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: A movie studio using AI to actually generate a movie-on-demand for Will Porteous. Because I know you like those rom-coms.
Will: [laugh].
Raju: I know you like them. You can hide all you want, but I’ll find out [laugh]. So.
Will: The infinite numbers of meet-cutes that are possible with generative AI. I guess that’s coming.
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.
Will: Welcome back to RRE POV. I’m Will Porteous. I’m here with my partner, Raju Rishi. Today we’re going to be talking about applied AI, and rather than having a guest, in the spirit of sharing the kind of conversations that we’re already having internally, I’m going to be talking to Raju about our thesis on applied AI, which is a big area of focus for our firm as we’re investing RRE [8 00:01:13]. So, Raju, to kick us off, how do you think about applied AI, and AI in general, in terms of the arc of technology inflection points that have driven value-creation opportunities and the like for the venture industry?
Raju: Absolutely, yeah. So, you know, if you look at the venture industry, we’ve made the vast majority of our money [laugh] as a result of three big inflection points. Trillion-dollar companies created lots of value to both small and big companies, lots of disruption. And so, the three technology inflection points that we’ve had in our lifetime are, first, it was the personal computer, and just, you know computers in general, the second was the internet, and the third was mobile phones. And they actually had three things in common that I think people kind of forget about.
The first thing they had in common is each of them had platforms upon which you could actually build, and that were highly leverageable. So, for the personal computer market, it was the Microsoft operating system, it was Apple OS, and it was Unix. And for the internet, it was the variety of different browsers that got created. And then finally, for mobile phones, it was Palm Pilot, it was BlackBerry, migrated to iOS, and then Android got added on top of that. So, they all had platforms.
The second big thing they had in common is that every business in the world was trying to figure out how to create value and/or leverage it. And so, you know, how am I going to use the PC to effectively improve the productivity of my people? What pieces of content need to sit on that device? And it had massive ramifications downstream. You know, for the internet, businesses were trying to figure out, what kind of presence do I need to have? Can I work with my business partners on it? What’s my relationship with customers on it? And then ultimately, you know, created this work from home kind of environment.
And then for the mobile phone, you know, the same kind of story because it was effectively taking the PC and the internet and putting it in your pocket, so you added on a layer of location-based services. So, you know, every single business in the world wanted to figure out how to leverage that more effectively. And the last thing they all had in common was massive amounts of open-source. There was a bunch of developers all around the world that were trying to figure out, and create an open-source capability for the personal computer on the variety of different platforms out there. UNIX leveraged that more than PC and Apple OS did, but frankly, it existed on all three platforms.
A bunch of open-source that we know about on the internet. Internet came about because of a lot of open-source capability. And then finally, mobile phones, there was a bunch of open-source capability that allowed you to create applications and program faster. And so, if you look at those three technology inflection points, you think about the companies that were venture-backed underneath that, and you see those trillion-dollar companies. But then you see adjacencies, applications like Uber. It’s not iPhone, it’s not, you know, Apple OS—it’s iOS—it’s not an Android OS as a platform, but it is an application that leverages it, and obviously, you know, a massive, massive company was created.
So, we were blessed—you and I, Will—to have lived through an era where we had three big technology inflection points. AI has the same three things in common, right? It’s got a bunch of platforms, Actually. It’s got, you know, Anthropic, Mistral, OpenAI, Liquid, you name it. It has a ton of open-source that’s being created around AI that has been leveraged for things like image recognition, and all sorts of things, and then every business in the world is trying to figure out how to leverage it.
The big difference between AI and the prior three is that AI is not going to be disruptive in the near term. And I say near term, that’s an important piece of qualification. I think the oil and electricity for AI is data, and right now, the big companies have that data, and so it’s going to be difficult for Goliath to come in there and take down David. An example would be insurance companies. We know all the big ones, right?
It’s New York Life, MetLife, Mass Mutual. They know about policies, they know about breakage, they know about the carrier fees, all sorts of structure around that. It’s hard for a startup to come in there and say, “I’m going to use AI, and I’m going to disrupt that industry.” They can disrupt it, but it’s going to take a while because the current schema is that those big companies own a lot of that data.
Will: You laid that out beautifully, and I didn’t want to interrupt you, but I’ve got a lot of questions [laugh].
Raju: [laugh]. Go for it.
