This Week in Startups - AI Demos: Big Drops This Week from OpenAI & Google | E1950

Episode Date: May 17, 2024

This Week in Startups is brought to you by… LinkedIn Jobs. A business is only as strong as its people, and every hire matters. Go to https://www.linkedin.com/twist to post your first job for free. Terms and conditions apply. Eight Sleep. Good sleep is the ultimate game changer. The newest generation of the pod, the Pod 4 ultra has arrived. Head to https://www.eightsleep.com/twist and use code TWIST to get $350 off the Pod 4 Ultra. HiddenLayer. Generative AI is revolutionizing industries. HiddenLayer’s AI Detection & Response Solution secures your Generative AI & LLMs from malicious attack. Helping you generate more – by enabling seamless & secure Generative AI. Visit https://www.HiddenLayer.com/TWiST to learn more. * Timestamps: (0:00) Sunny joins Jason to dive into this week’s AI news and demos. (2:10) OpenAI did their big reveal on Monday and review on Sam Altman’s appearance on the All In podcast (5:40) A dive into what OpenAI launched this week. (10:11) LinkedIn Jobs - Post your first job for free at https://www.linkedin.com/twist (13:12) OpenAI’s new model, where the “o” stands for omni, means it can now reason across audio, vision, and text in real-time. (17:21) Open AI’s “memory” setting and their new MacOS desktop app! (19:43) Eight Sleep - Head to https://www.eightsleep.com/twist and use code TWIST to get $350 off the Pod 4 Ultra. (22:37) Jason highlights the startup Athena. (25:53) Looking at the improvements on the new ChatGPT 4o model. (29:31) HiddenLayer - HiddenLayer’s AI Detection & Response Solution secures your Generative AI & LLMs from malicious attack. Visit https://www.HiddenLayer.com/TWiST to learn more. (32:34) Breaking down the pricing of the new vs historic ChatGPT models. (35:31) Jason demos Recall. (42:19) Sunny shares Labs.Google and NotebookLM. (46:03) Labs.Google and “AI Overviews” * Check out ChatGPT 4o: https://chatgpt.com/?model=gpt-4o Check out Labs Google: https://labs.google/ Check out Groq: https://groq.com/ Check out article “Apple Nears Deal With OpenAI to Put ChatGPT on iPhone”: https://www.bloomberg.com/news/articles/2024-05-11/apple-closes-in-on-deal-with-openai-to-put-chatgpt-on-iphone Check out Athena: https://www.athenago.com/ Check out Recall: https://www.getrecall.ai/ * Subscribe to This Week in Startups on Apple: https://rb.gy/v19fcp * Follow Sunny: X: ⁠https://twitter.com/sundeep⁠ LinkedIn: ⁠https://www.linkedin.com/in/sundeepm * Follow Jason: X: https://twitter.com/Jason LinkedIn: https://www.linkedin.com/in/jasoncalacanis * Thank you to our partners: (10:11) LinkedIn Jobs - Post your first job for free at https://www.linkedin.com/twist (19:43) Eight Sleep - Head to https://www.eightsleep.com/twist and use code TWIST to get $350 off the Pod 4 Ultra. (29:31) HiddenLayer - HiddenLayer’s AI Detection & Response Solution secures your Generative AI & LLMs from malicious attack. Visit https://www.HiddenLayer.com/TWiST to learn more. * 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 Instagram: https://www.instagram.com/thisweekinstartups TikTok: https://www.tiktok.com/@thisweekinstartups * Subscribe to the Founder University Podcast: https://www.founder.university/podcast

Let's cover what they really launched this week. And then we'll do a demo with some of their latest stuff. They launched a desktop app. Ironically, a Mac app before Windows app, given all their investment. A minute ago, you were showing me the charger. It can take a screenshot, actually, right? And here you take this and you see this. It can take the screenshot or you can take a photo.

A photo file, yeah. Yep. So I said, can you explain what you think is going on here? Let's see what it says. The image appears to show a video call between two individuals, person in the top frame is holding up advice, possibly explaining or demonstrating to the person. He looks somewhat confused or concerned. That's typically how I left.

It could be a virtual minute of tech discussion, perhaps a product demonstration. Yeah, I mean, it knows what's up. Yeah. So that is, yeah, kind of bonkers. This Weekend Startups is brought to you by LinkedIn Jobs. A business is only as strong as its people and every hire matters. Go to LinkedIn.com slash twist to post your first job for. for free. Terms and conditions apply. Eight sleep. Good sleep is the ultimate game changer.

The newest generation of the pod, the pod for Ultra, has arrived. Head to eightsleep.com slash twist and use code twist to get $350 off the pod for Ultra. And hidden layer. Generative AI is revolutionizing industries. Hidden layers AI detection and response solution secures your generative AI and LLMs from malicious attack, helping you generate more by enabling seamless and secure generative AI. Visit hidden layer.com slash twist to learn more. All right, everybody, welcome back to this weekend. Startup, Sundeep Madra, Sunny, my guy, is here with me. And we're going to do AI demos. It's been a huge week. You're going to want to get a notepad out or your chat GPT assistant. And you want to take notes because we have a ton of demos to

show you today. Welcome back to the program, Sundate Madra, from GROC, formerly definitive intelligence, which was purchased by Brock, which you can access at console.grock.com if you want to be one of the 250 billion developers who are working on. No. It's 150, 250, some great number. 175,000. It's a lot of devs. Okay, let's get right to it. We know Open AI did their big reveal on Monday after Sam Waltman was on All In and didn't tell us any device. How? all this stuff launched. Actually, true story, he was, I think, he told Freiburg that they had some stuff to demo, but then that they pushed it back. So that was our mistake. We should have rescheduled him so that we actually had the demos, but we'll give you the demos right now. Yeah.

Did you see the discussion with him on All-in? I did. What did you think? I thought it was really, really good. Okay. I think he shared- He got trashed in the comments and online. Yeah, I was kind of one for not saying anything. What do you think of that feedback? People were upset that he didn't say enough. I think like the people that watch All-Land are always expecting, you know, something big. And, you know, he said three or four important things. Here's what I'll kind of say. He said, the next versions of chat, GPT may not be numbered or GPT may not be numbered.

Right. That's big. That's big. He talked about the story of what happened. Right. And gave information he had never given before. That was one breaking news story. The second breaking news story was that all of these projects like the chips and the phone were

part of opening on. Well, that's where I was going to go. never told that to anybody and he was like yeah well i mean nobody asked me and i was like well okay just asked yeah and i thought the third thing that was breaking news was what do you think it was was there was anything else you that was kind of breaking or new thing the story of how he got fired yeah like he was at the f1 yes he got the call and then he sat around for a bit then he became okay with it right he was flying back like just that whole story was pretty good what else are people looking for well and then also the stuff he talked about was

with inference and training data versus and copyright, I thought was really revealing the fact that he wasn't going to touch music and that he respected people's copyright. He'd never talked about it in that kind of detail. So that's like four or five pieces right there. Yeah. I think people might be mistaken. His general, and I guess I've heard two different versions of this.

