Sam and Shaan explore the explosion of “right brain AI” — creative tools like GPT-3, DALL-E, and voice cloning that are transforming writing, design, photography, and music. They demo a stunning AI-generated Joe Rogan / Steve Jobs podcast, discuss Jasper, interior.ai, and AI product photography, then pitch three original business ideas: an AI therapist/companion, an AI Spotify with NFT ownership, and the opportunity in open-source AI. They close with a Billy of the Week segment on Emad Mostaque, the founder of Stable Diffusion.
Speakers: Sam Parr (host, co-founder of The Hustle), Shaan Puri (host, founder of Milk Road), Ben (producer/researcher)
Introduction: A New Era We’re Not Part Of [00:00:00]
Sam: There’s a new era and I’m not part of it — that’s kind of how I feel when I see this. How do you feel when you see this? How do you intend to get your hands on it? You know, we’re a bunch of scheming, greedy people — how are you going to get your paws in this game?
Shaan: Okay, this is the AI episode. It’s all AI — everything is AI. I was mind-blown. I’d say that’s the right word. To me this is the biggest “holy shit” moment I’ve had when it comes to technology, seeing what’s going on in AI.
Sam: For you — you got emotional, which is a strange response.
Shaan: I think I’m more emotional than you are. I get touched more than you do. Let me give the background really quick. There’s a company called play.ht, and one of their side projects is a thing called podcast.ai. Basically what they did was they gave their AI program Steve Jobs’s biography — I think they actually gave it one or two of his biographies, because there are two major ones — and then they gave the program every single Steve Jobs recording they could find on the internet, along with every Joe Rogan episode ever. And they made Joe Rogan interview Steve Jobs.
The interview is about 25 minutes long. They talk about all kinds of stuff. Joe Rogan does the intro just like he does in real life — he goes, “What’s up, freak bitches” — which I don’t even think he says anymore, but he said that in all the early episodes. He teases who the guest is without naming them, then he says something like, “He’s weird and brilliant and sometimes totally insufferable.”
Sam: Well, he says “insufferable,” which I thought was crazy. Let’s just play it.
[Clip plays: AI-generated Joe Rogan interviewing AI-generated Steve Jobs]
AI Joe Rogan: Hello freak bitches, welcome to another episode of the Bro Joe experience. On this episode I welcome my friend who’s difficult to describe. I’m fascinated by him, and I hope you will be too. He is weird and brilliant and sometimes totally insufferable. My guest today has made some of the great technological products of our age, always pushing the envelope in innovation. He developed a new programming language and operating system, then became even more famous for making three applications for that computer — a word processor, a spreadsheet, and an image editor. That just showed me this dude was brilliant, had amazing taste, and I would just hope I could be even one tenth of the genius that my friend today is. So yeah, I was super psyched about having him in the house today.
AI Steve Jobs: Good to see you, buddy. It’s been a long time since I’ve been on the show. I missed this. It’s always fun.
Reacting to the AI Joe Rogan / Steve Jobs Podcast [00:03:00]
Sam: Okay, just pause there. The voice quality is incredible. That sounds just like Joe Rogan. That sounds just like Steve Jobs. It has gotten so much better than this used to be. I remember when you had a Garmin GPS in your car and you could choose the Morgan Freeman voice — anytime you’d have that robot voice trying to say something new it sounded totally computerized. This doesn’t. So the voice quality is one thing.
The fact that it kind of makes more sense than it should — he’s doing an intro, it understands that he’s teasing, like “my friend who’s a genius but I’m not saying who it is yet.” That’s a showmanship thing. And the back-and-forth banter — he gives that long intro and then Steve Jobs just kind of laughs, like you would if you sat down for this kind of interview with Joe Rogan. I just thought that was incredible, the way it felt like a real conversation. Obviously there’s some stuff that’s like — he’s talking about word processors and documents a little too much, nobody would do that — but damn, if 80% of that is not amazing.
Shaan: And the reason I felt emotional listening to this was — I felt both awe and fear. As I was listening to it, I was agreeing with what Steve says. Ben, fast-forward maybe 10 minutes in. Basically Jobs goes on this long rant about LSD — saying it changed his life, he doesn’t think it’s for everyone, but it opened his mind and let him read all these books he never previously would have read. He says, “I don’t think I’d take LSD a bunch of times in a row because I only needed it once or twice and it opened my mind.” And he goes in-depth — he talks about India, he talks about God, he talks about all kinds of things in a way where you’re like, it’s not just like reading a Wikipedia page. It’s a fully-formed thought and philosophy, like something somebody would say when they’re having an in-depth podcast.
