Jeremy Levine on AI Hype, Market Cycles & Playing the Long Game

26:54But for this new front door of the internet, which is a chatbot, it's like open season. And so I'm sure there's a team at Booking trying to figure out aggressively, how do we make sure Booking is as present in these chatbot results as they are in Google search results. But it's like a brand new game. Are startups doing well at that, or what's even the strategy? And so in the early days of search engines, it was how do we manipulate the search engines to show our stuff high up in the results. And there were companies like Yelp, which did it really well. And anytime you search for a restaurant in any city, the Yelp results were among the first results.

27:30That was essentially the Kickstarter for how Yelp grew. And so it became an entire art or almost a discipline called search engine optimization. And there's an equivalent of that discipline now emerging for this new platform. And like I said, we've seen lots of people trying to innovate or reverse engineer or think about how to build their business. It's a good time to be in media. It's like one of the things that does well, I think in these spots is media. So yeah, happy to have that leverage, I guess. Yeah, but the point is, it's a new channel, and it's very immature, and it's a sprint to go figure out how to crack this new distribution channel.

28:09So that's one. The second and compounding force is that there are new capabilities that you can build with AI that get consumers interested. And so now, just the latter is sort of more exciting from a principled perspective, because like, do we need another travel booking engine? We don't. They work just fine. But there may be an opportunity for one just by dint of this new front door on the internet. If you can be the one who is right behind the new front door of the internet, you can build a replacement travel booking engine for booking. But the idea that you can build this wildly compelling new functionality because of AI makes it exciting.

28:50So now the consumers are also, in addition to discovering what's behind this new front door, bumping into these new capabilities that I think will be sort of eye-opening or potentially jaw-dropping. And when you put those two things together, I think you have the conditions for new consumer companies. The manipulation of the chatbots to produce sort of a company, I mean, it sounds somewhat depressing, right? I mean, I think what's great about the chatbots is like, if I'm like, I want the best pan, like, what's the best pan? And it's sort of like, well, here's the best expensive one. Here's the best sort of medium price one.

29:25Obviously, it's rooted in reviews and sort of different websites that trust, probably Reddit to a large degree. How much do you see these businesses as needing to sort of manipulate the chatbots versus accept sort of a paradigm where they're thinking about things in sort of a different way? I mean, we sort of saw the degradation of search over time, which has allowed some of this opportunity, which is just over time it got more and more. sort of gamed and degraded in my view. If you want to think about it cynically, you can get pretty far just manipulating someone else's front door and there are a lot of SEO companies that did that.

30:01I can't remember the name of the one that was, it was just like a massive farm of crappy content that would show up. It was called demand something. Demand media. Eventually, the front door, the search engine figured it out and like, this is crap and they just killed it. Right. So you can get pretty far but you probably can't build a great business. What makes a great business is that you may do some manipulation, if you will, these front doors to get your product in front of consumers, millions of them or billions of them for free. But then what makes your business special is that the product is awesome.

30:37And the consumer is like, wow, this is great. And next time when they want that experience, they don't go back through the front door. They come directly to your service. And that's like the holy grail of consumer. Are there any early businesses that have had success? I mean, I think Guillermo at Vercel tweeted at one point, you know, that's not a broad consumer business, but tweeted that he was getting a fair bit of his business through people sort of deciding what to build in the foundation models. I don't know. Are you seeing companies where they're getting a lot of customers through these chatbots?

31:07It's just beginning, but we are and we're just trying to invest in them. So I'm not going to share the names with you because the last thing I need is even more competitors in a competitive world. But I'll give you an example of a company that we haven't invested in that I think has done this. It's a business called Suno. I don't know if you know Suno. The music creation. Exactly. And so like there, the product is wildly compelling and it allows you to do things that you could never do before or at least most people couldn't do before. And I think they're also getting distribution through some of these chatbots.

31:39I think they're probably getting more distribution now because the product is so compelling. People are telling each other about it within the music world. But as an example, you have to build something that's good or you end up going the ways of demand media. But like I said, if you have to buy all your consumers one at a time, you need to have a massive sugar daddy with billions of dollars to fund you or you're never going to get there. And that's tough to do. one you know investor craze and honestly i can't remember if you made a bet in this category though like qvc uh you know the online shopping for video category did you make a bet in that space um uh if we did it was small really small material i don't know is that over or tiktok won it to the extent it exists i guess whatnot is still sort of fighting it out do you ever read on that space because that was such like a hot category yeah you know it's funny because i I think that I'm not an expert on the Chinese internet, but my sense is that that's a much bigger thing for Chinese consumers than it has been for consumers in the West.

