Ep 65: Marc Andreessen and the Case for AI Optimism

20:44That's scary. Right. And then they ought to have an override. The government ought to have a kill switch. And if the government sees what you're doing on your computer and then they don't like it, they should be able to turn your computer off. And then the more extreme form of this is what was in Time Magazine, which was a call for airstrikes on rogue AI data centers. And by the way, in complete contravention of international law, just like, yeah, screw it, bomb them. And then some people in this movement are saying we should be willing to risk nuclear war in order to prevent household regimes from using AI.

21:13And so at this point, like you're walking down a path basically saying, I want a global totalitarian super state omnipresent, right? Authoritarian dictatorship that's monitoring every computer and every person and every chip and every piece of software. And then I want to destroy anybody who deviates from that. And that's how I'm going to save things. Right. And so that's what's gotten just like for me completely surreal about this whole discussion. And I think this can only be happening because we're in a moral panic right now. but like rationality has left the room for the time being. And I think, I think we should bring it back.

21:46No, I think, and that's totally fair. I mean, the, the, the only excuse for that, for me, if we can talk about the Fermi paradox, the paradox of the fact that there seem to be trillions of worlds to support life and then we're not seeing life. And one of the possibilities is that as we get to this point and when every civilization figures out how to create thinking machines, it destroys itself somehow with the power of those machines, which is an argument they're making, which I realized maybe somewhat ridiculous, But if people believe that, of course, it would justify almost any of the things they're saying if they believe that.

22:12But I guess they're making a really big leap. Yeah, it's the precautionary principle on steroids. And it leads to stagnation and violence and madness and death. It's an incredibly anti-human philosophy. You know, it's basically anti-everything that's made Western civilization great in terms of like what you do as a consequence of that. And then look, I would make one final claim against that kind of position. And I do hear that. And it's a completely unscientific position. And by definition, it's a completely unscientific position because there is no way at all to prove or disprove the theory.

22:41Right. And so it is the worst kind of ivory tower thought experiment. And, you know, the history, if we've learned one thing from the last hundred years, it's that ivory tower thought experiments should not be the way that we determine the shape of society. So I think it's clearly me. Yeah. No, it's the kind of thing we talk about in our dorm room. We're drunk with friends in the fraternity at Stanford, which I think is very reasonable, but maybe not reasonable to use to create international policy. So it's a little bit scary if you do that. William Buckley famously said he'd rather be governed by the first 200 names in the Boston phone book than the faculty of Harvard.

23:14This is why. This is why. Because this is the kind of thing that professors do. And professors are fun and nice. And I'm glad they exist. And it's great. And I like reading the books. But they should not be in charge. So I was with some friends recently, my friend who ran Google for a long time. And he was saying how basically deep linking was illegal in the UK for a long time. and Google had to fix that to go there, which I hadn't realized. And all my friends in Europe right now are telling me that this new regulation on AI that's coming out will basically completely prevent anyone from using AI at all in Europe and just put them way behind.

23:44And everyone hates Brexit in the UK, where I am right now. We're not everyone, but a lot of people do. In London, they hate it overall. But one of the bright sides may be that they escape Europe's terrible regulatory regime on AI. Maybe it frees them from it ironically and helps out. What are the actual biggest dangers for countries with AI? Is Europe going to be left behind? And China in that regard as well. Yeah. So, I mean, look, Europe seems, I just take them at their word. They seem determined. They seem enthusiastic about being left behind. They seem to really badly want that. You know, look, the data, here's one where the data is clear, which is, you know, 25 years ago, if you did a list of the 50 biggest tech companies in the world, like a pretty significant number of them were European.

24:23It's like, I don't know, probably 15 of the top 50 or something. You know, today you're, you're basically down to Spotify, which is great. I love Spotify, but like, it's not, you know, that's not, That's not what the European tech sector used to be. And so they have sort of deliberately disarmed and turned away from technology and progress and capitalism in really profound ways for decades now. And yeah, look, the EU is designed to do this. That's kind of the whole point of the structure is to impose these kinds of controls and stop these kinds of things. And yeah, they're very vocal about it.