Did you hear the beautiful quote he says about India? Basically the AI is drawing on things Jobs actually said, weaving them together. He starts talking about India and why he loves that part of the world. He basically says, “It’s so old, and it influenced the rest of society.” And he goes, “There’s an Indian epic that’s ten times as long as the Bible, the Quran, the Odyssey.”
Sam: Yeah, that’s what he says. He’s basically saying Indian history influenced the rest of the world, and there’s an epic that’s ten times as long as all these other books. He doesn’t say the name, which kind of leaves you wanting.
Shaan: I believe he’s talking about the Mahabharata — which is like a 16-part series, incredibly long. But the point is, when he said that I was like, I’ve got to find that book. Jobs just sold me on this book. That’s the thing — the AI Jobs influenced me.
Sam: Ben, click play for just 30 seconds.
[Clip continues]
AI Steve Jobs: There is some kind of deeper meaning to life, and it can’t just be something that somebody made up — because if it was, it wouldn’t be compelling, it would seem contrived, and everyone would see through it. So I think the meaning and the purpose is of the cosmos — the nature of the cosmos. Which is pretty bold thinking. Taking LSD was a profound experience for me. LSD shows you that there’s another side to the coin, and you can’t remember it when it wears off, but you know it — it washes over you and tells you that everything is connected. You’re not here by accident. You were put here for a purpose, and if you can figure out what that is, then you’ll learn more about yourself than anything else could.
Sam: So that quote — “Taking LSD was a profound experience, one of the most important things in my life, it shows you there’s another side of the coin” — that is an actual quote of his. But it’s woven into the conversation. And then there’s a part where he’s talking about being a fan of Joe Rogan and he says something like, “I love to just sit back in the car and listen to you rant.” Where would that have come from? How does it know to say that to Joe Rogan in that moment? There are little moments like that where I just don’t understand this technology well enough to know how it could know to say something like that.
The first half of the interview he talks about India and these life-advice type things, which were pretty amazing. The second half he talks about Google — they actually say “Yahoo,” putting the emphasis on the wrong syllable — but he says things like, “That’s a problem I’ve always had with Microsoft. They’re smart people and they’ve done good work, but they’ve never had any taste, any aesthetic taste.” Then he goes on and talks about Adobe, criticizes their business model. Joe Rogan says, “What would you do to fix Adobe?” And Jobs says, “They’ve got 800 people working on this, and the fact that they charge just a little bit for a small piece of the product — it’s like buying a car but only getting part of the engine and having to pay more for the rest.” He then says, “I had lunch recently with Bob-whoever at Adobe and they actually agreed they need to fix it and they’re working on it.”
Shaan: It seems like he’s drawing on actual rants Jobs went on in the past, but it’s presenting them as current events. It’s pretty magical.
Sam: Yeah, this is basically “bring back someone from the dead” technology. Ben, do you have anything to add? You’re a history guy — what did this do to you watching this?
Ben: I really agree with Sam. It made me oddly emotional listening to it. It also scared me a little bit — just the thought of the application of, like, what if you did this to my grandpa? What emotions would that bring back? That frightened me. I don’t know if that’s a good thing to be able to experience — this representation of my grandpa that would feel like the real thing, but isn’t. So it’s equal parts compelling and scary, realizing it’s not really him.
The Deeper Significance: Steve Jobs Predicted This [00:09:00]
Shaan: It’s like a video, right? Today you can watch a home video — you see their face, you hear their voice, but it’s captured a moment in time. To me this is an extension of that, where now it’s going to be somewhat interactive. You can kind of interact with these people, or hear them talk about new things. There’s a Black Mirror episode that’s a lot like this where a woman uploads her boyfriend’s consciousness to a robot and is kind of still dating him even though he’s gone. I think there’s more good than bad.
Sam: Let me add two points about why this is going to get even stranger. In 1985, here’s a quote from Steve Jobs: “My hope is that someday, when the next Aristotle is alive, we can capture the underlying worldview of that Aristotle in a computer, and someday some student will not only be able to read the words Aristotle wrote but ask Aristotle a question and get an answer.” Steve Jobs wrote that in ‘85 — and what he just described is what just happened to him. It’s magical.
