SFBW19 - The Pseudonymous Economy - Balaji Srinivasan

So today I'm going to talk about something that I'm calling the pseudonymous economy. And, you know, I want to motivate it. You know, why should we want this thing? And then how might we build it? So further ado, let me just jump in. So first, you know, going to talk about what is pseudonymity? Then why pseudonymous economy? How might it work? How can we build it? So what is pseudonymity? It is. It is not anonymity. So, you know, if you're, if you're thinking about a real name, you know, for example, on Facebook, people typically use real names. On 4chan, people are typically anonymous. But on places like Reddit and sometimes Twitter, people are pseudonymous, where they have a name that is sort of a persistent kind of thing that gains reputation associated with it, but it's not a real name. It's not your global unique identifier they use on official forms. And the concept here of pseudonymity, the reason it's important to distinguish from anonymity is that pseudonymity provides accountability, but as we'll see, it's also a shield against character assassination. So one of the reasons I wanted to talk about this and direct people's attention towards this is I believe pseudonymity is as important as decentralization, albeit under kind of focused on. And the reason is that while the Bitcoin network was decentralized, Satoshi Nakamoto was also pseudonymous. And that meant that the state couldn't play either the man or the ball. Okay? So it meant that you couldn't go after the person who built the network and you couldn't shut down the network itself. That was critical in the early days of Bitcoin. And we're seeing some of the opposite of that now with government hearings on Libra and so on. If you have a highly named network and it's centralized, you've got kind of both. Both points vulnerable. So pseudonymity is as important as decentralization, especially for certain kinds of applications. And one important concept is pseudonymity is not a 01, it's a continuum. So there's this really powerful concept called 33 Bits by Arvind Narayanan. He popularized it as a prophet at Princeton. And the basic idea is that 2 to the 33rd power is 8 billion, or a little more than 8 billion. 8.6. And there's only 7 something billion people in the world. So with 33 independent bits of information, you can de anonymize somebody. Now, if they have zero bits of anonymity left, then you know exactly who they are. But if they have, let's say 10 bits of anonymity left, they're 1 out of 2 to the 10th people. So about 1 out of 10, 24. If, if they have 20 bits of anonymity left, they're one out of two to the 20th, so about one out of a million. And so in this fashion you can try to say, okay, I have a number of bits of pseudonymity, a number of bits of anonymity, and it's a continuum as opposed to being a totally zero, one discrete thing. Now pseudonymity is already mainstream. I guess everybody saw Pierre Delicto from last week, which is Mitt Romney, but James Comey was also using the pseudonym online. And this is happening at the very most senior levels of government, but it's also happening among Gen Z teenagers. Kids will set up so called finsta and rinstas like a fake Instagram account under a fake name and a real Instagram account under a real name. And what's funny is they're their fake self under their real name. They're in their Sunday best or their bar mitzvah clothing or what have you and they're smiling for the camera and they're their real self under their fake name. And so these, both kids and adults alike have fashioned these search resistant identities where you have hundreds of millions of people, like Reddit alone is several hundred million who use pseudonyms on a daily basis. And they do so not to commit crimes or anything like that, but simply to be able to discuss without retaliation. And so pseudonymity is already mainstream. Over the last 10, 20 years, we've built up hundreds of millions of people who use sometimes multiple pseudonyms. And I think the big next step is going to be not just pseudonymous communication, but pseudonymous transaction. So this is where I think society is going and there's several kind of correlated vectors towards this. At the top is Google multiple account support. Every major social media company, tech company, whether it's Twitter, Facebook, Google, et cetera, now offers multiple account support, if only to allow you to move between your individual account and your corporate account for placing ads or issuing corporate tweets or what have you. But that's now actually something that's very common. Many people have multiple accounts and they just use them in different contexts. So that's one big driver is just the technological support for it across many different services. The second driver is employment law, where increasingly people don't want to even inadvertently discriminate against somebody. So you won't ask about marital status or immigration status or other kinds of things in a job interview. And the third reason for pseudonymity is basically social media mobs, which can attack you and debit your reputation, and you have no recourse. So these are kind of three different vectors that are sort of pushing in towards why people would want to use pseudonyms online and in employment context and so on. Okay, so that's kind of what is pseudonymity, and give some initial motivation why a pseudonymous economy. So, you know, there's this guy, Yakov smirnov, in the 80s, he had this one liner which said, in the Soviet Union, we also have freedom of speech, but in America, you have freedom after speech. So the point of pseudonymity is it allows freedom after speech. You can't retaliate against a person. You can only retaliate against their ideas. And that's why with Satoshi Nakamoto, basically, the man or the woman or the people who did this, they can be denounced