Corruption.
Yeah, we're living in the greatest kleptocracy of your in my lifetimes. Someone gets you a billion dollars in crypto, you're bought. Period. So, the question is, where do they go as they get more power? And the answer is Elon became a narcissist. This is Gavin Newsom and this is Reed Hoffman.
Well, welcome. My pleasure to be here. First time in the governor's mansion.
Yes. Not bad. I mean, it's, you know, it's, you know, not I don't know. It's not, but it's, you know, we're blessed to be here.
But I'm blessed to be here. Look, was how many books you written? Six.
Yes. Well, you know, everyone, it was getting to the point where everyone had a podcast. So, I had two.
You had to have two. The most, you know, I mean, you've been around sort of OG, started a few companies. You're part of this mafia, this mob.
Yes. Although I try, you know, everyone likes the mafia kind of thing of it. We didn't do anything criminal. It's PayPal Network.
Did you? Well, I'll get to I want to get to, you know, you guys obviously have embraced that a little bit. And you know, we'll also talk about, you know, it's kind of broken apart a little bit. Yes. The old the PayPal version.
Yeah. PayPal started this company LinkedIn.
And you currently though not maybe for long we'll talk about that on the Microsoft board and you've got a number of AI companies that you've started up and you also have invested seems like everywhere in the AI space and you're doing a lot of angel investing so all of that.
And you haven't even turned 60 years old coming up but not quite.
It's not quite. But I really wanted to get you to talk about AI in the broadest sense because we've look we're just everybody's wringing their hands, the doomers, the gloomers. It's going to take our jobs. The anxiety is boiling. You're seeing it represented in polls and just anecdotal feelings. People are angry about data centers. People are furious that their kid who went to Stanford and got this unbelievable degree is not getting as many interviews as mom and dad expected their perfect child to get. But you have a different take on all this. You're a little bit more optimistic about this. So I mean first you got to acknowledge people are in pain having difficulties right I mean and yes there's difficulties with entry-level jobs and if you listen to the press including by the way a bunch of the AI people who make foolish statements like white collar bloodbath and other kinds of things they go well it's AI's fault and it's like actually if you actually look at it thus far it's not AI's fault doesn't mean there won't be an impact from AI and all that stuff in entry-level jobs.
But the pain that's being felt right now is overhiring from COVID, is tariffs, you know, screwing our economy and business planning. And when all of that kind of stuff happens, businesses go, I'm going to do no hiring until I figure out what's going on.
Right. So, you're saying some of those big headlines that we saw, Box and others, that somehow suggested or at least added in the press release maybe a number of factors including AI and then the mainstream press runs with AI. Overhyped.
Yeah. Well, overhyped, but it's a natural reason for a CEO to say that because if I was like, "Oh, I overhired and I mismanaged. It's my fault." No, no, I'm strong. I'm taking advantage of AI. AI is replacing jobs. I am just being a really good CEO and manager of my shareholder capital. CEOs are smart people. Pick which one do you want door A or door B? It's like I'll take door B. Thank you very much.
So what do you make then of the college graduate and the unemployment rates legitimately at least numerically statistically? What's the causation of that? What would you what's the factor that how would you factor AI in that respect or what do you think more broadly is the reason that those numbers are higher?
Well, I think what is factually the case is that a wide range of businesses are like, we don't know if we should be hiring new entry-level people right now. Some of it's like, well, we're doing layoffs because we hired too many people during the pandemic and then did this whole remote work thing, which is not really working out, and we're refactoring it. And then there's, oh, and now we've got a trade war going on with global tariffs, which are increasing cost to consumers, cost to businesses, and we got volatility central like, hey, it's July, why are we starting a new war? So in all those cases what businesses do the very first thing they do is let's not incur any new expenses. Let's not hire any new people except the very clear people we need. Entry levels are usually like oh these are people we need five years from now. Well we don't know what we need five years from now so let's not do that. It's also let's not do a new lease. Let's just chill. So that's the primary driving actual factor. Now that being said, and by the way there are a few places which are AI related like for example the Meta layoffs are we need to free up cost for spending on compute and so we're going to do so it's not that there's zero it's just that the broad brush college grad is not actually in fact yet. If it's a percentage from AI it's like five, it's not 30 or 50.
So you talk about some that have used some pretty aggressive language about a white collar bloodbath. But I mean you and I don't know if Dario said that or not but I mean it was not surprising on the basis of the question I'll ask which is I mean obviously he's because of his emphasis and he put a numeric about 50% of white collar jobs be eliminated at least new white collar jobs in the next well five years less than five years by 2030 again hyperbole from your perspective and you're a guy who's you know you're a fan of Dario.
Yes, I'm a huge fan of Dario and look, Garland gets criticized that he's doing it for like kind of promoting his own technology and this is how important my technology is and no Dario is very principled right he's like look I think that there's a real worry here and it's not on like society should be ready I'm trying to help society figure out getting ready so I appreciate Dario's principle his motivation etc. Now after he said that I called Dario and I said look what you said is not heard the way you think it's heard. Right? Because what people hear is, well, if that's going to create a huge bloodbath and that's the outcome, why are you building AI so fast? Why are you doing all this stuff? Right? Is it just to make yourself rich, right? Because what they hear is they hear the, "Hey everyone, I have this really great technology that's going to like ruin half of your lives and good for me and sucks to be you, right?" That's what they hear. That's not what you meant to say, but that's what they hear. And so I was like look I think the I said Dario the better thing to say is we're going to have a lot of job transition and there's a lot of uncertainty what's going on and it could have a really quick massive impact and we should be figuring out what to do on it.
