This Week in Startups - AI demos: Figma’s FigJam, AI drawing, dancing avatars, and more! | E1857

Episode Date: December 5, 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. Brave. If you’re building AI and search-based applications, train your models with the Brave Search API. Get started for free at brave.com/jason Lemon.io - Hire pre-vetted remote developers, get 15% off your first 4 weeks of developer time at https://Lemon.io/twist Today’s show: Sunny Madra joins Jason to demo Figma FigJam (2:08), a product that generates images from real-time drawing (31:05), and more while discussing the implications of IP theft writ large. * Timestamps: (0:00) Sunny Madra joins Jason (2:08) Sunny demos Figma’s FigJam (9:43) Squarespace - Use offer code TWIST to save 10% off your first purchase of a website or domain at https://Squarespace.com/twist (10:40) Figma FigJam demo continued (16:02) Jason shares his AI generated images and discusses how companies navigate copyright issues (24:58) Brave - Try the Brave Search API at http://brave.com/jason (26:12) Feedback and thoughts for FigJam (31:05) Sunny demos Fal, an AI that gives generated images from real time drawing (40:28) Lemon.io - Get 15% off your first 4 weeks of developer time at https://Lemon.io/twist (41:49) Sunny demos the Real-Time Latent Consistency Model (47:01) Sunny demos Alibaba’s Animate Anyone * Subscribe to This Week in Startups on Apple: https://rb.gy/v19fcp * Check out FigJam: https://help.figma.com/hc/en-us/articles/1500004362321-Guide-to-FigJam Check out Fal: https://www.fal.ai/dynamic Check out the hugging face video demo: https://huggingface.co/spaces/radames/Real-Time-Latent-Consistency-Model Check out Animate Anyone: https://www.youtube.com/watch?v=8PCn5hLKNu4 * Follow Sunny: https://twitter.com/sundeep * Follow Jason: Twitter: https://twitter.com/jason Instagram: https://www.instagram.com/jason LinkedIn: https://www.linkedin.com/in/jasoncalacanis * 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 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

Transcript
Discussion (0)
Starting point is 00:00:00 Why does the scale of the theft matter? Well, because you have to look at it and say, are they infringing on Disney's ability to do something in the world? So let's say, you and I decided we were going to create an event where we had a Jedi night party. Okay, great. Happy birthday. Jake Allen is a Jedi night party.
Starting point is 00:00:19 So that's the theme. Great. Star Wars theme. Everybody has a great time. Now, imagine we said, wow, everybody has such a great time at this. Let's make it a permanent installation at Calhalla. We'll get a building. And then we'll charge 50 bucks to come.
Starting point is 00:00:30 like, well, that sounds like Star Wars Town at Disney World. Yeah. Disney World, yeah. And so that's where the scale of something, the monetization of something comes in. 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. Brave.
Starting point is 00:00:57 If you're building AI and search-based apps, applications, train your models with the Brave Search API. Get started for free at brave.com slash Jason and lemon.io. Need to speed up your product development without draining your budget. Hire vetted engineers from Europe at lemon.io. Go to lemon.com slash twist to get 15% off for the first four weeks. All right, everybody. It's Monday. So it's time for our AI demos. This is what we do. We demo or every Monday. And with me again, Sondi, he's at Sundeepe on the Twitter slash X.
Starting point is 00:01:36 He's the co-founder of definitive intelligence. You can go check out definitive.io. They do all kinds of AI, data analysis, public data, private data, work with a lot of big companies out there. And I'm an investor. No hallucination.
Starting point is 00:01:49 All the good stuff. All the good stuff. Full disclosure, I put a little slicey poo. I'm an investor in Sunny's product, and we're close personal friends. We play poker every week. hang out and we geek out and Mondays are for geeking out. Let's geek out. What do you got first on the agenda for today?
Starting point is 00:02:08 Very excited about the first one because it's, we're going to be, you know, we've been saying this like AI all the way down. And we've been talking about doing a little this week in startups AI summit. Yes. It'd be a lot of fun to do a live event where it's just all demos of cool stuff. Yeah. And exactly. just, you know, kind of get it all going. And so I decided to use this, you know, so Figma
Starting point is 00:02:35 released this new thing called Figma Jam. And what I did here in all, Figma Jam. And so what you can go in here and I'll just, are going to kind of cut and paste here. So I'm going to say, I need a work full for planning the upcoming this week in Startups AI conference. And what it'll do in Figma right there is it was going to give you everything you need to get the this conference going. And yeah, I just did this live right in front of you. And you can see here as like, who's there? What are the goals in this particular meeting?
Starting point is 00:03:10 What's the speaker lineup? Conference venue, accommodations, transportation. And so it's just, it's really incredible. They can do this for any type of, you know, workflow or brainstorming or product review or anything you need. And then you can add to it. So did it miss something? because you've done a lot of conferences,
Starting point is 00:03:30 J-Cal. So did it miss something that you'd have, like as you were going to kind of sit down and do the work full planning for the start of the conference? Venue. I guess venue would be under, it says venue, conference venue right here.
Starting point is 00:03:44 Yeah, I can zoom in. So, yeah, and then you would need to do the format. So the editorial format, like you have speakers, but then there's the editorial format. Like,
Starting point is 00:03:56 are we going to do fireside chats? Are we going to do demo? So what's the schedule and the lineup, basically? So I see there's a speaker lineup there. But it didn't go drill down to that second thing. So I would be thinking, you know, hey, what is the most important format for us for? Is it going to be keynotes? Is it going to be panels, breakout sessions, whatever?
Starting point is 00:04:19 So wait, how did it do with that? The speaker line. So we said, can you add a section to cite on the editorial format? So it said keynote speakers, channel session, breakout sessions. So it keeps iterating with you. So the first one didn't have that. And so then it kind of basically...
Starting point is 00:04:35 You gave it another prompt. So just so people who aren't... For people who are watching, you're on YouTube.com slash this weekend. Just type in this week and Startup. She'll find us on YouTube. And you can see as demoing this. Now, if you're on audio, you have a generate box where you type in what you're looking for.
Starting point is 00:04:53 So the first prompt was, I need a workflow for planning and upcoming This Week in startups AI conference. And it gave us, would you say this as a con bond board or it's just a, I guess Figma designs like It's like a canvas combo board.
