Chris Dixon has been a general partner at venture capital firm Andreessen Horowitz since 2013. In 2018 and 2020, he led the creation of a16z's two crypto funds, their combined weight in capital breaching $865M. Chris was formerly the cofounder of SiteAdvisor and Hunch, acquired by McAfee and eBay respectively, and is on the board of Coinbase among others.
In this podcast episode, Chris and Daniel talk the future of AI, the fundamentals of blockchain, the likelihood of aliens and cults that beget industries.
Pioneer Podcast Episode #3
AI today and tomorrow
Daniel Gross: Okay, so maybe we could start off talking about AI. Obviously, you started SiteAdvisor which is awesome and actually, by the way, deserves a return in my humble opinion. Google results have become quite unusable. And then you ended up starting Hunch which is a master recommendation engine using machine learning. It's where I met you originally, at Hunch offices, I remember that very vividly. And then that got acquired by eBay I think in 2011.
I'm kind of curious to get your take, I guess, on– If we take a look at the past maybe, I don't know, five, ten years or so, and a five-year-old asks me, "AI. My parents read these articles about AI. How is my life changing because of AI?" I'm not really sure I could point to much. I can point to a lot of small, incremental improvements, but I can't point to anything–
Chris Dixon: Well, Siri. I don't know if it's changed their lives but certainly it's widespread and widely used, I guess.
DG: Right. So my question to you would be, in the next five years, it's now 2025, do you think my day-to-day life is significantly different? And if so, how?
CD: I'd love your view on this. I don't do AI anymore professionally, but it's not because I'm not extremely bullish on it. In fact, I think blockchain, crypto and AI, are by far the two most important things going on right now. I just happen to think AI is, for a variety of reasons, probably a lot of the benefits are going to accrue to incumbents, not startups. That's not to say there aren't tons of startup opportunities, but anyway. I can talk about that more later.
You tell me, but the stuff I see coming out of labs and things, it feels like there's a dam that's going to break, that we've only seen the beginning. My understanding is the quality of the results, so like GPT-3 and this kind of thing, just the sizes of models, that's growing very quickly and it's not slowing down. So like there's sort of a Moore's Law type thing going on here which is, "How big is your known network? How much data can you put in there?" And that seems to be going up at a very rapid rate and as far as I know, not slowing down at all, number one. So the core quality of the AI is getting better.
Number two, very importantly, it's being packaged in a way that's accessible to developers, so this is everything from Cloud kind of compute training services to TensorFlow to all of these really nice software packages to basically every client-side computer now has specialized chips for running models in deployment. That's very important, because you need it to get to the point where it's like AWS or something like this or just like a database where every piece of software, including accounting software but they don't have a big team of machine learning people, is able to use the latest stuff.
So you tell me, but I think that it's just a matter of getting this stuff out to market, a matter of the models increasingly getting better. I think another really interesting trend is so much of the last decade was on images, AI, ImageNet, identifying what's in an image. Second half of the decade, really cool stuff with generative images, creating images. Now, a lot of the action is moving to text which just unlocks a whole wide range of new applications.
So I think it's very likely that... I guess a couple things. If you go back in the history of AI, it was sort of fits and spurts. There were all these kind of AI summers and winters. I think we're now in the real thing. There's no more winters because of the things I was mentioning before in terms of the quality of the models and things like this, but also the economics behind it. There's now a business model for AI, search and ads and all these. There's all these other reasons why all this money will keep flowing in, whereas in the past they were dependent on government grants and kind of the whims of those various funding sources.
So I think it's still early. I got really intrigued by machine learning I guess back in 2007 or 2008. We sold our company to eBay, machine learning company, in 2011. I thought we were at the end. It's so ridiculous in retrospect. I thought we were late to the game and we better sell before the AI thing. Of course, it really began, the modern era, I think in 2013 with the Google cat video experiment that came out. I think we sort of consider it the kind of watershed moment when at least the epiphany happened. People realized. The underlying models, people were working on that for obviously decades, but that was when people were like, "Oh, wow." And if you look at the ImageNet results and all the other kind of metrics, it just got dramatically better.
That's the other thing. We're seven years into it. It takes a while to get these things built out, productized, deployed throughout the world, but I think it's going to happen. I think it's a major important area in the next couple of years.
DG: So five years from now I'm walking around the earth. Do you have a sense for what's going to be changed and different?
CD: I think the self-driving cars, we don't know exactly how far away they are, but they're going to happen. It's one of these things that's so hard to predict. It could be Waymo's– my understanding is they're running in Phoenixdriver. There's still limitations and there's problems around different lighting conditions and weather and it's not perfect yet, but it's just a matter of time. Three years to ten years or something probably. Could be 10 years before they kind of get all the kinks out. Maybe even longer, given that it requires cultural changes. It requires mindset shift to be willing to go into a car that no one's in. It requires a new system for insurance and regulatory and a whole bunch of things.
Maybe it takes longer, but that's going to happen. You're going to have autonomous drones. You're going to have autonomous cars. It's going to change transportation. It's probably going to change manufacturing, agriculture, any place you use vehicles and big, heavy moving machinery. That may take longer, I don't know. But anything where there's sort of physical interaction with the world will probably be mediated through some form of AI over the next coming decades.
Information work which is what most of us do now, you sit in front of a computer and you do stuff, having a cyber sidekick there fixing your spell check and your grammar, and the next thing is suggesting how to finish that essay. What's the best final paragraph for the blog post? I think maybe for a while it will be kind of this "centaur model". There was a period in which the best chess player in the world was a human plus a computer. I think the computer alone now is best, but I think you'll probably have an intermediate period where people are sort of assisted by AI in a whole variety of tasks and eventually, probably the AI takes over more and more of that.
