This Week in Startups - The rise of AutoGPTs and AI anxieties with Sunny Madra and Vinny Lingham | E1720

Episode Date: April 13, 2023

Vinny and Sunny join Jason to discuss AI’s blistering pace and compare its development to other product launches (2:47). They also break down the anxiety artists and developers feel from AI automation (14:47), the rise of AutoGPTs, Twitter’s reported LLM project, and more (37:01). (0:00) Jason kicks off the show (2:47) The blistering pace of AI (9:44) Developer builds Flappy Bird in 1-hour (11:24) Squarespace - Use offer code TWIST to save 10% off your first purchase of a website or domain at https://Squarespace.com/TWIST (12:53) Developers leveraging gains in AI (14:47) Ai anxiety (23:55) Instacart’s ChatGPT plugin  (26:06) Preserving your advantage against AI (29:10) Vanta - Get $1000 off your SOC 2 at https://vanta.com/twist (30:14) AI artists (35:54) Crowdbotics - Get a free scoping session for your next big app idea at http://crowdbotics.com/twist (37:01) Automating with AutoGPT (47:09) LLMs vs. Knowledge retrieval systems  (57:06) Is Google’s Bard behind?  (1:02:42) Twitter’s alleged generative AI project (1:08:30) Ethereum and Bitcoin updates FOLLOW Sunny: https://twitter.com/sundeep FOLLOW Vinny: https://twitter.com/VinnyLingham FOLLOW Jason: https://linktr.ee/calacanis Subscribe to our YouTube to watch all full episodes: https://www.youtube.com/channel/UCkkhmBWfS7pILYIk0izkc3A?sub_confirmation=1 FOUNDERS! Subscribe to the Founder University podcast: https://podcasts.apple.com/au/podcast/founder-university/id1648407190

All right, everybody, we have another great show for today. The Crypto Roundtable has been redubbed the AI and Crypto Roundtable. Why? Because 25% of people working in crypto have shifted their energy to AI. We're going to talk about why that is. And we're going to talk about how generative AI has had the fastest development cycle we've ever seen in the technology industry, even faster than apps on iPhones and Android, and how this speed is creating massive anxiety with developers and artists, because of how quickly,

whatever they build, gets replaced and automated by the AI. We're also going to talk about something incredibly important. The emergence of auto GPTs. What is an auto GPD? Well, it's an agent. It's something you set up for a language model to work on, but then you automate it so it never stops working on it, and it gets better and better at the test. So we're going to talk about how, this is going to change everything in AI. It's not just how you having a chat. It's chats occurring in the clouds with an automated agent and the results coming back and then triggering other AIs to do things for you. This is kind of singularity-esque in a way. Then we're going to talk about what the winning business models will be in general of AI and where the opportunities are for you as a

founder or a capital allocator placing bets. Finally, since this used to be the crypto roundtable with Vinnie Lingham and Sande Maudra. We're going to touch on some crypto, including what's going on with Bitcoin? Why did it double in value? Is it because of Bologi's million dollar bet? Or is it because the stock market's going up? Or maybe we're not in a recession? Or people are concerned we're going to go into a depression?

Why is it that it's doubled? And then what will Bitcoin be trading at by the end of the year? Vinny has a prediction and he's 75% confident of this prediction. So I want you to listen to hear his prediction. it's going to be a great show. Please. The love of God, stick with us. This weekend startups is brought to you by Squarespace.

Turn your idea into a new website. Go to Squarespace.com slash Twist for a free trial. When you're ready to launch, use offer code Twist to save 10% off your first purchase of a website or domain. Vanta. Compliance and security shouldn't be a deal breaker for startups to win new business. Vanta makes it easy for companies to get a SOC2 report fast. Twist listeners can get $1,000 off for a limited time at vanta.com slash twist and crowdbotics. Great ideas can change the world and crowdbotics is the fastest way to turn those

ideas into code. Get a free scoping session for your next big app idea at crowdbotics.com slash twist. All right, everybody. Welcome back to this week in startups. It's time for our crypto slash A.I. Roundtable with two of the smartest technologists in our industry with us again. Sunny Sandeep Madra, my good friend, and another great friend, looking really dope in his amber sun glasses, reading glasses. Vinnie Lingam, what's going on with those glasses, Vinnie? So I've changed the lighting in my office, yeah, and I have this bright light shining. You can probably see it there in my eyes. But I look better on camera and it makes things look good.

Do you? Well, it's very grainy otherwise, right? So the lighting is pretty bad. And then I had this like weird light coming through the middle, yeah. Oh, I see. So I turn the lights in the office off and look, this looks better now. But those are those glasses that take out like a certain type of light from your screen at night. Blue light glasses, yeah.

Oh, you got blue blockers. Yeah. Remember blue blockers from the infomercials? Those commercials, the infomercials. Oh, those still wears them. I mean Bono still wears them right

So you are the Cryptobano We'll call you Cryptobano I need to register that name Yeah Cryptobano All right

Listen Everybody knows Sunny is the co-founder of definitive Intelligence Who had a really special meeting this week

I saw it on the group chat A very special meeting We wouldn't say with who That was a very Very impressive meeting you landed Huh It was

It was cool It was really cool It was really cool. It was a legend. You met with a legend. And I saw the image. A legend, too. I won't say who it is, but who has been on this very program.

All right. Let's keep moving here. You have to guess from the 1,700 people who've been on this podcast in 12 years. Vidi, of course, is the co-founder of Civic, a startup that crips identity information on the blockchain. And he started the amazing waitroom. Go to waitroom. com. And, yeah, great for enterprise folks who are looking to do awesome.

meetings in the new remote world. So we got the plugs out of the way. You know, I have been watching this nonstop. I can leave now, right? You got the plug, that's it. You are, you are, now we have, now you have to earn the plug by giving us your knowledge. I saw auto GPT go by my Twitter feed the last 48 hours and some other studies. There was a great Stanford study that I talked about yesterday on this weekend startups.

essentially the pace, let's start with that, the pace of AI right now, can you remember a time in your career, Sunny, when you saw a technology moving this quickly, but also with user adoption this quick. So take a moment to think about that X, Y, axis, users adopting it and playing with it, and the technology advancing at this fast of a clip. What's the last time we saw something like this?

I have a couple of ideas, but for you, what do you think? It's the fastest, J. Kel. Hands out. Being a technologist, building. You know, we are not at the singularity yet, but we are, you know, approaching it quickly. What is starting to happen to increase the pace is, so we've all had this notion about a 10x engineer, right? Jekal, you're familiar with this, right? Vinnie, Vinnie himself is a tech.

it's, it's an, you know, I'm going to rewind back a little bit just to give some context here. So the notion in, in particular in software engineering is that it's one of the very few practices in terms of human productivity where one person can be 10 times more productive. If I create some, then the person beside them. So if you think about in the context of a factory, if you and I are working in a factory, J. Cal, like, I can't probably make 10 times more widgets than you. It's like limits of like, you know, speed of my hands and the supply chain and everything else. But in software, for engineering, it was like sort of the first time we were seen in a career, in a practice, in an area, that one person can be 10 times more productive than the person beside them. And that could be through their ingenuity of how they write code or the tools they use. And so this has been something that we all search for in our co-founders or, you know, engineers that join our businesses is to find these 10x engineers. Would that be a fair definition, Jacob? I would say that's extremely accurate.

and there are laws of physics. So if we both worked in a Ford factory putting tires and lug nuts onto a car, you know, there's a physics thing here. It doesn't matter how strong, fast, nimble you are. You might be 10% or 20% faster, but you wouldn't be a thousand percent faster, aka 10X. 10x, yes, exactly. But in software, if I am, you know,

I have more ingenuity in how I design an algorithm or how I leverage a set of tools, I can be. So my productivity could be 10 times more code because I'm using tools. My particular algorithm could run 10 times faster than yours because I've written it in a different, more unique way. And so we can see that, and we've seen that over the years. Now what's starting to happen is engineers are starting to use AI tools. They could be co-pilots, they could be open AI.

and OpenAIs powering out of these co-pililots as well. And this is allowing for more engineers to achieve the 10x improvement. And my belief is, and I'm seeing this hands-on as well, is that the teams that are leveraging these tools, whether it's GitHub copilot, whether it's OpenAI for debugging, whether it's a whole new set of tools in and around, you use a Notion AI plugin,

those teams are starting to achieve a 10-1. X across their entire engineering base. And this is why we're starting to see a huge acceleration, especially from AI companies, because they're natively building AI things and they're using AI themselves to make themselves faster. And this is something that is happening to many more people. So the average engineer, the average software engineer is now, maybe they don't hit 10x, but they hit five or six or seven because of autocomplete.