24:55They very much want it. I think they would be very happy if AI didn't happen inside the EU. I think they'd be very happy if social media didn't exist. I think they'd be quite happy to shut up the internet. And so, I think the citizenry of the EU have an even more kind of pointed version of the question that our own citizens have. I have a lot of friends who are optimistic about this or that, or the French are going to save it, or whatever. And it's like, I hope that happens. But the overall kind of vector there is in a very clear direction. And then look, China is sort of the dog that didn't bark in a lot of these conversations.

25:25So, what I find is when I talk to people in Washington about AI or tech policy, generally, there's two completely different conversations I can have with the same people, the same person, and on the same day. So, conversation number one is a conversation about tech in the US, right? And so, tech in the US, US government, US tech companies, US tech policy, US society, all these issues. And that's a discussion that's characterized by just enormous antipathy and loathing from Washington towards American tech and towards increasingly just like flat-out capitalism, market economy, innovation kinds of things.

25:58I can have a completely different conversation with that same person where China's on the table. And the minute the China topic's on the table, it's a completely different conversation, which is how does the American public sector and private sector team up to make sure that we win versus China in this sort of new Cold War that's underway? And so the cynical optimist in me would say, as long as China's a big enough threat, it will be sufficient to kind of glue us together to be able to actually make some rational decisions as a society. Maybe the presence of an external enemy is actually helpful to kind of galvanize, kind of your own side to organize and be smarter.

26:30The cynical pessimist would say China's not enough of a threat. And boy, it would maybe be actually nice if they were even a little bit more of a threat. Many other people would find this entire line of conversation horrifying. I would just say, I don't actually know how to have a productive conversation with people in DC right now if they don't have China in their head, because otherwise the negative emotion, fear, whatever, panic just turns internal. And I think it's very dangerous. I have seen some of this too. I can get together with Democrats, Republicans, and work on immigration and tech policy because we have to compete against China.

27:07And suddenly, they're willing to work with me on it and being coherent, which I love. So there is something to that. I mean, Xi Jinping does really seem to truly hate the US, the Maoists. I understand the Maoist generation he was in, the young people without poems about seeding your heart with the hate and distrust of America. And he seems like he really does hate us. But maybe as long as he doesn't NUCAS that could be useful. So it's a very interesting framework. I want to jump to a couple of things on the internet, Mark. Is the era of search over? What's the future look like when are these changes to pretend?

27:36Some people are saying Google's in trouble from this. Is that true? Yeah, it's a little funny because there are different layers of it. So how these AIs are trained, like how ChatGPT is trained, it's actually not that dissimilar to how Google's trained in the sense of step one is hoover up all the data. So basically scrape the internet. There's literally this thing called the common crawl. It's basically the open source version of the Google index. And just basically just like go get all the text on the internet and then train the AI in the same way that Google would go get all the text on the internet and train the search engine.

28:04So a little bit of like same data set. And then one is the chat UI. And then the other is the 10 blue links UI. Google has been pushing for a long time to get away from the 10 blue links anyway. Right. So they've been pushing to what they call the one box. And so more and more, Or if you do Google queries, it will give you the 10 blue links. But before it does that, it will either just answer the question for you, or it will give you some other UI to get to the thing that you want. And then the links are pushed further down the page. So I would say Google's already halfway there. And now, of course, they're starting to thread in full LLMs.

28:39And they have a pilot program now where increasingly the search engine is starting to give LLM responses to things. So there's probably a lot of shades of gray in here. Um, you know, look, having said that, you know, does the defining paradigm of the era go from, Oh, here's a bunch of links. Good luck to here's the answer. Yes. On that. I think it's, it's very, very clear that that's, what's going to happen. Um, and then, uh, my little wisecrack I give on this is, you know, uh, we still, um, you know, ride horses, um, you know, except we do it as like a hobby, um, you know, it's sort of an exception.

29:08Right. Um, and so one of the things that an LLM can do for you is give you the 10 blue links. Right. And so it just may be that these older paradigms just kind of get incorporated into the AI. And AI can also do that for you if you want it, but most people won't. And very interesting. You actually think it will? You think in the next five years, less people will search then for things in the traditional way? I think it's just like ask a question. So a couple of things. So one is just like, is ask a question, get an answer better than try to navigate the internet and search on links and find the answer?

