Mark Cuban’s investment strategy in the world of AI

0:00Starting a business comes with its share of ups and downs, which is why staying true to your vision is essential. a non-negotiable for Romeo and Milka Bregali, Capital One business customers and co-owners of Ra's plant-based restaurant in New York. Romeo and Milka took a leap of faith when starting their own restaurant, gutting an empty space and building it from the ground up. Every pipe, every wall, every detail. But building from scratch came with a heavy financial burden, which is when they turned to their Capital One business card. With the flexibility of the card's no preset spending limit, they were able to spend more and earn more rewards while bringing their vision to life.

0:36Today, Raz's success is proof that with passion and the right support, it's possible to make your dreams a reality. Learn more at CapitalOne.com slash business cards. If you've spent any time building AI products or leading technical teams, you know this. Transformation doesn't fail because of ideas. It fails because teams can't move together. Enter Atlassian's Teamwork Collection. It has planning in JIRA, documentation in Confluence, video updates in Loom, and now AI agents in Rovo, which connects the dots across your work so nothing gets lost. It's one AI-powered teamwork platform designed for how modern teams actually build.

1:21Learn more at Atlassian.com slash TeamChanger. That's A-T-L-A-S-S-I-A-N dot com slash TeamChanger.

1:41I mean, I don't know if you saw, but 6-7 is like the dictionary word of the year or whatever. And I heard it. I was like, what is this? Once sort of like half the videos with me are like, 6-7! Just doing stupid stuff yelling 6-7. It's insane. Looking back, the best financial decision I ever made was simple. It wasn't a flashy deal. It was sticking to what I understood, building it brick by brick. That changed everything for me. Six, seven. Sixty-seven! Sixty-seven! Sixty-seven! You might recognize that voice. It's Mark Cuban. Yes, that Mark Cuban. Known as one of the sharks on Shark Tank, owner of the NBA's Dallas Mavericks, and a legendary entrepreneur and investor, with a list of companies in his portfolio that would take this whole episode to read.

2:30And you heard him talking about all the AI videos he's appeared in lately, including lots with the 6-7 meme. More on that in a bit. But know this. Mark Cuban is going all in on AI. He's backing AI companies. He's thinking about how it will change everyday life. And he has strong views on what people need to navigate a world with AI. A big shout out to my friend, longtime venture and legal executive, Mike Chakarow, who connected me with Mark. Thank you, Mike. Mark and I connected and geeked out on topics like robotics and computer vision. And that's our episode today. It's a wide-ranging, no-holds-barred chat that I had a lot of fun recording.

3:16I hope you enjoy this look into the mind of Mark Cuban. I'm Rana El-Khalyubi, and this... Pioneers of AI, a podcast taking you behind the scenes of the AI revolution.

3:37Hi, Mark. Welcome to Pioneers of AI. Thanks for having me on. I'm excited to be here. So I want to start with Sora. So for our listeners who don't know what Sora is, it's OpenAI's video creation tool. You can create these short clips with yourself, your friends, and with celebrities if the celebrity has allowed it. And you were kind of very early in allowing people to create videos of you, and you kind of went viral. Why did you do that? Why did you decide to go all in? Well, so I went in, like, I try everything, right? I just love to play with everything. And so you have the opportunity to do a cameo is what they call it, right?

4:14Where you put it up, just like doing facial recognition type setup. And then they have these delimiters that you can ad to limit, you know, or define how you want things to proceed. You know, you can put don't do this or don't do that. But I was like, I wonder if I can use this for marketing. And so I put on there, always finish with the logo of costplusdrugs.com, one of my companies. And let's just see if it works. And damn if it didn't work. And it was just always there. So I'm like, if people want to put it out there, have fun. Because, you know, they do have guidelines and limits to what they'll allow people to make you do.

4:53They don't always go far enough, but they're still relatively safe. So I don't even know. I think there's been, you know, four or 5 ,000 videos created of me. And the funny thing is, right, I was sitting right here when I did it and I just cut myself shaving. And so I had this mark right there and I'm like, now I'm going to know if it's And if anytime you see a video of me with a little shaving cut, I know that it's from the AI. I know it's from Sora. That is funny because we made a video of you, which we want to show you. And we used it to announce that you were coming on the show. And we actually asked the audience to submit questions, which we have a few of, and we'll tackle that later.