under the name Satoshi nakamoto, praised. This cloud of millions of articles swirls around the pseudonym, but the person walks unscathed. Without a real name, you have no GPS location. You can't send in the drones. So, you know, when we talk about freedom after speech, social media mobs are now routine. You know, this is. This is something which anybody who's posted on social media has seen this, and the way of thinking about it is negative press on social media. It actually, it's not true that sticks and stones can, you know, break your bones, but words can never hurt you. The way that it works is negative press. Negative social media is an attack on your social network. And this visual over here, I think, encapsulates one of the reasons why one would want a pseudonym. If, for example, you have friends or partners, employers, investors, customers who read and respect a particular news outlet, that news outlet, when it publishes an article saying that you are now a bad person, has right access to their brains. If they read and respect this news outlet and they see that you're now a bad person, well, in the after scenario, some of them break ties. The investor doesn't close the deal. The employee quits, and the third person may not totally quit or break the deal, but they may retreat to what is contractually obligated. And so in this fashion, negative press, negative social media, it's like a mortar landing in your social network. It frays the ties, breaks the ties. But if instead the article is announcing your pseudonym now, it doesn't matter. You walk unscathed. There's no damage done to your social network. Nobody knows, Nobody can Map it. The concept of too much information is now applicable. And basically the way I think about this is pseudonymity defends against social supply chain disruptions in the same way that you've got a supply chain disruption for a component. If you're a hardware manufacturer, you know, like, is my rare earth elements? Are they arriving on time? You know, that kind of thing. My glass, your social network also has a supply chain. And your key investor, your key employee, your key customer, they can't go south on you as easily if what is being denounced is your pseudonym as opposed to you. Okay, so how might a pseudonymous economy work? Well, the way I think about it is your bank account is your stored wealth, but your real name is your stored reputation. And one sort of visual is as you accumulate wealth in your bank account, the number goes up on Twitter, which we're going to use as a working example. It's certainly not everything, but it's at least one slice of the problem. As you kind of tweet, you accumulate followers. So it's like your stored reputation. And while only you can debit your bank account, you know, by entering in your pin, anybody can debit your reputation. You know, there's a, there's a funny tweet which says on Twitter, every day there's a main character and your goal is to never be it. Okay, I think that's funny. So, you know, one way to deal with this is to separate out your earning, your speaking, and your real names. So your real name would just be used on official forms, like your Social Security number. You don't just disclose that casually. You use that on official forms when the government needs that official input. When speaking, you use a pseudonym, like Comfortably Smug or someone like that. And when earning, and this is the new part, you would earn under a pseudonym, you might have a GitHub handle under that. You basically just have a professional identity. And what you would tweet about or speak about is something that's relevant to that professional identity. And you don't have to have just one earning and one speaking and one real name. You might have only one real name, but you might have many different speaking names, many different earning names. Just like in the same way that you might use separate credit cards for every merchant that you buy from. Privacy.com allows that. The new Apple card allows that. It kind of gives a default privacy. In the same way that you might use multiple emails like your name +foo to go and sign up at a website to give some kind of isolation and scoping, you can use different names for different contexts. And so as a way of sort of formulating a technical problem, we can already move wealth to a pseudonym. Can we move reputation as well? In the top right, you can see, you know, take $100,000 in USD, you can buy Zcash at Coinbase, for example, and you can have 50K left under your real name, and you can move 50K to a pseudonym. So we can now transfer wealth to a pseudonym. The question is, can we transfer some reputation to a pseudonym? So, for example, in the bottom is Mark Anderson 700k followers. Could it be possible to set up an account where the original account stays with 700k followers, but there's a new pseudonymous account that has like 100k followers or something like that? Is it possible to trade off some reputation in return for some anonymity rather than all versus all? And so I'll give a cryptographic construction, kind of a thought experiment here as to how we might build something like this. So let's take the special case of Twitter. You know, in Twitter you have both real names and pseudonyms, you know, coexisting. And we're going to sort of describe, like, over the next few slides, a few constructions that'll get us to like, a special crypto Twitter. This is the kind of thing that people in crypto are thinking about. People are building. There's a few primitives, new primitives you need to build for this. So we'll start kind of application first and then describe what you need on the back end. So just to motivate, you know, we talked earlier about how pseudonymity can be quantified in bits on the X axis. And also we can. We're going to quantify reputation, the number of followers on the Y axis. And right now you kind