Now here’s where things get even more magical. Play.ht has a section where you can vote for new episodes. Some of the top ones people have voted for: Elon Musk interviewing Nikola Tesla, Kanye West and Bob Marley talking about music, Jesus interviews God, Einstein and Buddha having a conversation on science and spirituality, Trump interviews himself — twice — Lex Fridman interviewing Richard Feynman, and Joe Rogan mediates peace between Russia and the US.
That last one sounds like a joke — and it is — but it’s actually an interesting tool. Like, let’s hear Joe Rogan try to bring together two different people and find out where each is coming from, even if it’s make-believe. Maybe that’s actually how they feel, and you can work through the argument that way. So this stuff is really, really interesting and powerful.
Left Brain vs. Right Brain AI: A Framework for What’s Happening [00:13:00]
Shaan: Let’s switch gears. I want to give you my big-picture framework for how I think about AI.
I was having lunch with a guy who built an app called Wombo — it had like 100 million downloads, basically took a picture of you and made it look like you were singing. Then they had music licensing problems and got taken down. Now he has another one that’s a text-to-image creator on mobile, really popular. He was saying something that I kind of remixed, and here’s where we landed.
The last 10 years have been what I’ll call left brain AI. Your left brain is your analytical brain. That’s what AI could do: big data, machine learning, the computer plays chess and beats the best players, self-driving cars taking in sensor data and making decisions. That’s what we expected from a supercomputer — you should be able to drive perfectly, play chess better than a human. Great, I get that.
Then there was a game changer into right brain AI. Your right brain is your creative brain. This is where you got GPT-3, which generates text. You give it a prompt and it writes an essay, writes rap lyrics. That’s where you got DALL-E — you just say “give me a picture of a starry night but it takes place in Hogwarts” and it generates that scene. Then there’s what we just played — audio generation, whether that’s music or podcasts, just creating it from scratch. And now there’s video. There’s a company called Runway ML that does this — you describe a video, like “walking through the streets of Tokyo and it’s really busy,” and it creates that scene. You can say “add some rain” and it adds rain.
Sam: What’s it called, Runway ML? Can I go on right now and use it?
Shaan: Ben, pull up the demo for Runway ML.
Sam: Oh my God. So you highlight an object and say “remove this” — it just removes the street lamp. “Make a lush garden,” “make it look like it was hand-drawn,” “make it look like a jungle.” The last thing it did is so crazy — you dropped a video in, it goes green screen on the character, it’s a guy walking around, immediately removes the background. Then you say “add a sunny sky” and it adds one. Then “blur the background a little bit” and it blurs the background while this guy’s skateboarding. That’s crazy that it can do that.
Shaan: Now obviously a demo is generally a massive overpromise for what the tech can actually do. But we can all see that — if they’re not doing it now — it’s reasonable in the near term that this is going to be normal.
Sam: This is one of those things where you see something and say, “It’s not there yet, but it will be.” Almost like when I was in a Tesla for the first time. It was a practical car that also went 0-to-60 in 3 seconds. The range wasn’t good, other things weren’t good, but you get into it and you’re like, “Oh yeah — in 10 years this is just the norm.” It’s like puberty. The voice is changing but it still cracks. You’re like, all right, it doesn’t sound great yet, but it’ll get there.
Shaan: And it’s pretty magical when you see it. Right brain AI — creating art, music, audio, text — is bringing in a whole new wave of people. NFTs brought a new audience into crypto — there was a certain set of people interested in crypto as cryptography or finance, and then NFTs came along with the art angle, the celebrity access. This is that on steroids. The new AI stuff that’s creative is bringing in the Nerds and the Artists both, not just people who were interested in a chess bot. That’s my framework for what’s going on with AI and the next 10 years.
Interior AI and the Future of Design [00:21:00]
Shaan: So let’s talk about Peter Levels. He came on the pod — massive fan favorite, killer following, loves to just hack things together and make stuff. To me he’s like a craftsman, like a punk rocker who instead of playing on a musical keyboard is playing on a computer keyboard. He’s like, “I don’t do email or phone calls. If you have a question, here’s the FAQ.” He’s like, “I just want to work on my stuff and go swimming.” And he was doing that before remote work was popular.
He’s been building this thing publicly on Twitter for a few weeks — it’s called interior.ai, interior design. You upload a photo of any space, say what type of room it is, what style you want — ski chalet, tropical, minimalist, maximalist — say how many versions you want, click “render my idea,” and it just creates what an interior designer would create. Super realistic.
Sam: He took Kim Kardashian’s living room, uploaded it, said “give me some inspiration,” and it turned it into like a hot spring. A completely different look.