29:35Probably yes, for most things, most people. Look, there's another question, though, which is, is the actual model going to be ask a question, get an answer, right? Is that the optimal way to actually interact with these things? And that's the default way to do it. And that's the way people are doing it out of the gate. And so maybe that sticks. But maybe not. Because one of the things these things can do, they're actually really good at asking questions. So there's this idea of prompt engineering, which is how to build optimal prompts to prompt at LLM. And one of the funny things is, actually, the LLMs are very good at writing prompts.

30:08I'm actually very scared I'm going to have a Neuralink in my head. And it's going to be showing me prompts I may want to ask. Right? Yeah, exactly. That's what it's going to get to. Yeah. Well, you probably know this. The Bing version of GPT-4 actually does this, right? Not in the head part, but on the screen, it actually suggests prompts for you. And so, taking a step further, maybe the prompts don't even exist because maybe what happens is the AI just knows you so well that it just tells you what you need to know before you even ask. Right? And it basically is just more like, oh, it's just like your assistant.

30:37It's like an exceptionally well-trained, diligent assistant. You keep saying in your essay in here, assistant and partner. Sorry, I just want to push on this one more time because I am an optimist too. But there's something very scary to me. If you imagine that it gets smarter than you and it knows what you like, knows what you want. Doesn't it start to just give you direction and you say, oh, yeah, I like that direction. But doesn't it start to think for you and do things for you and lead you directions? Doesn't it start to become the script writer of everything you want? Isn't that scary a little bit to you?

31:03Is there some steel man of some scary thing here that you're at least a little bit worried about? So the implicit, and this is a big line of argument, right? And as people argue, this actually even the stronger version is this smart thing will be able to talk the, you know, the, the, the dumber human into whatever, whatever it wants. Right. So, you know, it'll be very good at persuasion. So this, this is one of the kind of doomer kind of dystopian arguments. My, my response to that is look at the world around you today. Does it seem like the smart people are in charge? Only in very narrow areas, right?

31:34Like, I don't know about you, but the world I see is a lot of high IQ people working for people who are not that high IQ. It's like, what are the PhDs running the society? And maybe to some extent, they have more influence on policy than they should. But for the most part, they work for people who are like 40 IQ points lower than they are. And again, you go right back to politics, which it's like, okay, the whole nature of politics, ask any political candidate, number one, are you smarter than your voters? And of course, the true answer is yes. And most of them will say that. And then second question is, can you get your voters to do what you want them to do.

32:05And the answer is no. So the whole thing where smart just wins versus dumb, put it this way, people are stubborn. You get one other aspect of the paradox, which is, okay, now it's going to be ultra smart. And then therefore it's going to be evil. And it's like, well, wait a minute. Why would something that smart be evil? You're not evil. Exactly. Yeah. I mean, it's like, look, once you're at super genius level, I don't know, maybe there's a handful of like real life, life's luthers in the world, but like once you're a super genius level, like you have better things to do than like go around and try to kill a bunch of people.

32:38Right. Like you were like plan, like your world conquest or whatever, like you're exploring, you're creating, you're designing, like you're writing, you're whatever you're doing all the things that you get to do. Cause you're super smart. And so you're fixing, you're fixing stuff. You're building and fixing stuff. You're building things. I want to hear just briefly. Everyone's obsessed with inequality, which I think is a really stupid thing to be obsessed with. Cause every time I create a company, I widen inequality, but I think I'm doing good things. but what does AI mean for widening or narrowing inequality?

33:04How do you think about this? Because it's a big topic. Yeah. So there's a couple of fallacies in the inequality discussion that I think are really important to understand. So first is the way that capitalists maximize their wealth is not by hoarding a new technology. It's by selling it to everybody. And that's been the case for everything from electricity to automobiles to you name it, the internet, computers, the whole thing, all of it. And so the maximum company, The maximum size tech company is always the one that gets to all 8 billion possible customers on earth or as close to that as it possibly can.