Starting point is 00:05:10 Or a whiteboard. It's kind of like a whiteboard session here. So kind of put the agenda, etc. But you said, can you add a section to decide the editorial format? And when you did that, it changed the speaker lineup
Starting point is 00:05:23 to a more granular view. of an event, which was keynote speakers, panel discussions, breakout sessions, and workshops. So this is really good work. And then on the logistics, it has accommodations, transportation, technical requirements, venue. And then it added, what's this final section on the right? Well, this is where it's interactive, right?
Starting point is 00:05:48 So it said, in order in this meeting, because we're going to use this board, how are we going to decide the editorial format? And so it kind of gave us some ideas on how to, how to think through the editorial format as well. So here we have live streaming as option one, recorded videos option two, written summaries and interactive slides.
Starting point is 00:06:08 So this is like a hodgepodge of ideas. It's actually kind of helpful because if you were doing your first time conference, and I was a conference producer, which I've been doing for 30 years, these are the things that I would tell you, like, are you going to live stream it? Are you going to charge for the live stream? Are you going to record the videos?
Starting point is 00:06:24 because if you do an event and you want a live stream, well, now you've got a cohort of issues you need to deal with. Can this place have a stable internet connection with an Ethernet cable? Do you have a high-powered machine dedicated to that? Because now you need a laptop to just do that. And now this isn't, when we're going through this board, it's not listening to us right now. So this isn't...
Starting point is 00:06:52 It's not listening to us. But then I did another one. one, I said, revenue generation ideas. And it gave us, right? Because you just said that. And so for those listening, I just typed that and said, we need to brainstorm monetization. So I said offer premium tickets, create sponsorships, post a virtual marketplace, yada, yada, yada, right? Right.
Starting point is 00:07:10 Amortizing space on the conference website. I mean, this is so, you know, think about this, JCal, right? Yeah, you know, if you had an event planner, what you just did, I would say, is the first half of what an event planner would do. a full 50% of what an event planner would do is they would say, here are the things you need to address. You know, here are the top level things, just like, you know, creating a spec, right?
Starting point is 00:07:41 And so there are worksheets that exist on the internet probably. There are probably some articles about this. And then what we could say is, how would you market this, right? So you could add that as a prompt. how would we market the event or how should we market the event? And then you could do, hey, we need a budget. And so when we use the term AI all the way down, what we meant was, AI will be, in fact, every single thing you're doing.
Starting point is 00:08:07 And so what would be very nice here is if while we were talking, Figma was listening. So we didn't have to type in the prompts. It would just be kind of another participant in the meeting. So I'll call this. type of AI, prompt AI. What I'm looking for is participatory AI. So this could be a blog post
Starting point is 00:08:32 for me. What I mean by participatory AI is the AI is the guide on the side. They're participating, listening, interacting with you. So what I would like here is some combination of the talking mode. What we refer to in the industry as
Starting point is 00:08:49 her mode from the movie her, where Joaquin Phoenix is talking to his AI love interest. I would love to have an AI here that would be in collaborative mode or participatory mode AI where they were listening to our conversation saying, sorry to interrupt, Sundip and JCal, I've added a section here for marketing. And these are the top marketing concepts. So right here. It should have listening in mode.
Starting point is 00:09:22 It should be listening mode. You turn it on as we're having our Zoom or whatever format we're having the meeting in, ideally in IRL. It's basically just pulling it all in and doing this real time and this, this con bond board, this figma jam board is iterating in real. Yeah, whiteboard. That would be really cool. If your landing page looks terrible.
Starting point is 00:09:46 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,
Starting point is 00:10:02 and sell anything you want. You know Squarespace's 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.
Starting point is 00:10:29 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. And if the whiteboard, right now, if I said to the whiteboard, we're not doing a live stream, it just erased the live stream and just took that out and said, you know, It's going to be a hundred. We already actually have a location. We're doing this at the Microsoft or the Oracle Conference Center, and they gave it to us for free in exchange for having a banner and a logo on the website.
Starting point is 00:11:03 And it would put that as bullet points, action items. Don't forget to put this on the, so we'd have a to-do list. Okay, I've added to the to do list to add Oracle's message to the website and have a banner printed for this event. And I've budgeted $0 in the budgeted $0 in the budget. budget for a venue location and do you know how many seats it's going to be? Oh yeah, they can accommodate up to 250. So our max capacity is 250. Great. Yeah. Do you have any existing? And then it would say, well, who's going to be in the 250? How many paid tickets? How many speakers?
Starting point is 00:11:34 Do you think how many days? So it should be asking us questions to kind of get through this checkbox. What it's doing right now, I believe, is it's index the world's information like Google did. It's taken some web crawler. Some people out there wrote. how to throw a conference medium posts or or answers or Reddit threads or wherever or blog lists
Starting point is 00:11:57 spreadsheets or something. Yeah. And it's pulling all those all that collateral. It's scraped. And now it's regurgitating it back to you in real time in this format. So it's basically stealing content,
Starting point is 00:12:09 liberating content. And then flaming it as its own and posting it up here for us to work with. What I would like it to do is the next steps. make a to-do list, make action items, and then actually start doing those to-do items. So it's a, would you like me to make a logo? So branding.
Starting point is 00:12:28 We should have branding and naming. So if you put into Figma Jam, come up with some names for an AI demo day or a showcase and come up with a logo. I've been playing with Dolly. You probably saw me like doing an all-in logo with hearts. Yeah, I saw that, yeah. Over the weekend.
Starting point is 00:12:45 They literally did that in 10 seconds. Now, you can't do anything. I also did one over the weekend, where I said, make me a bulldog Jedi Knight, make me a bulldog version of that operator. And I was trying to see if I could get chat GPT's Dolly on mobile, number one, to make the images. But number two, to make images that they did not have the rights to that were derivative works that were based on IP. This kind of takes us into, well, there's no way to stop this. This did it on open crawl, but there is a conference producer
Starting point is 00:13:19 whose work this is. I don't know if they're going to be able to find out who actually did this, Sunny. Like the model of whose work we're drafting off of here. You've got to remember, it's not the whole thing, right? Because it could be one, one hundredth is from this website
Starting point is 00:13:38 and another 100th is from this, you know, PDF and another, because it's just pulled it, it's pulled it together from a lot of different pieces. It's not like a holistic, you know. So it's death by a thousand cuts. They've stolen a penny from a thousand people's bank accounts for a million people's bank accounts. And here they are. Now they got thousands of dollars, but nobody misses the penny.