Like, cynical people are like, he's not being humble. He's being aloof or evasive and holding his cards close to the, chest and, you know, maybe, I guess, the most cynical interpretation was he, you know, he's being a sly or something. And then on the other side, people are like, he's being humble and he, you know, doesn't have all the answers. And he's trying to explain that. And then some people are like, it's both. Or he's being fake humble. Like, he actually knows what he's doing. And, you know, all this humility is like he's a James Bond villain. And I, you know, I was like, I don't think he's a James Bond villain. I do think he's figuring it out. I've known him for a long time. And, yeah,

I do think he's actually pretty calculated, but another word for calculate is thoughtful or strategic. So anyway, I don't think he had to share those things with you guys. He could have kept it. You know, and the reality was, you know,

some of the videos we're going to pull up today because all that functionality is not fully released in the public. They were recorded on the Friday. Ah, so the day before, day after we tape, we tape on horses. Because in some of the people are asking,

hey, what day is in? It's like, oh, it's May 10, right? So. I think that was the issue is that they were still working on it again working on it yeah yeah okay

let's get right into it then great job breaking down the all-in episode great job for freeberg for getting them on the pod I don't have any problems with it so let's get started so many different ways to start here and I'm curious where are you going to go first let's cover what they really launched this week and then we'll do a demo with some of their latest stuff right so here we are on the open eye YouTube channel and if you're not watching go to our YouTube I know this weekend startups on YouTube. And I'm going to pull up the ones that are kind of novel because there's a bunch here.

And some of them we've seen before. Like, you know, there's a demo here where someone makes it count. This gentleman makes it count. Then he makes it count slower and faster, which is actually very hard to do technically. But, you know, it's not that interesting. A dad joke, you've probably done this 100 times. Well, I thought the going faster and slower on the counting,

where he tells it, hey, can you count faster and it counts too fast? to somewhere in between, that actually is something that I think it does need to understand. So it doesn't feel like, and I talked about that on All In with Sam Altman, was right now, when you're talking to it, it feels like CB radio, where I'm like, can you talk slower, over? And then the other person can talk. But if you try to talk at the same time, I do think they're teaching it to be interrupted and talk more like a human. Can you count from one to ten for me, please?

Of course. One, two, three. Hey, actually, that's a little slow. Could you count faster? Sure thing. One, two, three, four, six, seven, eight, nine, ten. Okay, that was a little bit too fast. Can you, like, count somewhere in between those two?

Got it. One, two, three. And now even slower, please? Okay. One. Two. Okay, we get it. We get it.

I think it's important, you know, that the interface for voice works because when you use Siri, we all know when you use Siri, it's like you're giving it one command and you're waiting 30 seconds, 60 seconds for that to get deployed and come back to you. It just kills the whole experience.

Totally does. So I think the ability for it to understand and adjust itself, which we hadn't seen before, is a really big new feature. One of my favorite ones, which I hope a lot of people start using was math problems of GPT4.

I'm just going to start it. And this is Sal Khan. Yeah, in Khan Academy. Yeah, so I'm going to just start at this one. Open AI invited myself and my son, Imran, here, to try out some of their new technology. And so we're curious about how good it might be at tutoring someone in math on Khan Academy. So let's get it started. It's loading up.

So it's loading the iOS app, correct, on an iPad? Yep. Actually, in this form, I can't. I'm here with my son. And I love you to tutor him on this math problem, but don't give him the answer. You can ask questions and nudge him in the right direction, but I really want to make sure he understands it himself. And he's here in the room, so you can talk to him directly. Of course, I'd be happy to help.

Let's look at the problem together. Can you first identify which sides of the triangle are the opposite, adjacent, and hypotenuse relative to angle. Alpha. All right. So we get the idea here. And so I think, again, this is really important because they've done two things. They've made it such that it can look at your, you know, device screen. Right.

So the app, just to make this crystal clear for people, the chat GPT app is open. It's taking up about a third of the screen. The other two thirds of the screen is like a drawing program where they've drawn. I think he was on Khan Academy. It was just a website. Yeah, it's just a website. So he's on the webpage. but it's seeing what you're seeing on that web page.

So it's asking to read your screen, which is something that's allowed in iOS or on your desktop now. So it's watching what you're doing. Therefore, it's kind of the guide on your side. And it's there. And it's being able to watch the screen and then help you, which,

you know, for me, Sal Khan has a very famous TED talk where he talks about, you know, Two Sigma. And basically what that talk is about says, if you give any student at any level,

a tutor, they improve two standard deviations, two sigmas. And so imagine is now available for free for anyone on the planet to use pretty awesome. Now we just have to get them off TikTok or Fortnite. And everybody on the planet's going to be, you know, yeah, two years ahead in math. Hey, founders, hey startups. When you're hiring for your small business, you want to find quality professionals who are right for the role. understand it can be tough to find the time and resources to find that perfect hire. That's why you want to check out? You guessed at LinkedIn Jobs. LinkedIn Jobs gives you the tools to find the right

professionals for your team faster and for free. And LinkedIn isn't just some simple job board. It helps you hire professionals that you cannot find anywhere else. There's over a billion members on LinkedIn now. And they're in more than 200 countries. And I know many of you are leveraging different time zones to get work done, different geos to find talent. And LinkedIn is the best place to get eyeballs on your job posting. You know that. Even those who aren't actively searching for a job will see your ad. Why? Because LinkedIn knows who might want your job. You know, they've got a lot of AI and machine learning going on over there. So even if they're not actively searching, they're going to see your role. And you know what? Those are some of the best candidates.

