This Week in Startups - AI demos: Bard's YouTube summarizer, Stability's image editing tools, and more with Sunny Madra | E1853

Episode Date: November 28, 2023

This Week in Startups is brought to you by… Squarespace. Turn your idea into a new website! Go to Squarespace.com/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 SOC 2 report fast. TWiST listeners can get $1,000 off for a limited time at vanta.com/twist Northwest Registered Agent. When starting your business, it's important to use a service that will actually help you. Northwest Registered Agent is that service. They'll form your company fast, give you the documents you need to open a business bank account, and even provide you with mail scanning and a business address to keep your personal privacy intact. Visit http://northwestregisteredagent.com/twist to get a 60% discount on your next LLC. Today’s show: Sunny Madra joins Jason to demo Bard’s YouTube summarizer (2:27), an iPhone app that draws apps (41:14), Stability’s Clipdrop (59:20), and much more! * Time stamps: (0:00) Sunny Madra joins Jason (2:27) Sunny demos Bard’s YouTube summarizer (7:00) Jason's revamping his Substack! First topic; Jason's new ADD management framework: Automate, Deprecate, Delegate (13:58) Squarespace - Use offer code TWIST to save 10% off your first purchase of a website or domain at https://Squarespace.com/twist (14:54) Bard’s YouTube summarizer demo continued (27:09) Vanta - Get $1000 off your SOC 2 at https://vanta.com/twist (29:09) Sunny demos tlDraw and discusses the evolution of low-code environments (40:08) Northwest Registered Agent - Get a 60% discount on your next LLC at http://northwestregisteredagent.com/twist (41:14) Sunny showcases UIDraw, an app to draw and build a web app on your phone with AI (46:57) Sunny demos Trace, a tool to design a UI via AI (53:38) Sunny demos Runway TV (59:20) Sunny demos Stability’s image editing tools, including Clipdrop (1:05:18) Jason and Sunny’s AI predictions and bet * Links: https://www.wired.com/story/inside-story-of-pong-excerpt/https://en.wikipedia.org/wiki/Vin_Scelsastabilityhttps://twitter.com/jsngr/status/1728848624048853442?s=42&t=79J_alIF_jrz1wn1s5pboQclipdrop * Check out: https://www.trace.zip/https://www.runway.tv/https://makereal.tldraw.com/ * Follow Sunny: https://twitter.com/sundeep * Great 2023 interviews: Steve Huffman, Brian Chesky, Aaron Levie, Sophia Amoruso, Reid Hoffman, Frank Slootman, Billy McFarland Check out Jason’s suite of newsletters: https://substack.com/@calacanis * Follow Jason: Twitter: https://twitter.com/jason Instagram: https://www.instagram.com/jason LinkedIn: https://www.linkedin.com/in/jasoncalacanis * Follow TWiST: Substack: https://twistartups.substack.com Twitter: https://twitter.com/TWiStartups YouTube: https://www.youtube.com/thisweekin * Subscribe to the Founder University Podcast: https://www.founder.university/podcast

Okay, so here we go. So here's Pong. Okay. And you can see here. Oh, my God. Yeah, yeah, exactly. This is deranged. It literally wrote the code and made the game Pong from a drawing.

Yeah, and then you can create a link. Oh, my God, this is insane. Do you know how long it took Nolan Bushnell in 1973 to build the first Pong? I just looked it up. There's a wire story. It took three months for them to build Pong. You did it in three minutes. Yeah.

Wow. Wow. It's pretty wild. This week in 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 Northwest Registered Agent. When starting your business, it's important to use a service that will actually help you. Northwest Registered Agent is that service. They'll form your company fast, give you the documents you need to open a business bank account, and even provide you with mail scanning, and a business.

business address to keep your personal privacy intact. Visit Northwest registered agent.com slash twist to get a 60% discount on your next LLC. All right, everybody, welcome back to this week in startups. It's Madra Monda's, yes, every Monday, my good friend, Sandit Mandra comes on the program. He's from definitive intelligence. And, yeah, he's one of my besties. And we like to do demos of AI projects. So it's just AI Mondays here on this week in startups.

every Monday, make sure you see the show, because, hey, we got our finger on the pulse of what's happening. I see a lot of stuff coming in to my inbox, DMs, people now know that if you did something interesting in AI in the past week, that you just CC at Sundeep and at Jason,

first name club on Twitter, X, and you let us know what you're building or slide into either of our DMs or both and send us a link, and then maybe we'll talk about it on the show. Let's get right to it. All right. I noticed that Bard, remember we talked about, hey, they got to catch up and they got to hold people's feet to the fire.

Yeah. And I tweeted last week that Bard and YouTube are now connected in some way. So let's look at this. This is really interesting. I've said this from the beginning. Yeah. I think YouTube, Reddit, Quora, and Twitter slash X are some of the greatest, you know, pools of data, right? They are.

And so what we're going to do here in Bard now, and I have some pre-done in case we don't want to wait, but like we are going to ask Bard, can you provide a summary of the last this week in startups episode? Okay, this is a very specific search being done at bard.gov.com. Yes. And we'll see. Oh, look, and it shows a spinning wheel that it went on YouTube,

and here we go. Wow. Yeah. And then, which is great, it gave me a quick summary, but I can say, what were the key points made by Jason? Oh, well, that's even more detailed. So just to summarize what it showed us in the last screen, it gave us, let's look at the first response if we can.

Okay. From, we kind of glossed over it. I want to give them credit. Yep. Yeah. So what we see here, as it says, the most recent episode of this being in startups is titled,

Open AI, Chaos Continues. It was published on November 21st, 2020, I don't know if that's actually the latest episode, is it? Here is a summary of the episode. Five bullet points, Sam, was far from Open AI. The Open AI board is considering

spinning out a new company to focus on generative AI, yada, yada, yada, I hope this summary is helpful, and here's a link. 1852. Okay, 1852. And that's it. And that's the one when David joined. us. Okay, great. Yes. That was the latest was November 21st. Yeah. So it's right. Look at that.

Perfect. Great job. Okay. And then we did a follow-up. What was Jason's position. Yeah. And it picked up on something you said, which was very good, I thought. And I remember this from the episode. He said recent turmoil was unprecedented. And this one was, I remember this very clearly. He said the Open AI board made several mistakes, including not giving Sam Altman clear incentives and not having a clear secession plan in place. Sure, I did say that. I don't know if I mentioned secession, but I definitely mentioned the incentives. So I wonder if it's hallucinating about the secession or maybe Sacks brought that in.

He said the fallout from open AI situation could be significant for startups that rely on open A's model. Somebody said that. Calcanus also said that he believes Open Eye is still a valuable company and that he's hopeful that came recover from the crisis. Overall, Calcanus's key points with that open AI situation is. unprecedented that opening board makes several mistakes and that the fall off my message. Okay, I think it's pretty good actually. Um, and this would save you a lot of time. One of the things we do here, like as an example, I'm having Kevin O'Connor, the founder of

double click on the program. What I asked my team to do was to give me some bullet points from the last time he was on the show. Now I'm going to listen to the episode, but I like to have the bullet points there for me. Usually a producer, Sundip would listen to the episode, write the bullet points. I listened to it on two X speed. So this would take him an hour, right? maybe two, you know, if they're doing other things. So, you know, that's going to cost me whatever. If you, you know, call it $100 in producer time or whatever it is, you know, if an average producer is 50K all in fully baked with their vacation and everything,

you know, or if a producer, you know, general or a researcher generally in the world might cost. If you're going to hire a freelance researcher, I think to do this work, what would it be? 40 bucks an hour? I think 40 bucks an hour you could hire researcher. Yeah, you need someone, you can't just get the cheapest person, right? No, no, you can't do somebody offshore because they may not understand the context of all this. So, you know, a U.S.-based or North American-based person who understands tech would be $40 an hour to do this. And this does it basically for free.

So if you're a researcher and that's your gig in the world, there's two things that are possible here. You could just go do this and claim, you could go use Bard, claim you did the work and charge $40 for five minutes of work. Or you can then take this work and then do more meaningful work. which is take this and then come up with follow-up questions for the person, right? And so this is my general theme. I'm working on a blog post. I am so inspired by our discussions here, Cindy, that I'm going to break some news here. I have a new blog post coming.

