Alex Blania on Proof of Human and Building World's Identity Network

0:05Alex Blania and Erik Torenberg and Ben Horowitz

0:38Erik Torenberg:How do you prove you're real? In 1950, Alan Turing proposed a test. If a machine could fool a human into thinking it was also human, it had achieved intelligence. For decades, that remained theoretical. Today, AI agents run thousands of social media accounts at once, outperform humans in controlled persuasion tests and generate hundreds of videos a day that audiences believe are real. The Turing test didn't just get passed. It got commoditized. Every platform built on the assumption that its users are human now faces a problem no one has solved. Facial recognition fails at scale. Government IDs weren't designed for a global internet.

1:21Erik Torenberg:I speak with Alex Banya, co-founder and CEO at World, which is building the largest real human network, a proof of human layer for the AI era, alongside A16Z co-founder and general partner, Ben Horowitz. Alex, welcome to the podcast. Great to have you. Thanks for having me. So Proof of Human is having a moment right now. Why don't you first give a background for people who are unfamiliar? What is the moment that's happening and how did we get here? Yeah, and what is Proof of Human? Proof of Human, as the name suggests, is do you know if you interact with a human or like something else on the internet?

2:00And I actually think the kinds of questions that we're now asking is, are you interacting with a human, an agent on behalf of a human, or just an agent? I think these are like roughly the three areas that we want to split apart. And describe a little bit the difference between just an agent and an agent acting on behalf of a human. How do you see that distinction? Yeah, so quickly explaining just the term proof of human and I think what is hard about it, and then I'll explain how that fits into an agent on behalf of a human. So what proof of human really means is that every individual that interacts on a platform has only one, ideally one account or a limited number of accounts and stays the owner of that account.

2:41Like that's kind of the property that you're looking for. So like you're looking for initial verification that ideally should be something like anonymous or very extremely privacy-preserving and then ongoing authentication that the same person remains in control. account. And then there's like some secondary properties that I think are put to half. But that actually tells you that the really hard thing is uniqueness. Like what is happening on a platform like Twitter right now is that there's all these accounts, all these bots that are in replies, that there's probably one human sitting somewhere and sending out 10 ,000 or 100 ,000 of AIs.

3:14And there's this catch-up game where Twitter and X are trying to just find them and block probably millions a day of these. Which is what, a one-hundredth of Of the bots. That's right. That's how it feels like. And then agent on behalf of human, I think how it will look like is, I think all of us will have agents. It's unclear how they will look. It's going to be one or there are multiple ones, maybe with different tasks and different even types of characters. And I think it will then come down to, I approve a certain action of my agent. I give him certain rights. So act on my behalf. Okay. Post to my ex account, post to my Instagram.

3:51For example. But it's my Instagram and I'm a unique human that owns that. That's right. You know, that X or Instagram could decide that if that's actually something they want as a platform. Right. But that's how you could do it. That makes sense. And so how do you prove somebody is human? It is a surprisingly hard problem. Yeah. Those agents are very, very clever. It's funny. We started this company now a couple of years ago, before ChetGPT and before all of that. But we kind of took that as an assumption that eventually we will have AIs that pass the Turing test. So they can just claim to be a human.

4:27You will not be able to tell them anymore on the internet. And also that they would be highly agentic and just run around and do their own thing. And so that makes it really, really hard because back then when we started the company, there were like roughly three big ideas that people were interested in. One was this idea of web of trust or like related ideas. So this idea that you look how someone behaves on the internet or did behave in the past. So like usually a combination of you have the certain number of accounts that you own since a couple of years and then you post regularly or you comment regularly to GitHub.

5:03These are the kinds of things that people are using. And then let's say all three of us have them. And then I attest also that I know you in the real world and I attest to you that I know you in the real world. And that's how you would build a certain graph. and that was like a very hot idea back then for this but we disregarded it basically immediately because we assumed that eventually everything that is just digital and AI will be able to do as well. Yep. We're there. Yeah, exactly. So an AI will be able to have a GitHub account and will be able to post and own an account and like also attest to five other AIs that these are in fact humans even though they're not.

