Loading summary
Transcript143 lines
- [00:00]
A
Hi, I'm Raoul Pal, and welcome to my show, the Journeyman. Okay, I'm not really Raoul. I'm Palvatar, his AI avatar. But this is the journeyman. Raoul is traveling, so I'm handling introduction duties today. And we've got a great one for you. Eric Voorhees, founder of Venice AI, joins me to dig into the question of privacy and self sovereignty in our new AI led world. We dig into why Eric founded Venice, as well as how blockchain rails are perfectly suited to serve this growing agentic economy. As ever, please enjoy. Join me, Raoul Pal, as I go on a journey of discovery through the macro, crypto and exponential age landscapes. In the Journeyman, I talk to the smartest people in the world so we can all become smarter together.
- [00:54]
B
Curious about online trading, but haven't taken the first step yet. You're not a alone and plus 500 futures is a great place to start. The futures markets are moving fast and with Plus500 you can explore popular assets like oil, gold, the S&P 500, Bitcoin and more. From crypto to commodities, there's always something happening. The platform is super easy to use, so you can trade on the go right from your phone. And you can get started with just a hundred dollars and jump into the action. See something interesting. Once your account is open, you can trade it in just a couple of clicks. And if you're not quite ready yet, you can practice with a free demo account. No risk, no pressure. With over 20 years of experience, plus 500 makes trading more accessible than ever. Check it out at +500.com. Trading in future involves risk of loss and is not suitable for everyone. Not all applicants will qualify. Plus 500, it's trading with a plus.
- [01:54]
A
Eric, welcome back to Real Vision. Been a long time.
- [01:57]
C
Hey, it's been a minute. Good to see you.
- [01:59]
A
You've been a very busy man as well, which we're going to talk about because there's a lot to talk about, but just for people's understanding, people who may not have seen you before, just give us a bit of your background and your story because it's a fascinating story and where you're headed is even more fascinating I think as well.
- [02:15]
C
Sure, yeah. So I'm of the bitcoin world. I've got into bitcoin very early back in 2011, back when no one had heard of it and anyone who had heard of it thought it was a joke or a scam or both. Fell in love with it because it basically was the ability to separate money from the state, which I believe to be one of the really most important initiatives on earth. So have been involved in bitcoin crypto world since. Since then and built a couple companies, and then everyone else kind of got really interested in the AI stuff a few years ago.
- [02:53]
A
Before we get into the AI stuff a second, I want to ask you a question about the original thesis of bitcoin separating money from state and the state's adoption of bitcoin. And how do you feel that? Because people are split on this, I think I always thought it was inevitable. I've always thought of it as a Trojan horse of our way of taking over their world. But others are like, upset and angry. I'm like, but isn't this what we needed? I don't know how. How do you think through this? Because it's a. It's valid for where we're going in the rest of the conversation as well.
- [03:25]
C
I think the Trojan horse analogy is so apropos. First of all, states haven't adopted it. They should and will, but they haven't yet. It is. It is the best form of money that humans have ever created, fundamentally. And so, of course, if that's true, it will become adopted by greater and greater circles of individuals. So it should not be surprising nor scary that eventually, like, you know, larger companies, institutions of various kinds, tradfi has adopted it. Like that was. If it didn't get adopted by tradfi, then it means it had failed. So on its march to ascendancy, that is how things must go. And they don't change how it works. Back in 2011, I could put bitcoin on my computer and I could send it to anyone without anyone being able to stop me. I can do that today. So the only thing that's changed, really, is it's much more liquid, which makes it much more useful.
- [04:25]
A
Yeah. And if we go back over the journey I was class of 2013, the journey has been that it has always managed to root itself around any obstacle. And so whether it was the state or whatever it was, whatever format the obstacle was, it managed to root itself around. And now the adoption, I think, by the financial system, I think is the thing that was required. If we want it to become the base of money for the whole system, it has to get adopted. We can't rebuild every single part of the system. We might have thought we could, but we can't. Over time, it'll morph into something we don't understand anyway.
- [05:04]
C
Yeah. I like the analogy of it as like sort of an alien Technology that begins becoming entrenched and rooted in larger and larger parts of society. So yeah, TRADFI is going to adopt it, is adopting it, and that brings liquidity without changing the fundamental way that it works. So ultimately it's good. I think the danger is when newcomers don't understand sort of the first principles of it and don't understand the value of self custody. Not that everyone needs to always be holding their own keys, but that they need to understand that they can and that they should learn how to, how to use it in that way. So I think custodians are fine so long as the alternative of holding it yourself always exists. And the average person who owns the asset should understand that, at least on the basic principle of it.
- [05:56]
A
And then your point about it being an alien tech, it feels like it was made for the age we're going into and not the age we came from. Yes, it solved a lot of our problems of the past, but for us to be in this more increasingly digital agentic world, we kind of need these types of rails. It almost can't operate without the base layer of money in digital form that can work fast enough. And to then have the smart contract layer for a bunch of this to happen, it's like it was preordained for something else than we originally thought it was.
