OpenAI Forum
+00:00 GMT
Sign in or Join the community to continue

Worldcoin: Distinguishing Humans From AI

Posted Nov 13, 2023 | Views 26.9K
Share
speakers
avatar
Miles Brundage
Head of Policy Research @ OpenAI

Miles Brundage is a researcher and research manager, passionate about the responsible governance of artificial intelligence. In 2018, he joined OpenAI, where he began as a Research Scientist and recently became Head of Policy Research. Before joining OpenAI, he was a Research Fellow at the University of Oxford's Future of Humanity Institute. He is currently an Affiliate of the Centre for the Governance of AI and a member of the AI Task Force at the Center for a New American Security. From 2018 through 2022, he served as a member of Axon's AI and Policing Technology Ethics Board. He completed a PhD in Human and Social Dimensions of Science and Technology from Arizona State University in 2019. Prior to graduate school, he worked at the Advanced Research Projects Agency - Energy (ARPA-E). His academic research has been supported by the National Science Foundation, the Bipartisan Policy Center, and the Future of Life Institute.

+ Read More
avatar
Alex Blania
Co-Founder & CEO @ Tools for Humanity, Worldcoin

Alex is the CEO and Co-Founder of Tools for Humanity, the technology company building tools for the Worldcoin project. He is a Co-Founder of the Worldcoin protocol with Sam Altman and Max Novendstern. Alex is responsible for leading strategy, development, and execution of the technology and tools enabling Worldcoin to ensure everyone benefits fairly from the opportunities AI presents.

Worldcoin is building the world’s largest identity and financial network as a public utility, giving ownership to everyone. The protocol is proving humanness and uniqueness in the age of AI and is creating a future where everyone, regardless of their race, geography, or economic status can fully participate in the global economy. Alex led the development of Worldcoin’s custom, privacy-protecting biometric imaging device, the Orb, which has verified humanness and uniqueness of 2 million individuals worldwide; World ID, which is a first-of-its-kind decentralized, privacy-preserving identity protocol; and World App, which is one of the fastest-growing and most accessible self-custodial wallets in the world.

Alex holds degrees in Industrial Engineering and Physics from the University of Erlangen-Nuremberg in Germany and studied physics at Caltech before dedicating his full time and attention to supporting Worldcoin. You can follow and interact with him on Twitter @alexblania.

+ Read More
SUMMARY

About the Talk: Worldcoin is creating a new identity and financial network that distinguishes humans from AI. This recording captured the the final in-person OpenAI Forum event of 2023 featuring Alex Blania, the CEO and co-founder of Tools for Humanity and Worldcoin, and Miles Brundage, head of policy research at OpenAI. Attendees learned about the Worldcoin systems designed to generate proof of personhood and ensure democratic access and governance of these systems in order to fairly distribute their benefits. We even featured the Worldcoin Orbs on site! 

About the Speakers: Alex Blania is the CEO and Co-Founder of Tools for Humanity, the technology company building tools for the Worldcoin project. He is a Co-Founder of the Worldcoin protocol with Sam Altman and Max Novendstern. Alex is responsible for leading strategy, development, and execution of the technology and tools enabling Worldcoin to ensure everyone benefits fairly from the opportunities AI presents. Alex holds degrees in Industrial Engineering and Physics from the University of Erlangen-Nuremberg in Germany and studied physics at Caltech before dedicating his full time and attention to supporting Worldcoin. Miles Brundage is a researcher and research manager, passionate about the responsible governance of artificial intelligence. In 2018, he joined OpenAI, where he began as a Research Scientist and recently became Head of Policy Research. Before joining OpenAI, he was a Research Fellow at the University of Oxford's Future of Humanity Institute. He is currently an Affiliate of the Centre for the Governance of AI and a member of the AI Task Force at the Center for a New American Security. From 2018 through 2022, he served as a member of Axon's AI and Policing Technology Ethics Board. He completed a PhD in Human and Social Dimensions of Science and Technology from Arizona State University in 2019. Prior to graduate school, he worked at the Advanced Research Projects Agency - Energy (ARPA-E). His academic research has been supported by the National Science Foundation, the Bipartisan Policy Center, and the Future of Life Institute.

+ Read More
TRANSCRIPT

I'm Natalie Cone, the OpenAI Forum Community Manager, and I'd like to start our talks by reminding us of OpenAI's mission. OpenAI's mission is to ensure that Artificial General Intelligence, AGI, by which we mean highly autonomous systems that outperform humans at most economically valuable work, benefits all of humanity.

So welcome to the OpenAI Forum's final in-person event of 2023. I think we've all come to get really excited about our Thursday cadence of events, especially with the ones where we have the opportunity to commune in person. So I look forward to picking these up again in January. And until then, we will be hosting some virtual events in the forum.

Next week, we'll be hosting Terence Tao, the mathematician, and the following week, Ahmed El-Gamal from Rutgers University. So we'll still be here for you just virtually, but tonight we are here to learn from two important members of our community, Alex Blania and Miles Brundage.

