The Eyeball-Scanning Plan Designed to Save Us From AI

0:00You're being sold an AI future where you're obsolete or irrelevant. That vision is wrong. At Palantir, they're building AI that helps workers and unlocks their full potential. American workers are our nation's greatest strength. AI shouldn't eliminate them. It should elevate them. Palantir is here to tell their stories. From factories to hospitals, AI is freeing people from drudgery, letting them do what humans do best. Create. Solve. Build. Palantir, making Americans irreplaceable.

0:38So you're telling me that the AI that's meant to make everyone's job easier to manage just adds more to manage on top of the thousands of apps the IT department already manages. Funny how that works. Any business can add AI. IBM helps you scale and manage AI to change how you do business. Let's create smarter business, IBM.

1:15Hello and welcome to another episode of the Odd Lots podcast. I'm Joe Weisenthal. And I'm Tracy Alloway. Tracy, do you remember a few weeks ago or maybe a couple of months ago now, Someone made a fake AI version of the Odd Lots podcast, and they replicated our voice. Yes, I think they used ChatGPT to do it, and they basically asked ChatGPT to create a script for Odd Lots, and then they had someone replicate another app or something replicate the voices. It wasn't a someone. I mean, it was some other AI app. It was pretty good. It was good. I mean, it clearly was not us. How much did it worry you, Joe?

1:56I think I'm more of an AI doomer warrior than you are. Interesting. I mean, I have concerns, don't get me wrong, but what do you think is the worst case scenario? Is it robots taking over the world? No, I think that like in a couple of years, the AI will do a really good job of making the Outlaws podcast. And people will say, I don't really need to listen to Joe and Tracy anymore. The AI is, I don't know. I'm like, you know. But I feel like maybe that's a little paranoid. No, I feel like there is reason for concern. I hope. Maybe I'm just being an optimist here. I hope we are a few years away from the Odd Lots replicant.

2:29I know, but I want to do this for like another 30 years. I don't want to just have like three or four years or a few years. That's the optimistic scenario, is it? Doing this in 30 years time? Okay. All right. AI, lots of people expressing some worries about it. A million different kind of worries. Yeah. And we see all these, you know, good ways that the technology is being used, but also insidious ways. As you pointed out, someone made an Odd Lots episode. I think that was in a very fun-spirited way, a nice way. But there are people who use this technology for more nefarious things. Yeah, I think there was like a 60 Minutes recently where someone was able to impersonate someone's voice and say, hey, you could imagine a scam where someone's like kids say, hey, I need money or something like that.

3:15And they call and it sounds just like them, especially in a couple of years when the tech gets better. and it becomes a bit of a crisis in the sense it's like, do I know that a person is a person? When they talk to me on a Zoom, when they talk to me on a podcast, when they talk to me on a phone call, et cetera. Well, even on Twitter, on a much more basic level, you have all these bots. And I don't know about you, but there seems to be a new Tracy Alloway with the L as like an I and fooling everyone almost like every few weeks. And then you have to go through the whole bureaucracy of reporting to Twitter and proving that you are in fact the original Tracy Alloway.

3:49And not to get on too much of a tangent, but wasn't Elon going to solve this? Because he said like, oh, I'm buying Twitter to solve the bots problem. And I don't, and as you say, it's not been solved. I think that's a whole other episode. But even with like really rudimentary, like, you know, fake Tracys and fake Joes, like that's the most rudimentary thing. Just like taking our avatar and like changing like one of the L's and Trace in L away to like a I or whatever, or a one. Like, that's just going to get worse. Mm-hmm. You know what's a really good business model? creating a problem and then creating the solution to it.

4:26Yes, that's exactly right. So we have all of these AIs and they're going to replicate us and they're going to look like us and they're going to talk like us. But the good news, in a sense, good, is that others are working on the problem of how do you prove that you are who you are in a world in which AIs can replicate our voices and our faces and our expression? Technology can save us from technology is the thrust of this, right? Exactly. So we are going to be speaking to someone working on a technology designed to address this concern in part. And as you say, some of the people working on it are some of the same people behind some of the most cutting edge AI research that's sort of like blowing everyone's minds in both good and scary ways these days.