I think one thing that's important is the robots don't come looking like robots. If anyone here, the people watching are probably too young, but there's an old TV show called The Jetsons. It was a cartoon where it was the future and what happened and the way that AI was delivered was there was literally a robot butler who embodied the AI, and that was sort of... And they would go take the jobs, but in reality that's not what happens. You don't literally take the accountant and replace it with a robot accountant that looks like an android. What you do is you just have really good software that is AI-assisted. Suddenly, instead of having 10 people in your accounting department you only need two people. That's how robots actually take over the world, is very subtly through kind of boring-seeming enterprise software.
DG: Although, I do think in the Japanese version of the future it is an animatronic robot.
CD: Those Honda animatronics. What's that? Boston Robotics, those scary...
DG: Very interesting that that thing was just bought by Hyundai again. It seems to be the tumbling pin of Silicon Valley from one company to the other. Okay, so that's kind of interesting. We're talking a lot I guess about digital innovation. One interesting dichotomy I've been noticing is you kind of, of course, see a lot of the American AI laboratories, notably OpenAI and Google to some extent too working on the software stuff, and of course the holy grail here that everyone is secretly excited about it is code that can write code. A self-programming AI is kind of the penultimate goal. But what's interesting is if you look at the UK labs, it's quite different. DeepMind in particular is going very much after the material science world. I was kind of curious if you had any views on how synthetic biology or how the physical world might change as a result of AI as opposed to just the software world.
CD: Yeah. So I'm getting pretty far out of my area, but I will say... So I have a colleague, Vijay Pande, who founded our bio fund, so we have a bio fund at Andreessen Horowitz, and the thesis behind our bio fund is why... Our firm is very much orientated around software, and traditionally there was sort of a separation between software, venture firms, and bio investing, so they would do molecule development, things like this, and we would do software. Our belief is that the two are kind of converging or intersecting much more, which is why we felt like we had the competency to sort of go and enter biology.
Specifically, what Vijay would say is that because of software, machine learning, et cetera, we are now able to engineer biology instead of kind of going and doing trial and error and sort of a much more... Think about how, for example, drugs are discovered. It was in the past a much more physical trial and error process, and today you can actually go and using software do much more engineering in a much more predictable way. Again, I'm getting way outside of my... I'm only a hobbyist at best with respect to biology, but my understanding is that that's a major area. You mentioned material science. That's another important area. We have an interesting company that uses artificial intelligence to search for mineral deposits as an example. You're just going to see, I think, all sorts of interesting applications like that.
I think probably the gating factor right now is just expertise. It's just you don't have that many people yet, you will soon, who are deep into this stuff and can go and work on all these different problems. Solves itself partly through the market, but more... Now the most popular class at Stanford Undergrad is the machine learning class and it solves itself partly through tools.
DG: That makes a lot of sense. I feel like you are consistently early to scenes that end up becoming big. Really the only fault one could take with Chris Dixon is that you're too early to things. I mean, you were dancing at the AI party when the club was empty and you left-
CD: That hits home. I definitely feel like... Yeah.
Cults and innovation
DG: But it's amazing. You're always early and you were obviously early to crypto and more than made up for it I imagine with angel investing too. I feel like you were early to a lot of trends. How does one become early to things? How should I be-
CD: It's actually, I think, fairly straightforward. I wrote a blog post I called... I think it was what the smartest people do on the weekends, everyone else will be doing at work in 10 years. Something like this. When I got into machine learning it wasn't like there weren't people into it. There were a lot of really smart people into it, but it was kind of a cult. I find that's a really common pattern.
DG: What are the cults right now?
CD: Well, I think they're all over. By the way, they're not just in technology. So if you go back, my understanding is Chicago comedy in the 1970s was a genesis of all modern kind of SNL whatever, a whole bunch of... New York film scene in, I think it was the '80s maybe of like–
DG: Historically there's all these groups–
CD: Yeah. It's these small cultish groups. I think it happens in academia. I'm not an expert in this area. I mean, I know something about maybe historical examples, but today I bet you there are all sorts of interesting... Maybe they're happening on the internet now. They used to happen physically. But they're kind of cultish groups. They tend to be people who are motivated because of the interestingness of the problems, not because they see any kind of pay out or other kinds of things.
Oh, by the way, one of my favorite examples in that is computer graphics in the University of Utah in the '70s. You go back, there was one I forgot, but it's one rich endowment trustee who gave this grant to do computer graphics, and it was the only place, one of the only universities in the country that took this seriously. And if you go back, it's Adobe, Pixar, just Atari, Apple, everybody was there, and it was 15 people and the entire modern industry was built by that. Machine learning, by the way, we can talk about that, the fact that they're all Canadian and they have these Canadian grants.
DG: Geoff Hinton is the patriarch. Definitely. Or one of the patriarchs. Yeah.
CD: And you think about it like, "Why were they working on it?" Just going back to neural nets. Why were they working on neural nets in the '90s and 2000s right? I mean it was pretty eccentric to do it then. They had this view that that's how the brain works from what we know, it has sort of a neural network structure, and therefore this should eventually work, but that was very hard to stay excited about that in those decades when the results just weren't good. When we did Hunch, neural nets just didn't work. Turned out you just needed more GPU power and all those other things and it eventually worked, but at the time the results just were very poor.