So if you just want to imagine this as a civilian, when you're using iMessage autocomplete where it just guesses the next word, or g-mails, which guesses two or three words, now all of a sudden, you know, imagine that for a developer where it does two or three lines of code ahead or a paragraph of code. That's what developers are experienced on a day-to-day basis, yes?

Yeah, I want to share something. and it's a really cool example. There was a game a couple years ago, Nick, if you can pull this up, I was called Flappy Bird. And this game was interesting because the founder of, you know,

the creator pulled Flappy Bird from the app store because he said it was too addictive. This gentleman, in this medium article, we can share the link to everyone, recreated Flappy Bird, including the graphics,

including all the code in one hour. Okay, so that was a game. What do you think? think the game originally took to code? Probably a week. Let's say a week to, you know, a week, between a week and a month, depending on... 100 hours, whatever.

Hundreds of hours. Something along. Hundreds of hours. Yeah. Let's give it that. Well, this is a totally, yeah, game changer. And they were able to just describe this to the AI and have it build the code, et cetera.

Exactly. They described the graphics they wanted and they described the, you know, controls that they wanted to implement. And, you know, they leverage, you know, the unity underneath it. And that example and that link there, you know, you can follow it yourself, J. Cala, and I bet you've probably never written a video

game from start to scratch. No, of course. I haven't written code in 30 years, yeah. Yeah, but if you followed those instructions, you too could have your own version of Flappy Bird in one hour. It's incredible. And this is, this is, and so imagine,

you know, this is someone just using external set of tools. Imagine you're a native AI company and you're using the tools for your, you know, to accelerate yourself. And so that's why the pace is starting to move incredibly quickly. Okay, we're about to announce the winner of our second show-as-your-space competition. Elise runs a digital sports nutrition business built on Squarespace called, Whoa, Sports Nutrition.

You can go check it out at W-O-H-E Sportsnutrition.com. W-O-H-E sportsnutrition.com. Congrats, Elise. You're going to enjoy $1,000 in Squarespace credits. And if you want to be an amazing entrepreneur like Elise or start your side project or anything in between Squarespace is how you do it from personal projects to giant tech startups, anything can now be built on Squarespace. We love it so much.

We use it for remote demo day and many other projects. Here are some amazing Squarespace features that founders love. You're going to get unlimited, e-commerce, templates, analytics, inventory management, all of that stuff built in, SEO, 24-hour, seven-day-a-week customer support. everything just works inside of Squarespace and of course it's optimized for mobile you know about the gorgeous templates but you can also sell content there courses etc appointments and save the 15% tax that other platforms are taking from you that's your money don't give it to a platform use Squarespace instead Squarespace.com slash twist for a free trial and when you're ready to

launch use the offer code twist to save 10% off your first purchase of a website or domain we love Squarespace, and they hold the belt here at this weekend startups for our longest running partner. Thank you so much for supporting this weekend startups and our mission to support founders and inspire innovation. So what's happened is, Vinnie, the people who are working on AI are the first to leverage the gains of AI, which then leverage the gains of AI, which then makes AI advance even faster. So in a way, it's like we took, we took. We took. humans working in the factory, building robots, or robotic suits, like in this movie, Um, Edge of Tomorrow, which I watched again.

It really stands up, by the way. I watched it with my seven year olds. They loved it. So imagine you're like building edge of tomorrow robot suits. You put the suit on and then you start building suits in a suit. And that is in some ways like a singularity type event, which is the first people to experience the gains here are going to be developers. and they're going to put that, of course,

into developing AI faster. Vinny, your thoughts on this pace, can you think of another time technology? If this is the fastest, do you agree, it's the fastest? I agree, it's the fastest. I think it's also good. So what's the second fastest then? What would be, if we were to think about another moment in time?

Crypto was 2017 was crazy time. It was a lot moving fast. I would put mobile apps in front of that, Vinny. That's what I was thinking. I was thinking the app store. Mobile had a lot of moving fast. lot more of a longer run, right? So if you look at like instantaneously, when Bitcoin went from 2000 or 20,000 over like three or four months, that was just crazy, right? Like it was a massive

change. And mobile just took years to play out or even, you know, it wasn't in the sort of smaller time box. I will say the one thing we should we should definitely touch on and maybe speak about today is the psychological impact of this on everyone in the industry. Because it's not a simple, oh, everything's moving. People are getting anxiety from how fast things are moving. People can't catch up. People can't keep up. I have engineers.

I'm trying to, you know, it's like every day it's something new. Like, they cannot get their heads around what they're doing in the sense that like, you know, what you do today could be invalidated tomorrow because now there's a better method to do it. And all that work you've been doing for the past month or so to do is, it's just been open source by it. And all the AI is created. Like, everything's moving fast. This applies to AI artists, people who are doing, well, you know,

coders, engineers, people using the products. This stuff is moving so quickly in so many different directions. My fear is that the whole space gets highly fragmented and we get to the same point we're in with crypto. Crypto's biggest issue right now is there's no consolidation around a single or two or three chains. There's dozens of chains that actually have decent tech. The entire ecosystem is hyper fragmented.

And that's the problem with crypto. That's the one, you know, one of the things that makes Bitcoin so powerful is that it's 40, 50% of the entire crypto market and the critical mass that it's got there makes it a juggernaut and everyone else is kind of playing for second place. You know, or you can argue Ethereum is there and say people are playing for third place. But the critical mass that you have in Bitcoin, it gives it that sort of power. The problem with AI right now is that the lead that Open AI has can be eaten away at by all the other lawmakers that are coming out and everything else.

So it may be it may be the Bitcoin of this sort of space for a while. I don't know. I'm just part of the thing. I take it kind of, I kind of take it slightly differently. I agree with you with the anxiety. But like, I try to calm the anxiety with the following analogy.

When we first invented computers, we had to literally give computers zeros and ones, right? So we had punch cards, and those punch cards represented the zeros and ones and how the computers operated. And Jay College, back in your day, you remember this, right? It's interesting to bring up punch cards.

I remember when I went in 1988, when I went to Fordham University in the fall, they did have Vax machines and the internet was running on mini computers as well as desktop computers. I had just started to have some emulation software to do it on desktops. But it was mainly like you were on terminals, Vax terminals, these kind of things. So you know, there's a punch card machine and there were punch cards in the laboratory at the computer center. Yeah, still there. Yeah, 100%. We went from punch cards. Then people said, oh, you know what?

We got a, we got to, you know, too hard, right? You can't walk in with stacks and stacks of these things. Then we move to assembly, right? And so we, oh, we have this language that, you know, there's a compiler that that'll turn into the zeros and ones. Then we go higher to basic, right? We get the basic programming language. And then we kind of work all up there. Then we get C, C++, then we get all the exotic languages, right?

Now, what is the programming language? English. English. No, it's just English. Propting in English. And so we've just come all the way from zeros and ones, all the way to English. Well, this is 5GL. AI is the fifth generation language.

I mean, we went from first, second, third, fourth, AI is the 5GL. And if you can get the team comfortable that they can use English to accelerate themselves and get comfortable with that, then I think that anxiety. comes down because you switch your mindset to saying, hey, I don't have to write code for everything. I can just explain this in English and leverage these tools to get me there quicker. That's not what I'm talking about, Sonny. I think the anxiety comes in like creators, people building stuff are seeing replicants of what they're doing very quickly. So it's something which takes someone a month or two months or six months to build, let's a call a three months to build.