5:12So I'm just going to go ahead and lay out my priors here, which are that I find this sort of instinctively dystopian. The topic of this podcast, what we're going to be discussing, this technology creeps me out a little bit. But that said, I don't know that much about it. I don't know much about how it works. And so I am very interested to hear the bull case on eyeball scanning. So you just jump right. We're going to be talking eyeball scanning. We are going to be speaking with Alex Blagna. He is the co-founder of WorldCoin, an entity that was set up also by Sam Altman, a name that everybody knows now because he's the co-founder or the founder of OpenAI, which powers ChatGPT and other tools, other co-founder, Max Novenstern.

5:58We're going to be talking about how an eyeball scanner can allow us to theoretically prove our humanity in this future AI world. So, Alex, thank you so much for coming on OddLots. Thank you for having me. I'm excited to talk. So I'm trying to think, I mean, like the threat, how real is it that, you know, in a few years, do you think, not whether AI will be able to create a perfect replica of the podcast, but that let's say we're having this conversation. and I might really not know that I am talking to the real Alex Blania versus an AI that can perfectly replicate your voice and replicate your face.

6:44Like, is that two years away? Is that three years away? Like, how soon is that coming? Well, I think partially we are already there, right? The technology is not completely open source yet, so you don't have widely deployed systems that can do that. But rather, right now, you have a few entities that actually can do things like that or would be able to do things like that. But this is going to change because, of course, the progress is increasing. So it started with DALI, and you had stable diffusion and many other things that, on the image generation side, became increasingly impressive. And I think AI art is one of the coolest things of the last couple of the last year, to be honest.

7:31So that's on the image generation side of things. And then, of course, you have ChatGPT on the text generation side of things. So I think on Twitter, as you discussed in the introduction, you are, I think, already there. That it could happen that you just chat with a neural network and you have absolutely no idea that you chat with a neural network. And you might argue about odd lots for like an hour and don't realize that this is not actually a person. I'm just thinking like half the people I argue with on Twitter probably are bots. Neural networks. Yeah. Well, maybe just to step back in time a little bit.

8:03I mean, WorldCoin has been around for a few years now. And to some degree, as Joe mentioned, you guys kind of anticipated this threat. What was the sort of aha moment that made this all come together? Like, how did this actually develop into a company? and what was the use case thesis when you were just setting out? Well, we honestly did not, we didn't start off to solve a threat, right? That is not brief history in time. Back then I did research. So we started the company a little bit over three years ago and I was doing research in deep learning applied to quantum physics. So I I basically predicted quantum systems of neural networks.

8:50And that's kind of, I think, is a very promising field in physics. And Sam was already working on the idea of Roquan together with Max. So I think they've already been on it for like six months. But of course, Sam had a full-time job at OpenAI. Max had a full-time job back then. So it was all like in a early idea phase. Then I got an email from Max back then with an early paper describing the vision of WorldCoin and that I should come by and talk. And then I had a couple of interviews with Max, one with Sam. We spent a lot of time together and actually decided to proceed. But to go back to the initial proposition, and I think that is very important, that it's not about solving a threat, but rather how, I mean, Sam was already working on OpenAI, and this was three years ago.

9:41So back then it was not yet an accepted topic that AI is going to happen and it's going to be as powerful as it is today. It was a very niche topic, I think. Although OpenAI, of course, was already quite a famous entity. But of course, Sam was motivated by the idea of AI and what it's going to do to the world. And it's going to be this very powerful technology. And so the initial proposition of Rockland was really, what is the societal infrastructure that we can build that makes sure that AI is for the better of all of society versus a few? And so there's many things actually that come together here, like proof of personhood is what you just described is the ability to prove that you're an actual human being on the internet, which is basically just a much more capable identity system.