And did you know that 70% of LinkedIn users don't visit the other leading job sites? If you're not looking on LinkedIn, you're looking in the wrong place. And the best part is when you use LinkedIn jobs, 86% of small businesses get a qualified candidate in 24 hours. So, higher professionals like a professional on LinkedIn. LinkedIn jobs helps you get the right people faster. And guess what? The first one's on your boy, J-Cal. Post your first job for free. LinkedIn.com slash T-W-I-S-T. That's right, LinkedIn.com slash twist to post your job for free. Pairments and conditions do apply. And so that is the truth. As we move on and now what is your excuse for not wanting to get smarter becomes the question because not only is all the reference material on the internet, not only are

their videos. Now there's a coach that can walk you through those reference materials and do what's called adaptive learning. Adaptive learning has been something they've wanted to create on computers for a long time. And it's just been too hard because you had to manually hard code it. Here, a language model is doing the adaptive learn. Correct? That is exactly it. And doing it at your own pace. It's not a person. You don't have to worry about their time and their tutoring, you know, cost is just the smartest, most powerful tutor in the world that can be customized to your learning level. And so while we worry as a species about AI taking all of our jobs and all this kind of stuff,

AI is also going to make us smarter. Every single person who wants to learn how to play chess or do math or maybe cooking or play guitar, this thing's going to be able to watch you play guitar on your camera and tell you, hey, you know what? When you're doing that C-cord, there's two different ways to do it. Here's some visualizations of that and just, you know,

maybe press a little bit more gentle on your finger so they don't hurt. And like, this is stuff that a teacher would do and now anybody, billions of people on the planet are going to have a custom tutor. This is A plus for me. Well, one more just along those.

I want to show one. What would you rate this one? Did you rate, you want to rate this one? Don't rate yet. We'll rate the whole thing as a thing because otherwise we're going to end up a lot of ratings here. I like a lot of ratings.

Okay. I don't vote. So here, which is kind of going to your point, Hey, how's it going? You can see here, he has a video open. How about you? I see you're rock in an open AI hoodie. Nice, chilis.

What's up with that ceiling, though? Are you in a cool industry style office or something? Well, can you take a guess of what I might be doing based on what I'm showing you here? Hmm. From what I can see, it looks like you're in some kind of recording or production setup with those lights, tripods and possibly a mic, it seems. Yeah, so again, we get the idea here. The presenter has the iPhone app running the latest version of Chad GBT, and it now has a mode

where you can start a video, a live video stream to it. So it's not a picture anymore because we got used to that. So he just basically opened a camera, and he initially had the camera pointed at himself, and it described him and what he was wearing, and it got a little bit of the ceiling, so I guessed that he was in an industrial office. and then he flipped the camera around and he had like a recording setup and it said, oh yeah, it looks like you're recording

there's a lights and some cameras there. So is there a mode this is called? Is it called camera on mode? Is there a mode for this? Environment mode? Where do they call it? I haven't seen this released in the app yet. It's not in the app yet.

I looked for it. Yeah, exactly. So my guess is like, you know, today you have the ability to give it a picture or, you know, audio or other things. They're going to add them like a sort of of like a record button and it's going to open up and it's going to create that recording. Got it. So I wonder if it can turn on both cameras because I'll just call this environmental

awareness mode. Environment awareness mode, EAM. So when you're in EM mode, environmental awareness mode, it knows what's going on around you. So why would this be important? Well, if you had it in your phone in a cradle, let's say, you know, like I have a queue charger on my desk here. I'll actually show it to you. So I have charger, you know, like this from anchor. So this is my anchor charger and my phone's on it, right? So if I had this and it was looking at me, then I could, if I was holding my guitar, it'd be like, oh, are you looking to do guitar lessons? And we would be skipping the step of me opening, you know, the app and saying, hey, I want to do guitar lessons. Because it would be aware of what's going on around me, which I think is super powerful. If there was this environmental awareness,

mode. Okay, environmental awareness. So it's kind of setting the context for you saying, here's what's going on around. If you were in the kitchen, it would see that you're in front of the stove on your iPad and an iPad stand. And it would see if it was tilted down, all of the ingredients and it would see chicken and say, oh, you're cooking chicken. Do you need some help? Oh, I see there's breadcrumbs and I see mozzarella and I see some tomato sauce. Are you making chicken parm? Like, this is kind of like, whoa. just like a human might walk into your kitchen. You know, you might walk in and your spouse is cooking.

You say, oh, you make a chicken parham. That's kind of mind-blown when you think about it, just to understand the context of where you are and what's going on. Oh, you're in an airport. Are you looking for instructions of where to eat? Oh, I see you're in Tokyo. Oh, you're in Shibuya District. Oh, I see you're in, you know, Ginza.

Are you looking for a sushi place? I mean, this is crazy. When you think about it being aware of the environment around you, And just how much friction that takes out of the world. And plus sound, you can also listen to sound, right? It can now, yeah. It's going to have that full ability to listen all the time and provide that context to you. Yeah, chat chippy can now see, hear, and speak.

We're beginning to roll out new voice images, capabilities in chat chatsby. They offer a new, more intuitive type of interface by allowing you to have a voice conversation or show chat chatsypity what you're talking about. Voice and image give you more ways. They don't have a name for this. They should. It's actually an interesting thing because people, saying, hey, maybe they should name it because they have all these kind of new functionality.

Two more things I want to show. This wasn't part of the release, but we never got to demo this. So I want to pull this up. So one of the things that they've added, if you go on your settings in chat GPT now, Jason, you're going to really like this. If you turn on memory. Yes, I have memory turned on. Yeah.

Do you have it where it's starting to remember different things? Yes. Remember that I like this. Not custom instruction. It's different. It's different. In a table. We did that another time.

Yeah. But, hey, I've got a puppy and the puppy's name. And you remember all these kind of little things? Okay, that's another really interesting feature that they launched. And you can ask it, what do you remember about me? Correct. And it will tell you what it remembers about you, which is super cool. So the last thing we're going to show from them, and then we'll give them a grade on all these things is they launched a Mac app.

Desktop. A desktop app. Ironically, a Mac app before Windows app, given all their investment. Yeah. A minute ago, you were showing me the charger. It can take a screenshot, actually, right? In here, you take this and you see this. It can take the screenshot or you can take a photo.

A photo file, yeah. Yep. So I said, can you explain what you think is going on here? Let's see what it says. The image appears to show a video call between two individuals. Person in the top frame is holding up a device, possibly explaining or demonstrating to the person. He looks somewhat confused or concerned. That's typically how I left.

The background suggests in real music, relationship with some posters, correct. The person on the bottom is listening attentively, possibly involved in the discussion about the device. He's wearing glasses and has headphones on, earphones on with blur background indicating a focus on his face. It could be a virtual minute of tech discussion, perhaps, a product demonstration. Yeah, I mean, it knows what's up. Yeah. So that is, yeah, kind of bonkers. Now, is it continuously monitoring what's on your screen? No. It does not have a mode. I could not figure out a mode. Basically, all I could see here is pretty much this is like a chat GPT Mac app.

It has the same mode of like you can go in with your phone app or it can be constantly going back and forth with you, but it does not have a mode for... Continuously monitoring your desktop. Continuously monitoring, yeah, yeah. Because that would be very interesting that if you and I recorded this whole episode and instead of just giving us a transcript and a summary,

it edited the screenshots, it annotated the screenshots and all of that, you know, and now you start thinking... Three months away. They're going to have that. Now that they have this, they're going to have that. You've heard me talk about how great my sleep is. Sleep is a superpower and my superpower is eight sleep. They've got the amazing pod that you add to your existing mattress. It will cool you down or warm you up, depending on which side of the bed you're on.