DG: Even at us, for Cue in 2013, and this is after the famous Hinton paper which ultimately led to that cat situation with Google that you're talking about, it was still believed... I mean, we'd sold to Apple and we were working on machine learning on Apple's budget, and it was still kind of believed that yeah, the stuff is super dumb, maybe it can do cool toy examples, but it doesn't really work. But I love this idea of passion cults are the stories often of what creates future technology.
CD: Homebrew Computer Club.
DG: Exactly.
CD: Unix. Open source software. All these things were weird cultish clubs.
DG: Every person listening to this is only wondering, "What are the cults right now?"
CD: Pioneer.
DG: Pioneer and crypto. Great. We got that out of the way. So what are the other cults?
CD: I mean, I have my theories. I'm going to be wrong about a lot of it. I don't know. You have to be deep into an area. I bet you there are some kind of interesting new databases out there that are I don't know what. They're a streaming database or NoSQL. I don't know. I'm not a database person anymore. I don't really hang out there, but I bet you can pretty much take any area where there's a bunch of smart people working and there's some probably subset that's an interesting cultish group.
I can speak a lot about the cults I hang out in, which is crypto. What else? There's a lot of very interesting stuff happening around video games today as an example, but I think you could imagine a lot of those things. I think the video game industry probably, even though it's 150 billion dollars and people I think are starting to grok it. I think it's probably still being significantly underestimated in terms of how profound. I think it's by far the most important cultural medium and still is sort of considered, I think, as underestimated.
Look, I don't have all the answers here, but I think the characteristics are a bunch of smart people motivated by interestingness as opposed to near term profit, because usually things are just too far off to see that. And they tend to have this characteristic that the deeper you go into, I think crypto is very much like this, the deeper you go into it, the more interesting it gets. There are fields where the deeper you go in you're like, "Really? That's it?" The cult ones, no.
Another key characteristic of the cults, of the good ones, the ones that start off toys and become a big deal, is the rate of improvement. So this is kind of Clay Christensen's core insight, I think, in Innovator's Dilemma. His core thing is his chart where human needs stay relatively flat, like my need for, "How good of a payment processor do I want as a human?" or something like that, demand stays relatively, not constant, but it goes up at a certain rate, and then the technology goes up at a much faster rate. Those are the disruptive technologies. So it's this simple observation that these two lines that look far apart now and many people think it's a toy, those lines will get much closer.
The canonical example is the telephone versus the telegraph. Back when Bell first came up with the telephone, Western Union, which was the incumbent, passed on buying the patents because they had a big telegraph business, and they talked to their kind of enterprise customers who were railroads and other people and they were like, "This thing is great, and the telephone only goes a mile and it's a toy," and it's like why would you want to hear the person's voice when I just want to send them business info that I can do very efficiently through telegraph? But of course the phone, once you invest in it you can go more than a mile. You could go much further and you could do all sorts of... The handsets got better and all these things. The rate of improvement was so great. People focused too much on this gap between what you want and what there is today, when you should be looking at the gap plus the rate of change.
DG: The derivative. You mentioned earlier something very interesting which is people that are very focused on it for the interestingness of it, for the purity, often they don't really know why they're focused on it but they are. How do you think about that with crypto, because it does feel like with crypto there's also a very big commercial thing going on because the innovation is money innovation? Does that distort? How do you see through all the fake scenes into the real ones?
CD: Can I just say one last point on the other... To this point, I wrote an essay a few years ago that ended up being published in the Atlantic, and it was about the history of computing and how it related to logic. Actually, that essay was meant to be on this... The title got changed and things, but originally it was going to be called... The idea was that nothing interesting is a waste of time, that as long as you have a bunch of people that are really interested in something, it ends up having these inadvertent practical results.
And to me, one of the most fascinating examples of that is formal logic. So logic, there was a great quote, I actually have it in the essay, from a computer scientist. It said, "If you hired somebody in 1905 to survey the world and find the most useless, impractical subject on the planet, it would be these 10 people, who are Alonzo Church, Bertrand Russell, Kurt Godel, these 10 people who thought this was an interesting topic to do this formal logic stuff," which literally ended up being the entire Alan Turing, the entire foundation of computer science, and why after World War Two it all kind of came together and you had this explosion and became clearly the most important invention of the century. Any smart group of people would have said it's the dumbest, biggest, most abstract waste of time.
I tried to kind of trace out the genealogy of those ideas in that essay. If people are interested, it's on my blog, cdixon.org. I think you could actually go further on this concept. It's not just a technology concept. I think if you go and you do the kind of genealogy of ideas throughout history, I only know a few subject matters like logic and things like philosophy and computer science and things, but I think in general this is a pattern.
I think of it this way. It's about time horizons. Most things in the world operate on a two to three year time horizon. Business operates on a short time horizon. Public companies operate on quarters. If they have a really visionary CEO they can operate on a three-year cycle. There's very few institutions in the world that can operate on 10, 20 year cycles. One of those is universities, and another is hobbyists and hackers. The stuff they do on the weekend is because they don't have a time limit. They can actually do a 10-year time horizon. Anyway, so that's my... I feel very strongly about that thesis. It's frankly the only big pattern to look at the world through that lens.
10-20 year projects
DG: Before we move on maybe to the crypto thing, actually on that point, one issue I think with people looking at things that have a very long time horizon is it's not really clear what metric they should be optimizing. Say for example I wanted to build space elevators, which actually is a... It's very achievable to do. It does require 10 year, decade-level thinking. How does one even think about kind of raising capital for that and proving that they're making progress, when at the end of the day... Say if you're working on this, you're working on, I don't know, nanobots. Again, very achievable. Feynman laid out a path 60 years ago. It's just not really clear a few months in if you moved at all. There's no revenue. So when you have something long-term, what are the short-term bio markers for progress that you think of?