Alex is the CEO and founder of Tools for Humanity, the technology company building tools for the WorldCoin project. He is a co-founder of the WorldCoin Protocol with Sam Altman and Max Novenstern. Alex is responsible for leading strategy, development, and execution of the technology and tools enabling WorldCoin to ensure everyone benefits fairly from the opportunities AI presents. Alex holds degrees in industrial engineering and physics from the University of Erlangen-Nuremberg in Germany and studied physics at Caltech before dedicating his full time and attention supporting WorldCoin.

Miles Brundage is the head of policy research at OpenAI, which he joined in 2018 as a research scientist. He's been involved in AI policy research in various capacities for over a decade. And before that, he worked at the U.S. Department of Energy.

So welcome, Alex and Miles. And Miles, I will let you take it from here.

Yeah, so maybe one kind of framing for this conversation is, you know, helping people understand enough about WorldCoin so they can make an informed decision about whether to use the orbs afterwards, of which there are several outside. And so let's maybe start with the mission, and then we'll get into issues around, you know, privacy, security, implementation, and so forth. Why are you here, and why does WorldCoin exist?

Well, I'm here because I started WorldCoin three and a half years ago. I actually joined after Sam started six months earlier, and then started with tech, and later became the CEO of the company, Tools for Humanity. So we are 200 people, and we worked on this project for three and a half years. And why does it exist? We started with two big premises. One is, AGI is going to happen, and how can we build tools and infrastructure so that all of this kind of prosperity and wealth is distributed to all of humanity, all of society. And that is now a very common theme, but three and a half years ago, people made fun about that, as you probably know yourself if you work at OpenAI. And then the second piece is that we actually do believe that decentralized financial systems have their right to exist, and they will be empowering and delivering freedom to many, many people once they actually get to scale. But unfortunately, that did not happen. So these are the two things, like how can we get decentralized financial identity system to billions of people, empower billions of people, and then share the upside of AI with all of humanity.

Could you say a bit more about the kind of buzzwords that sometimes come up around WorldCoin, so proof-of-personhood and UBI, could you unpack those a bit more?

Sure. So proof-of-personhood is a pretty easy task, or it sounds like a pretty easy task, which means how can anyone prove to services on the internet that they are in fact unique and real human beings. And so in fact, when we started working on the project, and Sam had this idea, Sam had like a two-pager, three-pager describing the idea for WorldCoin, we did not plan to build orbs or anything like that. In fact, that was the last thing we planned to do. But we realized that one of the major things, if you, so one of the foundational ideas was okay, we can launch a digital currency and use the digital currency to align incentives around the growth of the network of real human beings. And so in some sense, PayPal spent $70 million to grow their initial network of people using the network. And one of the ideas that people in crypto try to make work since like 20 years at this point is what if you essentially would use such a digital currency to effectively use billions of dollars to bootstrap a really, really large network and get it to escape velocity. But that never worked because proof-of-personhood or civil resistance did not exist. And that's not the case only in crypto. But another kind of application that I think you're familiar with is just generally the existence of bots on the internet and the fact that they become increasingly powerful. So Twitter, for example, could use some system for proof-of-personhood. So that's one big thing that we have built. We spent a quarter billion dollars and three and a half years to solve the problem and really make it a way that is decentralized, eventually can scale to all of humanity, can work globally. And I think we will talk about why that is so hard and why we had to spend that much time and effort on it.

And then UBI, I think, is like something at least the opening people here think about quite a bit, is just the idea that we share a universal basic income of everyone. And in that group, I think that's not a problem. But especially outside of such a room, if you mention that concept, it relates often to political concepts around wealth redistribution and kind of rubs people a little bit the wrong way. But we are here because technology will change all of these things and we actually need to redistribute this wealth with people. So that's UBI.

And so essentially, WorldCoin does proof-of-personhood and an economic layer that connects everyone. So it can be infrastructure to distribute such a UBI independent of governments. And so could you walk through the process a bit and how the kind of overall system works? So starting from someone walking up to the Orb outside?

For sure. So you download an app right now. Only our app actually supports the protocol, but many others that you know will come soon. So at this point, you download the world app, it's called, in the App Store. And you have to verify once. So you go in front of a physical device. It's called the Orb. And it's a shiny Chrome ball, which was all not my decisions, but I think it looks actually pretty cool and it worked really well. But you go to one of these devices, hopefully if you operate in a city near you, and you show QR code. And what that QR code essentially does, it shares a public key. So you have a non-constituted wallet, if you're somewhat encrypted, you understand what it means. If not, it also doesn't really matter. And you kind of air gap that public QR code to the Orb. The Orb then has a lot of sensors in front and seven neural networks running in it to make sure that you're an actual human being. And you're not trying to fool the Orb or fraud the Orb in many different ways. Once that test is passed, we take a picture of the eye in the infrared, which is actually something that probably everyone here did many times before, because that happens at large airports. And then a neural network calculates a unique iris code out of this picture, signs it, and sends it right now to a centralized backend. At some point, it will be a blockchain. And now the really cool thing is that you as a user then prove that you're part of that group of people through zero-knowledge proofs, which is a kind of cryptographic technology that allows you to remain completely anonymous. So essentially, what it allows you to do is it allows us to issue a unique proof of personhood that you can use on the internet to verify yourself across different services while you actually remain anonymous, and also you are not traceable between the actions and between the platforms.

