The following is a conversation with Jack Dorsey, co-founder and CEO of Twitter, and founder and CEO of Square. Given the happenings at the time related to Twitter leadership and the very limited time we had, we decided to focus this conversation on Square and some broader philosophical topics, and to save an in-depth conversation on engineering and AI at Twitter for a second appearance on this podcast. This conversation was recorded before the outbreak of the pandemic. For everyone feeling the medical, psychological, and financial burden of this crisis, I'm sending love your way. Stay strong. We're in this together. We'll beat this thing. As an aside, let me mention that Jack moved $1 billion Square equity, which is 28% of his wealth, to form an organization that funds COVID-19 relief. First, as Andrew Yang tweeted, this is a spectacular commitment. And second, it is amazing that it operates transparently by posting all its donations to a single Google Doc. To me, true transparency is simple, and this is as simple as it gets. This is the Artificial Intelligence podcast. If you enjoy it, subscribe on YouTube, review it with five stars on Apple podcast, support it on Patreon, or simply connect with me on Twitter at Lex Fridman, spelled f r i d m a n. As usual, I'll do a few minutes of ads now, and never any ads in the middle that can break the flow of the conversation. I hope that works for you and doesn't hurt the listening experience. This show is presented by MasterClass. Sign up on masterclass.com/lex to get a discount and to support this podcast. When I first heard about MasterClass, I thought it was too good to be true. For $180 a year, you get an all-access pass to watch courses from, to list some of my favorites, Chris Hadfield on space exploration, Neil deGrasse Tyson on scientific thinking and communication, Will Wright, creator of SimCity and Sims, both one of my favorite games, on game design, Jane Goodall on conservation, Carlos Santana on guitar, one of my favorite guitar players, Garry Kasparov on chess, Daniel Negreanu on poker, and many, many more. Chris Hadfield explaining how rockets work and the experience of being launched into space alone is worth the money. For me, the key is to not be overwhelmed by the abundance of choice. Pick three courses you want to complete. Watch each all the way through. It's not that long, but it's an experience that will stick with you for a long time. It's easily worth the money. You can watch it on basically any device. Once again, sign up on masterclass.com/lex to get a discount and to support this podcast. And now, here's my conversation with Jack Dorsey.
You've been on several podcasts, Joe Rogan, Sam Harris, Rich Roll, others. Excellent conversations. But I think there's several topics that you didn't talk about that I think are fascinating that I'd love to talk to you about. Sort of machine learning, artificial intelligence, both the narrow kind and the general kind and engineering at scale. So, there's a lot of incredible engineering going on that you're part of, crypto, cryptocurrency, blockchain, UBI, all kinds of philosophical questions maybe we'll get to, well, life and death and meaning and beauty. So, you're involved in building some of the biggest network systems in the world, sort of trillions of interactions a day. The cool thing about that is the engineering at scale. You started as a programmer with C or building...
I'm not really an engineer.
Not a legit software engineer. You're a hacker at heart. But to achieve scale, you have to do some unfortunately legit large-scale engineering. So, how do you make that magic happen?
Hire people that I can learn from, number one. I mean, I'm a hacker in the sense that my approach has always been do whatever it takes to make it work, so that I can see and feel the thing and then learn what needs to come next. And often times, what needs to come next is a matter of being able to bring it to more people, which is scale. And there's a lot of great people out there that either have experience or are extremely fast learners that we've been lucky enough to find and work with for years. But I think a lot of it we benefit a ton from the open source community and just all the learnings there that are laid bare in the open. All the mistakes, all the success, all the problems. It's a very slow-moving process usually, open source. But it's very deliberate. And you get to see because of the pace, you get to see what it takes to really build something meaningful. So, I learned most of everything I learned about hacking and programming and engineering has been due to open source and the generosity that people have given to give up their time, sacrifice their time without any expectation in return other than being a part of something much larger than themselves, which I think is great.
The open source movement is amazing, but if you just look at the scale like Square has to take care of, is this fundamentally a software problem or hardware problem? You mentioned hiring a bunch of people. But it's not maybe from my perspective not often talked about how incredible that is to sort of have a system that doesn't go down often, that is secure, is able to take care of all these transactions. Like maybe I'm also a hacker at heart and it's incredible to me that that kind of scale could be achieved. Is there some insight, some lessons, some interesting tidbits that you can say how to make that scale happen? Is it a hardware fundamentally challenge? Is it a software challenge? Is it like is it a social challenge of building large teams of engineers that work together? That kind of thing? Like what's the interesting challenges there?
