--:-- Tom Bilyeu: Today, we're doing something a little different. We're dropping an episode from Impact Theory, a show hosted by Tom Bilyeu, featuring a conversation with Ben Horowitz, co-founder of Andreessen Horowitz. Ben rarely does interviews like this, and in this one, he goes deep on AI, on power, on the future of work, and what it really means to be a human in a world of intelligent machines.
--:-- Tom Bilyeu: He breaks down why AI is not life, not consciousness, but something else entirely. Why blockchain is critical to preserving trust and truth in the age of deepfakes. Why distribution may matter more than code. And why history tells us this isn't the end of jobs, but the beginning of something new. Let's get into it.
--:-- Tom Bilyeu: This information is for educational purposes only and is not a recommendation to buy, hold, or sell any investment or financial product. This podcast has been produced by a third party and may include paid promotional advertisements, other company references, and individuals unaffiliated with A16Z. Such advertisements, companies, and individuals are not endorsed by AH Capital Management LLC, A16Z, or any of its affiliates. Information is from sources deemed reliable on the date of publication, but A16Z does not guarantee its accuracy.
--:-- Tom Bilyeu: Revolutions don't always come with banners and protests. Sometimes the only shots fired are snippets of code. This is one of those moments. AI is the most disruptive force in history, and it's no longer a distant possibility. It is here right now, and it's already changing the foundations of power and the economy.
--:-- Tom Bilyeu: Few people have been as influential in shaping the direction of AI than mega investor Ben Horowitz. A pioneer in Silicon Valley, he spent decades at the center of every major technological disruption, including standing up to the Biden administration's attempts to limit and control AI. In today's episode, he lays out where AI is really taking us, the forces that will define the next decade, and how to position yourself before it's too late.
--:-- Tom Bilyeu: You are in an area making investments, thinking about some of the things that I think are the most consequential in the world today as it relates to innovation. But along those lines, you and Marc Andreessen are all in on AI, but how do we make sure that it benefits everyone instead of making humans obsolete?
--:-- Ben Horowitz: To begin with, we have to just realize what AI is. I think that because we called it artificial intelligence, our whole industry of technology has a naming problem. We started by calling computer science, computer science, which everybody thought, "Oh, that's just like computers." It's the science of a machine as opposed to information theory and what it really was.
--:-- Ben Horowitz: In the Web3 world, which you're familiar with, we called it cryptocurrency, which to normal people means secret money. But that's not what it does.
--:-- Tom Bilyeu: That's a good point.
--:-- Ben Horowitz: And then with artificial intelligence, I think that's also a bad name in a lot of ways. The people who work in the field call what they're building "models." I think that's a really accurate description. What we're doing is modeling something we've always done—we're trying to model the world in a way that enables us to predict it.
--:-- Ben Horowitz: We've built much more sophisticated models with this technology than we could before. In the old days, we had E=mc², which is amazing but a relatively simple model. Now we have models with 600 billion variables. We can model what the next word should be. That's amazing and powerful.
--:-- Ben Horowitz: But we need to distinguish the fact that it's a model directed by us to tell us things about the world and do things on our behalf. It's not life. It doesn't have free will. I think by default, we are the master and it is the servant, as opposed to vice versa.
--:-- Ben Horowitz: The question you're getting at is, "How do we not get obsoleted? Why do we need us if we've got these things that can do all the jobs we currently do?" We've gone through that in the past, and it's been interesting.
--:-- Ben Horowitz: In 1750, over 90% of the jobs in the country were agricultural. There was a huge fight with a group called the Luddites that fought the plow and these newfangled inventions that eventually eliminated 97% of the jobs that were there. But most people would say the life they have now is better than the life they would have had on the farm where all they did was farm and nothing else. If you want to farm still, you can. That is an option, but most people don't take it.
--:-- Ben Horowitz: A lot of the jobs we have now will go away, but we will likely have new jobs. Humans are pretty good at figuring out new things to do and new things to pursue, including going to Mars, which obviously isn't a thing today but could very well be a thing tomorrow. We have to stay creative and keep dreaming about a better future and how to improve things for people.
--:-- Ben Horowitz: For the people in the world that are on the struggle bus—living on a dollar a day, subject to all kinds of diseases—life is going to get radically better for them.
--:-- Tom Bilyeu: When I look at the plow example and the Luddites fighting against it, I think you'll see the same thing with AI. You're going to get people that completely reject it, refuse to engage with it. But when I look at AI, what I worry about is that there will be no refuge to go to. If you realize you can't plow as well as a plow or a tractor or a combine, there are still a lot of other things that technology can't do better than you. Do you think there's an upper bound to artificial intelligence—bad name or not? Or do you think it keeps going and literally becomes better than us once it's embodied in robotics at everything?
--:-- Ben Horowitz: We are kind of limited by the new ideas that we have. Artificial intelligence is really artificial human intelligence. Humans looked at the world, figured out what it was, described it, came up with concepts like trees and air. It's not necessarily real—just how we decided to structure the world. AI has learned our structure. They've learned human language, which is a version of the universe that is not an accurate version of the universe. It's just our version.
--:-- Tom Bilyeu: But you're going to have to go deep on that. I know my audience is going to be like, "Air seems pretty real when you're underwater. What do you mean that trees and air are not necessarily real?"
--:-- Ben Horowitz: It's a construction that we made. We decided it is literally the way humans have interpreted the world in order for humans to navigate it. Language isn't the universe as it is. If you had an objective look at the atoms and how they were arranged, those descriptions are lacking in a lot of ways. They don't predict everything about how the world works.
--:-- Ben Horowitz: What machines have learned—what artificial intelligence is—is an understanding of our knowledge. It has taken in our knowledge of the universe, and it can refine that, work on it, and derive things from our knowledge and our axiom set. But it isn't actually observing the world at this point and figuring out new stuff.
--:-- Ben Horowitz: At the very least, humans still have to discover new principles of the universe or interpret it in a different way. The machines have to somehow observe the world directly, which they're not yet doing. That's a pretty big role.
--:-- Ben Horowitz: In addition, we direct the world. I think Star Trek is a pretty good metaphor. The Star Trek computer was pretty badass, but the people in Star Trek were still flying around the universe discovering new things. There was still much to do.