I like it nice and cool. I sleep deeper. I am a sleep king and you might be a sleep king or queen too. And you can be one if you get the pod for ultra. That's right. Pod 4 Ultra is out, and it is definitively the greatest tool for helping your sleep. Pod 4 Ultra can cool down each side of the bed to 20 degrees Fahrenheit below room temperature. It's a better cool than the air conditioner. So I can put the air conditioner on a reasonable temperature and then keep my pod ultra, even a little bit cooler than that.

And they also introduced an adjustable base that fits between you and your mattress and your bed frame. So you know when you want to read and your partner wants to sleep? Well, now reading and sleeping positions are right there to help. you unwind after a hectic day. You can even elevate just a few degrees automatically to help with your snoring if that's one of your challenges. Not only that, it has sensors that will track your sleep time, all the phases of sleep, your HRV, and your heart rate. So it's basically a wearable that you don't need to wear. I think I get like an extra hour of deep sleep every night because of it. I'm coming into the game with good energy. So here's what I want you to do.

Get the sleep you deserve at 8Sleep.com slash twist and use the code twist to get 350 bucks off of the pod for Ultra currently shipping to the U.S. Canada, UK, Europe, and Australia. I love this product so much. I designed it to invest in the company. Well, now imagine we're on a call and it's researching it or let's say we were taping an episode of this week and started and said, hey, you know what? Give me all the links for the show notes and write a blog post about it and post it to my LinkedIn and put it in my draft folder. It could make like a whole blog post and email newsletter. And you can say, and tell us what we got wrong or give us suggestions for future topics for the podcast. I mean, this could get really powerful. It could be like a virtual podcast producer watching you work.

Or if, let's say, I did five meetings with startups today and I had two meetings with the investment team. Like I said, just give me a summary of the day and what I got accomplished. And it was just sitting there and it was doing that. And then you're like, and you're like, oh, you know what? Remember during that meeting? so-and-so was saying this about what was it again? Exactly. You could be like, hey, is there anything I forgot to put on my to-do list?

Yes. And it would just go back and be like, yeah, you know, you did mention the to-do list seven times last week or this week. Here are the seven items. And here's the video clip of it. I already put them on your to do-list. So like, just listen to my entire day with to-do list or just be listening and you and I are just sitting there working. And I'm like, you know what?

Remind me next week to do this. And those kind of things exist in task management software. to do list, but a persistent assistant listening to you is a whole different level of power that you could have. And I've actually, I'll just mention it here on this podcast, so I invested in a company called Athena. You can go check out Athenawow.com. And what it is, is it's assistants, like real assistants, but they're based in the Philippines and they're about $3,000 a month. It has changed my life. And so I will have this real world assistant, which costs me only $36,000 a year, I'll sometimes have her on Zoom with me, with my desktop up,

showing her what I'm working on, and she's learning what I'm doing. And then there's this learning one. There's a real life person. Okay. So we are in the process of documenting and learning everything. And I think there's going to be this merger between like my EA in Manila, who works with me for, you know, eight, nine, ten hours a day and this. And if you go to Athena, w.O.O.W. Athena, wow.com, they'll let you jump. There's a wait list to get on this thing. If you talk to any CEO in Silicon Valley, they're using this because it's hard to get a personal assistant. And they basically skim the cream of the top.

0.1% of knowledge workers in Manila. And it's by the same founder who did Thumbtack. And so you'll get like $1,500 off if you go to Athena, wow. Just, and there it is. Boom. I was the first investor in Thumbtack and I'm the first investor in Athena. So go to Athena. Wow.

And you'll see, yeah. And it just will make you so, so efficient. And then you add this to the mix that my assistant plus this virtual assistant would be sitting there, you know, collaborating on stuff. And that's what I'm really trying to do is super efficient. Yeah. It's so killer, right? Because think about, you know, Zoom has this feature now, right, where you can come in and you can do all that. But, you know, what's with me is not Zoom all the time, but Chad, GPT, or you.

take your pick of whichever one you want to use. So to be able to recap the day, how powerful would that be? Insane. Yeah, especially because then I was sure my Athena assistant, and then she would, Nikki would be like, oh, okay, I watched your day, I watched the recap of your day. And I guess I just give, she'd be like, give me a 10-minute recap of Jason's day.

And wow. Yeah. Okay. Let's keep going. Awesome. Okay. Well, let's give them a score because we're done on opening. I'm going to, I'm, you told me not to, but I'm going to do this anyway. I think the persistent

app watching what you're doing is a plus, plus, plus. The fact that it exists and it works and it's doing this right now is the game changer of all game changers. And I know a lot of people were thinking about like monitoring desktop and doing stuff, but this thing's out now. So this is

like one of the best things I've seen in AI to date because it's closing the gap between like a toy that gives you, you know, hallucinations and like some, you know, basic web research to now watching your behavior. I think the ability to talk to it better, I give that a B, like the speed, which is doing it. But the contextual awareness of your environment and watching us work, that's an A plus feature for me. Yeah, I'm in the A category. I was for some room.

Like, we don't have all those features yet, but like, we're close now. Yeah. I mean, it was a great release. Is it any better? Like, is the model any better? Or is just new features on the existing model? No, the model is faster, which is important, right? Obviously, we spend our time thinking about that.

So they've made it, they've made it a lot faster, which is good. It's low latency. But the important thing about the model is it's truly, and they call it Omni, it's truly multimodal. So a lot of times you see some people doing voice and images and obviously LLM type stuff. and what that's typically been is what's called kernel fusion, kernel fusing. But here,

they trained a model from scratch with all three of those things. So it's not three different models that they kind of fuse together. It's a model that was trained on text, on video, and on audio. So that is, you know, some underlying, underlying technology that I think you don't see in the interface explicitly, but if it is listening and watching you,

you know, that's what the O is for. The O is for Omni. Exactly. And the Omni means audio, visual, text, all in real time at the same time. And it knows how to move between those, which is kind of incredible. Yeah. And so with the MacOS app, yeah, you just can open it up and boom. And also on the desktop app, you now have the voice conversations.

It's only for plus subscribers, which I'm one, but I don't see it here. So I think I have to get it turned on or something. Twice as fast for 50% less costs for businesses. I think just talking about a big picture, the costs are going to continue to plummet on this thing. And they said they want it to be free for all users. And he wants to give everybody access to this. So is it now free for chat GPT4 on the website?

Like if you're not logged in, I noticed that everybody is now starting to let people use these services when you're not logged in. Did you notice that? Yes. are playing with these. So it used to be it had to be logged in. So I just launched a private window.