CD: It's a great question. And look, by the way, it could be that some of these things, just the floor profit capital markets fail. They don't work maybe. The counter example is stuff like SpaceX and Tesla or Elon Musk. He did manage to raise money, and continues to frankly, for years and decades with his big, ambitious ideas. I think, and by the way, a lot of the stuff I was talking about before like the crypto and software, those things aren't that capital intensive so you can have people just sort of do it on the weekend. So you could argue there's sort of this other category which is really long-term stuff that is capital intensive like space elevators that is not properly funded today. Some people argue that that's where government should come in or universities and others. I am kind of sympathetic to that view. I think generally people overstate the short-termism of venture capital. It's actually a very patient asset class in my experience.
DG: Totally. But here's the interesting thing about Airbnb. They can go and they can point to a bank account that's growing. What I wonder is if I come to you and I say here's my idea for nanobots and here's my Google doc, I've researched it 25 ways to Tuesday.
CD: Yeah. I think you have to do more project financing, like the way that people build oil wells. What they do is they have a very advanced and sophisticated kind of methodology for assessing the various sorts of technical and operational milestones. So basically it's traunched, so you do some big... I think big companies are pretty good at this with like the PlayStation 5, I don't know how much that cost but I'm sure it's billions and I'm sure they had many sophisticated internal milestones as to how it's going and at what point they should change direction. They probably had multiple teams working on various things. I assume they do things in parallel. Maybe you saw this at Apple. I don't know. Probably can't tell me if you did, but I think big companies are very good at this sort of sophisticated project planning.
I think it's an interesting question, should there be more venture firms that operate that way? Are there pools of capital that are willing to fund sort of 10 to 20 year projects that think about the world more in terms of technical and operational milestones as opposed to the way the venture world works today, which is all recurring revenue and all these other kinds of metrics that really work well with software but maybe not with space elevators.
DG: Yeah. I mean, I think the tricky thing is, at least from I guess what we've seen that I'd be curious to get your take on, is person basically says... Let's go with the nanobots. So they manage to raise $20,000, and then they come back to you and they say, "Look, I blew it all. Here's some nice renders of stuff," and you're like, "Well, I'm not a nanobot expert. I don't understand really." And then they say, "I now need five million dollars to go to the next step," and I don't think there's a normal venture firm in Silicon Valley that would fund that right now.
CD: Counter argument, Luminar. Luminar is a lidar company, but they just went public through SPAC, having raised hundreds of millions of dollars. I generally agree though. Or like Boom. They're building supersonic jets, right. I mean, there are a bunch of these examples. I think the founders of those projects have to be kind of larger than life, like Elon Musk style, to go and manage to raise that much money. But it's a good question. I do think there's an argument. I'm sympathetic to the argument that maybe there's too much short-term investing and too much stuff on software.
DG: One realization I had yesterday watching the giant explosion that was Starship, the most celebrated failure of all time, is you kind of mentioned maybe two axes and maybe there's a third. So you mentioned whether there's revenue in the short-term, you mentioned whether it's capital intensive or not, but then there's a third one that I think is quite interesting which is how visual the product development is. You look at what Elon's producing, it's visuals. I mean Starlink is not live globally. None of us are really benefiting from SpaceX, and yet people are extremely excited about it because it's visual. Boom, I think, only works because it's visual. I actually think you would really struggle to make a nanobot company because it's not visual, yet a space elevator is. So in my opinion, that's the real loser, something that's capital intensive. And this is where I think a lot of synthetic material biotech stuff really loses is it's capital intensive, there's nothing to look at, nothing to get excited about, and there's no short-term progress.
CD: I think this is the... It's sometimes called the World's Fair model, like Tesla coils and space rockets and Martian colonies. I think people, if you go back and read the Founders Fund manifesto, we asked for flying cars and we got Twitter. All the things they talk about are these World's Fair things, like Tony Stark, rocket ships, and they don't talk about nanobots and vaccines. There is a bias towards the spectacle. And so then the question is sort of the things that don't create a spectacle, that are capital intensive, how do you fund those?
DG: Yeah. Or even things that create the spectacle. I've been pretty disappointed just in terms of man's ability to manipulate nature and just command energy. I mean, we're not a Kardashev Type One civilization yet, and it seemed like... What I really wonder is 50 years from now, 100 years from now, will we kind of look back on this era and be like, "Well, we were really good with atoms. We were doing nuclear reactors in space, and then we took a detour through all this iPhone stuff." Imagine an alien civilization for a moment that is still using radio transistors but can fly wherever they want to, WX or WSAD, and they can just command nature to do whatever they want. Every person has terawatts of energy available at their disposal. I wonder if they're living a better life.
CD: Have you ever seen Back to the Future 2, so the one where he goes to the future?
DG: Yeah.
CD: It takes place 2018 or something, so now. It was made 30 years ago or something. What's striking about it is they have flying cars, he's got a hover skateboard, and they have no smartphones. They just completely missed. By the way, I was at the dentist recently and they had Friends on, the TV show, which I hadn't seen in decades or whatever, and it was striking. I think it's from the '90s, right? It's striking is it looks exactly like today. The world has not changed, except for one giant thing which is smartphones. All the plot devices revolve around the two people not communicating in a way that would ever happen today because they would just text each other. It's just striking how in 20 years, to your point, all these predictions were this futuristic world, Blade Runner, Back to the Future 2, flying cars and all this innovation on the physical axis, and we have none of it. Nothing. Literally. I think with the exception of a few buildings or something, you could–
DG: We've slowed down.