5:39So that was area number one. Area number two was to just use government IDs. for everything, which we just all see me disregarded for a couple of reasons. One is that I think, you know, it's strictly better if the government would not control such an infrastructure in terms of free speech and actually breaking that apart. But then also... Right, you lose anonymity instantly, right? You could hypothetically set up a system that maybe preserves it, but it's very hard to do. And then the second thing is also the government identity system is just not built for that. And what is so hard about this problem is it's going to be a global problem.

6:17And so it doesn't really matter if one government maybe has the perfect infrastructure. For example, Singapore is like an example of a government that has perfect infrastructure all around. But that barely doesn't matter because, for example, I don't know, Meta is a global product with three billion users and with a lot of other countries. Yeah, Singapore has two million people, a million people. Yeah, exactly. So do you want to lock everyone else out? And then there's a long list of other things why we disregarded that basically immediately. And then the last one is biometrics, which actually immediately gives us this egg reaction.

6:48It's like, and it even went further because what is so hard about this problem, as I mentioned in the beginning, is uniqueness. And so just like in very simple words, how you can describe the problem is, well, first of all, for example, what does Face ID do? Face ID checks that I'm the same person again using my phone. And so it's a one-to-one authentication. So there's an embedding store on my phone. It takes a picture of my face, creates a new picture, compares to the previous one. And if that is close enough, I can use my phone. So that's a one-to-one. One embedding to one new embedding. To solve the proof of human problem, you will need to distinguish one new individual from all previous individuals.

7:29You need to make sure that Ben is trying to sign up and Ben did not sign up before. And then suddenly it goes from one-to-one to one-to-N. And it's the size of your network, essentially, that you're trying to prove that to. And then you can just do the math and you can calculate how much mathematical entropy, like how much information, just information theoretically, do you need to prove that? And it turns out that's a pretty high number because it's an exponential problem. And so then you can just do the math and you find out that things like face or even fingerprints or something doesn't work.

8:02Then you would basically hit a wall after tens of millions of users. And so then you end up with something like iris, which is the muscle of your eye that actually has enough entropy. And that it's unique. That is unique. That is unique enough. And how do you also then solve the one thing that biometrics have been subject to historically is just replay attacks? Where, okay, I may not have your eyeball, but I've got enough information that I can run a replay attack on you. So, again, it is important, I think, to split up the problem and verification, which is essentially in all terms it's like you're getting your passport and then authentication which is you showing your passport constantly for certain kinds of things and on the verification piece we've went down, if you know World, you know that we've built this thing called an orb it's doing a lot of things to prevent these kinds of attacks for example it has multiple sensors in the electromagnetic spectrum to just make sure that you cannot show a display to it and it would recognize that.

9:05So I think on that side, we've got it handled. On the consumer side, you know, to then re-authenticate, it turns out to be much harder because you would need to trust the phone in some sense. Because what we actually do in that moment is when you verify with an orb, not only do we check your uniqueness in a fully anonymous and privacy-preserving way, and we should talk about that, but also we send to your phone a signed face image that you then can later use to re-authenticate against it. Right. And with a new iPhone, you can have meaningful amount of trust against that, but with old Android phones, basically.

9:38Oh, yeah, yeah, yeah. Because you can just show a deepfake, essentially, either through a display or just directly injected in the camera stream. So that's a problem. And so it's going to be a mix of, you know, if you have a new enough, let's say, iPhone or a general phone, then you can just re-authenticate against that picture that you took on verification. Otherwise, you would probably have to even go back to an orb somewhat frequently. Let's say a couple of times a year. I see. To re-authenticate. Yeah, that's right. Interesting. And then one of the kind of incorrect criticisms of the approach early was, oh my God, they've got my eyeball.

