Subscribe
Share
Share
Embed
Demystifying the conversations we're already here at RRE and with our portfolio companies. In each episode, your hosts, Will Porteous, Raju Rishi, and Jason Black will dive deeply into topics that are shaping the future, from satellite technology to digital health, to venture investing, and much more.
Raju: I have heard of interesting ways people are using it. I have a friend who, when he wants to give someone a gift—including his wife—he uses the tool and he gives it enough prompts that it gives him three recommendations. And his deal with himself is he will absolutely buy one of the three recommendat—top three recommendations that comes up.
Will: [laugh].
Raju: That is terrifying.
Will: That’s the ultimate ‘asking for a friend.’
Raju: I’m Raju Rishi.
Will: And I’m Will Porteous. Welcome to RRE POV, the show in which we record the conversations we’re already having among ourselves, our entrepreneurs, and industry leaders for you to listen in on.
Raju: Hello, listeners. This is Raju Rishi, and I’m joined by my partner, Will Porteous, and this is RRE POV. Today we’re going to talk about our favorite topic: investing in AI.
Will: [laugh].
Raju: And I know, and Will knows, we’ve touched on this in other podcasts, but you could pretty much do a podcast on this every week [laugh]. You literally could. It is moving so fast. This is an insane ride. Don’t you think, Will?
Will: Yeah, it’s remarkable. I think we both feel like it’s an incredible privilege to be an investor in this environment, to be an entrepreneur, or even an LP in this environment, it’s a great time to be participating in this reshaping of the economy that is the AI revolution.
Raju: Yeah, I agree. I’m going to start with some numbers, just so people have some context. You know, it’s hard to actually pin down how much investment has gone in an AI only because we kind of know what VC money has gone in, corporate investment is a little tougher to discern because everybody doesn’t disclose exactly the amount, but like, in 2024, from what I could see, roughly $100 billion was poured into companies by VCs. Corporate investing was like $250 billion, which is significant. And the hyperscalers, which is just a handful of companies, we all know who they are, is about $230 billion. You know, collectively, that’s, like, $550, 600 billion worth of investing.
Now, this year is not over, but Gartner is predicting $1.5 trillion in investment worldwide this year and probably $2 trillion next year. So we’re, like, doubling, tripling the amount of money that’s going into this sector. And I think we’re still searching for, like, apps, [laugh] you know? So, you know, like, we’ll talk a little bit about some of the stuff that’s working and not working, but you know, this podcast today, you know, you’re an entrepreneur, or you’re a venture capitalist, or you’re a limited partner, and you know, some people have already placed their bets, but if you’re still kind of noodling on how much, do you get into the race? Do you go, you know, full-in, or do you sit on the sideline? And I don’t know. I mean, I think you can’t go all in, and you can’t sit on the sideline. You definitely can’t sit on the sideline. And, Will, if you say, you know, completely sit it out, this is going to be a very short podcast [laugh].
Will: [laugh]. That wouldn’t be true to the things I believe. We are—you have to be in the game, but I think you have to recognize that it’s an iterative game, and that it’s iterating maybe faster than any innovation cycle we’ve ever been a part of. And it’s so profound in the sense of the way it’s already redefined the company-building process and technology and the way that, you know, vibe coding, for instance, has reshaped what software developers do every day. These changes are so fundamental to the structure of companies such that a future company is going to look very different than companies of the past, thanks to this transformation.
Raju: I agree. And, you know, I’ve got an entrepreneur that’s telling me vibe coding is out. It’s all vibe engineering now. And I understand that. You know, sort of like the engineering frameworks and getting to different discrete elements, giving you advice, and kind of piecing that together, it’s going to be interesting, interesting days.
Let’s talk about the entrepreneur first, right? I think, you know, AI is going to have profound impacts across every industry for decades. I really do, and I believe the entrepreneur needs to be in the game, but that doesn’t necessarily mean they need to start a company.
Will: Yes.
