't see a world, I don't see a future where you can't just say, hey, make this longer, add this next scene, you know, and just like stays consistent. You have the previous video, you just extend it. We're regenerating the whole thing.

Bilal Tahir (17:09)
Right.

Yeah, it is interesting. What do you think of the, I guess the big question is like right now, obviously this is the honeymoon phase. They launched this app. It's self-contained. Crucially, it doesn't let you share on other platforms as easily. you have to, you can't just share it to TikTok. So they're really banking on having an AI native TikTok style app, but self-contained. What do you think of this strategy? Like, do you think like...

it'll be sticky or will people just move on or as you know, it sounds like they're deliberately banking on the fact people will just be able to watch AI video, generated video, and they'll be interested, know, even when the novelty kind of wears you off.

Pierson Marks (17:46)
Yeah,

I'm a little bearish. And this is my reason why. So.

I think we have, we have platforms already that exist that users are there. you have Instagram, you have Tik Tok, you have X. think that those have already great network effects, existing. So there's existing network effects. And I think that when you have Meta and you have those people like Meta could very easily take cling or take some open source thing, whatever license fees they have to pay and they add something into Instagram. They haven't done that yet, just because do they not think that the market is ready? That might be the reason.

Bilal Tahir (17:56)
right.

Well, they

Pierson Marks (18:19)
Thank

Bilal Tahir (18:19)
did actually, we haven't talked about the Alexander Wiener launch for Vibe. It's called Vibe videos, right? Remember? Did you see that?

Pierson Marks (18:25)
Oh, interesting. Right, right, right. I saw something about

this. We'll go into this in more depth.

Bilal Tahir (18:29)
I think it's called Vive videos and it's basically that. It's AI generated. It'll be a separate tab on Instagram where you just get AI generated videos, which, so it's interesting. What you say, I agree with you. I, cause it is a, in a way it's a feature. mean, when you extrapolate like AI generated videos are just videos, right? I mean, so when it gets commoditized and stuff, you know, does it, in the long run, when I say long run, literally the next year, like.

does it matter if it's an AI generated video or a user generated video? mean, okay, there's an authenticity aspect, but in terms of entertainment, right? I just want to watch good content that I like, right? And so I don't understand, because they're making a separate tab on Instagram, this is AI native, why not just have a, just one single scroll? Like TikTok basically has a mix right now. So I guess they think that there is some...

some utility in having just an AI generated scroll feed versus a user generated scroll feed. And I'm not maybe as plugged in to see, understand why that would be different.

Pierson Marks (19:29)
Yeah,

my opinion here is that this is a short term play. think that where consumers are today is they want to feel, they don't want to break consumer trust. I think that's the key. It's like when you go into a Reels and you see somebody there, you see an influencer talking about a product, you see something, your mind still associates that with somebody, spend some time to edit this video and create this video. And that's like authentic. And then you see, hey, let me toggle on

Bilal Tahir (19:42)
No.

Right.

Hmm

Pierson Marks (19:58)
explicitly either a setting to allow AI or to have a distinct feed that's AI generated and your mind is prepped to be like, everything I see here is AI. You know, you could see some craziness and I think it's consumer trust because today I don't think we're ready to acknowledge the fact that whether it's traditionally recorded and created or AI generated, I don't think consumers are ready to mix those yet. And I think there'll be backlash. think there'll be I think if you do mix

Bilal Tahir (20:08)
Yeah.

Yeah

Pierson Marks (20:27)
them and people are like, they'll go to other platforms, they'll go to like other things and it's not ready yet for that and I think it will come.

Bilal Tahir (20:33)
Right. I think, no, I think

you hit the nail on the head about like they're not being ready. And because as you were saying, it occurred to me, I'm like, yeah, there are some videos that, you know, I'd rather be human. Like when it comes to a product review, like, hey, this mug, I don't want to watch an AI avatar, regardless of how good they are. You know, I want a real person to be reviewing the mug. But for watching a funny, like a Yeti joke video and stuff, you could argue that especially as, you know, if you extrapolate the tech getting better.

Pierson Marks (20:50)
Right.