It's the same way Flappy Bird was built in an hour, the pace of acceleration is giving creators this anxiety that if they don't build it right now, quickly put it out there, someone's going to beat them to it. And so the pace of development, whether you're creating art or, or project or code base or whatever, is so quick that people are now having,

you know, anxiety in the sense that like, what's depressing in a way? It's depressing. Yeah. It's exactly, yes. I am seeing this firsthand because at inside.com, we have a vibrant newsletter business,

makes millions of dollars a year. and the format I came up with, even before Axios existed, was read long stories and reduce them down to bullet points and rank the bullet points in order so that you get this quick summary. And I implored everybody, go buy Chat Chaptap4 and start playing with. I told all the riders, you're talking about like 40, 50 writers. And I said, you know, it's 20 bucks, so it's a thousand bucks for the whole company, just go play with it. and some of them had a really interesting experience when they took an article from Financial Times, Wall Street Journal, whatever, some paywall article, dumped it in, say, give me the most important bullet points, and it was 70 or 80% of what they would do. And I said, don't feel bad about it. Do it, check the facts, rewrite it, polish it, and then come up with insights. So take the same amount of time, three, four hours to write a seven item newsletter, and then, you know, just add the human touch to it. And that kind of freed them. But there was this existential dread moment,

which I think a lot of people are going to go through. And developers, let's face it, they felt very special the last, I don't know, since existence. It felt like once developers starting in the 80s when the PC came out, they went from being, you know, like low level, you know, people in a laboratory somewhere who were underappreciated to rock stars. Bill Gates, Steve Jobs, waas, all of a sudden,

developers were rock stars. And I think that is a hard thing to think, I spent all this time learning to play piano, learning to sing, and then somebody comes out with auto tune, and they put some beat

from a library collection in, and they say, give me a drum beat that's like this and a guitar solo that's like Mark Knopfler, and all of a sudden, my 30 years perfecting my instrument has been, duplicated by a computer.

Yeah, that's jump out the window moment for some people. But that's only if you're not willing to upgrade your skills, right? Like the way I think about it is. No, I don't, I don't disagree. It's not a skills issue. Even if you have the skills, it's the marketplace we're dealing with right now is become, you know, I grew up in a small town. Maybe there was like 100,000 people in the town. I was the best chess player in my state.

Like the moment you're going into a bigger world, like when nationals and, I'm not as good as anymore, right? Like, you feel inadequate. It's the whole Malcolm Gaddle, um, outliers book, right? Like,

you know, people who, so, so, so, the problem right now is AI is such a global marketplace with hundreds of millions of people doing it.

How do you compete? Yeah, you could build something. That's, that's the anxiety. You build something for AI, Vinnie, and you spend two weeks building it.

You think you got the greatest idea ever. And then it's a feature inside of, Exactly. The next AI tool, which is the same feeling we had going back to apps. And I agree with Sunny, I thought apps would be the answer we gave after this because those did move fast. And there were people who made flashlight apps, calculator apps, weather apps, all kinds of little apps. And you were in the app business, Sunny.

And then you just watched as like Apple and Android were like, oh, yeah, flashlight. We should build that into the little control center. Oh, weather. Well, we'll build that too. Oh, stock quotes. We'll build that too. Oh, news. We'll build that too.

people used to have news apps. There was a whole swath of news apps. Inside had one, smarter news in Asia, flipboard. And then all of a sudden I was like, yeah, Apple News, boom, we're done here.

And that is depressing when you put a lot of work into building something. It is. And look, I can completely agree with you guys. I'm trying to give the counterpoint to get people out there

to keep innovating, which is what you can do in this world is actually find those places of differentiation, right? And those places of differentiation have become harder now. So it's not just creating a UI anymore, right? Because that's been replaced by, like, you know, we just saw, right? Write a few lines inside one of these AI tools and you'll do that.

I think you have to think long and hard. And so I think if we think about a couple of businesses that have maybe something very defensible in this world that's not as replicatable because of just a UI. So let's let me pose a couple of different examples and maybe this will help people think about it. Uber is not replaceable. You could connect it to OpenAI.

You can say, hey, call me a car and all that stuff. But at the end of the day, that API interacts with drivers and vehicles. You can't replace that in a few days. So you have to build something actually defensible. Elon talks about, hey, build something real. Instacart. There's a really, really good tweet that came out yesterday

where someone asks, where and you know where someone asks exactly right uh gpt and uh the the picture down below is really good if you can click on to that so he goes hey i'd like uh you help buying groceries here's a hundred bucks breakfast lunch dinner i'm obviously summarizing this um hopefully you can see it in the video it comes up with the meal plan it then subsequently uses the instacart plugin to do the order fills up his instacart with the ingredients and you're off to the races This is also defensible because it's not easy to replicate the Instacart infrastructure, right? And so, you know, that's great.

A place where there will be some challenge. The guy who just built this tool now, okay, how does he monetize it? Affiliate revenue? Well, no, no, but now you can replicate it. You can make 50 different versions of this. Which means he innovated. He's an innovation. In the world today.

It means it's something that's not, doesn't have enough value. Exactly. So his innovation is this basically up for being copied. Well, I actually think it's slightly different. I think open AI is becoming the apex aggregator, right? And what they're going to do is, like, you have to understand, like, what he did, he actually didn't build anything. That's why I'm trying to be very clear. He just put something inside a prompt and use the Instacart plugin. And so, and that's the thing you have to really understand now. And so what you probably shouldn't be doing is building like a blue ribbon or something like that, because that's going to get displaced very quickly just by people's own preference with the Apex aggregator in OpenAI,

interacting with these plugins with real API services, right? So you're saying that the Instacarts of the world need to, so basically the real world infrastructure providers, because digital is easily copied, need to basically make it so that AI can interface with them a lot more easily. And we need people to create more real-world interfaces via APIs, right? And you know what happens to... Is I think there are jobs that we romanticize as humans,

and there are jobs that we, you know, marginalize. And so if we were to look at the jobs, you know, we marginalize, like a switchboard operator or like a film projection, or, you know, a travel agent, a toll booth operator. Anyone in farming before machinery? Secretaries. Hopefully that's not triggering language. I will get me canceled. We used everything called secretaries.

They were assistants. Film developers, right? These did not feel like they were complicated artistic tests. But what we're getting on now, putting aside this little, you know, prompt engineering to, you know, have your meal plan done. You start getting into like hand-drawn animators and, you know, what happened with Pixar or music producers or just painters, you know, now like, why would you hire a painter to make a painting of yourself and wait three months and pay $10,000 for a portrait in oil paint when you could see 20 of them in two minutes.

I mean, there's no doubt that like jobs are going to be disruptive with AI. It's like unquestionable. We're going to have. We're talking about the emotional state, though. Like, why are we emotional about certain jobs? Well, it's not an interesting part. Again, let me go back to my point. My point wasn't to be emotional about the creative destruction process. Clay Christensen laid it out.

I think it's very important that these jobs, if you can be automated, it should be automated. That's not what I'm talking about. I'm talking about the people who are participating in this ecosystem right now. This is like probably the most meritocratic, I wouldn't say the only meritocratic. There's a lot of luck involved as well, what takes off one dozen adoption and stuff. But the people who are contributing to the space right now in any way, shape, or form have essentially no abilities to build moats around their business. With the exception of what Sonny mentioned, like the offline. Or barrier to defensibility is higher, Vinnie.

That's all, right? It was barrier to defensibility was you had to know how to code. And that is now the barrier defenseability is just cool. The competition pool is bigger now. So in other words, in a world where you're competing with the 100 million engineers, you're now competing with 7 billion humans who can speak a language, you know, and type into a machine. Maybe it's 2 billion and 3 billion.

