Creative Flux

Bilal Tahir (00:00)
Hello, hello, we're back. Episode 19, we took a brief hiatus last week.

Pierson Marks (00:01)
Episode 19, we're back. I know. We had to take a

week off last week because last week was too busy.

Bilal Tahir (00:09)
But all the more content to bring you today because of that.

Pierson Marks (00:10)
Over here.

Right. I know we

have a lot of stuff to talk about. We got our first hater comment. And if you're listening, yes, we accidentally mispronounced and misspoke, think, on one of the episodes about X-402. It's X-402. It's not X-420. Even though Elon probably wants it to be X-420, it's X-402. So you're right. But we corrected ourselves in that episode.

Bilal Tahir (00:20)
Ooh.

Mm-hmm.

Pierson Marks (00:36)
So we love it. We'd love to see it. If it's your first time here listening to Creative Flux, I'm Pearson and this is Belol.

Bilal Tahir (00:43)
and I'm below, nice

to meet you. And you know what? I don't know, I like, I want haters, you know, it's like that quote, first they ignore you, then they hate you, then you know, you, then you win, but I was gonna come up with something, but yeah. That's what it is, yeah.

Pierson Marks (00:48)
I haters too.

Then you win. Just win. We win. No, we just win. We just win. That's what it is. We're winning. We're winning,

And yeah, so the last two weeks have been really cool. mean, there's been so much that has happened in this ecosystem, in the generative media space, in the AI space, in technology. mean, OpenAI just completed their for-profit conversion. That was very.

Bilal Tahir (01:18)
Yeah, they're rumored

to have a $1 trillion IPO apparently, is insane.

Pierson Marks (01:22)
All right, all right. That will be insane. And it'll

also be insane because I'm completely unrelated to generative media, but like Robinhood, how they back a few months ago, they announced like tokenization of stocks and they had, they were given like the ability to invest into SpaceX and OpenAI. And then OpenAI came out, I was like, this is not affiliated with us, like blah, blah, blah. And so it'll be interesting to see how that plays out.

Bilal Tahir (01:37)
Hmm

Right.

Pierson Marks (01:50)
And if

OpenAI does go public in 2026 at a $1 trillion valuation, which would be the by far like the biggest IPO ever. ⁓

Bilal Tahir (01:58)
Yeah, well,

second biggest here because Saudi Aramco was 1.7 trillion, is, but that was like a ridiculous outlier. And it was like a state owned, like biggest oil empire in the world. But again, yeah, for the US market is the biggest, but also, and I mentioned this previously, I've always said one of the, if you are a public investor, you don't have an in, you know, there's a Robinhood tokenization. And I think that's genius. And I don't know how openly I can fight that because it's like, you know, they don't really control it. You know, as I guess they can.

Pierson Marks (02:07)
Right, right, right, right.

Bilal Tahir (02:27)
try to not share that they can block their investors to not sell it on secondaries. But I think the second way is through Microsoft, because Microsoft, the other news that happened, and this was something on the cards, Microsoft was always going to get a chunk of OpenAI. We just didn't know how much.

Pierson Marks (02:28)
with it.

Bilal Tahir (02:44)
the guess was anywhere from a quarter to a third, and it ended up 27%. So basically what that means is if they go, if the IPO for a trillion dollars, $270 billion goes to Microsoft. And I think I'm actually kind of surprised Microsoft's stock is down. had an amazing quarter. Google had an amazing quarter. They blew it out of the park.

My Microsoft is down right now and I feel like the street hasn't caught up to that. So this is not a financial podcast, but I feel like Microsoft definitely will get a huge upside of opening eyes. So I think that's an interesting play there.

Pierson Marks (03:07)
Well, totally. Well, tying this into gender of media, mean, so this

is your first time listening. try to focus on gender of media, but we have interests in all this different space. We're builders. We work on an AI podcast studio, but we like the media space.

Tying that into media, mean, the interesting thing for me with Google's earnings yesterday is we've talked about this in the last 18 episodes, like Nano Banana, VO3, Genie 3. Even though Genie 3 is not publicly available, even Gemini. I mean, Gemini 3 is not out yet, but.

Bilal Tahir (03:40)
Yes. Yeah, yeah. In Gemini, I mean, pretty awesome. Yeah, what

the hell? mean, I thought it was out like there's every day somebody says Gemini 3 out today. Like literally I see that post, you know, I'm like, all right.

Pierson Marks (03:53)
I wonder where the polling market

is. This is the thing in living in SF and not saying that I do this at all and not saying that anybody should do this either, but like you have these prediction markets where if you just hear a like, you know, tweet in the wind, you're like, yeah, yeah, like Gemini 3 is coming out this week. VO4 is already being trained. This is something we were talking about the other day is like, hey, like there's some, some like rumors that VO4 is already trained or is in training right now. So it's like,

Bilal Tahir (04:10)
markets are crazy. ⁓

Pierson Marks (04:19)
There's a lot of arbitrage and alpha opportunity living in San Francisco on these. It's crazy.