Will: So, given that AI is not going to be disruptive in the short term in those areas, shouldn’t we be investing in the big platforms? I mean, it’s almost as if you’re saying that the winners of the big, trillion-dollar opportunity that comes with every inflection point, the winners have already kind of been identified in OpenAI, and Anthropic, and others. And a lot of our brethren in the industry are chasing those deals and chasing them hard right now. Should we be doing that?
Raju: You know, that’s a really polarizing question in the venture industry, and I’ve got a very strong opinion, which is, it is venture-investible, but best invested by corporate venture capital, not by true venture capital. I’ll give you an example. Microsoft put a bunch of money into OpenAI, and obviously Amazon has put a bunch of money into Anthropic and, you know, we kind of see that across the board. And there’s a bit of quid pro quo that goes on there, which, Microsoft says, “Hey, look, we’ll do this, but—you know, $20 billion, here you go—but we need you to spend $20 billion on Azure”—
Will: [laugh].
Raju: —“You know, over that period of time. And, oh, by the way, you know, we want some revenue sharing on that.” Traditional venture capital doesn’t have any of that. You know, what ask can I make of OpenAI? I says, “Well, you know, you got to give me leads for really good AI companies.” I mean, I’m not going to do that, right? I mean, maybe, but I don’t know how valuable that’s going to be in the end.
And so, I think the symbiotic relationship between corporate VC and these big platforms are real, and so they are probably going to make money, right? In fact, Microsoft’s almost definitely going to make money because the spend is going to be on Azure, and they’re going to get a revenue share, and they still have the equity, and if it kind of—so, you know, at a minimum, they get the money back. This reminds me a lot of the browser wars. You remember that, right? So, there was Lycos, there was Yahoo, there was Alta Vista, there was Google, and it was a bunch of different browsers out there, and people were pouring money into those platforms. And look where we are now, right?
And so, there will be one winner in that platform, and you have to hope that you, as a traditional venture capital firm, have made that bet properly. The second problem and difference—well, actually, now, let me take that back. The reality is, there’s actually something different than the browser wars in this because all of the platform companies that we have need to get trained, and the training costs on those are exorbitant, and the half life is short. So, I think, you know, we kind of all know that ChatGPT 3, when it was first created, required, you know, several million dollars worth of training costs, and you know, took a year or so. And this year it’s like, I think, $450,000 to train ChatGPT 3 equivalent, and it took, I don’t know, a few months. And shortly it’ll be $45 and it will take a minute.
And so, the deterioration of value on that spend is considerable. Now, if you’re spending all that training money on Azure, or you’re spending all that training money on AWS, right, they’re getting return value, but what’s the venture guy or gal getting? And so, I think yes, it’s investible. I think it’s better spent from a corporate venture capital arm than it is a traditional venture capital arm. And I think if you’re going to do it, you’d better be right because if five of those go away and there’s one left, I hope you picked the right one.
Will: I think your argument is very persuasive. So, let’s come back to the question of AI being disruptive because you also made a very strong point that if the oil and water of the AI revolution are data, big companies have the data, and therefore, AI is not naturally disruptive to established franchises. And so, what kind of early startups can be successful in this space, given those constraints?
Raju: Yeah. I mean, this is where we’re getting into applied AI, right? Applied AI companies, from my perspective, are using big company data and returning value to them. And they can be horizontally applied or they can be vertically applied. And I’ll give you example of both, right?
So, a horizontally applied AI company are doing things like leveraging your sales and marketing data and putting it through an AI engine and telling you what process you should be taking to close your sales prospects, or what’s the best spend from a marketing standpoint. Because it’s basically looking at prior history that you have for your product and your applicability, and then just grinding that, and giving you value back. Yeah, the big company can do it themselves, but I assure you that the small startups that are trying to do this on the behalf of many companies will do it more efficiently and faster.
And so, that’s an example. Another example is copilots for coding. There’s a lot of companies that are leveraging—that’s a horizontally applied application, and there’s a lot of open-source capability that allows you to program for free, or whatever it is, so they’re leveraging that data, but they’re also leveraging the data of those bigger companies to make the coders inside those big companies more efficient.
And then vertical is another example. So, discrete applications in industries. So, for instance, if you look at healthcare, you give me all of your radiology exams. I give you an answer, but it’s your data, it’s your radiology exams. So, I think that’s the way that these applied AI startups can be successful in the near term.