And if I type in chat GPT, I can ask it a question, hello, what can you do? Yeah. And in which model? Yeah, that's chat GPT 3.5. And then if I... You can change it to 4.

Oh, you should be... For 4, you have to log in and create an account. Oh, you do. Okay. But the point is, it does seem like there's a path here. But I think for chat GPT4,

not Omni, I don't think you have to pay now. So it seems like they're going to... going to, it's almost as if whatever the last version of Microsoft Office was would be free. Free. The new version is paid. So that's kind of an interesting model. It goes back to their nonprofit routes. I think what's going to happen is when we talked to Sam Altman on the All In podcast last

episode, I asked him about UBI and his experiments there. And he said he wanted to give a certain amount of compute AI or AI resources to people. That seems really interesting. I think Open AI may be part of their mandate is. going to be to make all this stuff free. So while they have a paid version, whatever the most recent is is paid. So you get a little bit of edge. So maybe like this year's version, you have to pay for it. But last year's version, everybody gets for free. Unless it's version shouldn't cost them that much to service because inference and everything's getting cheaper and they'll have

optimized the model. So it's expensive to serve up chat GPT4. Oh, I would expect, right? So Omni's going to be expensive. And 3.5 is cheap, right? What would you say the difference in cost is in a multiple for them. Well, in this case, 4-0 was a very specific, it's actually cheaper. It's half the price. Listen, we all know generative AI is revolutionizing

every industry. In fact, we talk about it here on this podcast every week. And companies that are slow to leverage this technology, well, they're going to risk falling behind their competition. But new innovations also bring new threats. How do you safely accelerate your company's gen AI adoption while also managing the risk?

Well, let me tell you about hidden layers AI detection and response platform. It's a security solution specifically designed for generative AI and large language models. LLMs, you know that. It provides your security team with visibility and tools necessary to detect and prevent threats like leaking of confidential information. We've seen that a couple of times, as well as malicious prompt injection. That's a serious threat. And model theft, right? People are starting to steal. That's just the truth of it, folks. People will steal your innovations. That's going to enable your company to harness the power of AI. and to do that securely and confidently.

It's trusted by leaders in finance, tech, healthcare, and even the U.S. Department of Defense, the DoD. Hidden Layer helps you generate more by enabling seamless and secure generative AI. So here's your CTA, the old call to action. With Hidden Layer, go from pause to possibilities and step into the future of secure AI innovation. Learn how to protect your generative AI today

by visiting hiddenlayer.com slash twist. That's HiddenLayer.com slash T-Div. WIST. There's a couple of interesting intersecting forces here. Let's just talk about them. People have been saying, and I'm not in agreement, that models will keep getting bigger. I feel like the folks that have been saying that

are trying to do it as a moat. Because the models get bigger, it becomes harder and harder for anybody else to enter that. But as the models get bigger, they become way more expensive to run. So it's also like a perverse incentive as in, oh, let's make one that's twice as big as GPT4. Well, then it's going to be twice as like,

expensive to run, right? So I think what we're seeing now is tech usually gets more efficient, not bigger and more expensive. Okay. And so we may see a bit of a shifting landscape here where future models are actually cheaper to run than older models. Right. Older models become worthless. So that's what it really is. It's not worth charging for them. Okay. That's bizarre. Hold on. I have to really grok what you're saying. With a cue. With a queue. rock with a cue. If you make a new model, they're becoming so much more efficient that the new one has better gas mileage than the last one, but not just better gas mileage, like 100 miles per gallon instead of 10. Therefore, if I gave you a new car and it was 100 miles per gallon and your previous one was 10 miles per gallon, you would stop driving the one that was 10 miles per gallon because you'd make up the cost so quickly.

Yeah. So in other words, all these previous models are going to get thrown in the garbage. They are depreciating super fast. Like I can tell you from our experience, like Brock. The day Lama 3 came out, 48 hours later, we deprecated Lama 2. We just took it off the site. Everyone moved over. And we can see the prices right here.

Producer Court will pull them up. The pricing is plummeting. What is the pricing, you know, trendline here when we talk about this, Sunny? Talk about price. GPD 4 is 50% cheaper than GPT4, where GPD4 was $10 for million tokens input and $34 million tokens output. It's now gone. If you scroll back up a little bit, it's gone to $5 and $15. But just to let you know, open source is still cheaper.

Open source is 60 cents for million tokens input and 80 cents per million tokens output. Okay. So that's a fraction of that. Yes. And the open source ones hosted are that price, right? Yes. And look at where the original GPT4 was, $120 and $60. Okay.

So $120 dropped to what? $120 dropped to $15. Over what time period approximately did that happen? So GPT4 came out last March. Okay. So in a year? Yeah, year and a month. So you think that these things will get 90% cheaper year over year?

I believe so. So that means over two years, it'll be 1%. So every two years, it will be 100 times cheaper. But we're going to use them 100 times more. Now, think about all the stuff we did today. And if it's running all the time, it's not viable to pay that much. If it's listening to everything you're doing, because we're entering a world where before it was sort of limited interaction, you're typing some text, you're doing it.

Now we're talking about a world where it's recording you all day long. So what you'd have to do is say, you know what, record every word I say, because that's just going to be 10 hours of audio. And maybe I only talk for two hours of the day. So it's two hours of audio. But take a screenshot of my desktop every second means 60 seconds, let's say 600 minutes, 3,600 screenshots or something like that. Processing 3,600 screenshots. That doesn't seem that expensive. Yeah.

It adds up. It adds up. Now, and especially if you were going to record video, you would need, at a lot, least, I don't know how many frames per second, but typically 30 frames per second, maybe less. Maybe you could do it in 10, skip around. You know, film is 24 frames per second. Yeah. You know, you probably do it.

Zoom obviously does less. Maybe like 10 or 12 would work. So what's going to happen is we're going to be processing a lot more data with these, you know, with these technologies. They should have like, you know, they have this, um, portfolio. FOLio company that I'm really enamored with. It's called Recall. And let me show it to you here. And I'll get your view on this.

This company, and actually I'll get you to give a letter grade. Hold on. Let me just share my screen. Oh, a J-Cal demo. A J-Cal demo here of this company, Recall. Because I know that you like this category. This is called Recall. So here I am on a YouTube video, right?

And let me just go to the Open Eye page here. Okay, here I am on Open AI. And here's that one we just played, right? But I will click on recall here. You see it's doing its work, and it's recalling this page. It's analyzing, parsing the data of this YouTube video. And then it gives me a summary of the YouTube video. And then I can save this.

