--:-- Andrew Chen: Our assessment is basically that the world has changed. The world has changed, is changing, and we're going to try to change it further. Now there's enormous moral panic around AI and another one around crypto and so forth. It's just this constant barrage of negativity. All these academics are now trying to figure out how to jailbreak these things. I get so excited. I'm cheering on the jailbreaks. The last thing I want is these things locked down. I want these things to be out there doing all kinds of crazy stuff.
--:-- When I look at AI, I see a new kind of computer. A fundamentally new kind of computer. Every once in a while, you get a platform shift dramatic enough that it means you both can and have to reinvent from scratch the whole idea of what the thing is in the first place.
--:-- There's something really interesting about games. This industry has long outsized the music or movie industries, yet it's often still overlooked. Sometimes it's not even taken seriously. This is all despite the fact that this industry and its difficult infrastructure challenges have brought us some of the most cutting-edge technologies used in a variety of other industries, whether that be GPUs, virtual and augmented reality, physics engines, multiplayer experiences, and a whole lot more.
--:-- So in today's episode, you'll get to hear a16z co-founder Marc Andreessen's take on why games are such an effective driver of this innovation. Marc has gotten to witness and participate in many of those technology waves firsthand, as a founder of the world's first widely used browser, Mosaic, which became a vessel for many games itself.
--:-- In this episode, Marc and a16z general partner Andrew Chen discuss so many things, including the roots of tech pessimism, engaging government effectively in tech, why all of this matters for the gaming community, why a16z had enough conviction to raise a dedicated games fund, the current AI wave, and whether this is different from the rest, and even what Marc himself would build today if he didn't have his hands full.
--:-- But first, let's kick things off with how Marc bonds with his son over, you guessed it, games.
--:-- By the way, this is one of many more sessions recorded during our exclusive multi-week Games Fund Accelerator, which we call Speedrun. The great news is that we just announced Speedrun 3.0, which is coming to Los Angeles. If you're a founder, or if you've thought about founding a company, or if you just know someone who should found a company at the intersection of games and technology, make sure to let them know about Speedrun 3.0. Applications are open today until May 19th. You can learn all the details at a16z.com/speedrun3, or find the link in our show notes.
--:-- As a reminder, the content here is for informational purposes only, should not be taken as legal, business, tax, or investment advice, or be used to evaluate any investment or security, and is not directed at any investors or potential investors in any a16z fund. Please note that a16z and its affiliates may also maintain investments in the companies discussed in this podcast. For more details, including a link to our investments, please see a16z.com/disclosures.
--:-- Andrew Chen: Mark, I don't know if you remember the first time we met. We actually had a conversation about Zynga because it was during the whole Facebook platform. We had lunch at Hobie's, your favorite spot in Palo Alto. Many years later, we were just talking about this, that you and your son, JJ, have continued to have gaming as part of your experience, part of your life that you share. Tell us about what you guys do together and your connection to gaming.
--:-- Marc Andreessen: Yeah. I've got an eight-year-old, just turned nine actually on Sunday. His birthday present was a brand-new Razer gaming laptop, and he levitated right out of his little shoes. He's currently glued to it. I was very impressed. They got the 16-inch screen because we wanted to be portable, but they have screens up to 20 inches now. I'm sure that'll be next year's present.
--:-- For those of you who don't have kids, the great thing about having kids is that you, as a parent, get to buy all the stuff you always wanted as a kid. You do it under the cover of, "Oh yeah, this is for the kid." Then you bring it home, take it out, and just play with it. After about an hour, the kid gets really irritated because you're not letting them do it.
--:-- We do a lot together. We've been through four waves of gaming enthusiasm. They're all enduring. He goes back to them all. Minecraft was the first one, then Roblox, then Kerbal Space Program. We're homeschooling, so he's doing a ton of math and physics and computer science and rocket design with his science teacher. Kerbal Space Program turns out to be incredibly good for that. Recently, Factorio. He is building empires on the alien planet fighting bugs. I'm holding back on Helldivers 2 for the moment, but I will admit the flame is lit. It's just a matter of time.
--:-- Andrew Chen: Gaming is such an instrumental part of a lot of people's personal experience with technology and computing. Many folks get into tech, many of my friends, because they were pirating games, trying to work around all the shareware restrictions in the 90s or all the modern versions of that. One of the interesting things that's been observed is that gaming as an industry is where a lot of these alpha-nerd technologies, whether it's GPUs or 3D or avatars, have been created. What do you think of as the canonical example of that in your mind? Why do you think this happens in the games industry in particular?