In the long-term, yeah, I do think generalized AI and generative AI will be successful, but I think there’s a lot of bridges to cross and a lot of problems to solve. I think if I was a company and I was, like, a founder, and I said, I want to start an AI company, I’d start an applied AI company. I would say, what are markets where big incumbents are trying to beat each other out, and I can work with one big incumbent, leverage their data, make them smarter, make them more efficient, and they become stronger than the other players in the industry. Maybe I can go to all the players, right? If I do my contracts properly, I’m never going to be able to, you know, limit myself from working with the competitors, but then you become arms merchants to the world. And if you can become an arms merchant to the world in AI, you can win.
Will: [laugh].
Raju: [laugh].
Will: So, that’s a fascinating set of elements to, sort of, sit between from a strategy standpoint. For the AI company to create sustainable advantage, they’ve got to not only figure out how to partner with incumbents who have the data—all the categories you name, there are other software companies today, so they’ve got to be better—much better—at innovating on AI with a partner’s data than anyone else who might come along. So, how does the applied AI company create sustainable advantage? Or are they ultimately going to be in some ways subsumed by the big platforms that we were talking about earlier, Anthropic and OpenAI and others?
Raju: No I don’t think Anthropic—I mean, there’s a lot of problems that are, you know, we got to, kind of, solve. I mean, big companies are terrified of putting their data into Anthropic, and OpenAI, and others, right, because, how do you backtrack them, right? Like, how do you backtrack that data? How do I create a private instance that my data is not going to get exposed? You know, my son works at a hedge fund, and, you know, they’re not going to use those open platforms, right, because they don’t want the data getting leaked out somehow. I went to this conference, right, called Ai4 a few months ago, and it was a really, really fascinating conference because it wasn’t the startups presenting.
Will: Interesting.
Raju: It was the large companies that were there talking about how they’re using AI. And they got a bunch of fears, right? One, data security. Is my data going to get leaked out, and then, you know, ultimately, my competitors are going to be making decisions using my data? They have fear of lawsuits from hallucinations. If I let AI do my customer service, what if it says something that is racist, you know, or whatever, and then all of a sudden I have a lawsuit on my hand.
They have issues with, like, copyright. Like, is the AI using copyright issues. And then can I get, you know, kind of in trouble around that? And then we haven’t even broached the cost and access to compute. There’s massive compute costs that need to happen. So anyway, you know, I think those are issues that kind of need to be wrestled with.
Will: Okay.
Raju: And I think the way a small company in the applied AI space can create longevity is the big companies, right, the platform companies are going to have trouble, but let’s look at a verticalized or a horizontal company—let’s just say Salesforce, for instance, right—Salesforce has access to data. It’s going to provide sales and marketing function. The bigger companies get encumbered by their own weight. They already have long-term contracts with the big companies, so in order to make the number they get accreted. They buy ExactTarget, they buy all these bigger brands and whatnot, and it becomes sort of an octopus.
And getting data out of Salesforce from all of their different components, and synthesizing into a single record for a customer or a prospect is difficult. And so, I think you got to have a little bit of a clean slate to do this, frankly, and you got to be dedicated to AI, not I’m going to be a CRM. I’m going to be your marketing engine. I’m going to be your service cloud as well. And, oh, by the way, I’m going to apply AI on top of all three of those things simultaneously. I think it’s really, really hard to do.
Will: Yeah. And a lot of implications for legacy products that are already out there, that already represent lines of business. So, you’re making a very coherent argument for the clean sheet of paper approach, and the new company approach for applied AI. What about generative AI? Is that part of applied AI? Is it venture investible? How do we think about generative AI in the context of an investment strategy in this area?
Raju: Yeah, I mean, I don’t know is the right answer. And I’m sure you have an opinion. Actually, you have a company that actually is relevant in this space. Actually, just as an aside, to give a little bit of a plug to some of our companies, I want to use—I want to take—before I answer that question on generative AI, want to go back to, sort of, the horizontal and vertical ones—which you know, are applied AI companies—and I kind of want to give props to a few of the bets that we’ve made right? One of them in horizontal space is a company called Avina.