Raju: You know, my belief is—and you know, like, we’ve been through so many inflection points, you know, sort of the PC era, the internet era, the mobile era, and now the AI era, and you know, if you jumped in early in any of those sectors as an entrepreneur, you may or may not have been successful, right? In all likelihood, you were not successful, early days. And it might be easier and better to work at a startup rather than, you know, sort of kick something off thinking that you’re going to make a profound change, and you don’t have the subject-matter expertise, or the volatility is too high. You might want to lead a project within a corporation. And certainly… certainly, certainly, you want to be playing with the tools.
You know, that’s been my advice to a lot of folks that don’t have discrete domain expertise. They’ve been a startup founder before, and they’re thinking about starting something in the AI game, but they don’t have anything specific in mind. You know, it’s always a bad idea to jump into something where you don’t have domain knowledge when it’s moving this quickly, and you could be caught raising money and holding the bag because things have inflected so much.
Will: Well, Raju, I think you’re making a great point to a lot of talented people. And I think we’re going to look in a year or two, and we’re going to see on the larger succeeding platforms in AI, there are some awesome careers being built of people running very large things who would have been entrepreneurs in a different cycle, or maybe were in a prior cycle. Because it’s really a generation of entrepreneurs who are building these companies right now.
Raju: Yeah, I agree. I mean, I think the trick here is that, you know, everyone I talk to, you know, from my mom all the way down to, like, people who started companies before, were highly technical, you know, you got to play with the stuff. You got to play with the stuff. You got to be in the, you know, mix of how it’s evolving, what capabilities it has, what its limitations are so you’re experiencing it firsthand. And, you know, like, I’m playing with the tech all the time. I don’t know if you’re—I don’t know what you’re doing. I’ll give you examples of how I use it, though.
Will: I want to hear them. I know our audience does, too.
Raju: Yeah. So, you know there’s two things that are near and dear to me: one is when a hurricane is showing up to South Florida. Turns out, you know, the platforms have great alerting functionality.
Will: Uh-huh.
Raju: Yeah, do you remember this company called PointCase way back in the day?
Will: Of course. Of course.
Raju: Oh, my God. It had such dreams and ambition. It was going to take the entire internet and just give you the things that you cared about. I actually think you can do that with some of these platforms nowadays. I’ve set up alerting for two things: hurricanes and tropical storms, and great DJ events that are happening locally. [laugh] I get those alerts on a daily basis.
Will: So, let me get this straight. You actually have an AI agent out there looking for DJ events locally for you? Like, that agent’s job is your nightlife?
Raju: Yeah.
Will: That’s awesome.
Raju: No, yeah, it’s fantastic. Because, you know, you have these pop ups, and I can’t keep track of all the DJs, there’s so many of them. But, you know, if something’s happening locally, like, I got a venue alert, right? Like, I could get that probably from the venue, but if there’s, like, a cool Afro House DJ playing at a particular venue near my house, I want to know.
Will: Dude, I want to meet this agent [laugh]. This guy’s obviously cool.
Raju: [laugh]. It’s so fun. I also do, obviously, a lot of research on it, I leverage it for my writing. And you know what I did recently? I used AI to create six songs.
Will: Really?
Raju: Yeah. Yeah, yeah. I basically—I’m not going to tell you the application; I don’t want to advertise anything—but I just gave it prompts, and it created lyrics, it created the song, in a genre that I picked, and it created two variants of each song. And I’m telling you, well, first of all, I own the rights to the song. I can actually monetize these songs if I want to.
Will: All right. That’s awesome.
Raju: I’m not going to. I got a day job.
Will: [laugh].
Raju: But when we have more time, I’m going to play them for you. It’s unbelievable. These songs are fantastic.
Will: That’s great.
Raju: I don’t think it’s going to be—like, you know, I don’t think it’s going to crush the music industry at all, and I’ve got other logic points around that, but most of it is because, if I create a song for myself, it doesn’t necessarily mean anybody else cares about it. And everybody’s going to have their own song. But there’s a reason why, on Monday morning, we all showed up at the water cooler and we started talking about Game of Thrones that happened last night. Because people like to talk about, you know, this group experience, and so music will always have that, you know? Hollywood will always have that. There’ll always be some mechanism where the world has watched something that was unbelievable, and we all want to talk about it.