I'm going back. I have a calicanastas, subsack.com. I'm going back to blogging and writing. Okay, okay. Just for clarity of thought. And is it AI enhanced? It is not. But they're going to be short and brief.

And I may, you know, I like to have my voice, but I do. do use Grammarly, which is an AI machine learning to, you know, make things clearer. And I set it to make things very clear. So, you know, shout out to our friends at Grammally. I love that product. Great. I have a new framework. I call it the new management framework. You know my old one of like the SOD and the EOD, start of day, end of day for managing real people. I just ask everybody on the team for five minutes. Most people do it. If they don't do it, I generally fire them. For some period of time, I don't keep asking them. I just cut them loose. But I ask everybody to do a start of week, end of week.

That should take 15 minutes at the start of the week, 15 minutes at the end of the week. And then your SOD, EOD should start of day, end of day, should take but five minutes. Just three or four bullet points of what you're going to do today and at the end of the day. Some people go crazy. They want to get credit for all their work. They're ambitious. They want to show the team that they're really crushing it. And I totally respect that.

If it was a young J-Cal, I would do that. Young Sandeep would do that too, right? You want credit for work done. What this does is it creates the ability for you as a young, young person in the company or a new person to see what everybody's doing and then say, wait a second, that person's more senior than me, I need to be doing some of their stuff, right? Yeah. Or now I know.

And then if senior people don't do it, I had two people who were like managing directors who refused to do it and they're no longer with me, I'm just not going to push people to do it. I just want people to do it because it holds you accountable. And then I realized from our discussions here, my ADD framework. People tell me I have ADD, ADHD. I don't know, maybe. We didn't diagnose it when I was a kid. You get diagnosed with that, by the way.

back then. No, no, that's well beyond us. Yeah, that was like a 90s phenomenon, right? It's a millennial thing, yeah. So anyway, yeah. Automate, delegate, deprecate. I'm now telling people, and I'm going to put this out on my substack. I love it. And I'm going to post the whole thing, ADD, but for the modern era. You look at your end of weeks. I took somebody who's going, I'm going to be on leave, and I just looked at their EOW for the last four weeks. And I said, of the items, that they got done each week. What could be automated with GPTs, with AI. What could be delegated to a remote worker?

So I am getting obsessed with remote workers in Manila, Portugal, South America. It's $1,000 a month, basically, for a remote college-educated, perfect English. There's a bunch of services that do this right here. Athena does it. There's a bunch of very cool services that I've been testing. And then you can also just go on Fiverr or other websites and find the Upwork. really great websites. You find the folks,

they cost literally $5, $6 an hour, $3 or $4 an hour is a great salary if you charge $5, $6, $7 an hour. So I'm pairing my people with admins from offshore. So every three people

will have one admin, I think. So that's about $12,000 to $36,000, depending on which service you should use. 12 to 36. Now, if you put three people's salary together all in, it'd be like $300,000. If you put a $25,000,

it was rounded up to 30, like a really high-end offshore person. That's 10% more. But how much could they delegate to that person? I wonder. So that's the experiment I'm running. And then deprecate is, you know, when you ask somebody to do something like two years ago and they're still doing it, but you didn't know they were doing it? And you're like, oh, my God, stop doing that. It's a waste of time.

We don't need that anymore. Yeah. And so I'm just asking everybody to have ADD. Yeah. Automate, delegate, deprecate. And if you do that, I think everybody. and the company can become 50% more efficient, which is to say they can automate maybe 10, 20% of what they do,

they can delegate maybe 20, 30% of what they do, and they can deprecate 10 or 20% where they do. Put it all together, half the work comes off your plate. What does that do? It means you don't have to hire another person or that person can do higher level things. So I just want to encourage people, as you see this incredible YouTube experiment,

if you were a producer of a podcast where you did summaries of podcasts or other news stuff, start with Bard now that it works. Check the facts, pull up the transcript, summarize the transcript, take the entire transcript, put the transcript in Claude, put the transcript in Bard, put it into chat, GPT,

and look at all three simultaneously. And then say to yourself, what could you do beyond this? So anyway, this is just my, you know, new framework based on what we are figuring out here every Monday. I think it's a framework for the AI era. We should, that's what, it's really because, you know,

the automate, delegate and deprecate is very much aligned to the AI era. Yes. Because before automating would require someone to, you know, maybe write code and getting resources across the world was harder. All these things are being enhanced by AI. And I can guarantee, Jason, that most people on those offshore services are going to be better at using AI than the FARCs.

Yeah. They already are. I've been talking to the people who are running some of these services. They are using AI already. They are using script. They've been using scripting forever. So as an example, we have an SDR that we're going to test,

a sales development rep. The sales development rep, one of the things they did was they built scripts. What do these scripts do? They go on to professional networking sites like LinkedIn, crunch base, pitch book, whatever it is. And there are tons of these different sites out there.

And then they look for targets for the salespeople. And they've just automated, qualifying them, and then even queuing up an email to them or queuing up, finding their email, and then queuing up an email to them or an in-mail, etc. So imagine you're a sales executive, you're selling SaaS software. This SDR could go into your LinkedIn, your crunch base, put into your drafts folder 20 emails for you to look at with those folks, follow them automatically on LinkedIn for you, maybe like a couple of posts, this is called social selling, well, that's all going to start to be done by, I think,

AI in some ways or offshore. And that combination is going to be wicked. And then if you watch the person working, so we've had people on the pod who are building software to watch the desktop, right? And Microsoft's watching the desktop with their new co-pilot, right? If they're watching the desktop and they see an SDR or somebody who's, let's say, booking meetings,

and they watch them do it enough times, the AI just watching your desktop and watching what's happening in your browser or in a headless browser, it's going to figure out what to do and then just go do it for you. If your landing page looks terrible, I'm out. We all know that. You see an ugly website, you skedaddle. You leave. You're done. So you need to stop selling for okay or good and start using Squarespace so you can be excellent and extraordinary. It's an out-of-the-box business solution to build beautiful websites, engage your audience, and sell anything you want. You know Squarespace is amazing features, gorgeous templates that are always optimized for mobile,

drag and drop web design with their fluid engine, advanced analytics, marketing analysis, sales data, and more. And with Squarespace, you can create an online store or start a blog at the click of a button. Create a subscription business for members-only content, and so much more. And you can do this all simultaneously. It's the simplest, most effective, and best-looking way to start a business online. So here's your call to action, Squarespace.com slash twist for a free trial. And when you're ready to launch, go to Squarespace.com slash twist for 10% off your first purchase of a website or domain. One of the Open AI engineers yesterday wrote like a little Gmail thing using OpenAI to go through all the Black Friday emails because obviously, you know, it's been crazy the last day or two, right? I remember three, four days.

I did a little shopping online, I'll be honest. Yeah, yeah. Did you do any black Friday deals? No, not really. I mean, it's like. I just had one. I had like a brand I really like this brand R-H-O-N-E

R-H-O-N-E R-H-O-N-E-R-O-N. Yeah. I like Rhone. Yeah, but it's Ath-Lezer, but they got into sweaters and some polos and nice stuff. Oh, nice. So I like Cuts. I like Rone.

Okay. You know, I like Mac Weldon, some of them. And I just saw the Rone go by and I had bought some stuff I really liked. Like they had this nice collared shirt that I like. Yeah. I just bought two more. They were 30 or 40% off.

They have a store in Hudson Yards, by the way. Oh, they do. Oh, they do. Isn't that interesting how all those direct-to-consumer brands were supposed to get rid of stores? And then they were like, you know, it would be cool if people could go into a store and try this stuff? But I guess they just look at it as marketing, right? I think if they break even on the store, but they got people to know the brand and they know they're going to go order online. It's kind of brilliant.

Yeah. Well, next time we're there, check it out. I give this, I'm going to tell you right now. I got to show you one more thing. We kind of got a little bit into. Can you do the comments? Can you ask it to, does it have? insight into comments?

Let's do that, but I want to show up what I said. I had made a prediction on the Monday pod that everything would get back to normal. And it got, this is a heavy hallucination. First of all, it thinks I'm the CEO of the Madrona Venture Group, which I am not. So, a little, is it saying you're part of, is a Madrone venture group, like an Indian group? Is this like a little? Is it, Madrona? What is Madrona?