Bilal Tahir (04:21)
I feel like the prediction markets are

wild. It's so interesting to me. And I feel like I need to, I've never put money on it, there's some, I'm like, how do I, like there's literally, there was a prediction market going on that with a 12 % chance that Satoshi Nakamoto is Donald Trump, which I'm like, you think Trump understands advanced cryptography, he made Bitcoin? Like, like, are you serious? Like that's obviously not true. So what is the,

like what's the play here? Like, you know, is it the fees or like, you know, like it sounds like you're just selling dollars for 90 cents. So those kinds of ⁓ prediction markets are interesting. The, again, not to get into politics, but there's a very interesting prediction market with the Mamdani and Como right now going on in New York because Mamdani basically is the favorite to win the mayor of New York race at 90%. But recent revelations showed that most of the money on the prediction market is coming from state actors.

Pierson Marks (04:49)
Right.

Bilal Tahir (05:15)
And

actually there's a very interesting second order effect where if someone sees a prediction market that's 90%, they might think, he's going to win. And so they don't vote or whatever. So you can actually, it's like kind of a, do call it? The George Soros call it reflexivity where the prediction itself can then impact.

how the race goes out. And so that a lot of people on the other side of whose commos backers, especially like the billionaires in New York, they're like, well, this, we need to investigate this. This is like tilting in his favor because there's some fishy thing here and the race might be closer than we think it is. very interesting stuff. Prediction markers, I feel like we're gonna know a lot more. It's like the wild, wild west right now. It's kind of like crypto in 2017, IKOLAND where there's no regulations. There's a lot of things are up in the air. It's a...

It's a very fascinating thing. I am very curious to see how the next two years play out, how this gets buttoned up. ⁓ Very fascinating field. ⁓

Pierson Marks (06:03)
Yes.

All right. All right.

Bilal Tahir (06:09)
Also, by the way, I mean, know I'm harping on prediction market, but I actually didn't know. So the founder of Carl Sheepe is these two MIT very smart people. They did a profile on them. I had no idea how these people like the founders. And they were young. And one of the founders, ⁓ the CTO, she actually decided to sue the regulator because they didn't let them. ⁓

Pierson Marks (06:30)
All right. This was last year. is right

before the election. This came out right before the election.

Bilal Tahir (06:33)
Yeah, and they won. And that's one of the reasons

Kalshi blew up, I feel like. So that was a clutch decision, but super smart. I really admire people who take those kind of gutsy decisions.

Pierson Marks (06:39)
Yes, Kalashii versus SEC, right?

right? You have

to have some money. have to like, have like, you need to raise a bunch of money for that. ⁓ Something to the Supreme Court. mean, it's crazy. So

Bilal Tahir (06:47)
Yeah. Oh yeah, they raised what? 100 something. Yeah, that was crazy.

But anyways, coming back to generative media, lot has been going on the last two weeks and I know you went to a really cool conference about it. So tell us like, know, which conference and what was it about? What happened?

Pierson Marks (07:06)
Yeah, totally. ⁓ so

last week, the reason why we skipped last week was because we are one, we're at the next JS and AI SDK conference. So that was cool. Much more developer focused talking about next JS talking about AI SDK, which is a way just to integrate AI into your apps more easily. but in the theme of this podcast Friday foul, ⁓ formerly known as features and labels, they had their conference called. yeah. Yeah. Yeah. Yeah. Yeah. So.

Bilal Tahir (07:30)
I actually didn't know that was their name. Is that? Featured labels. Interesting.

Pierson Marks (07:35)
They had their first ever generative BDA conference and that was really cool because it was small, was about like 200, 150 people and it was hosted at the Ferry Building here in San Francisco and there were just everybody in that room were people, either researchers building image generation models like Andres, I forget that, from Black Forest Labs. ⁓ He was there. ⁓

Bilal Tahir (07:53)
⁓ yeah, Cool.

Pierson Marks (07:57)
the Moore twins. And then there was like Jeffrey Katzenberg, who is the founder and CEO of DreamWorks. So he was there speaking about like Hollywood and how and gender media is impacting and he.

Bilal Tahir (08:12)
I know he probably

feels the what did the company went under the short form thing they made right. Yeah, I like the age of AI. If they do Colby now, I mean, that could have been could have been a different story right with AI.

Pierson Marks (08:17)
Quibi or right?

Yeah, yeah. No, totally. Quibi was like

this really funny, stupid. It kind of blows my mind. In the startup world, of the major, one of the important things as the founder is to sell the vision and sell this dream. But, I don't know. Who in their right mind was like, oh, we're just going to invest. It just didn't make sense to me. It felt like it was going to be a failure immediately. And I mean, that's like a...

hindsight sort of bias or whatever. But it just never made sense to me like how this Netflix competitor that was limited to your phone would like, oh, we're just gonna make shows specifically for vertical extreme. I was just like, what?

Bilal Tahir (09:02)
right?

Right.

I think

actually the idea made sense because of the trend. I'm not a consumer, I know you're not like of the short form, but you'd be surprised how many like people and Americans like they love short form dramas. And actually I think where they messed up was the cost model. Because what they did was they hired these celebrities like giving them like $10 million for like a 10 minute video and stuff like that. they really wanted, their whole thesis was kind of go big on the marketing and hire premium actors.

And you know, they just, the cost model never made sense. So as soon as, you know, they didn't get this immediate takeoff, everyone, the investors were balked and everything went under. However, it's funny because they were there, actually, if you go on the app store right now, the number one, I think if you look at the top 10 media apps, about...

four or five of them are these Chinese apps. They actually, you wouldn't know they're Chinese because they have like cool English sounding names and even English actors, but it's literally Quilby. And what they do is, and this was what Quilby should have done. What they did was they hired these B grade actors, know, who are up and coming. They don't have any, you who you work for next to nothing.