Right. And is that so I mean you've talked about it in the terms of a cognitive industrial revolution and you talked about not being Poland rather be England take advantage of this and that's and you've done that in the context of making the case that we you know we should be accelerationist though probably with the notion of a steer yes so you're not naive.
Sometimes you slow down on the curve.
And slow down on the curve but vehemently opposed to the sort of precautionary principle framework this notion of this is you know we've just got to put a damn pause on all of this.
Yes. So, so let's start with the pause because it's simple, which is say you issue a call for a pause and you have two groups of people. The group of people who pause and the group of people who don't pause. The group of people pause go, "Oh, I hear you. The humanist stuff really matters and I care." So, all pause, right? And then the people who go, "I don't give a shit." And they keep going. So, then what kind of AI is built, right? Not good. It's actually has a negative impact, right? That's the reason why I'm vehemently opposed to this kind of pause and things and it's like well what if we can get everyone to pause like well what's your plan to get everyone to pause including China and everyone like is this at all realistic? So my strategy tends to be how do we keep like you want the good guys to win, the things that are pro-humanist, keeping human beings at the center, keeping a notion of what is a positive to society. By the way, positive societies over years, that doesn't mean that there won't be massive transformation things that will be painful and difficult to get through, but it's like having this is how society's better with this as where we're going.
So when people hear that, I mean the good guys, I mean tech is getting slaughtered in the context of public opinion. People just don't trust anybody. Whole issue of truth and trust obviously has been accelerated social media but now obviously with AI deep fake this deep fake that voice and we can get to even your own interesting avatar in that space. But this notion of trust and you know someone may be listening go okay yeah Reed this is easy for you to say you know sitting on the Microsoft board and obviously investing in all these places you got two of your own AI companies you know come on man.
So I am not of the view that like frequently what it is is like just trust us right like you know it's like look we're in a very low trust environment overall, right? Low trust of almost every institution, even mixed trust of historically great trust institutions, doctors and so forth. So it's like there's an overall thing. So they just trust us. But then you have to say is like okay well you have to prove my trust before you do anything. Well that doesn't really work and it doesn't work in any of these places, right? So the notion is to say well do the things that ultimately are building trust over time and then when people really investigate like ah you're doing something that's fairly trustworthy. So like we were talking about Dario. One of the things that Dario did is they said okay we're going to build AI with a kind of a constitution. I actually don't think they should have used that word, but like you know it could be a set of principles.
What offends you about constitution, sir?
Well I for one am a huge believer in the American constitution right it is our thing.
I'm a little old school like you. I appreciate that. We'll get to that a little bit later.
Look that is our religion is the constitution. Right. So like we stay with that and it's a little bit of saying oh implying that this is a constitution is not really that like a constitution is a sacred document for me. So it's more simple as that.
And so the but here's how we're going to build trust is here is the document that is central to all of our training of our AIs. Right? So you can read it, you can comment on it to us, you can suggest changes, you can understand where we're going, what we're trying to do. That's the kind of thing that I think, you know, needs to be there.
Yes. That's the beginning.
And that's along the lines I mean you and I have had these conversations over the years and California's regulatory framework SB53 which we advanced was around this issue of transparency for frontier models, these large language models. What about this? You know, I think from your perspective it'd be interesting to hear your perspective about some of those contours. I mean, we talk about Dario as the, you know, the good guy and for a lot of folks that's through purely political lens because of some of the issues he had with Palantir and the Department of Defense, issues around mythos, which I'm interested in your take on mythos and the fact that he pulled that back, what that means or what that suggests. But it also has the origin story that played out very publicly with Sam Altman with Elon Musk all those origin stories of which you had a front seat sir.
Front seat front seat and participation in some of.
An actual board seat as it relates to OpenAI.
Exactly. So look most of the people creating AI are like kind of unique human beings.
So you know and part of it is because by the way geniuses frequently are.
Right. And so with folks who it's pretty easy to attack anybody. I mean you've had that experience. I've had the experience of it's like an attack by a negative person who is willing to invent shit or build on what is a small thing to make it into a huge thing is one of the things that kind of flows. Now, I have the fortune of knowing all of the heads, like every head of all these labs, and so I have a pretty good sense of what their actual moral characters are and what the things are. And many of them are good human beings. That doesn't mean they're perfect, right? But they're good human.
But it's a I mean, I want to pause on that because I mean, objectively from your perspective, many are good human beings.
Because they've been I mean, these guys really have been vilified across the spectrum. Yes. And perhaps more so even today with the first trillionaire being minted etc and all that issue which I want.
On that I didn't say all were good human beings.