CD: The modern world is indistinguishable. I think Peter Thiel talks about this–
DG: Yeah. Hasn't changed.
CD: But it is striking just how good the phones are. We've had an incredible run on that. It would be mind-blowing to somebody in the '90s to say that we have this super computer, internet super computer, all of the world's knowledge. Marc Andreessen always likes to say, "Of course you're staring at your smartphone, because it's a super communicator device with all the world's knowledge and all the human beings on the earth. What else is possibly more interesting that's in your physical room than this phone?"
DG: Something queens and emperors would have murdered nations for.
CD: Yeah. On the one hand it is a miracle, and the fact that it's accessible to most people on earth... You can get an Android phone for $15 or whatever now. So let's not discount that, the importance of software and smartphones, but you're right. How do we get that same kind of level of innovation everywhere else is a big question. There's different answers. The capital market is one we're discussing here. There's regulatory questions. There's an argument that a lot of things have just been banned. You can't do a lot of things. There's cultural questions. There's all this sort of pushback now any time there's new tech. Although, you could argue there always has been, but you go on Netflix and there's all these shows about how terrible tech is. The newspaper is full of how terrible tech is and are we creating the right cultural forces and I don't know. There are people that they want to go, they grow up somewhere, and my goal in life is to go and build space elevators. Are we creating people like that?
DG: On that front, I mean it's a question I've pondered, I'm sure you've pondered for quite some time, if at the end of day everything is upstream of culture and it's really Netflix is driving the software that is powering the humans, should we all just quit and be directing movies?
CD: I think that's another Peter Thiel argument, which is there's no... I think it's his argument. There's no positive science fiction. It's all dystopian. I think you could argue maybe that's just because it's a better story. No one wants to watch a story where everything is great.
DG: It all works out!
CD: Sit around. It's all great. [laughs] Like The Martian comes to mind as a movie where I think it was a very positive science movie and I think was a good movie too, so it is possible to have pro-science. On the other hand, if you go read a lot of the history of the early airplanes and cars, there was a lot of resistance back then too. There's always been a lot of resistance to change, so maybe that's just part of the process.
DG: It does seem like a fundamental human condition that it is easier to worry than it is to dream. I think it's just maybe metabolically cheaper, and so people are... It's easier to just pontificate about how Facebook is bad for you as opposed to think about how it's good.
CD: Yeah. It always sounds smart to be kind of critical.
What is a blockchain & democratized decision-making
DG: It's easier to sound smart. That's right. That's very interesting. So to crypto for a minute. I think as far as I can tell right now the world bifurcates into three groups when it comes to crypto. There's the believers, the people who are truly entrenched, the zealots and the priests of crypto. They're the extreme. Many of these people are of course the gate holders of capital over the world. There are the permanent nonbelievers. And then in the middle sits most of the planet which is, "I don't really understand it. I read about it occasionally and maybe I dabble in trading it, but I don't know how to think about it." So I guess for that group – and I'm in that group – my question that I wonder a lot is if say a decade from now we look back and there's one Wikipedia article on the big moment where crypto became part of everyday life, what do you think that Wikipedia article reads?
CD: Yeah. Okay. Look, can I get to that in a little bit of a roundabout way, which is I'll present my view of what blockchains/crypto... We kind of call it both things. Crypto is what most people in the business call it, however, it has this sort of unfortunate ambiguity with cryptography. So what is a blockchain? One way to look at a blockchain is a new type of computer. It's a computer, so something like Ethereum. Ethereum is a computer in the strict sense of what a computer is. It has a way to store information and a way to process that information. You can write code through the Ethereum computer, you can deploy on a computer, that code can do things. It can store information, it can process information, and it can interact with other code out there in the same way you can on a mobile device or on a Macintosh or whatever. It's a computer.
Then you ask, "What's special about this computer?" Well, what's special about a blockchain computer is... So with a typical computer, there's the hardware and the software, and the person who controls the hardware, if I'm running code on Google's computer, Google ultimately controls the software. They can pull the plug. They can change the rules. If Google came up with a system like Bitcoin and said I'm going to have Googlecoin and there's only going to be 21 million Googlecoins, would you believe that to be the case? You would know that in the end whoever controlled that computer could change those rules, and we've seen that through the history of internet services. Facebook and Twitter and all sorts of things have changed the rules.
What's different with a blockchain is they invert that power relationship, so the software is in charge. The way this is architected, and in the case of Ethereum, is anyone can run a node on Ethereum. It can be what's called a validator or miners in the case of Bitcoin, but those people are running code where there's a virtual computer on top and even if those people who are running the code decide to change the rules, they can't change the rules because of the way that the system is architected so that the virtual computer that sits on top is resilient to any of the underlying hardware providers changing their incentives or becoming evil or et cetera. That's sort of the core insight with Bitcoin.
So Bitcoin came out 11 years ago. It was a specific blockchain computer with a specific application, which is it's sort of digital gold. I'm sure people here have heard about Bitcoin. I won't go on about it. But then what happened after that, particularly in 2015 with Ethereum, is people generalized that idea. They said, "Okay, Bitcoin is a new type of computer that runs a specific application. What if we generalize that and abstract it to just a new type of computer and people can build new applications on top of it?"