--:-- Ben Horowitz: It's always difficult to figure out what the new jobs that get created are. We've had intelligence for a while. We've had machines that could do math way better than us. I remember when I was in junior high school, our school put on a play about how bad it was that there were calculators because nobody would know how to do arithmetic. Then all the calculators would break and we'd be stuck trying to fly around the universe in rockets but unable to do math.
--:-- Ben Horowitz: There is always that fear. We've had computers that can play games better than us. We currently have computers that can drive better than us. We have a lot of intelligence out there, but it hasn't created a super dystopia. It's actually made things better everywhere it's appeared. I would expect that to continue.
--:-- Tom Bilyeu: What do you think is the limiting function? When I look at AI, I always say unless we run into an upper bound where the computation just can't allow the intelligence to keep progressing, it seems like it will become not only generalized human intelligence and be able to do everything we can do, but it will become embodied as robotics.
--:-- Tom Bilyeu: If I ran the math on this one, Einstein is roughly 2.4 times smarter than someone who is definitionally a moron. The gap between those two is so dramatic. The army won't even draft somebody with an 81 IQ because they create more problems than they solve. Do you think there is something that's going to cause that upper bound, or do you have a belief about the nature of intelligence that will keep AI subservient to us?
--:-- Ben Horowitz: The smartest people don't rule the world. Einstein wasn't in charge. Many of us are ruled by our cats. Power and intelligence don't necessarily go together, particularly when the intelligence has no free will or desire. It doesn't have will. It's a model that's computing things.
--:-- Ben Horowitz: The whole general intelligence thing is interesting. Waymo has a super smart AI that can drive a car, but that AI doesn't know English and isn't particularly good at other tasks. ChatGPT can't drive a car. How much things generalize is a question.
--:-- Ben Horowitz: A lot of it has to do with the long tail of human behavior. The distribution of human behavior is fractal—it's not evenly distributed at all. An AI that captures all that turns out to be... we're not so much on that track. We're more on the track for really great reasoning over a set of axioms we came up with in math or physics, but not so much general human intelligence—being able to navigate other humans and the world in a way that is productive for us. That's a different dimension.
--:-- Ben Horowitz: You can compare the math capabilities or the Go playing capabilities or the driving capabilities or the IQ test capabilities of a computer. But that's not really a human. A human is different in a fairly fundamental way.
--:-- Ben Horowitz: What we end up doing is going to be different than what we're doing today, just like what we're doing today is very different than what we did 100 years ago. But the not having a need for us—these AIs are tools for us to navigate the world and help us solve problems, from preventing pandemics to dealing with climate change to not killing each other driving cars. Hopefully it doesn't create more wars. Hopefully it creates less wars. We'll see.
--:-- Tom Bilyeu: What I know about the human brain might be tricking me into painting a vision of the future that isn't going to come true. Let me put words in your mouth and you tell me if they fit appropriately. What I hear you saying is something akin to the way that we're approaching artificial intelligence right now—let's round it to large language models—is going to hit an upper bound where it's not able to have insights that a human will already have. They are trapped inside of the box that we have created, what you're calling the axioms by which we navigate the world.
--:-- Tom Bilyeu: They get trapped inside that box and thus will never be able to look at the world and go, "I'm not going to predict the next frame. I'm going to render the next frame based on what I know about physics. Water reacts this way in an earthbound gravity system, so it's going to splash like this." Is that accurate? Are you saying it is trapped inside of our box and will never have that capability?
--:-- Ben Horowitz: It hasn't demonstrated that capability yet. It hasn't walked up to a rock and said, "This is a rock." We labeled it a rock because that's our structure. A rock isn't probably what a more intelligent being would call it. Maybe the rock is irrelevant to how you can navigate the world safely.
--:-- Ben Horowitz: Figuring those things out or adapting to them is just not something it does. It's trained on our rendition of the universe, the way we have described it using language we invented. It is constrained to that in nature currently.
--:-- Ben Horowitz: That doesn't mean it's not a massively useful tool that can do things. It can derive new rules from the old rules we've given it. But it's a bit of a jump to say it's going to replace us entirely when the whole discovery process is something we do that it doesn't do yet.
--:-- Tom Bilyeu: The way the human mind is architected, you have competing regions of the brain. If you cut the corpus callosum—the part that connects the left and right hemisphere—you can get two distinct personalities. One that is atheist, for instance, and one that believes deeply in God, and they'll argue back and forth. This is in the same human brain.
--:-- Tom Bilyeu: That tells me that you have regions of the brain that get good at a thing, and then they come together to collaborate, and that is sort of human intelligence. I've heard you talk about there being something like 200 computers inside of a single car. If we already know that you can daisy chain all of these—deep knowledge about one thing—and as you daisy chain them, the intelligence gets more generalized. You don't see that as a flywheel that is going to keep going?
--:-- Ben Horowitz: What we can compute will get better and better and better. But that doesn't say that humans... humans built the machines, plug them in, give them the batteries. They've been created to fulfill our purposes. What it means to be a human will probably change like it has been changing. How humans live their life will change. But humans still find things to do.
--:-- Ben Horowitz: It's like a cheetah has been able to run faster than a human forever, but we never watch cheetahs race. We only watch humans race each other. Computers have played chess better than humans for a long time, but nobody watches computers play chess anymore. They watch humans play humans, and chess is more popular than it's ever been.
--:-- Ben Horowitz: We have a keen interest in each other and how that's going to work. These will be tools to enhance that whole experience for us. A world of just machines seems really unlikely.
--:-- Tom Bilyeu: You've got people like Elon Musk and Sam Altman who have both expressed deep concerns about how AI may make us obsolete. Elon has become fatalistic. He gave a rant that I absolutely love: AI is a demon summoning circle, and you're calling forward this demon that you were convinced you're going to be able to control. He certainly is not so sure.
--:-- Tom Bilyeu: It's interesting that both of them, despite saying these things, are building AI as fast as they can. They're literally in a race with each other to see who can build it faster. What do you take away from that? Is it just regulatory capture on both of their parts? Is Elon being sincere? What do you take away from the fact that they've both warned against it and they're both deploying it as fast as they can?