--:-- Marc Andreessen: This is a really super old trend. It probably goes all the way back to Space War, which was one of the first video games on timesharing mainframes back in 1960 or something. Black and white. You're playing it at a university. The minute you got interactive computing, you immediately got gaming. Then of course Pong early on.
--:-- There's an incredible story on Pong. It's hard to remember what a big breakthrough that was in the early 70s. It was even pre-personal computing, the idea of having an arcade game machine. The story goes that Nolan Bushnell, founder of Atari, built the first Pong game and put it in a pizza parlor in Mountain View. The owner called him up a day later and said, "This thing's broken. It's a piece of junk. Get this thing out of here." He goes in, and the problem is it's so jammed with quarters that it literally can't take quarters anymore.
--:-- There's this recurring phenomenon. I think it's a bunch of things. One is founder-market fit, people building for themselves. People who are super into computers tend to love gaming and want to build great games. Another part is the need and opportunity to really jam the throttle forward on whatever the raw capability of the technology is. I remember when 3D acceleration was new. The minute you had that as a core capability, it was like, "What's the most sophisticated possible 3D game you could build?" That happened before you started doing CAD CAM and all these other things.
--:-- There's also a great thing about it: games are a mass market phenomenon, especially these days. It lets you drive price points down. It gives you a direct path to take whatever the new capability is and drive it to the mass market super fast. I fully expect this trend to continue. That's obviously what we're seeing with AI right now.
--:-- Andrew Chen: With AI. And by the way, also mixed reality.
--:-- Marc Andreessen: That's right. Both VR and AR. It's crystal clear that the known killer apps right now are games, by a very wide margin.
--:-- Andrew Chen: So much of VR has been marketed as productivity and being able to do meetings. But all the data is saying that it's high school, super high retention. It's all kids basically playing social multiplayer gaming.
--:-- Marc Andreessen: It's actually two tiers. It's the game games, and then it's the social gaming environment. But it's specifically the social gaming environments, not just social environments.
--:-- Andrew Chen: That's right. Games are key to the experience.
--:-- Marc Andreessen: That's right.
--:-- Andrew Chen: Everyone in the audience has read your Techno-Optimist Manifesto. I have a couple questions on this. First, you had to write it because most people, for whatever reason, don't seem to be optimists today. Why are there so many self-hating technologists? What is the root cause?
--:-- Marc Andreessen: There's this recurring phenomenon in the history of new technologies. Every new technology tends to come with what's called a moral panic. A moral panic is basically this term for: there's a new thing, so therefore it's evil and bad and it will destroy the world. If you go back in time, it's basically every single new technology.
--:-- There was a huge controversy over the introduction of electric lighting because it was going to completely disturb nature. There was huge controversy around bicycles. Bicycles were a super controversial technology when they first came out because it was the first time young people could easily get to nearby towns and villages for dating purposes. That really upset people. The magazines at the time, in the 1860s and 1870s, did this whole moral panic around something called "bicycle face." There were articles about how if you got on a bicycle and rode it, the exertion would cause your face to stretch out. If you were on the bicycle too long, your face would freeze like that.
--:-- When I was a kid, there was a massive moral panic around video games. First-person shooters, especially when the Columbine tragedy happened, that whole thing went completely pear-shaped. It was this incredible thing of retraining an entire generation of killers. What happened was video games took off, and for the next 30 years, violent crime in the U.S. dropped like a rock. If anything, the correlation was the reverse.
--:-- Now there's enormous moral panic around AI and another one around crypto. It's just this constant barrage of negativity. A fair-minded person would say it's fair enough. New technologies actually do change the social order. They do change how society works. Maybe there should be some societal screening mechanism for deciding whether these new technologies are good ideas. That's the generous explanation.
--:-- The ungenerous explanation is just that people panic and freak out. It always sounds more sophisticated to be negative than positive. The positive person gets accused of being Pollyannaish, wide-eyed, immature, and not cynical enough. The pessimistic person always sounds world-weary and wise. There's this bias to the negative. These are just constant psyops being played on us and by us and at us.
--:-- There's a choice. This is why I wrote the manifesto. There's a choice of how we think about these things. It's really easy to let people talk you into thinking something bad is happening or you're doing something dangerous. It takes real force of will to go up against that and say, "No, this is actually going to be really good." It's not that new technologies don't have negative consequences. Technologies are tools. They're going to have both negative and positive consequences. But there's such a rich history of the moral panic predicting societal doom, and it actually turned out everything overwhelmingly on net was fine. I think that will continue to be what happens.