Shaan: And it’s like — you want a new one? Push a button. Get another one. Push a button. Get another one.
Sam: Which is what you were saying earlier — this is so much back-and-forth with a human designer. You’d sit down for design meetings, they’d create a lookbook, you’d give feedback, they’d give you a rendering, you’d say no and go back and forth. Here you just push buttons. And the creative process is totally different — it cuts all the friction out. Even somebody like me who doesn’t have design taste — this gives you design taste. But it also takes all the impatience out.
Look at this one — this is clearly an old factory someone’s trying to turn into an apartment. Industrial style, 1,500-to-2,000 square foot room, entirely empty. He says, “I found an original empty loft photo, wanted to add industrial-style decorating, wanted to auto-detect different parts around the building.” And then the next picture — it turns it into this modern loft. Replaces the floor, does this, does that. Multiple versions.
Shaan: Some of it’s outlandish. Like, replace the roof with bamboo and you can see through the roof — okay, that doesn’t make sense. But there’s this great blog post by Paul Buchheit who created Gmail, called “If You’re Great You Don’t Have to Be Good.” It’s kind of the motto of my life. The iPad — people were like, “Oh my God, it doesn’t have a keyboard, doesn’t have a USB port.” He said: yeah, but watch this. You tap a button and it instantly turns on. No boot process. You swipe and you’re on the internet. He said, “When you’re great you don’t have to be good.” You just need to be great in three things that matter, and everything else can be sucky and it’ll still win. He was absolutely right about the iPad.
Jasper AI and the Text Generation Opportunity [00:29:00]
Shaan: So let me give you another thing that I think is amazing. I just invested in a company called Jasper. Jasper AI.
Sam: The valuation was steep — I’m shocked you did it.
Shaan: The revenue curve is also steep. It’s working really well. I went in and told them, “Your numbers are insane — this might be fraud, but that’s okay, we’ll see.” I don’t think it’s fraud, but the numbers are so impressive you’re like, is this believable? And I said, “Sometimes companies that get off to really hot starts don’t sustain.” I mentioned Hopin as an example — it was the perfect thing at the perfect time during COVID, took off like a rocket, became worth $5 billion, then they laid off a thousand people because COVID ended and demand went down. They had forecasted they’d always keep growing like a rocket, and that didn’t happen.
Sam: And by the way, the founder took $1 million in secondary. Just had to secure the bag for his family. He don’t have kids, but good for him.
Shaan: I’m going to use that excuse all the time. “I got kids to feed.” I am such a good uncle.
So check out this demo of what Jasper can do. I teach a writing course called Power Writing, and the whole goal is: if you’re going to write something — on your website, an email, a Twitter bio, blog content — it’s either going to go nowhere, or it’s going to actually achieve the thing. Jasper is like a genie. You say what you want, you give it a little context — make it funny, I want an ad for my gym called Crunch, we have a welcome offer for $20, make it funny, I want it as a Facebook ad and also as a cold email — and it just generates 10 options. Then you can edit them, or literally highlight something and say “rephrase this intro,” and it gives you a new one. You’re like “give me another one” — it gives you another one. It’s like the best employee, super prompt, immediately delivers what you want.
Sam: Watch this — it’s the boss mode version. So imagine it’s kind of like Google Docs if you had an assistant in there and you’re giving instructions on how to do stuff. Say you’re a real estate agent with a lot of outbound sales emails to send. You give Jasper context — background info, tone of voice, keywords, an example of good behavior, and some merge tags to fill in. The agent is Dave, client is James, market is Austin Texas, it’s the New Year season, demand is up, inventory is down, you want to hop on a call. Then you activate voice mode: “Hey Jasper, write an email from Dave to James about an update on the Austin Texas real estate market and ask for a call later this week.” And it just… writes it. It’s perfect. You could ship it.
Shaan: That’s why these guys are taking off. Where you had employees, you no longer need employees — or if you have good employees, they can now be ten times more productive.
Sam: So yeah, I’m using it. For our Power Writing course I want to teach people how to use this. With our e-commerce thing I want to try it. Even with Milk Road — there are cases where writers can be more productive with this tool. At Milk Road every day we make an opener, like, “Hey, what’s up, this is the Milk Road, the crypto newsletter that brings you news so fresh it’ll smell like laundry straight out of the dryer.” I just made that up, but that’s the kind of thing. You give it three examples of those and tell Jasper to generate 500 clever openers — which is great because that’s one thing our writers have to come up with every day and it takes a little mental energy.
