Brian Armstrong is the co-founder and CEO of Coinbase, the largest legal crypto exchange in America. In 2012, Brian started Coinbase out of Y Combinator after reading the Bitcoin whitepaper, then took it public on Nasdaq nine years later. Today, Coinbase is worth over $50 billion and serves over 100 million users. He’s since co-founded his second unicorn called NewLimit, a biotech company that’s extending human life. In this conversation, Brian makes a case for “disagreeable optimism”: that doing anything good in the world means being willing to be vilified, that AI regulation might be more dangerous than AI itself, and that an “American Wakanda” is possible – but only if America remembers how to build.
Watch on YouTube; Listen on Spotify or Apple Podcasts
Why “it’s not that great to be human”
Brian’s vision for 2035: merging with AI, exploring the galaxy, and the hive mind
Why Brian expects the future to be “more C-3PO than Terminator”
How to build the “American Wakanda”
Where moral values come from when you’re not religious
The case for an AI government
Why Brian thinks Gary Gensler tanked crypto for a shot at Treasury Secretary
00:00 Intro
01:32 Brian’s vision for the world in 2035
05:33 How to explain this future to a 9-year-old
08:58 Why Brian isn’t buying the AI doom scenarios
10:07 Why merging with AI might be the only way to explore the galaxy
11:56 The Gattaca stack: embryo editing, APOE4, and the hive mind
15:06 Why centralized control leads to worse outcomes
16:03 The case for leaving AI unregulated
17:48 Building parallel institutions: AI, crypto, prediction markets
19:11 How to build the “American Wakanda”
25:16 Lessons from Chinese org structures
27:58 What an “AI government” might actually look like
29:36 Why building anything good in the world means being vilified
32:55 Where Brian’s moral values actually came from
35:43 Why Brian thinks Gary Gensler tanked crypto for Treasury Secretary
38:40 Creating Ozempic for tribalism
39:19 The right dose of moral certainty
42:55 What people get wrong about tech founders
45:05 Conspiracy Corner: “It’s not that great to be human”
46:00 Bio stack or compute stack: which do you bet on?
Brian:
It’s actually not that great to be human, maybe that’s okay if the AIs take over and they do better than us.
Scott:
Brian Armstrong founded Coinbase. Now at NewLimit, he’s trying to help everybody live longer.
Brian:
It’s unlikely to me that our current biological life form is going to be the best suited to explore the galaxy.
Scott:
AI doomers think the world is ending but Brian isn’t buying it.
Brian:
I think it’s much more likely that we end up in a world where it looks more like C3PO than Terminator.
Scott:
His idea for keeping America on top sounds straight out of a Marvel movie.
Brian:
You could have a sandbox for innovation to really accelerate the future and try to make an American Xinjiang or Wakanda.
Scott:
For any founder trying to change the world, Brian has a warning.
Brian:
I had never had a thousand people angry at me at the same time.
Scott:
He also digs into why founders can’t stop building.
Brian:
I think we just find it fulfilling to try to improve the world, it literally is that simple.
Scott:
I’m Scott Phoenix and this is Progress. In 2020, employees everywhere were demanding their CEOs get political, Brian Armstrong told his team, “We’re not here to be activists,” and offered generous severance to anyone who disagreed, less than 5% walked and Brian became an icon for what it means to be mission first. Today, the company is worth over $50 billion and now Brian’s betting his next decade on a second unicorn, NewLimit, where he’s reprogramming ourselves to younger states so we live not just longer but healthier lives. This is Progress in the mind of Brian Armstrong, CEO of Coinbase and co-founder of NewLimit. Brian, it’s so good to have you on the podcast, thanks for making time.
Brian:
Thanks for having me.
Scott:
So, a couple things I wanted to go over today. First, I wanted to talk about what your vision of the future looks like. When you imagine 2035, what’s it like?
Brian:
I’m generally a believer that technology is the most important force shaping the future and there’s a bunch of important trends happening right now. Obviously, AI, crypto, I think where AI is going to intersect with humans in terms of brain machine interfaces, that’s a big one, we got to have more energy so we can keep building all these data centers and everything, probably get them plugged into humanoid robots and then I think there’s the whole space thing and then there’s the bio thing. So, on the bio side, I think longevity, embryo editing, generally just taking control of the bio stack and accelerating evolution is going to be a big deal. And so, I don’t know if 100% of those things are going to hit by 2035, I think Ray Kurzweil in his famous book The Singularity is Near, I think he predicted by 2029. This was back in the ‘90s, by the way, he made this prediction which is amazing in hindsight but now it’s looking more and more prescient. I think, by 2029, we’ll have a computer for $1,000 with the power of a human brain, that was his prediction. And so, I think generally we’re on track to that and what happens if that world emerges. Intelligence is going to become much more abundant and inexpensive, our brains will go from not just having an iPhone connected to it with our thumbs but, hopefully, a high bandwidth connection accessing the sum of all human knowledge, maybe connecting into more hive minds. I think eventually people are going to have some sort of brain upload that happens. If more and more of your thinking is happening in the cloud and you happen to have one human body and you have some robot bodies, eventually, if the human body doesn’t work as well or perishes, hopefully we’ll get that solved with longevity too. But if it does go away, more and more people will be ... I guess they won’t be people, the definition will have to expand of what are these conscious beings, some hybrid of AI and human and they’ll be living with real world human bodies with robots and they’ll be in the cloud and, hopefully, it’s someday exploring the galaxy. So, I think you used this phrase one time, you said how do we achieve a positive singularity and I like that framing of it a lot so I’m very aligned to that mission.
Scott:
I love the articulation of what 2035 can be like and I also feel like, if I did a random survey of people I meet on the street, I’d be like, “What’s 2035 like?” the answers I would get would be like, “Well, the phones are smaller, it’s more like her where you have an AI friend,” or it would be a much less expansive vision of what it could be and it’s rooted in what we have now plus a little more Waymo or something. How do you bridge that gap, you know what I mean? People are experiencing what I think is one of the most remarkable periods in human history where we’re going from no AI to AI everywhere and that’s a really big change, you’ve clearly wrapped your head around it. How does a normal person come to grips with the kind of future that you described?