These are the Seattle guys that like, aren't they? I wonder if there's, I was just wondering if it was like being a little racist and it's like, oh, well, you're India and you must be part of the Madrona venture group. Yeah, who knows. Who knows? And then my prediction was that everything will return back to normal in that episode. And so it didn't get that right. So going back to, you know, our previous conversation.

Okay. I think it's cool that it's, it's making some things available, but then it is, it is sort of. Why is it getting it wrong? I know. Google has the actual transcript. This is why we, should not get this wrong. Chat, GPT, would not have gotten the, this wrong, I don't think. Okay. Okay. Uh, what are, okay, uh, what are the most popular

comments? Yeah. On that video. On this episode. Okay. Sounds great. Yeah. Let's see. It gets this. Yeah, I spelled episode wrong too, but it doesn't matter. I should know. Yeah, it should figure that out. So let's see. So it's searching YouTube. It's giving like a spinning wheel that it's searching YouTube. See, this is a thing. If the structured data can be organized, here are some of the most popular comments. I can't believe the board fires him. He is, a visionary. I'm hopeful that the EMT shear will be able to bring

some stability amateur, I guess. The whole situation is a mess. I wonder if these are actually really the comments. Hmm. Yeah. The drama doesn't know.

Okay. Well, at least it thinks it's getting the comments. So that's pretty good. Yeah. I'll have the producers go check if those comments were there. Ask it how many likes did the episode get?

Or how many views does the episode have? Yeah. How many views does this episode have? I think of Sachs is on, so it makes it pretty popular. Let's see. 80,000 views.

use. Okay. And I think I had it here. Yeah, it says 80K here. Okay, great. So that's pretty interesting. Can you ask it another question here? I want to ask Bard, with its integration with YouTube. Somebody can clip this and send it to the team over there. Ask it, what are the top five most popular videos on YouTube this month? Now, this is like a meta question, right? It needs to understand that YouTube is a website with videos and that there's a popular list somewhere. And let's see if it can actually Oh, okay. A Mr. Beast. Stokes Twins. Yeah. Mark Robert does get

a lot of views. Okay. The good wrench. They range from 2.2 to 23 million. Okay. I think this could be a game changer, folks. We could have a game changer. How about we do this?

Who is the most popular chef on YouTube? Let's see. This is like, something that you actually might get value from, is actually trying to figure out, instead of typing chefs, okay, yeah, it's giving you...

10.6 million subs. This one has 21 million subs, but yeah, it's really, okay. And it's giving a little picture, Jamie Oliver. Yeah. Has 5.8 million subs. So it's not ranking them by subs, but it's giving you a pretty good overview here. Let's ask, what's the most popular

French omelette recipe video on YouTube. This is interesting. See, like this, if should just, maybe they should just replace the YouTube search button. Like, YouTube search doesn't do this. The most popular French video YouTube is how to make, by Jacques Pepin. That is actually the one I've watched.

Wow. That is correct. It is a great video. You ever have a French omelet? You know the difference between a country omit or French? I know the French is like this slow stirring, right? It's like the test for chefs, isn't it? It is the test for chefs.

If you want to test a chef, you have to make a French omelet. It's very simple. You put the omel in the pan with butter, but you don't brown it on the outside. It is like a nice glossed color. You can put chives inside of it if you like. That's what you see here with the green one. And then you have to keep taking it off the heat so it doesn't burn. And then you roll it into a little cylinder.

And it looks like a managot or like a canoley or something. It's rolled. When you cut it open, it's moist on the inside, firm on the outside, but not burnt. A brown butter, country omelet is when you're a managot or like a man. when you mix it and it's almost like scrambled and it's hard, then you get in a typical diner. Yeah. And it's really brown on the outside.

Hotel breakfast buffet. Yeah. It's kind of gnarly. So the reason I know this is because I love a French omelet and I, we have a housekeeper and I taught the housekeeper how to make French omelet just by sharing that video. And I was just curious if they actually could go dig that deep. I'm going to go ahead and give this an A.

I don't know what it was last time, but this is like, I gave it like five or six things that I wanted. It gave me back a semblance of what I want. And this is 10 times better than YouTube search. So for me, it's either like B plus, A minus, or A, but compared to previous versions, I don't know what we gave it in previous versions or were giving letter grades on previous episodes. This is really showing what YouTube is going to be able. This is the first time.

I feel like maybe there's a J-trade coming in for, for YouTube shares. Well, private data and good experience. Like, I'm at A minus only because it hallucinated pretty heavy in this one of the summarizations. Yeah. But I think it answered everything. It pulled up videos.

We gave, you tested it in a quite a few different ways and it did a great job. And I think, you know, for me, what's interesting is there's all these tools that we've been using to summarize videos and now, you know, just do it in barred. You don't have to pay for another tool. can do it right here. And I think as long as they can start to prove the accuracy, I like the comment search too. That's a good one because sometimes I can help you figure out like maybe what you should

listen for or look forward. And the only thing I would say if I could ask for a feature, if they're doing this now, maybe it can even create a summary for you. Yes. If it's like one of these long Lex Friedman ones or, you know. I mean, I think it will do that if, and it will start to do that automatically or automatically does already like I think time stamps pretty easy. Let's ask you one more question. I want to

ask you a question that would be deep inside a video and this is going to be a little bit of a deep pull here but everybody knows I love the band Dyer Shreates. Have any of the members of Dyer Shreitz talked about doing a reunion tour in a YouTube video? Please jump to that timestamp

or please give me that timestamp. If so, please give me a link to that. Okay, so how about that? That's what you really want to do. Yeah, sure. a link to it. I mean, I would like to go deep. Okay, yes, Mark Knopfler has talked about doing a reunion tour in a YouTube video.

The 19-year-in-Rolling Stone, he said he would be open to the idea of reunion of the band, but that would be a lot of work to put together. He also said he would only want to do it if the original members of the band were on board. Mark Nau-R-Korneutored another interview with Mojo. He said he would never say never to go to a tour, but he didn't think it was very likely

to happen. I don't think this is true. I don't think this is true, because Mark Knopfler has just said he would only do it for charity. the Guardian novel said he was more open to the idea of doing an insurer that hadn't been passed and taken to others members about it so I see I don't think this is actually true

he said it's just like that error is over former Diochrates let me see that but it's can you probably link to the video Former Diochrates we'd love to reunion but Mark Naufler is busy that's the truth Diochrates leaves Diochrist's for the record by Mark Noffler

Diochrates Mark Novice in Spanish City These are all great videos I've seen of Mark talking about it but the truth is he has said they're not going to do a reunion tour. So this is pure hallucination. And it gave wrong information. This gives you the idea that it was going to happen and it is not. So it's definitely hallucinating.

It's giving you general information about other bands probably talking about a reunion. Yeah. Well, here it was very specific, right? He has talked about it and he said, we're open to doing it, but it'd be a lot of work. So I wonder if he said that. And by the way, producer John was looking for the comments that it said it was going to have. And those were hallucinations. Couldn't find those comments there.

Really? So it's making that up. So I'm going to a B plus right now. I'm dropping my score. I went from a real time from an A to a to a B plus. I like the fact that it knows what it's doing, but the hallucinations are far too wrong here to make this value. Actually, I'm going to a B.

I'm giving it a B. Yeah. Well, since the comments were fake and then the dire straits thing, yeah, I think definitely it goes down to a B because, you know, that can give you, I mean, imagine you're supposed to watch. something, you know, for like a, you know, something important and you go get one of these summaries that it just told you a bunch of things that weren't true. So that's, that's pretty

unreliable. We've got to be careful. Careful out there, kids. Yeah, be careful out there. Do not trust any language models right now. Trust but verify is the rule. They hallucinate. They're still hallucinating. We're just too trusting of technology. I think, isn't that what this says? The fact that we get so excited and then when we check the facts, we get so disappointed is that we actually believe in this technology a little too much? Well, I think it stems from the fact that we trust technology, Jason, right? We trust calculators. We trust spell checkers.