32:47I essentially think of there being two internets. There's a Chinese internet, and then there's the rest of the internet, and they don't mix very much. And I don't know why. For whatever reason, no one has been able to quite make it work. But I wouldn't rule it out yet. I think we've seen a bunch of people try. Nothing has really stuck. I think Whatnot is interesting. I think it's a scaled business. My sense is a lot of what's happening on Whatnot are people buying things speculatively so they can resell them as opposed to buying things they necessarily want for themselves. So that's like a different phenomenon altogether.

33:26But I haven't given up on it, but you're right to point out it hasn't happened yet. Yeah, I guess the gambling economy thesis, are you staying away from that? I mean, it's clearly, I mean, prediction markets loom super large. I mean, yeah, I think you're correct that whatnot is fueled in part by some of that same behavior. Yeah, is that a thesis you've embraced? So I think there are a lot of, if you're setting out to build a company, there's lots of companies you could set out to build. If you're setting out to invest in a company, there's a giant menu of companies you can invest in. And I think one of the criteria that I apply that not everybody does is I want to be proud of my association with the company.

34:06And so I'll give you like a classic example of a company that I admire the financial metrics on but would never invest in because I just didn't want to be associated. And that was a company called Juul. You may remember that they made vaping pens. By the way, people who invest in it really made phenomenal returns. It wasn't for me. and I think that when we look back on some of these prediction markets or gambling sites there are a bunch of parallels that make me somewhat uncomfortable that's not to say they aren't great businesses and that's not to say someone else may not be a genius for investing in them but but they're not for me and and by the way I happen to have three teenage kids two boys and one girl and I can see among particularly my sons and their friends it sort of reminds me of Juul.

34:55It's like Juul was purportedly a smoking cessation device for adults, but we actually looked at who was actually using Juul with all its various flavors. It was like teenage kids. They're essentially introducing teenage kids to a new form of smoking, which was not good for those kids. And I think in many respects, these new prediction markets, which are allowing betting on anything all the time, I suspect one of their big customer bases are teenage kids who are particularly susceptible to the addictive characteristics. Interesting. I hadn't really thought about the children angle. Yeah. They're not like gating it.

35:30So I think very recently they've introduced some KYC things, which require you to prove that you're of a certain age. But if you ask any teenager about some restriction that was imposed on them by someone or something, I can pretty much guarantee they have three ways around it. And so I'm not sure how effective the KYC stuff is. I confess to Mignor's I haven't looked into it carefully. But even if it's a modest percentage of their business, it's just not something I would necessarily want to be associated with. So you're staying away from the whole theme? I mean, I'm watching it. And by the way, maybe they find a way to eliminate that stuff entirely.

36:07And, you know, like if there were a jewel. Would you agree that prediction markets, which I agree in sort of quotes, which are becoming sort of sports gambling sites, are one of the major sort of consumer themes of the year? Oh, they've grown enormously. Yes. No question about it. I mean, I think what held gambling back was all the regulation. and a couple of really sharp entrepreneurs found a loophole around the regulation where they can make it available to everybody instantly, whether they're having to go through any of the regulatory processes or reviews. And, like, lo and behold, it's huge.

36:39So, yeah, not surprising. Do you think gambling was always in the crosshairs? I guess they did a good job of making it seem like, you know, elections and everything before it became gambling. I mean, betting and gambling. Sorry, I mean sports. Sports in particular. Sports in particular. Yeah, I mean, look, the difference, of course, as I understand it, is that they don't take a side of the bet. They're just matching bettors, so it's a marketplace. You're going to see the utility of predictions in a way that predictions about sports, it doesn't feel like you have any utility. Repricing the Vegas line, you know?

37:15Yeah. I suppose in certain instances, understanding what the world thinks on average about something is useful, but to whom other than a speculator? maybe not so clear. Any other categories or theses that you guys are really chasing at the moment? Within consumer specifically? Or no, consumer or more broadly. So the other one that I'm really excited about is robotics. And so we, as I've talked about with you and others in the past, we tend to be very roadmap oriented where we try to build a hypothesis and test hypothesis through experts about some change happening in the world. And oftentimes the most compelling ones are the simplest ones.

