Keith Rabois on Why the AI Bubble Fears Are Overblown

0:00Hi, I'm Connie Loizos. And this is Alex Gove. And this is Strictly VC Download.

0:15Hello and welcome back to Strictly VC Download. Today's conversation was recorded live at our Strictly VC event in New York on Thursday night, September 10th, where I sat down with Keith Rabois, managing director at Coastal Ventures. I don't get to talk with Raboy very often and I always enjoy it. I'm always stunned by how fast he speaks and how quickly he covers as much ground as he does. In this conversation, we get into his bullish outlook on AI. More specifically, as you'll hear, we discussed what he thinks of former anthropic researcher Jacob Coxon's recent comments about the dangers of AI and he doesn't think much of them.

0:54We also talked about why he's spending more time in New York, whether he approves of companies that use annualized revenue numbers to land press, even as these numbers are highly unreliable, and we spent some time on what Raboy looks for when deciding which founders to back. That sounds easy enough, but in a world that's rife with fancy degrees and people who are at ease telling a good story about their vision, I do think it's a lot harder than it looks, and it was interesting to hear about one of his hacks on this front. In any case, we hope you enjoyed this conversation with Keith Raboy as much as we did.

1:28If you missed this Strictly VC evening, we hope you can make it to the next one. And either way, we will see you back here next week.

1:38Thank you so much for coming. Pleasure to be with you. It's been a long time. It has been a long time. We didn't have an agenda, and you're always so great at your game for anything. I thought we could talk about maybe the news of the week, which has me freaked out, candidly. Which day of the week? I know. I guess, let's see, Monday, what is it? Was it Thursday? The former anthropic researcher, who was also at OpenAI previously, said on his way out to the door of anthropic in a tweet storm, these companies are gambling with our lives. After which, I'm sure you saw another anthropic researcher weighed in and said, yeah, he's not kidding, and we think there's like a 10 % chance that it's going to kill all of us in, what, like, I don't know, soon.

2:20I've been really unsettled by that. What are you thinking? What are your friends thinking or talking about in reaction? I mean, personally, I'm a tech optimist, and I don't think you should be in this profession unless you believe that tech is this magic wand that improves society. And so that's what we do. I give people capital and sometimes useful advice and ask them to take the magic wand of tech and fix things. And anybody who doesn't believe that tech is generally good and that tech can improve all of our lives, our kids' lives, my kids' lives, shouldn't be in tech. They shouldn't be investing and they shouldn't be an entrepreneur and they shouldn't be working at a company that's building things through tech.

2:59Fortunately, there's enough talented people that believe in the optimistic view of the world that almost all human progress for the last four centuries has been driven by technology. Now, it's more recent, the last 50 years, that's been driven through what we think of as startups or in venture capital fueled startups. But all human lives have been improved through tech, whether it was the Industrial Revolution or the railroad. Everything we take for granted is that we're not toiling on the fields from like 4 a.m. to whatever time and that we have choices in our life about moving around the globe or moving around the country and marrying people we like first people were forced to by our parents, is all a function of tech revolutions.

3:43And this is going to continue. And then also the zero defect philosophy just doesn't make any sense to me. If you go out the street right now, there's a non-zero chance that some Uber driver runs you over. Several people in this room sometime in the next 20 years, unfortunately, will have something like a heart attack. You can slip in your shower. Like living your life thinking about everything that can go wrong is just a nonsensical way to live in my view. And so I also don't believe that some chance that something goes wrong is a reason to hit brakes on everything. I think you can be thoughtful, careful, intelligent, but we need to improve the world.

4:24And if we don't improve the world through technology, other people that we don't like, and we don't like their values, and we don't believe that they celebrate or recognize human dignity, are going to take advantage of technology, and we're going to wind up in a world we don't like. This is the lesson of every history. And, you know, some people superficially allude to, you know, kind of World War II, and they watch some dumb movie called Oppenheimer. And you're like, oh, my God. The lesson from Oppenheimer is, first of all, what would the world look like if Germans had invented a nuclear bomb before we did?

4:55That would not be very pretty. And it could have happened. Fortunately, most of the German scientists who were any good were Jewish. and they were either forced or consciously chose to leave the country or it would have confronted a very serious problem. Or what would the world have looked like if the Soviet Union had nuclear capabilities way before the United States? The other thing about nuclear capabilities is arguably the last 50, 60, 70 years, 80 years of the modern era have been more peaceful than at any period in human history. As a percentage of the population, people dying through conflict has gone down after the advent of the nuclear bomb.