33:37The premium high end of the market is just never a very big market. It always has to be the mass market. So the basic capitalist incentive, self-interest to capitalist incentive of the capitalist with a new and special technology is to drive that thing to mass market adoption as fast as possible. By the way, that's what's already happening with AI. And if you think about this, this is actually fairly remarkable, which is you or I today can sign up for full GPT-4 or full mid-journey for$20 a month. Or we don't even need to do that because we can go to Microsoft Bing, which is free, or Google Bard, which is free, which is almost as good.

34:10Bing basically is GPT-4, so it basically is the same thing. And those are actually free. And so you've got the most advanced AI technology that took a huge amount of money and resource to be able to build. That's exactly the kind of thing that where the, you know, the sort of conspiracy theory would go that it's going to get hoarded. And already these companies are tripping over themselves to try to get as many people as possible to use it. And again, purely out of their own self-interest. And so one is just like, look, technology is inherently a democratizing force, including economically. So that's number one.

34:36And then number two is the way that wages get set in the economy is not some, you know, some sort of paranoid us versus them kind of thing between capital labor. Like Marx was wrong about that. The way wages get set in the economy is as a function of productivity of the individual worker. of what's called marginal productivity. And the reason why wages rise is because workers get more productive. How do workers get more productive? Because they're able to use technology. And so if you want the standard conventional worker, whatever, blue-collar laborer, white-collar worker, whatever, to rise up the income ladder, the A number one thing you want to do is arm them up with as much technology as possible to become more productive.

35:12And again, in that way, technology is fighting actually for more opportunity, not for less. That's very well said. You're making them more productive. You're making them so that they're worth more. They're going to demand a higher wage and create more. So Mark, we're speaking at the forbidding courses here. I want to shift a little bit to ask you about how Silicon Valley, which has really been one of the greatest engines of innovation in human history, how did it become so illiberal, woke, we call it? And how do we fix it? How do we break out of that? Yeah, so it's a real puzzle. I think there's a very straightforward explanation for the why it happened, which is basically it has to do with one of the fundamental personality traits of human beings called openness, trait openness, which is one of the big five kind of personality traits.

35:55Openness, which is by definition, people who are super into technology and entrepreneurship are, we're highly open people. We love new ideas and we love doing new things. And then there's just on the political spectrum, it just openness is a trait of the left. And then on the other side, on the right is conscientiousness. structure and order, which is a trade on the right. And by the way, this is not just tech. This is true in the arts. Anytime you have a clustering of people who are high in openness, inevitably, they're on the political left and usually very far to the political left. So I think we're just a classic example of what happens when very open people cluster together.

36:34You read these, again, these hysterical things in the newspaper or whatever, that Silicon Valley is a hotbed of libertarianism or incipient, whatever, crypto fascism, whatever. And it's just not true. It's like the Valley is like 99 % Democrat. And you can see that in the political donations and the voting patterns and everything else. And so it's just, it's a classic left-wing democratic environment. Look, I think people should have the politics that they want to have if they're voting for the people they like and they're getting the results that they want. But I didn't just say like left or liberal.

37:05I said illiberal too. Like, Why is it so illiberal? Because that wasn't the case in the 90s, right? Or was it? Well, but, you know, this is the paradox of tolerance, right? Like to be maximally tolerant, we have to be intolerant of intolerance, right? And so it is absolutely morally required. If you're going to maximize tolerance above everything else, you have to be intolerant towards the intolerant. And that's how liberalism becomes illiberalism. I quibble with the liberalism framing of it because I actually don't think, something you and I might disagree on, I don't think illiberalism, I don't think what's happening now is a betrayal of liberalism.

37:43I think it's just the full realization of liberalism. Like, I think it's what happens when you take liberalism to the extreme, which is you declare war on the intolerant, right? And then the fact that you yourself have become intolerant, like somehow you never quite like rub your head around that, but off and away you go. Interesting. I have some more, but I want to open it up. Lauren, I don't know if we haven't heard from you guys at all yet. Did you want to open up to some student questions? Yeah, that would be great. Guys, if you have questions, just come up to this podium here and just start lining up.