38:02And so, you know, by way of example, 20 years ago, as we saw the consumer internet emerging, we thought like, oh, consumers are going to contribute content to media properties online where the media properties have the benefit of being media companies without having to actually spend money on content. And that led to investments like Yelp and LinkedIn and Pinterest. And it was like a really basic idea. And once you have the basic idea, often an insight shared with an entrepreneur, you try to go find all the companies that are not competing with each other that are consistent with this idea. A really simple-minded, like one-sentence view on robotics is that there will be probably 10 ,000 times, maybe even 100 ,000 times more robots on planet Earth five to 10 years from now than there are today.

38:43And there are essentially three big challenges in robotics. One is, can you make hardware that functions as a robot? Two is, can you make the hardware move? They call that the locomotion problem. And three is, can you make the hardware manipulate things in the real world? And the hardware itself essentially works. There's tons of hardware. A lot of it is being made in China. Increasingly, companies are trying to make it in the US. So that was problem one. Problem two is the locomotion problem. And now, as evidenced by Waymo, another recent Bessemer investment, like we can make hardware move without issue.

39:21It's solved. And you see it in robot dogs. You see it in autonomous cars in all categories. And that was a machine learning problem, which is essentially solved. It isn't perfect, but it's close enough that it's usable. And the unsolved problem for robotics is around manipulation. So we have these things in the real world, but we need them to interact with objects to perform functions that in today's world, many humans do, but they're pretty crappy functions. Fold laundry. Yeah, I mean, that's the consumer one, but there's an enormous number of, I mean, there are many companies working on it and they have various approaches to it.

40:00And I'm quite convinced that the, ironically, the LLMs of today have solved a bunch of what I think of as white collar type jobs before they've solved the blue collar type jobs. And in fact, an MIT professor I met with said to me, the hard things are actually relatively easy in AI. Like, AI can do PhD-level statistics a hell of a lot better than I can. And yet the easy things, like picking up this glass, moving it here and moving it back without spilling the water, which is so easy for any human to do, is actually hard to do with AI. I believe it will be solved. And I can't tell exactly when, and I can't tell exactly what the path is.

40:38And so that is the basic thing. You know, I have a four-month-old right now, and obviously babies spend so much time figuring out, like, oh, how do I hold on to something that I think experiencing the human brain form. You're like, oh, yeah, obviously, it's a lot of work navigating the physical world. Are you bullish on humanoid robotics? I think eventually we will. We don't have to agree with it, but we were somewhat skeptical and newcomer about the humanoid form factor. So I think it's fun for humans because it reminds us of ourselves. I think the autonomy of robotic form factors will be wildly varied.

41:17And there may be certain instances where the humanoid form factor is particularly powerful. I think it will come later because there are a bunch of challenges with humanoids that don't exist in other form factors of robots. But that it's not going to get in the way of these robots being wildly helpful or productive for humankind. And so then the question is like, how do you invest in it? And we've started, we made an investment in a company called Foxglove. And if you believe there will be 100 ,000 times more robots, then we're going to need the same kind of infrastructure tooling for robots that we have for computer servers.

41:49And there's companies like Datadog that make some of the infrastructure tooling for computer servers, and we're going to want that same kind of capability for robots, and that's essentially what Foxglove does. It makes it very easy to... And so then you avoid the particulars of which robot format is going to work. It doesn't matter which robot companies or how they work, but if they have the leading solution for the underlying infrastructure so that the robotic makers don't have to go build this infrastructure software, they can just take it off the shelf, BoxCove should be a winner. So that's exactly right.

42:21So that's one area of investing. And then the second one is, OK, how do we figure out how to solve this manipulation problem? Once manipulation is solved, these robots become wildly more valuable to all of us. And so those are generally even earlier stage investing, oftentimes in academicians or inventors who are trying to solve this unsolved problem. And a bunch of companies who've raised billions of dollars. My general view is it is nice to have raised lots and lots of money so you can pay people in Mexico to mimic the behavior of a robot and record it and collect all this data. I suspect the winner will be much more innovative than these companies using brute force to collect data.

43:04But it's early. And obviously, I don't know for sure, but we're trying to find really smart innovators in this category. I want to talk about venture capital as an asset class for a second. Bessemer as a firm has been investing for more than a century, right? Is that? Yes. We came out of a family office, which has been, it's now, I think, since 1911. So 115 years. I don't know. I think Andreessen Horowitz just published a piece like defending sort of the mega firm or scale and venture capital. Yeah. What is your sense of sort of, I don't know, the rise of like the mega fund and how venture capital is being practiced today?