Well we'll get maybe go specifically to that human being but this notion that most are good human beings but they're also trapped by incentives right I mean we're all creatures of incentives and this notion the incentive as you suggest to accelerate at peril if it's not us it will be someone else would you rather the American stack in the context of what Jensen talks about in Nvidia, this notion of the American stack with the American values to the extent possible. But what I mean those incentives aren't necessarily even good people for good behavior.
Yeah. So in any incentive system you will get some good effects of it and some bad effects. No incentive system is only good effects.
Right. So for example you have an incentive system to try to make more money. Well, you try like it creates jobs and productivity, but you also do a whole bunch of other things like so it's a blend, right? And you get people selling cigarettes and other kinds of things. So, I think the incentive system, what you have to do is say how do we tune the incentive system so that the major incentives mostly align with good outcomes. Now, part of the reason why is you already noted the England versus Poland and the industrial revolution. If we are the society that kind of has the primary pole position in how AI does this next industrial revolution which I think is going to be bigger than the previous industrial revolution and where this then we can sort out how the economics and society works. Like if you say well we have the massive economic benefit occurs to our society and like well but it's only these seven companies. Well, we're in a society. We can broaden that out some. You say, "Well, any company that has profitability of above X per employee now has an extra tax that's used to pay for all the rest of society." And if you own that industry, you can do that. If you don't own that industry, it's a little bit like when people say, "Hey, I hate these social networks." And by the way, there's some real social network issues. Although I always, to my own defense, point to LinkedIn. One of the few, if not the only social network that actually had no I mean, right? Yeah.
I mean, kids are doing fine on LinkedIn if they're on that.
Yes. Exactly. Like LinkedIn works just fine. So, it's doable. But if you said, "Hey, we ban social networks." Even ignoring the LinkedIn question, it's like, well, would you want it to be only Chinese social networks? No. No. You'd rather have the social networks here and have our ability to shape them. Now, of course, we need to. There's the political melee around it and all the rest, right? But you want it to be a US property that you're shaping. And the same thing with AI. And that's part of the reason I'm an accelerationist. It's like one of the things I was saying a couple years ago is we want AI to be American intelligence, right? So it matters, right? And that doesn't mean at any cost. Nothing is at any cost, right? Like if you said, well, it would destroy an entire generation of children, that that's a horrific like can we avoid that and still be accelerationist the answer is I think yes but it's like okay what are the things we do in order to do that.
Now I do want to wrap back to one of the things before because I think this is very important given you know we've seen American college students going I hate AI right booing I mean have you given a did you do a commencement you probably smart enough not to do one this year who needs that.
Yes well because you're showing up to try to offer something to help them.
Right. Like the usual protest and commencement speech is sleeping. You sleep.
But what do you mean? What do you mean? But it reinforces the anxiety.
Yes. No. No. And I get it.
Highly educated people that have played by the rules.
Yeah. And I feel I look I understand I feel the pain. Right. So, but the question is, and this is why I think it's so important is look, it is completely legitimate for college graduates to be going, you know, hey, the older generations, the boomers, etc., you've really fucked it over for us. We got all this shit going on and this is on your watch.
And you suck. I get it. And I think it's a legitimate gripe, right? Like but in crisis, this is the thing I really want the kids to hear or you know the college kids. Crisis is opportunity. And by the way, you can be the AI generation. You should be using it. You should be going, "Hey, company X, you need people to help you like AI your company. I can be the person doing it." You shouldn't be trying to opt out. It's a disaster for you, right? And I'm saying it for them. Like I don't have any stake in them. Like the entire current graduating class could all say we refuse to use AI. It makes no difference to me, but it will be terrible for their lives, right? And so it's like, look, embrace and figure out how to make it useful to you and then have that help amplify your career and life path.
And you continue to make that point. It's about amplification. It's about having a co-pilot, but this notion back to turbulence.
This idea that the transition. So talk about the transition in the context. Look, if Dario is overrating the short term, are we understating the long term or is it just modest fits and starts a scale scope speed? Where are you in all that? How do you calibrate all that? If you're a policy maker sitting in this seat, how do you begin to anticipate where that stress, where that friction, where the real turbulence is going to come?
Well, some of what I thought you guys did that was really smart with SB53 was get data, start monitoring, start asking questions like because it's like, okay, what's really happening here? Not, oh, I watched a journalist wrote a story about one case, you know, blah blah blah blah. That's not the way to do it. That's like the dumb, oh my god, we should restrict airplane travel because it's unsafe. It's like, no, no, no. It's the drive to the airport that's unsafe, right? The plane is actually so much safer, right? Per mile, per minute, everything else just like no, the plane is much better. So, it's like look, making policy decisions based on data and we should be getting the actual data, not the hysteric kind of responses and start there is actually really important. Now that being said, you know, as a rough template, we want to be thinking about, okay, AI is this transformative tsunami that's coming and within a small number of years, there will be a large number of jobs that will be completely transformed. Some of those jobs will be they're just all done by AI. Some of them will now be clerical jobs.
What kind of jobs? I mean people have their theories of this and of course Anthropic puts out literally the pie chart of what kind of jobs are most likely to be impacted. What's your theory of that case?