And they do these dramas where it's, and I'll go on, I've got to be honest, it's basically catered towards middle-aged woman. So it's like literally rom-com stuff, like the billionaire and, you know, like, you know, who's a bad boy, whatever, kind of like a semi-erotic kind of a thing. And, you know, it's kind of like a...

Pierson Marks (10:32)
It's great.

Bilal Tahir (10:34)
not like HBO, like Hallmark. Hallmark, edgy Hallmark. And it's literally like Slop, it's the most generic plots, but they end every two minute episode on a clip hanger. know, am I going to marry him or not? And then boom, end, two minute. And then they literally charge ridiculous fees. So they'll do like five episodes for free. And then for the next episode, you pay 20 bucks or something. And they are killing it. They're making insane amounts of money. And so...

that and tying it to AI, I feel like there's a huge opportunity there where you can just literally generate these two minute short form clips and stuff. Because ⁓ as these Chinese apps have shown, people just want brain or they don't want to see Brad Pitt or whatever. I mean, they're fine if it's some no name actor, as long as the plot is something they're interested in, invested in. And the plot isn't that interesting. I mean, it's not original. You can literally chat to videos right now and it's something that'll hook them.

Pierson Marks (11:28)
Well, totally. mean, this is super interesting. mean, it ties back into everything, the theme of last Friday, like the what will media look like in five years from now? ⁓ You see we're starting to see this with Sora 2 and the app there cameos where, you know, if you listen to a previous episode, we're talking about how you can ⁓ upload your likeness to Sora and then other people can remix you and make one that went viral this week was like Mark Cuban was like, hey, just use my likeness and create as many videos as you can with it.

Bilal Tahir (11:37)
Yeah.

Yeah.

Mm-hmm.

Pierson Marks (11:58)
funny they took like Mark Cuban and they put him in like a McDonald's outfit and they made him flipping burgers getting fired and like that was funny and and it'll be interesting to see like as the cost of creation comes down like that was the theme is like hey the way that we've produced content in the past is dramatically changing like I've always believed this I think you believe this as well but

Bilal Tahir (12:18)
Mm-hmm.

Pierson Marks (12:21)
I think 2026 isn't gonna be this year of agents, it's gonna be the year of creators. I think that we are right along, I we're right before.

Bilal Tahir (12:26)
Yeah.

Pierson Marks (12:30)
this chasm that is about to break open where we have cheap, we have pretty cheap image models. They're good, they're functional, like can edit, like Nano Banana is awesome, you take an image, you can make edits to that image, it's the Photoshop killer, as people say. Like yesterday, Adobe, I think it was yesterday, but Adobe teased or released this like comfy UI sort of workflow product where you can take notes.

Bilal Tahir (12:31)
blessed.

Yeah. Right.

I think they added Firefly

or whatever in Photoshop. And I know Bilal Al Sidhu, who is great influencer in this space. made this tweet, Photoshop is the next Photoshop, which as much as I hate Adobe and their lock-in, I can slowly see Photoshop just absorbing, kind of like Chrome. It's kind of like Chrome at this point, where you just absorb the features. And even if you come up with the AI native browser, people are just going to use the thing that they're used to.

Pierson Marks (13:00)
Right.

Mm-hmm.

Well, totally. I think that the key thing is too, it's like you have to really focus on your demographic of people. Photoshop, I do believe that. I believe that there is going to be, you'll always have a trade off because Photoshop's gonna be a powerful tool. It's always gonna have the buttons. in order to be Photoshop, you need to have that granular sort of control when you need it. And that's what they've done. Photoshop for the last. ⁓

decades has just been like, you could do everything you can imagine. It's going to be hard, but it's very powerful. You just have to learn interface and then to layer on top of that or side by side, like this node based workflow where it's like AI native, where you, generate an image and you connect that image to like a text based, like prompt and you make edits and like stuff like that would be awesome. That's gonna be really good. But you have to build that experience where like

They're probably doing that so that they can expand their market, so they can get access to these newer creators. But how do you do that in a way that doesn't make it more difficult for their advanced users to still have that same powerful features and cycle between user-friendly, AI-native, easy-to-create interface while also giving the features that you need to you know.

is precisely crop, precisely select some edges and like doing all those layering workflows. So.

Bilal Tahir (14:46)
Yeah. Yeah. It's a delicate balance, but I feel like they are in a not as bad a position. I think they'll just lean towards the advanced use case because, you know, we, you know, we have so many, like the canvas of the world, FIGMAs. feel like a lot of, we have a lot of tools for the others at the beginner end of the spectrum, but you know, the advanced, it's kind of like a blue ocean for them, which makes sense, right? Cause you have to invest so much money. So it's kind of a moat almost. have decades of like investment they put into this tool, right?

Pierson Marks (14:47)
Yeah.

All right.

Mm-hmm.

fully agree.