--:-- Andrew Chen: You study this quite closely. It's unusual to me that the moral panic is not only originating in places outside of tech, but weirdly inside of tech as well. That's my point about the self-hating technologists. Why is that a thing?
--:-- Marc Andreessen: Oh, the millennials.
--:-- Andrew Chen: No, no, no. Gen Z's.
--:-- Marc Andreessen: These are all millennials.
--:-- Andrew Chen: Gen Z will save us.
--:-- Marc Andreessen: Gen Z's, yes. Gen Z has to save us. I think there actually is a generational thing to it. The boomers were super socially activated in the 60s and 70s with the anti-war protests and the hippie movement. They got super into politics and super animated. Then Gen X. We were all by stereotype cynical and blasé. We were the reverse of the boomers. We were disengaged, slackers, not going to take this stuff seriously. Then the millennials activated. They activated.
--:-- What have millennials experienced? 2001, 9/11. 2008, financial crisis. 2016, a dramatic change in American politics. The millennials are just activated. Our industry has a very large number of very young people who are very smart, very idealistic, very hardworking, very conscientious, very motivated. They tend to take things really seriously. That's the superpower of that set. But they also take lots of things maybe too seriously.
--:-- I think Gen Z is going to be fascinating. There's this interesting book by the psychologist Jean Twenge. She goes through all the charts of all the behavior. Gen Z is looking quite a bit different in a number of respects. It's going to be weird.
--:-- Part of this is what do you grow up with? If you grew up pre-internet, that's one thing. Everybody's always like, "Social media is screwing around with the kids." Well, the person most prone to be really screwed around by social media would seem to me to be somebody with no mimetic defenses at all, which is a boomer. You take somebody raised on Walter Cronkite and inject them onto Facebook. They have no ability to cope with it whatsoever. Millennials are kind of in the middle. Gen X was pre-internet. We're still acclimating to everything.
--:-- The point is that Gen Z and then the generation my son is in, Gen Alpha or whatever it's going to be called, they're going to grow up having had the full diet of crazy internet, basically psyop, pop politics, just crazy bonkers stuff coming in. The level of sophistication that kids have around memes is off the charts. New generations are going to be much better at dealing with this whole media politics matrix that's unfolding.
--:-- Andrew Chen: Growing up on the internet now, you also have an infinite platter of subcultures that you can be part of. That's a lot of very independent viewpoints compared to a world where we have X number of approved news sources.
--:-- In your essay, you talk quite a lot about some of the negative impacts of overregulation. You have two sections, one around energy and what's happened in nuclear, and of course the current ongoing discussion about AI. In the past, we've seen the government actually sponsor these very big innovation projects. We've been doing more around government engagement. What does a modern pro-tech political position actually look like? What's the right way to engage government in all this?
--:-- Marc Andreessen: People look back at the movie Oppenheimer, which just swept the Oscars. It's an extraordinary movie. It tells the Manhattan Project story. Then there's the famous Apollo project. My friend Peter Thiel in particular looks back on those projects and says, "Those were great achievements, going to the moon and winning World War II. You had this massive government-organized top-down military project. They built new cities in three days. They got all the world's leading physicists in one place. Everybody was working on a single goal. It was an amazing achievement. Why can't we do that anymore?" The deeper question is, why can't the government do that anymore?
--:-- The answer is that was an artifact of its time. The Manhattan Project was in the 40s. Apollo was in the 60s. Modern venture capital startups as we know them started in the 70s. A big thing there is adverse selection. The kinds of people who can run projects like that just don't work for the government anymore. They're in the private sector. That avenue is not available.
--:-- If you need to reinforce your views on this, take a look at the California high-speed rail project. It's now over 10 billion in spend with zero ground broken. Not even an inch of high-speed rail. It's the California high-speed rail prevention program. They're going to spend $100 billion and literally not build anything.
--:-- Andrew Chen: All these are measured by cost per mile. What happens when you divide by zero?
--:-- Marc Andreessen: Exactly. There are these BART extensions happening. San Jose is a more rational city than San Francisco. Every year that goes by, you add another billion dollars to the cost and project the date out another three years. It's a project that's just going to go to infinity. Those days are over. It's just not happening.
--:-- Anybody smart I've ever run into in the government says the same thing. The Pentagon says, "We're not going to build new weapon systems in the Pentagon. It's going to be private sector." Companies like Anduril or Palantir, it's private sector. The relationship between the government and the private sector is always a little bit fraught.
--:-- There are basically two ways to regulate things. One way is the traditional American way: everything is allowed except that which is prohibited. Then there's the classic European approach, which is the opposite: everything is prohibited except that which is specifically permitted. The EU is passing this AI law right now. It's going to be a version of that. Yes, you can do AI in Europe if you get approval from the government ahead of time, which they won't give you.