So those pie charts are usually limited in that they say if the tasks of doing the jobs are only the ones today and that there aren't new tasks this is what it look like there will be new tasks. Amen. Now that being said, for example, I think that there's a relatively small number of years before a customer service person, a customer calling customer service says, "Please put the AI on." It's a lot more helpful, right? Because you have a human being trying to follow a robotic script by looking at the screen or the AI is going to do that so much better. There's just like there there's rules for human beings, but they're like looking at what's going on with the pattern of AIs and trying to reimagine. That's not the bulk customer service jobs. So like that's one. I think there's a lot of like there's whole classes of sales jobs that I think will be much more AI and you begin to get to this weird sci-fi story where the AI salesperson, the sales agent calls the company AI agent to listen to the AI sales agent and there's a whole thing that kind of navigates through. It's a whole weird universe that will happen, right? It will happen.
Yeah. N years and not three, maybe 10, like not 20. Right.
Right. And so, those will be like kind of swaths of jobs. But here's the thing, and it's that new task thing I was mentioning earlier, which is, basically there will no longer be, there will be very few human jobs that are just humans by themselves. We will all have teams of agents that are working with us and doing. But actually in fact what you can see from the pattern of already looking at it is you actually get to some pretty incredible increases in performance where the human brings in the things that the AIs don't have and look one of the question is will the AIs not have those for a long time? Will they get them quickly? We don't know. It's all coming right. My bet would be we're very adaptive. Like this whole like human beings will be replaced and we'll have our cognitive faculties goes all the way back to the printing press and writing. Like I actually think there's areas of where we make good judgment. Like people say, "Oh my god, look at Claude code." Like there's no need for engineers anymore. Well, actually I talked to a whole bunch of people who use Claude code and I've used it myself intensively and I've used codex and OpenAI and I've used Microsoft Copilot and by the way some of it is just amazing like amazing like if you had told me like you would asked me even like say five to eight years ago would I be seeing this I would said no no no that not yet like yes here it is that's amazing.
On the other hand it breaks on weird shit. Right.
And so part of the human thing is like orchestration. It's like, okay, like for example, how does coding work at OpenAI? Well, they have a big screen and they have multiple agents working and they're like, oh, oh, that one's stuck. We got to get that one unstuck, right? Okay, this and and it makes them hugely more productive. Now, if you thought, well, it makes them hugely more productive. They're just doing all the work. We don't need anyone else doing all the work. It's like, look, there's infinite work. It's like now maybe there will be say for example tech companies may more naturally be at you know call it 50% of the current engineering crew that doesn't mean that those engineers it's like other engineering firms start more competitors etc etc it's just the productivity of how you operate it's like the humans aren't replaced in all cases by AI they're replaced by humans using AI and so you want to be the humans using AI and this gets back to I think is a central point is like well what's the thing you should be thinking about is one is monitoring and then the other one is how do we help AI be part of the solution because the thing that's driving it at scale and speed is AI the we need solutions that work on the same capability set so for example okay how do I help AI how do I have AI helping people make job transitions I want to be seeing that. Right. So like one of the things you could do right here at the mansion is you could say hey great you know California is the center of the AI revolution in the world. It's part of the reason why 53 mattered. Hey I want you guys to all come by and tell me about how AI could be useful in helping people all ranges. I want to hear about blue collar white collar multiple industries. How does that help those people, find the right work, do the right work, have pride in their work, we're going to have a day here in the mansion. You're going to come talk to us about it because I want to hear you thinking about it.
Right. Yeah. Because by the way ultimately if you're not delivering on this then I have to apply more pressure because like my responsibility is to the citizens of the state and look I'm very happy that we have the AI revolution and the economics flowing in from that and all the rest but I care about all my citizens.
Yeah. So tell me what you can do for them.
Yeah. Interesting. I mean, this notion of using AI, the thing that has displaced that worker to help find and place that worker exactly in a new opportunity, a new career that's not just training per se and the old sort of black and white movie community college construct.
Yes. But I want if I could I want to get back there's so many areas on compute and issues limiting issues in that space and what's going on with data centers etc. and where we are in terms of the competitive landscape. You I didn't know this. You were at Stanford back in the black and talk about black and white movie. Wasn't that it wasn't just a couple years ago, but way back when.
And you were studying what was simplistic solution. What the hell is this?
Solutions. Symbolic systems.
Yeah. What the hell is this?
Well, it's artificial intelligence.
So I was part of as an undergraduate I had an AI degree.
You had an AI degree?
When no one had ever heard of this nonsense. Well, not no one.
I mean, most of folks never heard of it.
Even though we started realizing we all had it and we've all been using it for decades, but this notion of Gen AI became the thing that we kind of conflated.
And so you was it just a natural inclination? I mean, did you see the future then? I mean, was it just a major that started to present itself? No one else was in the class and you figured, yeah, I couldn't get in the class. I get an A on this one.
No it was so what people assume people in my kind of role and many people in my kind of role are fascinated by technology I'm not I'm fascinated by human beings.