--:-- Ben Horowitz: It seems fairly contradictory. I won't question either of their sincerity. But I do think there are many reasons to warn about it. Any kind of new super powerful technology—they're right to warn that if we don't think about some of the implications, it could get dangerous. I think that's a good thing.
--:-- Ben Horowitz: Every technology we've ever had, from fire to automobiles to nuclear to AI to the internet, has downsides. The more powerful the technology, the more intriguing the downside. Without the internet, we probably would never have gotten to AI. Maybe that was the downside of the internet—that it led to AI. But I think we would take every technology we've invented and keep it because net-net, they've been positive for humanity. That's why I think they're building it so fast. I think they know that.
--:-- Tom Bilyeu: Anybody with a 17-year-old right now is thinking, "Where do I point my kid? What do I tell them to go study that's future-proof?" What can we learn about the way you guys are investing at Andreessen Horowitz that would give somebody an inclination of what you think a 17-year-old should be focused on now?
--:-- Ben Horowitz: It's really interesting. One of the things we're seeing in the smartest young people that come out is they spend a lot of time with AI learning everything they possibly can. I think you want to get very good at high curiosity and learning. You have available to you all of human knowledge in something that will talk to you. That's an incredible opportunity.
--:-- Ben Horowitz: Anything you want to do in the world to make the world better, you now have the tools as an individual to do that in a way that... if you look at what Thomas Edison had to do in creating GE and what that took, it was a way higher bar to have an impact. Now, you can very quickly build something or do something. Pick a problem.
--:-- Ben Horowitz: Sometimes in this AI conversation, we ignore the problems we have in the world. We still have cancer, diabetes, sickle cell, every disease. We still have the threat of pandemics, climate change, lots of people starving to death, malaria. Pick a problem you want to solve. Now you have a huge helping hand in doing that that nobody in the history of the planet has ever had before. There are really great opportunities along those lines.
--:-- Ben Horowitz: That would be my best advice: get really good with that. A lot of the things that have been valuable skills traditionally are going to change. You want to be able to learn how to do anything. I think that's probably going to be key.
--:-- Tom Bilyeu: When I look at the things you were just talking about, that feels right for people that have the inclination and the cognitive horsepower to leverage AI to extend their capabilities to tackle the biggest problems in the world. Certainly right now, that is the thrilling reality people should focus on.
--:-- Tom Bilyeu: But then I contrast that with the deaths of despair among largely young men. We have this problem in middle America where manufacturing jobs have gone away. For that normal, everyday person who wants to have a trade and go out into the world and get something done, is AI going to be useful to them, or are they going to get replaced by robotics?
--:-- Ben Horowitz: The truth of it is, there's only one robot supply chain in the world, and that's in China. We all need a robot supply chain. We need to manufacture that. I think there's going to be a real manufacturing opportunity coming up. It'll be a different kind of manufacturing. Certainly more will be automated, but there will be a lot of things to learn in that field that will be super interesting and likely very good jobs.
--:-- Ben Horowitz: Ironically, I would say going into manufacturing now as a young man and trying to figure out what that is and get engaged in it will probably lead to quite a good career—maybe in creating factories that have become insanely valuable and strategic to the national interest as well.
--:-- Tom Bilyeu: That makes sense. At the level of the guy smart enough to build the facility, yes. I recently saw a video of the grocery store of the future where it is a huge grid inside of a giant facility. There are bots that look like small shopping carts grid patterning across all the items, snatching up whatever you order. For the person savvy enough to build that facility, tremendous.
--:-- Tom Bilyeu: But what I think I hear you saying—correct me if I'm wrong—is that there are two opportunities here. Opportunity one is if you're the kind of person that can leverage AI to build that facility, massive opportunity. If you're the kind of person that would traditionally work at that factory, something new is coming. We know that because looking back at history, all these technologies unleash things we can't yet see. I have faith in the "we can't yet see it, but it is coming."
--:-- Ben Horowitz: For sure. Do you know what the biggest in-demand job in the world is right now? Data labelers. Data labeling wasn't a job not long ago.
--:-- Tom Bilyeu: I've even heard of this. What is data labeling?
--:-- Ben Horowitz: It's what Scale AI does. They pay armies and armies of people to label data. "Hey, this is a plant" or "this is a fig" for the AI to understand it. With reinforcement learning coming back into play, that kind of supervised learning is still very, very important. Right now, Scale AI has unlimited hiring demand. It's ironic for Scale AI to have unlimited need for humans.
--:-- Ben Horowitz: In manufacturing, there are going to be jobs like that. When you go into these robot software companies that are doing robotics, they have people managing the robots. Humans train robots to do all kinds of things. Folding clothes doesn't necessarily generalize to making eggs. They're super different for robots. You need these robots trained in all these kinds of fields.
--:-- Ben Horowitz: There's a whole new class of jobs that are a little bit hard to anticipate in advance. For at least the next 10 years, the number of new jobs related to making these machines smarter is going to increase a lot. After that, there just tend to be, throughout history, so many needs for new things that we never anticipated.
--:-- Ben Horowitz: One of my favorite examples: computers were going to kill the typesetting business. They did. Everybody knew that. Nobody said, "And then there's going to be 5 million graphic design jobs that come out of the PC." Not a person predicted that. It's easy to figure out which jobs are going to go away. It's much more difficult to figure out which jobs are going to come.
--:-- Ben Horowitz: If you look at the history of automation, it automated away everything we did 100 years ago. There's less unemployment now than there was then. Some of the employment will be much more enjoyable than the old employment, as it has been over time. You always talk about manufacturing jobs going away, but the manufacturing jobs that have gone away have been the most mind-numbing.
--:-- Ben Horowitz: Things evolve in very, very unpredictable ways. The hope is that the world just gets much better. I'm not so worried about anticipating all the horror that's going to come. The main reason we're making these things is the ways they're making life better. We finally figured out a way for everybody—we already have it in our hands—everybody can get a great education. That whole inequality of access to education is literally gone right now. That's pretty amazing.
--:-- Tom Bilyeu: It's certainly huge. Nothing that I ever thought I'd see. So hopefully things go well. The great irony. It's so crazy. I don't think anybody saw that coming. It was always going to go for the drivers, for all those hard, difficult, repetitive tasks. It's been very fascinating to see what actually is in danger—super creative jobs are very much in danger.