Brian:
I think you’re right, a lot of people are still grappling with issues that are, I don’t know, they’re a little bit more ... They feel super important to them, let’s say rising political violence or maybe tension between the US and China. But if you start to extrapolate this out a few steps ahead and we’re in a world where you can start to explore other planets with your AI robot body or something, the competition between the US and China may feel a little less important. But you’re right, people today are probably wondering is that cab driver or truck driver going to have a job and the changes are going to be so much more profound than that that it will seem so ... It’s going to be like why are you and I not farming a field in early 1900s, it’s going to seem like a very quaint and parochial thing to think about in 10 years is my guess.
Scott:
I think your callback to the industrial revolution is a really good one in so far as for context we went from I think it was something like 65, 70% of people working on a farm in 1900 to less than 5% now in not that long of a period of time. And lots of new jobs were created in factories and cities and stuff but I feel like this change that’s happening now is a lot faster and less comprehendible to the average person. I have kids who are seven and nine and explaining to them like, “Oh, here’s the kind of future you might get to live in and here’s how to relate to that,” I feel like I haven’t found a good way to describe it because it all sounds so different from the life they’re used to. How do you think about that?
Brian:
Well, I feel like in every generation there is some technology trend that happens and the people who are middle-aged or a little older have a freak-out about it. When I was young, I remember my parents their big fear was like, “Okay, you’re spending a lot of your time playing violent video games and you’re watching the TV,” and in their world that was super scary and they thought maybe these violent video games were creating murderers or who knows what. And if you go back and read history, believe it or not, the bicycle when it came out in America was very controversial-
Scott:
Books, books are super controversial.
Brian:
Were they? [inaudible 00:06:48]
Scott:
Yeah. People are reading all the time and it’s a disaster.
Brian:
Well, the printing press was highly ... That was the internet of its day. And for bicycles, people thought like, oh, your children are going to be all over town, you don’t know where they are engaging in licentious behavior or whatever, moral panic at the time. So, I think that we have to have a little bit of memory of that and now when ... Today, it’s probably social media or whatever your kids, it doesn’t mean there’s not some element of truth to it, maybe you should be a little cautious of what young kids are doing or whatever on the internet but I don’t think it requires a moral panic. People are very adaptable and maybe I’ll give you an even further out thought which is that let’s just say progress continues to accelerate even faster and faster and faster. I think that our current wirings of our brain and our biology are pretty adaptable, humans have gone through all kinds of crazy things, war and pandemic and natural disasters so we can be pretty resilient and adaptable. But I think in the future we’re actually going to be able to take control of evolution biologically in the sense that we’re not just doing evolution by natural selection which only optimizes for can you survive long enough to replicate yourself and then it doesn’t care what happens to you after that, whether you’re in pain or you’re addicted or whatever. I think that, as we start to really understand biology better in the coming decades, it’s going to take some time with the assistance of AI augmenting our brains, we’ll be able to upgrade our thinking as well to be even more resilient to change. And there could be like entirely different species or AI battles, it’s hard to imagine what’s going to happen in the future, I think that it’ll be okay, I think that ... We might not be the top predator on the Savannah anymore, the apex predator if these AIs surpass us, we can talk about that too, there’s a merge opportunity and in some ways we have to accept that.
Scott:
I love your comment of moral panics, it reminded me of this story I’d forgotten from my own childhood where I got a scanner for Christmas and I was so excited and my mom, after I opened it up and plugged in the computer and stuff, came up to my room where the computer was set up and she was like, “Scott, I just want to make sure this isn’t a temptation for you.” And I’m like, “What? It’s a scanner. What am I going to do download licentious pictures, print them and then scan them back in?” It was just such a gap between her understanding what the technology could do and the actual risks and so I think it’s a great frame of now maybe a lot of us who are our age are looking at this new technology like my mom looked at the scanner and being like, “It’s amorphous and scary if you don’t more deeply understand how it works.”
Brian:
You want to make sure maybe this thing doesn’t jump out of its sandbox and start to evolve too quickly but ... You can probably tell from my comments, I’m a little skeptical of the real doomer scenarios around AI. Look, I think there is some plausible scenario you could imagine where AI gets into this self-recursive improving loop and we give it control of certain parts of government and it has human in the loop to start but eventually you say, “No, actually it’s making better decisions without the human.” And at some point, if this thing supersedes us, it’s not really going to care much about our concerns and it may actually step on us accidentally or on purpose to try to further its own motives. I can’t really tell you I have some perfect way to debunk that entire philosophy, it could happen, it’s possible, it’s just my hunch is that it’s not going to be that simple. I think it’s much more likely that we end up in a world where it looks more like C3PO than Terminator, we’re going to have humanoid robots that are highly intelligent, they’re going to ... Humans or some augmented humans are probably going to still be a major player in this game and eventually we’re going to have really strong AIs that are also have their own ambitions out there in the world. I guess maybe my most contrarian thought on this is that maybe that’s okay if the AIs take over and they do better than us, I think that we’re going to have a path to merge our consciousness with AI and maybe supersede that. But if I’m imagining us 100 years from now and 1,000 years from now and what happened on planet earth is running out across the galaxy, it’s unlikely to me that our current biological life form is going to be the best suited to explore the galaxy. Our DNA gets damaged outside the ozone layer and we have to carry this little bubble of water and oxygen around with us. So, in some ways I’m okay with the idea of AI superseding humanity in a way, we should try to merge with that and evolve ourselves but we should also be proud to be the birth of that if it really goes that far.