Yeah. And where are you at currently in terms of scale of users and throughput of the currency and so forth?

So as you probably saw, we launched July 24th after working on this for three and a half years. So that was a long turn to get there. Thank you. And so that launch was much crazier than we thought in pretty much every dimension you could think of. So we actually sent half of the team around the world, like the HR team was with Orbs in Tokyo, and trying to sign up people. And we had long, long lines in front of these Orbs, sometimes thousands of people. So it was really, really hard to manage. And pretty much every guy...

Government, you could imagine, reached out to us. So that's, at this point, a more policy person than a tech person. But in terms of numbers, we are now at 2.3 million verified users. Within the app, around 4 million users. And while it grows quickly, we are only using 300 devices. And the goal is to get to 50,000. So hopefully, when we talk again next year, it will be a couple million a week to join.

BRAD FITZPATRICK Oh, congrats. That's a lot of progress. Could you say more about how you explain the security and privacy features to these users? How do you assure people that their data, and their potentially very sensitive biometric data, is safe? And you went through the explanation, but you run into criticism. And how do you respond to skepticism?

Yeah, so it really depends on who you're talking to. Because as you could imagine, an average user will not talk about their knowledge proofs and all the details. So they would just have a form that explains everything. And in fact, actually, the biometric data is not stored. So the images are not stored by default. You can help us to improve the protocol. And you can submit these images if you want to. But by default, that's turned off. And so that's the first part of the question. There is actually no data custody.

And then with tech people, it really matters to explain their knowledge proofs. I think that's a very important piece. And then you get all kinds of pushback about quantum computers. I did quantum computers before that, so that's actually not a thing. But it really depends. I can go more specifically. I'm sure at least some folks would like to briefly hear why quantum computing is not a thing in this context. Then we'll go back to more layman's terms after that.

Well, for one, I think we're far from making them work in such a way that they could actually crack cryptography of that level. Two, there is quantum safe implementations of such codes. So it's like a double stupid argument.

Say a bit more about how AI fits in. So we're at OpenAI. And in some sense, AI is kind of creating the problem and the solution of identity being a problem online, and then also wealth being available to distribute. How do you think about that? And what's WorldCoin's view on AI? Are you kind of hoping for this all to work out for the better? Or do you view it as a problem to be solved? Yeah, how do you think about all that?

We don't think about it as a problem. Obviously, AI creates so many great things. And the fact that Twitter has a bot problem does not at all outweigh the upside of this. So of course, we're very optimistic, given also Sam is a co-founder of this. And how it fits in really depends on the timescale. Meaning in the short term, probably proof of person is actually interesting. It's interesting to online services, online platforms. We talk to many people on Integrate once we get to scale. So in the short term, next two years, I think the narrative you will see emerging is kind of a decentralized human gate for services and where we need it.

Over the medium term, and some of which we talked about many times before, is, I think, decentralized governance and global governance over AI systems. As I noted, it's something that OpenAI thinks about quite a bit. So I think that is something that WorldCoin actually allows to do at a large scale. And then, of course, UBI. First, UBI experiments. And then at some point, global UBI distribution. And if this succeeds, and it very much is a moonshot, and there's many reasons why it will not work and why we will fail. But in a world in which it works, it will hopefully empower billions of people and be the largest financial network we have on this planet.

Yes, we'll get a bit more into potential failure modes a bit later. But for now, just kind of on the promise and the kind of mission, say a bit about what's in it for someone who maybe does or doesn't have a banking system, where they live, or financial access, and how you would kind of pitch the technology to different people.

When we tested WorldCoin, we made somewhat sure that we actually cover pretty much all geographies. Boy, did we get much flack for that. But I think it actually was very important. Meaning we went to, in the super early days, myself was literally traveling to Tromsø and signing on people there, or Tiago, with an orb, and kind of in Norway, to Nairobi, to Argentina, Chile. And so we really, over the last three years, I think we talked to users in pretty much every major geography.

And the first thing is, if you split it apart into the different fields it touches, crypto itself, here in the United States, obviously, I mean, also then last year, it was not a great year for crypto in many ways. So it really didn't help. But even before that, people in the United States don't really use crypto. It's just not a thing. I think swapping NFTs and JPEGs, that was not the original idea. But it looks very different if you're in Nairobi in a university, or in Nigeria, for example. Because there it actually solves problems in a meaningful way. That's something, somehow, like in Brooklyn, crypto conferences people don't like to talk about. But that is actually where people understand that part of the project very viscerally. Like you have a government currency that is failing. You have a financial system that doesn't work. So what I'm trying to say is, there's a large part of the world that actually understands the financial piece of the technology. While in the United States, or in Europe, I'm from Germany, especially the proof of personal part, I think, is much more interesting. Because people understand AI, understand all of these things that it will do. So really, I have 60% travel time at this point. And really, wherever I go, it's different things that resonate with people, which I think you experience also with AI.