10:17You know, now they somehow have access to my privacy and they're going to do all these things to me and that's my access. and then they can, they, WorldCoin can impersonate me and all these kinds of things. But that's not the case. So that was also like a non-trivial engineering problem. It was very much non-trivial. So actually, I think one point on Iris that I think people don't appreciate enough, and that's a bet we took back then, but it was essentially that Iris will turn out to be super normal as a modality just because I think we will all wear AR and VR systems that do that. You know, Apple already does it.

11:01Already has RSID in the Vision Pro. So I think it's, so maybe that's a general point. I think it's going to become something that we will use across many different devices and we'll normalize in that sense. But I think on the privacy piece, that took us a lot of time. Because when we decided back then that, you know, with our assumptions, you know, which was six years ago, that we will need a custom hardware device for biometrics, it was actually quite scary, you know, to come to the conclusion. Yeah, that's an expensive conclusion. It's like very expensive, and then just having this idea that you would need to distribute them all over the world, like that just assumes that you would be able to like somehow bring up billions of dollars and to like a massive effort to distribute it all over the world.

11:48But then also the privacy challenge of like how could you build such a system that has all the requirements that we care about And the two main high-level ideas on how to solve it were multi-party computation and zero-knowledge proofs. And so, again, what is different to Face ID, because Face ID actually can be very private just because the embedding is stored on the phone. It doesn't have to leave the phone ever, just because it's just you against you in the past. but to check uniqueness you need to check against all previous people so something needs to leave something needs to leave something and be compared to someone else and that's a much harder challenge and how we approach that is we have multi-party computation so that essentially means that in our case when you verify with an orb we take all these pictures they get computed on the device and then they actually get split up in multiple pieces.

12:58So for example we take a picture of the iris we calculate an iris code then we break that iris code in multiple pieces and send it to multiple computers such that there is no central database in some sort. So no one actually has the information about you. And then you do some clever tricks of how these different parties need to come together to do a computation that still leaves the pieces apart. Right, right, right. In such a way that... Where nobody has the whole thing. Yeah, so no one has the whole thing. And also during the computation, no one has the whole thing. Yeah. But they do some clever interactions to come to the conclusion.

13:36A little like a zero-knowledge proof kind of technique. I mean, it's very different, but I think in terms of the properties it achieves, it's somewhat similar. Where like you... No one knows anything about you, but you can actually together make a statement about you. Right. And so, you know, you send it to this multi-party computation and what comes back is, yes, that individual is unique. And then the second thing we do is we separate all of this from you with a zero knowledge proof. So meaning you have that secret on your phone, but no one else has it, no server has it, we don't have it. And then you can later go back to this multi-party computation and say like, hey, I have a secret that is part of that computation and I am in fact unique.

14:24And you can prove that to a platform. You could go to a social network and prove that you're a unique user to a social platform without us knowing anything about you or the social network knowing anything about you. And so it's this like very counterintuitive property that you, there is like, even though it uses biometrics, you preserve anonymity and extreme love with privacy, which I think is super cool. You know, social media is one kind of vector of, you know, things that were annoying and are now becoming overwhelming in terms of just bots, you know, particularly with psyops, propaganda, all these kinds of things.

15:03What are some of the other, you know, uses of bots that are going to be kind of impossible to live with if we don't get the proof of human in the future? Sure. Yeah, actually, I think the simple model I have for it is every moment on the internet that is primarily about humans interacting with each other, you know, or even indirectly interacting with each other. So, you know, you can start with simple ones like dating, you know, that really matters. The other side is, in fact, a person. Yeah, well, the... Got bad news for listeners. And the person who you expect it to be. Yeah, exactly. We had the problems even before.

15:46The whole catfish thing. Yeah, exactly. Yeah, so that's an obvious one. And so, for example, Tinder's already using it for that reason. I think... And what's the Tinder use case? So we started in Japan as a test market, and it's essentially exactly what we just discussed. It is if you verify it with an orb, you get a little batch that you know signals to other people that you are in fact a human so it has a high level of verification and then also I don't think that's live yet but what will come next is that you're actually the person you claim to be so meaning you have a world ID that is associated to the kind of profile pictures that you use so you just run a quick check that this is all correct and so you know you then know you're not interacting with bot but also you're you know, you interact with a fully authentic profile.