I very much view it, and I say all this because I don't think this is normally how people think about this and I think it's important, which is to locate blockchains in the history of computing. The history of computing, every 10 to 15 years you have a major new computing cycle. You had the mainframe computers in the '60's, PCs in the late '70s, early '80s, internet in the '90s, mobile phones last decade. Now you've got a whole series of interesting things happening. We talked earlier about machine learning, and I think blockchain computers are one of the profound things that are happening.
And why they're profound is they have new properties other computers don't have, and specifically, you can create internet services that can make strong commitments to their users or to the network. One type of commitment that they make is a commitment around digital money, so the commitment that Bitcoin makes is there will only ever be 21 million bitcoins. If you own a bitcoin and you have the private key, it's your bitcoin. No one can take it away. You can't double spend a bitcoin. It makes a variety of commitments that in turn give that currency at least potential value.
Ethereum is a much more general system where you can write arbitrary code that can make commitments. So I can write code, I could write a social network and that social network can make a commitment that if you're a developer and you build on top of it, that the rules will never change. I could build a marketplace and say this will be the take rate. I will take 2% of all the fees and that will never change throughout the history.
One of the very interesting things happening right now in crypto is what are called DAOs, Digital Autonomous Organizations, which is sort of a clumsy word for a new type of organization that exists only on a blockchain. So for example, we're involved in a project called Uniswap, which is a... I won't go into all the details of what the product does, but I think one of the really interesting features of Uniswap is it's owned by the users. In fact, they recently when they launched a token, they retroactively airdropped that token to all the users. So it would be sort of like if Airbnb back in 2015 after being out for a few years took 15% of their ownership and gave it to the Airbnb host. That just happened.
This is a reality that's happening today. There's all these new digital services that are both controlled by and ultimately the economic beneficiaries are the users, and they have that control and those economics through holding these tokens which control the system. They literally control the system. Uniswap is literally just autonomous code. The people that originally created it have no control over it at all. It's a very new concept. The code just runs autonomously, and it's controlled only by the token holders voting on changes to the system. So it's a new way to build an organization and it's a way to build an organization that's global from day one, that's digital from day one, that's owned and operated by the users. I think it's a very profound kind of new way.
Another interesting feature of those organizations is they use the token, kind of the ownership of the system, as a way to incentivize users. It was started by this other crypto thing called Compound which is a lending platform. It's another thing we're involved and invested in. What it does is whenever you use... Imagine a Facebook or an Uber where every time you use it you get tokens, which then in turn give you some ownership over that network. So it's actually owned by the users. I think we're actually getting close to being able to build these, and I think 10 years from now we'll have large networks built on the internet including social networks, marketplaces, et cetera, that are truly owned and operated by their user base.
DG: Our founding fathers of the United States of America very much didn't really agree with this approach just in terms of governance and very much were focused on finding a system where there were an appointed smaller group of people that was governing the nation as opposed to an actual complete virtual democracy. Do you think those people were wrong?
CD: I don't want to get into politics. I don't know about our–
DG: I'm joking with the Fox News question, but you kind of get my point. Would you actually want everything to be controlled by the-
CD: Well, okay. I think there's two questions. There's the question of do you need the people to vote on every single thing kind of like the way California does referendums versus representative democracy. That's question one. The second question is, let's just take core decisions like what should the rules be under which people are deplatformed, they're removed from a network? Do you want that decided by an opaque product management group that you don't know who they are, at one of these large companies, or should they be whether directly or indirectly, to your point, controlled by the user base? Like there's some process by which... And by the way, this is not that new of an idea. This is how the web works and how email works. There's no product organization in charge who can decide that, "Oh, this email, you can't send that email."
This is how the whole first year of the web worked and I would argue why we had such a great period of innovation and entrepreneurship and everything else. And then we moved to this other model which had other benefits, architectural benefits, because you could just make applications that were more advanced because the web 1 kind of thing petered out at some point. SMTP and DNS and HDP could only go so far. I think RSS is a whole long discussion we could have. RSS specifically was kind of the open version of social that failed to get traction and I think it's because it was–
DG: Although now it's having a total second coming with–
CD: Yeah. I think it could be. I think one of the things it's really missing though and the blockchains finally give it is state, the ability to store things. You never had a namespace with RSS. You could use DNS but it was clumsy. One of the reasons people flock to places like Twitter and Facebook is you go to Twitter and you can say, "I'm C Dixon and I'm going to follow these people," and that information was stored in a database managed by Twitter. RSS could never kind of keep up with that. Blockchains allow you to have, I think, feature parity with proprietary networks while also keeping the web 1 properties of being open.
DG: One curious thing. That's very interesting by the way. One interesting point I took from what you were saying is you could kind of construe actually that in the current paradigm, we do have a very distorted of total democracy in the sense where the most vocal people and indeed the most vocal users win. Even if they don't govern the platform, really all it takes is 20, 30 people who are super vocal and work at the New York Times to deplatform someone, whereas in a world where people could silently vote, maybe anonymously if the network supported it, you would actually get a much more accurate representation.
CD: I think it would be hard to design a worse system than we have today. You can criticize the things I'm describing, but I think that at least they're orders of magnitude better. Look, these are very important questions. I guess my proposition number one, I believe the internet is still very early in its development. I think we're 10% of the way, if not less. Number two, the killer app of the internet is creating networks on top of it. It's a network of networks, and those networks can be YouTube, it can be Twitter, it can be Facebook. I think some of the most important things built over the next 50 years will be new networks on the internet. And the question will be how do we build those networks?