Yeah.

Yeah, so certainly a lot of people get it. Other people are more skeptical. And you've run into issues in the US, in particular. Could you say a bit about that? You have a Chrome Orb that tries to distribute a global digital currency and solve proof of personhood. So yeah, there's many people that are skeptical. And that's to be expected. And we actually expected that from the early days.

But one of the major things when we started this company, it's actually literally my first interview, I think. Sam said, the central failure mode is that no one cares. I think at least we solved for that. People really care. And that is actually a big deal, the fact that everyone in so many countries knows about the project already at an early stage is a pretty big deal.

And so I think what now matters over the coming months is the technology is actually privacy-preserving. I really do think it really matters. No one else was crazy enough to spend a quarter billion dollars in four years on building something like that. So there is no competition. Meaning if we fail, it's very unlikely that something else like this will exist, which rarely happens. So at this point, I think it really is just about explaining what we do, why it is a good thing for the world, decentralizing everything, open sourcing everything. And I'm sure it's going to be amazing.

And so part of the reason why it wasn't happening before is that it took a lot of upfront investment to build the hardware and design the system. Could you say about why you chose that particular route of custom hardware and iris scanning as the kind of specific way of doing proof of personhood?

Essentially, the first year of the project, we really just started with, where do we want to go? And that kind of solving civil resistance will be very, very important to use a token to solve the bootstrapping problem. And it can actually grow to billions of people. And we actually did a lot of research. So I mean, I did physics before most of the founding team comes somewhere out of science. And so we looked at all the different options you would have fundamentally to solve such a problem, which you can broadly group in using government identities, so KYC, using social graphs and reputation systems, so using some kind of algorithms and ways to issue trust towards each other and kind of build up a scalable graph out of this. And then last but not least is biometrics. So you fundamentally measure what it actually means to be a human being. And we built implementations for all of these.

And the thing is, government KYC, I think, for a lot of reasons, is a really bad idea if you really think about a global scale, because it might work really well in some countries, but it won't.

will never work globally, and many governments will not be able to react to all the things that are happening in AI fast enough. So that was not really an option.

Then we went into social graphs. I met with many of the teams of the large social networks that you know, and came to a conclusion that they actually tried to build these implementations for many years, teams at Facebook, teams at Meta. And as soon as any economic incentive comes in, it's very easy to break the system. So in short, we never have seen a scalable implementation of these things, unfortunately.

However, you could actually think that if you have a large enough seat of trusted people, you could maybe even bootstrap something like that, which is something I know you want to talk about.

And then we went to biometrics. And within biometrics, you need to first understand that all the systems that you usually use, so Face ID, for example, does a one-to-one comparison. So there is an embedding stored on your phone of how you look like, essentially, a 3D map that is then compressed into a small mathematical embedding. And when I use my iPhone, it creates a new embedding measures that to a previous one of the measures I can use my phone. So it's a one-to-one comparison. If, however, you want to solve proof of personhood on a global scale, you need to compare one user against everyone else. So it goes from one-to-one to one-to-N. And that is an exponential explosion in complexity. And so you actually hit a wall with anything else that is not high-risk biometrics, because high-risk biometrics actually has a lot of information in it, a lot of entropy, mathematical entropy. Face could not do that. Fingerprints could not do that. So with the ambition of actually reaching everyone on this planet, it's the only way to do it.

And so just to circle back on the bootstrapping point, so obviously, there are other perspectives on proof of personhood. Some people believe that it's needed, but might want to take a different approach. So Vitalik Buterin, who I know you have discussed this issue with a lot, who co-founded Ethereum, wrote a blog post recently talking about the different approaches, and ultimately kind of landed on the bootstrapping approach of start with biometrics, but then use that as a seed for social graph and other kind of approaches. Could you unpack his perspective and say what you disagree with?

Yeah, his perspective is easy. His perspective is that distributing hardware globally is just a really hard problem. And so if we believe it's as important as we do, if you would have some way of kind of using a biometric way of issuing that proof of personhood to then switch to a social graph proof of personhood, that might be a good thing. And fundamentally, I agree with that. We discussed this for many hours now with him. And I think that's true. And we actually have people working on these things. I'm a little bit more pessimistic in that sense that I actually do think, again, AI will make that really hard. Because agents will be able to build up such a reputation very rapidly. So I just think it's a losing game. It sounds amazing. So that's my take.

And you mentioned KYC slash kind of government-issued IDs as an alternative. Why take the more decentralized approach and biometric approach rather?

Or yeah, the kind of decentralized approach rather than doing the biometrics through governments?

There's many reasons for that. I think the easiest or kind of the most obvious one is that just for 45% of the global population, there is nothing like a digital identity that you could actually use for such a purpose. Large-scale government systems already start failing under actually very unsophisticated kind of deepfake attacks and things like that. That's one. It's just accessibility, if you really think this has to exist globally.

And then two, one of the things, for example, as you read Vitalik's blog post, is if this becomes as important as we think, there's no reason why you should trust governments, that they would not actually attack such a system at geopolitical scale. So decentralized systems are the only answer we have to build something on a population scale that could sustain something of that importance.