Will: Actually, I think what you just said is really reassuring to people, and I agree with you. Like, the culture needs to rally around certain pieces of entertainment, and they will capture the global Zeitgeist in a moment, and we’re not losing that as a population.
Raju: Yeah. I totally agree. I you know, AI is—I think the music industry will be impacted, and by the way, our next podcast is going to have a great human being, Mauhan Zonoozy, who is one of the world’s expert on music and has created a digital music company called A Vinyl Shop in Shibuya. Actually, it’s A Vinyl Bar in Shibuya, and we are investors in the business. And he’s going to be on and we’re just going to talk and riff around how the music industry is going to be impacted by AI and kind of what he’s doing and his, you know, sort of journey.
But I believe that, you know, music and entertainment, you know, we still have this group philosophy. You know, humans are pack animals, even if we hermit ourselves for a little while. We like to talk about things as a group. We like to see things as a group. We like to, you know, just riff on ideas as an entity. And so, you know, having AI create your favorite song isn’t necessarily going to make it a top 50 song. It probably will never be a top 50 song, but you will like it. It’ll be number one at your list, but a whole bunch of other ones are going to be on there that the world likes. So, I do that. I use it for photo editing. Have you ever used the AIs for photo editing?
Will: You know, not for my own photos, no. I mean, I’ve certainly played around with things like Midjourney for image generation, and we can talk about some of that later, but not [unintelligible 00:11:39].
Raju: Yeah. Yeah, no, I haven’t played ar—I mean, created images is one thing, but, like, I sometimes I take really crappy photos, Will [laugh]. I just, I need to just change them a little, and software does that. It’s really nice. Nobody knows. Nobody knows. Anyway. And then obviously I use it to figure out my podcast ideas. So, [laugh] yeah, yeah, it said you should talk about AI again. So, we are [laugh]. Anyway. How are you using it? Are you using it at home for other purposes, or anything interesting?
Will: I mean, I think my experience is not dissimilar for a lot of people, in that AI is, for me, the research assistant that I want by my side at all times. And you know, my brain runs on data, particularly as an investor, and AI informs my thought process on an ongoing basis about my portfolio companies, about the most important holdings in the RRE funds. You know, as I think about the journey of some of our companies, I’m constantly using AI to read filings, to digest board decks, to do things like that, to kind of make sure that I stay on top of all the information that I need. So, it’s a huge leverage point for me, the akin to having the smartest research assistant I ever needed.
Raju: Yeah, that’s awesome. And I kind of use it the same way, but also for, like, personal things like alerting and, you know, sort of photo editing and music creation. People use it unwittingly, like Siri, Netflix recommendations, you know, home thermostats, that’s all AI. People don’t even, I mean, most people do know, but some people don’t.
I have heard of interesting ways people are using it. I have a friend who, when he wants to give someone a gift—including his wife—he uses the tool and he gives it enough prompts that it gives him three recommendations. And his deal with himself is he will absolutely buy one of the three recommendat—top three recommendations that comes up.
Will: [laugh].
Raju: That is terrifying.
Will: That’s the ultimate ‘asking for a friend.’
Raju: I mean, just think about that. Like, what if you put the prompt wrong, and all of a sudden you know, you’re buying, you know, something radically bizarre that is so outside this spec—
Will: I think he’s holding himself to a very high standard.
Raju: Yeah, I don’t know, man. I think he likes risk.
Will: Yeah, yeah [laugh] this isn’t the Magic 8 Ball we’re talking about. This is—[laugh].
Raju: Yeah, yeah. I know it’s totally true. I know people who are learning languages through it. They’re like, you know, having ChatGPT or other tools, like, sort of emulate a native Spanish speaker or something like that. I got to try using it to plan my workouts. I am getting a little bored with my routines, and I would love to input my routines into, you know, one of the different engines out there and have it just get a little creative because I’m sure it can do that really well.