10:30but it goes as far as we think eventually UBI will become a thing and it actually is like a big chance if you have like a very, very powerful technology and it's not about kind of political redistribution but you have a very centralizing technology that you want to share the upside with the world with and that was another motivation kind of all kind can actually be the infrastructure to do that because you have a very capable identity system you have an economic system that connects everyone So there were like a couple of very ambitious reasons why we think something like that. So basically creating and bootstrapping the largest identity and financial network that is actually open and privacy preserving versus relying on centralized government systems that might not work in large parts of the world is a very important proposition for the coming decades.

11:24So let's just jump right to it. People who are listening to this podcast, you might want to actually watch the video version. We're recording this on video as well. But here on set, I was going to wait a little while before unveiling it, but this is, I'm holding in my hand. I feel like we need extra music. A big orb. It feels about six pounds. I don't know. How heavy is this, Alex? I'm sure you know the exact weight. It's a little bit over one kilogram. Oh, okay. So I was wrong. So that's like just under like two and a half pounds. Okay. Okay, that's fine. About two and a half pounds and it's an eyeball scanner.

12:07And so here it is and it's not working right now or we're not scanning right now. So I can like tilt this in our face. It's not scanning. But let's like talk about what this piece of hardware that I'm actually holding it is. I'm going to set it on the table in front of me. it is an eyeball scanner. What does it do? So the short answer is one, it I will start with the short and then the longer answer. The short answer is that it first ensures that whatever is interacting with the device is an actual human being. So not a display or a photograph of an eyeball. Not a photograph or more sophisticated optical attacks because there's many things you can do here.

12:46So that's the first step. Checking that whatever we see is an actual human being. And then the second piece is it takes an image of my eye, calculates a unique embedding out of this picture. Everything happens on a device. And that embedding then is signed by the device. And that's then the only thing that actually leaves the device and gets compared against all other users. And if that uniqueness check is successful, then me, that I just verified with an orb, I then can later do a zero knowledge proof that I am included in that set. And if you're not familiar with your knowledge proofs, what it allows you to do is it lets you prove something without actually revealing the underlying information.

13:35So what that enables us to do is it builds essentially a self-custodial identity network where the data control is with the user and there's no centralized entity that actually has the information of who you are, where you're from, all of these things. But rather, this person is a unique human being and has verified Revolkan before, while the user could decide to attach more information to that. So that's kind of the short of it. I think we should talk much more about the why in a second, because that's not trivial, but yeah. Yeah, maybe just to help us get to the why, you could talk about what would be a practical use case for this technology?

14:18Because if I think about something like, you mentioned UBI, universal basic income. If I think a government starts some sort of UBI program, then I would assume that they identify people by something like a social security number or whatever the equivalent might be in that country. So what does this do that those traditional systems of identity do not? Well, so when you start with traditional identity, the really hard thing for governments is actually making sure everyone only has one passport. And that is with an identity is called deduplication. It's a really hard problem because all the systems, all the technologies you use usually in your life, let's say your iPhone, Face ID.

15:02what that actually does is just it re-authenticates you. So it realizes I'm the same person again logging into that phone. And that's a one-to-one comparison and that is fairly easy to do. What is really hard to do is making sure, okay, Alex did not sign up compared to a billion other people. And so that's why many governments actually have biometric systems in place because otherwise also their social security number system would completely break. But that's not rolled out globally. So in fact, a good mental model is about 50 % of the global population actually have a digitally verifiable identity.

15:41That's a huge number of people that just don't have it. And so I think United States is pretty good, or at least decent. It's actually not a really good identity solution. Europe in parts is pretty capable. The strongest is India, although there's definitely privacy concerns here because they don't use knowledge proofs and things like that. But for example, India, they have a program called Adhar that also used hours recognition to kind of roll out to the global population. And it really accelerated the whole economy like crazy. So they had Adhar and then they built UPI on top, which is a payment rail.