Could you talk a bit more about the failure modes of this particular path? So you talked a bit about the failure modes of alternative approaches. What are the things that keep you up at night in terms of WorldCoin's trajectory?

Oh my god, there's so many things, as you can imagine. There's many things that keep me up at night. The last three months have generally been the craziest three months of my life, I think. So right now, it's a lot of policy, meaning we meet a lot of government leaders and respond to many regulators much more than we expected at that stage, because the launch was just much crazier than we thought. That's kind of a short-term problem.

If we talk a little bit more abstract about the project itself, it's, for one, just operational challenge of getting from 300 devices to 50,000 devices globally, making it accessible. It will be really, really tough.

Second, engineering. So Vitalik actually laid out many things that we will have to solve around decentralization. It's a tough one. Decentralized protocol on top of a hardware standard is really hard to build.

Scalability. So there's many engineering problems that I think are all solvable, but it just requires perfect execution over a long period of time.

And then the last is adoption. It's like, get other major players in tech, governments, and eventually people to adopt this and hopefully find it useful. And so lots of different challenges.

Is there one that you'd pick out as saying this is the hardest problem that you have to solve?

I actually do think it is, for the team, it's write-up policy, because it's just a new muscle. I mean, I did theoretical physics before, not talking to presidents. So that's just a new thing we have to learn.

And on engineering, it's definitely decentralization, because it's a relatively short list with at least two fundamental research problems that no one has solved yet. So I think these are the top two I would pick.

And what can people actually do with the technology today, and how is that going to evolve over time?

So we are super early, meaning 2.3 million people is for a startup. That sounds cool, and everyone always cheers us on. But actually, that's not much for what we try to do. Meaning, you don't see much local density, or you see kind of whole, whatever, 30%, 40% of countries being verified. So that means that many of the super high value use cases are probably a year out.

Like, we're right now very much in the bootstrapping phase of this network. So as a user, you actually receive ownership in the network. Like, the whole point of the token is that it launches by giving ownership in all of its users. So that is fundamentally a bootstrapping force to actually get people going.

However, services can integrate World ID already today. Users can use it to log in, and kind of actually use the identity piece of that. And in November, we will actually have a lot of exciting updates there around both the protocol design, integrations, etc.

Could you say a bit about WorldCoin organizationally? Like, what are the different teams that are working on this? Sounds like a lot of engineers, also policy people, or engineers doing policy stuff. Somehow you're doing policy. And what are the other kind of projects that you have?

And I want to ask you to comment on one specifically that piqued my interest is red teaming the technology.

So the company is now around 200 people, 180, 200 people. We have this super weird background that during COVID, many of the founding team was locked in Germany. So I just lost my student visa, and then Trump locked the border. And so I was literally locked in Germany. And so we flew many people from San Francisco to this random town in Germany and built the company there during COVID. So that means we have 50% of the company right now is still in Germany. Turns out Americans love Germany, some parts of it. So that's one piece. So we have two major offices, San Francisco and Berlin, and then another one, engineering and manufacturing in a small town called Erlangen, which is where we actually built the thing.

And in terms of teams, it's very engineering heavy still, meaning I think 60% roughly is engineering. From engineering, most of it is still hardware, manufacturing, firmware, things like that. We have a AI team, which is like, oh, my god, hopefully they don't see that. It's not the kind of AI you do here. It's like it's.

It's deep learning to kind of do image recognition and things like that. But of course, it's very sophisticated stuff as well.

Then we have a protocol team. We have a legal team that is pretty large for the size, as you can expect. We have a one-person policy team, which is part of the problem. And then we have biz ops, finance, and all these other things, HR people that you could expect. And then actually, what is interesting, there's a nonprofit foundation that actually holds all the IP. So the company is a Delaware C-Corp with a German subsidiary for that reason. And then there's a nonprofit foundation that actually holds all the IP. And I really actually do not at all control.

All very cool things, but I still want to hear more about the red teaming part. We have an internal team that does these things and does all kinds of weird things you could imagine. So like animal eyes and mannequins and all kinds of weird shit. But actually, most of the red teaming happens with external companies. So we have a lot of audits ongoing pretty much all the time. We hire kind of white hat hackers to attack the protocol all the time.

It was one of these things, like, in the first two years, I thought that would be, I would expect one third of the company to be a red team, essentially. It actually did not turn out to be that way.

And is that because you can outsource a lot or because it's a very secure technology, or you can break it, but it's insanely expensive and not worth it?

I expected hardware security to be much harder for us. So for example, other than super crazy attacks where you actually need kind of a full on lab and you kind of have very, very sophisticated lab setups where you can actually attack the NVIDIA sensors, no one really succeeded with attacking the orb. So that was something I overestimated.

For what we call presentation attacks, so people actually showing something to the orb that is not a human, the attacks are way less sophisticated than we thought, meaning it's dogs and all kinds of things and rotated dogs and two dogs and all kinds of things in that direction. However, we very much expect that to change as the economic load on the system increases.