They take all your sales and marketing data, and they basically grind it. They actually can create a model. Like, a lot of people think they have in ICP. Their data might, you know, sort of educated guess, say that there’s four ICPs because you have four different models on how you close a customer. And it might look like the same customer to you, but they take all your pre-existing data and say, “Mmm, there’s four journeys to closing,” and then they map your prospects into one of those four, and then kind of tell you this is what you should do for this particular customer to move it along in the customer journey.
Another example of how that’s getting more powerful over time is they work with more and more companies, they figure out how those customer journeys work in a more efficient way, each and every time. But they’re taking the company’s data, munging it, and creating value back to the company. OpenEnvoy is another example of a horizontal one that’s creating value. That one does accounts payable. So, you have vendor contracts, and you have a bunch of accounts payable invoices, they basically take your vendor contracts and compare the invoices and say, “Hey, this invoice is incorrect,” so you don’t have to pay it. We’re going to send an email back to the vendor and say, “Invoice is incorrect. It doesn’t give us a 10% discount like we’re supposed to get.” And you know, when they send you back the invoice, they get another 30 days to pay.
If the invoice is correct, eh, you’re going to pay it on day 29, hour 23, minute 59, and you’re going to get all that float. And that’s a huge amount of value. And as OpenEnvoy works with additional customers, they get more and more value. So, those are both horizontally applied, you know, sort of creating value by taking the data from the big companies and giving them value in return.
A vertically applied one is a company in our portfolio called Anomaly. And Anomaly, basically, is in the healthcare industry, and they work with payers, and with, you know, sort of HCPs, and basically say, look, if there’s a claim for insurance, is this one fraudulent, right? Was there a duplicate one that was created? You know, is it one where there’s an upcoding that’s gone on by a doctor, and is that doctor consistently upcoding? And they are, once again, leveraging the data of the insurance companies and providing value back.
And so, those are examples of horizontal and vertical companies that are applying AI and leveraging the big company’s data. So, I just wanted to get that out there to create a little bit more clarity for our listeners on what the definition of applied AI is, and examples of companies that could use it.
Will: Beautiful framing, and those are some of the many RRE portfolio companies embracing AI that we’re proud of. And there are many others that we could name because it has had broad implications for our data-centric companies, our computer vision companies, and others. So, for all that, as a long-time student of these inflection points in technology, what are the headwinds to wide-scale deployment of AI? What’s holding it back?
Raju: I think, you know, kind of mentioned them before, right? There is a bunch of them. And I think when you look at mass deployment, at least by businesses today—and the conference was really kind of eye-opening because the litany of companies that were there talking about their AI projects, when I raised my hand and asked a bunch of questions, they’re like, “Yeah, we got AI working in, like, a small corner, you know, a walled garden, you know, we’re doing a pilot. It’ll, you know, go through stages. We got to get rid of the hallucinations, you know, we need more, like, live supervision to, kind of figure out whether that’s an appropriate answer or not.”
And they all mentioned, sort of, four things, and I mentioned them. One of copyright issues. Is the engine that you’re using leveraging copyrighted data or copyrighted information to create the answers, if you will, or to get to the answers? In which case, you know, do we have liability for leveraging that downstream?
And the second is data security, right? So, I think the corporate side of things does worry about data security. And I think that, you know, I don’t want my data in the cloud. I don’t want it in, you know, sort of the open-source environment. I don’t want the big platforms to be able to leverage my data, and so how do I create a walled garden where I’m getting value, but not necessarily losing my IP? You know, the fear of lawsuits from hallucinations.
The biggest one is, you know—and you and I talked to a company last week that may be solving some of this—is the cost and access to compute. And that may be for the mid-markets, right, like, the bigger companies can afford it, but like, the smaller companies, compute is really expensive right now. And you and I were in Germany last week, and I had the pleasure of drinking a lot of beers with my partner-in-crime Will Porteous at Oktoberfest—fantastic, by the way [laugh]—but anyway, we met with a company, and, you know, I won’t name their name because I don’t know if they want to be on the podcast or not, but they have a bunch of IP on distributed computing. And we talked to them, and they’re basically, if you really want to solve this problem, you’re going to have to, like, say, I don’t necessarily have dedicated compute capability in this massive data center, but I might be able to divide and conquer. I might be able to do a little bit of my compute here, a little bit of my compute elsewhere, and then in a third location, and then have this orchestration layer that actually is piecing all of that information together and getting the value.