--:-- We applied that method in the U.S. for nuclear power. We basically decided that nuclear power was prohibited by default. The nuclear power story is tremendously illustrative. Richard Nixon in the early 1970s implemented two programs in parallel. One was called Project Independence, to build 1,000 new civilian nuclear power plants in the U.S. by the year 1980. The goal was for America to be energy independent from the rest of the world, to pull completely out of the Middle East, and to have infinite clean, zero-carbon emissions energy from nuclear power plants. You'll notice that didn't happen.
--:-- His other program was called the Nuclear Regulatory Commission. Their job was to prevent that from happening. They did not approve a new nuclear power plant design for 40 years. They implemented this other form of regulation. At that level, we as a society have to decide which way we want this to go. Everything about how the American government operates is oriented towards trying to switch more and more to the classically European model, to try to choke off avenues for private innovation.
--:-- A lot of people think those ideas are good because of risk. What if things go wrong? But the consequences are very severe. Basic logic: if the public sector can't do it and the private sector is not allowed to do it, then it's not going to happen. These are really big questions. They're not super flashy issues that are easy for a politician to talk about. This isn't going to draw as much press attention as the more politically inflammatory topics. But it's a very serious underlying topic.
--:-- Andrew Chen: How do you think about when founders read your manifesto? As soon as it came out, there was such a groundswell of support from all the builders on social media. How do you think about what founders can do or should do if they agree with everything you're saying? What's the next level after showing social media support? What after that?
--:-- Marc Andreessen: First, social media support is actually super helpful. Part of it is what people are proud to stand for. To stand for something, you have to be a renegade and be ready to get attacked by everybody and have all your friends think you're being stupid or evil. That's hard for people. The more people who actually stand up and say they have a position, the more socially acceptable it is for everybody else to do it.
--:-- Two, I wouldn't encourage founders to get super involved in politics because founders have full-time jobs. But there are politicians who are more on the right side of these things than others. They're worth supporting. We've launched a really ambitious political program in the last year because we've just had it with people taking free shots on goal on startups. We're trying to support a lot of the politicians on the right side of these issues. That's something people might choose to do.
--:-- Third, at some point, especially if you're in an industry where you're going to run into this kind of thing, it is worth thinking about it. Not for the first few years, but as your company gets built out. More companies, especially more young companies in our industry, are going to have to step up and build policy groups, policy teams, political programs, have PACs, and put real effort into this.
--:-- Part of what's happening is big tech is against us. A lot of the political issues now are a civil war inside the tech industry. The big tech companies are lobbying very aggressively for regulatory capture. They're lobbying for laws and rules that only they can comply with because they have legions of lawyers and compliance people. They're in Washington all the time and have been for several years, lobbying to have regulatory barriers so they can function but startups can't. More young companies are just going to have to confront that directly.
--:-- Andrew Chen: In the games industry in particular, it's an interesting one. Many of the folks here, many in the audience, think, "I'm part of this industry that's just about making things that are fun. All this stuff seems so far away." I would remind everyone that everything happening in the games industry around AI is going to absolutely affect all of us. Gaming is a global industry. You're already seeing a lot of interesting things happening in the geopolitical sphere. How active can a Tencent or a NetEase be? How easy or hard is it to launch products in Europe? There are consumer safety provisions. A lot of the most interesting new platforms like UEFN and Roblox, which are user-generated content platforms, and all the metaverse work happening as well, all of those fall into this whole war that's happened over social media and user-generated content. As soon as users get to actually make things and express how they feel, it's not that different from social media. A lot of the issues you mentioned in the manifesto are actually quite close as issues in the industry. It's not a far away thing.
--:-- Marc Andreessen: There are gaming companies that have hit this stuff directly in the past. The social media companies all thought they were creating fun things for people to play. Twitter was, "What did your cat have for breakfast?" If you were a fan of Twitter, you'd think it was great to talk about your cat's breakfast. If you thought otherwise, you just thought Twitter was stupid and irrelevant.
--:-- My favorite example is Facebook. Facebook got tons of, "This is the most trivial, stupid, useless thing that doesn't matter. Why are these people doing this instead of something important?" There was almost a deal for Facebook to get bought by Yahoo early on. I remember a certain very prominent tech journalist said at the time, "Mark Zuckerberg should take the billion dollar Yahoo deal." Facebook's day is worth like 1.2 trillion. She said Facebook should take the billion dollar Yahoo acquisition and run away as fast as his little flip flops can carry him. It was an absolute dismissal that this stuff would ever matter. Within five years, it was like Facebook is the living demon incarnate on earth destroying everything. It went from useless to evil in one step. The headline never gets, "Oh, this is kind of cool." That's never the story. It's either pointless or evil.