And then when I save it, it saves it to my knowledge graph. And then it automatically is building a knowledge graph for me of all this stuff. And then I can click on knowledge graph. And I can see the All In Podcast. It can see Open AI. Say hello to chat, GPG4. and I can start to move around and see what are all these, here's universal basic income, right?

Which is the AI. We were just talking about universal basic income. Here's the UBI page. And this came to it because we were just on, I have the All In podcast on here from YouTube. And these are the themes that when I come to my knowledge graph, you can see here are the themes all in conversation with Sam. And here's what came up.

Here's Google, Masayoshi-san. I mentioned Stripe. mentioned Peter Thiel, Sam Altman. Wow. And that's Berkshire Hathaway, Y Combinator. Really from my introduction. Here's UBI. And then you see UBI can go out and here's chat GPT and here's OpenEI.

And I've only put a couple of things in here. But as I add stuff here, like if I were to do a search for UBI experiments and find a video about a UBI experiment, right, South Korea's UBI experiment here, Then I add that to my recall. And this is a company we've invested and we incubated. So now it's discovering the connections. It's summarizing the main ideas. Boom, it's going to put it into,

I've saved it into my knowledge graph. Now, when we go back to recall, it now has that one here. Did they do it via the YouTube transcripts? I don't know, but here we go. Now, if I find UBI, here's opening I. And when I find that UBI one, eventually there's Dolly,

Overton window. And at some point, there's UBI. It's going to have for UBI, you know, the two or three places we talked about it. And I can expand the card thing here, right? And see more things related to it. You get the idea.

And so what I asked them to do was to monitor every single web page that I visit in recall. And they said, that's too expensive. And I said, well, how expensive? They said it would be like hundreds of dollars a year. And I'm like, charge me a thousand. Yeah.

I would pay for that. Come on, what are we talking about here? If you're a business executive, let's, you know. You know where I would pay for that, J-Cal? GetReCall.a.I. You would pay $10,000 a year, a thousand a month. You'd pay $1,000 a month. You know how many times I've been to a tweet, and I forgot to like it and I can never find it again?

And it's like, oh, I really need it for something. Oh, my God. Rate that startup that we just invested in. What do you rate that? Get Recall. I'm going to give them a B. Perfect. Yeah.

I give it a B right now. If they had the feature where they did everything, I would give it in it. Yeah. You know, Chat ChbT will have some of these features, but I do believe as we look at this space that dedicated products that are going to, you know, that have features and interfaces for a particular vertical bookmarking and knowledge graph, writing a screenplay and doing storyboards, doing math, like, it's just going to enhance those things. So you'll have Chat Chipatty underneath Khan Academy, but you're still going to want to go to

the Khan Academy website, I think, and have like a full experience there. I think those things are still going to exist exactly. What's not clear to me just yet, because there's so much happening right now is, and, you know, what's really kind of got me thinking a lot is, you know, we're going to get into the next other demos, but what we just finished with is if I have that Mac app and it has now access to everything, and then I do something on my phone, it has access to a bunch of things. And then the flip side, you know, one of the things Google did today, or yesterday, right, when Google I.O. was on, they talked about the generative experience

being part of their court search engine. So everything you're doing from them. And as part of the browser. So search Gmail, YouTube, they're going to have their own knowledge graph. Oh, and in Chrome. Okay, wow. So here it is. Everything you do is going to be surveilled and inside of Google or OpenAI. Pretty scary and insane if that got hacked. But I guess the benefit is the language model knew everything about you. But at some point, the FBI could come or a quarter order to come and say, hey, tell me everything you know about Sundeepe. And it would be like, yeah, we know Sundip likes these websites and really likes this actor and

watches these TV shows, listens to this music. I mean, it's kind of crazy when you think about it. Yeah, they could probably get that from your search history. But if you go on. If your search history is on. For Google, it wasn't even on and they were still doing it for some people. They got in trouble for that recently. They got in trouble for it.

But already was in Chrome. Incognito mode, not the big joke inside of Google was incognito mode was not incognito. But Chrome has helped me write built in now. Oh, wow. You can create custom images. You can organize your tabs in a smart way. So they're already starting to add the high level AI features. Amazing.

So help me write is like grammarly inside of Chrome. That's interesting. Yeah. So I do think, like, I wonder. I love Gramerley but I think like 60, 70% of Gramerly will be built into other products

so Grammally is going to have to keep figuring out what unique features they can add for writers that aren't going to be in like sort of native products. Yes, exactly. So, you know, it's an entire journey that's happening very quickly.

I do think like you will be able to build apps that are 10, 20, 30, 40% better than whatever is in these. Just like, at GPT keeps making things that are 20, 30% better than what's in another person's model. So you just have to keep adding features and listen to your customers. Okay. All right, talk to me about Google.

So Google at Google A.O., a lot of things talked about there from developer stuff, which is at the core of what they announced on the developer things, which we won't bring up right here. They've expanded the context of Gemini to $2 million. They launched a new AI chip as well, which is awesome. and they're continuing to make more services available within vertex. From a consumer standpoint, we're just going to talk about a few of the things that they've put out. And the best way to kind of get access to these,

back at labs. Google. Okay. Once again, Google making it impossible to navigate to find stuff. I mean, this is something. I mean, bar. Dot Google, Gemini. Google.

Now it's labs. com. Okay. Yeah. So here you can see. It's not Google, right? Yeah, labs. at Google is not even.com now, right?

So I think we're, and so what we're going to do here is we're just going to start going into a few of the things that they launched. I know you like this one, J-Kal. No-Book L.M. So what they did in Notebook LM is it can take up to 1.5 million tokens and context now. So if you have a bunch of contact here, you can put it in here, and it can then basically create a summary for you on that content. Right.

I did this. I had Stephen Johnson on this weekend I started up, she can look it up. And I showed myself taking the screenplay for founder, the founder, and the book grinded out, the PDF of the book, grind it out about Ray Crock and a bunch of other Ray Crock business case studies I found from like Harvard and other places like PDFs I found. And I put those all in there. And when we did an episode, I asked it, give me key moments in the history of McDonald's and what lessons they learn.

and it really started to act as a research assistant, the job of a research assistant who would compile all that, read that, and then sit in a room with the writers, it kind of acted like that. And, you know, that's like a,

you know, basic research position in the world of 50, 60, $70,000 a year is what researchers get paid. You know, I've employed them for many years in publications and journalism, starting pay back, you know,

maybe 40, 50K, you know, if you're an expert researcher, you can get up to 60, 70, 80K in the world after 10 years. and those exist in all number of modalities,

like a magazine, a newspaper, a consulting firm, a law firm, whatever, just a general researcher. It seems like that function is going away in the world. Anybody who needs a researcher can use notebook LM, Gemini, Chat Chapti. The researcher mode. So that means we all get to do other things, right?

