Ep 121: Reid Hoffman on AI Optimism & How the Democratic Party Alienated Silicon Valley

0:00AI is coming for humanity, AI is coming for our jobs, AI is coming for our data. It's all this kind of fear and paranoia, which is a normal human response, but an irrational one. What could possibly go right with your AI future? People misunderstand a little bit of what the general frame of the AI revolution is. We build it towards a better future, and you don't get the futures you want by avoiding the futures you don't want. You're involved in Aurora. Tell us about this. Autonomous trucking. Aurora, a company headquartered in California, but of course, in modern regulatory environment. Where are we driving our first heavy-duty trucks?

0:32Texas? Of course. The Democratic Party really did alienate a section of Silicon Valley. There's a bunch of things that red states are doing better than blue states, and we should learn from them. I think one of the things that makes great founders, entrepreneurs, and investors...

0:54The American Optimist podcast has been featuring a lot of my friends. A little more on the right recently, we've had Peter Thiel, Marc Andreessen, Dave Rubin, Steve Hilton's running for governor of California. Just this week, we had Russ Vaudon, who's Trump's OMB director. I decided, let's try something new. Let's go over to one of my old friends on the left, Reid Hoffman. Reid is very controversial in right circles because he's been a major Democratic donor, but he and I also have a lot of things in common otherwise. We're both huge techno-optimists. We both see how AI is going to lift up hundreds of millions of lives.

1:26I think Reid would agree with me that there's lots of special interests, even on the left, trying to stop this from doing great things. How does he react to that when we talk to him about it? What are the areas of policy we disagree on and agree on this? And how are we going to get AI to be an optimistic future? What is Reid building in this area? Let's hear from one of the great builders of Silicon Valley. Welcome to American Optimist. We are very lucky today to have Reid Hoffman with us. Reid, thanks for joining. It's a pleasure to be here. I think we've known each other for quite a long time.

1:51At least 10 years. Yeah, at least 10 years. I was a little intern a long time ago at PayPal, but I wasn't senior enough to interact much with you then. But we've gotten to know each other in other contexts. We generally are like fighting when it comes to politics, but it turns out we agree on a lot of things in tech and in policy as well. Yes, 100%. So you're a partner, Greylock, chairman of Village Global. You have your podcast. It's very popular. You have a new book, Super Agency. I love the title of the new book. It says, what could possibly go right with our AI future? That's a good subtitle for American Optimist.

2:17I want to get to AI. What's exciting about the future? But before we do that, I want to go back to PayPal a little bit. It's interesting. All these characters are bigger characters these days than it used to be, whether it's Elon or Peter or David Sachs. Or you or me. Yeah, it's true. Or you or me as a junior intern there. And now I'm not going to say it's not as important still, but it's a great group. What was unique about that culture? What was your, in your explanation in your mind, for what brought those people together? What do you have in common? Well, it was a set of different things. You know, I got, when Peter and Max founded PayPal, each of Peter and Max said, okay, who's my closest friend who understands startups?

2:53Let's get them on the board. So for Max, it was Scott Bannister. For Peter, it was me. And so I was like at the very beginning when the two of them founded the company, joined the board, and then later joined the company as an executive. And part of the thing that they did masterfully was getting people who are just intense learners, intense risk takers and agents, and not interviewing of, have you had 20 years of experience previously doing this? But are you really, really thinking and working hard at doing this problem, being technical, etc.? And so, that's kind of the set of the culture. And there were just tons of things that I learned from Peter, from Elon, from Max, from David.

3:36There were things I learned from each of them. And I thought that kind of intensity of learning in terms of how we were doing things was simply amazing. So, it's more substance-based and experience-based, of course, what I've seen Peter is very focused on. Do you think a lot of them are still learning at a really high pace today? Do you think it's harder to do that when you've been doing these things for 20 years differently? I would generally say so. I mean, I think one of the things that makes great founders, entrepreneurs, and investors is kind of an infinite learning curve and a commitment to it.