Right and so what I was interested in was human thought and human language and so how do we understand ourselves how do we understand each other how do we communicate how do we get to a better understanding talking to each other etc etc and that was what most interested me and so like I looked at you know do I do anthropology or psychology like I was looking at the whole range but I was like what's the thing that in doing this I could actually make the most interesting things and I've always had a belief that we evolve ourselves through technology like this is a you know these glasses these are a piece of technology these podcast like and that's actually how we evolve our state as human beings and who we are and so forth right and so I was like okay this AI thing that's interesting right so I'll go do that and I did it symbolic system is a multi-disciplinary major. So it included psychology, included linguistics, included philosophy. So you know, did all of that as part of it.
And you went to Oxford and got it in philosophy as well.
Yes. Because well, what I concluded at Stanford was none of us understood what thinking and speaking was, right? So here we are trying to build AI and we have no sense of what that is. Maybe philosophers did. Went to Oxford included philosophers didn't either.
And so give me I mean this is what year were you studying this stuff? What roughly?
To 1993 in both Oxford and Stanford.
And so AI as we know it today in the context of the vernacular of this sort of notion of Gen AI etc. I mean the origin story they tend to lazily go back. It's sort of like the beginning of biotech may have been 79 with Genentech or something. There's a lazy punditry to that. But is that punditry accurate to suggest somehow in the minds DeepMind of Google that sort of the origin stories started to take shape early 2013 14 range or how would you describe what we are addressing or dealing with today in the context of that origin story? Where to really begin at a scale that we're now more accustomed to.
Well, so by the way, like many technological revolutions, including modern ones, there have been multiple like, oh, in five years we'll be having AI invent new science and that goes all the way back to the 80s.
80s. Yeah. So then here we are, we're there because what happens is some achievement happens and then people go aha. So like we beat human grandmaster at chess.
That was the hard problem and now it'll be inventing physics and you're like not so much.
And so it's like cycling through it. And that was actually part of my conclusion at Stanford was like none of these current technical approaches will work. So what's the thing that got me back in AI was a three-hour meeting with Demis the CEO and co-founder of DeepMind.
And what I realized that he had that he had like he is the I think the he and his co-founder is the original kickoff of DeepMind and it's not I mean again it's not Microsoft it's not Anthropic OpenAI it's not SpaceX's I mean that's the I mean if you really want to start to understand and unlock an understanding what Demis realized was we have scale compute and with scale compute we can create scale intelligence now his original idea has been less of the important thing which is self-play. So we had pong and go and it plays against each other. But the answer was we can apply a lot of compute to evolve a competent machine with cognitive capabilities and I was like you're right we now have scale compute and what's more we have scale data and we have scale engagement you know through internet and mobile. The revolution, a revolution, a massive revolution is here. Now, did I know when I walked out of that meeting, you know, and at King's Cross that I was like, "Oh, actually, it won't be the self-play. It'll actually be these things called transformers, these large language models that read the trillion and a half words on the internet and use that as a learning basis to create something." No, I didn't know that. But what I did know is this is now a revolution. And so it's like, okay, how do I help? And you know, for me, I'm not adverse to doing stuff commercially. Obviously, I've done a whole bunch of stuff commercially, but for me, I look across the whole spectrum. I go, okay, what's the stuff that I should do as a founder, as investor? What's the stuff that I should do as a content provider and author? What's the stuff that I should do? Talking to you know people who care about society like you and others and what to do like how do I go from humanity society industry and that's the order in which I care about them right humanity first society second industry third.
Industry can be very useful for society and humanity.
But you like part of the like when you were talking about that earlier steering and shaping well we want to steer and shape industry to go it should be on average massively positive for society over time right and if that's not the trajectory then it needs to be fixed.
Right now over time, not this year necessarily right. There's always costs of transformation and all the rest now. And so I went that's the revolution we're in. And so for me, I think that's 2013 or 2014.
Right. And it's and that's why when like Sam called me and said, "Hey, I'm thinking pulling this thing together with, you know, for AI for humanity, will you help?" The answer was yes, right? It was one of the things where I had just gotten my partners at Greylock really focused on crypto. And I was like, "Oh, I'm going to stop doing this crypto thing because this AI thing is coming." And they're like, "What AI thing?" I'm like, "Oh, it's coming." And actually, you know, I think my partners at Greylock were the first venture partnership that saw GPT-4 because I got Sam's permission. I was on the board of OpenAI to say, "Hey, come in and demo this and show this because there was a six-month period where GPT-4 was not public because we're like, "Oh, is it safe enough?" and all the rest. And you know, they were doing it. I was like, "Okay, look, I think it's safe enough. Is it okay to show the partnership?" And Sam's like, "Yeah, yeah. This since you're serving on a 501(c)(3) board of OpenAI trying to do this, this is the least we could do. Yes, you could go show them, you know, what's coming."
Google, there's or there's some interesting and I don't, you know, I don't want to get into gossip, but I think it's interesting because it's instructive to a lot of people and it's also I think connects a lot of your own personal relationships and dots. So, you you know, you were notoriously, as we were alluding to earlier, part of the PayPal Mafia. And those are the, you know, brand names today. Elon Musk, obviously, Peter Thiel, David Sacks, less of a brand name, but increasingly trying to be one. And in terms of his association with the Trump administration, Musk allegedly and Larry Page, co-founder of Google, had a conversation that as least Musk's telling goes, Musk was concerned that Larry wasn't taking this Gen AI seriously enough in terms of its downside risks. And decided to go off on his own. Made the same contacts you made with Sam Altman and others and went down this path of a 501(c)(3) and tried to birth OpenAI.