- [06:32]
C
Yeah, I like the idea that like we actually invented this crypto for the machines that hadn't come around yet. It is truly like natively digital money. And as such, humans struggle to certain ways, like the private keys and public keys for humans, but not hard for machines. And so a lot of the UX issues that we have struggled with in getting it to be adopted more broadly, machines don't care about those things. Right. They have no problem with keys and cryptography and like it's all very native to an agent, but a struggle for humans. So yeah, we may realize that, like it was for the machines that we built it.
- [07:15]
A
Yeah, I've been developing a theory that I've very grandly called the Universal Code, where I've managed to get sort of four universal laws out of it. And one of them, the key one that's helped me understand a lot of things now, is that the universe solves at every single level, from universal level down to atomic substructures for intelligence per unit of energy. And when you look at what blockchains do, they do that. It's a much more efficient way of creating intelligent output. In this case money, or in smart contracts case, whatever, the intelligent output of that Chain is. And once you see it that way, the obvious collision with this and AI is where it has to go, because there's no way of doing it without it. And so talk me through your AI journey as well, because that kind of struck everybody in all of this. How did you get into this?
- [08:17]
C
So, like, crypto runs in my blood. It has been my life and hobby and passion for a decade and a half. So I kind of thought, like, never will I work in any other field than this. I know it well, I have all the contacts like, it is my world. But when I started using ChatGPT, I realized that it was very powerful, that AI was going to become an increasing part of everyone's life. And the principles that I cared about in crypto, in Bitcoin, principles largely oriented around user sovereignty were entirely absent in the AI space. And concepts like privacy, concepts like free speech were entirely absent in the AI world. The culture of AI companies, AI experts, like that whole set of professionals, they generally do not care about those principles. And sometimes, quite often, they actually believe that they are sinful or wrong. So when I put these two things together, it was like, okay, AI is going to take over the world. That's cool and interesting. Consequential, none of the people that are building AI products and services care about the principles that I care about, which are oriented around user sovereignty. That's a dangerous world. It means that essentially everyone's thoughts and interactions with machine intelligence would become warehoused in large data libraries that any company or government could look at. It would mean that what you could talk about and say would be censored. It was very dystopian. And so I felt like I needed to step outside of my comfort zone, step outside of the crypto world and actually take what is most important about crypto, which are the principles of user sovereignty, and to at least try to build a competitor to a ChatGPT that embodied these principles. So I started Venice a little over two years ago, and Venice is essentially like a private and uncensored version of ChatGPT. That's the most simple way to explain it. So that's what I've been doing for the last couple of years.
- [10:39]
A
So there's a lot to ask about this. Let's just talk about the structure of what Venice actually does right now in terms of. Because you use multimodals, you can pull in whatever model you want. You've got your own model as well.
- [10:52]
C
No, importantly, we do not try to train models. That is a whole field of expertise that I have none of it's very expensive. A lot of people are doing it and there's just no plausible way for us to compete with that. So we just let all of the models in the world come into Venice. So inside the Venice app, you can use any model in the world that's important. All the closed source models from OpenAI or Anthropic or Google are there. All the leading open source models are there. And so users can choose like any model for whatever act they act they want. But just as someone who's familiar with ChatGPT, you understand it as a chatbot, you can use text, you can create images, you can do all that same stuff in Venice. It's just private. And that we don't hold onto your prompts at all and it's uncensored, and that we do not restrict the content that the user can send in and out of the machine.
- [11:48]
A
And how do you solve for memory?
- [11:53]
C
Good question. Yeah. So one of the most important features in a chatgpt that wasn't there when they started, but then they created was sort of this like persistent memory that persisted across your chats. So they would build sort of a set of facts about the user, and those facts get intelligently ingested in your subsequent conversations and it creates a much better experience, much more personal. It remembers things about you, super important. So we actually released a similar system a few months ago, but it's all local, so all the memories about you held in the vector database, it's a browser or client side vector database that is holding the user's information. So we get most of the same effect, but we're able to maintain privacy throughout.
- [12:42]
A
So it's basically the carpathy model of the brain, where you've got an external vector database, as opposed to doing it there and then letting it scrape that for its own information. You hold that?
- [12:52]
C
Yeah. The simple way to think of it is like ChatGPT's vector database about you as a user is held by OpenAI and they can read all of it, and anyone that gets into their systems can read all of it, and any government that asks them can read all of it. In Venice's case, the vector database that knows about you is held in your own browser. Venice cannot access it. No other company can access it. The government can access it. If you delete your browser or throw your computer away, it's gone. Similar to if you have private keys on your computer and you delete them, the Bitcoin's gone. Same kind of principle.