We trust grammarly, right? And we trust Google search. And I think the challenge with the technology, and a lot of people are working on us right now, is to eliminate the hallucinations because that's going to take away from the trust that we have of technology. These hallucinations are happening one. If you were to explain it to a neophyte. Well, because these models are inherently probabilistic, and you don't have a model that has every single thing that is on the internet, otherwise the model would be too gigantic in size to run.

So what it has to do, it has to effectively compress down the internet by a process of training. And when that's happening, it loses information. And so when it's trying to recreate things, it tends to make things up where these models are starting to drift is they've been further, I guess, reinforcement learned to always give humans answers. So one of the things that they've learned is humans don't like it when it says, I don't know. And that was the beauty of Google search. It never said, I don't know. It would just kind of give you a list of links. Yeah, start clinking. Look at whether those were good or not. And so, yeah. And so I, like, that's the reason is that

you can't take the entire internet and put it into a model, the, you know, the hardware underneath it's not able to do that yet today. And so, All right, good explanation. And so everything is compressed and in compression, there's loss. There you go, folks. Listen, selling software is hard enough right now, man. It's hand-to-hand combat out there in B2 land. The last thing you need to do is slow your sales team down because you don't have your

SOC2 dialed in. So if you're SaaS or a services company and you store consumer data in the cloud, you know what you need to do. You need to check out Vanta. They're going to get your SOC2 compliant easier and faster. And Vanta makes it so easy to get and renew your SOC2 on average. Vanta customers are SOC2 compliant in just two. to four weeks. Compare that to three to five months without Vanta. They're going to save you hundreds of

hours of work and up to 85% on compliance costs. And Vanta does more than just SOC2. They also automate up to 90% compliance for GDPR, HIPAA and more. You can't afford to lose out on those major customers, the lighthouse customers, the big fish, the whales because of silly stuff like lacking compliance. Just work with Vanta. I'm an investor in the company. It's a great company. Get your compliance automated. Get it tight. Tight is right. And close those big deals. Here's the best part. Vanda's going to give you a thousand off. Because they love this weekend startups,

they love startups. Vanta.com slash twist. That's V-A-N-T-A-com slash twist to get $1,000. So off your stock, too. We spent a lot of time on that one, but for good reason, because YouTube is the,

I think YouTube is the best data pool out there. Overall, I think Twitter's, well, I'd say X is good for real time, but YouTube has all these deep interviews, and I think that's going to be, that's why they want. Yeah, of course. By the way,

that's why they're making a big push for podcasts. a big push for podcasts and this is why Spotify could have something really interesting at some point or anybody podcasts are freely available to index when you think about it

now I think you probably need permission if you're going to build a well they're all on RSS feeds so you could just look for RSS feeds with attachments and just start indexing but I don't think you have the rights to do that now will a podcast or care will I care for this week in startups

to be in somebody's model I have to think about it if I got credit and links back I'd feel pretty good about it because I would get traffic back. So this is like a Google model. So we got to figure that out. We got to figure that out.

Yeah. Okay. All right. This one was making big rounds last week. And, you know, we had it queued up, but we did the podcast on the open AI stuff. So this one I find really fascinating. It's called TL draw. And what I did here, because I didn't want to do it real time, I kind of queued it up.

I drew this. I saw someone else do it. So I thought this was a good example. So I made up a game. It's a famous game, J-Cal's first video game. Pong. Yeah. And it's very monumental, right, to think about what's going to happen here,

because this was one of the first video games ever created for Atari, I believe. And so what I said is like, this, here's a paddle. It's controlled by the Q&Z keys. And this is just a line drawing you made on your computer. No, no, like I drew it in this app down here. You drew it in this app. You made a square. Yeah, I made a square.

You made two rectangles for the paddles. You made a ball. Yeah. I get it. And then I, and I, told it the rules and I said the, you know, the scores here. And then all I have to do is I'm going to take this, I'm going to highlight all these things that I did. And I'm going to hit this button that says make real. And this is going to spin for about, yes, this is going to spin for about

maybe 30 seconds. And, you know, think about the work that went into building the first poem. And what we're going to have here. And what this is going to do is, you know, we've talked about like low code and no code environments. This is the evolution of that. Okay, so here we go. So here's Pong. Okay. And you can see here. Oh my God. Yeah. Yeah, exactly. This is deranged. It literally wrote the code and made the game Pong from a drawing. Yeah. And then you can create a link. Oh my God. This is insane. Do you know how long it took Nolan Bushnell in 1973 to build the first Pong? I just looked it up. There's a wire story. It took three months. for them to go punk.

You did it in three minutes. Yeah. Wow. It's pretty wild. That is mind blowing. Yeah. This one for some reason is missing one of the lines. But when I did this earlier,

I had both paddles. Maybe I got to run it again or something. I mean, that's unbelievable. Yeah. Yeah. And you can see here I'm controlling the, you know,

and it's keeping score as well, right? So, you know, wild, huh? I mean, when we start thinking, about this, all you have to do is extrapolate from here, because that's how technology works. So Pong was one of the original games, then you get breakout,

then you get Defender, you get asteroids. Yeah. You know, and then you could say, you know, to the game, make a more interesting version or make a version that works in VR or make a version that works with Apple's headset and uses some IP from Disney.

You know, you're going to be able to just talk to this. Yeah. So you do a drawing, a sketch, It makes a computer program. And then you just talk to it back and forth. And I was in the car this weekend driving, you know, to Lake Tahoe with my 13-year-old. And I set up my, there's an action button. Do you know that iPhone 15 has an action button?

Yep, exactly. I program my action button. What do you think I program my action button to do? Chat GPT. You got it. So when I double-click my, or when I hold my shortcut, it opens up the chat GPT voice mode. voice mode. So now I'm driving, not looking at it, got it in the phone stand in the cradle. And I asked it about the conflict between, you know, I asked it about Hamas, the organization

share questions about that. I asked it about the Middle East conflict. I asked it about the solutions. And it just talked to us while we were driving. And then I asked it about the cure because she wanted to know about boys don't cry and their greatest sits. And then she said, well, what genre is the cure? And I said, well, I think it's new wave. And then I asked it. And it was like, they were post-punk. And then as they evolved, they became new waves. So I was right. But I didn't think they were post-punk so much. But then that kind of made sense to me.

Yeah. They gave a pretty great answer. And I was just thinking, this is like having a tutor, not just for my daughter, but for me, too. Yep. Because I'm guessing, okay, what genre is the cure? You know, like I kind of think it's rock and roll, kind of new-wavy. Yep. Kind of punk I didn't think of.

But, yeah, now that I think about, yeah, it was a little bit post-punk. wasn't it? And so anyway, you'd think about what we would do listening to, there was used to be a radio show on Sunday nights. I forgot the name of it. Vince Gelsen did it in New York. And it was Vince Gelsenna did this show every Sunday night where he just explained rock and roll to you. And people would record it and tape it. And I was just like, wow, this is like Vince Gelson's show on K Rock. Yeah. We've lost the art of that. Even like MTV back in our era, like, you know, they would go through kind of. do that. We've just lost that now, right? It's all about, you know, kind of advertising these days. But, you know, those experiences... Idiot's Delight was the name of the show. Idiot's Delight was the name. Delight. Oh, yeah. Okay. Finn Scalces, S-C-E-L-S-A-I-S-E-D-A-I's,

It a Delight. A really interesting show. And so what I take from what you just talked about there, Jason is, I don't know if you've seen Salcon's talk, the Two Sigma talk that he did at TED, which basically talks about, if you give anyone a tutor, they can improve their test scores by two standard deviations, right?

It's a great TED talk. It's from this year. And he talks about the impact of like chat GPT and LLMs for that. And so, you know, think about the amount of knowledge these things have, the costs, which is definitely pretty minimal, and making that available to everyone on the planet. Like that's a huge lift for our entire society, right? A lot of challenges we see come from lack of education. And so we're on the front end of this curve, but.

I love that use case. I'm going to try it out. Yeah. It's just a great thing when you're in the car. I think also it could play games with you. Remember Alexa had some games that could play? Yeah, of course. I think it's going to play some.

You can make up the games too. You can make up the games? Right. We could just say, hey, we want to play rock and roll trivia. Can you give us some rock and roll trivia? I bet you we can do pretty good job at that. So I just think it's like... Or lyrics.