--:-- Tom Bilyeu: But as you get down... I think I believe way more strongly than you do that robots are just going to get better and better and better. But that could be that I'm not as close to the problem as you are. Speaking of which, how are the insights that you've had into AI informing the investments that you guys make?
--:-- Ben Horowitz: The theory is there's this one super intelligent big brain that's going to do everything. The reality on the ground is even with the state-of-the-art models, they're all kind of good at slightly different things. Anthropic is really good at code. Grok is really good at real-time data because they've got the Twitter stuff. OpenAI has gotten very good at reasoning.
--:-- Ben Horowitz: With all of them, we're doing AGI, but they're all good at different stuff. From an investing standpoint, it's very good to know that. If something like that is not winner-take-all, that becomes super interesting.
--:-- Ben Horowitz: It's also interesting for what it means at the application layer. If the infrastructure products aren't winner-take-all... the other thing about the infrastructure products that's interesting is that they're not particularly sticky in the way that Microsoft Windows was very sticky. You build an application on Windows, it doesn't run on other stuff. You've got to do a lot of work to move it to something else. You get this network effect with developers.
--:-- Ben Horowitz: Then you go, "Okay, how does that work with state-of-the-art models?" People build applications on these things. But guess what?
--:-- Ben Horowitz: Like, to move your application to DeepSeek, you didn't have to change a line of code. They just literally took the open Python API and, like, it runs on DeepSeek now. Ta-da! So that kind of thing really impacts how you think about investing and what the value of having a lead in application is. And then, where is the mode going to come from?
--:-- Ben Horowitz: Of course, AI is also getting amazingly good at writing code. So then, how much of a lead do you have in the code itself versus the other traditional things? When I started in the industry, the salespeople were in charge. There's a great TV show called *Halt and Catch Fire*. If you watch it, the thing that's really stunning, if you're coming from the 2010s, 2020s world, is why are the salespeople so powerful? But they were the most powerful in those days because distribution was the most difficult thing.
--:-- Ben Horowitz: I think distribution is going to get very, very important again because maintaining a technological lead is a lot harder when the machine is writing the code and writing it very fast. Although it's not all the way where it can build super complex systems. But there are a bunch of things out now. Replit's got a great product for it. There's a company called Lovable that's got one out of Sweden that just builds you an app. If you need an app for something, you just say, "Build me this app," and there it is.
--:-- Ben Horowitz: Another thing is Cursor, where you can select what model you want to use for whatever thing you're about to generate. The ability to go, "Oh, I want 10 of these things. I'm going to use this one for this kind of code, this one for that kind of code," is really fascinating.
--:-- Tom Bilyeu: Now, the Biden administration was super hostile towards tech. When you look at what's going on now with the changes in regulatory, what do you think about the race between us and China? Were we headed down a dark path where if that administration had stayed with that—we're going to have one or two companies, we're going to control them, that's going to be that? Is it possible we could have lost that race? Is that race a figment of my imagination? Is that real?
--:-- Ben Horowitz: I think there are multiple layers to the AI race with China. The Biden administration was kind of hostile in many ways, but all for a central reason, I think. In AI in particular, we met with the White House. I should be very specific: we did meet with Jake Sullivan, and he was very good about it. We met with Gina Raimondo, and she was very good about it. But their position was super, I would say, ill-informed.
--:-- Ben Horowitz: They basically walked in with this idea that we've got a three-year lead on China and we have to protect that lead. Therefore, we need to shut down open source. And that doesn't matter to you guys and startups because startups can't participate in AI anyway—they don't have enough money. The only companies that are going to do AI are going to be, ironically, the two startups, Anthropic and OpenAI, and then the big companies, Google, Microsoft, and so forth. So we can put a huge regulatory barrier on them because they have the money and the people to deal with it. In their minds, I think they actually believed that would be how we would win. But in retrospect, that makes no sense.
--:-- Ben Horowitz: This goes to the next thing. There's how good is your AI, and then how well is it integrated into your military and the way the government works. I think that China, being a top-down society, their strength is that whatever AI they have, they're going to integrate into it. All the companies are already highly integrated into the government. So they're going to be able to deploy that, and we're going to see it in action with their military very fast.
--:-- Ben Horowitz: The advantage of the U.S. is that we're not a top-down society. We're a wild, messy society. But it means that all of our smart people can participate in the field. There's more to AI than just the big models. As you said, how important is Cursor? It's really important if you're building stuff. If you want to go build the next whatever thing that the CIA needs or the NSA needs, you're building that with Cursor. You're using a state-of-the-art model. But if the Biden White House had gotten their way, they'd eliminate things like Cursor. They'd eliminate startups being able to do anything in AI.
--:-- Ben Horowitz: The advantages that we have are that we don't just have a model. We've got all this other stuff that goes with it. And then we have new ideas on models with new algorithms. That's what the U.S. is great at. What China is great at is math. People are good at math, and AI is math, so their models are good. They also have a data advantage on us. They have access to the Chinese internet. They have access to copyrighted material, which they do not have the same deference for that we do in the U.S. So they're able to get to—if you use DeepSeek, you go, "Wow, DeepSeek really is a great writer compared to a lot of the U.S. models." Why is that? They train on a bigger data set than we do. And that's amazing.
--:-- Ben Horowitz: So I think what we want is to have world-class, first-class AI in the U.S. I think of it less as "Is it ahead of China? Is it slightly ahead of China?" I think that with our own state-of-the-art models, leads are very shallow. And I think that'll continue as long as we're able and allowed to build AI. Economically, you'd like to have a vibrant AI ecosystem coming out of the U.S. so other countries who aren't state-of-the-art adopt our technology. We continue to be strong economically, as opposed to everything goes to China. That was a big, big risk with the Biden administration, I think.
--:-- Ben Horowitz: What they were doing on AI was tough. I would say what they were doing on fintech and crypto was even tougher—they were just trying to get rid of the industry in its entirety. With AI, they were extremely arrogant in what they thought their ability was to predict the future. Mark and I were in there. Our job is to predict it. This is our job—to invest in the future, to predict the future. And they were saying things that were so arrogant that we would never even think to say them, even if we thought them, because we know that we don't know the future like that. It's just unknowable. There are too many moving parts. These things are really complicated.