So the competition is so much. No, it's great. But remember, capital is so scary. It's great for everybody except the people Vinnie is talking about who are going through a bit of cods of dissonance. I was special in the world last year, and this year I'm no longer special. Exactly. And I think for developers to come to that conclusion, I'm no longer unique and special in the world is like, it's like losing your Midas touch. It's like being a superhero and all of a sudden they just put a kryptonite necklace around you and you're like, wow, or everybody can fly.

So I'm not Superman anymore. Listen, it's 2023. The macro picture is a little shaky. It's uneasy out there and tech is getting hit super hard. As such, you cannot. to lose sales for silly stuff like not having your sock two right now. If you are unsure about your sock two, you need to check out Vanta. Vanta makes it incredibly easy to get and renew your sock two.

On average Vanta customers are SOC2 compliant in just two to four weeks. Compare that to three to five months without Vanta, huh? And they partner with over two dozen audit firms who have been trained to file SOC2 reports directly within Vantam. This is a total no-brainer. A bunch of my portfolio founders have used Vantam. Wenta, and they've had amazing experiences. And if you don't have SOC2 compliance, you can close major customers.

One major customer, that could be the difference between your startup thriving or going away. So get it done right now. Vanda's going to give you $1,000 off because you listen to this podcast. Think about it. $1,000 off Vanta.com slash twist. You got to write that down. Put it in your notes, V-A-N-T-A-com slash twist for $1,000 off your sock tube. Let's talk about AI artists for a second.

Okay, so you have these authors who can't use a, a pencil or a piece of paper. They have no ability to like, mechanically draw things, but they have these beautiful visions and views of what things should look like. And they're now able to use these tools to create this. So something that they, you know,

the disconnect, I think, is artists who had, you know, even physical limitations, right? You're an artist, you've got injured, you're paralyzed, whatever. Now you can use this. Using voice prompts, you can create the art that you want to create using prompts and whatever else.

So it's opened up a whole world. But what it has done is people who spent years refining their skills, their fine motor skills on painting and sketching and whatever. That's out the window right now for them, because you can't really tell the difference between something which was sketched and something which was just produced by a machine. And so the question is just... Let me just ask the question.

The question is this. if an artist drew something by hand and then replicated the exact same piece of art using a machine, which would have more value? The bespoke one created by the artist, of course. No, no, no, but they both... It's one of one. No, no, no.

The same artist, hand sketched something, and then prompted a machine to build it and let's assume it looked identical, right? Yeah. Which one has more value? If the original one is printed on like real paper, I would have more value because it's more bespoke and unique in the world.

And the craftsmanship that went into it was higher. But most people, if you were going to put it on a t-shirt eventually or put it on a billboard, nobody would care. So I guess it depends on the person. So the question is, is it the motion of creating the art or is it the artist that has the value? You know, I always liked the, I think the proof that the person did this manually would, for me, that I knew they put the work into it, increased the value of it. But if it was just for a T-shirt or if it was just to be put on a post, poster, I wouldn't care. But let's just say it's a print.

Let's say there's two prints and the machine. Can I... The print versus the original go for different prices in art. No, no, but let's say there's a once of print that the machine, you have a printer that produced a once of print, and then you have artists sketching it using pencil. You know, you basically produce exactly the same two pieces of works, two works of art just using two different ways of doing it. What I'm trying to sort of get to the bottom is, are we doing...

And this is the machine did it 10 times faster, okay? and then the human. So now are we gathering, it's the same artists, it's the same output. Are we validing the fact that the artist put 10 hours of their timing to the one and one hour into the other?

Like, where's the value coming from? We do this in a couple of places, Vinny. Think about farming, right? Farming before machinery. Today, and so, you know, obviously the same thing happened to millions of people as machinery came around. But today, to your analogy,

you can go buy stuff at Costco, which is definitely coming from huge industrial farms, right? And go get an apple or maybe a tomato, let's use that, right? Or you can go get a tomato at your farmer's market who, what may have been grown in like someone's backyard, right? And definitely you're going to pay four times more. Sunny, they're two different tomatoes.

The moment you go eat farming, you have. They can be the exact same variant. But then you have yield issues, right? So the, and then so they're using pesticides. The one may be organic. The other one isn't. They are indistinguishable in your example, but the effort put into it matters.

For a person looking for calories or eating the, tomato on a salad. If they're indistinguishable, they don't care. And then for a person who is precious and cares about the process, they much very much love the fact that, you know, this was a hand-shuck or hand-picked. And we literally have that. When you go buy eggs, you can get cage-free eggs. You can get free-range chickens or you can get factory. I guarantee you if I took the free-range chicken and I gave it to Tyler Florence and said, make a free range. And then here's one that was not free range and here's a bunch of eggs make me an omelet that were cage-free eggs versus regular eggs. I guarantee you 99 out of 100 people would be able to tell the difference

or care. They would just go for the cheaper one. So I think we have our answer. Like, do I care if I'm playing flappy birds? That's free because somebody made it in three hours or flappy birds for $10 and have to buy it because somebody put a thousand hours into it. I'll take the three. the thing that I think is most interesting that is not yet upon us is the ability to not just do these chat GPT searches one-off,

which I was doing, by the way, when I, just to do a little call back here, when we were talking about obsolete careers, I said to chat GPT, what are careers that have been totally replaced by technology? And it gave me the first set.

And then I said, hmm, can you give me more, artistic careers that were replaced by technology and I got that second list. So, less people think that I am so quick on the draw here that I can just rattle off these things, I'm literally using chat ChbT as a host of a podcast. The world's greatest moderator has been augmented already. And somebody did a tweet where they said, you're a 10x moderator. You're a 10x moderator now. But I mean, somebody just did a tweet where they said,

build the agenda for all in next week using the stuff in Sacks was like, oh, look, you know, it did a cold open and everything. Of course, it wasn't funny and, you know, what wasn't worth listening to. But eventually, who knows? All right, probably the most challenging thing I hear from founders is related to building. Either they aren't technical and they're searching for a technical co-founder or they can code, but they're just spread way too thin. This is one of the first major obstacles you're going to face as a founder. And it can be discouraging, right? We all know that. And when you're spread then, and you can't get the product velocity, it's very frustrating. You know what to do, but you're just stuck in the execution phase. So here's a solution.

Let crowd robotics be your CTO as a service. And boom, just like that, you can focus on building awesome products and delighting your customers rather than wasting your time on infrastructure, planning, architecture, compliance, and the boring stuff. But Crowd Botics also offers professional scoping to help you flesh out your project at the MVP stage and beyond. So cut out the hassle and get back to build. in that perfect product. When you think about CrowdBotics,

I want you to think getting time back to focus on your product and customers, product drives everything. So let the folks at CrowdBotics show you how it works.

Schedule a free scoping session and get your detailed build plan at CrowdBOTics.com slash twist. That's crowd, B-O-T-I-C-S-com slash twist. So then the question becomes automation.

So I need to get ideas every week for this podcast and for All In. It would be great. Instead of having my producers go out there and do stuff, and our producers, we have producer GPD is one of our jokes here. If I had a producer saying, read the following top 100 tech Twitter accounts,

pull the URLs of the stories they're talking about, summarize them across these hundred people, and build a docket based and sort it based on the engagement, the number of replies and likes to the tweets, and the number of news outlets covering that story. So, hey, there's been 30 people who covered this Twitter story, and there's 20 people who covered this chat GPT story, and 10 who covered this Microsoft story.

Give me some sort of order for these. That'd be an awesome thing to happen automatically. So tell us about this auto-GPT. Well, I was about to say great segue, so you did an awesome job there, moderator. Still some value to being a moderator. Yeah. You're going to have this within 30 to 60 days.

If not, you can already do it right now. So what auto-GPT, and there's a whole category around it, baby AGI, there's a few of these projects in the space. What they're really doing is they're taking the reasoning abilities of LLMs, and they're adding the ability to chain them together with agents. And so what you do is at the high level, you describe a set of tasks. those set of tasks for an LLM alone can't be solved because of, you know, limitations of the LMs not connected to the Internet or it can only go back so far in history.