By the way, you're the best dressed hacker I've met.
Thank you, by the way.
Of the enumeration you just went through, I don't think there's one. You have to kind of focus on all and the ability to focus on all that really comes down to how you face problems and whether you can break them down into parts that you can focus on because I think the biggest mistake is trying to solve or address too many at once or not going deep enough with the questions or not being critical of the answers you find or not taking the time to form credible hypotheses that you can actually test and you can see the results of. So, all of those fall in the face of ultimately critical thinking skills, problem solving skills. And if there's one skill I want to improve every day, it's that. That's what contributes to learning and the only way we can evolve any of these things is learning what it's currently doing and how to take it to the next step.
And questioning assumptions, the first principles kind of thinking seems like fundamentals of this whole process.
Yeah, but if you get too overextended into well, this is a hardware issue, you miss all the software solutions. And vice versa. If you focus too much on the software, there are hardware solutions that can 10x the thing. So I try to resist the categories of thinking and look for the underlying systems that make all these things work. But those only emerge when you have a skill around creative thinking, problem solving, and at being able to ask critical questions and having the patience to go deep.
So, one of the amazing things if we look at the mission of Square is to increase people's access to the economy. Maybe you can correct me if I'm wrong, that's from my perspective. So, from the perspective of merchants, peer-to-peer payments, even cryptocurrency, digital cryptocurrency. What do you see as the major ways our society can increase participation in the economy? So, if we look at today and the next 10 years, next 20 years, you go into Africa, maybe in Africa and all kinds of other places outside of North America.
If there was one word that I think represents what we're trying to do at Square, it is that word access. One of the things we found is that we weren't expecting this at all. When we started, we thought we were just building a piece of hardware to enable people to plug it into their phone and then swipe credit card. And then as we talked with people who actually tried to accept credit cards in the past, we found a consistent theme, which many of them weren't even enabled, not enabled but allowed to process credit cards. And we dug a little bit deeper, again asking that question, and we found that a lot of them would go to banks or these merchant acquirers. And waiting for them was a credit check and looking at a FICO score. And many of the businesses that we talked to and many small businesses, they don't have good credit or a credit history. They're entrepreneurs who are just getting started, taking a lot of personal risk, financial risk. And it just felt ridiculous to us that for the job of being able to accept money from people, you have to get your credit checked. And as we dug deeper, we realized that that wasn't the intention of the financial industry, but it's the only tool they had available to them to understand authenticity, intent, predictor of future behavior. So, that's the first thing we actually looked at, and that's where we built the hardware, but the software really came in terms of risk modeling. And that's when we started down the path that eventually leads to AI. We started with a very strong data science discipline because we knew that our business was not necessarily about making hardware, it was more about enabling more people to come into the system.
So, the fundamental challenge there is so to enable more people to come into the system, you have to lower the barrier of checking that that person will be a legitimate vendor. Is that the fundamental problem here?
Yeah, and a different mindset. I think a lot of the financial industry had a mindset of kind of distrust and just constantly looking for opportunities to prove why people shouldn't get into the system. Whereas we took on a mindset of trust and then verify, verify, verify, verify, verify. So, we moved, you know, when we entered the space, only about 30 to 40% of the people who applied to accept credit cards would actually get through the system. We took that number to 99%. And that's because we reframed the problem. We built credible models and we had this mindset of we're going to watch not at the merchant level, but we're going to watch at the transaction level. So, come in, perform some transactions, and as long as you're doing things that feel high integrity, credible, and don't look suspicious, we'll continue to serve you. If we see any interestingness in how you use our system, that will be bubbled up to people to review to figure out if there's something nefarious going on, and that's when we might ask you to leave. So, the change in the mindset led to the technology that we needed to enable more people to get through and to enable more people to access the system.
What role does machine learning play into that in that context of, you said first of all this is a beautiful shift. Anytime you shift your viewpoint into seeing that people are fundamentally good and then you just have to verify and catch the ones who are not as opposed to assuming everybody's bad, this is a beautiful thing. So, what role does machine learning throughout the history of the company has it played in doing that verification?