--:-- Tom Bilyeu: All right. Well, speaking of the future, fully accepting that it is very opaque and very difficult to see, what would you say is the most controversial view that you hold about the future?
--:-- Ben Horowitz: If we don't get too world-class in crypto, we're going to be—AI really has the potential to wreck society. What I mean by that is, if you think about what is clearly going to happen in an AI world: one, we're not going to be able to tell the difference between a human and a robot. Two, we're not going to know what's real or fake. Three, the level of security attacks on big central data repositories is going to get so good that everybody's data is going to be out there. There is no safe haven for a consumer. And finally, for these agents and bots to actually be useful, they need to be able to use money and pay for stuff and get paid for stuff.
--:-- Ben Horowitz: If you think about all those problems, they are problems that are by far best solved by blockchain technology. One, we absolutely need a public key infrastructure such that every citizen has their own wallet with their own data, with their own information. If you need to get credit or prove you're a citizen, you can do that with a zero-knowledge proof. You don't have to hand over your social security numbers, your bank account information, all this kind of thing, because the AI will get it. You really need your own keys and your own data, and there can't be these gigantic, massive honeypots of information that people can go after.
--:-- Ben Horowitz: With deepfakes, we're going to have to be able to whitelist things. We're going to have to be able to say what's real. But who keeps track of what's true? Is it the government? Oh, please, Jesus, no. Does everybody trust Trump now? Does everybody trust Biden? Is it going to be Google? Do we trust those guys? Or is it going to be the game-theoretic mathematical properties of the blockchain that can hold that up?
--:-- Ben Horowitz: I think it's essential that we regenerate our blockchain crypto development in the U.S. and get very serious about it. If the government were to do something, I think it should be to start requiring these information distribution networks, these social networks, to have a way to verifiably prove you're human, prove where a piece of data came from, and so forth. We have to have banks start accepting zero-knowledge proofs, and that be just the way the world works. We need a network architecture that is up to the challenge of these super intelligent agents that are running around.
--:-- Tom Bilyeu: We were talking before we started rolling that you guys have an office in D.C. Part of what you do is advise on that. What do the infrastructure changes need to look like? What are a small handful of things that you guys are really pushing to see the government adopt to modernize the way that the whole bureaucracy works?
--:-- Ben Horowitz: There are a few things. Because blockchain technology involves money, we do need regulation. It's not like we don't need any regulation. There are very specific things that we're working with the administration to make sure are done in a way that creates a great environment for everybody.
--:-- Tom Bilyeu: What are you guys hoping will get blocked out? Is that what you're about to cover?
--:-- Ben Horowitz: Yeah. One of the first things you need is electronic money in the form of stablecoin—actual currency. But we need that to not collapse, because then the whole trust in the system breaks down. Why do we need this kind of internet-native money? I'll give you an example. We have a company called Daylight Energy. We're going to run into a big energy problem with AI. AI consumes a massive amount of energy, much more than Bitcoin ever did, which everybody would call up in arms about. It's so much that you can't really even get it out of the power grid. I think Trump has been smart about this, saying you probably need to build power next to your data center because we can't be giving it to you from the central tank.
--:-- Ben Horowitz: Beyond that, individuals now have Tesla solar panels and Powerwalls. When you have one of those, you sometimes have more energy than you need and sometimes have less. Wouldn't it be great if there was a nice system that figured out who needed energy and who had energy, and you could just trade? There would be some kind of contract that says, "This is what you pay during peak. This is what you pay at different periods." That contract is probably best done in the form of a smart contract. But a Powerwall is not a human. It doesn't have a credit card. It can't get a credit card. It doesn't have a bank account. It doesn't have a social security number. But it can trade crypto. It can trade stablecoins.
--:-- Ben Horowitz: We need that kind of currency to facilitate all these automated agreements and automated transfer of wealth between entities in order to solve these big problems like energy. So we need a stablecoin bill that says, "Look, we need these currencies to be backed one-for-one with U.S. dollars or whatever it is so that we can have a system that works and is trusted."
--:-- Ben Horowitz: There's a really interesting side benefit to that. If you look at treasury auctions lately, the demand for dollars is not good. The two biggest lenders to the U.S. have been China and Japan. China's backed off a lot. The demand has backed off somewhat. We've done things to also dampen demand. When we sanctioned Russia and seized the assets of the Russian Central Bank, other countries had entities that had money there. Their money got frozen and they couldn't access it. That makes people more wary of holding everything in dollars. So we've done a lot to dampen that, which fueled inflation in the same way that increasing supply fuels inflation—killing demand fuels inflation.
--:-- Ben Horowitz: Here we would have this new major source of demand for dollars. And the dollars would be much more useful because you can use them online as well, and machines can use them.
--:-- Tom Bilyeu: Sorry, really fast for people that are trying to track that. The reason that would increase the demand for dollars is that the stablecoin would be backed one-for-one with debt? Is that the idea?
--:-- Ben Horowitz: Well, yeah. You would basically have to have a dollar. You would hold treasuries. You basically hold treasuries so that if somebody wanted to redeem their stablecoins, they could. That way, it's the equivalent of the gold standard in the old days when dollars were trying to get credible. We would need dollars to be the gold standard for the stablecoin. Probably we should never back off of that. Maybe we should never have backed off of gold, but it's easier when it's dollars because we did start to run out of gold a bit.
--:-- Ben Horowitz: So that's one thing. Secondly, there's a bill that went through the House known as the market structure bill. It was technically called FIT21. That's very important, whether it's exactly that or some form of that. When you talk about tokens, which are a very important instrument in the blockchain world because it's the way this amazing network of computers gets paid for—who pays the people for running the computers? That's paid in the form of these tokens.
--:-- Ben Horowitz: These tokens can be many things. You can create a token that's a collectible. You can create a token that is a digital property right that links to some piece of real estate or a piece of art. A token can be a Pokemon card. A token could be a coupon. A token could be a security that represents a stock. It could be a dollar. So which one is it? A very important set of rules doesn't exist.