But what these projects have enabled is for the LLMs to give instruction to the agent to go out to the Internet to go get the answers. And so in your case, if you used Auto GPD, and this, maybe we should just try this as a fun experiment. I can try it later this week, where you would describe what you did in the exact same way you described it, and then one of the agents would be like a Twitter API agent,

so I would know how to interact with a Twitter API, and that's, I think, all you'd really need, right? That's all you sort of asked for. And so when you asked in the prompt to say, hey, go get the top tweets from these folks, they would know how to use an agent that interacts with a Twitter API, in the same way we saw the plugin interact with Instacart.

It would go get those tweets, and then it can do the summaries and all that on its own. It already knows how to do that without that. So it's the really open AI plugins are like a sort of a version of what we're seeing with these auto GPTs. It's allowing the LLMs to have agent capabilities outside of its constrained world that it lives in. This is going to change everything. These were called intelligent agents in the 90s. It was a company called General Magic.

This was founded in 1990. I remember this. I was 19 years old when this thing was founded. Competitor to SGI. Competitor to SGI at that point. And what General Magic did was they made a small device. And man, this had some of the most amazing people. It was in Mountain View.

They made a, and they had this product called Magic Cap. This was the operating system long before iOS and Android. We're talking decades. and I had a programming language called Telescript. Now, what it did that was so fascinating was it would have agents. They came up with this concept

of a personal digital assistant and agents. And it would go out, you tell it, I want flights. This is before the internet and the web, before the web, go out and find me flights to Japan that leave on these days.

This is before those days. This is before those databases existed, and it would go out and try to go find the stuff. It actually went public on the NASDAQ in February 1995. And I don't know what ever happened to. There's a documentary about it, yeah. General Magic. But the Magic Cap OS was really, really powerful.

And this is in a way what that is, right? It's agents that are operating forever. And there, by the way, is this Sony version of it. And you see, they used a desktop. When I say a desktop metaphor, a literal desk on a, like, you know, 2D view of it. And you would click on phone, you click on schedule, you click on your files, your Rolodex, all this stuff. But it would go send these, you know, little agents out, genies or whatever they were called to do this. And I think this is what we now have is that you can just write a script to go do this forever.

and then these things could talk to each other, couldn't they? Can I pull up an example here? Please. Yeah, this would be great. Thank you. Now that I've taken everybody down, memory later. You make that two times bigger in terms of font and maybe come in plus plus. Yeah, sorry.

Yeah, there we go. How's that? All right, so you have the plug-ins here. So what you see at the top of Chat-JPT is plugins. So I've enabled to, in this case, open AI and Kayak. Open Table. Oh, sorry, Open Table and Kayak. And there's a store here.

And you can see there. How do you enable that? You have to apply. You have to be a developer of making plugins right now. And so here's my prompt now. And you know regular Open AI would not be able to do this, right? So find me at, let's call this oops, let's say direct flight tomorrow morning from San Francisco to New York and recommend me a restaurant in New York after I land. Just, you know, and so we'll kick this off.

I'll probably have to speed this up because it'll take a second. And so here you go. So what you see it's doing is now, to your point, J-Cowell's using an agent. In this case, the agent is Kayak to find out about the flight. And you can see it's basically, you know, given Kayak something here. And it'll, it'll, no, one of the issues is having

is there's so much usage of this, these API endpoints are not returning quick enough sometimes. And so we had an error here. So you can see Open Table is barfing. But the idea is that it's able to do sort of what you ask. And so here are the nonstop flights. And then after you land, oh, so it actually came back. I had the error the first time.

And so going back here was our prompt. And so here's the flights. And here you can kind of go and book them. And then here's some restaurants for lunch. tomorrow after you land. Amazing. And now you could do a follow-up there and say, which one has the highest-rated business class?

So let's try that. Which one has the highest-rated business class of those three? And which of these restaurants has the highest reviews on Yelp? Do we add Yelp to this mix? No, there's no Yelp. There's no Yelp. There's no Yelp. Oh, there's no Yelp yet.

See, Yelp is not going to want to give this up. Yelp is going to build their own interface, I think, for this. Maybe they'll have no choice. And which of the restaurants has... Yeah, but they're not real. Has highest rated business class. And which of the restaurants... You recommend with a high...

No, I do... And which restaurants have the highest rated hamburger? Just see if it can go down to a dish level. Yeah, I don't think so, but right, right? you never know because it might go out into the open web to find that right because it still has open web as part of it right you know

no you have to use a different version of open AI of the setup I'll show you up top if you wanted to go out to the web to do that I know that I can search the open web but does this also have the corpus of non open table

you know the previous like up to November crawl of the web so here's see here's what it's done here So it's giving us a quick overview. Yeah. See, it's not even giving you like the highest rate. So it's going to take a little time.

But it did tell you that mint is jet glues and then Polaris business class as United's, which is actually very true. Yeah. And it does give you what they provide. It's kind of neat. Actually, you know what? Like if you were to ask this question to a travel agent, this would be what a, this is simulating what a travel agent would be doing in the back.

ground. They would be checking open table. They would be checking you know. Remember, these plug-ins are only 10 days old.

Like, ask yourself where the app store would. Why did Siri ever add this, Vinay, to their product? Like, you look at Alexa and you look at Siri,

this would be so much better with a, you know, if I had my AirPods in and was talking to Siri. And all Siri can do for me right now is play a certain song when I'm,

you know, going down Westridge, you know, I'm like, play, you know, this song, uh, from YouTube music and, uh, you know, the live version of Sultans of Swing from Live Aid. And I, I can do that, but I can't say, make me a reservation or tell me what reservations are available at great gold and drucky tonight for four. Well, it's coming. Alexa can do that. So Alexa's got the skills that you can use. It's not great, it's clunky, but Alexa can do it. Siri, they've never really taken it that far. I think Apple is definitely looking to this and realizing they have to.

Can I share why? I believe it's because LLMs are actually better reasoning engines than knowledge retrieval systems. And I think that's the real breakthrough here. Let's explain what those two things are, LLM versus knowledge system. Yeah, so what we've mostly had to, to this point in technology is knowledge retrieval systems. So we give it a task and or we ask it something,

and it can go retrieve that information. But it was just following instructions and not reasoning through anything I've said. And so to your point, Jay Kyle, when you're like, hey, play me this song, that's like a knowledge retrieval task, right? It's not knowledge in that case,

but it's going and finding the song and playing it. What the LLMs are starting to, you know, be able to do is reason through things, through all of the training that they've acquired. And so that's creating this sort of superpower capability in terms of being, first of all, being able to interact with it in sort of much more natural language. And then beyond that, for it to kind of go through additional steps and reason through a

problem. And that's where we're starting to see this huge advancement. It's just not what Siri was ever, you know, built for or built around. And Google to that extent, right? It was, those were things that you had a specific question and find it for you. Yeah, see, when you're, when you're using a knowledge system, it's based essentially on a database. And it's based on heuristics and rules. And the rules are set in a, in a, what you're saying, it kind of rigid way.

These are restaurants. These are ratings. This is their location. And, you know, this is the number of reviews. Okay. you can act against those fields in some sort of, you know, and Boolean way. Give me restaurants that are within a mile of me that have five stars.

But you can't say what are the best restaurants in Brooklyn. Let me. Let me give you, exactly, or if you say, hey, which of these two is better? Because that better would require,

like you'd have to give your query some form of understanding around what better means, either rating wise or number of reviews wise. But when you ask OpenAI which is better, it can reason through it. And without giving it a definition of better, it will give you a response between two things. It's wild. All because so much knowledge has been indexed inside of it. And it's guessing the next word based on knowledge that it's seen before.

yeah it's you know kind of again the similar way to our brains work right a bunch of all the little pieces give us the ability to reason what's better than the other one it's so wild that you know one of the things i feel like's happening viny is it's not that we're learning how impressive chat gpt4 is and a i can be and lm's going to be i think what we're realizing is how simple our brains are that's what i think is what i think is actually the big revelation here. Our brains are not actually that complicated. We say, you know, the best hamburger you can get quickly when you're on a road trip is in and out. In and out. It's like literally in and out just came into all our brains.