It might be a lot less expensive than trying to go to one raw, high-powered data center to get this done. So, I think those three things kind of need to be solved for wide-scale deployment. I think for narrow applications that don’t require as much data munging, great. I think for some consumer applications where you’re just using it for yourself, and you’re not selling things on it, you know, the platforms can create a lot of that value today. And you know, what do it on their dime. That’s great.
You know, if you want to create a generated image of, you know, an elephant riding a bicycle, they’ll do that for you for free today. So, you know, kind of leverage all of that, but wide-scale deployment—and when I say wide scale, I’m not talking about the, you know, frivolous consumer applications that are, you know, for our own entertainment value; I’m talking about things that you can make money off of, like a company leveraging AI for a variety of their internal functions, or a movie studio using AI to actually generate a movie-on-demand for Will Porteous. Because I know you like those rom-coms.
Will: [laugh].
Raju: I know you like them. You can hide all you want, but I’ll find out [laugh]. So.
Will: The infinite numbers of meet-cutes that are possible with generative AI. I guess that’s coming. Yeah, I mean, the idea of custom entertainment is super powerful, but whether that’s a place where that’s venture-investible is another topic for another time.
Raju: Yeah, but I know Will Porteous. You’re going to be like, I want Adam Sandler, Drew Barrymore, and I like the one where she kind of got—like, what she had—50 First Dates, was it called?
Will: Yeah, yeah, right. Yeah.
Raju: It was really funny. And you’re going to be, like, you got a take on that, you know. You’re going to be like, oh, you know, it’s remote. She’s on the moon, you know, and then just generate that film for me.
Will: All of us who were raised with choose your own adventure as a book series recognize that, like, generative AI has always been a fantasy. Anyway, so in general, Raju, just to kind of round this out for our listeners, what should companies be thinking about relative to AI?
Raju: Yeah. I would say, you know, there’s two different viewpoints on that: big companies, and small companies. And I think, you know, if you’re a startup, don’t just try to swing yourself as an AI company. And be true to yourself because I think, you know, there was a moment in time where any company that had AI, you know, branding associated with was venture invested, you know, people throwing money at it, and I don’t think that’s the case anymore, I think. But just in general, I would say, you know, if you’re not, like, born as an AI company, think about how you can leverage it for non-critical functions, right?
If customer service is usable, and so instead of having four direct customer service reps, you can have one that’s overseeing how the AI is communicating, and, you know, kind of intervene if something stupid happens, I think, thinking about it for an internal function of maybe, you know, trying to build a [P&L 00:25:29] or something like that, there’s a bunch of companies that are helping you do that in the AI space. So, I think that’s really kind of, how can you leverage it with, sort of, offshoot parts of your business? I definitely would have a strategic conversation with the co-founders or the leaders of your organization, big or small, is what can AI do to my business, and really kind of understand where the risk factors are for another company or a competitor leveraging AI, and so you want—got to head it off. So, thinking about it from that vantage point is important. I think leveraging it for non-critical functions, or, you know, like, pairing it with human functions to give yourself some leverage is important.
I think locking up your proprietary data is pretty important, which is, like, I got a bunch of data and right now I’m sharing it, or it’s not, you know, locked up. How do I get it kind of locked up, and, you know, sort of like, from legal standpoint? And if you do work with other companies, how do you jointly own the data? I think those kinds of things are worth thinking through from a legal and contractual standpoint because there may come a time where you’re sitting there saying, “I’m now sitting on this nice pile of data.” And I will tell you, like, don’t do this, which is, if I’m a big company and I have tons and tons of data, I want to load it up in the cloud, I want to, you know because a lot of the older data is just not valuable.
Will: So, that’s an interesting observation, and I think a really good topic for another time. For me, I feel like we’re moving to an era where we will define a company’s value in terms of a data layer and an AI layer that sits on top of it. And we’re treating all data as good data [laugh] and valuable data, and yet, in many cases, the learning curve of the AI is asymptotic. And so, what is the incremental data that is really valuable will become an important question, right? When are you done with training? But that’s a podcast, perhaps, for another time.
Raju: Yeah. But I do have—I’ll make a couple statement. I do have Gatling gun section for us—
Will: Okay [laugh].