That's kind of the internal red team is mostly around presentation attacks. One question is, again, we're at OpenAI. How do you think about the potential for a partnership between our companies?

I do think there's three main, actually four main things. One is fundamental kind of almost like policy research around proof of personnel itself, meaning we work together in outlining what are the different options and kind of why does it matter and educate people around it. I think that's one. That's the most straightforward one.

I think two that aligns with current efforts of OpenAI, as I know, is kind of democratic experiments, democratic governance, which I think WorldCoin can really solve on a large scale.

Three is UBI, UBI experiments, first on a smaller scale so we actually understand everything, and then later on a larger scale, which is actually why Sam, a big part of why Sam started the company in the first place.

What was the fourth one we discussed yesterday? Identity verification, maybe?

Right, identity verification in your services, if that turns out to be a problem.

Say a bit about what would outrageous success look like and what does a day in the life of a 20, 30 person wakes up in the morning, how is their life better because of WorldCoin?

Outrageous success, I think on a three to five year time frame is that WorldCoin does something like Facebook gets to social networks, meaning it's the first technology like that that actually scales to billions of people, and as a result, empowers billions of people in a way we just did not see yet from such decentralized technologies because there's a huge community around these things.

But fundamentally, the big use cases we have not seen yet, and I think a big part of that is actually just network size. So that is one. You can really measure our success, I think, in the coming years around how fast are we able to scale. That's just fundamentally, I think, the whole point.

And I do think the craziest thing I could imagine is that WorldCoin itself becomes some kind of digital reserve currency itself. People actually transact with that, receive some form of UBI through that, and use WorldID to verify themselves online as they progress.

And I think perhaps the most important thing that is not happening in that scenario is people starving, not having an income, having elections overrun by bots, and so forth. So I think in some sense, that's kind of the flip side of the vision. Is that how you think about it?

Yeah, totally. Last question, how can people learn more about WorldCoin?

There's many people here, so you should just talk to them. I think that's the P0. And then two is we have, frankly, it's like 130 pages, unfortunately. A 130-page white paper that you're very welcome to read. It's not an easy read, but I think that's probably the best resource, and we constantly try to keep it up to date. If there's big changes to that, we do blog posts as well, so just like Follows on Twitter. I mean, Sam retweets, I think, most of the big ones as well, so you will likely see them.

Yeah, and I would make a plug for Humanness in the Age of AIs, did I get the name right? It's a great, great post.

Thank you. Cool, so I think we will now open up for audience questions.

Hi, I'm Artemis. I do policy at OpenAI, so if you ever need any help on the policy front, let me know. Just kidding. So the question is really about use cases. And forgive me if I don't formulate it very well, but hearing about your use case without being super familiar with the technology or your product before, the first thing that comes to mind is, oh, very cool. What this is going to do is counterbalance the risk that we won't know what's human versus what is AI-generated. It will provide a way for us to be able to tell. However, thinking about it a little more, the risk I see is that presumably you need to have a way to give people the right to use this technology but not the right to transfer it. Because if I can transfer this technology, then that means that A, I can give it to a whole lot of people, and B, that no longer means that that is associated with them. Specifically, it's just associated with a human IME. So how do you ward against transfer risk? And relatedly, what is the connection between this technology and centralized systems? So I can see this working really well in a government environment, say, where the government knows who is in scope for some kind of benefit and also has a way of identifying a unique identifier. Should only be one hit, right? But in a decentralized system, what stops me from using the system multiple times to get the same benefit if there is no centralized way to impose a unique identifier?

So to the first question, that is actually, there is a very good analogy to what we do in a centralized manner, which is the Aadhaar system in India, Aadhaar and UPI. And actually, one of the founders works with us, one of the founding CTOs. And so the hardest part is solving deduplication, which is what we kind of talked about at length, which is, everyone gets a unique identifier. Afterwards, you have a one-to-one problem. So how do you make sure, as you said, that someone does not transfer that image?

And how we actually do that, and we shipped this in November, is a local face auth. So meaning at sign-up, you get a face embedding on your phone. You self-custodially hold it. And whenever you use it, you actually have a face ID problem. So whenever you sign, you sign with your face embedding locally without it going through a centralized server. That's one.

However, that can be attacked with emulators and all kinds of things. So there's only so much security it gives. So you will likely have to reset once a year or something like that, depending on how crazy things get. Actually, that prevents selling. If someone steals it or if you lose it, you can just recover it with an orb. But actually, local face auth prevents you from selling it, because you will have to, at least you have to have the actual face, the face ID of someone to use role ID when you sign it. Does that make sense?

Actually, right after, you should go to send role, which is right behind you. A lot of folks from the company in the room.

Next question, and then there's another one over here.

Hi, I'm Dev. So I'm actually a founder of an agent company. And one question I have is, suppose in the future, you have a user who wants to authorize an agent to interact with the service on his behalf.

HP, like say DoorDash or Instacart, something like that. How do you think the proof of identity will work in that case? So essentially, the main thing this does, it gives a cap to how many bots or agents you can deploy. I think like, how we, I think all of us believe this will play out is that pretty much everything we do will be supported by some, by at least one agent. And so this does not stand in the way of that. You will just have to authenticate, okay, I'm now using that agent to do action, X.