5:35So things that people think are potentially catastrophic, they're potentially threatening, could exterminate all people, etc., have turned out to basically stabilize the world and make it more likely that you're going to live a normal lifespan. So again, I wouldn't overthink this. The last thing is the worst people in the world to sort of navigate this is often asking the government to intervene. Like, I don't trust the DMV. I don't trust the government with my health care. I wouldn't trust the government with anything. And so I know most of the people who serve in the federal government, actually, and the executive branch, certainly.

6:07And I like many of them. But would I rather take 10 people from Silicon Valley or 10 people from the executive branch? It's not so obvious who I'd want to make decisions about the future of mankind. Yeah, and I'm sure a lot of people here would agree with you. But I do think that we're talking about something that's unprecedented. I mean, I think how am I going to say... Well, obviously, also the details matter here. This is a six-week, this is like basically an intern. He worked there six weeks, most surely funded by China. I think that's all starting to play out. So there's an ideal... Is there six weeks this recently?

6:37Yeah, he literally worked at Anthropoc for six weeks. How long did he work at OpenAI? I don't know. That's interesting. I did not know that. But what I was going to say is, I wonder about how to talk to my kids about this. And what's difficult for me is the fact that the founders of these companies, Sam and Dario, say also there is this risk. Well, I think they just need to communicate that. First of all, Sam, particularly, has got much better at this. And AI is going to transform so many positive things. You're seeing this play out in medicine and bio. But the net effect also is very important.

7:06Like the way I think about it, I have a five-year-old, actually boy and girl, but the boy is into rockets. I believe that by the time he's finishing high school, 13 years from today, his high school project will actually be put a rocket onto some planet, like the moon. The rate of progress and the ability of access to information, that's going to be his high school project. And think about what the world's like when high school kids are putting rockets to the moon. So, okay, in terms of regulation, there's different flavors. You know, Anthropic thinks it needs to be regulated by states, which are a little bit heavier-handed.

7:41I think Sam has said basically he wants for the federal government to take a look at what they're using or developing in return for not regulating it more heavily. But you think zero regulation? I mean, what you must think. For the most part, I think there is a transitional moment right now that is being dealt with reasonably well by the federal government and some of the leading research labs, which is the ability to crack things that we thought were secure. You can talk about different versions of this, but fundamentally everybody five years ago was worried about quantum computing, basically undermining all the cryptography, and that still may happen too.

8:18But it was going to take a long enough time that most systems now, They're kind of pseudo-quantum computing. The same thing is going to happen to some extent in the short term with AI. The offensive capabilities and the ability to crack things will make systems vulnerable. That said, the defensive uses of AI actually may turn out to be superior. The reason why things can be hacked is usually a human flaw. Someone who is building a system didn't think about certain interactions or corner cases or made a pure mistake. AI-based code and development should eliminate those and reduce the surface area for vulnerabilities, or at least you can make a pretty strong argument.

8:55So as the offense and defensive capabilities sort of normalize, it's not clear that things are going to be much more vulnerable to attacks in the way people are thinking about vulnerabilities right now. This always happens in military. It's playing out with drones right now. Right now the offensive capabilities of drones are exceedingly the defensive capabilities for the most part, certainly if you adjust and normalize for cost. Where that stabilizes, nobody knows. and this is the battle through any potential technology that has military applications forever. And so I wouldn't assume that just because today it opens up some exploits, that it's going to drive more offensive interactions than defensive over time.

9:34I hope you're right. I also just wanted to ask... That's why I'm an optimist. I wanted to ask, speaking of that, you're a friend of Sam Allman. Your firm is the first institutional investor in OpenAI. I saw the trailer for Artificial, this week, and it looks bad. I mean, it's obviously going to portray him as a kook. I just wonder if you worry about him because I feel like the narrative in the public sphere is Dario, good guy, Sam Altman, like you know, Satan. Really? I mean, at least the people I hang out with think Dario's bad and Sam's good, but certainly the people I hang out with in D.C. believe that.

10:14You know, Anthropoc's very unpopular with most elected officials, actually on both sides of the aisle. for different reasons in some ways, but their ideology is very offensive in just the way Dario communicates. Both of them talking about job loss was probably a massive mistake, and it's actually factually false, which is more interesting in some ways. Every single prediction about job loss associated with AI as of today is absolutely empirically false. Now, that doesn't mean that in 10 years, 20 years, 30 years that could change, but then the burden is, if you're predicting these changes, identifying what's the inflection, What are the drivers?