Starting point is 00:14:03 Yeah, I wouldn't say stolen because it's sort of the way we operate. It's the way we learn, right? And the way we create. We've read a bunch of books. We've had a bunch of experiences in real life. and then we stitch those together and it turns into something new and creative for us. It's not plagiarism.
Starting point is 00:14:21 If you've taken enough source material that it's not a large percentage of anyone, it's a collection of learning. Can't argue with that. It's a hard thing to argue from with because it is wholesale. If you look at the end result, if I read five books on how to do a startup
Starting point is 00:14:42 and I did a startup, would I be stealing that IP? And the intent of that person was to have you use it. Now, if I made a bunch of templates on the web and I said, you know, these are my templates. I make money from advertising on this. And these are my conference templates and my PDFs on how to do it. And I sell an advanced one. And, you know, I put a little bit out there to get people to the events.
Starting point is 00:15:03 But now if you ingested that, you now have created a derivative work based on my work. And if this all came from, you know, a smaller subsection of people, you know, that it would feel like stealing. So this is where fairness comes into play. Does this feel fair or unfair? If you're the conference producer who made a conference website to try to attract people to your conference website and the AI is putting it out there, this feels profoundly unfair.
Starting point is 00:15:29 Just like you roll your eyes, you don't think so. Yeah, because if it only had one piece of information, then to your point, you could feel like that's plagiarism. But because these things have access to most of the internet, And let's, you know, and I think in your, in your David Luan interview from ADEP, he also called out that they licensed a bunch of material. So what if they went to a bunch of folks and paid them and said, Oh, I feel great about it.
Starting point is 00:15:56 I feel great about it. If you're licensing it and the person gets 500 bucks a year, you need to create a derivative work. I feel great about it. Okay. So this is what I did this weekend. So check this out. Oh.
Starting point is 00:16:07 And these are beautiful images. And so I said, make a Jedi bulldog. And it made it. Yeah. Great. Now, the next image I did. So now look at this interaction. This is very interesting.
Starting point is 00:16:19 I want to get your response to the Sipfler. I said, now make a cat version of Darth Rader. Dolly said to me, I'm unable to generate images based on your request due to our content policy. No link to it, no explanation of what the content policy is. If you have any other ideas or concepts you'd like to explore, feel free. I said, make me a Sithor cat. Boom. Here's an image of a Sithlor cat.
Starting point is 00:16:42 So it must have, in its database, that Darth Vader is protected IP. Correct. Don't let people make Darth Vader. Just like it doesn't do real images. Like if I said, make something based on Jason Calacadis as a Jedi, it won't do it. Yeah. But it would make that. Then I asked it, check this one out.
Starting point is 00:17:00 The next thing I asked ChatGPT to do was I moved over from Dolly to chat GPT. And in that same thread, I asked it to make me a sequel to The Shining. That's my very next tweet. And it did an amazing job. I said, write a short story about the characters from The Shining. The Shining. And it was like, certainly, here's a short story featuring the characters from The Shining. So it knows the Shining is protected IP and it just created a derivative work on this. After the harrowing events at the Overlook Hotel, Wendy and Danny Torrance sought a new beginning. They moved to a small,
Starting point is 00:17:32 peaceful town in Maine, far from the nightmarish memories of past. The town's people were warm and welcoming, intrigued by the newcomers, but respectful of their privacy. And I was like, wait a second. Where does this collection land on, you know, in our debate, me being very sensitive to IP and creators, in this case, Stephen King, George Lucas Disney and their IP. And in your case, hey, it's learning from the internet, you know, well, there's the fact that this is protected IP, like explicitly protected high profile IP that has proven value in the world, change your position. What's your position on this? Yeah, so these to me are distinctly different than the example we were putting in Figma Jam, right? Because the process or method or the template to create a conference, there could be, it's hard to nail it down to one particular one. But a Bulldog Jedi Knight is very clearly, you know, there's no other Jedi Knights out there because of the copyright and there's no other Darth Vader's.
Starting point is 00:18:36 And there's not even other Sith Lords, right? And so I feel like that, you know, I don't think I'm kind of having two different stances, but I think in that case, it does feel like a violation. Now, what I recall, and I want you to chime in here, because I'm not really a copyrights, you know, person. I don't really understand it well. And I think this is allowed in music as well. Someone is allowed to recreate something.
Starting point is 00:19:01 Like you can do a drawing of a Jedi Bulldog, Jedi Night Bulldog. Non-commercial use, non-commercial use. There's a fair use doctrine. Yeah. And so if you do things that are non-commercial, now Chad GPT is a commercial product. I'm paying 20 bucks for it. So Chad CPD is making $240 a year to allow me to make Disney characters or Disney inspire characters. So that's where we get in a rub here.
Starting point is 00:19:27 Now, if you have an open source project that nobody really is responsible for, it's like, who do you sue? There's no target there. So that falls out of it. Now, IP and fair use and the laws here in the United States are very different than many around the world. But people tend to follow our lead. Yep. And, you know, the truth is that they don't make bright lines. Why don't they make bright lines?
Starting point is 00:19:51 Well, because there are some uses that are clearly, like if a child draws a Jedi night of their bulldog, you're not going to have the police come and arrest them. And Disney's totally cool with that. But once you introduce commercial benefit, That's where you have this doctrine of that was the the original IP owner should have the right to that opportunity. So if I am Star Wars,
Starting point is 00:20:19 they should have a Disney Plus AI that allows you to do this. And then if you want to produce it in the world, it would put at the bottom, created by Disney Plus. So this should be built into Disney Plus. You can make any characters you want. You can put your face on it,
Starting point is 00:20:33 my face on it, make birthday cards and you can go print them. So all of this should happen there. And Etsy is an example of people creating bespoke, single-use products, etc. And when they do stuff, yeah, they can get in trouble for this. And so there is a constant tug of war with Disney going to them and saying, listen, you can't make Jedi robes or whatever or they'll let people do it. And it's up to the IP owner.