Bilal Tahir (21:00)
I would rather watch AI videos rather than real videos, because they can have cool effects like dinosaurs in the background and stuff. Like so for pure like just entertainment value, it might be, there's a bifurcation where if I just want to be entertained, I want AI videos, but if I want real authenticity, something, you I want human videos. I do think though, having even just having the tab, because I've noticed, I don't think most people want to make the decision of, ⁓ I'm in a human activity mode, I want this. I feel like, so maybe it's a short-term thing where people learn.

you get accustomed to it and they get good, but ideally there's a router, like a video router that just based on my scroll, they're like, oh, you're in an entertainment video mode. Let me show you more of AI stuff. Oh, you want authentic videos, here's some human stuff, you know?

Pierson Marks (21:44)
Yeah, totally. And I think it'd just be like very, just like the algorithm just.

Prioritizes content based on what you're looking into and doesn't matter if it's AI or not and then but I do like I do see that let's say there's a group of ten people friends your friend group of ten and Maybe nine out of ten will never click that AI tab But one time one friend clicks on that AI tab just to check it out you scrolls He has a good experience and he sends that to one of his other friends and they're like that will open up that AI tab you open that video that was a funny video because there was dinosaurs pop like eating your head off or something crazy And then you scroll and you say it's the next one

Bilal Tahir (22:15)
break.

Pierson Marks (22:17)
And that next video that's shown knows your preferences from your normal traditional feed and then it shows you an AI generated video that you'll probably like because it knows your preferences and then you're like, this is cool. And you kind of are able to slowly get in to the minds of people like, hey, like AI generated videos could be good and cool. And then once you start to see like the time spent on that feed, get to some threshold, you start to integrate it in and try to make those businesses.

Bilal Tahir (22:21)
Hmm

Pierson Marks (22:43)
from Metta because I think you have to do it carefully and just because consumers aren't ready for the reality that is quickly approaching. so, but tying this back into like the original question about Sore and why they did this and do you think it's going to be like a long-term play or whatever? I mean, I don't see...

Bilal Tahir (22:44)
Hmm.

Pierson Marks (23:08)
Personally, Sora being more than like a...

gimmick right now. don't think that OpenAI is absolutely a consumer product. They do consumer very well in addition to research. I think that they're trying to stay in the minds of consumer. have a, I don't know their org structure. think it makes sense to say OpenAI, Sora, people are familiar with Sora. They're going to be like, no video, Jen. think eventually what's going to happen is you're going to have

the ability to create videos with ChatGPT, a unified interface, ChatGPT can do it all. And I don't necessarily know.

if a distinct app for just AI video is going to work. think you're going to, think what's going to work is going to be a hybrid like Instagram where you can have both real people, AI videos, and it'll kind of dynamically change based on what you're interested in. And after I think there's like having just AI might not be a long-term play, but there was a podcast I was listening to and they were discussing like, like why is open AI spending GPU and compute resources on building a slop generator when they can be curing cancer. And that's like kind

Bilal Tahir (23:50)
Mm-hmm.

Right, Yeah.

You

Pierson Marks (24:11)
of this thing that people have said this quote and what people don't realize I think is that OpenAI at the foundation, it's a consumer company but it's also a model lab you know and where we are today in terms of model capabilities you can very much so picture a the frontier models as a person the smartest person in the world

give them every single piece of work in the world and maybe video but every single piece of written work in the world.

and put them in solitary confinement. Put them in a room, no windows, nothing. Give them the entire library of the world's knowledge, the smartest person, put them in the room. That's where AI models are today. They know everything. They're smart, but they've never been outside the box. They've never seen the waves crashing on the ocean. They've never felt, you know, a car collision and how that collides. World models and video models training with that type of data gets the models outside the box, gets them to AGI. And I think

Bilal Tahir (24:44)
and

Hmm.

Hmm.

Pierson Marks (25:06)
very, very much that we cannot achieve AGI until the models understand the world that they live in. Because right now they're living in solitary confinement in a box and they need to get out. And Sora is just a stepping stone that they can create as a cool consumer product as they approach that vision, but not like a new track that they're pursuing.