Raju: Both to go through. Because you can’t do a podcast without a Gatling gun. The one area that I did not talk about in this podcast is the companies that are generating new data, and effectively are using AI for that. And you, my friend, are the king of that. So, I’m going to turn it around a little bit and maybe ask you a question or two on those companies.
And the ones I’m thinking about, we’re going to do a podcast shortly on a company called Nanit. Nanit is a baby monitor. It’s got video capabilities, and, you know, sort of image capabilities, and it’s got years and years and years of training data on how to evaluate babies sleeping in their crib. Another company are your two satellite companies, right? Which, one of them is Spire that is, kind of, getting all of this really rich, unique, proprietary data that helps them decipher temperature on the planet, and data no one else has right and can leverage that using AI or machine learning, whatever you want to call it.
And the third is a satellite company that’s taking imagery around the world and, you know, kind of creating value. Because images are useless—well, they’re not useless; they have some value, sort of a static image, one time, but the real value comes in being able to interpret changes that are happening on this planet with the result of multiple images. And you’ve been really kind of prophetic in, kind of, thinking through those kinds of investments, where these companies are creating new data, and are able to really become powerhouses in the industry. So, how do you think about data with those three companies in particular?
Will: Sure. Well [laugh], in some respects, it’s about sidestepping all of the copyright issues that you talked about earlier. If a company can generate its own data, it owns that data, it has total control about decisions about licensing that data or not licensing that data, and it can create its own vertically integrated data chain, so to speak, a value chain that takes you from the data production layer right through to training the AI. But you’ve got to have a way to collect data, and computer vision, broadly speaking, is a great way to generate a new primary data set.
And so, I think the big unlock for computer vision companies, generally, has been AI. And you mentioned Nanit, and since we’ll be talking to Nanit CEO Anushka Salinas next week, I’ll just say Nanit cameras monitor babies in over a million households today, and all of that imagery data helps the Nanit AIs understand the development pattern of babies, understand what baby sleep looks like, what baby distress looks like, what normal baby breathing looks like, and out of that comes a huge amount of predictive data, too. The AI knows typically when babies are going to stand up for the first time, and when they’re going to actually talk for the first time. We can recognize the patterns of life that are common across a large set of infants, and that is the largest store of imagery on infant development in the world, and it’s all come as a primary data set from the Nanit camera. They don’t sell that data, and all they’re doing with it right now is making the AIs better and looking for ways to help Nanit customers. So, it’s a very powerful value chain, and it happens to be self-contained, and it creates a ton of strategic advantage because at the scale that they’ve gotten to, nobody else’s product can come close to the level of insights that they have.
Raju: I love that. I love that so much. We didn’t talk about it because I was trying to kind of stay short and narrow, but yeah, if you are a company that is generating its own data, if you are a startup or a big company that’s generating its own data that is proprietary to yourself, I mean—and it has applicability in a wide range of decision-making, man, that can be super, super valuable. So, if you are a startup, reach out to us. We would love to invest in companies that have that level of proprietary capability—
Will: A hundred percent.
Raju: And we can help you. As Will has shown, he’s done with at least his three companies, if not more. So, we’re going to move to Gatling gun, William, because this needs to be done at every podcast.
Will: [laugh]. Every—every one.
Raju: Every one. It’s got to be done. It’s just, like—it’s our, like, what, a signature, a finale, or whatever you want to call it. So, Gatling gun, for uh, listeners who haven’t listened before, is just where I ask a bunch of, like, yeah, quick questions and get quick responses back from my partner and/or any other speaker that’s on there, and then I’ll chime in as well and give my answers. So, this one, the year that you think AI engine will fully replace search?
Will: Will fully replace search? Oh, wow, I think that’s, like, within the next 24 months. I think search is going to die quickly, probably the next 12 months.
Raju: Yeah. I kind of agree with you. I mean, and you know, back to the former question that you had around, you know, the different platform companies, one of them is probably going to be an outlier winner here, but then you got to worry, like, if they’re all kind of the same answers coming out, then nobody wins, and there’s no value creation. And if only one of them wins, then what happens to the other ten?
Will: Yeah. If you think about page rank, and sort of the wisdom of crowds that’s implicit in Google-powered search, I mean, AI is basically taking that to its maximum expression, but it’s going to happen elsewhere. It appears like it’s not going to happen at Google, at least today, as we sit here in early October 2024.