Starting point is 00:21:00 But I did buy a Jedi outfit because I went to a samurai party. once. And I said, I thought it'd be fun to come as, because I know the Jedi were inspired by the samurai. Yeah. So I bought a samurai, a Jedi outfit, an Obi-Wan Jedi outfit. And yeah, so you can buy all this stuff on Etsy. Yeah. And just for those listening, we're just scrolling through just Jedi night merchandise on Etsy. Now, you could be an artist and you could make a pop culture reference like Andy Warhol did of a Campbell's Soup thing. And, you know, they they don't have recourse. But if you were to do it at scale,
Starting point is 00:21:37 they might have recourse. Yeah. These are very refined. Are you saying that Disney doesn't care because these are individuals that are, you know, maybe selling 10 or 20 T-shirts or 100 at most? But it, because Open AI is a now a,
Starting point is 00:21:51 you know, a $100 billion company that is distinctly different. The scale of the theft matters. Why does the scale of the theft matter? Well, because you have to look at it and say, are they infringing on Disney's ability to do something in the world. So let's say, you and I decided we were going to create an event where we had a
Starting point is 00:22:12 Jedi night party. Okay, great. Happy birthday. Jake Al, here's a Jedi night party. So that's the theme. Great. Star Wars theme. Everybody has a great time. Now, imagine we said, wow, everybody has such a great time at this. Let's make it a permanent installation at Calhalla. We'll get a building, and then we'll charge 50 bucks to come. And it's like, well, that sounds like Star Wars town at Disney World. Disney World, yeah. And so that's where the skis scale of something, the monetization of something comes in. So when you look at Figma's, what do they call
Starting point is 00:22:40 Figma Jam? Figma Jam, yeah. Good name. I give the Figma team, a lot of credit, I like the name. They're using a language model to go find best practices. There might be 10,000 documents of best practices out there. It doesn't feel like any one person is having
Starting point is 00:22:57 too much stolen from them, you know? But, you know, this is where like the images in Dolly or I guess it was stable diffusion when it had the Getty logo on it and they indexed on that. This is where the problems come in. It just depends on you need to have somebody who feels exploited in order to have copyright lawsuits happen. And so that's why a lot of these go back to Disney because Disney is a very old company
Starting point is 00:23:24 and the rights to Disney's characters. We're going to fall into the public domain and Mickey Mouse, a lot of these laws literally revolve around protecting Mickey Mouse. It's going to be very interesting. It's tough for me because I like to see the advancement, but I also respect copyright holders. So in the case of the Bulldogs, and I see the difference between sort of the Etsy use case
Starting point is 00:23:47 and then your individual use case as well. And the shining one is fan fiction. So now if you were to do fan fiction, you can do that. It's no big deal. Now, if you try to sell the fan fiction, it turns out 50 Shades of Grey, which became a huge movie franchise
Starting point is 00:24:05 with a lot of money. It started, my understanding is, that's Twilight Fan Fiction. That was Twilight Erotica. So there is a great company called Wapad. Wapad, Canadian Company. Canadian Company.
Starting point is 00:24:17 It's done very well. And they help people monetize, monetize their fan fiction. And, you know, Star Trek, I think, took a heavy hand. Star Wars took a, we'll give you the sounds
Starting point is 00:24:30 to make a lightsaber. video and we'll give you the collateral and teach you how to do it. But you can't commercialize it. So that's like there's a kid who does Star Wars theory and he did a movie about Darth Vader. He's doing a second part on it. Wow. And as long as he doesn't sell the movie, he can raise money as donations and make the movie, but as long as he doesn't sell it, Disney has given people the okay. Wow. Are you building the next great AI product? Well, if so, you know how expensive APIs can be for their model training data. Training AI is pricey. That's a fact. We all know it. So you have to try the Brave Search API. Yes, I am talking about Brave, the privacy browser that I am obsessed with. Braves browser has 65 million users. Think about how much data that drives for Brave Search, which is the only global scale independent search index outside of big tech. And that index is available to anyone with the Brave Search API. The Brave Search API can power your chat, bots and train your models, inform answers to real-time queries, and it will serve images,
Starting point is 00:25:36 web results, and even rich text snippets. The Brave Search API features an easy, intuitive data structure, and its data is populated by real human interaction, not web crawlers, all for a fraction of the cost of the major players. It's free for up to 2,000 queries per month, with paid plans for as little as a $3 CPM. That's cost per thousand. So if you're building a NextGen AI app or chatbots, you've got to try the Brave Search API. Get started today. Brave.com slash Jason. Oh, Jason's brave.
Starting point is 00:26:05 I like it. Brave.com slash Jason and get the browser while you add it. It is awesome. It's also got a VPN built in. That's pretty cool. Let's get some grades going here.
Starting point is 00:26:14 Figma Jam. Figma Jam. Yeah. I'm going to tell you, I love this a lot. I could see myself using it. This feels to me like a way for non-designers to want to use Figma more.
Starting point is 00:26:27 I love whiteboarding software like this. I've used it before. It's very creative. It's good for remote teams. The potentials here. Yeah. And I like this kind of being built into project management software. I like it being built into any number of pieces of software.
Starting point is 00:26:49 And if you want to impress your boss, if you're a young person or young in your career, like in a new position, like you're new to a position, if you worked for me and said, hey, I know we're doing the Angel Summit again. We're going to call it liquidity this time. I'm changing the branding of Angel Summit to liquidity. So I want to make it go beyond Angels because I have all these LPs and hedge fund people coming now. So it's going to be liquidity. The liquidity summit. Liquidity conference. And I'm going to probably do a podcast
Starting point is 00:27:19 called liquidity for like 10 episodes where I put together a little bit of a roundtable or it may be solo. I haven't decided yet, but this liquidity podcast I want to do, I want to have an LP. Okay. A GP. And then like a hedge fund manager or an economist. Okay. And just talk about capital allocation writ large.
Starting point is 00:27:41 What do you think of that idea? Interesting. It's definitely upmarket. And I think, yeah, and I think it's niche. It's got to be smaller. If you can tie it back in, like you don't want to, I mean, I spoke at the last conference that you did, right? The one in that was great. And I thought it was excellent.