Yeah. Well, this concept of a personal tutor, I think is going to be very powerful. In fact, I was thinking about doing like 20 years ago, a little bit of, fiction writing and never got around to it but my first story that I wrote it wasn't like a black mirror story

but it was a little black mirror-esque that somebody had put implants of headphones in their ears and that they had a personal assistant that would talk to them constantly and it would read to them stock prices and it was about some internet version of the pin

well oh yeah exactly but it was an implant actually and it was a dot com person and I had written it about the dot com era but like 30 years later and they went and sane because the stock market went 24-7, which was a prediction I had that the stock market 24-7, and he became obsessed with the stock price and trying to change it and manipulate it.

And the stock price was being read to him every 30 seconds. And then he would be doing things in between. He went mad. Yeah. Because I just had interviewed somebody as a journalist about, you know, this flow of information and then somebody was pitching a 24-hour stock market in the dot-com era. It never happened. And people figured out this is a really bad idea.

Well, when Neurrelink were not too far away. Well, and that's an interesting part. You know, 23 years later when I wrote that little short story, I wish I could find my laptop with it. You know, now I think this AI and the TV, then movie horror, kind of gives you a path to that. You've got to be very careful with this because you could just be in constant dopamine hit mode, constant learning mode.

And I have a little bit of that now. I don't know if you have this addiction to podcasts, but I have probably 100 podcasts I listen to. And I go to sleep, listening to them. And sometimes I stay up for two hours listening to podcasts. It's very dangerous. Yeah. Yeah.

I have to really... I wish I'll call on the pot, yeah. Yeah, for that. And, you know, definitely watch the talk, Jake, the two-signment talk. Yeah, I'm going to watch it now. I remember when that went. You know, it's another example of like, there's just too many things. I'm going to dump that into...

Actually, how long is I'll probably listen to it. What if we should go back to Bard and ask for the summary? Yeah, well, you know, interestingly, the way I do this, literally this morning I was talking about my producers and I was asking producer, Nick, how do we do this? The way we do it right now is we take the... a YouTube video. There are kind of websites out there that let you take a YouTube video

and save it as an MP3 file. They're kind of like Russian websites or whatever. You can, and there's one called ss, YouTube.com. So if you're on a YouTube video, if you put an SS in front of the word YouTube, so you just go into your URL and your browser,

it'll take you to a website. I'll show you what it looks like. Oh, looks like it got discontinued in the United States. Anyway, what this video saved from the net did was it would let you save an MP3 file. This is like a cat and mouse game. And then you can take that MP3 file.

You can upload it to Descript or another service. It makes a transcript. They take the transcript. You upload it to Claude and make a summary. So what we're showing here from YouTube is maybe you don't have to do that two or three-step process. But Google's Bard is not a high-functioning LLM yet. Is that what we've come to? That Claude, Falcon, and Chat Chappet, and GROC are better?

Oh, man. I think CHAPT is in a league of its own. Okay. Still. And I think Palm is improving all the time. I'd love, you know, hopefully maybe some folks get back in touch with us and explain why we saw such wild hallucinations. And if there's a way for us to understand.

Yeah, let's get the barred people back on. Yeah. We had somebody from Bard on back in the day. So let's do that again. So here is the TED Talk. You just summarized it. Let's take a look. We had Jack came on from.

is this actually good here? Yeah, I'm just kind of reading it because I pulled it up. But in general, this seems pretty accurate to what he says. Basically, it's like AI has a potential to provide personalized instruction students at a global scale, what we're talking about. And it's been unattainable because of cost and scalability issues. And then he talks about like the impact of tutoring. This is, yeah, it's a good, it's a good high-level summary. I don't think it made up anything here.

Yeah. So, okay, congratulations, Barb B-plus. On this one, making Pong, that's an A-plus. Okay. All right, making Pong's an A-plus. What's the name of this? That one is called, the service is called T-L-Draw. I'm just going to pull it back up for-T-L-Draw.

T-L-Dra. Yeah, so T-L-Dra.com, and that's it. Now, something related. Who makes it as a startup or something? Yeah. Wow. Just incredible. Starting a business used to be such a painful process.

You needed to get a lawyer. There were tons of fees. It was a mess, but not anymore. Just check out Northwest registered agent. They're going to help you form your company fast. Remember, speed matters. And then they're going to get you the docs you need to open a business bank account instantly. Then they're going to provide you with mail scanning and a business address.

And they're going to do all that keeping your personal privacy intact. Northwest can form LLCs, corporations, and nonprofits, and here's why founders love Northwest. There's no hidden fees. There's no upselling. You can call them or cancel at any time. And Northwest has the best of both world solution. It's simple and self-serve, but they can be hands-on if you need help with their amazing registered agent service. Northwest provides everything you need to start and maintain your business. And they're giving Twist listeners a 60% discount. For just $39 plus state fees, they'll form your LLC. Corporate. or nonprofit. So visit Northwest registeredagent.com

slash twist today. Northwest registered agent.com slash twist. I want to show this gentleman Jordan singer. He did a similar one. And so all of these are powered by the GPTV API. So, you know, the vision that we've played around with a little bit. And I think you've even tried it. You gave some, I think you were doing some food things as great JCal.

And so this person created basically, an iPhone app, which you can go and download if you go to his tweet. He has a GitHub link to it. And I'm just going to play the video. And in the video, what he does is, and I'll just explain it for those listening, he's basically drawing like a number keyboard. So one, two, three, four, five, six, and then putting plus minus equals. And then he gives it a name here. And what you're going to see is he's going to say create. And this is on the phone itself. And watch this. And so this is like app creation. And look, I built a business making apps.

I remember, yeah. Yeah. And, you know, this is, it'll be done here in a second. And so it's spinning and there you go. And he's got a working calculator from a sketch. So he literally drew a calculator, one, two, three, four, five, six, seven, eight nine, zero, plus minus, etc. And then basically told it to compile it and it did it. Yeah. Yes. And a working iPhone app. A work. Well, this looks like a browser app to me, but I guess. No, it's not. No, no, no, because if you go, he gives you, this is like a Swift app. It's not, it's not a, yeah. So it's, so this is a Swift app building Swift apps? What is he in? Is he an ID or something? What's this? No, no, no. This, his app is the app is the app that builds apps. It's very meta. Oh my God. So it's an app. You draw on it and it makes an app inside of the app. Correct. That is bonkers. Well, you know, this is what I love because, you know, for the past couple of years, we've been doing this founder university concept. And one of the things we did is we have, we have.

had bubble and some other no-code platforms. And we have Zapier and Notion Coda, a bunch of these ways type forms, SurveyMonkey, et cetera. You can kind of stitch together a Frankenstein. Now, that's a little bit of a derogatory term, but you can stitch together a no-code prototype. And what's happened is we've gone from no-code prototypes

then to actually production products with Webflow or bubble. And now we're going to actually native apps with AI. What this does is it gives a group of non-developer, but product manager, designer types, the ability to actually make a prototype and throw it out there

and let somebody play with it, right? And we saw this like, you could build Angry Birds now. And, you know, if you play a video of Angry Birds to an LLM, could it make the code, do you think? Well, and you know what this ushers in? I think it could.

Like, Pong, you know, this is just, it's happening really quickly. And remember we were giving predictions? Like, it's happening even faster than I would have thought. I think we're going to enter an era where we're going to have a lot of apps instead of that you'll just create on the fly for what you need. And I think people are going to start doing that. Give an example. Well, like, and I'm going to do another demo in a second.

So I'm going to lead up to this. But like in the case of you want to do like a meme editor, right? And so you'll kind of just be like, instead of searching in the app store for the best one, you'll be like, like, hey, can you quickly give me an app where I can pull up memes and then, you know, edit them and then post them? And so instead of looking through the 50 or 100 meme editors there are, you'll just get one created for you on the fly for the thing that you're trying to do. Interesting. Yeah. So if you and I were going to go skiing, we use something called slopes, really cool app, where we track our runs. You could say, make me a slopes app. Now that has GPS.

That's a little more sophisticated, but we could actually. say, hey, we want an app to just track ourselves while we're a mountain biking around Lake Tahoe, make me an app, share it with my friend, and it'll be our own personal, you know, just between us. Summer app for, or summer activity app. Yeah. We're going to be skiing and, like, make us like a communicator for each other or something, just like a walkie talkie.