38:11And we'll take them one by one. Hi, my name is Shlomo Clapper. I'm one of the founders of WeBio. We are building AI, NERF symbolic AI that can automate a lot of the regulatory filings for the FDA. And I'm personally a huge accelerationist. Earlier today, we heard from Peter Boghossian, and he has something called street epistemology, where you check your own epistemic processes and thoughts. And one question that he had was, what evidence do you need in order to change your opinion? So, Mark, you complain in your article that the AI warriors don't have a falsifiable hypothesis, but I didn't see that you had one in the article.

38:51So my question for you is, what would convince you that AI can be dangerous? Yeah. So of course it's a question of like which form of danger. Right. Um, so, but let's just like take the, take the existential version of it. I mean, yeah. I mean, like how about a rogue AI, right? Isn't really the obvious answer to the question. Um, and so, um, you know, the minute one of these things tries to take out, you know, um, you know, one of these, uh, AI risk people, you know, develops a, you know, custom unique pathogen and, you know, somebody drops dead in the street and it turns out a rogue AI did it like at that point, like I'm going to reconsider my priors.

39:23But I would like to see that. It's like the Fermi Paradox thing. I would like to see that happen once, right? Like the Fermi Paradox thing. I would like to see one alien species that died because their AI's turned on them before we conclude that that explains why we haven't found intelligent life in the universe. Like just one. Okay. Thank you. By the way, the good news, if you guys know about, there's this principle in AI called Rokos Basilisk. And the idea is that because the AIs are getting trained on the data set of everything that everybody has ever said and done before they, before they emerge, um, that it's in your self-interest, even if you believe that AI is going to be rogue killer, you know, dangerous, you still want to be on record as saying that it's going to be very positive and beneficial, right?

40:01Because you want the new AI to view you as a friend and not an enemy. Um, and so let me just say like, if it turns out the evil AI is like really, really evil, I I'm, I'm, I'm fine. I think teaching us about that was, uh, just made us all complicit, but Exactly. That's right. Thanks, Mark and Joe. My question is, I think a lot of technologists are into, you know, building things and they don't necessarily like to venture too much into the political and social realm. And I think a somewhat rational assumption might be that if you increase consumer surplus a lot, then people will like what you've given them.

40:40And, you know, voters will sort of push for regulation that opens that up for people. But of course, people like Elon and Jeff Bezos and other people that have increased consumer surplus haven't necessarily been treated with a ton of appreciation from some people. And I wonder if you think it's naive for sort of AI to take a similar let's just build things approach. And if they basically if tech needs to rethink its strategy and get more politically involved from the start. Yeah. So there's an old joke in the Soviet Union, which is you may not be interested in communism, but communism is interested in you.

41:15And so, the challenge... I mean, look, most technologists are not political. Most of the ones I've ever met and worked with are not political. They just want to build new things and have people use them and like them. And so, I think by default, that's what most of us want to do. That's certainly what I would prefer. But look, that only is a path if they're not interested in you. And when they, being Washington or regulators or politicians or whatever, decide they're interested in you, then at that point, your options are limited. You can either ignore it and hope that they don't destroy you, or you can confront it and try to deal with it.

41:48Look, the tech industry, I think, sailed by for 50 or 60 years, basically up until probably 2010, just not having to really pay attention to politics. One interpretation of what's happened is where the dog that caught the bus, it turned out that we were right. It turns out tech is a big deal. It turns out it's not proceeding in the world. It turns out the internet actually is a big thing. AI is going to be a big thing. And so, a cynic might say, okay, we created this problem by building these technologies that actually have a big impact on the world, a big impact on society, and a big impact on the economy.

42:22And so, of course, the broader world is going to perk up and want to express themselves on this through their elected representatives. And so, I think that's where we're at. I think the world has just changed in that way. It's part and parcel with just being more successful as an industry and being more influential as an industry. And I think we have to confront it. Having said that, the other thing that is just very complex that makes this hard to figure out the right thing to do is it's not like tech is unified. And so a lot of the people I am opposing, at least rhetorically, on these AI issues are actually people in the tech industry who are showing up in Washington asking for various kinds of things that I think would be horrifically destructive.