This is why blockchains are so important. It's a new, very powerful way to build networks such that you can in a very granular way allocate control and money. Prior to blockchains, you had either the completely open thing like the Web which just ran out of steam, I think, or you have the corporate model. You have another kind of model which is like a Wikipedia model, but they have to beg for money every year and there's a bunch of other kinds of challenges. Occasionally works, but this is a very powerful new way to architect networks where you can take these very important questions, I think, of who gets deplatformed, what is misinformation– Shouldn't there be some open process in the same way that we have an open process for other things in society that are important? Shouldn't that be something that's done openly? And by the way, I'm not trying to take up the issue of who should be platformed or not. I do think people should be deplatformed, but there should be a process, and it should be an open process.
Satoshi, aliens and Keto media diet
DG: Certainly better to have a clear one than an opaque one. So here's a question that I feel like people were really curious about and then it kind of dropped, but I really think we need to be endlessly curious about it. Do you think 100 years from now, 200 years from now, let's imagine various forms of cryptocurrency are in fact running our space elevators and our life on Mars, are we ever going to find out who the author of Bitcoin was?
CD: Satoshi. It's a good question. My bias is sort of I'm glad that we don't know. I think it would sort of undermine the neutrality of the technology. The more informed guesses are groups of people, some people that were involved early on. I don't know. Yeah, you're right. I think people sort of stopped the parlor game of trying to figure it out.
DG: But it is incredible that... I mean, I don't think in the history of-
CD: By the way, it's not just... Bitcoin is certainly the big one, but one of the really cool things happening right now is a lot of the most interesting projects in crypto are made by anonymous teams. It's a very common thing. So if you went and polled people who are into this stuff right now, probably Yearn would be the current kind of hot thing. Some of them are known on Twitter but a bunch of them aren't.
DG: Let me ask you this. As a person who finances some of these things, often what keeps founders going and what keeps founders in check is at the end of the day whether they want to or not, they bind their ego with their company. They're public, the name of which they have only one, the identity of which they have only one, and that keeps them very motivated. At times crushingly and cripplingly so, but it does keep people motivated and also keeps folks accountable. Do you worry at all that either having a pseudonymous identity which you can create and destruct at will, will either cause immoral behavior or cause kind of less motivated behavior?
CD: I think that's definitely a risk. One of the mitigating things in crypto is everything is open source. Everything is run by communities. People get booted from the communities all the time. The whole point of the architecture is there's no need to trust individuals because you can trust the code. That's one of the reasons Bitcoin works, is it doesn't matter who Satoshi is. It doesn't matter if Satoshi is the NSA or evil. It doesn't really matter because all the code is open. Hundreds of thousands of people have read the code in detail and worked on it. At this point, it's sort of this evolutionary thing where now it's not even the same code anymore. It's been reworked and there's something on the order of hundreds of bitcoin developers who are just constantly kind of reworking it.
DG: Yeah. I mean, it spawned a whole genre of classical music but it's still very weird to me that we don't know who Mozart is or Schubert or take your pick. Fascinatingly so. I don't know if you have a registered advisor who can talk about this, but a related question that I'm very curious about is what's your take on the most interesting plot line of 2020, which is not the elections nor is it the pandemic, it's this, "The aliens keep on hinting at us that they exist"?
CD: The military keeps coming out with these statements that seem like they believe it exists too, I think, or something.
DG: I guess more crisply, do you believe in aliens now?
CD: Well, aliens that have visited or aliens in the Fermi paradox sense of it? It's Fermi, right?
DG: That's his paradox. Yeah.
CD: Yeah. I think just statistically they have to be probably out there. There's a, I think it's like... What is it? Some cartoon somewhere which was UFO sightings and then everyone suddenly had a camera and then it just dropped dramatically now that we all have video cameras around. There is that question of everyone has a video camera now, shouldn't this be the golden age of alien sightings? I don't know. I read all this stuff for fun but I don't have really an informed view on it.
DG: What do you think of the fact though that the United States military, this does seem to be a fact, is reporting that there is something that's been seen by multiple pilots that are obviously extremely trained at sighting objects that is moving at a pace that would indicate it's using a technology that is completely unknown both to this nation and every other nation, and they have no clue what it is and no explanation for it? What is your best 'it's not aliens' rationale as to what's going on?
CD: I think the argument I've heard is that it's effectively an optical illusion. They happen to catch something at a weird angle. Look, I don't know. I've watched the YouTube videos of the–
DG: So fast.
CD: I don't know. What do you think?
DG: What I find so funny is that the reason I'm struggling is because when I was growing up believing in aliens was basically being a crack pot, and so my mind is really struggling to adapt to that. And I'm wondering, I grew up orthodox, I always wondered this about religion, what level of miracle do I need to see before I'll believe and it's impossible that there is nothing there.
CD: Well, the other thing is that I think just the basic intellectual humility should be that we're probably wrong about half of the important things in the world. People historically have always been wrong. Germs were thought to be just absurd until whenever it was.
DG: Lobotomy.
CD: Yeah, and like radio waves, so I think just the default assumption has to be that of the 20 things that we say are ridiculous, probably some reasonable portion we're wrong about, whether it's ESP or aliens or I don't know what. New types of fusion energy. Whatever. Scientific kind of inventions. If you go read books about the Wright brothers, it was just absolutely scientific consensus that flight would never happen. It's just shocking the level of conviction that people had about that. So I think the default assumption is we're wrong about just a whole bunch of stuff. I don't know which things we're wrong about, but I think we must be wrong about most of them. So I try to just sort of use that as my baseline. It seems very unlikely that we're at this magical point in history the first time where they've sort of had everything figured out.