It was immediate. I mean, we weren't calling it machine learning, but it was data science. And then as the industry evolved, machine learning became more of the nomenclature and as that evolved, it became more sophisticated with deep learning and as it continues to evolve, it'll be another thing, but they're all in the same vein. But we built that discipline up within the first year of the company because we also had to partner with a bank. We had to partner with Visa, MasterCard, and we had to show that by bringing more people into the system, that we could do so in a responsible way that would not compromise their systems and that they would trust us.
How do you convince that this upstart company with some cool machine learning tricks is able to deliver on this sort of a trustworthy set of merchants?
We staged it out in tiers. We had a bucket of, you know, 500 people using it and then we showed results and then 1,000 and then 10,000 and 50,000 and then the constraint was lifted. So, again it's kind of, you know, getting something tangible out there. I want to show what we can do rather than talk about it. And that put a lot of pressure on us to do the right things. And it also created a culture of accountability, of a little bit more transparency, and I think incentivized all of our early folks in the company in the right way.
So, what does the future look like in terms of increasing people's access? Or if you look at IoT, Internet of Things, there's more and more intelligent devices. You can see there's some people even talking about our personal data as a thing that we could monetize more explicitly versus implicitly. Sort of everything can become part of the economy. Do you see, so what does the future of Square look like in sort of giving people access in all kinds of ways to being part of the economy as merchants and as consumers?
I believe that the currency we use is a huge part of the answer. And I believe that the internet deserves and requires a native currency. And that's why I'm such a huge believer in Bitcoin because it just, our biggest problem as a company right now is we cannot act like an internet company. Open a new market, we have to have a partnership with a local bank. We have to pay attention to different regulatory onboarding environments. And a digital currency like Bitcoin takes a bunch of that away where we can potentially launch a product in every single market around the world. Because they're all using the same currency. And we have consistent understanding of regulation and onboarding and what that means. So, I think, you know, the internet continuing to be accessible to people is number one. And then I think currency is number two. And it will just allow for a lot more innovation, a lot more speed in terms of what we can build and others can build. And it's just really exciting. So, I mean, I want to be able to see that and feel that in my lifetime.
So, in this aspect and other aspects, you have a deep interest in cryptocurrency and distributed ledger tech in general. I talked to Vitalik Buterin yesterday on this podcast. He says hi, by the way.
He's a brilliant, brilliant person. Talked a lot about Bitcoin and Ethereum, of course. So, can you maybe linger on this point? What do you find appealing about Bitcoin, about digital currency? Where do you see it going in the next 10, 20 years? And what are some of the challenges with respect to Square, but also just bigger for our world, for the way we think about money?
I think the most beautiful thing about it is there's no one person setting the direction. And there's no one person on the other side that can stop it. So, we have something that is pretty organic in nature. And very principled in its original design. And I think the Bitcoin white paper is one of the most seminal works of computer science in the last 20, 30 years. It's poetry. I mean, it really is.
Technology. I mean, that's not often talked about. Sort of there's so much sort of hype around digital currency, about the financial impacts of it, but the actual technology is quite beautiful from a computer science perspective.
Yeah, and the underlying principles behind it that went into it, even to the point of releasing it under a pseudonym. I think that's a very, very powerful statement. The timing of when it was released is powerful. It was a total activist move. I mean, it's moving the world forward in a way that I think is extremely noble and honorable and enables everyone to be part of the story, which is also really cool. So, you asked a question around 10 years and 20 years. I mean, I think the amazing thing is no one knows. And it can emerge and every person that comes in the ecosystem, whether they be a developer or someone who uses it, can change its direction in small and large ways. And that's what I think it should be because that's what the internet has shown is possible. Now, there's complications with that, of course, and there's, you know, certainly companies that own large parts of the internet and control it more than others and there's not equal access to every single person in the world just yet, but all those problems are visible enough to speak about them, and to me that gives confidence that they're solvable in a relatively short time frame. I think the world changes a lot as we get these satellites projecting the internet down to Earth because it just removes a bunch of the former constraints and really levels the playing field. But, a global currency, which a native currency for the internet is a proxy for, is a very powerful concept, and I don't think any one person on this planet truly understands the ramifications of that. I think there's a lot of positives to it. There's some negatives as well.
It's possible, sorry to interrupt, do you think it's possible that this kind of digital currency would redefine the nature of money, so become the main currency of the world as opposed to being tied to fiat currency of different nations, and sort of really push the decentralization of control of money.