43:00And so even within tech, we're not organized on this, nor do I think we will be. And so in practice, it's going to be this multidimensional thing. And it's just going to be complicated and messy. And I think that's the world we're going to live in. Yeah, that's very well said, Mark. But I'll add something to that. John Adams and Thomas Jefferson, two of our founders, had a series of letters. They went back and forth a lot after they were both president. And they talked about their theory of who should run society. And both of them agreed that there's a sort of a natural aristocracy. That's kind of the view of our founders.

43:30I think it's probably the view of everyone, but based on how they define it, where there's going to be some people who are more talented than others. And the whole idea was that society should have the most talented people involved in government. Now, if you're an anti-federalist, which Everson was, you didn't think people should be running things. You think people should be making sure things don't break themselves. But you still want the most talented people there, like whether or not you're a federalist or an anti-federalist. And I think there's something very scary in our society where this idea of the natural aristocracy, being involved, having the debates, being part of running things has really fallen out of favor.

43:59And tech is amazing in the sense that it's attracted a lot of the most talented people in the world. And yet it's also like has a culture where basically none of them have been running things for 30 or 40 years. So we now have an area that clearly has some of the very most talented leaders, entrepreneurs, thinkers, innovators. And yet the vast majority of them have totally avoided running things. And so even though a lot of us disagree, I think it's very important, we do get involved and we do run things. My view, I think our founders would believe that as well. Because otherwise, things are going to be just very terribly done.

44:27As we're saying, it's comically bad right now, the people who are in charge. Hey, thank you, Mark. Thank you, Joe. I'm Jason. I'm the co-founder of Story Protocol, which happens to be an A16Z portfolio company. So just wanted to reassure you, Mark, that I've been doing work throughout the week and not just talking about philosophy. But it is good to have a chance to think about these things in the startup world as well. So my question is about the intersection of AI and political structures. I think AI is easily on the level of something like the Gutenberg printing press, which of course had a huge and profound impact on the way that people thought about themselves and also the political possibilities in Europe and abroad.

45:03So I'm just curious, as AI turns us almost into gods over the years, do you see that having an inherent tendency to push our political structures, whether it's liberal democracy at home or whether it's, you know, Chinese sort of form of government towards one or the other, or a completely new form of governance? Or is it more dependent on who gets there first? You know, if we get there first, a certain form of political structure will be crystallized. Or if China gets there first, it'll be different. Yeah. So I look, I think these are very big questions. And I think I don't, it's going to be hard to predict the out year kind of versions of these.

45:35But I think there's a fundamental tension that's kind of at the core of what you're asking that I'll focus on, which is an old tension, by the way. It's an old tension in the use of prior technologies, which is, is technology to be used by the state in some sense in opposition to the people, right? So this is sort of the Chinese model is to be used for basically the state to be able to surveil, monitor, control, right, the people, right, you know, including, you know, in a visible form of things like security cameras, but even, you know, also things like social credit scores and so forth, you know, or is technology a tool in the hands of the the individual to be able to express, to be able to have an increased level of power influence, protection from the state.

46:20And so then this is like the idea, for example, of the mimeograph machine in one's home. So the Soviet Union famously, they outlawed mimeograph machines, basically pre-photocopiers. And if you were found with a photocopier in your house, you were shot, because it was such a threat, the idea that a normal citizen would have the ability to express themselves in a way that people outside their immediate earshot would be able to hear their ideas. In the US, we don't worry about that. We never did. Everybody who has a photocopier at home or an internet blogging platform or whatever is free to do that.

46:53And we would be incredibly horrified if the government were to tell us that we can't do that. And so it's this basically tension, technology leveraging state power, technology leveraging individual power. Look, for sure, AI is going to create a whole new set of dreams of state power. And like I said, the Chinese have one version of that. But on the far left, even in the US, you have this idea. There's a book called Fully Automated Luxury Communism, which is this idea that finally with AI and big data and computers, we finally have the level of sophistication necessary through technology to be able to centrally plan and organize the economy and create a communist utopia on Earth.

47:32And so, you know, AI is going to bring those ideas back by people who are into central planning. And by the way, everybody who have, you know, there's not everybody, a certain set of people who advocate for UBI, you know, that is basically what they're advocating for. That's what that means. You know, look, I think there's a great push and a pull there. I would love to tell you that in the U.S. we haven't all figured out. I think this is all, like I said, I think this is all up for grabs. But I think that would be a major tension point. And depending on where that tilts, you can imagine this being the greatest liberation force for human freedom and human empowerment in all of history.