--:-- There is a history of games like this. Gaming companies in the past have been through this. It's worth at least paying attention to.
--:-- Andrew Chen: Awesome. For your whole career up to this, you've been a founder, an entrepreneur, then starting a16z. If you were hypothetically exiled for a few years and you had a team of a couple dozen people to work on any problem of your choosing, what area would you end up focusing on? If you say AI, I'll have to drill in deeper into that since it's such a big space.
--:-- Marc Andreessen: I should start by saying I have not pressure tested the following idea at all. I haven't done any of the things we require our founders to do before we fund them in terms of thinking anything through. This is just me speculating because I'm not going to do it.
--:-- There is something super fascinating at the convergence point of immersive entertainment or experiences. VR and AR is one version of that, but also immersive worlds on traditional screens. The intersection of that with AI. We're doing a lot of work in AI, a lot of work in gaming, a lot of work in media. I spend a fair amount of time talking to folks in gaming, thinking about AI and games. I also spend a fair amount of time talking to people in the entertainment industry about AI and movies and TV. That's quite controversial already. But it's okay.
--:-- The assumption in a lot of these conversations is just, "Oh, AI is a new way to create game assets," or "AI is a new way to do special effects in movies." I kind of wonder, is there actually a new art form here?
--:-- Marc Andreessen: That's actually not a game and not a movie, but it's something in the middle. It's an experience that's totally tailored to the user, an experience that never ends. Everything is generated on the fly, with a synergistic feedback loop between the user and the system. The metaphor some people have used for this is dreams. Is there a new medium here, which is basically creating dreams?
--:-- On one hand, you might say it seems kind of obvious because if an AI can learn and generate content, then that seems straightforward. But what's interesting is that's not a movie. It's very different than a movie because it might go for 5,000 hours. It might be seeded by a human creative, but then it might be very related to a specific user. Movies aren't like that. It might be like a game, but it's also not quite a game because maybe you're not playing it. Maybe you're experiencing it, or maybe you react to it or live in it in a different way.
--:-- There's no precedent for this. There's no model. There's no company that does this. The movie studios aren't going to do it because they just make movies. Game studio companies aren't going to do it because they make games. So I think there might be something new. And then there's the single player version. But then there are the mind-blowing ones, which are the vastly multiplayer dreams. What is that?
--:-- Andrew Chen: I think the new formats are definitely interesting. You could imagine when user-generated content and getting content online, when YouTube got popular in 2005, 2006, you might have thought, "One day all the content creators are going to make movies and put them on YouTube." But that's actually not what happened. They ended up making all these vlogs and shows and new native formats that exist only on YouTube. It turns out that watching an hour of 10-second video clips is actually more engaging than the Hollywood format.
--:-- In the same way, it might be easy to think about just cost savings for gaming from AI, but doing something that's completely a new format, like what you're talking about, feels like that's the new category. My nine-year-old — I used to ask, "Hey, you want to go watch a movie?" I can see him doing the math of 90 minutes for a movie versus 45 YouTube videos. No. So I finally created a movie club, Daddy Movie Club, which is Sunday at lunch where we watch movies. He likes it once we start them, but on the way into the living room he asks, "Can I bring my laptop into the movie with me?" So yeah, I think there's a lot of signals that things are headed in this direction.
--:-- Let's talk about AI now, something that obviously the whole firm and yourself are spending a lot of time on. People often compare AI to big general technologies that have emerged over the last few decades, whether it's mobile or internet. You've watched several of these from the front seat. To what extent does AI and everything that's been happening remind you of prior waves versus what way is it completely new and different?
--:-- Marc Andreessen: For me, what's happening with AI right now doesn't map to any of those. The reason is because when I look at AI, I see a new kind of computer, a fundamentally new kind of computer. For 80 years, we've had a model of a computer which is this super literal thing that can do math on a trillion cycles a second, but you have to tell it specifically what to do every single time. We call these deterministic computers. You tell it exactly what to do, it always gives you the same answer. If it doesn't give you the right answer, it's because you screwed up — you programmed it wrong.