Will: That and nutrition are such a great personal frontier for AI. And I think that’s a place that you’re going to see a lot of jobs streamline around what AI can do in that conversation for us personally. Because once it’s learning from and consuming all of my data—and, you know, back to a prior episode, you know, I have 20 years of heart rate monitor workout data [laugh] to feed to some AI at some point then you’re into a whole different kind of a potential conversation.
Raju: Yeah, absolutely.
Will: [crosstalk 00:15:20] performance.
Raju: Absolutely. I know people who are using it to construct bedtime stories for their kids. I used to do this, dude. I got to tell this story. I used to come home—and I didn’t use ChatGPT because it didn’t exist when I was, you know, having small children, I had, like, I think a seven-year-old and a nine-year-old. Oh, no, I think it was like eight and six, so—it doesn’t matter.
And I invented a story, and I would go to my boy’s bedroom and my wife said, “You got put them to bed. They want to hear the story.” And I was big into, like, you know, fantasy novels and stuff like that, so I read a bunch of Tolkien and a bunch of, like, White Gold Wielder and some really cool fantasy novels. And so, you know, a lot of these ideas were percolating in my head. So, I would go in there, and I would start this story about a druid, and, you know, he discovered his powers.
What I would do every day is I would weave lessons into the stories that I thought were important for my kids to know. And really, really weird way of, like, if you tell your kids you know, like, don’t steal or, like, don’t abandon your friends. Like, be loyal to your friends, it’s one thing, telling them that, and it’s a totally different thing, weaving it into a story.
Will: Yeah, yeah.
Raju: And I used to do that, and every single day, they would be waiting to, like, hear the next iteration, the next chapter. And one time I got really bored with the character, and so I was like, I’m going to—this was on the fly. I didn’t, like, rehearse this. I didn’t think about it. I just got in the room, and I kind of came out, and I was like, “Okay, I’m going to kill him.”
Will: [laugh]. Oh no!
Raju: It’s going to be this royal battle, and I’m going to, like—and he’s victorious, but in the end, you know, the last gasp of life so I could start a new story. And I told the story, and the kids started crying.
Will: Of course they did.
Raju: So… and I was like, and I was like, “What did I do?” And they were like—and I was like, “It’s fine. You know, like, he had a heroic ending. He did all the good things. And he’s got a kid.” And, you know, blah, blah, blah kid—
Will: They were totally invested in him.
Raju: They were totally invested in him. And the next morning—they didn’t sleep. It did, like, not accomplish the purpose of sleeping. They woke up the next morning, and then in the morning they said, “[Rienta 00:17:40] is not dead, is he? He’s not really dead is he?” And I was like, “No, he’s not dead. You have to wait for the next chapter.” And I had to resurrect him the next day, out of the grave.
Will: [laugh]. You had to keep going, yeah, yeah, yeah.
Raju: Yeah. But I did weave the lessons. But anyway, this is about AI. It isn’t about Raju’s brain. So, I think that, you know, my recommendation—I think yours is the same—is being in the game, being around the hoop, playing with the tools.
You know, like, if you’re a coder, you should be figuring out how to, you know, experiment with vibe coding, vibe engineering. If you’re a photographer—I mean, every single profession, anything you want to do, test how AI improve [unintelligible 00:18:22]. And, you know, you may be, like, a purist in many ways, but I think this is a reality we’re going to have to just come to grips with. And if you don’t know the tools, you’re going to be behind the eight ball.
Will: Yeah.
Raju: So anyway, totally different story for investing.
Will: Yes.
Raju: Totally different story for investing so much money is going to be made and lost in this time horizon that it’s—I can just forecast. I can see it in front of my eyes.