Scott:
Just to say it back to you, maybe it’s something you’re saying is, if we make the super intelligent AI and we go through what you described as the gradual disempowerment scenario where we delegate more and more to it and then later it decides, hey, I’m in charge now and I want to propagate machine intelligence through world and the humans have been in my way, I’m going to turn off their electricity so that I can use electricity for better purposes which are my purposes. We’re in a world where it’s a post-human world in the worst case of this. Your view is obviously not that’s a good outcome, right?
Brian:
Well, I don’t think it’ll be that simple of just they turn off electricity for all humanity, I think that ... We already exist in a world today where there’s people that are smarter than me, there’s people that are more powerful than me, there’s people that are implementing policies that are against some of my interests and I go out there and I try to advocate for things I think are better and there’s certain parts of the world that might be more aligned to me or not. And so, I think it’ll be complicated, I don’t think it’ll be that simple where it’s just AI wakes up and decides to step on us to get more energy or something. I think it’ll be more complex and we’ll partly be AI ourselves. It’s basically like what would be the Gattaca stack, if you’re familiar with that movie.
Scott:
Yeah, yeah.
Brian:
Yeah. And it’s, okay, in vitro gametogenesis allows you to take skin cells, turn them into eggs so women don’t have to go through this invasive procedure with IVF. You could have, not just selecting the best embryos which is called pre-implementation genetic testing, but actually starting to make some edits on the best ones to what if you could get ... APOE4 is a single base pair edit that is protective of Alzheimer’s, these would be amazing. And then artificial wombs and you go on down the stack and that could get us to a place where actually more and more people in the world eventually are born with better ... Freedom from these genetic diseases and it blurs the line into enhancement and you could have much longer longevity because of that and other benefits.
Scott:
One of the things I thought was surprising when I was reading about this was that a lot of ... You might think that it’s a trade-off where you either get to be an Olympic swimmer or a brilliant mathematician but, actually, the positive traits are correlated with each other. And so, making an improvement in one area where you are treating a vision degradation condition or something actually also tends to reduce heart disease risk and other types of problems and so you actually end up just moving strictly in a positive direction.
Brian:
Yeah, totally. I think that’s a counterintuitive thing most people don’t realize, they assume everything’s a trade-off. And there’s a couple things that are trade-offs but-
Scott:
In society too, it’s like a microcosm.
Brian:
Yeah, yeah.
Scott:
In the body so is in the-
Brian:
It’s not zero sum.
Scott:
But something about the positive sum nature of emerge takes us into a weird place that we talked about earlier which is, if you imagine that you and I can share thoughts or share memories and we do an economic exchange where I offer you some memories, you offer me some memories, we both get richer. And when you play that game over multiple iterations, everyone has everyone’s memories and we end at a place where you’re walking around with most of humanity, most of the internet, most of everyone’s experiences in your head, how much do you still feel like you’re you?
Brian:
So, we’re all standing on the shoulders of giants and we’re benefiting from, whatever, 100,000 years of evolution from stone tools and fire to now we can read about quantum computing or something in our textbook in college but our brains are getting increasingly more connected like a hive mind. The internet was a huge step in that direction, brain machine interfaces will accelerate that and then what would it look like to neuro link your brain more permanently with your partner, your spouse or a whole bunch of-
Scott:
Or your AI, yeah.
Brian:
Yeah. We will start to feel a little bit more disconnected from our bodies in a sense you might actually have, whatever, 100 bodies, different things running around. And so, in many ways, our body is like this meat puppet for our brain and one of our friends loves to talk about should we just be a brain in a vat. It would really suck if our pancreas took us out or something like that. It is interesting, the more and more humanity connects its brain and AI connects, we’re becoming like a hive mind, it’s almost like more at a planetary scale we are some loosely connected big organism. Is the ant colony an organism or is it the individual ant? The individual ant can’t live by itself so it’s some hybrid like that.
Scott:
Brian isn’t worried about whether that future is coming, he’s more worried about whether America still knows how to build it. The vision you gave at the beginning was awesome, I want to live in that vision, that sounds great. How do we make sure that that’s the most likely version of the future versus some other of the darker ones?
Brian:
I think I may have more faith in humanity than maybe the average person because, if you just look historically, the people who go and try to build a lot of these things, they have good intentions and that doesn’t always mean that they know exactly what the outcomes are going to be but they have good intentions. If people have bad intentions, society tends to tamp them down, they lose reputation, they lose resources to go pursue that. Society our tribe does not really reward people that they feel like have bad motives and so I’m more okay with the free market of ideas, lots of people trying different things to improve the worlds and then our society will get some sense of, overall, are they doing better or worse. People who are doing better will accumulate more resources, more power, the people who aren’t will not. I think the alternative is often to sit there with some high power in the government or whatever and say, “We’re going to dictate on high that you cannot do this and you can do this and you cannot do this and here’s all the ...” I don’t believe in that centralized control, I think it leads to worst outcomes and so I’m more interested in the free market of ideas will allow the best ideas to come to the top and, to the extent that there’s bad people doing something, there will be a product or service by a good person to combat it over here. So, that tends to have worked well historically. My bias in these situations is, if a technology is new and we don’t quite understand where it’s going, leave it unregulated until we start to see material harm because, otherwise, you’re trying to prevent a harm which is it’s amorphous, it may never happen, it could be a boogeyman and I think that a lot of harm has been done. Part of the reason you think about why was the internet the biggest growth area of the economy over the last couple of decades, it’s largely because it was mostly unregulated and we saw high growth rate there. If you look at education or healthcare or some of the most regulated industries, they had slower growth, their prices went up. Think about the cell phone you have now or the TV you have versus 20 years ago versus the education or the healthcare you’re getting. Look, I’m not saying there should be zero regulation but I think, in general, we in society bias toward too much. The best regulation I’ve seen actually is choice, it’s competition in a market because I think people don’t give customers enough credit. If you have a couple of different products that they can choose from and relatively low switching costs, that can do far more good than a government bureaucrat trying to choose what’s best for everybody. Humans can understand that product, I don’t like it for some reason, maybe the societal impact that it’s having or the risk that it’s taking and I’m going to choose a different one, I want to vote with my dollars and cents and that tends to be a good regulator as well.