Shaan: How are they different from Copy.ai? Because I invested in Copy.ai and they do kind of the same thing.
Sam: Just imagine Copy.ai with more revenue. No offense to those guys — they’re cool. It’s kind of like LeBron James — what’s the difference between me and LeBron? Just imagine me taller, faster, stronger, more handsome, more successful, and richer.
Shaan: I’ve been an investor in Copy.ai, I love those guys. Paul is the founder. They’re at $10 million in revenue. They’re actually quite similar, but I think geared toward a different user.
AI Product Photography [00:38:00]
Shaan: Here’s another example of things I think are amazing. For my e-commerce brand, we spend maybe $5,000 to $10,000 a month on photography. You have to take product photos, model photos — so you do castings, book models, they come to your studio, use your products, you take photos or videos, use that for your website and ads. That whole process takes multiple people — a photographer, a casting person, an editor. It takes weeks. Things don’t always turn out how you wanted.
Check out DreamBooth. There are two here — Osmosis AI and DreamBooth. This guy — his Twitter handle is Strange Native — I think he works in the futures division at Shopify. He shows: AI can unlock unlimited product photography. You take a generic image — the left image is just a shoe on grass, not particularly good, kind of a glare, looks like you just went out to your backyard — then you say “make this shoe look epic,” and it takes the shoe, cuts out the background, puts it on this lightning background. But look at the second example — that’s better. The training image is a dude taking a selfie in a hoodie. Looks bad. You can’t use this on your website for e-commerce. Then it generates a model-looking studio photo of the same hoodie. It gave this dude pecs, better shoulders, a jawline, fixed the lighting, put him on a background image. What the hell?
Sam: If I could just take me wearing my product — don’t care about the lighting, don’t care about the background, don’t care if I have my hair done — and it would just generate a studio-quality photo? Dude, how do I use this?
Shaan: You click the tweet for now. I don’t think you can use this yet — I think it’s proof of concept, not a real product product yet. A lot of these are doing demos and training models just to see what happens. Some of them open-source the code.
There’s also Osmosis — Mickey Friedman is her name — and what she’s doing is you give it an image and say “turn this into an ad,” and it turns it into Facebook ad creative for you, making it look cooler, turning it into a video, that sort of thing.
Sam: Now, I haven’t tested these for real to see if they’re actually good. But the concepts are good, and whether these exact products are the ones that do it, someone’s going to do it — because these are really valuable business use cases. If a process that’s costing me $10,000 a month now costs $29.99 a month and is instantaneous and better, you’re ten times faster, ten times cheaper, and ten times better. That’s the makings of billion-dollar companies. This is just 100% frame-breaking when you see this stuff. And there are still hundreds of thousands and millions of people who know and care about this — but comparatively to how much of an impact it’s going to have, nobody knows about it.
Sam and Shaan’s Position: Investing, Not Building [00:45:00]
Sam: So where are we in this? I am not capable — I don’t have the ability to work on this. I simply don’t have the horsepower. The engine I’ve got upstairs is more of a go-kart.
Shaan: I’m very much either going to be a spectator or only semi-active in this game. You and I were joking — we’re obsessed with TikTok, we watch it constantly. You made a joke like, we’re content creators in the same way a horse-carriage person in 1912 just saw Henry Ford drive by. You’re like, “What the hell was that?” When I watch TikToks, I see an 18-year-old with Vans who makes this spectacular video that’s just the funniest thing ever, and I’m like — there’s a new era and I’m not part of it. That’s a little bit how I feel when I see this.
Sam: How do you intend to get your hands on it? We’re a bunch of scheming, greedy people — how are you going to get your paws in this game?
Shaan: That’s a great question, one I’ve been asking myself. I’m a little different than you in that I have more self-delusion. I’m like, I could do this. If I hired the right people, I could be the vision guy. I’ll have the idea and all these geniuses will make it. You previously had a company that I would tell my parents was a tech company but was really just an email newsletter — you actually had a proper technology business.
Sam: Right — I was basically writing brochures, whereas you were running a Silicon Valley enterprise. I’m like a restaurant with an online ordering menu calling themselves a tech company.
Shaan: My plan is: invest in everything, because investing is easy and great. I’m a believer in this wave. A lot of the people who can build this stuff don’t know where to apply it — what’s the actual pain point they should be solving — and they don’t know how to build a defensible business or go-to-market strategy that makes sense. So I’m like, I’m going to invest in a bunch of these companies. That’s plan A, already in motion.