Right, right. mean, this is philosophy I think that we also have in Jellypod where we've you build a foundation. Like if you if you build your AI native company from the ground up, understanding where the failures of AI is today, where you give like the human the ability to go deeper like AI is not perfect. It's non deterministic. Like you can't necessarily always guarantee that

what AI is going to produce is going to match exactly what you want it to. And you have to build, you'll always have to build around it some sort of safeguards, some sort of harness, or a way for you as a user to go deeper and do some manual tweaking. Like if audio doesn't sound right, like just having the ability to regenerate. If there is a thing in...

and an image is generated to be able to select that part of the image and crop it out or select the butterfly that was weirdly placed in this image and just like select, remove this. so giving more powerful features that allow a user to go deeper, that's the hard thing. I think that like everybody's focusing, oh, the models are gonna get better. The outputs are just gonna get better. We could build the UI and the UX to connect all these things and we can build upon that. But if you don't have the ability to go deeper and kind of...

Bilal Tahir (16:11)
Right.

Pierson Marks (16:34)
win over market share now with the limitations of the current models, then it's going to be a hard uphill battle because when the models just get better, it commoditizes the playing field. And so if you don't have like the engineering beneath these models to allow like when the models get to 99 % accuracy to go from 99 to 100, because everybody's going to be at 99. And the last 1 % is going to be the differentiator.

Bilal Tahir (16:44)
Right.

right yeah

Pierson Marks (17:04)
to actually be production grade, be used in Hollywood, to be used in a business setting. So Photoshop is very well positioned because they already have that infrastructure.

Bilal Tahir (17:13)
Yeah. And as you

and I know, it's software, I feel like pretty much any field, the difference between 99 and 100 is actually huge. It's hard as what is the quote? Like the first 80 % and then the last 20 % is another 80 % or something, right? Like, cause it does make a huge difference. so, yeah, I agree. think that's that gap. It increasingly will become the differentiator of like the, you know, small, so it's almost like getting to seven nines or whatever.

Pierson Marks (17:28)
Right, right, right.

Totally. Totally.

Bilal Tahir (17:42)
But I

want to circle back to something you said though, very interesting, because you said like 2026 is going to be the year of creators and stuff. And I believe that. And I actually want to tie this to this conference, is this, would you say, was this the first generative media specific conference? feel like this is like a sign of a come, like, do you feel like this is going to increasingly carve out its own space within generative AI where there's generative media of, you know.

Pierson Marks (18:06)
100%.

Bilal Tahir (18:07)
creators and influencers. It's almost like, know, YouTube kind of gave a rise to influencers. So now there's literally YouTubers, right? That's what they do full time and they have their own conferences and stuff, podcasters. And I feel like maybe there's an AI kind of a...

like creators or gendered media who are half technical, like they understand the tools, maybe they're not into the codes, but they understand the models. They're jumping between Kling and Vl3 and all that, Like the more twins, et cetera. So it's become its own niche almost. think this is going to probably, for anyone listening, feel like this is a great time to get into this community because I feel like this community is going to 10x probably next year, right? It's going to be like, my God, mean, some of these names like PJ and Justin Moore, they'll probably

Pierson Marks (18:33)
Well, absolutely.

Totally.

Bilal Tahir (18:48)
become like huge influencers, right? And people will, you know, they'll be on TV and all that. And then, you know, you get to say, wow, I saw them when they only had 50,000 followers or something, right?

Pierson Marks (18:59)
Right. Yeah,

I completely agree. mean, it's like before today, you had to, if you wanted to create any sort of film or any sort of content, you had to either record it or you had to be like an expert with a team of CGI and, you know, like a whole production studio to create some animated film. Like ask anybody who's ever created an animated short, like how much effort goes into that.

Bilal Tahir (19:23)
Yes.

Pierson Marks (19:23)
or any sort of content that's like high production value. Like it's a lot. And is it a lot because it needs to be a lot? No. Like what matters? It's just the realization taking, you know, your idea and that storyboard, that good story and the idea that you have in your head for your characters and their outfits and the scenes and everything and getting that into something that's consumable by others. Like that process in between.

Bilal Tahir (19:35)
Mm-hmm.

Pierson Marks (19:48)
is gonna go from a hard, very hard thing to easy. And so what's gonna matter is just those ideas and stories. I responded to somebody on LinkedIn the other day and I was just like, I fully believe that giving more people the ability to be creative and share their story.

Bilal Tahir (19:53)
Mm-hmm.

Pierson Marks (20:06)
without the technical burden overhead of figuring out tools. It's just a positive thing and we can just get more high quality content. Movies are going to get better. Shows are going to get better. They can get cheaper produced. I think that we're entering the age of multi-billion dollar budget or not billion multi-million dollar budget movies. Marvel, Disney, like they know this ⁓ and I think you're going to see now where to have a blockbuster you don't need to spend

Bilal Tahir (20:11)
Mm-hmm.

Yes.

Yes.

Pierson Marks (20:35)
a bunch of, it changes the business model of Hollywood because you can spend five million dollars and have a 10x return on a 50 million dollar box office. Whereas now you can't do that because you can't create a 50 million dollar box office movie on a five million dollar budget. Like it's impossible. You have to kind of go like, I'm gonna.

Bilal Tahir (20:38)
It does, it does.

Mm-hmm.

Pierson Marks (20:54)
I'm gonna spend a hundred million dollars to hopefully get a five hundred million dollar budget and you go big and because you have the actors and you have all this stuff but like you're gonna see all these indie films and people that have smaller budgets to have a massive ROI because the films are good like squid game or like these things and they'll get bought up by Netflix so they'll be on YouTube exclusives you know so

Bilal Tahir (21:15)
yeah.