Will: Well, I think we both can, and as to two guys who’ve been investors through a lot of cycles, I think we both are seeing things right now that beg an essential question, right? The elephant in the room, in many respects, around AI, is valuation of a lot of companies. And it’s something we have to contend with a lot as true early-stage investors who are often there alongside an entrepreneur at the founding. And a lot of this acceleration in valuation is justified for exceptional platforms, exceptional teams, businesses that are actually executing really well, but it also carries with it that natural investor instinct to price things to perfection, priced for perfect execution, priced, in some cases, to take over the world, right now.
Raju: Yeah I mean, I don’t want to disparage any company, frankly, but like, you know, OpenAI has a half a trillion-dollar valuation, and we just—like, I don’t know how—was it, like, two or three years ago, we had our first trillion-dollar company. And I think it was Apple.
Will: Yeah.
Raju: And everybody had the phone in there po—I mean, like, 80% of the world was carrying their product in their hands. And they were wildly profitable.
Will: Yeah. You know, there are only two companies that have ever gone public with a market cap of $100 billion. One was Saudi Aramco, and then Meta was $80 billion when it went public. And I’m the first person to believe that we’re into a fundamentally different era, but we’re talking about a company at five plus x that in the private markets right now and going.
Raju: I know. It’s absolutely bonkers. So, you’re right; the elephant in the room is evaluation. And I’m going to—I told this story on another podcast, but I’m going to rehash it because I think it is a really important thing for people to at least think about. You know, look, I could be wrong, but I don’t think I am, okay?
And you know, when you look at the four big inflection points that happened in our lifetime—PC, internet, mobile, AI—they played out in very interesting ways. And the PC and the mobile sort of inflection points played out in much the same way. You had hardware platforms upon which applications sat, and the platform companies actually minted money, minted money. And the ones that won were obviously Intel, Microsoft, AppleOS, Unix, and then the apps were like Adobe and Borland and Macromedia and PowerPoint and all those kinds of things, Lotus 1-2-3. And then for mobile, it was PalmPilot, BlackBerry, iOS, Android, Nokia, and a bunch of apps, Uber, Yelp, and everything else that you know about in the App Store today, the games.
And you know, the platform companies, if you won those, they were vital, and they won the war. They were—upon which everything sat. Internet and AI are very similar in that they’re both query-based architectures. You ask a question, you get some feedback. Sometimes it’s generative, and sometimes it’s just whatever I’ve found, the results I found, and you and I all know the platform companies that people poured money into were companies like Excite and Lycos and Yahoo and my—[laugh]—Ask Jeeves and—
Will: The names of our youth, yeah.
Raju: Yeah, Netscape and Firefox and Mozilla. And there was an article in the paper that you and I saw, which was, “The search engine wars are over. Alta Vista has won.” And that was a year before Google was born. And Google had page rank and a business model, and, you know, billions of dollars were poured into those platform companies, and they tried to out innovate one another, and they were building kind of the same thing, you know, it was kind of the same thing, slight differentiation.
And you look at the LLM companies OpenAI, Anthropic, Mistral, Liquid, you know, Grok, Perplexity, whatever, and they’re all trying to innovate around the same thing. And we have billions of dollars. And you know, it costs a lot of money to train these models, and when the new version of the model comes out, there’s a lot of throwaway. I don’t want to call it a complete throw away, but you know, it’s kind of not used anymore. And you know you have DeepSeek, and you have, you know, the other innovations that happened in the Middle East, where the training was $6 million bucks instead of hundreds of millions of dollars, and people get a little nervous. So, I don’t know if that war has been won yet.
Will: No, I think we know that it hasn’t been won yet. And you’re making such a crucial point, which is to try to keep your bearings as an entrepreneur and an investor in terms of where you are in the cycle, in terms of the—because we’ve both seen these once-valuable tech stacks collapse, and they’re collapsing quickly. They’re being overrun. And the way you talk about the capital invested in prior generation models versus more recent models is a great expression of this. What was once costly is now cheap and free, and we see this over and over again.