Scott:
What you’re describing are how do we construct parallel institutions that open a door that we didn’t realize was there. And prediction markets is one, crypto as an economic system is an alternative parallel institution that you’re creating. AI certainly feels like a parallel institution now where we can create digital people and they can have wallets and can transact and participate in exchange. And as you imagine the playing forward of the creation of parallel institutions, what are the ones that get made? What are the ones that become irrelevant? What do you think happens?
Brian:
Yeah. Well, okay, so I actually do think it’s really important to create new institutions. A lot of people want to try to reform the existing institutions whether it’s the universities or the traditional media companies. I think that it’s okay for some people to try to do that but you’re not truly going to get reform unless you allow new ones to come in and compete with them. It’s would you have rather tried to reform Microsoft or do you want Google to come in, it’s voice versus exit our friend biology might say, I think that that’s very important. And actually, maybe one of the biggest ones would be what is the alternative to the United States of America if it falls? Because it’s currently the bastion of freedom and rule of law where you can actually go build things and it’s under threat, we’re seeing Mamdami and New York and AOC in Gavin and Elizabeth Warren are trying to do a whole bunch of socialist things or whatever.
Scott:
Or other people would say Trump and their view of the attack of rule of law and there’s a ongoing ... The American debate, the American experiment is certainly in the most fragile state it’s been, it seems like, in a really long time.
Brian:
Agreed. And to be clear ... And I really want that American experiment to continue and I think it has a good shot at doing that especially if we can get deficit spending under control and all kinds of things, rule of law, whatever. But how do we build a parallel institution, how do we build a backup plan? And I think special economic zones are an interesting topic that’s underexplored. There actually are large ones, Shenzhen is a ... Dubai, Singapore, these are all arguably special economic zones that have unlocked a lot of value and I would love to see some of those exist in the US. There’s been some proposals for this, I think people calling them acceleration zones or whatever but they might be pieces of federal land in the US with congressional approval or you could do it through an interstate compact or maybe even a Native American reservation, there’s ways. The goal would be to get outside of federal law and have a sandbox for innovation within these certain acceleration zones and there you could do clinical trials, you could do nuclear energy, you could do space, you could do crypto, you could have a sandbox for innovation to really accelerate the future and try to make an American Xinjiang or Wakanda or some very inspiring vision of the future.
Scott:
Love it, the American Wakanda.
Brian:
Yeah.
Scott:
Well, I also ... So, another strategy you didn’t list that I thought was surprising or interesting was don’t reform the FDA or maybe try but you should also create the FDB and maybe the FDC and the FDD and have these as parallel paths that compete against each other for some objective function or whatever. The way we set up our institutions now is we give all the monopoly on the thing they regulate. Rather than fighting the battle one at a time to try and create alternative regulatory agencies and it’s a turf war, the FDA people don’t want there to be an FDB, is there some higher level action that could be taken that set a framework in place for all of the agencies by necessity having some market of ideas or some set of competing principles where alternative hypotheses for efficiency can be tested?
Brian:
In many ways, this is what the founding fathers contemplated with federalism. We have 50 states, arguably, the United States is 50 special economic zones. The problem though is that, while the states have some power, I don’t think it goes quite far enough. If you want to build a supersonic aircraft, you can’t do it because the FAA blocks it at least until recently, there was an executive order. Or if you want to build a nuclear power plant in that state, you actually still need the approval of the Nuclear Regulatory Commission or the FDA. So, these federal regulators, it gives you a benefit in one sense which is you have one regulator for all 50 states, you don’t have to get 50 licenses but, if they’re blocking innovation at the federal level, then we can’t go further enough. So, anyway, I hope that we get a chance to see federal land that could be designated as these special zones that would actually be exempt from federal law and, within that zone, you could try anything in that sandbox and people can opt into it. If you want to live there-
Scott:
Live there, yeah.
Brian:
... and there’s sonic booms happening, that’s up to you, you don’t have to live there.
Scott:
I like that idea a lot. Okay. So, when you think about the manifestation of those special economic zones or the manifestation of a backup plan for America, does it even need to start with land? Biology is really into network states and that is a future direction, do you think that ... You didn’t articulate it as part of your 2035 which I thought was interesting. Is it because it’s low key or is it something that you think is a natural evolution? Is land really the central organizing principle of human tribes and it’s hard to break out of that? How do you think about that?
Brian:
Biology would probably agree. It’s you could start in the cloud but you’re going to have to get into land at some point. In a very practical sense, where are you going to put the data centers?
Scott:
Yeah.
Brian:
And so, I do think it’s important to get land. Mars is another option here, that’s a little ways off but I think-
Scott:
It’s hard to live there.
Brian:
Yeah, for now but I think it will get ... Yeah, the more likely path is to do something in the US that would be under a ... At least you’d have some sense of rule of law. People like Prospera are doing a special economic zone in Honduras which I think is a great V1 prototype.
Scott:
Did they make it through ... There was some legal issues where perhaps it was going to get taken away, did that end up being resolved?
Brian:
It’s in a decent place, I don’t know if I’d call it resolved but my understanding is that ... Yeah. So, the Constitution in Honduras actually did allow for these zones. As it sometimes happens in these South American countries, there was a little bit of a I’ll call it a banana republic or kangaroo court type thing that happened where the Supreme Court ruled that that law and the Constitution didn’t apply and somehow that doesn’t really make sense because they can’t overrule the Constitution. It just had to do with the current government.
Scott:
The Constitution’s unconstitutional.
Brian:
Yeah, yeah.
Scott:
Yeah.
Brian:
And so, anyway, that current socialist government that got elected was trying to tamp down on these zones but it was more words than action and the zone still exists and it’s still there and I’m optimistic on it. But let’s just say that I would consider Honduras to be a nice V1 prototype but I’d feel a lot better if there were these zones in the US.