It's wild. Yeah. Well, yeah, think about this. Hey, we've got a group of 10 besties. Yeah. Make us an app that shares our location and what stores we're going. going to and just creates a running list and then deletes it after 10 days so people can see what stores we were going to.

Whatever. You know, like, we're a bunch of parents, you know, put us in a group and we can talk about our kids, but we want to track our kids and see where they're going and, you know, they're going to have a play date this weekend. It's going to be very bespoke. I think we're going to look at that. I love that. That's just kind of like how people use Google sheets when you think of it.

Like you pop up a spreadsheet. What do you use it for? People use spreadsheets for like planning a party, their schedule. you know, voting on trips. And you just use it for anything. Budgeting, recipes. People will use, you know, little Google sheets for anything. And I think that's what's going to happen here.

Yeah, you're going to talk to your AI, make some sort of app and just use it amongst friends and toss it in the garbage. Just like you might toss a Google sheet in the garbage. It's not using it anymore. All right. So. I give that an A plus, by the way. That was mind blowing.

Yeah. That was mind blowing. Yeah. Like an app with it. Now, I think one of the crazy thing is Apple won't approve that because they don't like, but I think it's a really, really interesting era we're going to see.

I give both of those in A plus. They're two different companies. Yeah, they are. Well, one's not even a company. One's just someone who created it like a developer. So it's not even available. So I'm just going to give them a shout out. No, no, no.

They have a GitHub that you can go and use it and download it yourself. It's for developers. Got it. Okay. Yeah. All right. Okay. Next one, which is also really interesting.

this one gives you like the UI. It doesn't implement it, but it gives you a really nice looking UI. So down here in the text box, what you can see here, there's a iPhone. And I said, design a restaurant reservation page,

a number of guest date, and special request. And I'm going to hit that. And basically, um, yeah, we knew this was coming. We talked about this last year.

And it just starts writing the code for this app. Yeah. It's just bonkers. I mean, and then I guess is the, what is the fidelity? be able to code. Have you looked at that? Yeah. Totally usable.

Totally usable. Totally usable. Yeah. And it's just loading up here, so I'm just doing it live. Okay, so you can see that. Here it goes. And there you go. And there it is. Make a reservation. Think about, we used to use,

I forgot the name of it, the app to just mock up. We use Figma. You could use InVision. Yeah, InVis. You can use balsamic. Balsamic. Balsamic was like the old school one. Look what's happened.

Like, you don't, and this is, you can download this. Yeah. Right? And there it is, right? I've just pulled up the code for those folks listening. That's nuts. Yeah. So let me ask you a question.

Somebody who ran a dev shop and you built lots of apps for large social networks, dating apps famously to Tinder when you had your dev shop in Canada. What's the future here? Well, I think it's super powerful. You know, so I'm talking to a couple of people who are reimagining that business right now. You can, you know, they came and wanted to speak and the business is this about reimagined. What used to take months will take days. What used to take different sets of resources can be collapsed into these tools.

And I think we're the same type of innovation we saw when the App Store came out. We're going to see like another kind of step function in innovation right now. Okay. So the App Store. Yeah. made the world into software buyers whereas previously

the average person bought one software package a year maybe putting video games aside people didn't buy apps all that they would go to Comp USA back in the day

and buy some software off of a package you might buy a browser you might download something off the internet freeware shareware but you got Microsoft office and you were kind of done and then you did cloud computing

you might buy two or three SaaS products maybe if you were a junkie you had 10 SaaS products, but call it five for the average, you know, executive maybe you bought. But then the app store came out and people had collections of dozens of apps on their phone. Probably the average phone has 50 apps on it, I would think, over, you know, if you had some number easily. Easily. Yeah, yeah. So now you think a step function, it could be 500, it could be 10 times a number of apps available, 100 times a number of apps available.

Well, I think what you're going to get is a combination of these two things that we just saw is like what you need. in that moment created for you. So mind-blowing. Right? Yeah. So, like, you know, I think it's a challenge to like an open table. Now, obviously they have a back-end integration, which is quite important. But, you know, think about the experience of a booking restaurants and like having people

pick. And so you could quickly just create something, send it around. Everyone picks a date. And then, you know, that feature is really hard for, you know, the company to build. People just put that together and maybe it connects to the open table backend and gets a restaurant book. So it sends it out. Everyone picks their date. It finds the common date. You know, like those tips splitting things.

Sure, sure. Sure. Foring that for more complicated use cases. Coordinating. It's like Callumley or whatever. You know, find us the date that works. Or you're the restaurant itself. And you say, you know what? I want to have my own booking app. I want to give that precedent over places that charge. So make me for, you know, my favorite restaurant in Tahoe. I want to get more people fighting for reservations there.

But they can make their own app for beep, beep. and then they could say, you know, hey, we want to have these three tables or the VIP tables. Don't include them. These three tables are for tasting. And if you want them, you have to do the tasting menu and then the rest are first come first serve

and build me a database of all of my best customers. And this is your product spec. What you just said, that's all that would be required for the product specification. It's just so mind-blowing. Because you think, like, there are some restaurants that probably create their own apps. Like Alina, we had the famous co-founder of it,

chef and everything. They have a huge wait list. I think they were the restaurant that in season two of the chef episode, I think it was seven or eight, the one where he goes and works in and polish his somewhere. It was based on that restaurant. You know, they actually built their own software for buying, because they had so many reservations.

They built their own software platform. They wound up selling it to Square, I think. Okay. To book their restaurant. They're, and I think actually even bid on reservations, but you had to, you had to buy them like sports tickets, basically, was his big innovation was,

okay. You know, like when you buy tickets to the warriors. It's not Dorcia, is it? Is it Dorcia? No, Dorcia is, is kind of inspired a little bit by that, but what Dorcia is doing,

which, um, you know, if I was younger, I would be using it all the time, but I don't really go out to eat as much as I used to. Um, they will let you bid on like,

really prime tables in prime markets. And you just agree to $500. I've seen $500 a person in Dorcia. Yeah. Like a pre-committed spend. Pre-committed spend. Yeah, yeah, yeah. And so it's not an auction.

It's just you're pre-committing your spend. And when you go to the restaurant, my understanding with Dorsey, is it like, it shows you in the app in real time as you water, you're like, oh, I'm going to get this two bottles of wine. This bottle of champagne, this bottle of wine. And that's $600. And we, you and I went on a date. It was a thousand dollar pre-commit.

Now it's like, you walk out. And you don't even have to pay a bill at the end. You just walk out. Yeah. And then whatever you have left, I don't know if you can just give that as a tip. I think it ends up as that, yeah. No, I think the restaurant takes it. You don't get it back.

I know you don't get it back. What I'm saying is like if I spent 600, I had 400 left, could I put that 400 towards the staff? I'm going to look into that. I've used it a few times. I haven't thought about that. You have used it. See, I would be fine with it because I'm a big tipper anyway. Yeah.

So if I really wanted to go to a restaurant, I was only in that city for that weekend. I kind of appreciate that as a little. So when you get nice wine, it's a no-brainer because you go past that. anyways, right? Yeah, see, I don't, I'm not a big drinker, so like, I'll have a social glass of, like, an ounce of one, but not really my jam. All right. Got any more demo? So, let's get one more demo in before we got a couple more, but I want to jump forward to one here.

Okay. So runway just launched this thing called Runway TV, and basically it's an AI generated channel. So everything in here, I have a little guide, and basically this entire service is AI generated. Okay, so they've made a cable system with cable channels that are all AI generated. Correct. They just replaced the entire corpus of cable television. Well, no live sports, but you can get anything else that's scripted. And they have like, you know, kind of here.

Yeah. Yeah. So let's go here, Gen 48, Film of the We Creative Partner Programs. I want to be careful here because I haven't pretty test. listed these, but here we go. Sci-Fi. Just another cat. Okay.

So you clicked on that from the channel guide. Yeah. We hit play on it. So it's like a channel guide. Now, just it comes up with what looks like a 1970 TV show of like a, looks like a detective or something in a 1970s car. Yeah. And so now it's just like full episodes of these things and it just made TV shows?