--:-- Ben Horowitz: This is one of the most insidious things the Biden administration did. They basically said, "Well, everything is a security. Everything's a stock or something with asymmetric information." That basically undermines the whole power of the technology. It was a scheme for them to get rid of the industry. It was a very dark, cynical way of legislating things. Then they would make these fake claims about scams.
--:-- Ben Horowitz: The market structure bill is very important in that way. Another thing that was in the original market structure bill is important: there are also scams. We call it the casino. You can create some coin like the Haktua girl did. She creates a coin, lies about her holdings, says she's going to hold them, but then sells them after people buy it in a short time period. There are no rules around that. In the bill that passed the House, it said you can create a token, but if you hold it, you can't trade it for four years. That takes a lot of the ability to scam out of it and forces people to do things that are their real utilities. If it's a collectible, like the Haktua collectible, it's got to be a real collectible where you don't just rug the users of it right away.
--:-- Tom Bilyeu: Right away. Yeah. It's okay to do it later, but—
--:-- Ben Horowitz: Well, in four years, it is what it is.
--:-- Tom Bilyeu: Yeah, no, I'm just giving you a hard time. I just know how that's going to sound to people.
--:-- Ben Horowitz: Yeah, no, thank you. But these kinds of things are going to be really important to making the whole industry work. We're working on that, trying to make it safe for everybody. As I said, it's such a critical technology in an AI world. If we don't have it, it's going to be a very problematic, cyberpunk, high-technology, difficult society.
--:-- Tom Bilyeu: It was shocking to me, the level of backlash that the blockchain Web3 community got. What do you think drives that? Is it just the perception that it was only scams and there's nothing real? What was that all about?
--:-- Ben Horowitz: There were multiple factors. The first one is the one that hits all new technology: "Oh, it's a toy. It doesn't do anything new. The old way of doing things is better." We saw that with social networking. We actually saw that with the Internet. I think Paul Krugman famously said it would never have more economic impact than a fax machine. So that's just a normal thing that happens with new technologies as they start out looking not that important.
--:-- Ben Horowitz: With crypto in particular, one way to think about it is that blockchain is a new kind of computer. New kinds of computers are always kind of worse in every way but one than the old computer. If you look at even the iPhone, it was a bad phone. It had a horrible keyboard. If you compared it to anything, it wasn't very powerful. It had a little itty-bitty screen. But it had a feature that was pretty awesome: you could put it in your pocket, and it had a GPS and a camera. Now you could build Instagram. You could build Uber, which you could not build with a PC, and you still can't build with a PC. That was enough. Eventually, it started to add the other features, and it's an awfully powerful computer these days.
--:-- Ben Horowitz: If you look at blockchain, it's slower. It's more complicated to program. There are a lot of issues with it. But it's got a new feature: trust. It can make promises. You can trust that when that code says there's only 21 million Bitcoin, you can absolutely count on that in a way that you can't trust Google. You can't trust Facebook to say, "Oh, these are our privacy rules." You can't trust the U.S. government to say they're not going to print any more money. That's for sure. Here's a computer that can make promises that you can absolutely count on. You don't have to trust a company. You don't have to trust a country. You don't have to trust a lawyer. You just have to trust the game-theoretic mathematical properties of the blockchain. That's amazing.
--:-- Ben Horowitz: Now you can program property rights and money and law and all these kinds of things that you could never do before. I think that's hard for normal people to understand who aren't deep into technology. They get confused and say, "Yeah, it's nothing, blah, blah, blah."
--:-- Ben Horowitz: Then the next wave was a very odd thing with the Biden administration. I think it's come out now that he wasn't really the president. He wasn't really making any decisions. You couldn't even get a meeting with him if you were in his cabinet. In terms of domestic policy, that was run by Elizabeth Warren. The confusing thing is Elizabeth Warren is always calling people fascist. Her whole push with fintech and crypto was to make sure that she could kick people out of the banking system who are political enemies. To do that, you have to outlaw new forms of financial technology because those would be back doors or side doors or parallels to the GSIBs and the banking system, which she could comprehensively kick people out of.
--:-- Ben Horowitz: When you use a full top-down hierarchy and you can use private companies to enforce your will, that is the way fascism works. The way she does it is she sells this fake story about it funding terror. It turns out USAID was funding the terrorist groups, but that's a different story. It was doing all these nefarious things, which was a very unfair portrayal. The whole industry got this reputation as scammy.
--:-- Ben Horowitz: Then we had Sam Bankman-Fried, who didn't do us any favors. This is another issue with what Elizabeth Warren did: she blocked all legislation. The criminals were running free, and the people doing things that should have been legal were getting terrorized by the government. They should have been looking at FTX. They were looking at Coinbase, which was totally compliant, a public company, begging for feedback: "Tell us what you want us to do."
--:-- Tom Bilyeu: Exactly. That whole thing was crazy.
--:-- Tom Bilyeu: Given that AI is putting us on a collision course with not knowing who's real or what's fake, do you think that blockchain is about to have its day in the next 12 to 24 months? Or is this still something so embedded deep in the infrastructure that it's going to take a long time to have its "I told you so" moment?
--:-- Ben Horowitz: No, I think it's within 24 months for sure. In the last wave of blockchain, there were real technological limitations that slowed it down from getting broader adoption. Very obvious usability challenges. The fees were really high. The blockchains were slow. There were a lot of use cases that you just couldn't do on them. I think that's changing very, very fast. The chains are much faster. The layer-two stuff makes them very fast and cheap. People are doing a lot on usability for wallets and these kinds of things. I think we're getting pretty close.
--:-- Ben Horowitz: The needs are very high. If you think of something like WorldCoin, the difference between that thing being very broadly adopted and where it is now—I think half the people in Buenos Aires use it daily. It's very widely adopted where it's been legal. If they are able to get integrated into some of the big social platforms, then everybody needs proof of human. It would make the experience online so much better if you knew who was human and who was not. Right now, you can't tell at all. That problem is going to get worse, and the solution is really here. I think it's going to start to take off.