48:04Or you can imagine this as a tool of oppressive state control that'll be very hard to get out from under once it's in place. We have a quick model just really quickly for people. I think the answer is neither. I think the greatest constitutional scholar, one of the greatest ones alive today is Philip Bobbitt, who wrote The Shield of Achilles. He traces the last thousand years of history, and he shows the forms of government mostly changed based on what's dominating warfare, which makes sense. So if you have cannons, they can knock down walls. You can't have little city-states anymore because jerks go around building empires, because they could do that.

48:31Or if Napoleon has to give everyone a rifle, then the Prussians can't have their aristocracy anymore because they need to give everyone a rifle too. So I actually think that the changing of warfare and defense is probably the dominant question for how government has changed over the last 1 ,000 years. And I think what this does is it empowers small groups so dramatically to be more productive, as well as the asymmetric nature of success in warfare, that this actually empowers smaller city-states yet again. And then this is a longer discussion, but I actually think that's where it goes versus either one.

49:03But anyway, we can talk more about it. Let me add on that point. It makes a lot of sense. Let me add one thing on that, which is I think arguably this swings, AI swings very much the advantage of the defender over the attacker. And I think maybe we're already seeing that play out in Ukraine, which is, you know, there's this classic, you know, military kind of, you know, sort of a tactical conflict in Ukraine as a Russian tank up against a Ukrainian soldier with a shoulder mounted fly by wire, you know, anti-attack weapon, which literally like steers itself effectively with AI. right and it's just like literally you just like fire it vaguely in the direction of a tank and it like comes in on top of the tank and destroys it um and of course that shoulder mounted weapon it's it's technologically more sophisticated it costs you know i don't know a thousandth or a ten thousandth or something the price of a battle tank um and you know you definitely on that battlefield want to be the guy with the shoulder weapon not the guy in the tank um and so all of a sudden it's just like this question of like okay are those tanks actually worth even having anymore like are they just always sitting ducks and so my my guess is this tilts in the direction of defense and therefore in the direction of probably actually reduced levels of violence.

50:06And probably a preservation of the status quo, Joe, to your point, probably smaller states. But I would also say I'm not an expert on this. So that's just a guess. Hi, Mark and Joe. Thanks so much for being here. My name is Ezra. I listened on the way down here to your podcast with Lex this week. And you said the best founders don't take advice, then what is the role of a VC firm beyond investment? And really, in other words, question for you too, Joe, why would the best founders take investment from 8VC or A16Z? So the reason is because every founder is setting off on a 10-year journey where they are going to have challenge after challenge after challenge after challenge.

50:51These stories always get told later as like everything was great and wonderful and everybody got along and all the customers were happy and everything was great. And inside these companies, these startups, it's always a sausage factory. It's always just disaster after catastrophe, after mayhem. It's just constant mayhem. And it's just because you're trying to do something very unnatural. You're trying to change the world. You're trying to bring something new to market. A lot of people don't want you to do that. You're trying to organize a team of people who are a lot like you, who don't necessarily get along by default.

51:17You're trying to convince customers who have many other things they can do with their money to spend the money on you. You're up against the government half the time now. And so you just have this mountain of challenges that you have to climb. And look, I don't know, maybe some people can do that by themselves, or some people can do that with a small number of people without broader support. Most founders decide that they want to be supported. They want to have somebody who stands behind them. And then very specifically, they want their investors to be the kind of people who stand behind them.

51:44And when the going gets tough, they don't freak out and panic, but rather they actually have constructive ways that they can help and help improve the situation. And so that's generally the conversation. Generally, when the founders talk to each other, you know, and when founders are doing reference checks in a venture firm, those are usually the kinds of stories that they're telling, which is, yeah, my company got out of the shit and, you know, my venture firm was really able to help me get out of it. Or my venture firm didn't do anything, right? Or even worse, my venture firm actually, like, took advantage of the situation and screwed me, right?