4:04And so, you know, it's part of the reason why, like, for example, you know, the vast majority of us are all very focused on AI these days because of that same thing. And you have to be learning new things to do it. You can't say, oh, the tech patterns of 10 years ago, I'm just going to roll over and kind of start applying them now. That's lunacy. I mean, part of what's so exciting about what you and I and they do in technology is you have to reinvent the game every three to five years. And that reinvention of the game is that you have to have this learning cycle in order to do it. So before I jump into that, one more question on PayPal.

4:37So PayPal culture, you know, as big tech became ascendant, these cultures seem like became a little more monolithic, a little less intellectually diverse. Was there something about PayPal that would prevent that from happening? Or was it just like everything shifted a certain way over time? And do you actually think intellectual diversity is important in these cultures? Do you agree that they became too monolithic in some ways? Well, for sure, intellectual diversity is important, but it's what's most important in each of these cultures is that they're truth-seeking cultures and that truth-seeking part of learning what is their game.

5:08Now, each culture is different. So I do think that there's a coherent PayPal culture that's different than a Microsoft culture, that's different than a Google culture, that's different than an Apple culture. And each of these cultures are wildly different. And it's part of the, like there's these truisms, aphorisms that we have in Silicon Valley, like A players only hire A players and B players hire C players. And, you know, we use that as a heuristic, but it's not all A players. Like it's the A players at PayPal who would be massively successful there are not the same players as the A players at Apple who can be massively successful.

5:39So you actually get these areas of coherence and cultural identity. Sometimes you have characteristics that irritate people. like for example you know the like apple is very command and control i don't know how much you know about their culture but almost all the decisions are made by executives by long memos written their way all the way to the top which to you and me as entrepreneur people seems it drives us crazy yeah drives us crazy right so but there's a reason why the cultures are different and there's reasons why they match to different kinds of play and different kinds of excellence that get to scale ecosystems it feels like the paypal culture you're actually supposed to maybe argue and debate and have really strong opinions a lot more than that one.

6:19Yes. And I don't think you are at Apple, really. Yeah. It doesn't seem like it anyway. No, it's very interesting. So, what could possibly go right with our AI future? And we're all having to learn, again, everything's new in the last three or four years about what's possible. First of all, what are the best arguments for AI optimism? There's a lot of people who are afraid right now. What makes you so optimistic? So, part of the reason I wrote Super Agency is because as you and I both experienced the dialogue is like AI is coming for humanity, AI is coming for our jobs, AI is coming for our data, you know, and it's all this kind of fear and paranoia, which is a normal human response, but an irrational one.

6:54And so it was like, no, no, the way we get to our amazing future, which we have, by the way, the dialogue around AI is very similar to the dialogue around the printing press, you know, et cetera. It's like, no, no, we build it towards a better future and you don't get the futures you want by avoiding the futures you don't want. You have a vision you know what you're going for. Now, as you and I know, we are line of sight from a medical assistant that's 24-7 on every smartphone for every person who has a smartphone. That's huge, right? Lots of people do not have easy access to doctors, easy access to clinics, easy access to emergency rooms.

7:32Big, big deal. And even in countries where you have single payer systems, imagine when you had a medical assistant that was on the phone that said, oh, no, no, I'm going to jump the queue. You're going to be in there tomorrow morning. This is really important. Or no, no, this is fine. Let me help you with it. And you can check in six weeks later. Medical assistant, tutor, legal assistant. I mean, just all of these ways of elevating humanity. And it's part of the advice I give government officials because it's help people see the benefits they're going to get from AI so they're not just focused on what are clearly going to be a lot of job transformations.

8:07But it's like, that'll be hard, just like the industrial age, but also beneficial in the medium and long term, just like the industrial age. Let's talk about the healthcare one a little bit more, because I think this is, I do a lot in the policy world. And, you know, I think it's interesting because the way you talk about it is very similar to how Mark Andreessen talks about it, who's obviously he's gone really into the Trump administration and different politics. But I think we all agree this is really good for us if we can deliver better healthcare much cheaper. Healthcare is 20 % of our GDP, give or take right now.