10:49What are the inputs that cause the equation to change? As far as I can tell, none of these people are doing even a reasonably good job at articulating what are the drivers that transform net positive job growth. The economy in the United States right now is being powered by AI. Fortunately, because the rest of the economy is a little more shaky, AI is producing productivity gains, which is leading to employment, which is leading to high-paying jobs for people who have actually been excluded from many previous tech revolutions. Like the best opportunities right now are more in the trades. And these are high paying 2x, 3x the median income of the United States.

11:24So this is really exciting. People need to talk about that, plus the medical benefits. Like the kinds of things that kill us and that killed my parents' generation, grandparents' generation, probably are not going to affect us. Now, yes, humans may still have some lifespan. You may not live for eternity or something. But I think we're all going to live healthier, happier lives because of AI. Yeah, absolutely. the biotech startups that we are starting to see are mind-blowing. And I also think the explosion of entrepreneurship driven by AI is going to be really exciting if we survive. But on a lighter note, we are sitting here in New York.

11:59What are you doing in New York? So a complicated story. I mean, New York's a great entrepreneurial environment. Over the last six years, the number one geo that I've invested in, five of the six years has actually been New York. This year we'll see, depends whether you wait by dollars or kind of investments, but it's a little bit more open this year. More personal reason, my family works in D.C., goes to school in D.C., and it's very difficult to commute from either SF or Miami every week to D.C., where it's very feasible to go every week or every weekend to D.C., so that's the primary driver.

12:32It's a good reason. Yeah, but there are great companies here. Like, you know, Ramp is here, as an example. Rogo, you know, we just announced financing this morning. We have lots of application layer. We have lots of financial services, innovative companies like Imprint that are wonderful companies, all based in New York. So I think this is a great entrepreneurial place to invest, but I don't really have a choice where I am right now. Do you think the density of talent is the same here? Depends on what role and at what level of seniority. Individual contributor level, right out of school, absolutely.

13:05Arguably, this is what is Power and Ramp. We've been tapping into right out of school. graduates and been able to create a critical density of talent from the intern class on that is extraordinary. There's these rankings of intern classes, and we always rank number one. Senior engineers who are architect level, no. I think that's a challenge. Fortunately, maybe in the modern age, you need less of these people per company than you have historically, but that's a challenge. Senior executives is also a challenge. If you have an in-office culture, Most of the more senior people that live and reside in the New York area live outside the city.

13:42And the commute in and out of the city for an office environment can be very painful. I grew up in a commuter suburb of New York, so I'm very familiar with this. We were like 32 minutes express train into the city. But many people live two concentric circles further away. So when you need to recruit proven executive talent and you really believe in an in-office culture, those two things have been very challenging. The way we solve this at RAMP specifically is we don't hire senior people. We just build from the bottom up, ground up. It's been a very conscious strategy, very intentionally for the last three years.

14:14That can work. But if you need a CFO, SVP of sales, someone who's got a lot of gravitas and experience, it's really hard to have them in the office five days a week because unless they're very independently wealthy, they really can't afford to raise a family like right in the middle of the city. And are you going to be a lone wolf or is there more to come? We're opening an office. So Katie is opening an office on 14th Street. It's actually allegedly being built out now. We'll see. This fall opening date is very vague in my mind. I'll have a couple colleagues here with me, and then we're also opening up what's called an executive briefing center, which is where we matchmake a set of portfolio companies, say 10 or 12 at a time, with a Fortune 500 company.

14:56And we're going to do that four days a week. And the portfolio companies love this. You know, they get pilots and customers. And so it's going to be a very vibrant office because of that. There'll only be a handful of investors here, hopefully. This is the first time that KB's really ever left 2128 Sand Hill Road. We don't even have an SF office. So this is a very big step for us. You know, you mentioned Ramp. I know that's a big investment of yours through COSLA and Founders Fund where you spent some time. I think Ramp is so brilliant in ways, including getting press. But sometimes I find it a little unsettling.

15:30I mean, the talk about ARRs, which I know a lot of founders in the audience probably also rely on as a means of attracting attention, annualized revenue. What do you think of that strategy? It depends. Founders kind of have different preferences. The value of transparency is you show the momentum, you get press coverage, you get recruiting. The downside that I cautioned founders in the beginning about that level of transparency ultimately is every company has blips. And it's not so fun to have your metrics public when you're suffering through a rollercoaster ride. So yeah, the success case, it's great.

16:04but you should plan for how are you going to manage. And public companies obviously deal with this all the time, but many companies, when they're earlier, they're not set up, they're not insulated, they're not immune, they're not sophisticated enough, they're not mature enough to deal with that downside risk. So I think sometimes founders are over the optimistic about sharing metrics until it's painful. And so I do give everybody who wants to play that game a cautionary tale. But you see it playing out in the legal domain. You can't open X these days without Harvey posting some metrics and the Lagor's investors are all publishing their metrics.