52:13And so that's the general pattern of recommendation that gets made and it seems to work. Yeah. Yeah. So Mark's on the board of a company I started. And I think there's a few big advantages you get having somebody like that around. First of all, you do get very useful intellectual frameworks. Maybe you don't always take his advice, but there are some things. Mark's convinced me over the years not to raise insane amounts of money because it breaks cultures. I didn't agree with him 10 years ago. I've now seen it enough times. I know he's right, embarrassingly. So that's worked really well. But you also get the fact that Mark is involved or the fact that his firm's involved.

52:44Serious executives take you a lot more seriously. It helps you win deals. They help you bring these people. And really what you're going for with investors, you're trying to get unfair advantages. And there's all sorts of unfair advantages you can get from being in a community with people who've done it before, with people who have deep networks where you need to be. So it's pretty obvious to me having built a lot of things. Like you want whatever unfair advantages you can get. You want whatever supporters you can get. And so I think it still does make a difference. Thanks, Mark and Joe. My name is John Kidenda.

53:11Um, one of my favorite parts of the, the AI articles where you say something to the effect of, um, the AIs are not going to want to kill you because they're not going to want. And my question is given that, do you think that, uh, the whole conversation around AI alignment is, is, is net negative because it kind of tends to push people towards anthropomorphizing at the technology and creates this sense that it's going to have these wants that are similar to the wants of a human being. The whole field of AI alignment and AI safety starts with the assumption that there's a problem, right? You also see this in the academic world because like the origin point of a lot of these ideas at Oxford, I think it's called the Center for Existential Risk, right?

54:02And notice they didn't say the Center for Existential Benefits and Risk. They just said the Center for Existential Risk. And so, you know, these things are wired from the start. You know, I have like nightmares that there had been like a center of internet risk in like 1993, which there wasn't. But had there been, like, would we have ever been allowed to even do what we did? Like, I don't know. You know, imagine a center for microchip risk in 1960, probably also a bad idea. A center for automobile risk in 1930 would have been a bad idea. So, so, so they're just, they're sort of inherently inclined to the negative.

54:33They're, you know, they're sort of critical. The in thing in academia is critical everything. And this is sort of critical technology studies. And so they're inclined to the negative. And then the other thing that just happens is there's a tip-off for what's happening, which is actually in the term AI alignment, which is the AI alignment is not the original term that was actually used for this whole field of things. The original term was AI risk or AI safety. And it was literally, it started with people who were literally worried about things like killer robots and so forth. So So it was that real, like actual risk to life.

55:05The flip of the vocabulary was very telling. So when it went from AI risk to AI alignment, which was about 10 years ago, it was very telling what was actually happening there. And the reason is because with risk, it's just like, okay, risk, safe, unsafe. That's like, okay, very clear. Alignment, it's like, okay, alignment with what? And then what they'll tell you is, oh, it's alignment with human values. And then it's like, well, alignment with whose human values? And basically what happened is there were a set of people who basically took what had been a literally life and death existential question, they turned it into a morality and ethics question.

55:34Of course, why did they do that? In part, it's because they have legitimate concerns about the future of morals and ethics, but in part because they envision themselves as being able to answer those questions. This is why in my essay, I started with risk number one, which was existential risk, but then I went to risk number two, which is societal risk, the risk of what they call societal harms. The harm word is another one of these tip offs when you're dealing with this kind of thing. And for that second risk expressed as AI alignment, what you're dealing with fundamentally is social engineering.

56:05And when you're dealing with social engineering, you're necessarily dealing with politics. And so that's the sort of collapse that's happened inside the AI risk field. By the way, the original AI risk people who are worried about the killer robots, they actually are very upset that this shift happens. And they think that their movement has been hijacked by basically social activists, which it has been. And they now have proposed that the name of the field should be changed to something even and more vividly clear about what they're concerned about. And the leading candidate is AI, not kill everyone-ism, which has not taken off yet, but give it time.

56:39Thank you, John. Thank you very much, Mark, for taking your time to join us. Lauren, I think we're going to wrap up unless you're something else. Thank you so much, John, Mark. Everybody, thank you. All right. See you soon, Mark. Thanks a lot.