Would you say this is the most interesting sector right now?
Sam: For sure. I kind of feel like an idiot, like, oh cool, now I’m interested in the new thing. On one hand — the more crypto prices went up, the more I invested, and then I created the Milk Road and turned my content attention to it. So now oh, you know, AI is that new thing. But what am I supposed to do? I just saw a flying object. I just saw a UFO. What am I supposed to do — pretend I’m not interested? No. Count me in. Beam me up.
Shaan: It is exciting. It’s an unlock — like we got new toys to play with, I can’t leave them in the box. I got to unbox it and see what I can do with it.
Three Business Ideas in AI [00:51:00]
Shaan: Let me give you three ideas I think somebody could build that aren’t in the demos we just talked about.
Sam: All right.
Shaan: Idea one: the AI therapist slash AI companion. There are a lot of people who would benefit from therapy, and there are apps that connect you with a therapist on your phone — Talkspace and things like that. These companies are doing well, like $100 million-plus a year, but it’s expensive. What they were doing was connecting you with a real licensed therapist — they take an in-person visit that might be $150 or $200 and turn it into a $60 or $90 a month subscription with three visits or something.
But there are tens of millions, if not hundreds of millions, of people who would benefit from having someone to talk to — a therapist, a companion, somebody to confide in, somebody who won’t judge them, who’ll be positive and helpful, who’ll offer good advice, who’ll never share their secrets, and who has no self-interest in the matter. You could now provide that for maybe a dollar or something. The cost and accessibility of helping people talk through what’s on their mind is going to drop dramatically. AI therapy slash AI companion and friend — I think that is a tens of billions of dollars idea. I think maybe there’s a lot of competition and not clear network effects, but it’s a very, very big idea.
Sam: That’s pretty plain and obvious. It’s a pretty obvious, straightforward solution — obviously challenging to pull off, but straightforward. We could all predict that. Yes, I agree.
Shaan: The next one is a little sexier. I call it AI Spotify.
Traditionally the music business has been like — what’s a situation where people run into a burning building making terrible decisions? That’s basically what the music business was. You’re competing with everybody who loves music. Even if you break out of the thousands of people trying it, here’s your prize: a lawsuit. You get sued out of existence, whether you’re Napster or one of the many companies in between. Okay, you survived the lawsuit — you’re Pandora, you’re Spotify. You built the best-in-class product, got everybody to use it. After 15 years, Spotify does not have impressive economics. They don’t spin off cash the way Google or Facebook do. Why? Because to deliver music you have to pay the record labels — they own the pricing power.
Now, why is AI Spotify different? Basically you can create a really cool music app that doesn’t have to pay musicians a dollar. Here’s what the service looks like: the same way interior.ai works — you want an industrial look, you want a minimal look — what you’re going to be able to do with AI Spotify is pick a song that you really like, like Pandora tried to do, but these songs don’t exist. The AI is just going to generate a new song on the fly for you. Like TikTok’s algorithm, it’s going to learn when you skip, what songs get liked, if you like this song will you like that song. The algorithm generates music on the fly for you.
Then you could add crypto to it — by default all the songs are ephemeral, here today gone tomorrow. But if you like the song and want to keep it, you don’t just save it — you own it. You click “mint” and mint an NFT. Maybe it splits royalties with all the artists that inspired that music, or the algorithm owns 50% and the user owns 50%.
Sam: Did you just come up with that, or is someone working on this?
Shaan: Off the dome, baby. Got that big brain. No small boy stuff.
Sam: This is wonderful. I think this is finally — you web3 nerds are finally coming up with interesting stuff. We did it. Nobody believed in us for ten years and a trillion dollars, and we got an idea that makes sense.
Shaan: I do think this idea is great. What happened was: with GPT-3, OpenAI basically downloaded all the text of the internet to create a text generation engine. With DALL-E they did the same thing with images. The speech guys downloaded all of Joe Rogan’s back catalog to generate his voice. Somebody is about to do that for music. They’re about to download all the music off Spotify, train an engine on it, and release an open-source model that lets you generate new music. That’s what’s going to happen.
Sam: Brilliant. Finally, brilliant.
Shaan: “Finally brilliant” is the best backhanded compliment ever.
Sam: I’m like Beavis and Butt-Head making a few bucks. You might have artificial intelligence, but we got that real stupidity over here.
Shaan: All right, what’s the third one?