I mean, you're going to see some guy, you know, like, ⁓ like one of these like Dave Clark's or whatever, you know, like people who have made amazing films, they're going to like put out like a series of show and then you'll get a headline, ⁓ Netflix buys them for like 50 million. And suddenly it's going to be like, you're like, wait, he just like prompted his way to 50 million. Like, can, I can do that.

Pierson Marks (21:34)
It's so, so

cool. Like to be honest, I've always told people that engineering, specifically software engineering, is the coolest career because it's like playing with an infinite set of Legos that you just, but like infinite set of Legos for free. Like all you need is a computer in front of you and you can build anything. Like you could build with a single computer and a good idea, you could build a billion dollar company.

Bilal Tahir (21:49)
Mm-hmm.

Pierson Marks (22:02)
That's all you need. There's a little computer. have an infinite set, infinite playground. We're entering the same age of creators where all you need is a computer and an idea and you could build a movie. You could build a show. You can create a hit song, a Billboard top 100 song like...

Bilal Tahir (22:13)
Right.

Pierson Marks (22:21)
And it's going to be amazing. It's going to be so cool. You're going to see new opportunities for creators. That's more jobs. People are going to go and go back into what makes us human. Besides, know, people are like so worried about AIs coming for our jobs. Amazon, they had 14,000 people laid off. All this stuff. Like, what are we all going to do? mean, nobody's meant to be sitting behind a desk, pushing paper, responding to emails. It's like...

Bilal Tahir (22:32)
Right.

Right.

Pierson Marks (22:46)
Go build that movie. go pursue those passions, those creative passions. That's what makes us human, you know?

Bilal Tahir (22:48)
Right.

It's so fascinating. ⁓

It's interesting. It harkens back to, you know, in the 60s when the whole monopoly of the cable television was starting to break and people were, camcorders became small enough that people could record. I remember Andy Warhol had the famous saying, in the future everyone will be famous. And similar, I think he also coined the 50 minutes of fame phrase. I think that might be wrong. But both of these courts come from that era. And the whole idea was like these people saw how the democratized

democratization of like creating film.

is going to lead to more people just having ideas. And now I feel like, we saw that if you, it's like kind of the same curve because, you know, it used to be like three channels, only big budget movie stars, Elvis Presley, et cetera, like Beatles. Then it was like, ⁓ it was like you invented like, you know, the Ableton or whatever. And suddenly, you know, you got all these like garage bands and stuff and you get people creating shows and then YouTube happened and you had a small indie directors, et cetera, being able to do more shows, more shows.

amount number of celebrities quote unquote has increased you know they've kind of like it's almost like the pie of attention has been like shared by more people and now and grown and grown and grown and now it's just gonna that this is the next step where literally

Pierson Marks (24:00)
and grown and the pie grow.

Bilal Tahir (24:08)
you and I or anyone with an idea in our head and we don't need to understand the nuances of film and stuff or have heavy equipment or stuff like that. We can just be directors and we can create these shows and put them out there and maybe most of us don't get millions and millions of views but we can carve out our own niche and maybe a hundred people care about your weird steampunk, cool, dystopian kind of fantasy, whatever.

right? And you know, you go from there. So it's gonna be, the world is gonna be wealthier for it, you know, because you're just gonna have more ideas out there, more people sharing. ⁓ So yeah, it's gonna be an exciting time. I'm very excited for all this stuff. If anything, I wish now they invent the other cyber, I can just have some sort of wire in my brain that can just absorb all the information because there's just so much cool stuff out there. I just want to see all of it, but I just don't have time.

Pierson Marks (24:45)
Totally.

I know, I

I was talking to somebody about this the other day. It's like, we're talking about prompting and this will be the same issue in coding. This will be the same issue in like media gen. Like you have an image in your head and you have to express that image in your head via prompt. And that's very hard, know, still like.

Bilal Tahir (25:17)
you

Pierson Marks (25:20)
That's a massive skill like whether if you're coding I mean you take a paragraph you kind of have to be very specific You can't be like, you know, you're stupid like fix this bug or like please fix or like it's like

means nothing to these things. You have to be very, very explicit and be like, hey, like this is wrong. Really like, like look at this, do this. And with image generation too, like you also have to be very, very explicit and verbose and kind of in your intent. And that's difficult because we're never, we've never really.

Bilal Tahir (25:42)
Yes.

Pierson Marks (25:50)
had to do that because in creative fields, it's more of this flow. Like you get into flow state and it kind of just comes out and you work on, like if you're painting, you don't really think about how you're moving the brush and like the, it just kind of, you're moving the brush in the way that feels right to you and you paint your trees and you paint your things. And it's not like like a actual.

You don't have to put into words about what you're doing. But right now, like the medium is words. And I always thought it'd be super cool as just like, this is completely sci-fi and I don't think it's on anybody's radar, but like Neuralink is so interesting to me. I think it's super cool. But imagine if you could put on like a Neuralink or whatever.

Bilal Tahir (26:14)
Yeah. Yes.

Right.

Right, and just think the prompt. Well, we had the,

talked about the startup who was making that thing where your thoughts can be transcribed or something,

Pierson Marks (26:34)
Right, right.

Totally,

But think of it, because your thoughts are very messy as well, but what if you can combine your messy thoughts with a prompt and then a language model is trained both on tokens from your prompt input, but also on your brain waves? It kind of sounds crazy, but...