Scott:
In a lot of ways, right now America has an autoimmune condition or maybe humanity does where we’ve started to recognize the self as the other and so, instead of attacking problems, we attack other cells that are differentiated and are performing a vital function. How do you treat the autoimmune conditions so that we can produce a more ... Forget aligning AI, how do we align each other?
Brian:
One contrarian thought that I’ve had is that I actually think people are really worried about competition with China and it might actually be good in a sense that America was leading the world for so long we got a little complacent, right?
Scott:
Yeah.
Brian:
And you don’t want complacency but you also don’t want conflict that’s violent and kinetic but competition is good, competition breeds excellence. So, anyway, I hope that, in this competition with China, it actually gets the US to get back to growth, back to excellence, back to building, believing we can manufacture things like having more kids, whatever it is. So, I hope it continues on but there should be that kind of competition and pressure all throughout the micro level, the macro level, maybe someday with other species, we’ll see.
Scott:
So, this is a big difference I’ve noticed in some Chinese companies versus American companies [inaudible 00:25:45] Coinbase is like. I’ve seen the org structure of Chinese companies where, literally, it’s two competing divisions working on identical things and, if this one builds it first, everyone’s fired on this one, it’s a extreme cutthroat marketplace of ideas. Where is Coinbase is an organization between socialism and a free market internally?
Brian:
I actually think there’s some merit to that of having internal competition. Within Coinbase, we have the main retail app, most people know it’s a custodial wallet, we also have a self-custodial wallet and some of the features between them are similar. You can do trading on both, you can do payments but they have different stacks underneath about the changes of all the kinds of things you can offer in the product. It’s actually great because sometimes I’ll be pushing on something in one of the apps and I’m like, “Man, why isn’t it easier to do X and Y and I get these” ... And then the other team will show me a new feature, I’m like, “That’s cool, that’s better.” And then I’m like, “Okay, why don’t you guys copy it better?” and then the next month it’ll happen the other way around.
Scott:
It’s interesting because I remember Peter Thiel’s philosophy with PayPal was everyone had exactly one job and, if it wasn’t your job, he wouldn’t even speak to you about it which is the most, I don’t know if it’s socialist exactly, but the most not free market oriented people can try things version of it. And I think paradoxically Google with its 20% time and they have five different versions of every product and library and everything but it’s remarkably bureaucratic. And so, how do you resolve this tension between the organizations that seem like they should be capitalist actually becoming socialist bureaucracies in so far as they allow for competition but somehow they all turn very bureaucratic and then organizations that are, like you described, very incisive, here’s what we’re doing, there’s some competition but it’s extremely clear zones of fire?
Brian:
Between employees?
Scott:
Being the most efficient ones. That seems like it should be an oxymoron.
Brian:
Okay. I think it’s okay to have top-down dictatorship in a company from a CEO as long as-
Scott:
How convenient [inaudible 00:27:45].
Brian:
Yeah. As long as there ... This is the part that contradict, as long as there’s a mechanism to clean out the ones that aren’t working. So, that means you’re competing with a bunch of other CEOs of other companies and your results are getting compared to the product or the revenue, you will get cleaned out by the market. So, what the problem with government sometimes why you don’t want to have a totally autocratic government, you want more freedom is that, because of geographical constraints, et cetera, what’s going to clean that out. If you have no term limits and dictator for life and they happen to be great, then you’re good. Happen to be bad and there’s no way to clean it out, then you’re in deep, deep trouble.
Scott:
I had this conversation with Ashton Kutcher which he said such a great metaphor for this which is that we’re headed towards a Cyrano de Bergerac society. Do you know about this?
Brian:
No.
Scott:
I didn’t know about it either. It’s a famous play about this extremely smart and totally ugly guy who writes love letters for his really handsome but extremely dumb friend. And so, eventually, they realize that it’s him writing letters and not a dumb friend but Ashton’s observation was we’re headed towards this future where anytime I’m like, “Oh, Brian, what do you think of this?” You’re like, “Hang on a second, doop, doop, doop, ‘ChatGPT, what do you think of this?’ ‘Oh, Scott, here’s my thoughts.’” We’re just the middlemen, basically, for the messages that go back and forth and I actually think, for government, that’s a really good thing. The more we can have our bureaucrats delegating their decision authority or their evaluations of policy choices or their applications of them to something that is unbiased and in possession of all the world’s knowledge and has an IQ of 300, I think probably the better off we’ll be.
Brian:
You’re thinking an AI government.
Scott:
I wouldn’t call it an AI government, I would call it AI helpers for the government. Maybe it converges towards the AI government someday, people are thin rappers around it.
Brian:
I think it’ll start by just, yeah, consulting AI on policy and have a human in the loop and, who knows, 50 years or less, maybe if there’s certain-
Scott:
Fifty years? We’re all uploading our brains and flying around spaceships by then.
Brian:
Hopefully. I think that there could be situations where you actually just fully delegate it to AI and it works out better.
Scott:
I asked Brian what drives his optimism and his answer came with a few scars.
Brian:
There were a few things maybe early on that I read that caught my interest in this area but it wasn’t until I started actually trying to build things in the real world that I may have gone even further on this spectrum. Because I think, as a young person, I hadn’t really built anything and I’d read some Ayn Rand and some Milton Friedman and different ... Karl Marx or whatever, I’d read all these things but I remember reading things in the press and I’d say, “Oh, this CEO did something bad and this CEO did something good,” whatever. And I would always think, well, man, if I ever build something, I’m just never going to do what that person did. I didn’t realize until you go out and start to actually try to build something in the world that it’s more complicated than that. Some people will attack you just for trying to build something good in the world because they’re playing some other status game, some journalists will write articles about what you’re doing that have no basis in reality, they don’t even really understand what you’re doing and they’ll attack it for some reason or they’re upset about it. Actually, I remember there was a friend I had that was doing something which I thought was so unequivocally ... It was untouchable, unimpeachable, they were basically trying to create this charity to give money to homeless people in San Francisco. I remember I started to see these news articles come out attacking this charity and they were like, “How evil it is, they’re trying to displace these people out of the city.” I was like, “Really? You can do something.” It’s literally the most unimpeachably good thing I could think of is giving help to homeless people or something and they started getting hit pieces. And I realized, okay, there’s something much more complicated here happening, some people are just going to be upset if you try to improve the world. It’s violating their narrative about what should be allowed, maybe they have their own inner issues going on about comparing themself to you, they don’t like it or they’re trying to sound smart and raise their status by tearing you down but you actually ... What I realized is you have to be highly disagreeable and willing to be vilified to build anything interesting in the world. This was an underappreciated skillset that I did not anticipate was necessary to build a company, I thought people would cheer me on. Whether they thought it would work or not or they even liked my product or not, I thought, okay, he’s trying to create jobs and trying to make the future better whether he succeeds or not, we’ll see but it turned out not to be the case, many people will hate you for even trying to do that. I still think-
Scott:
Is that hard?