8:36There's a lot of market failures in healthcare. You know, it's interesting if you're in a red state, the health systems are in control. And if you're in a blue state, the health systems are in control. So, these are very powerful interests. They'll be on whatever side they need to be on to lock out competition. So, it seems like it's going to be a huge battle. And it's as weird, it's going to cross party lines. It's going to be like the innovators and the people who want cheaper healthcare and better healthcare against this like really powerful special interest. You see it this way too? Like, what are we going to have to do to fight this battle and win?

9:03Unfortunately, all institutions, including large companies, tend to want to set up the system such that they are enshrined in power and control. That's a regulatory regime is going out of control. That's large companies that sometimes if they are too embedded with politics, you know, affecting regulation in order to do that, healthcare is the canonical one because there's all kinds of things. If you or I were given kind of like a carte blanche to say, here's things we could do to change drug discovery process. Here's things we could do to change the creation of an AI medical system. Here's things we could do.

9:33like for example you were talking about companies but plaintiffs attorneys associations here's things we could do powerful yes here's things we could do to change legal liability to enable innovation that helps americans american citizens there's just there's a long list in each of these things and it's really important to do and it's one of the things i'm you know frankly you know i'm always hopeful for can we can we kind of roll back for example medical liability in selective ways to allow for innovation, for allow for startup. This is dangerous for you because these are big donors to Democratic Party.

10:06Are you sure? I don't think they only donate to Democratic parties. They donate to both. But they've two more than left. I don't know. Maybe. They do. They do. But that's fair. But they're very powerful interests. And sometimes, and by the way, I think people should get sued and have to pay money if they're negligent, right? Of course. But they make things infinitely expensive by having it be ridiculous, right? And by the way, this is also like, for example, like, why is it so expensive for companies to go public? because we allow essentially wide-ranging plaintiff lawsuits against public companies.

10:35We actually want more companies being public. We want people participating with that. We want to allow plaintiff lawsuits but have fixed lanes for them. Another example of this, the head of the California Labor Union, I think, attacked me personally in some random tweet for supporting autonomous construction. And this is something I'm very optimistic on, actually lowering costs for everyone. It's just like obviously positive sum. But of course, we're going to have attacks like this. In this case, it's a core stronghold in California. What do we do against these? There's going to be coming from both sides.

11:05Do we have a responsibility to fight back on our own side when they're crazy like this? Well, actually, I mean, look, I try to fight against these kinds of... Look, I think it's a very good thing that there are institutions that are trying to advocate on behalf of, call it middle-class American workers, including unions. But unions can get it wrong. I'm frequently, on whatever side I'm on, fighting back against that, whether it's a autonomous vehicles, whether it's autonomous construction, because that is, in fact, the future. And an abundance, I'm sure you love it as well, is exactly right. And it is funny to sometimes talk to an occasional Democrat who goes, I'm for abundance.

11:41You understand that this is a change in regulatory posture, part of which there's a bunch of people on both parties who have incremented that regulatory structure. Well, they want to keep the regulation to help their own company. Yeah. So it's like, no, no, we have to factor back the regulatory structure in order to enable it. And like, for example, if we're not looking for automation, that's how we get GDP growth. That's how we get prosperity. Construction, awesome. Transport, awesome. You're involved in Aurora. Tell us about this. Atonement trucking, I think a big early investor. Is it starting to work?

12:14Like, what do you have to do there? Well, I mean, look, so Aurora, a company headquartered in California, but of course, you know, in modern regulatory environment, where are we driving our first, you know, heavy duty trucks? Texas. of course right and look it's it's it's it's we should be doing more of this like i yell at california people say we want this in california too because that's actually going to do that's going to unlock incredible amounts of productivity it's going to unlock basically like safety i mean there's just lots and lots of reasons so we've got the first heavy duty trucks working on routes in texas and it is awesome what's it going to take so this is interesting because i have a lot of friends who, even though they're on the left, when I have something innovative in policy, we'll usually get it done first in a red state and then bring it to the blue state.

13:02What's it going to take for the blue states to let you do these things there? Like, how's that going to happen? Well, I send everyone a copy of Abundance because, you know, Abundance, you know, while you could argue that it's written from a progressive standpoint is very clearly, there's a bunch of things that red states are doing better than blue states. And we should learn from them. We should do that. We should say housing prices and construction, exactly what you're mentioning. Regulatory, allowing things like, for example, autonomous vehicles to develop and get better and learn from it and deploy it.