Starting point is 00:28:00 And I thought like that was going in the right direction. So as long as you don't get too far away from what you were doing there because I thought the content was excellent. I thought what you explained to everyone on the vision of the fund and everything happening was good. So just don't get too far away. Don't leave everyone behind, J. You're the you're out there fighting for the little guys. So my idea is to have the angel investors there, high net worth individuals, got it. Have LPs there, people who back funds, have some GPs there.
Starting point is 00:28:30 who could help pay for the party, basically, subsidize it, and who get benefit for meeting LPs. And then maybe, you know, like I had Gavin there, you had Gavin Begaret. I had Steve Jervison did a great talk. You did a great talk.
Starting point is 00:28:45 Really good. So, you know, the three of you did these incredible talks. And so anyway, that's my thinking on it is like to do that. But now my point was, if somebody on my team who was doing event production, used one of these tools, and then had this all organized and then formatted into question,
Starting point is 00:29:00 and made a document or brought this up. I was like, hey, can we do a jam session on this with Figma Jam? I set it up already. I would assume that they did all that work. And I would then give them credit for it. So if you want extra credit and you're a remote employee, you could probably use these tools to make it look like you're doing a ton of work. Do 15 minutes of work and then screw off for three hours and 45 minutes.
Starting point is 00:29:23 And your boss is probably going to think he did a long. Or you can work for somebody like me who does it and then says, this took me 15 minutes in chat GPD people ask me questions now remember that website let me Google that for you let me Google it for you can somebody in the audience make
Starting point is 00:29:39 let me GPT that for you yeah because people are like how do I do X and I literally take I cut and paste their question into chat GPT barred and send it back to them as a share
Starting point is 00:29:50 and I send it back to them as a share and I take three screenshots and I put it in the public slack yeah and I'm like I don't know I just did this yeah I give this a B plus. It was a B plus.
Starting point is 00:30:02 This was inspiring to C. Okay. I'm at A minus. I think this is really, I'm at A minus for this. I'm at A minus for this because I like the feature. I was probably a little bit higher, but I like the features that you said,
Starting point is 00:30:13 if it could just iterate while you're, you know, talking or something like that, that would basically take it to being a perfect product essentially. Why not just add, they can do this with like their, you can either do this with Zoom.
Starting point is 00:30:24 You could make this a Zoom plugin or you could add your own audio to it. And then just to the team at Figma, they're awesome. Yeah, just let us do a little discussion while we're doing this and talk to the AI. Yeah. Much better experience. Yeah. Or have the AI listening to us.
Starting point is 00:30:42 And then actually, a good bridge step would be have the AI listening and have it just give a running list of potential prompts. And you just click run prompt, run prompt, run prompt. Yeah. And this is. You like to create a budget. Yes or no. Yeah. And J.
Starting point is 00:30:57 as you always do as the World Greatest moderator, it's a good segue into the next thing that's happening in AI, which, you know, you had the founder of, I think it's pronounced CREA on recently. And so this is what, you know, what are called these like... Episode 1850. Cria was spelled K-R-E-A,
Starting point is 00:31:18 real-time image manipulation. So think of Dolly, where I showed that Jedi Knight, in Dolly or stable diffusion. But when you, if you wanted to move the lightsaber or change the smile or something, you can move the objects around and it changes in real time. It's really beautiful. Yeah.
Starting point is 00:31:38 And so I'm doing a share here now for those watching. And this is becoming something quite real. So I have a picture with a moon and I kind of squiggled a, you know, maybe look what I wanted to be a mountain in the background. And you can see its creativity change as I'm moving around. So you're dragging a circle around on the. the left, that is like your heads-up display. It's just a little circle made in like a Microsoft 8th window. Looks very bitmappy, basic. On the right, you have stable diffusion, creating beautiful
Starting point is 00:32:10 real-time images, and it thinks in one of these that you have a woman blowing a kiss to the moon, which was super weird. Over here, yeah, over here. And over here it thinks it's a mountain, which is what I was kind of going for. And what this does for like input and kind of real time discovery, editing and creativity, this to me is well beyond now the point that we were talking about earlier, because here you're just seeing the power of the models and the humans kind of interacting together in very, very creative ways. When it's moving the image, it's giving an input to the large language model, correct?
Starting point is 00:32:57 Yes, correct, correct. Is it giving a text-based one, or is it just saying recreate the image with the moon here, or with these pixels here? Like, if you were to explain to a layperson, a five-year-old, whatever, what's happening behind the scenes here, how is this accomplished? And what is this called? Yes. So the starting point is this prompt, right, a moon and a starry night.
Starting point is 00:33:23 So that's kind of given the frame of reference. And then what this picture on the left is doing is basically tweaks to that. So as you've probably done a hundred times, right, as you were creating those bulldogs and cat vaders, you know, this is a different way to provide an input that's not just words. because ultimately words get translated back down into numbers for machines, right? And so this is just another representation. And that's why, you know, it's funny, I had this tweet on the weekend saying,
Starting point is 00:33:59 we need to just stop calling them large language models because there's just sequence predictors. And so in this particular case, it's using it to kind of predict what I'm looking for by giving it an input that's not a language in any way, shape, or form. Now one could argue that this picture is maybe perhaps a representation of language because I'm trying to draw a tree. And what we're really seeing happen here, which I think is just fascinating, is the creativity explosion goes up
Starting point is 00:34:25 because a lot of people don't have the ability to describe exactly what they're looking for and then iterate it in ways. And then you see here the creativity of the model. When I'm on the left here, and for those folks just listening, my intent was a moon with a mountain in the background. But when I move it over to the right,
Starting point is 00:34:45 I get this person because it's interpretation of my... It drew a silhouette of a person. It thought your drawing was a person's face. Yeah, exactly. And that's what it... Yeah. And I think... Wild.