Is that reasonably accurate?
That is reasonably accurate. I think it was a conversation I think was you know Larry was kind of saying was you know kind of it's probably too simplistic but you know kind of articulating a transhumanist perspective to Elon that you know naturally freaks someone out because it is the rational response to that is it's a little strange right and so got very concerned about it now you know in Elon's case in particular it's the you know Elon has the definition of a messiah complex. So it's like no no I have to be the one building AI not anyone else. So there's a little bit of the I'm worried about it generally but it's also the I'm worried it's you not me. I think that's sort of proven itself a little bit.
This sort of notion of going back even when you said the glasses which is interesting technology right can see better. We've been aided and embedded and as a human that technology it's not just you know ones and zeros. Yes. Though you know now the or now they're becoming ones and zeros as Meta and others I think and now Apple's coming out with their own version but what this this notion of technology trans I mean you remind when with what what allegedly Larry said to Elon that freaked Elon out is what guys like Peter Thiel couldn't even answer you know I think it was a 17-second pause in an interview about you know human beings and technology. But when you you sort of invited that in with the glasses. I mean, what what is all that? I mean, for folks that may not be familiar with some of this, but maybe familiar with the anxiety that some of these conversations have induced more broadly. What is the suggestion? Is this are we just all in a simulation or something? Is that the point? There's that kind of.
You've been in a few of those conversations. A number of them. Yes. It's like actually I was sitting with two of my friends at the time. Elon was one of them, right? And they're like, "Oh yeah, everyone smart I know is living in a simulation." And I was like, "Am I smart?" And they're like, "Yes." I'm like, "Well, then not everyone."
So, and it wasn't I remember the origin that we had Second Life. We had some of these virtual realities. Maybe they were alluding to a version of playing around with that, but they mean quite literally we're in a simulation.
And so, I think simulation is the kind of classic theory that gets people to like, the kind of Christian theory of intelligent design about how we get created or the Fermi paradox like we have questions we don't know the answers to. Therefore, right, simulation and you're like, "No, no, we have questions we don't have answers to. Stop right like we will try to figure out the answers to those questions." Like the simulation theory stuff is like, "Sure, there's unknowns." Like for example, the classic one that's usually used to argue with this is we don't have signals that there's other intelligent life in the universe. And so you know like with radio waves and statistics and all the rest we should have that well okay that's an interesting puzzle that's an interesting question why do we not have that and but like and therefore we're in a simulation is like that's a lot of cognitive theoretical infrastructure to explain this a similar way like the hey we have an unknown about the complexity about how human brains evolved and what the intermediate steps looked like. Well, therefore, there must be a creator. Like, well, there's a lot of it's an unknown. And by the way, therefore, there must be a creator and therefore by the way, the creator must be my religion, not any of the other 20s. Like, it's like what? Like, it's like no, no, pause on the mystery, right? And then go, okay, we have an unknown to answer. So, there's a whole bunch of craziness. But I think coming back to your the first part of your question I think is very important because I think this like I've been giving some speeches in Italy like I gave one at Bologna and Perugia because I wanted Silicon Valley to remember some of the important humanist nature of the Renaissance which is what is it to really be a humanist when you're thinking about this technology and it's not just self-declaration. I am humanist you're like no that that that's nice. It's better than I'm not. Yep. Right. But it's like what is it to be humanist to say look I have a theory about why the work I'm doing will cause a much better result for call it at least 80% of humanity. Then the next thing is and I'm transparent about it. I'm talking about it. I'm giving you a chance to critique me, dialogue with me, etc., etc. And then this is a super important thing is I'm not like a narcissist about it. I don't come to my theory of humanity because I go, "Well, this is good for me, right?" Or I'm the most genius person in the world. Everyone else's theory of humanity is bad. Mine is the good one. But no, no. Engage in dialogue, get critique, understand kind of what's going on for this. That's what being a humanist for technology is, right? And we have some, right? Great. It doesn't mean that they don't make mistakes, Dario, Sam, etc. But they're actually in fact they are committed. And you can tell they're committed because they do things that are not in their own economic self-interest and they invest in it in order to try to make it work. Doesn't mean they're perfect. Doesn't mean it's not worth speaking up and say, "Hey, do X do also more of X or less of Y. Perfectly good thing to do." It's part of the reason why we have the dialogue about it to say, "Hey, the way that we try to make decisions about because the short answer is these technologies will be built by small groups of people, right? and will have humanity level impact and you're like okay you can't then say hey we're going to have UN global voting day that's not the way it's going to work so the dialogue and bringing up the concerns about what it is is that iteration is the way that we get to being more broadly humanist.
One of the great anxieties and you triggered it this notion that it's a handful of people and you know We talked about the American version and at least having a chance at regulating something within our quote unquote jurisdiction, but regulatory capture is real. You have just massive obscene off-the-charts wealth.
Comes with massive obscene influence to write the rules and regulations to and the incumbency protection racket. To lock everything out and lock everybody out.