16:21And so whenever you talk to people that are in that kind of area of technology, everyone always brings up India as the most prominent example of what identity plus a strong financial rail can actually do to accelerate an economy. And this is what RollCoin is doing. Basically, RollCoin is able to create a deduplication set. So it provides me with the ability to prove that I'm actually unique against everyone else. but then also there's world id which is an identity protocol on top of that and let's start with the basics so the first thing i could do is i could log into twitter and i could say i don't already have a twitter account without actually reeling my name um and then if you think a little bit forward you could even cryptographically sign every tweet you make so that you you kind of lay a trace of like okay this is actually alex versus anyone else and i I think this is really where this is going to go, because this progress around AI is definitely an exponential, so we will get to a place where you will have to authenticate pretty much everything you do on the internet.

17:29So that's the basic, but then with WorldID, which is a protocol that lets you add verifiable credentials and many other things within identity space, you can actually bootstrap a whole identity system on top. So this was like a quick rant. No, it was good. So it's not that it's proving that you are a Joe Weisenthal or a Tracy Alloway. It's proving that there is only one of you without necessarily the ID attached or the name attached. That is exactly. So that is the whole point. But then what WorldID lets you do, so if you're not into crypto, but this whole idea of self-custody is very important, right?

18:14Of course, I'm pretty sure it came up before. What that lets you do is you basically, you verify for Volcon, you receive your World ID. And then it might be that for some applications, you actually want to prove that you're over 18 or you're Joe Weissetal. Then the World ID protocol, which is literally just like an open source protocol and every developer, every company could implement with that. you as a user can then decide to also attach additional information to your world ID. So that is my name or that is my KYC information. And then you can decide to reveal that to some applications with zero knowledge proofs as well.

18:53But no one other than you actually has that information. So I think, God, I have like hundreds of questions. There are so many different ways. So, you know, one basic, but one idea, you keep mentioning zero knowledge proof And just to sort of like the basic idea, and this is the short is like, you know, I go to websites sometimes, government websites, you enter in your social security number. But the idea is with a zero knowledge proof, it would be I prove to the government that I know my social security number without telling the government what my social security number is. Well, I mean, that one is actually tricky because in that case, you would reveal the social security number.

19:30What is a much better way to explain it is the website asks you, are you on any kinds of sanction list? Or are you over 18? Or are you actually a US citizen? And then these things you could prove of zero knowledge. And then basically what comes out of this is like a proof runs on your phone that you're not included in a sanctions list. let's say, and that the only thing that comes out and ends up with the website is yes or no. But the website does not know what is the social security number you checked against.

20:25Silicon Valley is selling you a future where you're obsolete, or worse, identical. At Palantir, they're witnessing something different and revolutionary, From re-industrializing the nation's defense base to shipyard workers building faster and frontline workers boosting productivity, AI is transforming work across the nation. AI is not replacing American workers or flattening them into conformity. It's unleashing what makes each one irreplaceable. Their judgment, their craft, their creativity. When American workers become more powerfully themselves, they own the future. Palantir, making Americans irreplaceable.

21:33hr and more no more app overload no more juggling logins just one seamless system that makes work easier and the best part odoo replaces multiple expensive platforms for a fraction of the cost it's built to grow with your business whether you are just starting out or already scaling up plus it's easy to use customizable and designed to streamline every process so you can focus on what really matters running your business thousands of businesses have made the switch So why not you? Try Odoo for free at odoo.com. That's O-D-O-O dot com. So then the other question I have, and the obvious thing is, look, scanning your eyeball, it's going to make anyone anxious.

22:16And I'm like totally up for it. And even I'm like, you know, a little bit anxious about it. But how do you establish – so there's no – unlike, say, Clear at the airport, which also uses an eyeball scan so you can go by security, there's no centralized database of our irises. And how do you sort of like establish or prove to people that there's no database? Because if my understanding, there has been reporting that at least at some point in the process of the development of WorldCoin, as you've grown as a startup, that at one point, or at least currently, there was a Iris database. Okay, so there's many things here.

22:55Yeah. I'm going to start with explaining how to develop comfort around the idea and why it's necessary. Sure. First of all, I hated the idea of building the orb. When we actually started working on Rollcoin, it was absolutely horrific because we've been four people sitting in a small apartment in San Francisco. And the idea of building hardware devices that take biometric data and roll them out globally was a horrific idea back then. One, because it's very complicated, expensive. Yeah, I imagine pitching that one to VCs would be kind of tricky, right? Oh, it was horrific. It was horrific. It was in the middle of, it was also like, I mean, Bitcoin is also a crypto project.

