#228 Elad Gil: How to Spot a Billion-Dollar Startup Before the Rest of the World

AI is the only market where the more I learn, the less I know. And in every other market, the more I learn, the more I know, the more I'm able to predict things and I can't predict anything anymore. I think in a couple years we'll start thinking about it as we're selling units of cognition. AI is dramatically underhyped because most enterprises have not done anything in it. And that's where all the money is, all the changes, all the impact is, all the jobs, everything. The people that I know who have been very successful were driven solely by money, end up miserable because they have money. And then what do you do? What fulfills you?

I think really the way I think about investing in general is that there's two dimensions that really matter. The first dimension is what people call product market fit, or is there a strong demand for whatever it is you're building? And then secondarily, I look at the team, and I think most early stage people flip it. They look at the team first and how good. As a founder, and obviously I've started two companies myself, I think the founder side is incredibly important and the talent side is incredibly important. But I've seen amazing people get crushed by terrible markets, and I've seen reasonably mediocre teams do extremely well. And what are very good markets? And so in general, I first asked, do I think there's a real need here? How's it differentiated? What's different about it? And then I dig into like, are these people exceptional? How will they grow over time? You know, what are some of the characteristics of how they do things?

For things like consumer businesses, you're just looking at organic growth rate and retention or people using it a lot, are they living in it every day, that sort of thing. That'd be early Facebook. Right. The usage metrics were insane. And then for a certain B2B products, it could be rate of growth and adoption. It could be metrics people call like NDR and a dollar retention or other things like that. Honestly, if you're investing before the thing even exists in the market, then you have to really dig into how much do I believe there's a need here? Right. Or how much is there a customer need?

For an early tech team. I almost use, like this Apple framework of Jobs, Wozniak and Cook, right? Steve Jobs and Steve Wozniak started Apple together. Steve Jobs was known as somebody who really was great at setting the vision and direction, but also was just an amazing salesperson. And selling means selling employees to join you. It means raising money and means selling your first customers. It's negotiating your supply chain. Those are all aspects of sales in some sense, or negotiation. And so you need at least one person who can do that. Then you need somebody who can build stuff and build it in a uniquely good way. And that was Wozniak. Right. The way that he was able to hack things together, drop chips from the original of Apple devices, et cetera, was just considered legendary. And then as a thing starts working, you eventually need somebody like Tim Cook who can help scale the company.

It's really fascinating because if you look at almost every major movement throughout history, and that could be a literary movement, it could be an artistic movement, it could be a finance movement, economic schools of thought, it's almost always a group of young people aggregating in a specific city who all somehow find each other. And I'll start collaborating and working together towards some common set of goals that reflect that. And similarly that's happened for tech. And even within tech we've had these successive waves. So one big thing is just you have a geographic cluster and you have that for every single industry.

Remote work is generally not great for innovation unless you're truly in an online collaborative environment. And the funny thing is that when people talk about tech, they would always talk about how tech is the first thing that could go remote because you can write code from anywhere and you can contribute from anywhere. But that's true of every industry, right? You look at Hollywood, you could make a movie from anywhere. Like you film it off site anyhow, or on site in different places. You could write a script from anywhere, you could edit the musical score from anywhere, you could edit the film from anywhere, you could write the script from anywhere. So why is everything clustered in Hollywood? And it's because there's that aggregation of people, there's the people helping each other, sharing ideas, trading things informally, learning new distribution methods that kind of spread, learning new AI techniques that spread.

I think they do a really good job of two things. One is plugging people in, as mentioned, particularly you wanna have a bunch of customers instantly, your batchmates will help you with that. But also it teaches people to ship fast and to kind of force finding customers. And so because you're in this batch structure and you're meeting with your batch every week and you hear what everybody else is doing, you feel PE pressure to do it. But also it kind of shapes how you think about the world, what's important, what to work on. And so I think it's almost like a brainwashing program, right beyond everything else they do, which is great, it sets a timeline that you have to hit and it brainwashes you to think a certain way.

As the AI wave was happening, you know, I started getting involved with a lot of the generative AI companies maybe 3 ish years ago, maybe 3 and a half years ago. So before Chachipiti came out and before mid joourney and all these things kind of took off and the people starting those companies were uniquely good and you felt the shift. You went from these kind of plain vanilla meoo almost larpers to these incredibly driven, mission oriented, hyper smart, very technical people who wanted to do something really big. And you felt it. It was a dramatic shift. And if you look at it, there's basically been three or four waves of talent coming through the AI ecosystem.