Bilal Tahir (25:17)
Hmm.

Right. Right. Yeah.

Yeah, I mean, I agree. mean, the angle of using it as training data for, you know, more smarter models, robotics, et I think is huge. I think the other thing, which is I'm not as bullish.

I mean, I don't know necessarily if it'll pan out. It's kind of like the browser analogy where, you you got perplexity making comment or whatever, right? They're making AI. what if we make an AI native browser, right? But I still believe I'm a Chrome guy. I'll probably still be a Chrome guy because, but there's, understand the play, like what if from the ground up you make a browser that just has AI built-in features? Similar to Vitsora, think there, if I have to kind of make a case for that is, what if you built an app?

from the ground up on AI. You can do stuff like we talked about cameos, but other stuff like I can take a video, can remix it, but because I know all of it is generated with the same model, maybe there's stuff like embeddings and stuff which make it much easier to remix stuff, share that, add another edit, video, another, swap out. You see somebody's videos, you're like, I like this, but I want to swap out a character with me, and you do that and you take it your own. that, not saying TikTok can't do that, but maybe there are some, you know, just because OpenAI.

has one model and it's AI native first, maybe they can roll out these features faster and better than other apps and their banking on that fact, that it can create engagement that a TikTok or Meta won't be able to keep up with. That's, I'm just, okay, and I still lean like as with Chrome, I still lean on the network of XT's app that they can still do it with a good enough feature, but that's like the other side.

Pierson Marks (26:41)
Mm.

toilet.

But I mean, this just goes back to what we were just saying before this call about like sometimes it's important to know what features not to release. And when you already have an existing user base that's come to trust your product, understand your product, when you get so big, it's much more difficult to roll out new features because of the innovator's dilemma, because you've had some feature set and now you're like.

Bilal Tahir (27:04)
Right.

Right.

Pierson Marks (27:18)
If you launch a botched feature or something that doesn't work, you're annoying a lot of people that already came to trust your product, that your product's already doing well and you don't need to mess around with something that's doing well. But when you're a new product, like Sora, that everybody is a first mover, everybody's like at the very left of that technology adoption curve.

You could just do whatever you want and things could break and everybody kind of knows that you're at the top. You're the point one percent of the people that are ever experiencing this at this stage and you're like, okay, and accepting of trials and changes and things and new features and everything changing all the time. And it's just like, it's so interesting about like business in general is because like, can't a big company do this? And it's because of that real reason. It's like they're big and they just can't have these small experiments because they're done.

Bilal Tahir (27:59)
Hmm.

Pierson Marks (28:06)
shareholder value and they don't actually matter in the grand scheme of things or you could potentially lose more than you can achieve. It's like why won't... it's like that famous quote like no CEO ever got fired by choosing IBM or something.

Bilal Tahir (28:20)
Right.

Pierson Marks (28:20)
It's like, yeah, like, you're the founder, you're a founder led company, like you're not taking these big bets because like, Hey, you know, the worst thing that could happen is I lose my job and shareholder value goes down to nothing. Cause I made some stupid bet. But if I just keep going in with my own playbook, like, Hey, we're doing well. Like, I don't need to really innovate that much. And I don't need to add.

Bilal Tahir (28:32)
Mmm.

Pierson Marks (28:39)
like AI slop into TikTok algorithm and then I get fired because there's so much backlash and they put in a new CEO or something. yeah, it's interesting.

Bilal Tahir (28:49)
Yeah, I I agree. think it's interesting. I agree with you. I mean, I definitely lean on the fact that, you know, I've actually criticized OpenA because of that. like, I think I said before that I would be surprised if they came out with their own Laboo doll because I feel like they're all over the place. They're trying to build chips. doing this. And that makes me my instinct to that is like a little bearish because I feel like they're just diluting themselves. They're not really focusing on one thing. But at the same time, like you said, you know, they, you know,

they built awesome product. mean, they've delivered before and so, you know, maybe all, and maybe there is a grand master plan on using the data and training data and all this leads to the next frontier model, et cetera. So who knows.