5:34But we want to show you the video. Okay. What questions about AI should we talk about on the Pioneers of AI podcast? Well, I saw this one. I saw this one. Basically, it's an AI twin of you, and you are in a robot factory building a humanoid robot. And it actually, you're a little stiff, to be fair, but your voice sounds exactly like you. And I feel like if somebody doesn't know you super well, they could really think it's you. Oh, no, it's crazy. Like, people have used me to, like, do promotions, come to Destin, Florida, you know, to vacation. And I got an email from a friend, and he was like, you're going down to Destin?

6:12I have a place there. We should meet. But my latest ones were me winning the somersault competition of the Olympics and being crowned world champ. Can you actually do a somersault? Could I? Yes. Will I? No. But Sora knows how to make me do a somersault. So it's been fun. That is so fun. Well, to your point, though, you have all of your AI generated videos include Cost Plus Drugs, which is really genius because it's free advertising. Do you envision that that will be the future of advertising in a way? No, not really. That was more of a kludge on my end just to see if it worked. And now because it's new, people are willing to do it.

6:51And because it's 10 seconds, it's not worthwhile to go to another platform and remove the Sora or Cost Plus logos. I mean, even though some people have. So it's more just about learning for me because, you know, the beauty of this is that you can iterate so many times and so quickly. And then if you want to edit, you can edit them together into something. And, you know, but right now it's just the first inning and it's a chance to play for it. Do you have a digital twin of yourself that's out there that say, you know, I want to book a meeting with you that I can just go talk to your twin? No, but I did.

7:26So I was the first investor in a company called Synthesia. Wow, they're doing really well. They're doing really well. That's a British company that creates human-like AI avatars. They just raised a new round of funding, and they're reportedly valued at$4 billion. And so seven, eight years ago, I invested because I wanted to try to get in front of all these different AI things and computer vision and neural network stuff. The Mavs went to China in 2019, and I had an avatar that had me speaking Chinese that they put up on the Jumbotron in China. So, you know, it's not, I'm not responsive in anything that I'm doing, but I've used Synthesia.io and it's been great.

8:08That's amazing. So we're both passionate about this idea of the health span revolution and the combination of new sensor technology, new types of data, and of course, both predictive and generative AI will unlock preventative medicine, personalized medicine. Would love to get your thoughts on this space and how are you thinking about it? A couple of things there. One, your body is just one big math equation. you know, but we don't know all the variables. We don't even know how many there are. But over time, we'll develop new ways to uncover all of that, right? And new sensors to be able to gather new information, new algorithms that'll be able to pull things out of that data.

8:48And I think that will make us better at preventive healthcare. You're starting to see more blood tests and analysis of MRIs and x-rays and everything. So I think all those are good steps forward, But we haven't really even got into the computer vision side of it, right? Because computer vision, particularly for robotics and capturing things and analyzing them, is really compute intensive and really hard right now. And it's really hard to get enough stuff to train on. And so I think as we start getting more into CV-driven things, we'll see more impact on healthcare and the results will start to be more dramatic.

9:23You know, I invested in a company for sports. It uses computer vision to track a player's gait while they're running in a game and to look for nuanced variances from game to game to game, and then using algorithms to try to look to see how it correlates to injuries. and they're getting a lot closer. They started with soccer in the UK and Europe and now we're starting to do things with basketball. And so doing those types of things where it's more preventive than reactive, I think we'll see more and more of as we get more in-depth compute with video. Yeah, I think that's fascinating. My son is a squash player and he's gotten a number of hamstring injuries and Achilles, you know, tears.