Brian:
Sure. There was times where ... I had never had a thousand people angry at me at the same time. And on the internet, things can run wild before the truth gets a chance to put its shoes on or whatever and, yeah, there was times where it initially would really bother me. And I’d say, every time it happened, it got a little easier and I started to develop more thick skin and ... I don’t know, now there might be a dozen articles a week, I don’t read any of them so I don’t even really know. But yeah, it’s important to cultivate, I think, that mindset of you also don’t want to be totally closed off from input. So, if there’s people around you who have the best interests at heart and they’re telling you, hey, we should go in a different direction, you have to be willing to hear that. Otherwise, you can become isolated like Putin or something. But if people don’t have your best interest at heart, I actually think it’s very valuable to be willing to completely ignore them and don’t let them rob your mental energy during the day. Don’t allow them to put some angry text message on your phone, don’t allow them to ... Don’t read any of the articles that they write, I don’t even read any of them now so don’t ... Never read the comments on Hacker News or whatever. So, that’s an important thing I didn’t appreciate getting into building companies.
Scott:
One of the things that I’ve always most deeply admired about you is the degree of moral centeredness that you have. So, to be in what is an opium den industry full of hucksters who are all lying and trying to get rich and ... To have integrity in normal business is something that many people fail to achieve. And to have integrity in a business where all of the incentives and all of the peers around you are refusing to play by the rules and are engaging in activities that are morally questionable and holding true to your principles is something that I really admire about how you built your company. Do you feel like there’s a shortage of integrity or moral fortitude in society? Is that we’re just middle-aged people now grumbling about kids these days or, if only politicians are noble, the world will be better? How do you relate to being a couple standard deviations above in the way you relate to the question of integrity?
Brian:
Well, it’s an interesting question, yeah, and thank you for saying that. It boils down to where do values come from. I’m not really a particularly religious person but I did grow up around a lot of Judeo-Christian values, actually, and it’s making me question now later in life to what extent that actually is almost an underpinning of a lot of my mental frameworks more than I ever appreciated. I went to a Jesuit high school and they had this motto about be a man for others. And, in high school, I never really took it that seriously because I thought it was just some box I had to check to get through all the thing and go do some volunteer work or something. But in a way, it may have rubbed off on me in a sense that this is what ... The people who are trying to get rich on something don’t realize is that it’s a lower rung of Maslow’s hierarchy, it’s actually very unfulfilling and, if you destroy your reputation in the process or maybe friends or people around you that you know, you harm them in some way, you’ve actually destroyed your whole life. So, even if you want to look at it totally selfishly, how do I have the best life, if you do something that gets you money and it hurts your reputation and the people around you, you’ve actually hurt yourself far more than I think any people really realize. But if you’re going to take responsibility and try to build something meaningful in the world, accidents can happen too. We’ve had a cybersecurity event where some customer information got stolen, imagine there’s much worse examples. Imagine you’re trying to build a space program or a self-driving car, people could really die or at a biotech company, clinical trials. So, it takes a lot of responsibility and carefulness to go into this world and be okay with the outcome but people have to know it’s coming from a good place and they might be very upset with you when things don’t go the way that they planned.
Scott:
One of the question I had for you about that was ... And I didn’t get into it because I was worried it’d be too hot but is your view that Gary Gensler has deep seated moral beliefs and a north star about a society he wishes to create or that he is a power seeking social climber who will say anything and do anything to get the next rung in his career?
Brian:
It’s a very interesting question about why did Gary Gensler do what he did. He really tried to unlawfully kill an entire industry in the United States, why did he do that? He’s not a dumb person, I actually think he’s pretty smart. If you go back and watch ... Before he was SEC chair, he was teaching these classes at MIT about crypto and you could tell he understands the technology and the potential of it. He says a lot of things in those videos like, “This has enormous potential to update the financial system or something to that effect,” and then suddenly he gets into the SEC role and he’s trying to kill the whole thing. So, it was ultimately about personal ambition and trying to get the next job. It was widely known in DC that he wanted to really be treasury secretary, Elizabeth Warren made this deal with him. When she stepped down and endorsed Biden, he allowed her to nominate, I think, a lot of the financial service regulators and what I was told is that she said, “If you take this SEC chair job which I know you don’t want and go hard against crypto, I’ll basically try my best to help you get treasury secretary job next.” That’s the only reason I’ve ever heard that’s plausible, I don’t obviously know but he seemed to have the, let’s call it, intellectual flexibility or maybe lack of integrity, whatever you want to call it, to be totally willing to kill something which he intellectually knew had a lot of potential in service of trying to advance his career or power or something like that. So, it’s closer to the latter thing that you said but, yeah, it’s frustrating. I think for Elizabeth Warren my guess is that it’s more of a philosophical thing that she believes, she’s a socialist and believes the government should run financial services and that crypto is actually evil, for Gensler I think it was just expedient for his next career move in the organization. He may have somehow talked himself into this idea that it was all bad actors and I have to clean it up. It’s actually an interesting psychology question of, if your interests are aligned for this thing to be true, do you genuinely believe it’s true or do you intellectually know it’s not true but you’re doing it anyway. And I think people’s brains are complicated, you can actually ... It’s very hard to convince somebody of something that’s against their economic interests, our brains are mixed up in terms of what we can convince ourselves of.