Definitely, but I think the bigger ramification is how it affects how society works. And I think there are many positive ramifications.
Outside of just money, just...
Outside of just money. Money is a foundational layer that enables so much more. I was meeting with an entrepreneur in Ethiopia. And payments is probably the number one problem to solve across the continent, both in terms of moving money across borders between nations on the continent or the amount of corruption within the current system. But the lack of easy ways to pay people makes starting anything really difficult. I met an entrepreneur who started the Lyft/Uber of Ethiopia, and one of the biggest problems she has is that it's not easy for her riders to pay the company, and it's not easy for her to pay the drivers. And that definitely has stunted her growth and made everything more challenging. So, the fact that she even has to think about payments instead of thinking about the best rider experience and the best driver experience is pretty telling. So, I think as we get a more durable, resilient and global standard, we see a lot more innovation everywhere. And I think there's no better case study for this than the various countries within Africa and their entrepreneurs who are trying to start things within health or sustainability or transportation or a lot of the companies that we've seen here. So, the majority of companies I met in November when I spent a month on the continent were payments oriented.
You mentioned as a small tangent, you mentioned the anonymous launch of Bitcoin is a sort of profound philosophical statement.
What's that even mean? Like there's a pseudo...
There's an identity tied to it. It's not just anonymous. It's Nakamoto. So, Nakamoto might represent one person or multiple people. But...
Let me ask, are you Satoshi Nakamoto? Just checking.
And if I were, would I tell you?
But a pseudonym is a constructed identity. Anonymity is just kind of this, you know, random, like drop something off and leave. There's no intention to build an identity around it. And while the identity being built was a short time window, it was meant to stick around, I think, and to be known. And it's being honored in how the community thinks about building it. Like the concept of Satoshis, for instance, is one such example. But I think it was smart not to do it anonymous, not to do it as a real identity, but to do it as a pseudonym because I think it builds tangibility and a little bit of empathy that this was a human or a set of humans behind it. And there's this natural identity that I can imagine.
But there is also a sacrifice of ego. That's a pretty powerful thing from your perspective.
Would you do, sort of philosophically to ask you the question, would you do all the same things you're doing now if your name wasn't attached to it? Sort of if you had to sacrifice the ego. Put another way, or is your ego deeply tied in the decisions you've been making?
I hope not. I mean, I believe I would certainly attempt to do the things without my name having to be attached with it. But, it's hard to do that in a corporation. Legally. That's the issue. If I were to do more open-source things, then absolutely. Like, I don't need my particular identity, my real identity, associated with it, but I think, you know, the appreciation that comes from doing something good and being able to see it and see people use it is pretty overwhelming and powerful. More so than maybe seeing your name in the headlines.
Let's talk about artificial intelligence a little bit, if we could. 70 years ago, Alan Turing formulated the Turing test. To me, natural language is one of the most interesting spaces of problems that are tackled by artificial intelligence. It's the canonical problem of what it means to be intelligent. He formulated as the Turing test. Let me ask sort of the broad question, how hard do you think is it to...
pass the Turing test in the space of language?
Just from a very practical standpoint, I think where we are now and for at least years out is one where the artificial intelligence, machine learning, the deep learning models can bubble up interestingness very, very quickly. And pair that with human discretion around severity, around depth, around nuance and meaning, I think, for me, the chasm to cross for general intelligence is to be able to explain why and the meaning behind something.
Behind a decision or a set of data. Sets of data.
So, the explainability part is kind of essential to be able to explain using natural language why the decisions were made, that kind of thing.
Yeah, I mean I think that's one of our biggest risks in artificial intelligence going forward is we are building a lot of black boxes that can't necessarily explain why they made a decision or what criteria they used to make the decision. And we're trusting them more and more from lending decisions to content recommendation to driving to health, like, you know, a lot of us have watches that tell us to run or stand. How is it deciding that? I mean, that one's pretty simple, but you can imagine how complex they get.
Being able to explain the reasoning behind some of those recommendations seems to be an essential part.
Although, it's also a very hard problem because sometimes even we can't explain why we make decisions.
That's what I was... I think we're being sometimes a little bit unfair to artificial intelligence systems because we're not very good at some of these things.
Do you think... Apologize for the ridiculous romanticized question, but on that line of thought, do you think we'll ever be able to build a system like in the movie Her that you could fall in love with? So, have that kind of deep connection with?