And I can tell you, my producers have been using Chat Chepti since it came out at my insistent to summarize things. And the summaries that they do versus what chat chipped tea can do, essentially the same. So then what it comes down to is them understanding my personal taste or the the gestalt of the show. Fancy word for the vibes of the show. The vibes. The vibes. And then kind of cross-referencing them, right?

And so eventually we'll put into one of these LLMs all 2,000 episodes and say, what would J-Cal find interesting about this startup or this person? and it will go out, find the 20 interviews this person has done and then surface the things that they think I might be interested in. It's really going to be very interesting. Yeah.

So now it's all of that just with 1.5 million tokens of input context. You can imagine J-CAL, that should cover almost all the things you've ever published. Everything I've ever published, wow. You think 4,000 hours of podcasts would fit in that? Maybe no, let's leave the podcast out.

Oh, okay. How about everything I've ever published? blog posts, etc. Everything you've ever written blog posts. Maybe all the transcripts, right?

Yeah. Yeah. I mean, a book is basically six to ten hours, right? Yeah. So if you think about it,

like every podcast, maybe every five podcast is a book in terms of the transcript. So there you go. 60,000 words, whatever. Amazing.

Wild. We're going to keep going through their stuff. So the next thing they did is they've made generative search experience. I don't know if you use this, but like this is sort of my

go-to these days more so, which is the, the gen search experience. And so they've made this generally available. And so when you do something, and we're going to just show this example here.

So again, I'm in the Google Labs is showing enough AI overview. On more searches. AI overviews and search. Okay. And this doesn't work. You have to use your Gmail address

for this to work. Once again. And you have to use Google Chrome. God damn. So you can say, how can I get better at improv? How can I keep bananas fresh longer? What you're going to see is and see they've got it,

the AI overview right here now. This is interesting now, J-Cal, right? Because they're calling it. Perplexity. Yeah, it's perplexity. Yeah. Why about perplexity? Yeah.

Well, you know, I think they're innovating too. So don't count them out. But this experience is zippy. So yeah, do it one more time. I want to see. I didn't see it. Where should I go on vacation? with my 10-year-old in January.

That's warm and fun. I mean, I got really detailed here. Let's see. That was too much for it. Okay. Tell me about Turks and Kcos. There we go. There you go.

Perfect, yeah. And then if you show more, it's kind of instant. I wonder if it's pre-cash. I don't know. Yeah. What's interesting is this one is different than the banana one.

This one's called AI Overfew. It was different color. That one was blue. That was a little bit different. Interesting. So they may be just cashed up all the most popular ones, right? Yeah. See how this one is red?

Oh, yeah. Look at that. Huh. Yeah. I mean, this one was instantaneous. They make everything so hard. God, it's just. Yeah.

Oh, see, this one went into generating interest. AI overviews in search. Okay, I got to sign in. I got to use my Gmail account. I talked to Sundar about this, actually. I said, do you see, Jekal, like, so much stuff has gone below the fold? Like, previous Google was so much harder to use. Yeah, I mean, this is the end of SEO.

So if you wanted to get traffic from Google, the reason Google didn't do a lot about this, or one of the reasons they didn't do this was because they were in a very fragile balance between indexing people's content, leveraging it to make money, intercepting search traffic, Google becoming the starting point, and then sending traffic to people. Now they're going to train on your data,

give the answer, and not send anybody to your website. And this is going to result in people saying, I want to get paid for my content if you're just going to use it to rehash this stuff. You know what I'm saying? Because here when it's doing it, let me show you mine,

I just used one of theirs or actually yeah so if you show more here one of the interesting things is they I think they'll put uh scroll down let's see if they put uh links so see what they did here they made an overview but because they're in a very fragile relationship with publishers they gave a carousel at the bottom of their sources they didn't quote them but click the first one and you know on that list and I bet you of the six they did here you know this AI are article from Forbes was probably where they got the information from. So I wonder if they're just giving you the top six Google results or the top six Google results that created that summary. What do you think? I don't know. That wouldn't be useful. I mean, I think what they have to do if they want to be competitive here with folks like perplexity, they have to look through a few of the results, summarize it, and then spit that out. I think that's what they're doing. I think that's what they just did. So if you typed in here, what are the top science fiction films of 20, 23? And if it does summarize this, you know that it doesn't know that without having indexed a page, right?

That's not the type of result that they're going to know. Oh, interesting here. See this? Yeah, well, that's because, yeah, you got the, it asked you to generate it because they already have what to watch. Like, they have their own thing here. But here, look, it did, it, it gives you some of them. them and then it gives you immediately the IMDBA. And then it says according to rotten tomatoes. And look, there's a down carrot there. If you click it, it gives you the link. So they're doing sources and they're making the sources a little robust. Do you notice that? Why are they doing that? This is the fragile relationship that happens when you steal people's content without their

permission. Google has stolen this content and made these summaries without the permission of those sources. So what they're hoping is, and I'm using the word stole to be a little provocative here, but you could say they've leveraged, they've utilized without permission. We know that they don't have permission to utilize these. Okay. They utilize these without permission to create summaries. In order to not get sued,

they're hoping slash film and IMDB are good with that high profile link. Yeah. See what I'm saying? What the ultimate test will be is how many people continue to click these links and how IMDB looks at it and says, oh, you know, they're still sending 80 or 90% of the traffic they were before fine. But if that goes way down, there's going to be chaos.

I'm investing in a startup that is creating a clearinghouse for the rights for content holders in AI. Mark? So like ASCAP or something. Yeah, exactly. Someone was going to say someone does it for music, right? Yeah, and people do it for movie clips. People do it for social clips. Like, you know, those viral clips where you see when somebody posts some breaking news story on

X.com formerly known as Twitter. And then all of a sudden, all the reporters reply back, hey, it's ABC News. Hey, it's NBC News. Hey, it's CNN. Can we have permission to use your clip? There's a company I think it's called Storyful. And what Storyful does is they look for those moments. They tell the person, we will monetize your clip.

They'll put the clip on a marketplace and tell everybody you can license this for this amount of money to play on the local news or your website. And like, I don't know if you follow. I take a picture of like a car on the fire or something on the highway. And it's, I get the video. I can monetize that. Wow. Of course you can. And so here is Storyful.

This company makes a lot of money. and the only reason you would know about Storyful is because you would have shared one of these but Storyful sources Verify and License Breaking News and viral video content in real time and so if you are a news program

with editors, a New York Times a CNN, you know that you have to get permission for these social videos and so they have relationships and they pay you can see here they have 300 plus broadcast clients

etc. and so when these crazy, you know, like that whale, hump back well coming out of the ocean, all those citizen journalists. But the person that did that submitted it to them to say. Storyful has people out there watching and then you can submit the video. They'll get in touch with me. They'll get in touch with you. Or you can just go in there and submit it and make money from your video.