So, follow-up question. So do you think this will be on the side of the merchant where they have to integrate, like, do we support this agent or not? Or will this be some, mostly on the user side?

That one, I honestly, I don't think any of us knows that yet how that will work. Yeah.

Hi, thanks for coming. My name's Julia Godinez, and I'm a fintech policy advisor for the Federal Reserve. And I see so many implications and potential use cases for traditional finance, non-traditional finance, especially with the proof of personhood and mixing zero-knowledge proofs with biometrics and encryption and all that. But what I'm not seeing is the link from a competitive economy, capitalistic structure, to then this smart, circular economy, which would be the promise of initiating a UBI, leveraging this technology. So can you talk about that link? And, you know, let's say you get scalability, and everybody has it, but how do you transition to a smart, globally-linked, circular economy? Can you explain more what you mean by circular economy?

Sure, so with UBI, universal basic income, the idea is that you would have equally distributed amongst the global population. So the implication of that is that we are all equally entitled to an equal share of whatever the world pie is, whatever the land is, whatever the everything is, all the masses, everything. So that's what the implications of a basic income would be. So how do you make that link to even linking this technology to universal basic income?

Well, a company like OpenAI, for example, could decide that it wants to distribute a certain percentage of its revenue with all Rokin users. That's pretty much it. So it doesn't come out of the system itself inherently, but what Rokin does is it at least gives a way to reach all of the Rokin users through identity and both financially. So we are not solving UBI in any way or shape. That has to come through other technologies like OpenAI. So this is more like the method to distribute the UBI.

Correct, it's the infrastructure to do that, yeah.

Thank you. Just to follow up on that, you are kind of distributing a little bit. Is that right? On like every two weeks or something? But you're saying eventually the money will run out, there needs to be some kind of economic productivity boost from AI in order for it to be a large amount. That's kind of what you're saying?

Yes.

Hello. I'm interested in the geo differences that you talked about in the very beginning. So you said when you launched, you went to all geos and some give you flack. I would love to hear who gave you flack for speaking to whom and for what?

It's interesting actually. So like on the user side of things, adoption was pretty uniformly actually. So like the first sign-up actually happened in Berlin and in Germany. Berlin and Germany is like, as you would expect, is like very conservative around many of these things. So on the user side, it's like you don't get much flack. People ask questions, sometimes critical questions, but in the end, they either are excited or not and leave. So a lot of the discourse that happens is of course with journalists or on social media. And broadly speaking, I think most of the debate actually happened in the US. In Asia more broadly, I think it was just broad excitement. I think India also, a lot of excitement because like Aadhaar already exists. People are familiar with like many of these kind of implications. What else matters? Oh, you have like these local clusters, for example, Argentina, where you have hyperinflation and because of that, you have a high crypto penetration. And so then let's say Argentina, they are excited about not for personal, but because it's a crypto wallet. And so then that is very surprising to the team. So you have like local clusters depending on what are the circumstances.

Thanks so much for the presentation. I met Santiago Siri, who's I think one of the advisors who started Ubicoin along like 10 years ago. And they were doing this thing in Argentina where you'd record a video of your face holding up your Ethereum code and that would like verify your personhood. That seemed sort of sloppy, but the iris scanner that looks like a robot's eyeball scanning my iris, like that's scary to me. It's hard, like I really like the universal basic income part. And it's hard for me to imagine that a lot of people will be okay with the iris thing. Maybe I'm wrong, but is there some alternative way? Do we have to have the iris scanning bound up with the universal basic income part?

Unfortunately, I think yes. I do think you will see, you will see different forms. Like, I mean, I totally get that a Chrome orb is somewhat provocative. And I do think like a couple of years in the future, you will probably see other, hopefully other tech companies building completely other form factors. And the thing is like, what is interesting about it, I understand the instinct. I mean, I had the same when I actually started working on the project, but there's companies like Clear, there's of course governments with way worse kind of security implications using the same technology. So I think a lot of it is really just education, understanding what this actually means, and then the whole stack being fully open source. So that people can actually really trust it more than they could ever do with a private tech company. So I think like I cannot, yeah.

So just to dig a bit deeper on the transparency thing. So, you know, you mentioned that, you know, the biometric data never actually leaves the device, it's just a hash. You know, why should people believe you? Like what if like an orb is swapped, or, you know, you release the plans, but actually there's, you know, some different engineering. So how do you kind of build that trust?

Well, there's a lot to that. I mean, it starts with fully open sourcing the whole stack, which right now we did not do. Meaning most of the hardware is open source, but for example, the security critical pieces of the hardware are not, for obvious reasons. Because we're kind of still red teaming that part, and we'll release it later in the year. But I think that's the first step. Step number two will be to kind of roll out some kind of auditing scheme, and economic incentives around that. So you kind of have globally in the countries you operate, you actually have audit firms that are incentivized and paid for just like going around, checking these orbs for validity, and killing them otherwise. So that's the direction we will take.