Yes. I mean you know so how does that play in I mean that certainly plays that is playing into people's anxiety in a profound and outsized way how do we address both end so when you have a kind of a tsunami like this and this is to some degree talking to you talking to the Dems on this who tend to be the oh fight all the wealthy power and you're like you need capital on your side too.
Right politically right so have that as a good theory of the game, right?
So, the issue is say and like for example, and I'll come back to AI, but this is like the mistake the Dems made with crypto, which is like there's a huge amount of crypto wealth. If we just go attack it and go, "Hey, we're just going to try to kill all of you." They go, "You're trying to kill us. Okay, we're going to vote the other way with not just our votes, but our dollars and all the rest." And it's like, okay, that's really painful and causes a whole bunch of bad outcomes for society. Yeah. So, again, you go, well, how do we shape it? It's like, look, we're not trying to like extinguish all of you. We're trying to make sure that crypto has the right positive function in society. And they said, well, I can't envision anything. It's like, well, I come call me. I can envision things, right? There's payments, there's much cheaper banking systems, there's all kinds of things. Let's just shape it that direction. Yeah, I get it. You don't like because who does like terrorist ransom coin, right? Like you know, like there's a bunch of bad shit. We should do bad shit about that. Or the question of well it's this random speculation. And it's like, well, look, we do allow betting in various ways. Let's try to contain it and make sure it's not too bad. You know, same reason why we go, you know, alcohol has a bunch of bad outcomes, but it's better for society if we go, sure, it's okay if when you get to age, you drink, don't drive. But it's like that kind of stuff. You have to shape it, right? Because there's certain ways. Now, you get to AI and you say, "All right, an AI wealth generation." And it's like this is creating a massive amount of economics that will have massive impact politically. And by the way what's shared across all this economics is a interest in not having the economics of the AI industry like trimmed right and by the way some of that trimming is a terrible idea which is let's just reduce our own American wealth. Some of it like taxation for helping the rest of society can be a good idea, right? So, it's, and you know, getting it right is hard, but I think it's doable. And it's like, well, then you go to, well, who are the people who would be with me on, hey, we should tax AI in some good ways, robot tax, whatever else in order make it happen. What are the ways we do that? So, we take off some of the AI wealth that fights on our side for helping society and we fight against the other AI wealth that's like, "No, no, no. I should take my trillion dollars and be able to do whatever the hell I want."
So, easier said than done. I mean, this gets to and we talk about that turbulence and we talk about the transition and using your vernacular and you've there's a lot of conversations about all right if if this does have a more pronounced impact sooner than we anticipated and we don't have the answer to what to do about that and we don't have the AI solution to address the AI problem in terms of that transition this notion of universal basic income versus universal basic capital this notion of equity this notion of ownership have broadly shared so that we can address some of that anxiety. We all have a stake in the game. You tend to be on the UBC side of this equation. But the question I have for you is not where you are on that simple question, the binary, at least in that context, which is hardly the binary, but the question I posed is how the hell you do it? Who's going to do it? You think Sam's gonna He says two and a half percent. You know what? They're about to go public. We've called for it in California. Why doesn't he do it here in California? Let's we'll happily take two and a half percent a year in this his home state that help birth this damn industry. Think he's going to pick up the phone after hearing this?
He might. I like Sam cares about this stuff.
To his credit, he put out.
Yes, he put that out. He's been running universal basic income experiments funding it himself in Oakland. Like he's.
This is the moment with all these IPOs coming Dario, here's the moment.
Yes. No. So like I think that's a great call to ask right now. I actually believe of all of the OpenAI folks and all of the Anthropic folks, they are very philanthropically minded.
So I think like I.
Well the foundation I mean objectively you can you can dismiss these I get the populist these guys.
But they did put a foundation with the largest in the world.
That they set up that exists.
And has tens of over hundred billion dollars of capital today.
Yes. Exactly. And so and what I think is unlike for example take the OpenAI Foundation. I don't think they're going to be sitting on their hands, right? Right. And I actually think one of the things that you know you should do and I think California should do you know it should be is hey you guys you know were only possible because of California right like we provided the platform by which you guys could do this could you guys you know disproportionately focus on helping our state and doing stuff we'd love to talk to you.
Supposed to take your capital gains to Texas.
Texas Florida etc. It's like.
Like you couldn't have made this without us.
So please contribute back.
Yeah. Or or contribute back without the please.
Yes. We're grateful for the job creation, the innovation. We're grateful for all the energy and daring. We're very very proud of your success and please.
Yes. And time to Yes. Yes.
Right. Because like for example, you know, SpaceX forward impossible without this state.
Yeah. Tesla wouldn't exist without the regulations.
Impossible without this state. Thank you. Right. So it's like.
You're welcome Elon and thank you for your innovation in Darian as well.
Yes. Like by the way extremely important like absolutely.
I'm pretty sure you're gonna ask me a question where I'm about to dump on Elon but.
You for good reason. But this is amazing.
Reason for me to have to ask.
Yes. But why isn't I mean so let's you open that door. Why isn't he feel this way?