They're short. Right. Or it's just two minutes. Yeah. No, no, but no, they have a whole storyline. The audio's coming in. Maybe I got to reshare it. No, it's okay.

Don't worry about it. We won't play it because we don't want to get dinged with anything. But so anyway, it just basically plays a short. So it's just random AI shorts that were created. That's nuts. Yeah. What they all fit into a genre, that's the sci-fi genre, right? And so it's really nuts.

That's wild. And look, like, we're, again, we're in the early innings. Like, what does this mean for, you know? And, you know, we obviously seen, like, offset at ESPN and other places, but man, for content, and we just saw the writer strike end, this was pretty wild to have already this early on have like a, you know, a whole network, like a cable network. Yeah, and I'm doubtful that any of this is actually really that interesting or refined because

it's not like, doesn't have the human touch yet. Yeah. But given time, maybe some of these will actually be worth watching. Now imagine also. Well, you take your stories, WAPAD plus this, right? You take really great stories and combine it with this. Human written stories on WapTad and then move them over here. Yeah, because, you know, actually 50 Shades of Grey, my understanding was I met the woman who wrote that at a party years ago.

And somebody told me that the 50 Shades of Grey was originally Twilight fan fiction. And she changed the names of it to make like new characters. But it was kind of, I guess, in this romance. They would add a lot more romance to a, you know, fill in the blanks here. To, you know, something that didn't have as much role. romance in it. And so that's kind of interesting. I'm thinking about this and I'm just wondering about some IP licensing and some big IP

brands. It would be very interesting to me is to take Mike Tyson and have Mike Tyson fight Floyd Mayweather. And this is, I think we talked about this back in the day, but imagine if you took all of the Floyd Mayweather fights in history, you took all of the Mike Tyson fights in history. And then you said, what was their peak? Okay, Mike Tyson versus this person. was the peak, Floyd Mayweather versus this rose the peak.

Okay. Now, give me those two fighters. Put them in the ring. All these debates. And now the debates are over. Let's play. Floyd Mayweather versus Mike Tyson. I think Mike Tyson is going to get a 70% chance of absolutely just destroying him.

But I don't know because Floyd's defensive strategy might have tuckered him out. And then Floyd's a heavyweight. And heavyweights can't fight lightweights like that. Okay. So then what you say is, move Mike Tyson down to, same Mike Tyson strategy, move him down to middleweight, take Floyd, move him up to middleweight, show me, yeah, so just change your body types,

but keep their style. Wow. Yeah. That is incredible because now you could have people go across it and you'd say, hey, let's actually have Michael Jordan play LeBron in a game of one-on-one, and let's see what that's like. Yeah. Which I guess is what the NBA video games do to a lesser extent with machine learning. They try to proximate that.

What is the NBA game that everybody plays, the video game, where you make your own teams and you play? Oh, like 2K, NBA 2K. Like 2K. 2K. Yeah. I wonder how much AI or machine learning 2K is doing when they play, you know, teams. I understand you can play like, you know, whatever. And you can, yeah, and there's career modes and you can play, yeah, the legendary Bulls versus legendary Lakers and or whatever you want.

Or NBA live, I guess. Yeah. That's a big thing. The ratings are provided. by humans because there's a guy that everyone kind of, you know, kisses up to, I think is Ronnie 2K, that's his name. And he's the guy that ultimately is in charge of the ratings. And so a lot of players go to him when your rating is off. And they wait for ratings, get updated and all that. So it's a whole kind of sub-ecosystem. Yeah. I'm going to give that a C-plus right now.

Okay. It's like an interesting proof of concept. I'll give it a C. I'm not even giving it a Plus, interesting proof of concept. It doesn't look real. It doesn't pass the Uncanny Valley. I think it's going to need a lot more work. But I think it shows some promise. And it's interesting.

It's tied into storymaking. So I'm kind of in the same place as you, like, you know, C++. But look, kudos to them. And now you can see why they have such a huge evaluation because they could become like the next network. Okay, this next one I've actually been using quite a bit. So stability AI, these are the guys. behind a bunch of these really good models that we see for generative image, like generation. But they have created a whole bunch of new tools.

And so they've got tools here that can take an image and uncrop it for you. I love this. You can basically start a doodle. You can clean up an image. You can remove a background. So I'm not going to go through all these. You can upscale an image. This has been really useful.

And what's the URL for this? It's just clipdrop. dot co. Okay, and who's this by this by stability? Stability, yeah. So this is like things that you would pay a Photoshop graphic designer to do for you. You'd give them a picture of yourself and say, I want this to, you know, have me take the background of my house out, my backyard, and put me on a beach. Or just take the background out and, you know, put a color behind it, replace the background.

So I'll do one here. Like just, and I was, you know, this is a classic, uh, guy here and so you know this guy. That's the weird looking guy with a mug who's got like the grin. Yeah. And he's like a forced grin. And so I just did that. Yeah.

Put him on a. The stock image. I forgot what they call that guy. I don't know. He's got like he's got the white beard and like the bald head wearing the blue shirt and he's kind of doing work. There's a name for this meme, but the guy looks like he's constipated or something. Well, he's always buying like the thing late.

Like, oh, I just bought, you know. He's like old dude. laggard. He's like, I'm going to call him like Whitebeard. So now what you're doing here is, but were you also making memes move around on the internet this weekend? Oh, that's a different one. I can show you that too if you want to see that. Okay. I thought that was the most interesting demo. So this one I'm going to give B plus to this collection is incredible. Yeah, it's incredible. Yeah. And honestly,

like, I'm using it all the time for whether I'm creating decks or something like that. And there's a lot of tools in here. So I recommend everyone to go through it, like removing backgrounds, reimagining pictures, uncropping something. Upscaling is, this is a real pain in the butt. Like upscaling pictures has always been... By the way, the name of that guy? Hide the pain, Harold.

Just for people... Oh, hide the pain, Harold. Yeah. All right. And, okay, so then, so that was stability. So this is a model from stability, but they haven't made it available on their site yet. And so I'm running this on Replicate.

So Replicate is a place where you can take models and they'll run them for you. And basically, you know, they've got farms of GPUs available. So in this case, instead of the memes, because I did animate some memes, but something more important is I took this very famous picture. Raising of the flag at Eo GMA. Yeah. And basically, you know, you run it and we can do here. Did you tell it what to do?

Did you tell it to raise the flag? No, no, no. That's the beauty. You tell it nothing. Oh, you tell it nothing. You give it an image. It has a few things that you can modify here. And basically, these are, how long you want it to be?

How many frames per second? Motion bucket. These are some very, this is, again, this is, we're running it live on the GPUs ourselves, right? And basically, you click run. And this may not go super fast because it depends if we have access to it. It goes, it started. And so this is, this is actually what happens when, you know, AI runs in the background. So you can see this.

doing the work on the the GPUs here directly. Yeah. What is this costing you to do? It's costing you 10 cents to do this or once a end? You can go look at that. Yeah. So once this is done,

or maybe I'll just open this, maybe in a new tab or something. So you can kind of, I don't know if it's doing them. Yeah, so it just goes like second by second. So they take about 60 seconds. Right. And the pricing for that model,

which is NVDA, N40 large, is, you know, basically Yeah What is that? Two bucks an hour So it's like I don't know

One one hundredth of You know A cent or something Two bucks an hour Yeah So that's nothing Yeah It's like three cents a minute

Yeah Yeah exactly So that's what it takes And you know This thing will run here I don't know if this one will come out as good As the one I just shared

Because sometimes you have to tweak Some of the settings Got it And it's amazing The idea that you could Take a meme So we just did that live And this one, the flag is waving a bit different than the one I had posted.

Yeah. Yeah. Crazy. Absolutely nuts. Yeah. I mean, and this is, again, we're still in, we just hit year one. So if we just consider this year one, this is nuts. And what we're seeing here is entire industries.