43:43Do you think there is just sort of, you know, you're a huge, you're a big multi-stage fund yourself. Like, how do you think about scale and venture capital? I think scale is critical in venture capital. Does that mean it's required to have a successful fund? No. Do I think it's required to have a successful firm and a collection of funds? Yes. A bunch of things have changed even since I've been doing it. So I've been doing venture capital for 25 years. And 25 years ago, if you were an entrepreneur, you found out about venture capital and venture capital firms through word of mouth and some friends.

44:17and it was a little bit of an old boys network and it was literally boys at the time. We've gotten wildly more diverse, thankfully, although it's still predominantly an old boys network. And therefore, having scale and building brand and marketing awareness and content, like none of that mattered back then because it was - Now you're forced to come slum it on my podcast. I mean, so now there's the internet and people discover and learn in so many ways. And if your firm or your brand is only in the conversation 1 % of the time and someone else's is in the conversation 10 % of the time, that's an advantage.

45:01And so I think there's this counter narrative, which I think is wrong. And that is if you want to have a 10x fund, it has to be a small fund because it's virtually impossible to have a 10x fund on a$3 billion fund. But if it's a$200 million fund, you could actually end up with a 10x fund. I think that's right, except what it misses is all the$200 million funds that are going to be a 0.1x fund. And that if you have somehow the luck or pressions to invest in the$200 million fund that does become a 10x fund, good on you. but that's maybe as much luck as it is skill. Whereas if you want repeated sources of competitive advantage, you need scale and you need to be in the conversation all the time.

45:44And it's impossible to do that as a boutique firm with four or five people who do all the work themselves. Or you're leaving money on the table. I mean, I sort of recently gave Benchmark a hard time in a story, which is they sort of have the brand and they're leaving it somewhat on the table by not investing in these later stage rounds? There are a couple of firms, you know, USV or whatever, that have built brands that if the reason you need scale is because you need to stand out in people's minds, there are firms that built brands that are smaller. There are, but I would argue that the brands of the small Boudic firms relative to the rest of the industry 20 years ago were much bigger and they're increasingly smaller, getting crowded out by larger firms.

46:29But that's just one reason why scale matters. The second reason why scale matters is because timeframes to evaluate opportunities have shrunk substantially. 25 years ago, I might meet a company and we might get to know each other over months before they were ready to raise money or we were ready to invest in the company. Increasingly, that number of months is shrinking to days. And if you want to make good decisions, you have to arm yourself with meaningful data. and even for an early stage company that has 16 early trial customers, like that's the whole business, you want to talk to 10 of the 16 to really know what's going on.

47:11And if you're in a small firm with limited resources, it's kind of impossible to do that in three days' time, whereas a larger firm with more resources is going to have talked to every customer and 10 others to understand, like, is this an easy thing to sell? And what do prospects think? And that's just an incredible information edge. And I've seen it recently. And I don't want to name names, but there was one instance where we were looking at it. It was a classic Series A investment. And I was working on it with several colleagues. And so we actually called every company that had its logo on this prospect website.

47:50And we found out that not a single one liked the product. And these were the logos on the company's website. And another really well-known, respected firm at a smaller scale than us ended up investing in the company. Because I think it was impossible. They had to sort of go on more faith and kind of make a judgment, a snap judgment. Is this entrepreneur talented, truthful, and so forth? And I think they made the wrong call. I think if they had the same data that we had, they would have made a different decision. And so I think there are other, it's not just brand and being in the conversation that scale gets you.

48:29It also gets you information. Does honesty so matter? Or I feel like there's a nihilism. I mean, it came up sort of in the prediction market piece. But, you know, Yoni Reckman in my last podcast was sort of, or in a recent podcast, described venture capitalists as sort of amoral financiers. Like, I mean, I guess you can be amoral and still think sort of honesty matters. But how much in this sort of world where, I don't know, people are playing pretty fast and loose, do you think it matters to back honest founders? I mean, look, do you want to be friends with honest people? Do you want to be married to an honest person?