Sam: I lost my list — I misclicked. But okay.
Carter AI, NPC Characters, and AI Friends [00:59:00]
Shaan: So I just talked to these young guys yesterday. I wanted to invest in them but didn’t love their idea. One of the founders had started a company out of college, raised a little money, didn’t work out, failed. Then he was joining different communities and met the other founder through TikTok.
Sam: He met his co-founder through TikTok? Kids meet through TikTok nowadays?
Shaan: Yeah. The other guy, Hugh, has been building what he calls Jarvis — like from Iron Man. He’s been building it publicly on TikTok for a year, trying to create a real-life AI assistant. He’s got about a million followers on TikTok just from doing that in public. They decided to create an AI company together.
The first thing they made is called Carter. In games, you’ve got NPCs — non-playable characters. They’re just there to fill the game so it’s not empty, but they don’t do much. You can punch them or talk to them, they say the same three words. What these guys said was: hey, any game developer who wants their in-game characters to actually be able to talk — just plug in this line of code, and all your NPCs will be able to hold conversations with players. The player asks for something and the character guides them: “If it’s the sword you seek, you should check behind the waterfall.” The character guides you automatically, you don’t have to hardcode the response.
I don’t love this use case, but it is a cool idea — in the future, stock characters in games are going to be things you can actually interact with, which changes how games work.
This guy Hugh retweeted something from Alex Wang, who’s the CEO of Scale AI — a $10 billion software company that helps label data for machine learning. He says: “We’re at a critical turning point for humanity. Children born today are likely to have more AI friends than human friends. AI friends are going to be more reliable, considerate, and agreeable. What does this mean for childhood development and social norms? We will find out.” And then: “Loneliness is an epidemic, it’s a real public health problem. This technology has a lot of potential for good.”
Sam: I completely agree with him.
Billy of the Week: Emad Mostaque and Stable Diffusion [01:03:00]
Shaan: Last thing I want to leave everybody with is my Billy of the Week. The guy I think should be the Billy of the Week is Emad Mostaque.
Sam: A billion dollars isn’t cool. You know what’s cool? A billion dollars.
Shaan: Emad is the guy behind Stable Diffusion. Do you know what Stable Diffusion is?
Sam: I know that the tweet you referenced earlier was talking all about it.
Shaan: Stable Diffusion is basically an open-source competitor to DALL-E. OpenAI created DALL-E — everybody was excited about it, but you had to get permission to even use it. We were asking, “Who can get us access to DALL-E? Oh, you have to pay for all these credits.” Stable Diffusion is like — here’s their tagline: “AI by the people, for the people — designing and implementing solutions using collective intelligence and augmented technology.” They’re the FUBU of the AI world.
Sam: Anybody who does the “by us, for us” thing is generally full of shit. But this guy — I watched one interview with him and I can confidently say from the bottom of my heart: after one interview, I think this guy’s the real deal.
Shaan: So let me tell you about this guy. He put in $600,000 to fund the development of Stable Diffusion. The company is worth over a billion dollars already. What was he wealthy from? Born in Jordan, British accent for some reason, math background, quant type. You can tell in the first few minutes of talking to him — this guy could say more than five prime numbers. He works at a hedge fund as a quant, doing trading. He’s like, “I was winning that game.”
Then he has a life change — his first kid is diagnosed with autism. He takes a few years off, retires early after winning the hedge fund game. He goes to the doctor and asks, “What do I do about this?” The doctor says, “There’s no solution.” He’s like, “I’m an engineer. I have a problem, which means there must be a solution.” He doesn’t accept that response.
So he starts doing his own research. He realizes all this literature — research papers — would take forever to read one by one. So he creates an engine that ingests all the papers, does a semantic search, and uses machine learning to figure out what all the research on autism is actually telling him. He basically finds that autism isn’t one thing — it’s a phenotype, a behavioral description, with many underlying causes. Different body chemistry imbalances. He says doctors are correct that you can’t just fix it or cure it, because it’s so multimodal. But on an individual level, if you can understand where that one individual’s body chemistry is slightly different, perhaps there’s a rebalancing you could do. Over a couple years he creates a process to help his son. He says in the interview his son is doing so much better. He didn’t claim he cured autism, but he learned deeply about these biological systems and where AI could help in the future.