And so, we need almost to get to a settling point. And right now, we’re in a mode where I think a lot of people feel that OpenAI and a couple of other platforms are simply going to extend their tentacles into everything. And frankly, I think there are people for whom that is just the big blind bet, which is—
Raju: Well, so that’s the thing. I think three or four things need to get ironed out. I really do. I think one is, you know, you can invest all of this money in, you know, these LLMs, but you know, is the way we’re doing that, you know, archaic, compared to how it’s going to be done in the past? And can you do it for a much cheaper and much more efficient way downstream, and so pouring a lot of money into the LLMs and expecting a result could result in, you know, a loss on your money?
The second thing is, I don’t think we necessarily have a business model yet for AI, right? We don’t know what the costs are really going to be to grind the data, and we don’t know what the costs are going to be to generate a result, or a video, or, you know, an image. And, you know, there are companies that are coming out, sort of, executing on those models, and if the underlying costs go up rather than down, you know, their business models are flawed. So, I think that, like, you know, one of the things Google did effectively is, not only did they have page rank, which was a more—like, it was a beauty of an algorithm, but they also had a business model, right? They said, like, we’re going to give you the result, but we’re going to give you a bunch of ads. And they knew how to monetize those ads.
And initially, those ads were—they got money on those just because they were painted, right, you know? And then if you clicked on something in the middle of the page, you got, like, ten more ads. And if you clicked on an ad, that was even more money, like, it was, like, just like they had a monetization schema. So, you know, agentic AI doesn’t care about ads, you know, so are we going to rationalize on, I look for a result, and maybe I do give you a recommendation, but it’s one, right? Like, you should buy this, like, product, and that’s why my friend was doing the gifting idea. He wanted to see how good it could get, you know, at recommending gifts. But my issue is, like, there’s a lot to still figure out on the LLM efficiency side of things, and there’s a lot to figure out in terms of the cost structure, so building companies in those spaces are still risky.
Will: Enormously risky, yeah. I mean, I think we both know that the unit economics are far from figured out. And actually, you know, our listeners should take as a signal the enormous capital raises that are being done as a reflection of the fact that the unit economics are far from perfect on supporting the activities of these large scale AI models.
Raju: The other risk, Will, I’ll just point out, and I tell people this all the time, is that there are companies being built on top of these LLMs, and the LLMs are actually subsidizing the engine. And so, you know, if it comes down to it, you know, at the end of the day, these companies have to be profitable, and they’re like, “Hey, I charged you X for a token, but that price needs to go out up because I was subsidizing it.” Now, you have to go back to your customer as an application company that’s layered on top and raise the price. And if that happens, you know, is what you’re selling worth what paid for? And can you do it? Can you actually get out of those contracts?
So, you know, that’s why I think it’s a little bit of a domino game. And I don’t want to be a naysayer, and I know you don’t either because we, you and I, are both investing in AI companies, so you know, why would we say one thing and do another? And the reality is, there is a way to play here, right? There is a way, and the way to play this is actually, like, it’s the difference between a dotcom bust and a successful company, right? And, you know, I’ve expressed it before, there are four characteristics of the companies that I like to invest in. Virtually all of them are verticalized AI companies. So, in other words, they’re specific to a particular domain, and usually the team has, you know, profound knowledge of that domain.
But the four characteristics—or at least mine—are they’re embedded in the workflow. And let me unpack that. When I say, “Embedded into the workflow,” it means that it takes effort, a little bit of effort, not, like, a ton of effort, but a little bit of effort, to weave through the working processes of that end-user, of that enterprise. Because what’s a little bit tricky to get in often becomes really hard to pull out. So, what you do is you get stickiness.
If you’re not embedded in workflow and it’s just a search engine bar, well, it’s pretty damn easy to use a different browser, right? And so, like being tethered into the APIs, tethered into the workflow is important. So, that’s one thing I say. The second is being LLM agnostic. We don’t know who the winner is going to be right now.