49:06Like, that's a value judgment. I mean, I do. But, like, you can have your own values there. That said, I think more people than not resonate with people who do what they say, say what they do, are honest. And so I think, is it impossible to build a company as like a pathological liar? It's not impossible, but I think it's probably harder because it's harder to have loyalty and have people bought into the mission and so forth. But like I will say, there's also this age-old idea in Silicon Valley of faking it till you make it. And so could you fake it and have 15 fake-ish customers raise a bunch of money and with that money turn things around?

49:45It is possible. It has been done. But it's not the kind of gamble that I want to take. And so it may well work out for this company that we didn't invest in that now that they've raised money from a really well-known investor, they're going to use that to reform their ways and become a great company. But I'd bet against it. You know, one advantage you're saying you get with venture capital scales, you have, you know, all these people running around and doing all this diligence. Do you still need all the people or how much are you relying on analysts today versus using AI? I just published some research that Venture 5 had done and they saw that analysts at venture capital firms saw their salary fall.

50:26Like, do you think you're hiring fewer of them because of what AI can do? In the short term, no. But advantage three of a venture capital firm with scale is that we have the resources to invest substantially in AI tooling. And AI tooling is both a source of data. As a silly example, imagine if you knew any time someone of a certain seniority level left Google and changed their LinkedIn profile to be a stealth founder. That's really valuable information for a venture capital firm to have. we have tooling that tells us that for every single person. If you're a little boutique venture firm with no resources to spend on tech and innovation yourself, you're not going to have that information.

51:12You're depending on your network to get you close to those people at exactly the right time. That worked really well when that's what everybody was doing. But now when there's a group of competitors who have access to this incredibly rich information generated by their own AI tools, you're at a disadvantage because they also have the same personal networks, but they have tooling on top. And so to date, we've used our tooling to make all of us more productive. And I would say we're in inning one or two in a nine inning baseball game that probably even goes into overtime. So it's still early. But I have a hard time imagining in the short term, the people in venture capital mattering less because fundamentally companies are still run and founded by humans.

51:58And humans don't really want to interact with or build a partnership with an AI bot. They want another human. And so having people who are smart, ambitious, hardworking, and compelling as humans within our firm is really important because those are the people who have to build relationships with today's founders and the founders of the future. Because fundamentally what we do is we become partners. It's why it's in our name, Bessemer Venture Partners, not Bessemer Capital, because we're looking to partner with these people in a very human partnership to help them through the ups and the downs, solve problems along the way, be the shoulder to cry on, be the person to bring them back down a little bit when they're celebrating a little bit too much.

52:39And those are all very human things. If the venture capital industry is, you know, inevitably booms and busts over hype and under under hype sometimes. Where do you think we are? I don't know. I think we're still rooted in the dot-com story, but where do you think we are today in that sort of hype journey? I think we are, well, I mean, it is cyclical. And the growth of AI and the emergence of these companies growing at faster rates and reaching bigger scale than anything we've ever seen before echoes to me the early days of the internet, the consumer internet. But the scale and speed is an order of magnitude or more bigger and faster.

53:23And so I think we are on our way up a well-justified hype curve. Now, much like in the early days of the internet, there were hundreds if not thousands of internet companies that failed. They had the wrong model. They had the wrong ideas. They were focused on the wrong things. but got funded in what was objectively a gold rush of sorts, we will see the same thing happening here. But I think much like some phenomenal companies emerged from the advent of the internet, like Google and Meta and many others, we will see massive companies emerge from this. And I think they will be even bigger than the ones that we saw in the internet.

54:04And so I think it's a great time to be investing. It's an amazing time to be alive. And I'm so curious what some of these things will actually evolve into. But nonetheless, venture remains the same, which is that it's a long game. And when we make an early stage investment today on an idea like there'll be 100 ,000 times more robots in the future than there are now, that may not fully come to fruition for seven, eight, or nine years. And if as an investor in venture capital, you're looking for returns every quarter or every year, you're in the wrong asset class. This is a patient person's game, even with these incredible momentum ideas like AI emerging.

54:43And so there'll be some instances where things just work immediately and it goes up to the right and you can actually see and taste the value right away. Most of the time, it's a slow burn process until suddenly something appears to be really large when, in fact, it was growing a little bit along the way and just no one was paying attention to it. Jeremy, thank you for coming on the Newcomer podcast. My pleasure. Thanks for having me. Thanks so much for listening to the Newcomer Podcast. I'm Eric Newcomer. Follow us on Substack at newcomer.co. Hope you enjoyed the episode. See you next week.