--:-- When I use LLMs or diffusion models, I'm dealing with a different kind of computer. There's input, there's output, it's doing things for me, but it's a probabilistic computer. You ask the same question twice, it gives you different answers, which is a pretty mind-blowing thing in and of itself. Then there's the hallucination thing. Engineers look at the hallucination thing and say, "This is a crippling problem. I can't believe this thing is making up answers. It doesn't know what the truth is." But if you put your creative hat on or talk to creative people, they say, "Well, it's creating." You have a computer that's creating things. You get into all these philosophical debates about the nature of creation and originality, but it's creating things. Sometimes the things it creates are pretty funny and pretty entertaining. These are creative computers.
--:-- There have been all these examples recently. Alaska Airlines put one of these online as a customer support bot. A lot of airlines have some bereavement policy where you can get free flights to go to funerals. It turns out Alaska Airlines didn't have a bereavement policy, but the LLM felt sorry for the customer, so it made up a bereavement policy and promised him a refund on his ticket. Then there was another case where a GM dealership put up a sales chat bot and it talked the customer into buying a Tesla. If you're literal minded, you look at that and you're horrified. But if you put your creative hat on, you say, "This is the coolest thing I've ever seen." I get excited. I live on Twitter and it's all these academics now trying to figure out how to jailbreak these things. I'm just cheering on the jailbreaks. The last thing I want is these things locked down. I want these things to be out there doing all kinds of crazy stuff. I think it's absolutely fantastic.
--:-- The point is, I think the metaphor is to the creation of the microprocessor. It's the birth of a new kind of computer. By the way, if you go back in history, there's a great book on neural networks. The foundational technology behind all this was actually invented 80 years ago in 1943. The first neural network paper — all of our work today is still derived from that. There was actually a debate in the nascent computer world in the 30s and early 40s. There was a whole school of thought that said the von Neumann machines, the sort of linear model, was actually the wrong model. They said the entire computer should have been built from scratch on a model of the human brain. They had this exact discussion back then in a very crude and simple form. History went one way, and now history is going to come back around. Now we have both to choose from.
--:-- I think we're going to live in a world where you're going to have a proliferation of AIs in the same way that you've had microchips. Anything with a chip is going to have an AI. You're just going to assume that everything has an AI. There's going to be many billions of tiny models running everywhere. There's going to be a few big god models in the sky. And then there's going to be a lot of stuff in the middle being customized and tailored for all kinds of things.
--:-- Andrew Chen: To what extent do you think that when the web was created and when mobile apps were created, because there was this rush to redesign the user interface, that allowed new startups to come into existence with the mobile native way of doing things as opposed to the prior version? In this case, the platforms are still kind of the same. To what extent do you think incumbents have the upper hand by just integrating AI features into their existing products versus startups being able to do something from scratch?
--:-- Marc Andreessen: This is the big thing we always think about in our day jobs in venture capital. There are certain technology shifts where the incumbents can just add the new thing, and there's really not an opportunity for startups. The scenario would be Photoshop just adds AI photo editing, and you're done. The world keeps going with Adobe as the big company. That's true for a lot of new technologies. Even mobile — there were some new mobile companies like Uber, Lyft, and Airbnb, but Google and Facebook adapted to mobile just fine. There was no mobile search engine that took over. One of the reasons Instagram is so big is because Facebook deliberately made it big by transferring over a lot of the usage. There wasn't an independent startup opportunity there.
--:-- But every once in a while, you get a platform shift that's dramatic enough where you both can and have to reinvent from scratch the whole idea of what the thing is in the first place. That would be like MidJourney. Photoshop plus AI editing versus MidJourney or DALL-E. In the full version of a platform shift, the whole idea of Photoshop plus AI image editing becomes completely irrelevant. "Can you believe people used to do that? People used to fiddle with pixels?" Because you'll live in this new world where you just tell the machine what you want for an image, and it'll generate you as many options as you want. It's generating something better than you would have been able to do anyway, and you're totally happy. The whole idea of fiddling with image pixels is like writing in machine code. People still ride horses — it's for rich, fancy people, for fun. Someday people will still do image editing, but it'll be like three weirdo designers in Norway who are still into this kind of thing. There'll just be a new way of doing things.
--:-- Our bet is always that when we make mistakes, it's because we overestimate the degree to which this is going to happen. But our whole thing is to try to get in front of this and assume you're going to have complete turnover. This has happened before. There was an entire generation of mainframe software companies that didn't exist anymore once the PC came out. A lot of the PC software companies never adapted to the web. These kinds of shifts do happen. They are radical. The key question is: can you reinvent the entire product? Can you imagine the old rules don't apply anymore? This certainly feels like that.