16:36It's like nonstop. It's actually kind of funny. Who would have thought that the most popular thing on X most days of the week would be a debate about legal AI? It's so true. Also, Ramp is pretty savvy about sharing transactional data, sort of to say, oh, Anthropic is being adopted more aggressively right now by enterprises, or no, no, it's open AI. Anyway, I give them a lot of credit. I do wonder if you are worried in any way about rippling. It really does seem like whether or not he says it explicitly, they've got RAMP in their crosshairs. And Parker Conrad... I love Parker. I've been friends with Parker forever.

17:10RAMP is not worried about Rippling. We're going to compete with Anthropic. That's the goal. I want to be larger and more successful and more important than Anthropic. And you've seen some of the steps that we've taken recently. The future of the world is driven through tokens and equivalent. And RAMP's going to be the primary way people manage their token spend, how they optimize it, how they measure ROI. and that will lead us to collide probably with Anthropic in many ways. That is a bold statement. I do wonder about some of the other tactics that we're seeing, which is companies that announce valuations.

17:47I mean, I've seen a couple of instances where it's been within weeks and the valuation is multiple billions of dollars more than it was the last time. What do you think of that and how should people interpret what's happening out there? I mean, I would say we've done a few of those fast follow-around it's not our favorite thing to do. We're probably pretty cautious and have a lot of debate about it. But there is a belief that in the verticals that matter a company that is successful will be worth a hell of a lot of money. Like when I was starting a venture which wasn't that long ago, 2013 a home run success in the back of my mind anyway was between a$10 billion and$50 billion outcome and the extreme end of like super success is a hundred bill.

18:30Now, I'd put at least one zero on that goal and possibly put a second zero. So two orders of magnitude. So if you believe that that's true, so we talked about two companies that in venture timeframes, like within a fund cycle, plus or minus, have created a trillion dollars plus of value, it's now possible to dream that way. If that prediction is right, that there might be more trillion dollar outcomes created within a venture life of a funder, so. Picking the right company and figuring out how to get capital in it at almost any price is not as crazy as it seems. I'll quote my former colleague at a Founders Fund, Brian Singerman.

19:07He did a podcast recently with Jack Altman of Benchmark. And what Brian said is like, look, at Founders Fund we were the 30th investor in Airbnb, but we made the second most money,$3 billion of returns. Does it matter really that we spent$150 million to generate the$3 billion of returns? And some Some people could have done it for 50 or 15. Not really. It's the 3 billion that really matters. So that's what most people are thinking. That said, I think identifying the winning companies is much more challenging than people realize, including my colleagues. For example, in the Ramp case, many people thought Brex was going to be a successful company.

19:42If you had chased Brex at every valuation, the last couple rounds did not make money. Maybe they got their capital back or something like that, but they risk-adjusted and time-adjusted, not particularly attractive. Who knows how open AI versus anthropic is going to play out long-term. But a year and a half ago, everybody's like, open AI is a monopoly. Then everybody's like, now, oh no, it's anthropic. And now people are shifting back a little bit more open AI. Which of those companies are you powering money into? I don't know how to choose. Or legal AI. Would you give, Winston just announced, they're financing a$550 million, or a$15.5 bill, like Tuesday.

20:16Would you power a billion dollars into Harvey or Lagora or none of the above? or some startup in legal AI that none of us have heard of yet? I don't think that's such an obvious call. And so I'm sitting around, and would I write a$300 million check to any of them? I don't know how to do that right now. Do I take Dekacon versus Sierra versus whoever? Almost in every AI vertical, the art of choosing where to do that is much more difficult, I think, than people realize. I'm sure that's true. And unfortunately, right ahead, but I did want to ask you, because I think there's founders in the audience, I did see you appear with Vinod on Jack Altman's podcast, which I really enjoy.

20:56You talked about A-plus players and basically your ability to discern them in like a few minutes. And of course, in Silicon Valley and here too, so many smart, polished speakers who can tell a story and articulate a vision. Like there's so many superhumans in this industry. What makes somebody more superhuman? And where have you made a mistake on that front? So there's one formulation, there's a second formulation, maybe a third formulation. First formulation is I am personally looking for someone who is in the top one to ten basis points on any particular trait. It doesn't matter, like, polished.