Starting point is 00:34:55 That's... And this, to me, ties back into what you were talking about with Figma Jam, where we now are entering this world where it should be iterated, how we're iterating this image on the right hand side by moving this picture around. Imagine we had that happening in Figma Jam with kind of the... inputs that we were giving. We're right there. I mean, on this one,
Starting point is 00:35:20 I'd say we're less than three months away from that happening. This is inspiring. Producer John was playing with this. So if you unshare your screen, he'll show what he was just doing with this. But I give this, you know, right now,
Starting point is 00:35:34 I feel like this is very solid. I give it a B. Okay. I feel like the images are really good, and I would almost be able to use them. I, too, would like to see this have some while I'm dragging
Starting point is 00:35:50 stuff around, if I could say if I could give a prompt and say make it brighter. So I'm moving the sun around. He's moving the sun around here with the ocean. Yeah, so sun over ocean is the prompt here. And so I would like to say, oh, make it daytime or make it brighter.
Starting point is 00:36:06 Or make larger waves in the ocean. Add a surfer. So while I'm dragging things around, I would like to talk to it. And this reminds me a lot of Blade Runner where he is examining photos and he's like zoom in on that to the left to the right, recreate this, whatever. And then also in, of course, the incredible film Minority Report where he's got special gloves on and he's pinching and zooming and moving objects around a virtual desktop and AR basically. Yeah. You know, projection. So again, I like having multiple
Starting point is 00:36:40 inputs here. I like talking to it. I like dragging it around. And then you can see creativity getting really wild. And then you could say, make it trippy or make it like a 70s photo or, you know, put some people in, you know, bikinis and surfers out there. Yeah, make them make it more like the 60s, you know, style, fashion style. Okay, yeah, put a barnfire on the beach. And then you start like really getting into massive advances in creativity. So I think this is like a 1.0 product. I give it a B.
Starting point is 00:37:11 It's a solid B. All right. But again, I want to see multimodal input here, and I want to see more participatory AI. I want the AI talking to me. It would be great if the AI said, how could I make this better? Yeah. Or is this what you intended? So, do you have a location in mind?
Starting point is 00:37:30 Maybe we tell you, do you have a specific beach in mind? Like, what are the questions the AI should ask me? It should ask me, what time of day? Do you want anything else in the photo? Is the sun, you want the sun brighter? is you want the sun higher or lower. It seems like you want the sun setting. Like, it should be interacting with you more.
Starting point is 00:37:49 And I find like chat GPTs could easily do that today, but I feel like the people making them are not thinking holistically of the relationship between the creator and the other creator, you know? The really good point is that good creation is always highly interactive.
Starting point is 00:38:08 And this is interactive in one way. Collaborative, yeah. You're going to call it. If it's Figma jam, if this is image jam, well, then make it like a jam session, which is you and I make eye contact. I'm playing the bass, you're playing the guitar, we've got another piano for here.
Starting point is 00:38:21 We're kind of looking at each other, speeding up, slowing down, giving some vibes, you know? Give some vibes. Okay. Well, what's your grade on here? You're very forgiven. You're probably going to give a B plus, I'm guessing.
Starting point is 00:38:32 No, I think it's a B plus. Like, I think these are, look, what I'm really blown away by is in less than a year, right, where we've gone mid-journey, Dolly, we're seeing it's going, now this is just real time.
Starting point is 00:38:46 You don't even have, like, I have a mid-journey with maybe 5,000 generations in it because I've iterated trying to create an image so much that, like, I'm,
Starting point is 00:38:54 you know, I've done it like 100 times to try to get the right image. And now having something like this is just a game changer. And, you know, I'll tell you how much of a game change is this.
Starting point is 00:39:05 In September or in the summer, I was trying to make the all-in posters. Yeah. Using the stuff. And it just wasn't coming for the summit. Like we had the Bestie Who Loved me, like the Bestie Royale. Yeah. We had Fast Times at Bestie High, the beach party.
Starting point is 00:39:21 The beach party. Just wasn't working. And then this weekend, when we were doing our hearts up for all in, when I said, make me put the words all in with a bunch of hearts and unicorns around it, it made it to me perfectly. And it actually kind of put it in the right size. I guess it just defaults to the, you know, I think Instagram size or whatever, four by three. Yeah.
Starting point is 00:39:41 whatever that is. And I was the first time, I was like, Dolly isn't the best model? I don't think so. Yeah. Is there an app I can use that would allow me to use copyrighted stuff and uploaded?
Starting point is 00:39:54 Like if I wanted to upload my image and make me into a Darth Vader, how would I, can I do that? Can I put up two images and have it do it? Because I looked on the app store and every single app in the app store tried to upsell me on a $7.99 a week subscription. Wouldn't let me use it if I didn't do that. Yeah.
Starting point is 00:40:11 And I understand. like the servers are expensive, but I couldn't find one that did what I wanted, which is what Dolly does without the content restrictions. Yeah. Clip drop. Remember, we brought it up really quickly last week? I think clip drop. Do it better? Yeah, my feeling is they'll do the best job.
Starting point is 00:40:28 Imagine this. You got an idea for a tech startup. You're going to change the world. I know it. But you got a problem. You don't have any engineers. Engineers, hard to come by. They're very busy.
Starting point is 00:40:38 They got jobs backed up. Well, you need to find. find great engineers. You need to find them quickly. And you need to reduce your burn rate, right? Because you can't be spending like a drunken sailor. You have a limited amount of resources as a startup. Now imagine there was a partner out there waiting to help you who had a thousand on-demand developers and they were vetted, experienced, results-oriented, and passionate about helping your startup grow. And what if they charge competitive rates, you know, reasonable rates? Does this sound too good to be true? Well, you need to head to lemon.io right now. Startups choose lemon.i.o because they only
Starting point is 00:41:11 offer handpicked developers with three or more years of experience with strong portfolios. And if anything goes wrong, lemon.io will replace your developer as soon as possible. A bunch of launch founders have worked with lemon.io. They've had great experiences. So here's the call to action. Super easy. To learn more, go to lemon.com slash twist to find your perfect developer or tech team in 48 hours or less. And Twist listeners get 15% off the first four weeks. What a deal. So stop burning money, hire developers smarter, and visit at lemon.io slash twist. Easy peasy.