Raju: Yeah, true. Year people who want one can have an AI assistant that’s almost as good as a human assistant?
Will: Ooh, almost as good as a human. You know, that pushes it out quite a bit. Like, I think we’re going to have useful AI copilots for a lot of things, and frankly, I think we already do. I think many of us are generating first drafts for things, or mock-ups of things with AI copilots today, but you’re talking about a different level of sophistication, so I would say it’s probably three-plus years from now that we’ve got somebody, an AI that is really delivering steady assistant-level value, day in and day out. And, you know, it’ll be a long time before they’re human in equality.
Raju: Yeah. Yeah, I’m going to say five years, and I’m going to give a prop out. I don’t think any AI ever is going to be as good as McKenna.
Will: No [laugh].
Raju: So, you know, you and I know McKenna intimately because we have worked with her for so many years because she’s been assisting us. She’s awesome, and like, if you want to train on the perfect human, you would train on McKenna.
Will: A hundred percent [laugh].
Raju: But there’s no way any AI is ever as good as her. She rocks. And so anyway, just to prop to her. Top three tasks that you’d like AI to do for you if it was available?
Will: So actually, the first one is a really mundane one, which is relative prioritization at any given time. I think that we’re going to reinvent the whole organization of work with rapid calendaring, if you will, AI-driven prioritization, and be able to actually match up relative importance between parties or people who need to do things in a given moment in time. I’m excited for that change away from our very analog way of organizing professional work.
Raju: So, that’s one.
Will: [laugh]. Okay.
Raju: [crosstalk 00:35:24] three things, if you have two more. If you don’t have them, you know you don’t have to. I’ve—
Will: That’s the one I want most when I look around my desk. I mean, I’m already using AI for drafting things, and for research and summary, which I think gives me a lot of leverage. And then I actually think that, kind of, AI as the super assistant for hard things in your life, how to fix things, how to cook certain things, like, techniques, like, AI is going to find great insertion points to improve people’s basic quality of life and how they enjoy their life.
Raju: Yeah, I like those. I mine my three—and I need to qualify; I would love AI to do trip planning for me. And the way I would like to do it is, it needs to know me enough to do the trip planning. So, if I do a family trip, and I have, like, I’ve got 48 hours in, you know, sort of Italy, in Rome, and you know, what are the things that I should be doing? I don’t like to, sort of, be moving from, you know, here’s the top five things in Rome. I got to go here, here, here, here. I like to meander sometimes. Like, it needs to know that, right? Like, oh, here’s meander time for you, this particular area. And I’ve read all the reviews, and it basically, like, what you don’t want to do is you don’t want to meander into this particular restaurant or this particular area. It’s really kind of you’re not going to get any value. It’s going to look like a US shop or something like that. So—
Will: That’s really cool, actually. That’s that’s insightful because you want it to know your yoursel—you that well.
Raju: Yeah, and my family, frankly. Like, I would love to say, like, what’s the perfect restaurant? Like, I’ve had a family of six people, and the biggest problem we have is picking a meal at the end of the night to go to. It is, like, I don’t really like that food, or, you know, this one’s good, but I wish it had this. And why can’t we get this specialty. Just, like, you know, everybody throws their stuff in there, and the AI says, “You’re going to go here,” you know [laugh]? Like, actually, basically eliminates my ability to make a decision. Let the AI kind of decide where my family is going to eat.
Will: [laugh]. I like this AI that knows you well enough to know what you want on a trip. Do you think it could maybe find a place for us to get our lederhosen cleaned before we go back to Oktoberfest?
Raju: Never clean your lederhosen. It’s a rule. You don’t clean your lederhosen. And it’s got to have—it’s like a tree. It’s got the history from three or four years, like, the barbecue sauce from the particular, you know, bratwurst you ate, and then, you know, just a little bit of wiener schnitzel, that kind of just sat there, and a beer spill. Like, man, that’s—
Will: Definitely a beer spill.
Raju: That’s a h—[laugh] yeah. Lederhosen never be cleaned. Okay, so the second thing I would like is shopping analysis. I like to buy products, and I do my own research, and a lot of the stuff, reviews are fake, and I have to read them to figure out they’re fake. But I think AI could get through the fake ones pretty quickly, and say that one looks fake, you know, and this one looks real.