Then in 2020, he’s at a dinner in Davos — as you do when you’re a person with a British accent and a huge vocabulary — and people are talking about this virus coming out of China. He looks into it, and the way they’re describing COVID-19 seems very similar to autism to him. He calls it a multi-system inflammatory disease. His main thing was: developing a vaccine for this was going to take some time, and even if we did, it wouldn’t work for everybody, because this is a multimodal system. He starts an initiative using data to help policymakers do a better job with COVID. He ends up dealing with the WHO and UNESCO and World Bank, it gets too bureaucratic, and the whole initiative collapses.
Then he comes up with what he calls his Promethean mission — Prometheus, the guy who gives fire to humans. His mission: open-source all the powerful AI and ML tools that exist, so they’re not controlled by large private corporations. OpenAI is called “OpenAI” but it’s a private organization, close to the vest about what they’re working on, and when they release things it’s not to everybody. If you look at the charts, Stable Diffusion has far surpassed OpenAI. The tweet that went viral was: “Open AI — meaning the actual open-sourced AI — is beating OpenAI.” The people who took an open approach are crushing the private organization literally called OpenAI.
Sam: How do they make money?
Shaan: I think it’s the same kind of thing — you pay for some credits or usage. Like Linux and Red Hat — open-source companies have these weird models where the model is free but services on top make money.
He starts Stable Diffusion because his daughter asked him if he could do the same thing he did for COVID for art. Oh, and in between — he wins a grant, an X Prize, like a $15 million prize for who could create a system that teaches kids in foreign countries mathematics for less than, like, $75 total cost. They created something on an iPad where kids could learn and pass certain tests within 75 days. He wins this $15 million thing. Again, doing good in the world.
He also has synesthesia — actually I think a different one, maybe aphantasia — anyway, he has one of those things where senses are linked in unusual ways.
So he gets together 20 of the best engineers, Manhattan Project style, self-funds the thing with $600,000, and they create Stable Diffusion. Now valued over a billion dollars.
Sam: Isn’t it crazy just how bold some people are? When I talked to Ryan Holiday for the pod the other day, I felt inspired because I was like, your type of success is awesome and inspiring to me because I too can work hard and achieve what you’ve achieved. Then I hear a story like this guy and I’m like — he just slapped me in the face. I am nothing and I am nobody and this guy is going to steal my lunch money. That idea just bullied me.
Shaan: He’s not going to steal your lunch money — he’s going to give you lunch money. You’ll be like, “Who are you?” and he’ll just be like, “Don’t even worry about it, son.”
Sam: Yeah, I feel inadequate. The Billy of the Week section is really the “Inadequacy of the Week.” Ladies and gentlemen, are you ready to feel like shit? I’ve got a story for you — this guy who’s younger, better, smarter, richer, did it faster, was doing the same thing you said you wanted to do and he actually did it. And he’s also ripped.
Shaan: There is a world where he’s full of shit. I didn’t know of this guy’s existence till like three weeks ago. But I watched the interview and my feeling was — this guy has the same energy as the first time I heard Vitalik talk. Or Zuckerberg. People who are extremely mission-driven while also being slightly on the spectrum and extremely impressive in their past accomplishments. Very matter-of-fact about the way they think the world is going. Not trying to hype you up, not trying to sell you. You ask them a question, they give you their answer, and then your frame is broken. The tone of their voice is like, “I said what I said. This is what I believe and this is what I’m doing with my time. Yes, this is what I think is important.” And you just look at yourself — you’re speaking a different language, the language of being silly old you, trying to understand what the hell they’re talking about.
Wrap-Up [01:16:00]
Sam: Dude, this is one of those episodes where I record it and I have to take a nap at the end from excitement. Like when you feed my dog too many treats — he gets so pumped up he has to take a nap just from the excitement. That’s how I feel. I need a dad chair. I need to be reclined at a 34-degree angle with my feet above my hips.
Shaan: At the end of this episode I’m going to get up and make a grunting noise — not from pain, just out of habit.
Sam: This is just so fascinating. I think this episode has had a lot of mind-blowing things. I’m eager to see what people say. I feel amped. This is almost a “quit and dedicate your life to this topic” type of thing.
Shaan: Almost. It’s like when someone has worked at your company forever and it’s their last day, and they’re all the way across the room, and you’ve worked with them for ten years and you know their wife’s name and their kids’ names, and you’re so excited to give them a hug and tell them how much they mean to you — but the couch is really comfortable and they’re just all the way across the room. Like, it’s going to be at least 50 steps.
Sam: AI is all the way across the room.
Shaan: That’s so good. All right — we’re ending it on that. We’re out of here.