And are you able to cancel a world ID if it turns out that it was generated from a bad orb?

Yes.

I have a question. I haven't had a chance to actually check out the orb itself. I saw it from a distance, but I'm curious what sort of accessibility features are enabled in it for folks with disabilities?

That's actually something we did not address yet, unfortunately. It actually affects an ultra small percentage, because even most blind people have still a functioning iris. So it really affects a very small subset of diseases. And the only way we can actually solve that is there's some kind of oracle, human oracle, that has some amount of tokens, some kind of role IDs, and can distribute them to people with disabilities. So that's something we need to roll out in Q1. As you can imagine, it's not the highest importance, but it truly matters. So we'll get to that.

My name is Sanjit Dung. Can you talk about the developer ecosystem that you intend to build maybe over the next few years after you have certain user base? Because you're gonna have a lot of privacy related information, and developers could build fantastic applications on top of these.

Yeah, I mean, that is a very similar problem to kind of all major decentralized applications. Protocols and tech, right? It's like, I do think what will make Rokin very unique is that there's a large user base at the other end of it. Like usually, if you build such things, it's really hard to actually build a functioning business on such things, on crypto. And they kind of need to hold space. So that's mostly why developers are really excited about it. But then once it's at scale, it of course unlocks a lot of completely new things that right now we just cannot build.

In the financial space, under-collateralized lending, et cetera, et cetera. So many, many things that I think people will actually build companies on top of that in the coming years. So how to get there is, I think, just solve a very unique problem, get to a meaningful scale, and the rest is probably easy.

I had another two really quick questions. One of them is actually another developer question. Can you talk about the cost profile for verifying somebody's identity? If I'm a developer, we actually have this problem at a massive scale at OpenEdge. It turns out if you give people free compute, they want to use it to do potentially bad things and malicious stuff. So we try to figure out who is a real person, who is a bot. And being able to actually solve that problem would be immensely useful to us. So how much does it cost to verify one person through the developer platform? One analogy, I think that the short answer is $6. The longer answer is that I think the coolest analogy that I've heard actually from Vitalik around this is that Bitcoin uses billions of dollars a year, essentially, to secure the network, solve a really hard computation problem. And as a result of that, the network is somewhat secure. We have never seen such a protocol to allocate similar resources to actually scaling the network in terms of real human beings. And so I think if all of this works and we're going to look back at it in five years, I do think that will be the one thing that will have led to all of these profound outcomes, is that effectively, the network right now is valued at $15 billion. And 80% of that will go to users. So you will have tens of billions of dollars allocated around the growth of the network. And what this means is if you use WorldID, obviously, you will not have to pay for that, but rather actually the protocol itself and somewhat pays for it.

And one other quick follow-up, have you thought about in a lot of the countries where there's pushback around the iris scanning, have you thought about changing the form factor of the orb? I feel like that's where a lot of the distrust comes from. You see a little chrome ball, and you're like, this is just kind of weird and futuristic. But if you make it a pink mannequin or some other weird random form factor, I feel like people would just inherently be more trusting, potentially. So yeah, yes. And actually, that's coming. The next version of the orb comes in different colors and looks very friendly and nice. Still, there's going to be this kind of ball, the chrome version. So yeah, this is coming for sure. But the really interesting piece is when we started testing, we had every Saturday, we had shipping Saturday, essentially. And every Saturday, we would go out with a new version of the orb. And Sam or Max or any of those two hired this designer that worked with Connie West that came up with this crazy idea for an orb before my time. And at some point, we just, in the beginning, it looked like a bowling ball. And it was blue, and it already was a sphere. And I signed people up. And if you know me, I'm not the most extroverted person. So this was inherently a failure mode. So it didn't work really well. And then we started changing the pitch. We started changing the color. And we went to this chrome. And it actually started suddenly working. So the thing is, the pushback or this, oh, this is so weird, that's mostly actually a Twitter phenomena. It's not something that you actually see when you stand next to an orb in the operating markets. So in some sense, I do actually think that's a big reason of why it works. It's just like it's so weird and so futuristic that some people are annoyed, and others are super excited. And I think that's what really matters at that stage of the project.

Thank you so much, Alex and Myles. That was really lovely. And it was such a pleasure to have you here, Alex. Thanks for having me. Thank you. Thank you.

+ Read More
Sign in or Join the community

Create an account

Change email
e.g. https://www.linkedin.com/in/xxx or https://xx.linkedin.com/in/xxx
I agree to OpenAI Forum’s Terms of Service, Code of Conduct and Privacy Policy.

Watch More

59:10
AI Art From the Uncanny Valley to Prompting: Gains and Losses
Posted Oct 18, 2023 | Views 34.6K
# Innovation
# Cultural Production
# Higher Education
# AI Research
OpenAI's AI Trainer Community Mixer with Special Appearances from Research Leadership
Posted Jul 12, 2024 | Views 11K
# AI Research
# Expert AI Training
# AI Safety
Red Teaming AI Systems
Posted Mar 08, 2024 | Views 17.2K
# Expert AI Training
# AI Literacy
# AI Research