10:08Anytime you have that quick movement like that, right? You're always putting strain on your, and particularly if you play a lot, you get these repetitive stress injuries. So can you detect that early on and have him and then personalize the training that he needs to do to stretch and whatever? Exactly right. All that stuff, it's great training grounds, but we're just in the early, early, early days. So we're both passionate about both AI and entrepreneurship. So I want to talk about that next. Do you think AI is lowering the barrier to entrepreneurship? Oh, yeah. Oh, yeah. By age and access. I mean, there's so many people.

10:45Like, I gave a speech yesterday. So we have this thing, Mark Cuban's AI Foundation Boot Camp. Tell us about that. Okay. So basically, I had this guy that I did it with yesterday, six years ago, maybe five years ago. Somebody just sent me a cold email. I'm from Pittsburgh, and they were from Pittsburgh. And they're like, we're trying to come up with ways to introduce AI, neural networks at the time into schools for middle school and high school. I'm like, that's great. Somebody needs to come up with an idea to help train kids and get them excited about it. I'm like, I should just create this thing called a boot camp and just set it up.

11:17But yesterday we had 500 people in, including a bunch of kids that were all involved with the boot camps. We started it as a way of going into schools that might not have otherwise had advanced STEM programs or the resources to do this and making it available to them. We went from basic neural networks, here's JavaScript for a three-layer neural network, you know, six years ago, to, you know, to machine learning, you know, here's how, here's six different trees, here's the attributes, you know, which tree, you know, guess which tree is the next one, to now, you know, doing prompting with some understanding of machine learning underneath it and the probabilistic, you know, approach that AI takes in generative models.

11:59I love that you're not just teaching these kids to a prompt engineer or a vibe code, but you're actually teaching them like what's inside the box. I mean, it is a great democratizer. I mean, few kids, even with the least amount of resources, don't have access to a smartphone. I think it's not easy, though. So I sit on the board of my kids' school, and, you know, it's a New England prep school, so it's been around for hundreds of years. And it's been interesting to kind of see how we're going about this. Like, I think there's interest in being at the forefront of using AI in the classroom. But we've also, as a school, struggled on how to bring that to practice.

12:37And of course, all the kids are already using AI, right? Oh, yeah. They're using it no matter what. Like, I was talking to the kids and they're like, they know how, you know, they know if you go into chat to UPT and you have it rewrite something, all the different tells, and they know how to erase the tells, all that stuff. They're way ahead and they follow all the Instagram and TikTok accounts that have all the AI hacks and all that kind of stuff. So they're way ahead. But the teachers are terrified, right? And that's what it comes down to. And that was a big purpose of what I was doing yesterday, to try to get them excited that they can do more with kids.

13:06I want to talk about young people a little bit more. So I have two kids. Like you and I? Yeah, exactly. It's funny. So my daughter's 22. People actually often say we look like twins. And I'm like, yes. And she rolls her eyes and she doesn't like that. But my daughter's 22 and my son is 16 and a half. And they are on the... Yeah, my son's 16. Yeah, I think we have like similarly aged kids. I have a daughter who's 22 and one that's 19. Yeah. So I'm curious. My kids are on the opposite sides of the AI spectrum. My daughter, Jenna, will not use AI for anything. She loves her brain. My middle daughter will not, hates it.

13:43Really? But my son is so AI forward. He uses AI for everything. Well, my son will use it, but he thinks all AI videos are slop. My middle daughter is concerned about the environmental concerns, right, of AI and all that. And then my oldest daughter is like, yes, anything that's a hack that makes my life easier, I'm all in. So it runs the gamut in the cube. Yeah. Yeah. So how do you, I mean, I guess what's your advice to young people? Try it. It's like, look, it's always, there's always going to be something new, always. And the kids adapt the fastest. And it's the adults that say, oh my God, I'm afraid.

14:21And then when it's your kid that knows, you know, back in the day, it was like, when a kid could write HTML code to create web pages, they were like a computer genius, you know? And then if they used up their game to JavaScript, they were like the best of the best of the best. That's just the nature of the beast, right? And kids are just going to adapt more quickly because, you know, like music, like anything else, they want to be in the forefront. And then for our 22-year-olds who are, you know, joining the workforce, what's your advice? I mean, there's a lot of headlines about how these entry-level jobs are going away.