21:30They can be completely introverted and completely unpolished, but they're, like, the smartest person. Or they can be extremely polished and it's the greatest salesperson you've ever met in your life, and you just know instantly. Or it's a combination of two traits that are very rarely found together. This is my Max Levchin example. Max is a first-rate technologist, and he's got a very top-tier business acumen. Those do not normally match. In fact, this is in hindsight. Reid Hoffman told me this about Max in December 2000. I was preparing for my first one-on-one meeting with Max and Reid Hoffman sits me down and says, this is what you need to know about Max.

22:09He's as sharp as they come in a business acumen sense, so don't try to compete with him that way. And he's a first-rate technologist and there's five people in Silicon Valley that have that Venn diagram overlap. Proved to be a pretty good call by Reed, honestly, in the next 26 years. So Jack Dorsey was kind of that way. Jack's pretty first-rate business strategist. Actually pretty top-tier designer, too, or at least design taste, and pretty good technologist. So he has three. So I think the rare overlap Venn diagram. I was actually talking on this podcast recently about my husband who just showed up, about he has an unusually Venn diagram overlap of he can speak to tech people and speak to political people, and almost no one can speak both of their languages.

22:49and that became his superpower initially. And then third is there's these rare people that can walk you through what Chris Dixon and Balaji used to call the intellectual maze, and they blogged about it, where they can just tell you where they are today and how they're going to build a company in a vertical industry and walk you through every single move and account for all the trapdoors and mistakes that everybody's ever made through history. They tell you this story, and actually it's even more impressive if you have contacts, like someone like me in financial services or something, and you're just, wow.

23:18Like this person has a map of the planet. That is instantly fundable too. That's actually the rarest of the three, by the way. But there was a company recently, let's say in the last year in New York, that does AI infrastructure, which is something I thought I would never touch in my life. Met this founder. It is urgent that I sort of meet the founder because I was the only partner in New York. And he just walked me through this map and I was like, oh my God. Like I've never heard anybody in an impressive way walk me through how AI infrastructure works and should work in account for all the moving pieces of what other people are doing or not doing.

23:52And I texted my colleague who sent it to me, and I was like, no-brainer, please invest now, close. It's an area I know nothing about, but the map was so impressive. It just blew me away. There's a third thing. So I recently got asked by my colleagues. This question comes up a lot, but I got asked on an off-site to sort of teach my colleagues how to assess founders. I was like, oh my God, this is... They've heard me say all this stuff. I'm like, it's not working because you guys are not copying this. And I was frustrated, but they're like, no, Adam, you've got to speak about. So I went back to my last chief of staff.

24:24And he's phenomenal and very talented. Shadowing me for like 18 months, like 24-7 sort of thing. And I wrote to him, like, from your observation and vantage point, what am I actually looking for? And he wrote me five bullet point texts. And I wound up using three of the five instead of my normal speech. And he had picked up on certain things. The one that most applies to your question was, he said, the first thing you're going to do is people who are trying to raise money basically store in cash mode in a technology sense the answers to every investor question. He's like the first thing you do at every meeting is you're going to invalidate their cash right away.

24:59You're going to find a way to throw it out and you're going to notice a time to first new token and then the quality of that and that's going to be like half your assessment. And I was like wow I'd actually never thought I was doing that but as soon as you wrote it to me I was like oh my god that's exactly what I'm doing. And then I could teach that as a skill. Now the problem is invalidating a founder who's pretty good to cash requires a lot of generally experience. It's not too easy for a junior VC to do that. And also you have to do it in a deft enough style that people will tolerate it. You can't just tell a founder like, oh that's stupid.

25:31Like you just start from scratch. You just like ask some orthogonal question at work. Yeah, it's like basically find something that you know a fair amount about ideally and basically probe pretty aggressively in, I don't believe that's true, or it's not consistent with this data point, or it's not consistent with this company, or it's not consistent with this experience, and force them to re-derive it and see what they do. And can they absorb and build on it? The ones that are most impressive actually will remix some of your critiques and then have a new answer, and that's just, like, incredibly impressive.

26:03That's, like, almost an instant invest decision if they can take your ideas or critiques, absorb them on the fly, and come up with a refined answer that's still mostly theirs. It's like remixing a song on the fly, but taking a David Guetta song and remixing it successfully is very challenging. Well, founders, now you know the secret. Pretty much. Keith, thank you so much for coming. Pleasure. Thanks for the invite. Hi, Jacob, if you're back there somewhere.

26:29The Strictly VC Download podcast is hosted by TechCrunch Editor-in-Chief Connie Loizos and me, Alex Gove of Strictly VC. Strictly VC Download is produced by Maggie Nye with editing assistance by Teresa Lancansolo and Kel Thanks for listening We'll see you back here next week