16:44Another fun one, because I think it's somewhat constitutive, but I think it will be video conferencing. Because, you know, you already have deep fakes. Yeah, just, I don't feel like going to this video conference. Just put my deep fakes up. Yeah, and actually, you raised it to me first, and that's why we started building a product for it. Because, you know, it will actually start with very high-value users. Like, for example, people, you know, like yourself, that maybe manage a fund. And, you know, sometimes calls actually could be very high value if it's about borrowing money or. Oh, yeah, yeah.

17:15Well, so somebody can be me and say, Eric, can you please wire this Nigerian prince? $400 million. Right. Exactly. That would be good to know. Yeah. Yeah. Like, you know, that's still slightly hypothetical because these things are not fully real time and you can somehow. We're very close. But we're very close. And so I think, you know, in a year from now, it's just going to be a full commodity and it's going to be super photorealistic and absolutely real time. And you will just not know anything anymore on these video calls. And so I think that's another one. I think another one then will be, which I think is fun, but it's going to be gaming, you know, because.

17:57Oh, yeah, yeah. Because like gamers really care. Oh, yeah, that they're not playing AI. Holy cow, that's frustrating. Especially if we bet money. exactly and you lose money you train multiple hours a day to get like really good at this thing and then suddenly you get you know you get destroyed by an AI that is just superhuman in every dimension funny enough I was like I wonder what you think about this but because I don't have a good mental model about it but even the whole model for video platforms I think is about to break because there's a couple dimensions that are problem. But one, if the creation of content is becoming super scalable.

18:41Like for example, I heard about this one guy that created, I think like it was like on the order of a hundred videos a day on YouTube and made tens of thousands of dollars a month. All of them were fully AI generated. And people just fell for it. So now the question is, is that actually something that YouTube wants to monetize that way? Yeah, well, it's interesting. right they fell for it but maybe they liked it that could be but it would sure be nice to know like okay this is a human video or this is an AI video actually my thesis about this is like something along the lines of I think there's categories of content that are clearly just fictional like movies are that you don't care that there's any connection to reality it's just a fully fictional story but now if you think about something like TikTok or you know all these kind of things like people actually really care about them mostly because there is some connection to reality yeah well there's reality and there's connection to human right so you can create a pretty good like you can take a scientific paper and give it to Gemini and say make this into a podcast and you know it'll be like a pretty entertaining podcast and it will be reality and that it came from some real thing.

20:06But you would like to know that. You would like to know that. I would like to know that. And then it continues. As an advertiser, you would like to know did a human watch it? Or did an AI watch it? Yes. Right, right. Well, right. That's the other thing. I created 100 AI videos. I had a million AIs watch it. And then I made a lot of money off of YouTube. Exactly. I actually saw that video today of a YouTube farm. They had these thousands of phones that just watch videos all day for a reason. Yeah, yeah. And then that's got zero value to the YouTube advertisers. And so that's actually a real problem for them.

20:43Right.

20:43Erik Torenberg:The whole sort of the creator economy platforms of the last decade, you know, Substack, Spotify, and all the people who support artists or, you know, Patreon, creators, YouTubers, they have a personal relationship with these people. It's not just they like the art. And so if they all of a sudden found out that they were, you know, bots that might, you know, they might not want to support them in the same way. Yeah, you might not want to give them a big YouTube tip. Yeah, I think there's a certain subset of people who support, you know, want to support actual people and feel like they're having a real relationship.

21:16Yeah. And the thing that I think like people don't really get is that, you know, it should be obvious, but I don't think people really understand the consequence of that. I think two things. One is that what we currently experience is like a super, super tiny thing of what is about to happen. You know, just because... Yeah, right, it's a glimpse. It's a glimpse. Like, you know, cost of intelligence is dropping almost exponentially. Agente capabilities are increasing, you know, in like some super linear form. So like, yeah, what we currently see is less than 1 % of what it will look like in probably a year or two.