13:32That's part of how you're part of the future. And so, you know, whenever I meet with a blue state governor, I point to these things. So, do you think there's an argument that people like you should maybe be backing pro-innovation abundance candidates on the left, if you're on the left? Is there a way to do that more? Are you doing that? Oh, I certainly do that. I mean, look, I'm fundamentally more of a centrist than anything else. So it's kind of like I dislike the extremes on both sides. And so, and you know, you know, you know me, just about everything I say and do is about innovation in technology and the economy.

14:02Let's talk also about AI, inflection AI. You built this with Mustafa Solomon. I think there's a big bet on personal intelligence. I think you got acquired by Microsoft. What were the lessons learned there? So we were right about chatbots. And so, but we got the timing a little wrong. We launched five months after ChatGPT. If we had launched just a little earlier, maybe it would have been better. And there's a limited number of startups that are going to be able to do the scale compute infrastructure that is required for frontier models, right? Not zero, right? Then we're going to open AI, obviously moving the scale, Anthropic, et cetera.

14:38But it's going to be a very limited number that's going to be doing the mega computers in order to do this. Zuckerberg is like, I saw this thing, it slides in Manhattan he's planning on building now. Have you seen this? It's like, no one's going to be able to build these. It's crazy. Yes. And by the way, that's still a good thing because that scale compute. I mean, I think people misunderstand a little bit of what the general frame of the AI revolution is, which is scale compute of learning systems that learn off mass amounts of data with scale teams doing them. And that's essentially the unlock that we've had with revolutionizing, adding intelligence to everything.

15:10Like we're going to add intelligence to these microphones. We're going to add intelligence to the lights. We're going to add intelligence to the cameras. There's going to be intelligence everywhere. It's going to be a really, really good thing. And some of that will be done through massive scale computers. Now, the reason why people say, well, is it only going to be the Microsofts, the Googles, the Open AIs, the Anthropics? Actually, I think everyone's going to win because I think startups can do applications. Startups can do distilled models. Startups can do things off open source models. I think there's going to be a bunch of different things.

15:37That's part of you're in my business. We invest in startups. but I think it's going to be, I think it's one of the mistakes we made with inflection was, oh, we're going to go after the scale compute side. And that's something you should be very careful about doing if you're a startup. That is very hard. I guess Elon did it right, but he had to raise a lot of money to do it. And I don't have any inside information, but I hear like a billion dollars a month is an expense line. Difficult as a startup. I was having this debate a little bit with Peter the other day. I don't want to mischaracterize substitution but here's i think if you use framework you'll be telling what you think so so so a lot of people are investing in ai uh they want to be in part of it i have like there's six levels level zero is energy and all the energy infrastructure level one is chips nvidia level two is data centers level three is like all the model companies right which is the meta and google and x is that a thropic level four i'd say is software infrastructure which is like horizontal to deploy ai uh and you can argue palantir is kind of level four and level five for me is actually the applications and services, like directly doing the services in the economy.

16:38And so as a venture fund, most of what I'm doing is level five right now. And I think that a little bit of level four, but I'm worried level three is going to eat a lot of level four. And that's the question. And I'm curious if that is similar to how you think about it. Well, I do think, I think it's generally speaking, most of our business should be focused on kind of level five with elements of level four out of this framework. But it's also that contrarian, but right, like if you find an interesting chip company, You might go, oh, I'm actually going to do this one. It could be something level one that's still worth doing.

17:08Yeah. So you can actually do level one. Or for example, it's, oh, actually, everyone's really focused on this narrow set of techniques, you know, mixture of experts with LLMs and, you know, neural networks or diffusion models. And if you find something that's different. Then way to deploy it. You know, then you might go, okay, I'm going to operate in different levels. We call it a level four or something to do infrastructure, this spiel miss. Because I have done a few level four things. Like we were in Olama early on. I don't know what it's worth. It's amazing. Everyone's using it. Yes, it's great.