To recap for the episode, we started with what a researcher does for 40 bucks an hour. And we showed that that can be done at a 90%. 80 to 90% of the work will be done by it, is already being done by AI. Then we showed making a basic video game or a calculator for a developer. And again, I think 80, 90% of the work's done. You probably have to massage the last 20%. Then we showed video and images. And in the case of the static images, we saw maybe 70% of what an image processing person

who gets paid 20, 30 bucks offshore could do, right? Cleaning a background or whatever. and maybe 40 or 50 bucks here in the U.S. And then we did video with that network show, and that would be, yeah, you know, I don't know. That's a lot of creatives doing a lot of different things, but it's not ready for prime time, so it's probably doing 10% of what needs to be 20%.

Yeah. We could check back in a year from now. 10% of millions of dollars. Yes. And so the question is, when can it do a Pixar movie? So I'm going to put that out there. When could it make something? And let's even go with a short.

When can it make the short at the beginning of a Pixar movie? You know how they have the shorts at the beginning of a Pixar movie? Summer 2024, I'll put it out there. We'll see a full movie Pixar quality created by AI. Wait, wait, you're saying in six months. You're saying August 1st, 2024, I'm taking it over. It might not be in the theaters because of the system. But you're saying indistinguishable from the movie X or Rat Tatooey.

We're talking about just the fidelity of the video where the storytelling The animation, the voice over, all that. Everything. Okay, so we're going to complete. Hold on. I'm going to let you reset the bet here. Yeah. I'm going to reset the bet here because it's going to be a real bet.

Yeah. I want to just set it properly. Indistinguishable from Ratatoui in my favorite film. I see Ratatatoui and I see this new one and I cannot tell the difference. I watch five minutes of it where we show somebody five minutes of Ratatoui versus five minutes of this, the best five minutes. And the consumer, it passes the Uncadney Valley. They believe this is the next Pixar film. They believe it's the next Pixar film

Yes Will be the bet I think that's three years from now Because you gotta do storytelling too And I gotta do voiceover on the app Voices can be copied as we know right We've seen that through the AI tools

And it's gonna go on like YouTube Because the I'm gonna say 2026 I'm gonna go 2026 Oh my God Jake In January 1st 2026 I would set the line at So I'm taking the over I'm taking the over

It's easy win for me Yeah. And if you're the person making it, go ahead and prove it to us. Make us a film. Make us Rat Tattooey Part 2. You said the short. The short that goes before, though. Okay, you make a Rat Tattooey short.

So it's son of Ratatatoui, daughter of Ratatoui. Ratatoui makes shake shack, you know, Ratatoui does a cooking show, whatever your jam is. The AI has to come up with the story, the images, the voiceover, the whole package, 10 minutes, short. of Ratatoui and it's better or it's indistinguishable from any other short that was done by Pixar

at this beginning of something okay there we go we got it we got it you say okay and if you do that and you make it you can come on the show

did we give um we gave our letter grades to everything producers we didn't give a no what do we give a letter to clip

we didn't give them a letter for the meme for the meme one not the meme one I'm giving a B plus B plus okay yeah

yeah Because it's, listen, is it perfect yet? No, it looks like it's AI generated. But I'll give it a B plus because I think next year it will not look. It'll look like that was actually a video clip taken from a camera. From that scene. Yeah. From that scene.

Like they just discovered the video. Like it would, I think next year at this time, you show it, you do the same exercise. That would look like, oh, it's the lost footage. And you just say, that's a lost footage. That's what happens with one of the memes I posted JCal. And someone said, well, this is already a video. And I'm like, yeah, that's the. point. I had animated some memes, and that was a point that it does such a good job that

people couldn't figure. And this one had a little bit of wonkingness in it, but wonkiness in it, but people can't figure out that that's happened. Let me see if I can pull this up here. This is one of the more, I think this was the most famous meme of 2020. This, you know, the two women screaming and it's a cat ignoring them. It's like some reality show books. Yeah. And they're actually now screaming. People say, well, that's the actual video. And it's not. Yeah, it's not. It was just generating from the woman's finger is like her hand is super weird.

And her, and her face here is a little bit weird. But, you know, yeah, we're, I almost believe it. That'll be fixed. That's what I'm saying.

This thing is in beta still. Amazing. Okay. So I give it B plus. I give it to me and generate people. But would you give it? I give it, yeah,

just needs to be go to production. So B plus, but it's, you know, with a caveat that we know it's still in beta. Oh, here we go. Yeah.

That guy is called hide the pain, Harold. Hide the pain Harold. He just leans forward and like spills his drink or something on his keyboard. It's like very weird. He feels the pain because of, you know, what's happened. So. Yes, his life is horrible. He's just slave to his. He's lost all his money buying the wrong crypto. He literally bought the wrong crypto and on his windows like 98 laptop. Top. Yeah. From like HP. But he's, but he's happy, but he's sad because his retirement savings gone.

and now he's going to have to go work at TGI Fridays and put flare on his things at the age of 70. Maybe he'll welcome her. American tragedy. Walmart. Welcome her at Walmart. I mean, literally, isn't it amazing?

You were deep in the crypto game, and now you're deep in the AI game. And more happens in a week in AI than happened in 10 years of crypto. Yeah. I think it's safe to say. A month, maybe. My retrospective of it is false. So one, like I'm a still big believer in the underlying technology.

That's why we spent time, you know, building there and being in the space. But when you can allow people to financialize things, which inherently crypto was, because you could start trading them right away without their being value, that was the problem, right? And so it was a combination of like there was really great ideas and there still are great ideas being built every single day. But when you can financialize it, meaning people can trade on it before it's gotten traction. So imagine all these things that we've been demoing, you could go and buy it right away. Like there was a token associated with it. That would be chaos and AI.

The good news is, like, AI and crypto are not intersected in that way. So you can't go off one of these demos and buy it. I like the idea of a more fluid funding environment for projects. But there has to be some constraint. And so it would be very easy to just say, I fell in love with crypto for that reason. And when I saw the abuse, I fell out of love. There's a very simple constraint here. Any adult in the United States can spend $1,000 as an investment in a project 10 times a year.

So there's an exception. Every American can blow $10,000 a year, $1,000 at a time per deal. We just call it the Angel Investment Exception. Angel Exception Act. By the way, if you want to do it and you want to go above that and you want to go to a hundred thousand a year and ten bets of 10K, you have to take a course and you have to get a certification like a driver's license, gun license,

barber's license, whatever, somewhere in that range of you have to do five hours of coursework and answer a 50 question test. Just so you know what diversity is and you know, that 90% of startups return zero dollars, yada, yada, yada, that would be actually great for society

because you'd have people be able to transition from lower class to middle class to upper class. It would give upward mobility because you have equity upside. And you'd have some protection because you could only invest so much. You can't take your life savings and just blow it like you can on sports betting. Yep. That would be a fair, I think, way to do this. Now, there's some people, you know, apology might believe that you could just do whatever you want.

And you can't trade the thing. Like you can buy into it, but. Like in the similar way, if you invest in a startup, you can't go in like secondary markets do get created, but they get created much later. People don't start going and trading on the angel investment you just did, right? One, the investors really don't allow for it. I would be okay with trading it if it still has those rules. So you could trade it to another party, but they can only do 10 deals per year. So it's throttled.

In other words, like I bought into, you know, the, the, the, the, Kong app for a thousand bucks. I could sell it anytime I want, but if I sell it, that's one of my 10 transactions for the year. So I get 10 transactions per year, or I could sell up to 10, I can buy up to 10, something like that, where you have some throttling of the velocity of this? Because like you said, when it's unlimited velocity, it just becomes chaos. Well, speculation and chaos, yeah, which is unfortunately what took over. Human nature.

I mean, I respect people who are libertarians and say it's my. money, I should be able to take any risk I want with it. I actually respect that. And we do have that in the United States when it comes to gambling. Now, other places where there's gambling, like my understanding in Korea, when I played poker at Walker Hill Casino in Seoul, there was all non-Koreanians. And I was like, why are there no Koreans here? And they're no Koreans aren't allowed to gamble in Seoul. They weren't allowed to gamble there. Yeah. There was a ban. And Koreans, then, what was one of the most popular places for crypto? Korea

because humans want to gamble and they actually remember they tried to ban it in Korea and then they relented because people wanted it so badly and so human nature to want to gamble we just have to have

some thoughtfulness around it yeah all right there's been another amazing episode follow at Sendip on Twitter X X.com slash Zendip X.com slash jason

the site formerly known as Twitter see you all next time bye-bye