Hasn't that already happened? Hasn't someone in Japan fallen in love with his AI?
There's always going to be somebody that does that kind of thing. I mean, on a much larger scale of actually building relationships, of being deeper connections. It doesn't have to be love, but it's just deeper connections with artificial intelligence systems. So, you mentioned explainability.
That's less a function of the artificial intelligence and more a function of the individual and how they find meaning and where they find meaning.
Do you think we humans can find meaning in technology in this kind of way?
100%. 100%. And I don't necessarily think it's a negative. But it's constantly going to evolve. So I don't know, but meaning is something that's entirely subjective and I don't think it's going to be a function of finding the magic algorithm that enables everyone to love it. But maybe. I don't know.
But that question really gets at the difference between human and machine. So you had a little bit of an exchange with Elon Musk. Basically, I mean it's a trivial version of that, but I think there's a more fundamental question of is it possible to tell the difference between a bot and a human? And do you think it's... if we look into the future 10, 20 years out, do you think it will be possible or is it even necessary to tell the difference in the digital space between a human and a robot? Can we have fulfilling relationships with each or do we need to tell the difference between them?
I think it's certainly useful in certain problem domains to be able to tell the difference. I think in others it might not be as useful. I think it's possible for us today to tell that difference.
Is the reverse the meta of the Turing test?
Well, what's interesting is I think the technology to create is moving much faster than the technology to detect.
You think so? So if you look at like adversarial machine learning, there's a lot of systems that try to fool machine learning systems. And at least for me the hope is that the technology to defend will always be right there, at least. Your sense is that...
I don't know if they'll be right there. I mean, it's a race, right? So, the detection technologies have to be two or 10 steps ahead of the creation technologies. And this is a problem that I think the financial industry will face more and more because a lot of our risk models, for instance, are built around identity. Payments ultimately comes down to identity. And you can imagine a world where all this conversation around deepfakes goes towards a direction of driver's license or passports or state identities. And people construct identities in order to get through a system such as ours to start accepting credit cards or into the Cash App. And those technologies seem to be moving very, very quickly. Our ability to detect them, I think, is probably lagging at this point, but certainly with more focus, we can get ahead of it. But this is going to touch everything. So, I think it's like security. We're never going to be able to build a perfect detection system. We're only going to be able to... you know, what we should be focused on is the speed of evolving it and being able to take signals that show correctness or errors as quickly as possible and move and to be able to build that into our newer models or the self-learning models.
Do you have other worries like some people, like Elon and others, have worries of existential threats of artificial intelligence, of artificial general intelligence, or if you think more narrowly about threats and concerns about more narrow artificial intelligence. Like what are your thoughts in this domain? Do you have concerns or are you more optimistic?
I think Yuval in his book 21 Lessons for the 21st Century, you know, his last chapter is around meditation. And you look at the title of the chapter and you're like, oh, it's kind of, you know, it's all meditation. But what was interesting about that chapter is he believes that kids being born today, growing up today, Google has a stronger sense of their preferences than they do. Which you can easily imagine. I can easily imagine today that Google probably knows my preferences more than my mother does. Maybe not me per se, but for someone growing up only knowing the internet, only knowing what Google is capable of or Facebook or Twitter or Square or any of these things, the self-awareness is being offloaded to other systems. And particularly these algorithms. And his concern is that we lose that self-awareness because the self-awareness is now outside of us and it's doing such a better job at helping us direct our decisions around should I stand, should I walk today, what doctor should I choose, who should I date. All these things we're now seeing play out very quickly. So, he sees meditation as a tool to build that self-awareness and to bring the focus back on why do I make these decisions? Why do I react in this way? Why did I have this thought? Where did that come from?
That's a way to regain control.
Mhm. It's more awareness, maybe not control, but awareness so that you can be aware that yes, I am offloading this decision to this algorithm that I don't fully understand and can't tell me why it's doing the things it's doing because it's so complex.
That's not to say that the algorithm can't be a good thing. And to me, recommend our systems the best of what they can do is to help guide you on a journey of learning new ideas, of learning period.
It can be a great thing, but do you know you're doing that? Are you aware that you're inviting it to do that to you? I think that's the risk he identifies, right? That's perfectly okay. But are you aware that you have that invitation and it's being acted upon?