Pierson Marks (29:15)
It's.

color.

think you have

to also look at who Sam Altman is. Before OpenAI, he was the president of Y Combinator. He was surrounded by every year, 100 or so different companies that were able to cross-pollinate with each other. You created a microcosm of really intelligent people that were able to talk and created the environment for them to communicate and share resources and ideas. And it's very similar in the way that OpenAI is organized, where you have a group of core

Bilal Tahir (29:41)
Right.

Hmm.

Pierson Marks (29:57)
researchers, have product engineers, you have PMs, have like all surrounded by like bottleneck by compute and they all have they all are like together working on their own little projects kind of autonomously but cross pollinating ideas and it just I think that's just an interesting and cool approach because it's less top-down and more of hey we're gonna do a lot of like little cool things everyone talked to each other and some things are gonna work some things aren't

Bilal Tahir (30:23)
Mm-mm.

Pierson Marks (30:27)
gonna work like and you forget about the things that don't work like you really do like I was just trying to think like right now it's like what's something that they released that really didn't work and I there has been stuff that they kind of like slowly rolled back but I just can't think of it on the top of head right now nobody remembers 4.5 exactly

Bilal Tahir (30:39)
I mean, the 4.5 was a big, like, I mean,

they kind of still wrote it out, but then they were like, oh, we need the GPUs. I mean, there were other stuff too, I'm trying to remember. They renamed a lot of stuff, like, what was it? Similar to, not Sora, the, well, Sora is funny because they released it, but then they kind of took it back. mean, Sora was kind of like just dead for a long time. They were still building the new model, obviously.

Pierson Marks (31:03)
All right.

All right. All right. Yeah.

Bilal Tahir (31:05)
So yeah, I mean, I agree with you. mean, they're

just trying to move fast, you know, experiment a lot. And there's something to that approach for sure. So, but in any case, there are two sounds. mean, at least the output looks great. I don't know what the generation time is, provided it's reasonable, I think this could be a you know, great new way for them to get engagement, get the data, you know, keep, keep, keep in the conscious AI consciousness in the world of, you know.

Pierson Marks (31:12)
So and.

Bilal Tahir (31:32)
in the eyes of the world. So we'll see.

Pierson Marks (31:34)
Totally, totally.

And I know there's something else I just want to touch on before we wrap up soon. Because last week we talked about X4 or 2 and payments this week also, OpenAI, alongside Shopify, Etsy, was it Stripe as well? I forget. But they rolled out essentially shopping and ChatGPT. ⁓

Bilal Tahir (31:43)
to the room.

really? I did

not see this.

Pierson Marks (31:54)
And so, yeah,

it was pretty cool. Like they've talked about this in the past that you want to be able to.

Bilal Tahir (31:58)
Right, like you buy,

you say buy me whatever and then it just does it, an agent goes and buys it on Amazon for you or something.

Pierson Marks (32:05)
Yeah,

so I don't know if Amazon is in there. should eventually be, because Amazon would probably be shitting his pants if you could go to Jatchabee Tea and buy me toilet paper. And Target will always be the one that pops out versus Amazon. But yeah, think Etsy was the major one. I know Shopify is coming down the line, if not already there. But.

Bilal Tahir (32:15)
Right, right here, right.

Pierson Marks (32:23)
You're just going to have your credit card information. Chetchi Petit is going to be signed into your Etsy account, signed into your Shopify. Because Shopify already has the Shop app, so every Shopify store is available on the Shop app if you enable that. so it's just like a unified payment. like a distributed. Shopify is a distributed Amazon, essentially, even though it looks like everyone has their own website.

Bilal Tahir (32:35)
No.

Right, right, yeah. No, makes sense to bother a shop assistant with that

because they're scattered all over and so, you know, can, it's a great way for them to, it's kind of like GEO SEO, like, you know, it's another way to buy stuff. I don't know, I mean, I'm just, I can, maybe I'm just too much of a control freak, but even with that, I don't trust an agent to go buy it. I'm like, you can take me to the site, but I need to see it before I click buy. I don't know. I've always been like that.

Pierson Marks (32:46)
Alright.