But there are higher end hamburger joints that emulated in and out at twice the price, including Roan, shake shack and five months. I don't like shake check as much. But Roanberg is pretty good down in union. Jay Cal, can I do a quick show something from definitive, not trying to plug, but that shows us as an example.

That's why you're here. I know definitive, which you really were focused on knowledge bases when you started. You were focused on knowledge bases in crypto and trying to find information there. Yeah, exactly. And we've just started to expand and look at it.

And so, all right. So here's a question. to a crime database that we have connected the system to. And say, what are the top violent crimes in Chicago? But this database doesn't have these things category as violent. What our system was able to do working together. And there's, you know, you can see here, like a lot of interaction with OpenEI to pull this off and our own LL. Say what you're showing.

So people listening can understand. We're going to get an SQL query. We have a prompt. We have a prompt. No, we have just a prompt. The SQL queries generated by our system, right? So we have a prompt to a data set, which is what are the top violent crimes in Chicago in 2023?

What our system was able to do working with our LLMs plus, you know, some external LLMs, is turn violent crimes actually into the subcategories that are listed here. And those are assault, battery, criminal assault, criminal sexual assault, homicide, and robbery. And these are the actual primary types in the database. And so this is reasoning that's occurred, right? It's reasoned its way to say violent and looking at the data set that exists to expand it out to these things. Well, and you put this is sort of the violent crimes. So if you were to take the word top out, it might expand into a couple of other types of violent crime?

It might add two or three more. It might add two or three more. Or if we expand it to more years because in 2020, there's probably, you know, less, right? And so that's what's really fascinating and interesting here is that this is, and this is like an example. start embracing what it can do for you, you can simplify and you can do more. And that's, that's, you know, this is a classic example now that we're starting to see in what the, the, the, the capabilities are of, of the technologies. Well, and then you also have visualization as a component there.

So you can visualize this in some way. Well, and, and we can also have the LLM generate the visualization for us, right? And so, um, so that, that was actually dynamically created. The JavaScript for this is dynamically also created from the LLLLLM. LEN. Yeah, but you could say, and you could ask it, what would be the best way to visualize this? Or show me 10 ways to visualize this.

To visualize it. In order of what is most evocative or most whatever, most accurate. What is the, yeah, it may be a table, it may be a pie chart to your point, maybe a line graph. Yeah. Wow. It is just amazing how this is progressive. That's the world we're entering.

Yeah. And this is all happening, like in, counted in weeks or months. So if we were to sit here, I think chat GPT5 is coming out in the fall. I think November or December is the word on the street. And we have Google Bards is doing pretty interesting stuff. Bloomberg just came out with their LLM. And it's really good for finance stuff. Well, they haven't released it yet.

I think they've built it for internal use. Right, yes. But people are playing with it. Let's talk about vertical LLMs versus general LLMs with plugins. What's going to win the day here? Or is there just going to be room for all of these things? And then we're just going to be connecting them together. I think you nailed it.

You're going to be connecting them together. I think at the topmost level of any interaction will exist the large public LLMs, whether it's, you know, Googles or Open AIs or whoever else kind of, gets there. And that'll be the first LLM that most systems will interact with. And then they will chain out things to either agents like we saw in Kayak to an API,

or there may be another LLM that they interact with that as much as access to proprietary data as, and we'll see that chaining occur between them. That's where I believe we end up pretty quickly. Got it. So you basically saying it's going to go meta. It's like AI talking to AI.

It's, we're already there. We're already there. I mean, many, yeah, many systems that are out there now are, are being built as such, right? Even our system, like, we have LLMs talking to LLMs. Now, all of the questions and things that are inputted into these LLMs are owned by the owner of the LLM, right, Vinny?

So if I were to ask questions to chat GPT and it gives me answer, that information is the proprietary that all becomes ownership. The learning all becomes ownership with chat chvety. No, if you have an enterprise contract with OpenAI, they have do not train provisions. So they will not capture or train. So they've put that in place, in fact.

Because this is where it's sort of interesting is if Google releases this, and I had said this on all-in. I think chat, GPT, just to be clear on opening I, this is one of the 10 greatest technology launches of all time. Five.

Probably. Top five. Okay. What might wind up being top five, right? And so, you know, iPhone, Tesla Model S,

PC, whatever it is. But I also think, you know, if Google puts this on YouTube, Android, Chrome,

and Google search, and they start having billions of queries coming in. Those queries have massive value, and that interaction has massive value in terms of training the data set, just like Google, knowing what ads people click on, etc. So if you were to look at this Vinny and say,

you know, if we say Google's two years behind, I'm just picking a number here, maybe 18 months, I think if you ask the average technologist now who's played with these things, they might say a year or two years. So let's just pick 18 months. If they're 18 months behind,

if they were to allow this to be used on YouTube, on Android, on Chrome, and it just became a default box. So when you open your Chrome browser, imagine this. It has URL slash question. And it tells you type in a question or type in a URL.

You put in a question, it doesn't give you a Google search anymore. It gives you, if it's got a question mark or something, it gives you the bar. answer then the Google results under it. How quickly could Google catch up here? So here's what I think everyone is, being human, we have a problem. We don't know how to look up the slope of an exponential curve very well.

Okay, so when things change exponentially, we just don't see it. We're living in the moment. We don't have good context. This is not something I'm saying. This is something like Bill Gurley said, everyone said this. It's like there's a sons of God, in terms of people who have been in technology for decades who know that when there's an exponential curve in front of you, you can't see how high it goes, right?

So I think we're in that phase right now. I think this is exponential innovation. It's like it's logarithmically, it's like it's just through the roof, right? And so I think that Google is going to have a tough time catching up in the areas that open AI has been focused on, which is quite honestly enabling developers and enabling a lot of people to use the platform. And the network effect and the learnings from that are going to give them a huge,

huge head start over Google. Google can come in later on with something. It may not be whatever it is that Open AI is offering at the time that they get to market with something decent, but it might be something totally different. It might be like, hey, you know what, that's great, but now we have to go destroy our search business and that our search business turns into a personal life assistant

in business where you say, book me a flight, lowest prices, this time, this place, and maybe they get better at doing like direct integrations. I think as far as the developer

ecosystem play is concerned, I think Open AI is going to win that. I don't think Google's going to be able to pull people into that. Yeah. It's going to be an interesting race. I think the pools of data are going to be quickly restricted. Reddit,

Cora, and Twitter have already signaled, you're going to need permission for this data set. Those things are going to be ripped out, right? If they trained chat GPT or barred on that information, they're going to have to rip that out or face significant lawsuits, I think. Is there going to become a market to license all this data

and, you know, chat GPT will have to give Reddit $50 million a year, $10 million a year, to have access to their data or rip it out? What's the size of the search deal, right, between Apple and Google? A couple of billion a year. I think we're, exactly. I think we're going to start seeing that for access to those data sets very, very quickly. This is a whole, you know, there was this concept that we kind of laughed at in the venture

community capital allocators. People would come to us with the business model of, we're going to sell our data. And nobody was ever able to use data. So it was like, you know, unless you had some data about, I don't know, the yield of corn and wheat and then you were able to sell it to some hedge fund who used it to make trades. There was just no way to sell data. Personal informational data and advertising, of course, would be another exception. But it just generally didn't work. Now we're going to look at this, I think, and look back and say, you know what Reddit's businesses?

Turn the ads off. It's meaningless. Cora, they spent 10 years building that business. and it was like, if this thing ever going to make money, it's like, yeah, I think these things are just worth hundreds of millions of dollars a year for their data to train LLMs.