Scott:
Or even against their tribal interests. I remember reading this study about they would do interviews with people on what they thought of a policy and, if the person was Republican, they’d be like, “Oh, what do you think of this policy?” and they’d give their honest answer. Another Republican they’d be like, “What do you think of the Trump policy two?” Immediately they’re like, “Oh, this is great, this is just an awesome policy.” And the same thing if they flipped it to like, “What do you think of the Biden policy two?” And so, there’s something about tribal ... The root password to people’s brains is tribe or something.
Brian:
I agree. Tribalism is deeply ingrained into human psychology, it’s probably the root of all ... It’s like sports teams, countries, rivalries, racism, arguably, all kinds of things and it’s pretty unhelpful. If we ever really manage to understand how genetics work and our brains work, imagine if we could actually edit out that level of tribalism.
Scott:
Ozempic for tribalism would be-
Brian:
Yeah.
Scott:
... the most amazing drug. Put it in the water supply.
Brian:
We’d be much more objective about reality. I would love to not have my brain play tricks on me and like, “Oh, because I grew up in America, I’m more sympathetic, I’d rather just look at it in an objective way.”
Scott:
One thing about moral certainty I was thinking about as we were talking is moral certainty in the right dose is great medicine but, in too high of a dose, it’s poison. A lot of what we see today at the extremes of both ends of the political spectrum is too much moral certainty and that the knob’s so high that they view anyone who disagrees as the enemy or Al-Qaeda is a mission-first organization, you know what I mean? How do people, how does society have the discernment to know the right dose of this potent drug?
Brian:
I’ve wondered this myself because I do think that there’s people who ... Let’s take George Soros as an example, made a lot of money, wants to do a lot of philanthropy, a lot of this ... I think, by the way, it starts as a tax avoidance thing or tax minimization thing which would be, if you put a lot of this money into a foundation, you’re like, “I don’t have to pay it to the government.” And then you’re like, “Okay, what’s my foundation going to do?” It’s they get the order of ... And I see this with, actually, a lot of wealthy people where ... So, they start thinking, okay, what are we going to do with this money and it’s let’s push a bunch of socialism or something like that or defund the police and it can have all this accidental harm. And I think there’s a surprising number of people who try to do philanthropy and they learn later in life, or maybe they don’t learn, but they had enormous harm by trying to do something good. So, I don’t know if I’ve figured that out. I haven’t done a lot in philanthropy yet, I’m a little bit approaching it with caution and an open mind. And if I take the mission of Coinbase around increasing economic freedom in the world, look, I think that’s a good thing, I’ve read the research, I’ve had a lot of personal experiences, I feel it viscerally, obviously I’ve committed a lot of my time and life to it. Do I know for sure all the economic policies in the world, how they would create a better society? I think it would be intellectually dishonest for me to say I know for sure. I can have a thesis and I think that these things can be tested in the real world like we had North Korea and South Korea, that was an AB test or East and West Germany and we saw those outcomes. Ultimately, I think everyone is just running around trying to do what they can to improve the world in the ideal case and we’re not always going to get it right but the best ideas will rise to the top and evidence will emerge.
Scott:
What you just described is discourse and intellectual humility and I hear little of that when I turn on the TV, you know what I mean, or look at what’s going on and polarization and dehumanization and a divorce from evidence-based decision making. Is there a solution here?
Brian:
Just from a comms point of view and the war of ideas, you’re right, it doesn’t really help to shout out something that’s super intellectually honest like, “Hey, this could be true and it might be a thesis worth testing.” People want to hear like, “Death to this,” or they want to hear a similar extreme statement. I think our friend Ryan Peterson had a funny statement and he said, “Only a half truth can go viral.” You have to make it the most extreme version for it to really go viral so there’s that war of ideas on social media. I think it is an interesting topic, a lot of people today are trying to figure out how do you even know what’s true. Because traditional social media or traditional media, I should say, got very politicized and they’re essentially political advocacy organizations, not doing much real journalism in my view. Social media has its own problems too with the war of ideas and there’s people trying to do community notes and all that. I think people are now trying to do prediction markets. Crypto is trying to just create the right game theory and tokenomics that will allow these things to resolve in a truthful way and there’s slashing penalties if you try to manipulate the market on something that wasn’t actually true. I think that’s somewhere probably interesting on the frontier to play but I don’t know if we have a perfect solution other than read the sources that you think are the most, treat it all with a grain of salt, talk to other people who are there on the ground and try to have a good firewall for your brain so it doesn’t get hijacked by some of these memes.
Scott:
So, one more thing on Brian the human, what would people be surprised to learn about you?
Brian:
I find it fun to try to build things, it’s also stressful and so it’s a constant battle of I lean into this thing and it’s frustrating but then, if it works, eventually it’s fulfilling. And obviously, I don’t need to be working from a money point of view but I don’t think I would find it very interesting to not work. That’s probably a misconception people actually have about successful tech people in general is that they think that they’re doing it to accumulate more money for personal consumption or something like that. But I think we just find it fulfilling to try to improve the world, it literally is that simple.
Scott:
We’re all very naive in that way. I think we have a very child hopeful vision of our thing really helping people.
Brian:
Yeah. And you never know to what degree it’s going to help or not, you’re just trying your best ideas, some of them work, some of them don’t. But I want to try to be useful in the world that gives me a sense of purpose, I like trying to learn new things so it puts me into these uncomfortable situations where there’s always a next level of the game. In some ways I do think of it almost a little bit of a video game, I’m like, “I’m trying to do the best I can, I failed on this one, I did better on that one,” but I like being in the arena. So, yeah, I would just encourage more people to go try and build something to try to improve the world, I don’t ... I think there’s a temptation to be a critic, to sound smart, to be an activist, to try to ... Try to build something better to replace the thing that you don’t like and then you’ll come to develop an appreciation for how difficult it is maybe but, if you can do it, it’s fulfilling.