Totally. ⁓

Yeah, I

think that's more so what it is. It's like it shows up a card. It's like, hey, can you buy me toilet paper? Or can you buy me like this new thing? And they'll go search. I'll find a, hey, here's like these three things I think you'd like. I don't know if I can improve search, I would say. And then you don't have to say, you don't have to plug in your information. Say, yeah, just buy it once it looks good to you and you approve it. But.

Bilal Tahir (33:14)
Right.

Right.

No, that's a good

idea. I bet you somebody is going to do this, but I think it's fun. Somebody is going to build like a bot that focuses on a particular niche where you can get better results. But what they the way they monetize it, I think this is how you monetize. You make it all affiliate links to your account and then it shows and then the agents are purchasing just through your affiliate account. Or maybe there'll be a SQL injection cooler where as soon as an agent is trying to go to link, you you.

Pierson Marks (33:37)
Mm. wow.

Bilal Tahir (33:49)
add in your affiliate link or something. You're gonna see all kinds of stuff like this happen.

Pierson Marks (33:51)
Hmm. Do

agents when open AI searches Google or whatever or Bing, do they ignore ads? Do you know?

Bilal Tahir (33:58)
do they ignore ads? Like, you mean like, do the agent itself like, or? that's a good question. That's a good question. Whether the results of Scrape are they sponsored or not? ⁓ That's, I have no idea. I've never really looked at that.

Pierson Marks (34:02)
Yeah, does the agent see when.

Right.

Yeah, because I like, yeah, that'd be interesting. I know usually it's against terms of service to use it affiliate link and sponsor ads. But that'll be interesting there. And also the ads that like there is going to be an ecosystem of AI agent ads where.

Bilal Tahir (34:28)
I mean.

Pierson Marks (34:29)
the copy that you write for an agent is going be different than the copy you write for a human. Humans are a little bit more hooky, clickbaity, whatever, and agents are going to be more of like verbose and kind of is this actually the right thing to buy? And so if you can nail that. So I think there will be like a...

AI agent ad ecosystem where you're buying ads so the agent can see your products. I was also talking to a guy, his name's Andrea and he has this company called Thread and it's like inserting ads into chat bots. And so as an advertiser, when somebody's like saying, like what...

what's the best AI podcasting platform? You could buy JellyPod and you put JellyPod in there. It doesn't work with Check GPT because they're not exposed, but there are a lot of like third party apps on the app store that are just like using APIs behind the scenes that create chat. And they're like in there. And so it's just interesting because that's a great, like if you're searching for something in GEO or AEO or whatever, and you're asking for a specific product recommendation, you know, as a brand, you want to be there either organically or paid.

Bilal Tahir (35:16)
Right, Yeah.

Pierson Marks (35:32)
because they're very, very high intent. So once chatgbt adds ads in, in search results queries, like that's going to be every marketers like number one priority. How do we get

Bilal Tahir (35:41)
Yeah. I believe

Gemini already has it and they actually don't even make it opt out. So you have that which I mean, it makes sense to me. I don't know. People are freaking out about it, but it's just better ads. It's like search. It's very good. I think it's equivalent.

Pierson Marks (35:48)
Interesting.

I don't think people dislike ads if they're relevant. If they're not relevant, then people hate them. But they're relevant.

Bilal Tahir (36:01)
Right. I mean, as long as it's clear,

it's a lie. I I think the bigger thing is where the AI, you ask an AI question, it says, do this, but admin knows to you that was like a sponsored thing. What I think what happens is they recommend stuff and then you see it with searchers, like you can see that's a sponsored post. As long as you have that distinction, some sort of badge or something, makes sense.

Pierson Marks (36:12)
Right.

Right.

Right. There's probably some interesting techniques right now to create a lot of pages that are paid pages or promoting on Reddit or doing some shady gray hat stuff to get into the trading data by paying today and existing forever in the future models. Well, episode 16.

Bilal Tahir (36:31)
yeah.

So I know you have

Pierson Marks (36:46)
at 20. We're almost there. But I'll get this one out, and ⁓ we'll talk next week.

Bilal Tahir (36:51)
Yeah, sounds good. All right, take care.