--:-- Ben Horowitz: You only need one or two big use cases to start getting the whole infrastructure deployed. Once the infrastructure is deployed, we'll certainly rely on it. If you look at the curve of people who have active wallets and the curve of internet adoption, they're pretty similar. I think blockchain is growing a little faster than the internet did initially. I think we'll get to a place where certainly everybody in the U.S. will be on it.
--:-- Ben Horowitz: That could be great from a government standpoint. Elon has talked about putting all the government payments on the blockchain, which I think would be really, really good for transparency.
--:-- Tom Bilyeu: Oh, my God. Yes, please.
--:-- Ben Horowitz: We'd never get into this weird situation now where half the country wants to tear down all the government services and half of them wants to keep it because nobody knows what the hell the spending is. That would be great.
--:-- Ben Horowitz: Beyond that, why is there so much waste and fraud? Part of it is that I get taxed. I give my money to the IRS. The IRS gives it to Congress. They do whatever they do with it. How does it get to the people who need it? That's a very lossy process. We don't even know who they are. One of the things we found out during COVID is that the government is not very good at sending people money. It's good at taking money. It's not good at sending them money. We lost like $400 billion trying to give people stimulus checks.
--:-- Tom Bilyeu: Ridiculous. Yeah, crazy.
--:-- Ben Horowitz: Ridiculous. But if everybody in the U.S. had an address on the blockchain, you could just tell me, "Okay, here are 10,000 people who need money. Please send them $5,000 each." Well, probably that's too much money for me. But something like that, whatever my tax bill is or whatever that portion of wealth redistribution is, that would be 100% zero loss. By the way, I'd feel a lot better about it because I know I'd be helping people. Maybe even somebody would go, "Hey, this is great. Thank you." Maybe we wouldn't have this crazy class warfare because everybody would go, "Hey, we're all integrated. I'm helping you. You're helping me."
--:-- Ben Horowitz: If you had that, you'd fix the whole democracy integrity problem because everybody could vote off that address. Everybody would have an address because everybody would want the money. What bigger incentive to register to vote than that? In order to get money, you have to have an address which registers you to vote. That kind of thing could get us to just a much higher trust in our own institutions.
--:-- Tom Bilyeu: Speaking of that, I wanted to absolutely scream into the abyss when I heard that people couldn't retire from the government faster than the elevator would lower their records down into a mine. I was like, "What is happening?" What do you take away from why Elon wants to do this? Why is he sleeping in hallways? Why is he doing this? Is it just to get government contracts and it's nefarious in the way that so many people think it is? Or is there something positive there? What's the game?
--:-- Ben Horowitz: I think there are a couple of different things. The strong thing is that he truly believes that America is the best country in the world. He is an immigrant. And it's not guaranteed to stay that way. We have been in danger of losing it. The most important thing for him to do in order for his companies to be relevant, in order for going to Mars to be relevant, in order for anything he wants to do in life to be relevant, is we've got to stabilize the U.S. government. I think that's the main thing driving him.
--:-- Ben Horowitz: How did he get to that conclusion that the whole country is in jeopardy? It was a pretty interesting thing to watch. Right in 2021, I think he was a Democrat and he was certainly pretty apolitical. I was actually in a chat group with him when he got the idea or posed the question, "Should he buy Twitter?" A lot of it stemmed from the U.S. government just harassing him, which was a very odd thing.
--:-- Ben Horowitz: Like, here's your – I mean, I think you could very well argue he was our most productive citizen. He was our entire space program. He advanced the state of electric cars by 20 years. He's still like 95% of – there was something like that percent of the electric cars sold in the U.S. Jesus. You know, he's done the things with Neuralink to help people use their arms and legs who have been paralyzed. So, you've got a really kind of remarkable person to want to pick on.
--:-- Ben Horowitz: But what happened was because he got this PR for being very wealthy, the Biden administration targeted him. And again, they're fascists. So it's really a power struggle always with the fascists and anybody who looks like they're becoming powerful. One of the things they did that's talked about a lot was they sued him. The Biden Department of Justice sued him for discriminating against refugees. But he had a contract with the U.S. Department of Defense that required him to only hire U.S. citizens. So he was breaking the law either way. And they never dropped the lawsuit, even after it was pointed out by Congress. So it was clear harassment.
--:-- Ben Horowitz: I think his conclusion from that was, ironically, that we are losing the democracy. We're going into this very strange world where the incentives are all upside down. And the way Elon thinks, it's up to him to save it. So he got extremely involved. Then I think the more involved he got, the more he realized a lot of the things really were dangerous, and secondly, that he personally would be somebody who would know how to fix it.
--:-- Tom Bilyeu: And you go, "Why the hell would Elon Musk know how to fix the government?"
--:-- Ben Horowitz: This is the thing everybody's saying now. It's funny because I told this to Andreessen years ago. I'm a big fan of Isaac Newton. We always talked about who Elon is like, what entrepreneur comes to mind. It really wasn't Thomas Edison. But Isaac Newton was the one I always thought he was most like. The reason I thought he was like Isaac Newton was at the end of Isaac Newton's life. I think he was in his late 60s, maybe 67 or 68. For those of you who don't know Isaac Newton, he figured out how the entire world works and wrote it down in a book called *Principia Mathematica*, which is probably the most amazing work in the history of science. He did it entirely by himself. He didn't even talk to anybody.
--:-- Ben Horowitz: So he gets to be in his late 60s, and the Bank of England has a crisis causing a huge crisis for the whole country. There's a giant counterfeiting problem. The currency is going to be undermined and England's going to basically go bankrupt. They had no idea what to do, so they call Isaac Newton because he's the smartest man in the world. Isaac Newton, 67-year-old hermit physicist, goes in and says, "Okay, I can help. Make me CEO of the Mint." So they make him CEO of the Mint, head of Doge, whatever. He reorganizes the Mint in like a week, fixes the technology in a month, and completely makes it impossible to counterfeit. Then he becomes a private eye, goes into all the pubs where the counterfeiters are, arrests all of them. Then he learns the law, becomes the prosecutor, prosecutes all the counterfeiters, and has a 100% conviction record. And that, by the way, is Elon.
--:-- Tom Bilyeu: I didn't know that part of his story.