So you have a submit thing. So that's storyful. That's crazy. Yeah. So, you know, if you or I as non-commercial entities, retweet something, yeah, you're not going to get in trouble for that. A retweet is not taking it and then commercially using it. Yeah. And so nobody's really going to sue you for that. And if you put it on a social media platform that has retweeting, that's considered like safe harbor, no big deal.

But if I take it and I put it on CNN or I put, I download the video and then put it on my up, re-upload it to my account. Yeah. And I monetize it. Then you're in trouble. And that's what StoryFold does is they've created a way to be a clearinghouse for all of this. So those 300 broadcast channels go on to Storyful and they say, you know what? We need something for the 11 o'clock news. Just grab, hey,

you're never going to believe what happened when this couple went to SeaWorld. Oh, my God. They got splashed with a bunch of water. How crazy is that? You know, that kind of kooky, you know, morning show. You should do a whole episode in that voice. Oh, God. I can only keep that voice up for like 90 seconds before it actually hurts.

Because it's so high energy. Oh, whoa. It's crazy. this is the innovators dilemma. When you're open AI or an open source model, you can do whatever the hell you want. You got nothing to lose. When you're Google and you already have these relationships

and you have deals, they're just really scared of getting sued or having a class act of lawsuit and they know better because they've got 20 years of lawsuits and legal actions and the people at Google indexing stuff have had the people at the New York Times

come and say, here's what you can take, here's what you can't take. And, you know, But this is what happens after multiple decades. Open AI is like Yolo, bring on the lawsuits, we're brand new. They don't have the institutional knowledge relationships and or scar tissue. So I was going to say scars are the biggest ones, right? Yeah, they make a gunshot, right?

Yeah. The one thing Sam said is, hey, we're not touching music right now. And he was like, yeah, I just believe an artist. I wouldn't do it in the style. I wouldn't say do it in the style of dire straits or rolling stones because they know they're going to get sued. Everyone understands the history of Napster and, you know, all the route. The music industry understands the value of their content and their rights, and they are predicated on protecting those rights. Yes.

The news industry and content sites, they don't bother. Photographers, sometimes they bother. Sometimes they don't. Video. If it's Disney, yes, they do, right? Well, you know why? If you think about content as these, like, concentric circles or pyramid or whatever, There's not that many a big artist, so it's easy to protect them.

Yes, there's a small number of artists on the fat tail of the long tail. Yeah. Whereas Forbes articles, they produce 10,000 a day of the schlocky articles on Fortune Forbes, all these places. Yeah. You know, whatever, you know, there's a 10,000 substacks get published today. That kind of stuff, those people are not fighting to protect their rights. But, yeah, every single Rolling Stone song is violently protected. Every single Star Wars and Marvel character is violent.

protected. And he said, I said, you know, I noticed you can't do Darth Vader, but I did know you could do the concept of a Sith Lord. And he's like, yeah, that kind of makes sense, doesn't it? Like, one's a concept, a Jedi, one's a specific character. And so, you know, I guess their red teams are doing all that. But anyway, I think, you know, Google putting. Yeah, I get it, you know, kind of. I think Google. They're all concepts within Star Wars. Yeah, but the concept of a hero. or a villain is not. The concept of a laser sword

probably existed before Star Wars, or a sword certainly did, or a sword that's got superpowers, Excalibur has superpowers. Yeah, of course. So, like a powerful sword, or a sword made a fire, has existed before Star Wars, right?

What's the tail on content protection 50 years? I think it's 75 after the life of, yeah, length of, let's ask, what is the length of protecting characters in the United States, whatever? copyright. There you go. Life of the author

plus 70 years. Yeah, that's what I said. Life of the author plus 75 years. 70. So the reason that keeps growing

is because of Mickey Mouse. So they kept, they called it the Mickey Mouse rule, right? So for works published before

1978, the term is 95 years from publication, right? So they kind of grandfathered everything in. Yeah. So whenever you

published it, 95 years. And the Sorcerer of Mickey will not enter the, for example, 1940's film Fantasia version of Mickey Mouse known as Sorcerer Mickey will not enter the public domain into 2036.

The cartoon style shorts will not enter until 2100. So Steamboat Willie in black and white is now. So see that? I know more about this than I should. The Steamboat Willie version and black and white

features a more simplistic style. That was the one that just became public domain on January 1st, 2024 in 1928. So this is why they keep updating the style of the characters. They keep updating the style so that they reset it. That's why you see new versions of Thor or slightly modified versions of Iron Man.

If they keep changing it, they can keep protecting it. And then you might be using some old one that doesn't look like the old one, right? Clever, right? Yeah. Jake, I think we have to part two this. We're definitely part to this. Thank you everybody for tuning into this week in startups. Part one.

Please. This is part one. We got a lot more to do. Consul. groc.com with a cue. sign up as a developer. Let's get Sonny's equity position

and my equity position in Grock to the moon. You can follow Sunny. X.com slash sundee if I'm X.com slash Jason. Joining us as co-host of this week in startup shortly. Alex Wilhelm,

formerly of TechCrunch, X.com slash Alex. I cannot wait to get a new co-host in here. Sunny will still be doing AI, by the way. But the co-host for the other couple episodes every week. We'll see you. And by the way, Mark Suster of Upfront Ventures

has been doing four episodes for me. So like Johnny Carson or Jimmy Kimmel, I have some people sit in and do hosting for me. Mark Suster, thank you to my pal, Mark Suster. He's doing four episodes. I think two of them are up already. They're amazing. Go check them out and you'll see them in the feed. So if you don't see me on an episode, don't worry, not going anywhere.

But with four episodes a week, 200 a year plus all in, that's 250 episodes a year. I'm going down to like 150 and then 100 will be done by people in my network. Can I say something? And J-KEL's team makes it quite easy. you got to do your work, you show up with your microphone and your computer, and that's it. So everyone that's wanting to do pods

that has to go through the entire arc of the content, then you have to record it, then you have to edit it, then you have to publish it. Marketing. You've got a machine.

Yeah. I'm trying to make this week in startups twist into a platform. So I think I'm going to start referring to it just a twist and then it'll be like twist with Mark Suster,

twist AI, whatever, just you have this weekend startups And, you know, if it goes to seven days a week, because it was that six days a week at one point, if it goes to seven days a week again, that would be okay with me. I just can't do seven days a week and all in because that'd be eight, and then I have no life. And then, you know, you look at these people who do a daily talk show at night, they lose their minds. Yeah.

They eventually, James Corden, you know, Charlie Rose, Howard Stern, anybody who does this every day, the burnout is too much. Yep. See you all next time. Bye-bye.