Bilal Tahir (26:41)
Yes.

Mm-hmm.

Pierson Marks (26:55)
It might be able to, if you combine those two inputs, because when you're typing, you have a thought in your head, but you're not being verbose enough about that. You can combine the verbosity of what your brain's kind of thinking, but that's super fuzzy and super hard, and I'm not a neuroscientist at all to know how that works. But you combine your thoughts with a prompt into the model, and it's like, like.

Bilal Tahir (27:04)
Right, yes. ⁓

Thank you.

Pierson Marks (27:20)
this is like, I get what you're trying to say. I know you're not saying it well, but like, I know the image you're trying to create, because you're imagining it and you're also telling me and you're kind of confirming it. And it's just like, wow, you put on your headset, you type and you go like, create my outer space spaceship.

Bilal Tahir (27:28)
Yeah. ⁓

Yeah. 100%. I mean, I would

love that. As someone who struggles with, for me, it's like pulling teeth, like trying to write down exactly what I'm thinking. It's so hard. And it's why I feel like currently the people who are amazing prompt, it is a skill because some people are just very good at articulating what they are thinking or at least doing the work and putting, writing it down. It's such an alpha, but I would love for a world where that...

gap is closed. And actually on that point, there's a very interesting, it's funny you said that because there was an interesting model that came out, it was called Structured Prompts by Bria. what they do was, and it was like such an obvious idea, I thought, because what they do is essentially the concept of a picture is worth a thousand words. So what they do is, let's say you generate an image like a flying dog across the New York City skyline, and that's your prompt.

kind of get a flying dog, you get some building in the background, okay. And what they do is they take that and they literally generate a thousand word, a really long structured prompt where they're like, let's get into the details. Bottom left, you see a street, three people walking, one is wearing a...

a woman with a red purse and green, and then top left, the buildings, the sun is shining. Like all those little details that you were too lazy or you didn't even think about, it takes that basic prompt and really divides that image into all these like small subsets. And I'm like, that's a more granular way of getting the exact image you want, right? Because then once you have that, it's much easier to then edit that prompt. can be like, all right, yeah, but bottom right, I actually don't like the sun is too bright. I was actually thinking it would be like a sunset. You can just update that.

part and the beauty of a structured prompt is everything else can remain pretty much the same and you can just edit that. And I wonder if you use tooling like that until we get the Neuralink bandwidth, we can just have these tools where it lets you much more easily manipulate the image by getting more comprehensive about using the LLM to just have a much more ⁓ bigger prompt. So check the structured prompt out. I think that was very cool.

Pierson Marks (29:25)
Totally. Totally.

Yeah, it was cool. They were there

on Friday. And they also, the other cool thing about this was that it was trained in 100 % licensed data from FreePic, from Getty, from like all these image providers. So they have the licenses for all the training data. So as an enterprise who's like worried about the copyright issues, they're like, hey, this is 100 % licensed. All the data we own, we got the permissions to train a model on this. And then they also have this attribution.

Bilal Tahir (29:35)
Yes.

Pierson Marks (29:54)
And so I think the reason how they were able to get these image providers to agree to share their data with Bria to train this model was that not only will they pay them upfront for the use of the image, but every time that an image is generated, a percentage of that cost will go back to those image providers. So they built this attribution sort of

I don't know what they're called, attribution, detections, whatever. like, it knows, hey, this image that you just generated had these kind of base images that would influence the output. So it was trained on these base images. And you have a percentage of, hey, this image influenced the output by 50%. So it gets 50 % of the attribution. And then you scale that.

Bilal Tahir (30:33)
interesting now.

That's so interesting. That's

fascinating. So if you generate an image and it knows, OK, 1 20th of this image came from images you contributed. So 1 20th of that. And then they kind of total their sub and add it. And it's kind of like the Kindle library. They take percent of pages read. It's like how many people read pages of your books, basically. And then they divide it among the total dominant. And then that's your numerator and then the denominator. So that's your denominator. And then the number of pages read. And then they give you that percentage.

Pierson Marks (30:53)
Exactly. And that makes total sense.

Bilal Tahir (31:14)
very good way of dividing this attribution, know, royalties up.

Pierson Marks (31:19)
Totally, totally, because you have to incentivize those people because you know that they'll sue. because especially if it's just not like, you, everyone's like, just pay for your data once. mean, you like you have to like, why would they just shoot themselves by saying, hey, give us a lump sum payment upfront to train. And then you're like, okay, well now your business is dead because they'd never need to come to you anymore. So.

Bilal Tahir (31:35)
Yeah.

Yeah.

I mean,

it's a scary time. feel like, you where I was, you know, it's been a very, a lot of people, companies have been laying off people like this week's scary. And one of the thoughts I had like, you know, imagine maybe AGI is like, anyone can build a SaaS by like one prompt, but then everyone's broke because they're all unemployed. So nobody's there to pay for your SaaS. It'd be a weird mid curve thing, you know, I wonder what's going to happen next two years with a lot of these layouts. I feel like

I do think there's going to be a mass unemployment. Unfortunately, think that's like, you know, it's coming, it's really a white collar one. So.

Pierson Marks (32:14)
Totally, no, I

completely agree with you. But the thing is that, there's going to be laughs. And we've talked about this in the past, and we've talked about this personally. mean...