I think it's gonna be an ongoing wave of kind of everything. Right. There's still a lot to build, but I think we'll see more and more application level companies. We'll see fewer, what are known as foundation model companies. There will be specialized versions of that, right? There's gonna be the same thing done for physics and material science. We've already seeing it happening in biology, right? So at that layer there's a bunch of stuff. There's the infrastructure, what is the equivalent of cloud services, and then there's the apps on top. And then in the apps you have B2B and then you have consumer. And so I think we're gonna see a lot of innovation across the stack. But I think this next wave is a mix of B2B and consumer. And then I think the wave after that is Very large at enterprise adoption. And so I think AI is dramatically underhyped because most enterprises have not done anything in it.

I think the thing that people misunderstand about artificial intelligence is that, you know, people are kind of viewing it as what you're selling as like a cool tool to help you with productivity or whatever it is. I think in a couple years we'll start thinking about it as we're selling units of cognition, right? We're selling bits of person time or person equivalent to do stuff for us. I'm Gonna effectively hire 20 BoAH programmers to write code for me to build an app, or I'm gonna hire an AI accountant and I'm gonna basically rent time off of this unit of cognition. On the digital side, it really is this shift from you're selling tools to your selling effectively white collar work.

I think that I have opposing fears in the short run. I worry that there's the real chance to kind of strangle the golden goose. Right. I do think AI and this wave of AI is the single biggest potential motivator for versions of global advancements in health and education and all the things that really matter fundamentally. This AI revolution is a Great example of something that could basically provide that for every child around the world as long as they have access to any device, which is most people at this point, right. Globally. So from an education system perspective, a healthcare system perspective, it's a massive change. So in the short run, I'm really worried that people are going to constrain it and strangle it and prevent it from happening, because I think it's really important for humanity in the long run.

No, it's already surpasses on many things. Right. Just even look at how people play GO now and the patterns they learned off of AI which can beat any person to go. I mean, gaming is a really good example of that. Where every wave of gaming advancements where you pitted AI against people, people said, well, fine, they beat people at checkers, but they'll never beat him at chess. And then they beat them at chess and say, well, ll find chess, but they'll never beat him at go. They beat him at go. And they're like, well, what about complex games where there's bluffing? They'll never beat him at poker. And then Noam Brown had his poker paper. I say, well, okay, poker. Well, they'll never beat him at things like diplomacy, where you're manipulating people against each other. And then a Facebook team solved diplomacy. Right. And so gaming is a really great example where you have superhuman performance against every game now.

If you look at it, about half or more than half of all biology research and top journals is not reproducible. So you have a big data problem. Half the data is false. It's incorrect. Right. And this is actually something that Amgen published a couple years ago. They showed this because they weren't able to reproduce cancer findings in their lab because they're trying to develop a drug. And they're like, wait a minute, this thing we thought could turn into. Into drug isn't real. Right. And so there's this really big replication issue in certain sciences.

There's people publishing things that are just bad. And the question is, is it bad because they Ignored other data. Did they throw out data points? How would you know as an AI system, right. That somebody threw out half their data to publish a paper. And so there's other issues around how science is done right now, or you just rush it and you have the wrong controls, but it still gets published. Cause it's a hot field. That happens a lot. If you look during COVID like there were so many papers that in hindsight were awful papers, but they got rushed out because of COVID And unless somebody goes back and actually redo the experiment and then publishes that, they read it it and it didn't work, which nobody does. Cause nobody's gonna publish it for you, how do you know that it's not reproducible?

Yeah, that's called self play. And as long as you have enough rules, you can do it. You need a utility function you're working against. Right. And so in the context of a game, it's winning the game. And there's very specific rules of the game. You know when to flip over the go piece, you know what winning means. Right. And so it's easy to train against that. Cause you have a function to select against. This game you did well. This game you did badly. Here's positive feedback or negative feedback to the model. They're starting to do that more and more. So if you look at the way people are thinking about models now and scaling them, there's three or four components to it. One is ongoing data scale. Second is the training cluster. People always talk about all the money they're spending on GPUs. The third is reasoning modules.