Starting point is 00:41:45 Lemon squeasy at lemon.io slash twist. Let's do another demo or two here. We're going to lightning round, folks. Lightning round. Okay. So this is another one of these models, these latent consistency models real time. And what this does is it just takes your video webcam
Starting point is 00:42:04 and then you put the prompt down below. And in this case, it was just like turning me into like a version of Arnold. Okay, so there you go. Wow. Yeah. Okay, on the left, in real-time latent consistency model, you have made yourself look like Arnold from Terminator 2. Pretty convincing.
Starting point is 00:42:28 And it's like a hybrid of you in Arnold, and it's doing it in what looks like pretty close to real-time. Yeah. And we're sharing like a GPU, and this is like on a free thing. So you can imagine, you know, and I think this is what you were asking for. like something like this, but, you know.
Starting point is 00:42:44 But what I like about this is, number one, it makes you look like a badass. And that's first and foremost what I love about it, too. And you're pretty badass already. But it lets you in real time kind of play with it. So again, if you were to put, you know, add a shotgun or holding a shotgun and wearing, you know, Raybans, would it do that? Could you say, hold a shotgun?
Starting point is 00:43:13 Holding shotguns will be hard because my, my Zoom is there, but maybe we can say wearing sunglasses. Let's see what I know. Yeah, so it added the sunglasses. Yeah. Maybe put raybans, mirrored sunglasses. Mirrored raybans sunglasses. I'm wondering if it would.
Starting point is 00:43:31 Well, I might not know what raybans are, but what are they? Well, I mean, it should. It's a language model and it's been indexed things. It should have pictures of raybans. It would have every product in the world index. Yeah. So I wonder if it would do mirrored Raybans.
Starting point is 00:43:44 No, it's still doing the same kind of sunglasses. So, yeah, I mean, it's, this is like a C plus for me. Oh, wow. I think it's like a C, C plus. It's like, it's a novelty.
Starting point is 00:43:58 It's not usable for anything in the real world yet. But we could be sitting here next year and, you know, we could do the pod as characters from our favorite movies. Movies, yeah. And or if we were doing here, I'll give you a really good use case.
Starting point is 00:44:13 Let's say we're doing a table read and we are actors in the Avengers and I'm Iron Man and you're Hulk and we decide we're going to do a table read. Well, we could do the table read from home and it could make me into the Hulk, make you into Iron Man and we could have the table read in character. Yeah. It'd be really accretive to the performances and really help the director, right? And now we can say, hey, put us on the battlefield, you know, from end game with Thanos. and I'm playing Thanos and you're playing Iron Man.
Starting point is 00:44:41 You know, we do our scene. And we kind of get a feel for that moment. And the director is giving real-time prompts. Hey, make the scene a little more gritty. Okay, add some dead soldiers, add some dead aliens, put some smoke. Yeah. You know, and the director is directing the AI set creator. So imagine the set creation is being done by, you know, the art direction team and the director.
Starting point is 00:45:08 The cinematographer is giving the AI Christian, hey, can you make this a shot from above, make it letterbox, okay, now do a cut shot to Thanos. And they would be essentially taking what is a table read where you're just kind of working on performance and dialogue, but then you add the shots. You add the story words.
Starting point is 00:45:28 And so you can see where this is going. So I think it's a C plus. Oh, that's a harsh grade. I'm going to tell you that is... No, because it's not usable. I'm bringing usability into here. We are past the time. You're like the product manager and I'm the engineer.
Starting point is 00:45:43 Like I'm like, wow, that's really hard to do. No, no credit for technical achievement here where we're talking about, hey, in the real world, does this apply? At least from my life. Well, that's just a model on hugging. That's just a model on hugging face. So you're giving people ideas to take that model and go build around it. Yes, I would like to see. We have a company that's working on like storyboarding and, you know, doing scripts and stuff like that.
Starting point is 00:46:06 and it's going well. And I think this is kind of like eventually going to be the next vision, which is from screenplay to storyboards, to table read, to fully produce movie. Yeah. To our bet last week.
Starting point is 00:46:22 Yes. On when some people, I'm going to bring it up on the next weeks, but there's a couple people that have messaged us. I have one more to show and then let's wrap up. Okay. You gave it a B or B might, what did you say? C plus.
Starting point is 00:46:34 Oh, man. I gave it like a, I give it from a technology perspective, an A, but someone's got to take it and build something around it. So hopefully one of our listeners, we want to build a startup. So you're giving it an A and potential. Yes, exactly.
Starting point is 00:46:47 I mean, it's not meant to be something usable. It's like a model right now. And you're giving it an A. Okay. Yeah. There it is. Folks, we've got a big gap here. Yeah, on that one.
Starting point is 00:46:56 That's very hard to do, and it's really well done. And it's running for free. It's incredible. All right. Last one. This is, I wasn't able to demo it because I took it off GitHub. But I wanted to show this one because I wanted to get your thoughts on it. So these folks created this model, and they just released it yesterday, which basically all you do is give it a picture of yourself.
Starting point is 00:47:19 And then you give it the dance as described through some kind of stick figure animation. Wow. And then it will animate you. And I think this is great because the TikTok dances need to come to an end. It's a waste, like, you know, you've said, this on the all in pod, like, people are wasting their time doing this stuff. And so if this becomes commoditized, I think it's very good for our society because it would be like, well, we don't know if it's real or fake and everyone has them and people will
Starting point is 00:47:47 stop wasting their time doing dances. Okay, auto tune of dance. Yes, good way of call. That's a, you know, auto tune of dance. On the left, just a static image of a woman wearing a sweatshirt and sweatpants. In the middle, a stick figure doing the dance, but the dance is derived from a text prompt. Is that right?
Starting point is 00:48:11 So they haven't released the co-for. I don't know what was the driving force for the dance. Maybe it was like it took someone. The right way to do it would be someone's already done the dance. And then you use an AI to... Exactly. You take an AI to turn it into a stick figure and say, well, make me do the same dance. So essentially, if you'd think about like, I saw somebody complaining,
Starting point is 00:48:31 I think it was Elton John, was complaining about Madonna, lip-sinking in her concert. and he thinks, like, if people are paying for tickets, they should be told that they're getting lip syncing. That was kind of an interesting concept, which is, like, if I'm paying to go to a concert and I think I'm seeing a live performer. You don't know if it's live or semi-live or recorded.