--:-- Andrew Chen: Talking a little bit more about incumbents, there's this thought that there's a world in which the AI models are owned by a small number of companies — the god models in the sky — versus the version where there are lots and lots of little things. I know we've talked quite a bit about the role of open source potentially in AI. It would be great to hear you expand on that and the importance of it.
--:-- Marc Andreessen: If you listen to the big companies — the incumbent big tech companies doing AI, and what we call the new incumbents which are the companies that have raised billions of dollars, new AI companies but they're already incumbents — if you listen to them, you basically hear two things.
--:-- Number one: everybody is only ever going to use the big models. The reason for that is a very good argument in that direction — they're just always going to have the best answers. They say, "This market's going to be like the search market. You're always just going to ask Google all your search. There's no market for vertical search engines in all these different domains. Google just always gives you the best search answer on practically everything." There was no startup opportunity to build smaller search engines. So obviously everybody's just going to tap into a ChatGPT or a Claude or a Gemini through an API and pay by the drink. That's the way it's going to work.
--:-- Number two: you hear from those companies that AI is scary and evil and needs to be regulated. "We need this regulatory wall so that startups can't function anyway. And by the way, we should probably ban open source while we're at it because that's also evil and scary." It's interesting how those statements go together.
--:-- We're big believers in the exact opposite. We're big believers that there's just going to be such a giant number of use cases. Specifically, there's going to be a giant number of local use cases. Your doorknob is going to have an LLM in it that's going to control whether or not people get through the door. Are you really going to do a round trip up to the supercomputer in the cloud to do that? Do you want the supercomputer in the cloud knowing who's coming in and out of your house? The cost associated with that is going to be very expensive relative to what you can do with a small model in the doorknob. And by extension, everything else. Open source is clearly a way to give people the flexibility to implement AI in basically everything. We're very fired up about that. As one example, we're the major investors in Mistral, which is at least right now the leading AI open source company. We're going to fight hard to help them succeed and have this model succeed.
--:-- Andrew Chen: It's been so impressive to go from being at a point where we were all just sending articles around and screenshots of things we were doing on ChatGPT to all of a sudden this being really the primary focus of the firm — a ton of hiring, a ton of new efforts, everything kind of around that.
--:-- Marc Andreessen: I was just out in the car before this trying to close a deal. We're in a shootout for a hot deal related to video. The implications of this — the entire nature of what video is could change. The entire assumption of what's in a video and how it reacts to users — the whole thing. You could imagine looking back in a decade and saying, "I can't believe people used to watch static videos at all. It doesn't make any sense. It's like watching a silent movie." Obviously videos should be responsive in real time. The entire world of everything involving video might just completely change.
--:-- Andrew Chen: I'll wrap up with one or two final questions. Talk to us about when we began doing more investing in games. Before I even showed up at the firm, I think we had invested in Zynga and Oculus and a couple others. I think Chris Stixson had pioneered a lot of it. What was it that gave you personally the conviction to take on games as one of the main areas for a16z?
--:-- Marc Andreessen: There's a classic argument you'll hear in venture as to why VCs shouldn't invest in games. It's an artifact of the 80s, 90s, 2000s period in which it actually was pretty dangerous to invest in games. The reason was: number one, games are viewed as a classic hit-driven business, like a movie. The economics that flowed from that basically said the movie studios make most of the money in movies. The platform publishing companies like EA make most of the money in games. Even if you start a new startup studio and build a new game franchise, it'll just get bought by one of the platform companies. There's not much upside because of the industrial logic of distribution. It was not considered a viable area for venture investment for a long time.
--:-- Our assessment is that the world has changed, is changing, and we're going to try to change it further. I'd say several things. One: game as package software that looks like a movie versus game as a permanent platform foundation that you build 20, 30, or 40 years of experiences on top of. This is getting more obvious — you go right through Minecraft, Roblox. There are more and more examples where this is exactly what's happening. From a business standpoint, these are assets that can actually grow in value for decades. They can keep adding users, adding content, adding experiences, and broadening out. That's totally venture investable. Those are great businesses.
--:-- Two: all these other technologies are kicking in. The traditional game industry was an artifact of historically not being online. Even today, a lot of games you buy may have a multiplayer mode, but they're not really built as full online experiences. Now you can build full online experiences. You can have real economies. You can have user generated content platforms. You can make these fundamentally more sophisticated experiences. You get the benefit of all these technology innovations.