The economic structure of the world that's worked in the last 30 or years, maybe even longer, probably even more like 50, 60, 70 years, was that hey, graduate high school, go to college, study something in college, and a lot of times you'd get like people like become a lawyer, become a doctor, become an engineer, study like STEM kind of or whatever, then go to grad school, go to the corporate workplace, climb the corporate ladder and go up. that's kind of just like...

Bilal Tahir (32:37)
Right.

Right.

Pierson Marks (32:53)
Every, just that model's broken. I am a huge believer in college. I think college is amazing. I had a personally amazing experience. I think everybody should go through that. It pushes you in its heart in certain ways and I will always be pro college. But the debt that a lot of people take on for a degree that doesn't necessarily lead them to that corporate workplace. And if it does, then you also.

Bilal Tahir (32:57)
Right.

Pierson Marks (33:18)
in a world that is increasingly becoming more automated, you go to a big company, the big company is probably not the place to go. Like I do not see places like Google, Amazon, Microsoft, like the big fanguy's of the world hiring like in any sort of more than like a single digit, like just fixing attrition. Like I do not see them growing for the next decade. Like I think that

Bilal Tahir (33:41)
Right. Right.

Right. Like, these

are like primarily the junior software engineers, right? I cause they're just, they're finding either they can automate the whole job or more likely what they do is the senior guys can just take on more roles cause they can automate stuff. can run agents like auto, whatever. they're like, yeah, I'd rather have 10 cloud agents or open a GPD agents versus trying to train some 21 year old kid. you know, this is how get her works. This is, and I mean, I guess the game is that if you believe it's just only going to get better.

Pierson Marks (33:47)
Exactly.

Bilal Tahir (34:13)
then the game theory is like, I don't need to train it. The other side is if you, it can actually shoot you in the footwear. You know, we end up in a world where there's just dirt. These guys were never trained. And so suddenly you have no mid-level engineers anymore. yeah. ⁓

Pierson Marks (34:27)
Totally. No,

completely agree. And I think there'll still be some hiring. just don't think it will ever be as big of a growth because yeah, like you'll see fewer promotions.

you'll see, and you'll see like forced attrition. You'll try to, you'll like prevent promotions and you'll force people out of your company. You'll do remote, like you'll say, come back to your office. You'll force people out of your company. You'll do all that. You'll do layoffs and get people out. And then you wonder, okay, so like the big companies aren't hiring, but what about the mid tier? What about the smart, the startups? Like I think this is just forcing this kind of.

Bilal Tahir (34:48)
All

Pierson Marks (35:04)
reorganization of the economy. Like you talk about like a big companies are getting bigger. Cool. But like small companies are going to get bigger. We're going to have more entrepreneurs because I mean you can build something like if and that's it's not for everybody and it's hard. So it's like especially not starting that you're on startup but like working at a series a series B startup that's kind of proven product market fit like

Those are the companies that you're gonna have to go to. They're not the shiny ones that everybody's going to, but they're gonna be easy. If you know somebody, if you could actually do research and you're passionate about a space, it's gonna be much easier to go to those companies like, hey, I love what you guys are doing. I'm passionate about it. I wanna work on green energy. I wanna work on renewables. I wanna work on these things that like...

Maybe it's no longer like, I want to work at Microsoft. Like I want to work on a more niche thing that like more directly aligns. They're growing super fast. There's a lot of venture capital money in there. They're going to be paying you a good amount of salary. You're going to get equity. And it's just going to be like, I don't think that the post-college pipeline.

is going to ever lead back into these big tech companies anymore, except if you're a researcher or something that's like super, super niche and you have like a PhD or master's, or like PhD honestly, to those companies. and obviously it's a generalization, but I think it's gonna, we're gonna enter in a massive period of encouraged entrepreneurship, a lot of small businesses, a lot of creative works, people going back into the creative fields and.

Bilal Tahir (36:22)
Right.

Pierson Marks (36:35)
We'll see how it plays out, but. ⁓

Bilal Tahir (36:36)
Yeah,

I mean, it's interesting. agree with you and I hope we get to that point. mean, maybe there's some sort of a UBI or something where, we just like, you know, it is an interesting thought, but when people don't have to worry about paying the bills or stuff, know, what will they do? I mean, I have a very...

I'm very skeptical that most people will do anything productive. I feel like it'll kind of be more like Brave New World where most people are just kind of like taking their Soma and just chilling and watching shows, which is fine. But then there's a certain percentage of people I think they need to be work. They're just driven and they need to be doing something. And I feel like we're gonna have this bifurcation in our society where most people will.

Pierson Marks (37:00)
Well, whether.

Alright, alright.

Bilal Tahir (37:16)
get the bare minimum and they just kind of are almost like zombies, unfortunately, and like kind of like the matrix there, they take the blue pill. And then some people will be like, no, I actually want to do stuff and they'll become ultra producers. So the producers will just produce 10x as much and then the consumers will consume 10x as much.

Pierson Marks (37:33)
Right.

Totally,

I mean, there's been so many studies, like OpenAI is their UBI study. They hoped that it would show that UBI increases happiness. never does. Like UBI does not increase the well-being of a society, does not increase the happiness because eventually you have a temporary boost in satisfaction and living, but then you kind of just, your living adjusts and ⁓ it becomes your baseline again. And people need a pursuit. It's the pursuit of happiness. People need a pursuit.

and a purpose to life and whether that purpose is like building a family, which is a great purpose and we need more of, know, raising children and everything. And then also just.