But it’s got to get smarter, but it basically says, you know, if you’re trying to buy, I don’t know, a battery pack that the Qi2 that sits on the back of your iPhone, and you want this size and this format, this much battery life, it can help me decide that. And the third one is, we all have a lot of meetings, and you, my friend, are magnificent at follow up. In fact, I don’t know anyone better at doing follow up than you. I would love to emulate you, but I cannot. I just don’t have the fortitude, and you’re just amazing. I think you think about the follow-up while you’re having the conversation, right? Like—
Will: I do [laugh].
Raju: —you’re like—yeah, I mean, maybe you don’t. I don’t know, or you have a log or something. I don’t—I sit there, and I take all the notes and everything like that, and then I do follow up, but I follow up, like, you know, too late. You know, way, way in the future, and you’re, like, immediate. So, if something could follow up right away, for me, I would love that. Okay, so what’s one task you would want AI to never do for you?
Will: Communicate to my family on my behalf. That’s scary to me. I think what the areas that AI will touch that frighten me most are really in the context of human relationships and authenticity in human relationships.
Raju: You know, I had the same answer.
Will: Yeah, good.
Raju: I had the same answer. And not just family, but conversations. Yeah, I never want it to converse on my he—I learn so much from having a conversation with someone. You know, a lot of times they’re just speaking, but, like, sometimes I’m also listening; it’s a real bidirectional conversation. And AI does that for you, it’s learning; I’m not. And that’s not going to happen in my life, ever. Okay, would you ever want to be preserved in AI?
Will: [laugh]. Yeah, that’s a really interesting question. I think… I don’t know yet. I with all the technology that we have, it’s amazing to me how little people have gone for recording their loved ones before they die, gone for filming them, et cetera. And I think that AI presents even different questions about that because you have the opportunity to represent someone without them knowing after their death. So, I think it’s fraud as a topic. So, I’m not—I wouldn’t be comfortable with it, I don’t think.
Raju: I would do it. I would do it.
Will: That’s good, because we’re going to need you [laugh].
Raju: I don’t know if they’re going to need me. I’m going to inflict myself on everyone for the eternity. That’s [laugh]—
Will: Well, it does create this possibility of you saying, “Well, let’s go ask your father.” Like, “He’s not here right now, but, you know, we’ve got his AI, and that’s pretty good.” You know, so you can start to be in more places.
Raju: Yeah, maybe. Maybe. You know, I wouldn’t mind doing it, just because I think there’s a lot of lessons in there that, you know, you’d want your grandkids to hear and stories and stuff like that. And maybe not as an AI that’s actually thinking and productive, but maybe just like the querying thing, where it’s like a library book, or, like, you know, memoir. I don’t know exactly how would want it, but I’d be comfortable with it. Okay, best AI non-robot movie ever?
Will: [laugh]. Well, for me, it’s still War Games [laugh].
Raju: Not 2001: A Space Odyssey?
Will: The computer in War Games is so… it’s so personal, and I can still hear him say, “Shall we play a game?” It’s Joshua’s desire to play a game with someone. It’s actually a very human impulse. And that, to me, is a seminal AI moment, more than HAL, kind of, locking the door.
Raju: I’ve learned my favorite lines of all time was in that movie, where that—I think it was a five star general—he goes, “Boy, I’d piss on a spark plug if it would help.” [laugh]. Because they can’t turn the AI off, you know? He’s like [laugh]—yeah. It was a great line. Okay, worst AI non-robot?
Will: Oh, I don’t know. What’s yours?
Raju: I’m going to go with Lawnmower Man. I didn’t really love it all that much. You know, he’s like… The Lawnmower guy. I mean, like, there’s so many better ones out there, so that’s all I’ve got. And you know, we can wrap it up.
I’ll go ahead and wrap this one up, Will, but I appreciate all the questions, and we both think about this a lot. So anyway, to our listeners, really appreciate you guys joining, listening to our thoughts, the conversations that we’re having inside of our walls, and we would love for you to, you know, give us five stars. Maybe—is there a six-star rating? We need a six-star rating. I think, like that’s—we’d love for you to give us a six-star rating. But anyway, until next time. Thank you.
Will: Thank you.
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.