23:40And we started in like in the middle of a bear market. So I remember back then when we talked to the first investors, they're like, okay, like other than Ethereum or Bitcoin, nothing will exist. Like all of this nonsense. What do you talk about AI also? No one cares about that also. Like, what are you even doing? But again, so neither me or anyone else was really excited about building a hardware device. But really, we spent almost a year doing research in pretty much everything you could do to solve the problem. And solving that problem is, okay, issuing a privacy-preserving identity that works globally, not only in the United States, not only in Europe.

24:24What do we have to do? And there's basically like three big answers to this. One is you use government KYC, right? Which might be a sufficient answer for some parts of the world, but it was pretty shocking to realize for how few, right? So kind of that very quickly disqualified. Second is what people call web of trust. So this whole idea that you're just like, you build reputation up between people and that also like in short, just didn't work. And then the third is biometrics and everything that surrounds that whole field. And within biometrics, we then build prototypes for pretty much everything you could imagine.

25:08We build palm scanners. We build kind of a face ID-like implementation of it. We build fingerprint scanners, all these things. But the short of it is to solve this deduplication problem, so this uniqueness check, you need a lot of entropy. so information about each user. If you don't have that, your error rate explodes exponentially. And so simple example, if you would use Face ID to solve the same problem, after tens of millions of users, you would have to reject everyone. So you don't have a constant error rate. You don't have like 5 % that we constantly reject, but you just hit a wall and your system breaks.

25:49And so faces work, same is true for fingerprints. and the only thing that really works is and is proven to work is IRIS recognition. So that's why we ended up there. And then even with IRIS recognition we even had to build our own hardware. We had to build our own lens, which sounds absolutely ridiculous, but we kind of had to build this whole device from the ground up to make that happen. So much to the why. And how do we get people or how do we think about this whole headline of comfort around this? there's a couple things. One, actually built a privacy-preserving system that goes way further than what you're used to.

26:30And that's, I think, really what we did because there's no information about you that you actually reveal by using WorldCoin. And I think that's pretty remarkable. It's definitely counterintuitive because you're like, okay, we use eyeball scanning. So this is headline one. Headline two, and that's a journey for sure, is open sourcing everything. I think that's one of the core parts of the crypto ethos is just you don't have to trust, but rather you can verify. And the hardware is completely open sourced already, the software to meaningful parts, the protocol completely. So everything will be open sourced in the coming months.

27:15And so that's answer two. and then three, it actually is way less of a thing than you would think. So we have now 1.7 million users, roughly 600 ,000 of them that use the app on a monthly basis. I travel the world very frequently. I talk to a lot of users on the ground and just try to understand what's your feeling, what do you understand, what do you not understand. And it is very much, yeah, like broadly speaking, Globally speaking, this is way less of a concern, which obviously does not lower our bar, but it's the case. So I have two things on that, and I think they're sort of related. But one, why does open source seem to be such a thing in AI in particular?

28:04And then secondly, what is the crypto role in this project? Because, you know, we mentioned WorldCoin. I can only imagine what it was like pitching to VCs in the midst of crypto bear market, saying we want to scan everyone's eyeballs. And oh, by the way, there's also like a crypto element to this as well, a token. Why have that component of it? I mean, at this point, I'm actually not an AI expert anymore, because I think I'm basically two years out of the field and pointed to crypto, which is an interesting journey also. But soon you'll be going back into AI, probably. That's what all the crypto people are doing now, right?

28:46Well, no comment. I'm joking. So why is open sourcing such a thing? It is basically around AI specifically. It is that you have a technology that is probably as foundational as electricity, as literally, it's still, I think, incomprehensible to people. What a big deal this is, right? This is one of the biggest paradigm shifts in technology that happened ever. And there's different approaches of how to deal with this and all the underlying risks. And basically, two different lines of thought is, one is you open source everything. Everyone should be able to verify and understand how everything is happening.