17:35You're building a law investing a lot here. Tell us about Manus AI because that's an application layer company for this. Well, sort of. It does have application because the one thing about the framework is companies can actually have a combination of them. So like no, Manus is not going to go try to build a 700 billion or a trillion parameter model. But it might build a model still for what it's doing for drug discovery. Exactly. Right. So it's still going to do models, but it's going to be doing this kind of specific thing with specific data and using other models as a way of doing it. And that's part of the thing that is the reason why it's like, look, as you know, when the whole flood of everyone's moving this way, be very careful about moving where the whole flood's moving.

18:15Look at where the other interesting things are. And so part of, it was like three or four years ago, I went to my partner's gray lock and said, there's going to be a ton of coding agents. There's going to be productivity software. There's going to be, you know, kind of a robotic process automation. There's going to be all this stuff that's great. I'll help you on investment. on it. What I really want to look at is the areas that most people in Silicon Valley are not looking at. And I actually think drug discovery could be one of those areas. I talked to Sid Arnamukarji and he said, okay, cancer is the totally right focus here.

18:43And by the way, because I'd known Sid as the Pulitzer Prize winning author of Emperor All Maladies, a celebrated international oncology researcher. It was news to me when we were at dinner at a sushi restaurant in New York when he was like, you know, I brought cancer drugs to market. I was like, no, I didn't know that. Oh, let's talk about how this could work. And that's how Manus came about. Can you tell us any more about it? Like what's the progress or what are we going to do? Well, we've got some very good progress. We've hired an amazing CTO, although we haven't announced that yet. Right.

19:16And we've actually got even, we've made so much progress that we found some interesting drug candidates already, even though we haven't finished our tech stack for possibilities for things like triple negative breast cancer and other things. And so it looks very promising. I love it. I love it. We're doing it. We're doing a lot of AI applied to bio too. It seems like just one of like the purely good things for the world for us to understand this stuff and save lives. I guess you've been doing interviews lately with your AI twin. Can you tell us a little bit about this? What's that all about? Well, so, you know, similar, I think in the ways that we think about things, it's like every technology, there's a way that you can apply it well, as well as always applying it badly.

19:56And part of what we're trying to do as techno optimists say, how do we shape it? So it's applied well. So you have this technology that most people call deep fake and deep fake is kind of like, boy, that just sounds like it's bad. So I was like, all right, well, let's start experimenting with it and see how it can work in better ways. And literally like my staff was like, Like, oh, no, everyone's going to hate this. You know, it's going to seem like, you know, you're creating a deep fake, et cetera, et cetera. And I'm like, I am, but it's of myself and it's transparent. And I was like, well, let's just make it as a conversation.

20:26Like the very first one is I'll have a conversation with it. Like, well, we'll just do that and see where it goes. I mean, it's literally just raw experimentation. And you just gave it all of your writing, all of your everything. Yeah. So we did a custom chat GPT model for the read brain. We did 11 labs for The Voice. And then we're changing. It was our one initially. But as different models happen, we're doing different video. Do you give it your emails too? No. Too much. No, too much. Too private. Yes, exactly. You know, read AI should know my emails. It's not that I would necessarily be opposed to it as much as I didn't need to in order to get something that was good.

21:08And for example, here is the kind of thing that's very interesting as you start just experimenting. It's one of the reasons why as techno optimists to say, how do we do small experiments in order to learn how to steer to get in the future? And that's part of the thing that's great about innovation on startups. And it was like, actually, in fact, I gave last year a speech in Perugia for an honorary doctorate. And then I had Read AI give the same speech in languages, because I really only speak English, in Hindi, in Chinese, in Japanese, in Italian, and talk about human connection. It's like, well, here's a good use, right?

21:44And I'm sure there's many others. That's very interesting for your global PR is to have your own AI just read things in our language. I hadn't thought about that. Yeah, exactly. And my voice, it's really funny. You're listening to like, I've never heard my voice speaking Hindi. I've never heard my voice speaking Japanese. And it was like, wow, it is clearly my voice. I hope that's not too bad on the accent. I was told I had a Southern Indian accent in Hindi, but it's like, okay. We're going to have to try this from American Optimist to see if we can get some listeners in Hindi or something. Yes, exactly.