14:55There's something that just came out today. The unemployment rate for recent college graduates is 9.2%. That's scary. But my advice is stop going after the big companies. Go to the small to medium-sized companies. The big companies have IT departments and they have all the AI resources they need. Whereas if you're going to a small or medium-sized company, particularly one that's not a tech company that's entrepreneurial at all, you're like the new great thing. And what I tell kids is like, talk to whoever, manager, owner, whatever, and just say, I know how to do AI agents. Let me find all your manual processes or all the things that people hate to do at work here, right?

15:32That just are grind and boring and nobody wants to do them. And let me automate them. That's it. You're saying, these are the things everybody hates to do. They take too long, keep everybody late after work. I'm going to create agents that automate that. And we had one of my Shark Tank companies that Rebel Cheese, they sell nothing but vegan cheese, great stuff, rebelcheese.com. And so I'm only selling all day. But in any event, so they ship a lot of these cheeses. And when you ship something to consumers, you know, you're putting it in UPS or whatever boxes, right, USPS boxes, and they're set rates for the boxes.

16:11and then where you're going to send it to depends on the distance, that'll impact the rate as well. And they'll come up to a total charge of what the shipping company should charge you. They're always wrong. But most small to medium-sized companies that ship, they know that it's wrong and they have people that do nothing but review those invoices for accuracy. Here's what they want to charge. Save$10 ,000 a week by just creating a basic agent. One of my investment hypotheses, is then we will get to investing in AI in a second is vertical AI. So like looking for all these antiquated industries that have a lot of these manual processes and build like really like focused solutions.

16:52And yeah, it can be like manufacturing, like at Cost Plus Drugs, you know, we use robots and agents to, you know, just in sensors to say, okay, if anything's out of kilt, right? If there's an error message at all, beep, beep, beep, report, you know, recalibrate, do what we got to do with as little human intervention as possible so that we can get the price of manufacturing drugs here to be as cheap or not cheaper where there's cheaper labor costs elsewhere in the world. All those things, anytime there's an industry that's been doing it the same way for years, that's ripe with opportunity. Like you said, for the vertical industries, the crazy part is like, I get pitches all day, every day.

17:36So all, you know, HVAC and, you know, auto repair and plumbing, all these things, people are trying to do roll-ups and automate those. Getting them to understand what agents are sometimes can be, you know, an obstacle course, but those things are happening, right? People see them as opportunities. Coming up, Mark answers some of your questions and we dive into our philosophies around investing in AI. Stay with us. Thank you.

18:36of the most fascinating scholars and thinkers as we tackle big topics like how whales became the face of environmental activism, how to succeed at failing, and whether public transportation should be free. Go ahead, listen to Freakonomics Radio wherever you get your podcasts.

18:57So we will get to AI investing, but before we do that, I wanted to get to a couple of the audience questions that we got from that Sora call out. And I want these to be lightning round style. So let's go. This one came from Joseph Wehbe on LinkedIn, and he builds these AI ecosystems all around the world. So his question for you is, the U.S. is ranked number one in the global AI rankings. What does the U.S. need to do to win the global AI race? That's a great question because it calls on so many different questions, you know, the ethical, regulatory aspect of it. And I think if we can stay out of the way, but be very persistent in how closely we look at outputs, right?

19:41And don't over-regulate people who are creating, but over-monitor the output. I'm just saying, if this is what it's designed to do, it might be military, right? Show us so we can review it to make sure we know what you have so you don't go sell it to somebody that shouldn't have it, right? I think that's one. And I'm concerned about the over-reliance on these big data centers that we're creating because it's not even so much how much money is being spent, even though that's important. I just can't imagine over a 10-year period that we aren't going to improve the technology enough that if you overspend on today's technology, and particularly now with the administration that's opposed to green energy, and you're seeing even a state like Texas where you've got a lot of solar energy and even wind, that's 27 % of our energy consumption.