Bilal Tahir (38:10)
Right.

Pierson Marks (38:12)
Just like working on things and until, I always say this, until the world is perfectly efficient, until there is no more disease, until there is no more inefficiency in the way that we live and work and communicate and all these things, until it's perfectly efficient, there will always be things to do, things to work on, problems to solve. There are billion people, more than a billion people in poverty, still in the world, don't have access to fresh water, electricity, like still a lot of problems to go around.

until all those are solved. mean there are things to work on and yeah I mean that's what excites That's cool.

Bilal Tahir (38:46)
Yeah, I

mean, it's crazy. How did we get from gendered in media to this? But here we are. It's all related. I mean, gendered media is at the heart of it. I mean, it's going to play a huge role, actually. mean.

Pierson Marks (38:57)
generative media is the heart of it. about humans.

It's becoming human. It's like, what are humans? And we're creative beings.

Bilal Tahir (39:02)
I, yeah.

Well, generated media, feel like, you it's interesting. We talk, when we think of generated media, we think of software, we think of actual content that's created on screen, but you know, like one, there's more to it. Like if you think about, especially robotics, right? I mean, there was an interesting launch called about this company called Neo this week. And this is like this very kind of cute looking robot humanoid, but the...

I guess the big thing was it was only $20,000, which is relatively very cheap for a humanoid robot. And you could have it at your home and it can do dishes and stuff like that. And I found it very fascinating because it was tele-operated. actually, a lot of people call them out on that. They were like, oh, this is a fraud because you just tele-operate a human robot. It's not like full autonomy. But I'm like, you know, that's a great on-ramp, you know, because even if you tele-operate a robot.

you can imagine, it's kind of like having, when Upwork came out, you can certainly hire someone in the Philippines for like, you know, a fraction of the wage it would cost to do white collar work here, but for them it's a great wage and you can get that work done. And I feel like this teleoperation humanoid hybrid opens up that for white, blue collar work. Cause you can have someone in the Philippines or you know, whatever, Bangladesh, like hire them and they can do the dishes. They're tele-operating from Bangladesh, whatever. You're paying them a good wage.

Pierson Marks (40:14)
Totally.

Bilal Tahir (40:23)
from their market and you're getting cheap labor and it's great and maybe there's a generative media angle to it where now these human humanoids can be your companion as well where you can like actually you know feed them input about shows and your things you like and stuff and they can kind of you can almost like bond them like with animals so it's kind of like this weird like have

kind of a pet half household mate kind of a hybrid thing. I think we're gonna have more of those. It makes people uncomfortable, but this is the world we're gonna be living in.

Pierson Marks (40:53)
I

know it's interesting. I know I responded to your tweet about this. talked about this, I think, in the past, but in China, there are remote construction companies, teleoperated bulldozers and cranes and things. mean, when you're a worker and you're using heavy machinery, you sit inside of a track or you sit inside something, bulldozer or whatever, and you...

Bilal Tahir (41:02)
Yes.

Pierson Marks (41:19)
pull some levers, push some buttons, you look around. You're essentially sitting in a remote control like Xbox controller and so you could do that remotely. As long as the interface is recreated.

Bilal Tahir (41:28)
Right, yeah.

Pierson Marks (41:31)
in an environment that like matches and you have some screens, have cameras, like there's no reason why you shouldn't be able to do that remotely. And I could see that being really cool, especially as we try to encourage building and then we revitalize manufacturing and building in the country, which I really, really am optimistic and hopeful of. But when wages skyrocket and we're not able to do that, like how do we enable more building? And like not necessarily replacing, but like there's not

people that go into construction and labor works. Like if you could snap your fingers and say hey every construction project that is currently in progress and planned for the next five years can get done in half the time and half the budget. I think everybody would want that and just without replacing anybody like just like hey you still get paid the same amount as what's going to happen it's just going to get done faster and I think I don't know how you can be against that. You could just do more things ⁓ but there's probably people that are against that for some reason so. ⁓

Bilal Tahir (42:23)
Right.

I mean, there

always will be people like, know, whether it's like, it's just going to depress our pages and stuff like that, or, you know, just resistance to change. But this is the future. I think it's an exciting future and it will be a very interesting future. you know, we'll see how the U.S. ⁓ look like for generative media in the future. Maybe, I mean, we haven't even talked about like glasses and stuff. feel like glasses are going to be unusual with this too, but so many angles.

Pierson Marks (42:28)
as always. ⁓

All right.

Yep.

Yeah, totally, totally. Well, maybe

we'll talk about glasses and everything else next week. ⁓ Today was a fun one. We caught up on a lot. Two weeks of things to talk about.

Bilal Tahir (42:56)
Yeah.

I know we haven't even touched like

there's so many updates that happened, but you know like there's so many more video models were released. See that released a model of min max helio 2.3 came out LTX V2 amazing. we're going to get all of that. Hopefully next week we'll get to it, but.

Pierson Marks (43:09)
Yeah, I get to them all. I'm to get to

them all. So stay here. Get everything you need to know about gender media and life and stocks and not stocks, not financial podcasts. But yeah, so join us next week for episode 20.

Bilal Tahir (43:25)
Yeah.

Pierson Marks (43:29)
We'll see you. We'll talk next week.

Bilal Tahir (43:31)
All right, cheers.