--:-- Ben Horowitz: Oh, yeah. So it's an amazing thing. If you look at Elon and DOGE, the most remarkable thing is how he's done it. If you or I were to say, "Let's go in and get the waste and fraud out of the government," what do we do? We'd audit the departments. No, no, no. That's not how he thinks. He thinks, "How do the checks go out? How is the system designed? When does the money leave the building?" It all comes out of one system. "Let me have access to that system and I'll look at all the payments. I'm not asking anybody what they're spending. I'm looking at what they're spending. I'm getting to ground truth, and I'm going to work my way backwards from there." Not only is he not unqualified, he's maybe the only person qualified to figure out how we're spending $7 trillion. He's a very unique individual. He's also a troll. He likes upsetting people. I get all that. But what he brings to the table is pretty interesting.
--:-- Tom Bilyeu: Yeah, I would say very extraordinary. You have also written about another extraordinary historical figure from the Haitian Revolution, a guy named Toussaint L'Overture. Tell us about him. There's something about this moment, about being a master strategist, about using what you have, being creative, that feels very apropos to this moment. What made his story special?
--:-- Ben Horowitz: Toussaint was another one of these characters in history. I call them once in every 400-year type people. In the history of humanity, there's been one successful slave revolt that entered into an independent state. If you think about the history of slavery, which goes back thousands of years, there's a lot of motivation to have a revolt if you're a slave. But why only one successful one? It turns out it's really hard to generate an effective revolt if you're slaves because slave culture is difficult. You don't own anything. You don't own your own will. You're at the pleasure of whoever's running things. Long-term thinking doesn't make sense. Why plan for next week? It doesn't matter what you plan. It's not yours. So everything is very short-term. Short-termism is difficult in a military context because in order to have an effective military, there needs to be trust. You have to be able to trust people to execute the order. But trust is a long-term idea. "I'm going to do something for you today because I trust that down the line you'll do something for me." That doesn't exist in slave culture because there is no tomorrow.
--:-- Ben Horowitz: If you look at Haiti at the time, you had the French army, the British army, and the Spanish army all in there fighting for it. Really well-developed, strongest militaries of the era, all very interested in the sugar, which was quite valuable. How in the world would you ever get out of that? He was probably the great cultural genius of the last several hundred years. Although he was born a slave, he was very integrated into European culture because he was so smart. The person who ran the plantation took him to all the diplomatic meetings. He mastered European culture and the different subtleties. He started adopting those things and applying them to his leadership. Furthermore, he incorporated Europeans into the slave army. He would defeat the Spanish, capture some guys rather than kill them, and incorporate the best leaders into his army. He built this very advanced hybrid fighting system where they used a lot of the guerrilla techniques he had brought over from Africa, combined with the more regimented discipline strategies of the Europeans. He ended up building this massive army and defeated Napoleon and everyone else. It was quite a remarkable story about how he just figured everything out from his principles. In a way, that was very much like Elon in that sense.
--:-- Tom Bilyeu: One of the things I heard you talk about that I thought was so ingenious was he would basically use song and sound as encrypted language.
--:-- Ben Horowitz: Yeah, that was a very cool thing. Remember, this is in the days before telephony or the internet. It's pre-Alexander Graham Bell. The Europeans were on notes, carrier pigeons, guys running back and forth. As a result, you needed the army together in one place just to communicate the order. Toussaint had these drummers and these songs which he could put on top of a hill, who could be very, very loud. He would separate his army into six or seven groups. In the song would be embedded the order of when to attack, when to retreat, all these kinds of things. He had this super advanced, wide area communication system that nobody else had. That was a big, big advantage for them.
--:-- Tom Bilyeu: That to me, the reason that comes up for me now is we have all these new technologies coming online. The person that's going to be able to get outside that box and see something new and fresh is going to be able to use this in totally different ways. While in the final analysis, I think you and I see it very differently in terms of AI's ability to ultimately gobble up what humans can do. But right now, AI is this incredible tool. As an entrepreneur, it has been ridiculously exciting to see how much farther each of my employees can push their own abilities by using AI. It does not take much to prognosticate out 12, 18 months to understand where the tools are going to be and how much more they're going to let you do. We're largely an entertainment company. For us to look at that and the revolutionary changes, you can't be trapped inside the old way of thinking. You've got to build up from first principles.
--:-- Ben Horowitz: It's a new creative canvas. I think that's a really great way of thinking about it. Is your creativity going to be used on frame-by-frame editing of a video, or will it be thinking of incredible new things you can do in a video ad that you could never do before and have the AI do that for you? It's a readjustment of where you put your creative energy and the things that are possible. We're really seeing that across the board. In our firm, we're applying a lot of AI. You'd think, "Does this mean you don't have human investors anymore?" It's been totally the opposite. Instead of painstakingly collecting all the data needed to put the investment memo together, it just does that for you. Then you're just thinking about what the really compelling things about this are. Rather than trying to track every entrepreneur and great engineer in our database, the AI is just tracking all those people and letting you know, "Hey, that guy just updated his LinkedIn profile," or "That guy just put out an interesting tweet. Maybe you should call him." That's a much more fun part of the game.
--:-- Ben Horowitz: I would say the best predictor of how things are going to go is more like what's happening now than the most dystopian view of it that we can possibly think of, which is where a lot of people go. I think some of that's the name: artificial intelligence. We hate everything artificial, so why do we name it artificial?
--:-- Tom Bilyeu: That's too true. Ben, I've enjoyed every minute of this. Where can people keep up with you?
--:-- Ben Horowitz: I am @BHorowitz on X. That's probably the best thing. We're a16z.com. Hope you enjoyed it. It was great fun catching up.
--:-- Tom Bilyeu: It was indeed. And you also have multiple books that people can read that are extraordinarily well-respected in the field. So thank you for those.
--:-- Ben Horowitz: Absolutely. Awesome. Thanks so much, Tom.
--:-- Tom Bilyeu: Well, thank you, brother. I appreciate it. All right, everybody, if you have not already, be sure to subscribe. And until next time, my friends, be legendary. Take care. Peace.
--:-- [Outro music begins]
--:-- Announcer: Thanks for listening to the a16z Podcast. If you enjoyed the episode, let us know by leaving a review at ratethispodcast.com slash a16z. We've got more great conversations coming your way. See you next time.