I don't know, am I crazy? I mean, well, I think you're partially right. I think there's a couple of points that are probably going to create some challenges here. One, if someone's LLM's already been trained with these things, it's really hard to rip those things out, right? And so, because the training doesn't usually start from scratch as they're going forward.

that that's a kind of a big technical debate. So they settle with? Yeah, I think so. That's what I would bet. Two, I think what really happens is we probably don't need, and this is what's going to sound crazy, we probably don't need much more training data beyond five and say, whatever the equivalent of GPT5 and six is. And at that point, it becomes more about the type of things that we've been talking about

today and experimenting with today is around, you know, what is the business model? around the agents and how does that work? So I think the cat's been out of the bag for a while and people have been doing, you know, you remember, like this stuff was started back in 2017. And so I think the incremental data training, of course, for proprietary data sets will be huge, but I think for the stuff that's already out there is not as important. And I think what you're going to start to see is that is being handled through this world of agents and plugins.

You might have heard. Listen, I'm reticent to bring up any Elon Twitter stores with people think I have inside information. I do not have inside information. I just saw on Twitter that he bought 10,000 Nvidia. GPUs. I forgot the name of these A100s, maybe.

Yeah, something like that, yeah, yeah. These things aren't cheap. It's like $250 million or something, right? Oh, he bought $250,000. $50 million worth. Okay, so this is a, even for the world's richest man, it's not cheap.

So what do we think's going on here? And also some open AI people are coming to work for him reportedly as well. I have zero inside information. Please do not reaggregate this business insider, whoever. I mean, Elon's been very, very clear

what he's doing. You know, he's basically turning putter into a, you know, an app for everything where you can do payments, you can do feeds. It's like what QQ does in China. He's trying to basically turn

Twitter into that. So he's going to put Dogecoin in as a payments layer. He's going to build AI into it. He's going to improve the algorithms for feeds, rankings. He's in the probably connected to some of the shopping API services, so you'll be able to do what Sunny was doing, look for flights, book it, all in one app,

never need to leave the app. That's what he's doing. I mean, this is not a big secret. I had a really clever idea. I just tested this as a joke. I put hashtag Twitter AI, give me blank. And I was like, because Twitter is in a way, a chat interface.

So I took this picture of Trump, where he is being indicted, I guess, and pleading not guilty. And it's like four attorneys talking to each other with him looking directly at the camera. And I said, make this a renaissance painting with 25% more drama. Now, I know that there's no AI. And I put, how cool would it be if you could tweet an AI request on Twitter and have a reply to it?

Again, I have no inside knowledge of what Elon's doing. But if you go down and you look at it, you will find there was a couple of people who then used different ones. And so mid journey or something to do this. Mid journey or whatever. And like, literally, I was like, well, that's incredible. If you're not watching this, YouTube.com such this weekend. And people started making versions of this that were extraordinary. And so what I thought was, how interesting would it be if, you know, when you did the hashtag Twitter AI,

anybody who had any language model, you know, could be chat GPT, it could be barred, just knew Twitter AI was a general call for help. And they all just responded. Now Twitter would be filled with humans, bots, interacting with, spam. which they're getting rid of pretty quickly, but then also legit AI accounts. This becomes singularity-esque, where like our whole world is us talking to different AI agency. You ask a question to Twitter AI,

like, what's the resolution to Russia's invasion of Ukraine? You know, and Sachs responds and the libs respond and the right response. But then the AI starts inserting itself into the discourse. Brave new world, huh? wild. It's a crazy role. Here's my next one.

Sacks was like, they wanted to make a Batman movies realistic. It would be the mob that takes, it wouldn't be the mob that takes over Gotham City, it would be the progressives. Pelosi, Pelosi, Biden, Biden, Biden, you can add all that stuff for you know, Sacks. So I said, please generate a picture of David Sacks's Batman crouched down

on the peak of the Golden Gate Bridge. And it started making and it's just quite unfair because Sacks has lost a lot of weights, but they made a chubby version of Sacks wearing different, like a blazer, but with underneath it, like a Batman outfit. And some of them,

like these ones were not so great, but there were a couple further down as the maniacs on Twitter started getting into this. They put one with Montclair logos on it. And then there were a couple, that one. This one, it looks like a thinner,

buff version of Sack's, accent looks great. He's literally on the Golden Gate Bridge with a real concerned look on his face. I thought this was incredible. Then they put me as Robin. I mean, it'll just be incredible though, J-Cal.

If that's the world that we're, I mean, we're heading towards where every, I mean, what does it mean for social media or Instagram or any of those kind of things? I mean, literally Instagram could just be us posting our pictures

and saying, here's a generic picture of me I took today. make me look better looking, thinner, more hair, sexier, and then put me in whatever the most exotic, interesting event in the world is happening right now. And I could be like at Coachella this weekend without having to be at Coachella. Yeah. Yeah. We're close to that moment, right?

And then what ends up happening when everyone, like, there was some level, and I'm not in agreement in this, but there was some level of like, you know, FOMO or whatever gets created by people posting those things. That's why people post it. But now if everyone can just post whatever they want, then why would you ever go there? All you'll be seeing is everyone doing the next greatest thing.

To your point, everyone at Coachella and everyone at Burning Man or everyone at, you know, pick your next event, the all-in summit. Let me ask people about the Shanghai upgrade to Ethereum. Supposed to happen Wednesday. This long-planned Ethereum mainnet is expected for April 12th.

Upgrade allows withdrawals from the staking contracts on beacon chain, where 15% of supply has been locked. Many have been debating, but what happened once the upgrade happens, and the supply can be unlocked. There's a bearish look, there's a bullish look. What do we think? Is this like a major moment for Ethereum or not? I think it's a nothing burger.

Nothing burger, okay. Why are people talking about it? What do you think, Sunny? Yeah, I think, to be honest, I think kind of like even our focus on energy right now is elsewhere. I think this would have been a bigger deal if more there was like, look, people in crypto are still doing amazing things. I just think the energy of the broader ecosystem is elsewhere right now. And because of that, it'll be sort of a nothing burger. We won't see any well-priced fluctuations or kind of any huge advancements.

The tech community's energy is shifted in the direction of what we spent the first hour of this pod talking about today. And so because of that, I kind of align with Vinny. I think these are important advancements that need to keep happening. I think they, you know, continually make the platforms go in the direction that, you know, hopefully continue to make them become more mainstream. But I think in terms of like it being like some kind of, big event. I don't think we see that happen. What do we think about Bitcoin

surge? The $16,000 last year, or maybe even earlier this year, earlier, yeah, it's December. Now it's 30. So it only has $970,000 to go for Balaji to win his two individual $1 million

bets. He's obviously not going to win that. I think he's probably a month into that bet, and I think he gave it a hundred days or something or 90 days. Why has Bitcoin doubled in value while Operation ChokePoint, as it's been dubbed, by crypto people. In other words, it's hard to get money into Bitcoin now. Banks have been shut down and, you know, even the mighty Coinbase has gotten a well's

notice. What's going on here? So I've been chatting about this on Spaces and a few other things. There's a couple of things. I think what Choke Point is doing is you think about who is Chowpoint really effective. Is it affecting net buyers or net sellers or market makers? Okay. Okay.

And so let's go through the category. The net buyers are retail, me, you, Sonny, whatever. We go and use our Coinbase accounts and we go and buy Bitcoin. It goes to our JP Morgan or Wells Fargo or Bank of America accounts. That's still happening. That's still happening. No one's shutting off the rails for us to go buy Bitcoin and Coinbase. what's happened is they've shut off the rails for the big traders and the market makers that are out there to go and arbitrage these platforms from one exchange to another, right, and move money around and pay out their clients and do big sales and all that stuff.

These guys are not net buyers or sellers. They're pretty much market neutral with the exception of the opportunity that they take. So cutting off the rails, the only other group that has really been affected is companies that are crypto. companies operating crypto accounts and they employ people in crypto and they're being cut off or whatever. Now, how does that affect them? Well, they probably have to go buy more crypto, stable coins, Bitcoin, Ethereum to pay their employees around the world and say, look, we can't send you dollar wires anymore.