I mean what what what is it? You know I I remember there was a different guy though. I was with someone the other day. I said, "No, he was always this guy." And I said, "Well, he didn't appear that way in the 15 interactions I had. You know him better than anybody." So, look, people are dynamic. So, the question is, where do they go as they get more power, more sycophants around them and so forth? And the answer is Elon became a narcissist who he wasn't a narcissist. I think he was a small narcissist and now he's a big narcissist.
Right. And like he lies through his teeth constantly. I don't know if he tells himself stories about it like and it basically like the whole pitiful OpenAI lawsuit.
Was it pathetic and courts agreed.
Yes. Well and it's pathetic written out large for everyone to see.
Yeah. That's that too, right? It's like it's a court of law with truth in documents, depositions, etc. It's just pathetic. It's like the no no like I'm trying to do anything possible to argue that you guys are wrong and I was right. And it was like no, they were right and you were wrong.
Yeah. But it's just I mean it so it just I mean I guess it I mean it's just a classic clinical definition of narcissism.
And and like for example he goes I like for example he helps found OpenAI as like we need to have AI for the benefit of humanity not just Google.
Okay. Then he goes oh shit I should have my own.
And in a desperate bid for market share he's like oh yeah that's fine if we create a whole bunch of like pornographic images some of which are children. You know, journalists have been reporting on this, etc. And that's because it's more important that I have this AI thing than this damage that we're doing to children and all the rest of stuff. It's like it's crazy.
So how do I mean when by the way just I mean accelerating and fast forwarding I mean space is a pretty remarkable company on the basis of the facts and evidence Starlink in particular but from a sort of cash flow perspective an ROI perspective.
And a goodness for society perspective.
Yeah and and by the way I've seen that firsthand in the middle of the Amazon which actually really inspired and surprised me to to be that in fact universally awkwardly connected in the middle of the rainforest. Even more so than frankly out here in Silicon Valley. But I was curious just you know with SP with regard to SpaceX in the IPO. I mean there's other component parts of that that are a little more dubious and questionable. There are folks out there deeply more cynical than I am that just call it a damn Ponzi scheme. Period. Full stop where you are in for example.
Look I so I'm a small investor. But I think it's nuts, right? I think it's nuts, right? I mean it's like look for example you look at the you just look at the perspectives and the the the biggest part of it is payments from Anthropic for these data centers that are horrific for the environment are badly set up in their networking equipment that nominally he's trying to restart xAI for the third time by cursor to go somewhere by cursor $60 billion San Francisco based company that five years ago didn't exist.
Yes. and and it's like okay and I'm glad he's buying cursor you know great you know hopefully something will come of it.
What are they a vibe what what is vibe coding sorry to go down this rabbit hole no worries.
But it's a vibe coding company.
Well it's one of the early revolutionaries of how do you have AI accelerate your coding and vibe coding is a part of that it's not the whole thing it's like you know basically anyone who is coding who's not using AI right now or not everyone call it 90 plus%. If you're coding and not using AI, you're making a mistake, right? So, and it's only getting better, right? And both Claude codex sorry Claude code, OpenAI codex, I think Microsoft Copilot I think we'll have a good entry here. are all like I think like the fundamental way has changed the entire world in terms of how we operate. So Anthropic's willing to buy the revenue but like that is your primary revenue thing in the thing that you're going public on like this is.
And the government.
Yes. Well and then you're like oh the DOJ is saying they've got to drop the lawsuit because this is part of our national security defense. Like, well, that's none of that's in their prospectus. And by the way, the entire xAI thing, Elon himself said, it's a complete train wreck. We're rebuilding it. I was like, well, you know, if you had like honor integrity, you'd say, well, we'll hand back those government contracts because we just said our product was a complete train wreck, right? It's like, you know, so just it's all corruption.
I mean, you just call it.
Yeah. We're living in the greatest kleptocracy of your in my lifetimes.
And it's I want to circle back on that. Stay tuned for those that are tuning in on that because that deserves some attention. But as we we sort of bounce back and forth, the Anthropic data center point you made and this notion of compute and you made the point the biggest limiting factor in terms of the acceleration is less the chips as important and profound as they are. It's really about the compute. And so it gets to the question of data centers and all the anxiety in data centers which has actually been sort of a leading indicator of one of the reasons people have such animus towards AI because they connect the dots and there's been a few studies DC Maryland that 60% of the growth of those retail or rather residential utility bills have been attached to the issue of data center growth. So where are you in all that? I know you've said we blamed AI for everything but the issue of utility costs which have been growing.
Yep. separate from AI certainly haven't been benefited from centers. Look, I think that the clear thing for you and other folks to do is say, look, we want to enable data centers that are net very positive contributors to communities. So, for example, a very easy one is to say, hey, if you're building a massive data center, you have to provision power at like 130%. Right? And that 30% is that's amongst the things that you're contributing back, right? And so by the way, who should benefit from that? Well, the local residents who should have a choice of whether or not they want that or not, you know, their power bills go down, right, as part of that. I mean, the irony is the data centers are getting subsidies on their bills.
Yeah. And we're socializing an increase on everybody else's.
No. And so what it should be is like no, you should contribute. Look, we will help you because by the way, look knowing from the inside Microsoft board and other things.