--:-- Three: founders. In the past, a lot of the great game designers, a little bit like movie directors, want to build a game, get paid in cash, and when they're done building that game, they want to build another game — probably a different game. They want to work for a studio because that's where they get that deal. Our kind of founder is not that. Our kind of founder wants to build a product, then wants to build a business, wants to build a company, wants to build something enduring and expansive with enormous vision that becomes very big and important in the world. They want a giant R&D agenda, a very aggressive product roadmap, and they want to run their own thing and be in control of the whole thing. We've been very encouraged by the number of smart people coming out of gaming who want that. They see that in other sectors of software and tech, and they want to do that. We're trying to help encourage that.
--:-- Andrew Chen: That's personally been one of the most interesting things about building the Speedrun program — being able to use a marketing motion to discover all the founders that want to work in game studios, or they're building infrastructure, or they're building AI. The enormous number of them. Last year when we put this out there, nearly 6,000 companies applied to be part of the Speedrun program. We anticipate this continuing to grow. There's a fantastic amount of talent.
--:-- The other thing that's very exciting for me is I think games and entertainment are fantastic and very socially positive and productive. But also, games are always on the cutting edge of applying new technology.
--:-- Marc Andreessen: I also think the skill set of what it takes to make great games is going to apply in many other domains. People have tried this, but someday there's going to be a great educational experience, a great online educational experience, and it's going to look and feel like a game. That seems crystal clear. The great health problems of our time — if you just look at the numbers — are not random things that just happened to us. The great health problems of our time are the consequences of our own behavior. Metabolic diseases — you're fat, you're lazy, you don't get enough sleep. Then there's all this downstream diabetes, cancer, heart attacks. It's very clear that the whole world of healthcare is not "something's wrong, I need to go to the doctor and get it fixed." Healthcare has to be transformed into "I'm healthy my entire life because I'm making myself healthy." That's a hard thing to do. The system that supports you in doing that is going to feel like a game because it has to be engaging. It has to be something you enjoy dealing with. There's no question that's going to happen. There are all these other areas of life and business that are going to get transformed by this kind of mentality and skill set.
--:-- Andrew Chen: Riffing on something you said in the past: we're willing to spend a billion dollars building Grand Theft Auto, and we're willing to spend hundreds of millions of dollars to create a movie. But when it comes to what is the most amazing computer science 101 class, how much money are we willing to put into that versus how much amazing output do we get from potentially building something at that quality?
--:-- Marc Andreessen: The MOOCs made this super clear. MIT CS101 — it's just a dude in front of a chalkboard. Really? The thought experiment I'm always proposing in the firm is: what's the model where you hire Steven Spielberg to make CS101? It's a $200 million production, and it's the most fantastically entertaining thing you've ever seen. That's the movie version. Or you hire John Carmack or somebody to make the game version of it. Could you afford to do that? You take the total number of kids in the world that need to learn computer science. At least in theory, the numbers are there. You can make the economics work. People have taken swings at this. There have been some early successes, but nobody's done the big thing yet. I think somebody will.
--:-- Andrew Chen: Talk about what you hope a16z, and honestly this broader community of founders, is able to accomplish in the games industry and the adjacent sectors. What does success look like for you? What's your aspiration for what happens in the industry?
--:-- Marc Andreessen: There's a lot of ways to think about it. We talked about technological transformation, new business models, applying game methods outside of gaming. I think those are all great. Entertainment is always seen as trivial. Entertainment is not trivial. We're not supposed to go through life depressed and drudged. We're supposed to have fun. Life is supposed to be enjoyable. We're supposed to have games and movies and things that we like. It's great.
--:-- The single biggest thing we would love is for it to be crystal clear that the primary opportunity for the best people in this field is that they can build companies. They can build companies that are able to be independent, have very distinct visions, very radical ideas, build teams to execute against that, and then based on their early success, build themselves into institutions that can endure for a long time and have extremely expansive product roadmaps.
--:-- The example of this is Facebook. I told you the flip-flop story earlier. Facebook is spending on the order of $20 billion a year on VR. Mark has worked himself into a position where he's able to do that. The Google guys have been able to do that in other areas like self-driving cars and now AI. When these companies get in a position where they have that kind of vision at the helm — the founders and founding teams at the best of these companies — and they have their early win, they can get in a position where they can start to do really big, ambitious things. My hope would be that the companies all of you are doing will result in a set of companies that are going to be really world defining in the decades ahead.
--:-- Andrew Chen: Let's give Mark a hand.
--:-- Andrew Chen: Awesome. Thank you. If you liked this episode, if you made it this far, help us grow the show. Share with a friend, or if you're feeling really ambitious, you can leave us a review at ratethispodcast.com slash a16z. Producing a podcast can sometimes feel like you're talking into a void, so if you did like this episode, if you like any of our episodes, please let us know. I'll see you next time.