Starting point is 00:48:52 And he was saying, like, I sing my songs. I want credit for that. Madonna presses the play button with his claim. And if you're paying 100 quid to go, like, that's unfair. this basically will take it to like the pop star doesn't need to be able to sing can they just dance crazy right well they gotta be able to dance because you're gonna go there and they have to you know they can't put a video of them so dancing sorry about the music video yeah oh the videos yeah just the video it's like you're just going to be able to do
Starting point is 00:49:23 crazy dance moves yeah I mean more from a societal standpoint too many people are spending time doing dances and I mean if you had a fun if you had a joyful time making a TikTok, I'm okay with it. If it's like your full-on purpose in life and like, I don't know. I mean, it's just, it's great.
Starting point is 00:49:44 Yeah, everything in proportion. You know, you and I like to ski. Yeah. Talking to a guy who told me he's skiing like 200 days a year. And I was like, I think if I got to. He's moving around then, right? He's following the northern southern hemisphere.
Starting point is 00:49:57 He's a snow chaser, snow chaser. Yeah. And I'm like, yeah, you know, I, I think that would be enough. It was like four times as much skiing as or five times as much skiing as I'd want to do in a year. But okay, sure, go for it. So, yeah, I give this like a B.
Starting point is 00:50:12 I think it's really, I think this is really impressive. The output, as I, as seen, it's really impressive. Like, that looked pretty good to me. Like, I don't know if it passed the Uncanny Valley for you, but it was pretty close. Yeah. That's the thing I'm noticing is, you know, like the Dolly, I did one of a postcard this weekend from San Francisco.
Starting point is 00:50:30 I just gave it the prompt and Dolly. Make me a modern version of like Tokyo, Dubai and San Francisco. Or make me San Francisco in 2050 postcard and then make it as modern as Dubai and Tokyo. And it did like a really, I think a pretty good job. And then when you looked at like it had San Francisco spelled wrong and then it just threw in weird copy. But it did a postcard. So it didn't have the polish, right? And the fit and finish.
Starting point is 00:50:56 Look at this now. You know, like if you think like if I said I want to do for the. next conference, you know, collateral. We're talking about making collateral. And I wanted to make a welcome to San Francisco. This made like San Ran Cisco with two eyes and welcome to is not correct. It's all kinds of musugina in here and craziness, but fun. The image itself was acceptable.
Starting point is 00:51:21 And so this is the thing. I really think the next job of creatives is going to be everybody gets to be creative. And I think this is going to be, when we look at things that are bad for society, things that are good for society. Everybody being able to be creative, I think, is going to be wonderful for society. The fact that you're going to be able to make a movie, write a better screenplay, make beautiful images, make collateral for your conference. That's just upscaled. You do your restaurant menu. It looks like, you know, somebody spent, you know, a month on it, but you did it in a minute.
Starting point is 00:51:59 All that, to me, is great for society. Now, it's bad for the people who get paid a lot of money to do that or get paid an average amount of money. But as we discussed, people used to spend on average, the average funded startup would spend $5 to $20,000 on a logo and a branding thing when they were a Series A company. Yeah. Five to $20 is pretty low. I mean, probably more than that even. Well, I was talking about you raised $2 million or $3 million in 2000, 2010. And back in those times, yeah.
Starting point is 00:52:30 Back in those days when you actually spent money on logos. And then 99 designs came out. And, you know, then you had beheads and dribble and fiber and, you know, all these upwork, all these incredible sites for finding somebody in Manila or Latin America for $35 or somebody at home in Canada who would do it for 500, $5 to $20,000 turned into $500 to $2,000 is now going to turn to five minutes to 20 minutes, not even money. just minutes. And I think that's wonderful.
Starting point is 00:53:01 I think it's wonderful. And then those artists can work on higher level things. So I think of this, like, this is to me as a solid B. I'd like it to teach you dance moves, like Steasy does, you know, teach you the dance moves. So I can see this, like, I do the move, and then it shows me what I should look like. So imagine it's like, here's me attempting the move. Yeah. Here's what perfection would look like in my body.
Starting point is 00:53:27 and then I can see myself doing it perfect. So, oh, I got to raise my arms a little higher. My elbows need to go up. Oh, I need to smile and put my... Or put it into like Peloton or for your workouts. Like, here's how you can fix your form or whatever. Yeah, that's actually really interesting. If you could show me as I would look if I did weights four days a week,
Starting point is 00:53:45 and I would say, hey, do these things. And here's what you're working towards. And this is the percentage you are. Like if FitPod did this, it'd be amazing. Yeah. If you commit to this number of workouts, 100 work, out's over the next year. This is what you could look like.
Starting point is 00:54:00 Yeah. And then it shows you... And this is what your form would look like as well. Yes. Yeah. And then it shows what percentage you're trending to it. You could say, hey, you know, you're 3% faster to it than, you know, whatever. This is amazing.
Starting point is 00:54:13 I mean, the world is changing so quickly. If you've got great AI demos, just CC at Sundeepe and at Jason, S-U-N-D-E-E-P. First Name Club. Because he's so excellent. At Sundip and at Jason. First Name Club, you know? That's where we're... part of Definitive.io if you've got a huge company or a well-funded company and you need somebody
Starting point is 00:54:33 to help you solve these kind of at-scale problems for Walmart or Target or Disney. If you're a Fortune 500 like or if you're an S&P 500 like, the latest S&P 500 company, Uber. Somebody asked me if I invested in Uber. You know what I told them? I got a check. I have a Google sheet where I track all my investment. It might be in there. It might be in there.
Starting point is 00:54:57 It might be in there. It might be up, you know, 5% today. Congratulations, in all sincerity, to the Uber team, alumni, 1.0 team, 2.0 team, the people who were the 1.5ers. Congratulations. Anyone who bought at the pandemic low. $16. Yeah. Now you're 4X.
Starting point is 00:55:15 And last year it was 25 bucks a share. Someone from our group chat bought at that low and I bought at the same time. You did too. Congratulations. Awesome. Well, we know who's going to be paying for lunch when we go skiing this winter. Let's do it. We'll see you all next time.
Starting point is 00:55:27 Bye, bye.

There aren't comments yet for this episode. Click on any sentence in the transcript to leave a comment.