We'll just send you Bitcoin. And so that's probably more, I'd say, you know, a net buyer situation, but over the long term because it's a burn, it goes towards market neutral. And then the miners are the other ones that are affected in the sense that they go and sell their bitcoins, but they probably just don't, they probably don't do it in the U.S. They're probably doing it on finance or wherever else. So, you know, the one thing it arguably could do is reduce liquidity in the system because market makers are less able to move money around arbitrage. But, you know, the flip side to that is I think that there's too much liquidity as you can look at the trade volumes per day. I think it's like, there's an over, the guys I speak to have got hedge funds, crypto hedge funds and doing arbitraging. They're making a couple of percentage points a month, like maybe four or five percent.

It's like an, it's an overtraded space. There's just too many guys out there doing the orb. Yeah, and the ob for those listening is like, let's say Bitcoin's trading for $30,000 on the dot right now on Coinbase and it's selling for $30,100 on a different exchange. They will buy the coin on Coinbase. They'll sell it from the other exchange simultaneously, and they'll make a $100 margin in between. And that's what they do.

And that causes that the prices are more stable worldwide. Cutting these guys off, I don't think changes the equation. Are there more buyers than sellers? I mean, economics is very simple. Why do prices go up, more buyers than sellers? Why do prices go down more sellers than buyers? Like, that's it. It's not that people can't get their money out

or can't easily get their money out and they're sitting on it. I think a lot of people who bought Bitcoin over the last decade just thought, I'm buying this to hold it in case it goes to a million dollars a coin. And I don't want to feel stupid. And if it keeps going from $16,000 to $30,000 to $45 to $10 to $20,000, so don't assume that everyone, like not everyone has the same exit price point, right? So Sunny's extra price point and mine are probably two different ones to yours. You have some Bitcoin.

Like maybe, you know, it depends on how much. you have. Maybe at a million dollars, it's just, you know, it's too much. There's no way you can take that risk. Maybe at 100,000 to write a point to exit, everyone is different. And so you should never judge people what the exit point is because the biggest issue is, as the price of one asset, especially a big asset in your portfolio rises, how does it actually balance or imbalance your portfolio? And if you get a point of Bitcoin is now 95% of your wealth and you're like, well, maybe I should sell 20% and buy a bigger house and rebalance my portfolio. So that's going to happen and it's different for everyone.

I hate the whole view of like, you should just hold it forever. I think you should hold to a point where, you know, it becomes more than meaningful to you. And if it's too meaningful to you and it'll be too disruptive to your portfolio, you rebalance. And some people have the view that you should never rebalance and ride it to the moon. Great. But like, you know, you should have other income sources then. Everyone's financial situation is different.

So advice is a very personalized thing. Yeah. I give a simpler answer. You want a simpler answer? Yes, please. equity markets are back where they were last May and Bitcoin prices back to where it was last May. I think we continue to just see that Bitcoin price and other cryptos is highly correlated to equity markets. And if you pull that up, we'll see both are back to last May levels.

We've had a run in the market this year. And that market has created a correlated run in Bitcoin. It's really interesting to me that there are still buyers for Bitcoin right now which means people believe it's going up people believe it will ultimately go up I think it's going up I mean I think Bitcoin will hit all-time highs this year

you believe Bitcoin will break the 68 or 69,000 Yeah it might be later on the year but I think it'll do it But then we have to see the equity markets do or you would say without the equity markets ripping, you'll see that happen. I can't say, if you had asked me, like,

what would I bet percentage-wise? I'd say there's a 75% chance that Bitcoin heads towards, you know, or hits, at least gets very close within, call it 20% of the all-time highs this year

and maybe goes above it. Because for the same reason that biology, actually, but I think biology scale is wrong. I don't think it gets a million. I think that, I think it rises.

100,000 would be the more. more likely, 75,000, even 60,000. The reason is that I think there's a lot of, like, if you look at what biology is saying, okay, and this is, let me make that. So he's saying basically Bitcoin can rise to a million dollars without any, you know, without any opposition from governments,

trying to clamp down even harder, right? Now, if Bitcoin got to the point where it's worth two, $300,000 and it's effectively draining money out of the banking system, I guarantee you, even retail traders wouldn't be able to trade on Coinbase. Everything would be shut down. They would just be like, you know, if there was a panic outflow of capital from dollars into Bitcoin at a rapid pace, the governments would react to that. Now, are they going to react to 30,000?

No, they've kind of done it, but it's not going to be meaningful, I don't think. I think becomes, so that's the problem. This is like reflexivity type of thing that happens. What percentage of people who are working on Bitcoin projects, I'm sorry, crypto projects have redeployed their energy into AI? Yeah. I would say, I would say, at a guess, no less than 25%. Hmm. 25%.

So AI's gain is crypto's loss. You know, the thing about AI, and look at me, okay, I'm a crypto guy for a decade. and I'm doing Waitrum, which is not a crypto project, and we're using AI. And the reason is because I want to build something that I think is used by people, where as crypto has a very, very small market share, like maybe 100 million people worldwide use it. It's a very, very small, and as a product guy building products, I want to be able to use a product, at least have the ability to get a product to a billion people, right? I may not get there, but at least has got the shot and people can use it and people can understand what I do.

for the past like, you know, decade to explain to your parents and family what I do is just impossible. And I'm like, oh, I'm building like a Zoom type competitor, but it's got an AI assistant like Siri built into it and it can take, oh, that's very cool.

So now you can have all your meetings in this and you never have to take notes. Like, that's very powerful. And for me, that's more exciting than crypto right now because I think crypto has got a lot of U.X problems, I have an idea for you.

You've heard the term a super cut, you know, so people are making, you know, summaries. So when we meet with the founder, Zoom now gives you a transcript automatically. Yeah. We do that. We do that too.

We cut the transcripts. We put them into Notion. Notion has summary tools based into it. It's summarizing it. But I want to see when we do a 20, 30 minute, what we call an introductory meeting for a founder university or launch fund or the syndicate, what I want to see is a supercut,

a three-minute version of, the 20-minute interview that one of my associates did and make a video version of the transcript. I'm sorry, of the summary. So a super cut video. So we, I mean,

we already do all that. We have a clipping built-in. We're doing all that. The feature, I think, which is interesting, which I mean, you probably haven't heard this one,

but we have something called catch up, which is launching soon. I listen to the demo over today. It's looking really good. You walk into a meeting with 20 people and you, you're five minutes late or even 15 minutes late or 20 minutes late. Now there's a catch-up of what just happened in the meeting. What are the action items?

And before you even get into the room and people see you there, you walk in knowing what actually happened. Oh, Sunny and Jason decided that we're doing dinner on Friday at 4 o'clock. Cool, we'll try to add it to your calendar, click a button that adds your calendar. Like these sorts of things happen. So we're basically moving to the point where, you know, like the – and this is why I'm saying this is very exciting for me coming out of crypto because I can build stuff that normal people can use.

In crypto, it's been so hard to build things which people can use. And every single crypto project out there is struggling with this right now. And when I say normal people, like, you can get, there are a lot of crypto projects that can get crypto people to use it, but the moment they try and go over to the mainstream, it just doesn't carry. The only thing it carries the mainstream right now is stuff like Coinbase because people will buy the Bitcoin. But the NFTs, it was a nice little ride up.

And now it's like, it's kind of, you know, it's still stuck in the crypto community. Maybe it's pulled in a bunch of artists or whatever else. But so I, to answer your question, I think 25% of people in crypto have probably moved on to. All right. I'm officially renaming our weekly or biweekly or whatever we're doing here. Roundtable, it's now the AI and Crypto Roundtable.

Let's come every week with a couple of AI stories. Absolutely. Let's do it. Let's do it. All right, everybody. Four. Crypto, bonds. Pidding-Lingham.

And for new money, Sunny, I am the world's greatest moderator. We'll see you next time. Bye-bye.