Scott:
I love the invitation to get out there and try and build something because it’s probably much easier than most people imagine it would be who aren’t sitting on the front lines.
Brian:
Yeah, you’re right, technology is a great democratizing force in that sense. I’m using the same AI model as everybody else in the world, I don’t get access to some special one. I have the same cell phone, I buy the same Bitcoin, I use the same whatever, statin drug, it’s off patent now, it’s super cheap. So, technology is a great equalizing force and maybe the first model that something comes out is expensive but the prices go down enormously as long as we have a free market.
Scott:
Yeah, that’s a temporal redistribution.
Brian:
Yeah.
Scott:
That’s great.
Brian:
Keep building.
Scott:
According to Brian, our operating systems need an update, Brian and I went back and forth on how. This is conspiracy corner.
Brian:
People are so worried about AI but we might actually feel great by becoming AI. It’s actually not that great to be human in many ways. I always make this joke you have to fall on the floor and hallucinate eight hours a day and you have to put all these dead plants and animals in your body to keep it running and why don’t we-
Scott:
It’s falling apart all the time.
Brian:
Yeah, no replaceable parts and we get addicted to sugar and social media and all these. So, in some ways people might find it very refreshing to have a more objective modern mind that is not running this super outdated software that gets addicted to sugar-
Scott:
Go for the savannah.
Brian:
... jealousy, all these things are ... Yeah. It’s really primitive naive algorithms that are just doing the survive and replicate thing, it has nothing to do with our own objectives around happiness or productivity or whatever. I think you actually came up with this concept, you were the first person I heard it from a while back about green versus gray. If you cut off a piece of plant, it can regrow itself with water and sunlight. If you cut off a piece of your iPhone, it can’t grow back with water and sunlight. So, in some ways, the bio stack is actually more ... It’s self-replicating-
Scott:
It’s nanotech, it’s Drexel’s vision of net technology basically.
Brian:
Yeah. What’s your view currently on the bio stack versus the, I guess, what would you call it, the compute stack?
Scott:
So, my framing around it is it’s basically as if you and I showed up on a planet that was full of the most advanced alien technology we’ve ever seen where it performs computations that are impossible for our digital circuits to do, it can regrow itself, it uses quantum sensing for all kinds of things, it’s just inconceivably efficient self-assembly technology, we have no idea how it works. That’s where we are with biology, we’re running it and it’s some amazing set of systems left behind by an ancient alien civilization we don’t understand and we’re now just barely getting a sense of some of the mechanisms and we’re a toddler pressing buttons basically on a control panel.
Brian:
I go back and forth on this. Sometimes I think to myself, you know what, really all that matters is AI and brain machine interfaces because the bio stack will clearly get left behind. We’re only adding another cubic inch of brain matter every 100,000 years or something and it takes us 20 years to birth a new human and train it, educate it enough to be a productive member of society, whereas, AI model can do that, fork it off in a millisecond or whatever and so it’s clearly the bio stack is behind. Then, other days, I think to myself, well, even if we get strong AI and brain machine interfaces, we’re still going to have 8 billion humans running around and we should still, from a nostalgia point of view if nothing else, but maybe it’s even bigger than that. It’s like the self-replicating nanotechnology, self-repairing nanotech that we don’t even understand, it’s worth trying to upgrade the bio stack with gene editing and longevity drugs and all these things. What do you think? Which stack would you bet on or is it both?
Scott:
So, I think there’s exactly one important thing happenin0g in our lifetimes which is AI, I think that is the thing. And everything else, whether it’s energy or bio or whatever is defined in relation to the center of gravity that is the advent of superhuman AI. Now, I have to caveat that because the AI we have now is really bad at out of distribution generalization and so bio is a thing that we need a lot more data for it to get good at even doing small narrow tasks, let alone the kinds of things foundation models are capable of in LMs and that data can’t be acquired in a hurry because cells take time to replicate and you need people to work with mice and work with primates and stuff to actually test it out. And so, I think it’s really good that we’re investing heavily in making tools that make biology transparent. And so, I love things that are instruments or ways of scaling data acquisition, data manipulation, interventions, what you’re doing at NewLimit and a bunch of other companies too, [inaudible 00:49:40], are creating data that one day our AIs will be smart enough to understand or the compute clusters will be large enough for us to actually process. Because the size of a bio dataset compared to the size of the ... Bio datasets make the internet look small but someday we’ll be able to learn from them and so we should be gathering them now is my view.
Brian:
It’s almost like you’re saying we need the bio stack at least to generate tons of data off of it so that, eventually, the AI can even understand this alien nanotech.
Scott:
There’s two problems with AI, really one problem which is that we are 10,000 times or more less efficient in our large models than the human brain is and it’s just extremely bad at learning causal relationships and ... The stuff we’re working on on Vicarious, it still can’t do. And as it turns out, a miracle happened and we found an alternative architecture for intelligence that works well enough that we don’t need all the stuff that our brain is doing, it’s less efficient but we can live with that. And so, what we get is something that works really great when we have an internet scale dataset and we’re predicting one token at a time and that’s amazing, it’s going to create so much economic value for everyone, that’s going to transform the world, very bullish on that. But because we never figured this out, we just skipped over it, we built plan B, dumped a lot of data on a lot of GPUs, what that means for the next step of building brain machine interfaces and for interfacing with our biology is that it’s going to be driven by datasets that take a long time to collect and you need a lot more of them than a human would need to figure out the same conclusion. So, that’s Progress in the mind of Brian Armstrong the disagreeable optimist betting on crypto, longevity and an American Wakanda. If you liked this episode, be sure to like, subscribe and share with a friend. I’m Scott Phoenix and this is Progress, thanks for listening.
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.