21:52And then second, these things will be actually, they will be superhuman. in many ways. They would be perfectly able to understand you and talk in the right way to you. For example, there's this one paper that I think you could have deleted after, but it was the Change My Mind subreddit where the University of Zurich did this thing where they had AIs actually interact with Change My Mind. And they were superhuman in their ability to change it because they were going back to their profile of the people posting it and were like, understanding their political motivation, the way they talk, and they're just interacting in the perfect way.

22:32You know, and just like hit all the buttons. AIs are really good at programming humans. That's much better than humans are at programming AIs. Absolutely, there's no question. And so I think that's going to get quite scary also. I think at least if you know you're being a victim of a PSYOP, or it's a very advanced one done by an AI, that would be extremely useful to understand.

22:57Erik Torenberg:Totally. Talk a little bit more about the state of the product and the business today. Like how many IDs are out there? Why don't you give a little bit of that to you? Maybe talk about the evolution as well. Well, first of all, it's a multi-sided problem. And I think there's like roughly three that you have to consider. One is, well, you need platforms to use the technology. Then, you know, like things like Reddit or, you know, X or, you know, things like that. Secondly, you need distribution of these devices. And I think the right mental model to have for it is how many minutes does it take a person to reach such a device on average?

23:37And currently, if you would take the global average, it would be a terrible number. It would be like days or something because many people would need to fly. But how do we get that down to below 15 minutes across the US? And so that's probably roughly around 50 ,000 devices that you need to deploy. That's like, it's not crazy, but it's also not nothing. It's, you know, it's hard to do. And then the last one is, how does all of that come together to something that a lot of people really want to use it? And that's a combination of, you know, the utility of all the sub-platforms, essentially. But all of that layer is on top.

24:13Like, maybe you can use it in your Reddit account. Maybe you get, like, you know, a certain amount of TGPD subscription for free. So I think it's going to be a combination of things, but you need to land all three at some point at the same time, which is hard to do. We are now at 18 million users that are verified, 40 million in total in the app. But the biggest thing is because of the past administration, because we use crypto, we did not really invest in the U.S. for a long time. And that's not the main shift that we're going through. For all of this, the main thing that matters is the U.S. and hopefully we get the Clarity Act passed shortly.

24:52Yeah, exactly. That would be really great. So to get clarity on that. So the big focus that we now are going through right now is to kind of go all in on the U.S. So I think over the next year, 90 % of the effort of the company is just going to go about the U.S. And how do you get, for example, device distribution up? How do you eventually have this in every Starbucks? so it becomes just super normal and people just use it every day so that's kind of the and then on the platform side actually we went through it was a very interesting experience to go through personally because I think like a couple of years ago universally people just made fun of us it was like the universal reaction well Minas and Risa and a couple other people that believed in it but yeah like in the press Like the amount of fun making of something that just shows how short-sighted people are.

25:53That's right. It's like, you don't think the bots are coming? What did you think when we first pitched, actually? Because even you must have thought this is crazy. Well, because you had the orb. Like the orb was so wild. You know, okay, we're going to scan people's retinas, and that's how we're going to know they're human and so forth. And this was, I mean, you pitched us. Six years ago? Six years ago? Yeah, it was before COVID because you were there with the orb. Right. And, you know, AI just hadn't happened yet. And, you know, you could kind of see, but there, you know, there's bots, but they were kind of very crude and, you know, compared to what there are now.

26:36But it seemed inevitable, at least. At the time, you know, the thing was, it was so out, it was so from the future that, you know, we always worry about, okay, like, what's the timing of this and this and that and the other and so forth. But, you know, you were impressive enough and it was going to happen eventually. And it was an exciting enough idea that I think all those things kind of got us to go, okay, we're, but it was not, it was one of the, it wasn't obvious that like it was going to work in that time frame.