20:31I don't know that we're approaching energy the right way, but I'm not an expert in it. It just doesn't feel right to me. Yeah, that is interesting. How do we become kind of the hub for the entire AI infrastructure, whether it's data centers, manufacturing? Here's the ultimate problem. You've got five, six, whatever it is, companies that are trying to create the ultimate foundational model that we all depend on. It's almost like in the 90s when all the search engines were competing. Pre-Google, right? There was Lycos, Yahoo, whatever, whatever, whatever, right? There were all these different. And you didn't know if it was going to be a winner take all or a top five, right?

21:10Now we know with search engines, it's Google. And then there's Bing as like one or 2 % and DuckDuckGo has got a half a percent. And so it's effectively a winner take all. And so you've got all the, you know, Google, Meta, et cetera, OpenAI going all in, spending everything, consuming every resource that they can just in case it's winner take all. And that's going to be really scary because there will come a time where they have to live up to the economics, where they're not going to be able to just use their cashflow or lack thereof and just borrow money because people are going to start to see the business aspect of it.

21:47And that's part one. Part two is they're going to need, if it's a winner take all, they're going to need to figure out how to differentiate themselves. And you're starting to see that now that everybody's trained on everything that's on the internet, right, that they're licensing content. And I think you're going to see some of them start to become part of silos, right? So like when I go talk to hospitals and I go talk to researchers, I'm like, the days of publish or perish are done, right? If you just put your shit out in the internet, you're a moron because you need to keep all your IP and protect it and maybe even consider not even patenting it.

22:24Just keep it as a trade secret and then either use it for your own model, because I think there's going to be millions of models, right? Or auction it off to the highest bidder, right? So you saw Open Evidence do a deal with the New England Journal of Medicine and JAMA, right? So you're going to see these companies try to, you know, learn from what happened on the internet and search engines and protect their IP and then auction it off. And so how does that impact all of the foundational models trying to win the game? You just don't know yet. Should MD Anderson have their own model? Should Mayo Clinic have their own model, et cetera, et cetera, et cetera?

23:02Those are the things from the business side that we still have to figure out. Yeah, absolutely. I think this idea of data as the moat, right, is becoming more and more powerful. All right, here's a more philosophical question that was asked by Steve Song, Hi Steve, on Instagram. How do we embed AI with maternal instincts? That's a really weird question. I don't know. But I will say this, maternal instincts relative to models, the thing about AI, it doesn't know how to say I don't know. Right. It doesn't know how to say I'm wrong unless you tell it why it's wrong. and it agrees with you, right?

23:43The best it can do is set a probabilistic threshold of why it might not be accurate and determine that if there's a 20 % chance I'm wrong, I'm going to say I don't know, right? But that's not knowing that you don't know. Right. So going to maternal instincts, right? You talk to your mom and my wife talked to her mom and women talk, you know, because that's a unique experience. But how does AI do that, right? You know, love and all that kind of thing that goes into the equation. You can pretend to create it, you know, and we've seen some weird shit start to happen when people think they've recreated it, but you can't.

24:19And so just the latency of real life versus models, you'll never be able to replicate all of them. But you can use it as an information source. Yeah. All right. So when we FaceTimed a few weeks ago, you mentioned that you're really interested in founders who are really taking big swings with AI and kind of reimagining how we do life, how we go about life. So I wanted to ask you, what's exciting you most in the AI space? Less and less every day. Interesting. Really? Why? Because in every new technology swing, you get this rush of entrepreneurs. You know, when the app store came out, there was a zillion apps that everybody was raising money for.

25:01And so we're already through all the low-hanging fruit. There'll be anecdotal great ideas that are simple, that people use existing open-source and generative models. But in terms of trying to compete with the foundational models, ship is to sail. So where there's great opportunity, I think, is what happens with video. Now, a bunch of people are trying to get into the labeling business with video. I'm going to take videos of 50 people folding their sheets and making their bed, right? So a robot can do it, right? I don't think that that is anything to write home about. But, you know, trying to understand the laws of the universe and trying to figure those things out, like there's one company I looked at that I can't go into too much detail about, but they basically use their version of a spectrometer that looks at what something's made of and then looks at its physical attributes and its movement attributes and starts using algorithms and AI, try to figure out the relationships to know what comes next.