of have two choices. Either you can have zero pseudonymity and all your followers, so you're operating under your real name, or you can have total pseudonymity and no followers and just set up a new, you know, like, fake account or pseudonymous account? And so an interesting question is, is there something in the middle? Can we trade off some degree of pseudonymity for some, you know, to gain some followers? Is there, there's something between those two extreme points. And that's what I meant about the information you need to identify the third or three bits. So can we turn off some pseudonymity for some distribution? And here's a naive approach. So this data structure here, it's called the adjacency matrix. In a graph in the upper left, we see kind of an Individual, in this case Mark, and he's got four people pointing into him, and those are represented in the first row by four ones, because those folks are pointing him into him, they're following him rather than vice versa. So it's asymmetric matrix. Now if we did the most naive approach, where a new pseudonym was booted up and all of the followers were ported over to that new pseudonym, the problem is that anybody who was analyzing this graph would see that the vector of followers was the same between these two people. Basically that you know that they're really part of the same network. And you could, you could see that those two patterns were extremely similar zeros and ones. And so there would be effectively zero pseudonymity. You wouldn't have any bits of pseudonymity. So that's bad. And so the naive approach of simply setting up a new account and transferring all your followers to it doesn't work. Okay, can we do something smarter? What if we transfer attestations, fungible attestations, rather than the entire non fungible list of followers? And so how would that work? So if we go back to a Twitter profile, there's kind of three types of data there. There's a data that is put there by the platform, like the user's verification. There's a data which is stated by the individual themselves, like their name, their location, their bio. Then there's data that is input by third parties, like the followers, the likes and the RTs. The first and the third are actually relatively hard to fake. The second, like the name, location and bio, you know, that's just self attestation, so it's not as secure. Let's focus on the first one, the verification. What we could do is we could set up something where, you know, Mark sets up a new pseudonym and then does a zero knowledge attestation. If, if the backend of this new crypto Twitter allowed it, that would push over the bit that says that he's a verified user. Okay, so we linger on that. So essentially you're moving something from one account to another account, like a zero knowledge attestation. And what's cool about this is that if we think about it from the perspective of how much pseudonymity you're giving up, in the first case, you're completely unique. You have zero bits of, you know, pseudonymity, anonymity. In the second case, if you're just a pure new, you know, fake account or pseudonymous account, you could be any one of the 330 million MAUs on Twitter. So you have about 28 something bits of anonymity. But there's about 330,000 verified users. So if you move over that bit of verification, you're now one of 330,000 verifieds. You've given up 10 bits of anonymity. And so we can now quantify the trade off between pseudonymity and reputation. And that's new and that's really, I think, interesting and powerful because variants of this can be applied to all major social networks and so on. Okay, so we're making progress. We've got a construction that will move over the fungible fields where other people share them. There's other people who share that verified flag, as opposed to the non fungible follower vector, which other people do not share. Okay, so now, you know, the obvious question arises. Well, can we do more? Yes, of course we can do more. You could also port over, you know, the fact that you have more than 100k followers under one account, or that you're followed by Jack or something like that. And each such attestation that you pour over, you can determine prior to porting it how much of a pseudonymity decrease it is, of course, if Jack only follows, you know, one person, you have de anonymized yourself by doing this. But you can see that prior to porting over that attestation. And of course, it's not just some degree combinatorial. If you have 100,000 followers and Jack follows you, that is less anonymity than either one of those independently. So you just have to calculate the joint distribution. Okay, very doable. But there's another problem that now arises, which is, okay, we've set up this new thing. It's got this verification and these other kinds of attestations attached to it. How do we get followers for this? And so there's another kind of construction which we'll want, which is something I call auto follow. So the idea here is that folks would be on a network like this in order to hear from pseudonymous accounts in part. And so they would set up a flag which is like, okay, I'll auto follow anybody who has poured over a verification, who's got more than 10k followers, who you know, if Jack follows them. Any pseudonymous account that boots up like this is kind of interesting. It's just like you could follow a person or you could follow a hashtag. You could follow pseudonymous accounts that have a certain level of these attestations. And so now you set up a pseudonym and you move over verification. And now, because those three people have autofollow set up for Verification, boom, you get some followers. And this is not that hard to implement if this is how the network is set up. But this would now start to have something where people could use pseudonyms on kind of a more reasonable level and they could actually pull