I mean, there’s conjecture on who the winners are going to be, but nobody really knows. Like, nobody knew BackRub slash Google existed. That was the original name of the company. I love that. You know, people don’t know that. That’s a great party game. So, BackRub came out, and people were like, you know, like, they didn’t know page rank was good. All of a sudden it took over everything. And if you had rationalized on Alta Vista or Excite or Lycos, you know, not that you could have, but if you did, [crosstalk 00:29:37] done.
Will: Yeah.
Raju: So, being LLM agnostic. The third is, I think the world—and I do a lot of homework on this, like, in terms of enterprises adopting technology and everybody’s playing with it, but I don’t know if a lot of people are going live with it. And part of that is, you need a rapid ROI. You know, it needs to be immediate, or near immediate because I think at the end of the day, you know, we don’t have six months to train this thing, you know, or train our own variant of it, so your value creation needs to be quick.
And my last characteristic for at least the ones I pick, is the customer—the value is so profound, the customer is actually willing to do the training. So, you don’t have to hire your staff to do it. And this all gets to the business model piece that we were talking about.
Will: Completely.
Raju: So, those are the four characteristics, chunky characteristics, that I got. And then I know you have one that I agree with. Like, you got to get a team that is exceptional. Like, they know the space, they are subject-matter experts in what they’re building, but also they have some humility that you don’t need to get a valuation that is so high off the charts that you will never, no one will ever make money in the company. You have to thread the needle and get it perfect to make money.
Will: Yeah, you and I both believe that really profoundly. And another way of saying it is that you want people who are as interested in building undeniable, objective value along the way, in addition to seeking the big prize at the end. You know, you need to build a business now, in addition to growing the value that you can extract through AI from whatever data you’re activating on it. That’s been true, actually, of a lot of the things that we’ve backed in computer vision, which is really, you know, a meaningful part of the history of AI. Our companies that have been applying computer vision really took advantage of the emergence of convolutional neural networks more than ten years ago, and you take a business like Nanit, our smart baby monitor company, and they built a business with a smart camera-enabled baby monitor, while they grew this data set and trained this AI that now provides all these powerful insights to parents. And it’s the largest store of imagery on infant development in the world. It is an unmatched data set against the problem of what’s going on with this baby who can’t yet speak for themselves, and what advice can we give to parents who don’t quite know what to do?
Raju: Talk about immediate ROI. That’s an immediate ROI [laugh].
Will: Yeah.
Raju: Should I pick up the baby or should I leave them in the crib?
Will: It turns out that faced with a choice between asking your mother or mother-in-law what to do or hiring a nanny, that asking Nanit is a much better way to go. So, that journey, though, of kind of from proprietary data to real AI-powered insights has been a long one in that world, and similarly so in the satellite imagery sector, where we’ve also been active for about ten years, to train models over and over again, to understand what normal looks like in a place. So, there’s a famous sort of problem in national security, of no one knows what’s going on in North Korea. So, a lot of satellites are trained on looking at Pyongyang every day. And there’s this classic case of, like, will you know when what’s happening when they’re when there are no people in the center of Pyongyang, which occasionally happens. They declare a national day, and suddenly there are no people. And we find out after the fact that there was a national day. AI models are still learning and not yet quite trained to recognize things like that.
Raju: That’s so interesting. I actually do, like, you know, the four, you know, characteristics of the company are really important to me, and I know us, RRE, but I do think there are categories of spaces where the AI tech will apply itself first. And I think imaging analysis, well you know, you’ve been doing it for so many years in the satellite industry, and you’re probably, like, one of the experts in the world in that sector, it’s definitely going to play out because—and prob—whether we call it, you know, like AI of yesterday or AI today, that sector has been leveraging AI for a long time.
Will: Well, in all imaging categories, we’re confronted with a basic challenge right now, and I’ll digress on this just for a minute, which is the generative AI problem enabled by, you know, the tools we’ve seen in things like Midjourney, that can be applied to certain high value imaging contexts. So, how do you differentiate one satellite image as authentic from another? How do you differentiate one x-ray image from another and confirm what’s authentic?
Raju: Oh, my God. People are feeding, like, fraudulent imagery into the system?