DG: I think that's right. It's funny to think of what the Wright brothers went through and Da Vinci effectively in terms of the flying machines that didn't work and the wheel flying machine. It's very similar to what happened with deep learning, obviously in a much shorter timescale because everything is moving much faster now so that was 300 years, and with deep learning it was about 40 or 60 years or so delta, where you kind of get the first wave, that gets proven wrong, gets written off, and then the second wave makes it happen. It's very similar to your famous strong versus weak technology blog post.
Okay, so we're slowly running... I mean, we're fairly quickly of course running low on time. Any good content Chris Dixon has consumed lately? And this can be anything from a TikTok video to a YouTube video to a podcast to a book.
CD: Yeah. I've read a lot of books this year. It's one of the silver linings in this whole otherwise bad year. I've read a lot of science fiction. I should have read this before. I read the Foundation series. I loved that. Asimov. Lately I've been reading historical fiction.
DG: Ken Follett has a new one.
CD: Yes. Used up on my Ken Follett. Have you read Alfred Bester? This is great. The Stars My Destination.
DG: Oh. I should check it out.
CD: It's kind of like Phillip K. Dick style, kind of half hallucination, half science fiction. I feel like I just need to read the science fiction cannon, so I'm just trying to do that lately. I don't know. I guess I'll just tell you my framework, which is I kind of try to apply the keto diet mindset to media consumption. I believe you can't change your reaction to media, but you can change the meta, what you do, how you spend your time.
My mental framework is books is protein and social media... I just got through talking about how everyone is so negative about technology and I'll just be negative now. I love social media, et cetera. Blah, blah, blah, all the good things, but I do find that too many hours a day is not good, so I think of that as sort of the sugar. And I have all these sort of rules around it where I only can check Twitter from the desktop various times of the day and you have to read books instead. I read almost no news. That's very deliberate. I have a rule about primary sources. I read news, but I don't like secondary sources. I like only primary sources and I try to be really disciplined about that. So if there's some news article about whatever, something, I try and find the data or read the original. If it's about a study, I'll go read the pdf or whatever.
DG: It's probably a right thing for the world to reward the original source with your eyeballs.
CD: It's just a nutrition thing or something. I think it's just a more nutritious way to go about it and you learn more and you develop a-
DG: Wait. We've gotten maybe the carbohydrate, we've gotten the protein, what's fat in this metaphor, which on the keto diet is your primary source of calories?
CD: Yeah, I don't know. What is it? I don't know. Maybe books are fat. It's a loose analogy.
DG: Yeah, I would want books as fat because fat actually satiates very slowly and books are kind of a slow thing. Maybe protein would be maybe email, something productive.
CD: Another thing I really like doing now is YouTube videos with academics and authors. I find that's really nice because when they're interviewed for example, like they're on a book tour or something or they're just being interviewed by somebody who's not an expert in the field, they're forced to get out of the jargon and speak... One is just sort of summarize things in a digestible way that people with busy schedules can follow, but two, I think it sort of forces them to not be kind of in the insider speak. I guess I kind of believe every field probably has a whole bunch of interesting things and if I could just somehow get those out without having to read a whole bunch of technical papers and things... Anyway. I find just reading or someone has a new history book and you go watch the YouTube book tour interviews and it's just a great way to kind of go and discover new things. Some of them you go read the book and some you don't, but that's been a useful hack.
DG: That's very interesting. It's like your best AGI summary you're going to get is from the author themselves.
CD: Kind of. Yeah. I think so.
DG: Ultimate GPD.
CD: How about you? Anything you recommend I should read?
DG: Things you should read is interesting. I think you would enjoy... Look, I mean all the sci-fi, and I'm just going to presume you've read all the classics. One thing I started doing recently, sorry for not being that exciting but it really works, I imagine you're a sucker for Sorkin material. I just started rewatching The West Wing. I know it's not that new, but it's worth rewatching every three to four years.
CD: Is it good?
DG: Extremely.
CD: It doesn't date?
DG: Sorkin dialogue does not date. It's so pleasurable to watch and it's so clearly above its pay grade. It's miles above everything else, especially for people that consume content at two to three to five X, it's organically faster. Anyway.
CD: I haven't seen that since back when it was on. I have to go check it out again.
DG: Definitely watch it. Yeah. I don't know. I find it incredible. It's tough to find good sci-fi. What else do I... I mean, I watch a lot of weird aviation stuff on YouTube. I'm not even a pilot, I just find it fascinating.
CD: That's interesting.
DG: I'll send some to you. Just F16s going haywire and doing emergency landings.
CD: Nice. You've got to create a playlist there.
DG: I actually do have a playlist. A bit embarrassing in reality. But look, hey, this was awesome. Actually on the top of content, I've told you in this past, in my opinion the world had Joe Rogan for a while. I do think it's gotten worse since he went over to Spotify. A huge opening in the market if you want to be the next Joe Rogan.
CD: Told you my plan. I don't know if I want to be Joe Rogan, but I did tell you my plan is someday I'd like to do more– have a video, YouTube show or something. Probably more tech focused.
DG: You have one viewer here.
CD: All right. That's how you start off.
DG: Exactly.
CD: A small cult and then... Start with a cult.
DG: Exactly. It's limited to 10 people, like Gmail invite.
CD: Awesome.
DG: Good stuff. Thank you so much for the time.
CD: Yeah, no. Thanks for having me.
DG: I hope to see you and catch up more soon.
Subscribe to the podcast here: https://pioneer-podcast.simplecast.com/.
For clips and highlights, visit the Pioneer Podcast playlist.