over some of their distribution. Moreover, they could know of their existing followers how many were set up to auto follow accounts that had particular attestations associated with them. So you could calculate how much followers you'd get for a particular pseudonymous. Okay, so let's assume that like 150k of the 700k followers have this auto follow set up. In this case we have a construction which yields an intermediate identity. It doesn't have the full 700k, it has 150k that have auto follow set up. It doesn't have full anonymity of 28.3 bits of being one of 330 million people, it's one of 330,000 people. But still that's pretty good, right? We have, you know, we can quantify the reduction in anonymity due to porting over the attestation and we can quantify how much distribution you've retained due to auto follow. And this gives you a way of sort of configuring a anonymity reputation trade off and you can negotiate other points along that line. Now, this means we can now move both wealth and reputation to a pseudonym by kind of implementing these two new sort of backend constructs. There's different ways of doing this. You might set up a new crypto Twitter where you get a bunch of folks to move over there. There's a degree of strength in numbers, so you want to have a large enough critical mass. So that part I kind of leave as something we're going to have to build over the next 10, maybe 20 years. But so, you know, I'm well aware, kind of, if you're a physicist, there's this concept of a sumospherical cow and everything's become simple. So this is sort of assuming a crypto Twitter makes things simple, but this is sort of a motivational kind of thing. And I think it's interesting for crypto devs to think about an important feature of a crypto social network. So a bunch of people are building these things. So just in summary, we discussed what is pseudonymity? You know, you've got your real name, your pseudonym and your anonym, and those are three different things. The pseudonym has persistent reputation. We talked about, you know, why you'd want a pseudonymous economy. And the answer is by Separating out your earning and your speaking names. Having some social media attack on you doesn't affect your daily economics. It doesn't affect your social network supply chain. How could it work? We separate out your earning, your speaking and your real names and we build it with this cryptographic construct or something similar that allows you to trade off some anonymity for some reputation. And I'm open to other ways of accomplishing this beyond the one that I outlined, but I just wanted to at least set up the problem and provide one sort of thought experiment solution. Okay, so that's it. And you can follow me on Twitter if you like this stuff. Apologies. Thanks. I think we have time for a question or two if people want to. Yeah, how do you. Yeah, so the question was, how do you handle offline to online? And I think the thing is, if you're offline, it's sort of hard to operate under a pseudonym because you can be face recognized and so on and so forth. There's folks who worked on this with like CV dazzle, like makeup that makes you face recognition resistant. I'd say that, like, I don't have a good solution for that yet. Just thinking about the online part right now.

How do you handle the fact that you do content analysis? So you come through anonymous and you're talking about the same topics and the same kind of voice.

Great question. So her question is, how do you handle the issue of content analysis? For example, Joe Klein in the 90s wrote this book Primary Colors and was eventually de anonymized with something called stylometric de anonymization, where you can go and analyze the frequency of stop words, other kinds of things, and figure out who they are. Well, one answer is there's new AI techniques called style transfer. And just like you can transfer the style of like a, like a wallpaper to another image, you can also have a style transfer thing that would have you write like Hemingway or write like Jane Austen. So you write your normal text, you punch it through style transfer, and it will translate it not into another language, but into the argo of another kind of speaker. And that's kind of an interesting way of sort of defeating styleometric de anonymization. Look at my Twitter. There's a paper that I linked biologist styleometric de anonymization. I don't think it's like, I don't think there's a production one yet out there, but the papers are out there so you can implement it. Sir, how do you keep people from selling the child accounts? They spin up how do you keep people from selling the child accounts they spin up? Well, it is certainly possible to do that. Actually, I don't even think that's necessarily bad. It just means that you have somewhat less trust in that user's utterances or what have you. You might do something where, at least at the platform level, you link it to the same private key with their main account. And then if so, you know, you could. You could do something where they would be giving up their money or their access to their money by doing that. There's various kinds of constructions. I think that's a good question. What could be our identity in free private cities? Oh, what could that identity be in free private cities? Stay tuned. So the identity in free private cities. I think. I think what's gonna happen is that social media and cryptocurrency are causing the unbundling and rebundling of the nation state. And that's like a 20 or 30 or 40 year kind of voyage. And we're just at the beginning of that. And I think over time what you're gonna find are you're gonna have cryptocurrencies that are integrated with like the fabric of a nation state. And that's, you know, Estonia is kind of like one of the first along this line with the digital residency that you're gonna see much more like that. Okay, thanks.