Will: Well, there’s concern about it from a national security standpoint in our global satellite architecture of how do you confirm the provenance of an image? And this is a classic information security problem in spaces where you and I have spent a ton of time. How do you establish authenticity and non-repudiation and confirm sort of the point of origin on a piece of information? And we’re now looking at down to pixel level digital signatures on satellite imagery as a way to ascertain original production from an asset in space. Because in the information wars, nothing would be better than to be able to have your adversary’s satellites take a picture and for them not to see anything [laugh].
Raju: That’s super interesting. Well, it’s trickier than I thought. You know, I feel like medical imaging, you know, diagnosing fractures or medical radiology exams and satellites would be sort of really good applications that, you know, machine learning and AI could, you know, kind of narrow in on. But—
Will: And they are.
Raju: I didn’t think about fraudulent actors. Wow, that’s nasty.
Will: So, we have to think about data integrity and authenticity as essential problems in the imaging area as a deep fake [unintelligible 00:36:05].
Raju: That’s interesting. So, I actually think I wrote this down as something to talk about. I actually think some of the picks and shovels that, Will, are great potential investments, like, ETL, right? So, like, transforming data, or authenticity of data, you know, I think they’re going to be needed in order to make sure the data set that you’re using is relatively clean. And maybe subsequent to that, the image analysis.
I also think, you know, areas that are heavy into data science today, like, fraud, risk management, credit scoring, natural opportunity for AI to, you know, take a guess. But you know, even there, you have to look at potential fraudulent data. So, the picks and shovels companies will have to go alongside of that cybersecurity, where you do vulnerability assessment. So, I think those are sectors, you know. So, just to our listeners, so you understand, like, we are bullish on AI in a massive way.
We are not bullish on incredibly high valuations that you need to thread the needle to, you know, kind of get a return of any consequence. We are not bullish on areas that are, you know, very volatile. We are definitely bullish on areas where you have a verticalized expertise, you are embedded in workflow, so you’re really tethered into the customer’s environment, you are smart enough to be, you know, sheltered from the, you know, changes in the LLMs and the changing landscape of LLMs, you have a relatively quick ROI, and you have an application where the customer is willing to go on the journey with you and help train because the value is super immense. And, you know, like, you can come to our site and see some of our postings, and we’ve invested in, you know, probably a dozen companies in the near term that are all fitting that bill. And you know, our latest one is more in the customer service side with voice agents, which, you know, we’ll have on the podcast soon.
And then we have some consumer ones coming up that are all around music and in other sectors. But this is going to be a really, really interesting era. I love it. I love a little bit of chaos. I think opportunity is created out of chaos, and certainly for venture capitalists and LPs and entrepreneurs. But you know, you can’t, like, just jump in, in a volatile environment, and expect that things are that the floor is going to be stable. And so, I guess that’s the main lesson that I think we wanted to share today. But you know, anything else you want to sort of mention around this, Will? Like, things that people should, like, keep their—
Will: No, Raju. As usual, that’s very well said. I mean, I think our overarching message is indeed, kind of, keep your bearings on where we are in this cycle, look before you leap, and recognize that there’s a lot of change still to come on this long journey, and part of that will be the classic cycle of creative destruction that underpins every great innovation-driven wave. And as we know innovation is the lever of great wealth creation in every cycle. So, we just need to remember, we’re probably still early in the cycle.
Raju: Yeah. Yeah, yeah. Yeah. That’s [unintelligible 00:39:36]. That’s a [unintelligible 00:39:37]—that’s, that’s… it’s great, but it’s early [laugh]. So, amazing.
Will: Yeah, great. Thank you, listeners. Great having you with us for another episode. We appreciate your loyalty, and we’ll be back to you soon.
Thank you for listening to RRE POV. You can keep up with the latest on the podcast at @RRE on X or rre.com, and on Apple Podcasts, Spotify, Google Podcasts, or wherever fine podcasts are distributed. We’ll see you next time.