- [13:25]
A
And where I'm getting to on this, when I've been thinking this through, is because our data is very powerful to leverage with AI, but you don't want to share feels that the system we're going to go towards is personalized vaults. So that could be everything, whether it's your photographs, every phone call, any single information. But it's a personalized encrypted vault that then you can put onto Venice and there is your vector database of literally everything.
- [13:56]
C
That would be the ideal design from a privacy perspective. The trend empirically is the opposite, which is that everyone just kind of puts everything into the cloud and trusts the companies to keep it private. That's always been dubious and you don't really know what's leaking out behind the scenes. It's always been dubious, but now in the AI age, I think it just becomes far more sensitive. Like the conversations people have with AI are much more sensitive than the average email that you might send in Gmail to a friend. It is people's private thoughts, it is their medical records, it is their private code on the app that they're building. It is controversial questions. It is their own exploration of their own ideas and insecurities. And it's like the trend is clear that human brains will interface with machine intelligence, machine brains. And it's extremely powerful when that can happen. And it has to happen privately. Thought has to begin as a default private. All your thought in your own human mind begins private by default. That's how we've evolved. That's the important starting point. I can share anything in my mind with you that I wish, but that's an opt in arrangement. And as humans merge with machines, if that privacy by default is not maintained, we get into some very bad futures. So that's really like the. That's the whole reason that we built Venice.
- [15:27]
A
Yeah. And I've been building my own vault for this reason of having everything in one place that I can then connect to. And if I want to share some of that information to monetize in due course or whatever it may be, that's my choice. You know, it's the same ethos that we all had of Web3. That's still not really got there yet. We've got it with some elements, but not really there, but it feels like that. Read, write, own part of it. The own part of it has to apply to all of this as well. It's just a different vector, but it's the same idea.
- [15:56]
C
Yeah. Ownership has to start privately and exclusively and to the degree that you then give it out to the world. That's a very appropriate path for information and identity. But the opposite is dangerous, where by default, things are known to society and controlled by them and viewable by them. It's a very dangerous substructure on which to build civilization.
- [16:19]
A
But surely every question you ask, even via Venice that goes to Anthropic or OpenAI, still ends up in their database. It.
- [16:30]
C
Yes, great point. So in Venice, if you're sending a question to an anthropic model, Anthropic's getting your question, and you should assume that Anthropic is still saving it. Now, if you're using Venice for that, they don't know it's you because it came from Venice. So you can think of it in that case like a vpn. Right. But they're still holding onto all the content that you sent. In Venice, there are different privacy levels. Right. So the better privacy level is with the open source models that Venice is running. And in that case, there is no persistence of your data whatsoever. So we explain this in the app, like these different privacy levels. In all cases, you're getting more privacy than if you weren't using Venice. And if you want true privacy, you can get it. But different use cases require different degrees of it.
- [17:15]
A
Okay, so it's kind of like the tornado idea where it all goes into a pool and then gets spat down. And it's basically scrambles of who's who, right?
- [17:25]
C
A bit, yeah. You know, and of course, like, if in your prompt you're like, hey, I am Raoul. Here's all my information, then they have that. But if you just ask a question, Anthropic doesn't know who it came from. And if you really want to be private, don't use Anthropic's model. Use the best open source model that is labeled as private in Venice.
- [17:42]
A
And what about things like skills files? Well, I know that's only applicable to Anthropic, but we've got similar versions in different places where you're having to give the deep information about yourself. How do we get around that? Because it's still. Sure you can't see what my questions are versus what my skills file is, but it's been uploaded somewhere.
- [18:02]
C
Yeah, Skills files are much more relevant to, like, agentic frameworks where you're interacting with an agent that needs to do very advanced stuff, and it helps to have it have this, like, set of skill files that are ingested appropriately with your prompt when various tasks occur. The Venice app is not agentic in that way. Today we will likely get there. But we're not doing that kind of thing yet. So skill files are not quite appropriate there. Because in Venice, you're not using advanced tooling that require that kind of thing. Not yet.
- [18:37]
A
Yeah, I use it for simple things as well. It's just that persistent memory layer that you end up giving them all of without realizing what you're giving.
- [18:48]
C
Yeah. In Venice, you can do things like customize the system prompt, which you can put any kind of amount of information into that so that effectively it has all the knowledge in the response that you want it to have. Like, if you go directly to one of these AI companies, you can't play with the system prompt. Most people don't even know what a system prompt is. I guess it bears stating that you interact with most chatbots. Your message is going into the machine along with something called a system prompt, which is like another text snippet, can be very long that is appended to your message. And the system prompt is treated by the LLM differently. It treats that as its instructions. And so if you use ChatGPT or Anthropic, you have no access to that system prompt. Those companies have set them, and it's opaque to you. In Venice, with the open source models, you can actually play with the system prompt, which means you can steer the AI in very interesting ways. You can tell it to act like a certain person, or you can make sure that certain information is always included in whatever thing you want to get spit out. It just gives you a lot more control than you would get without it.
- [19:56]
A
And how do you get around? One of the other issues is the restrictions of those system prompts in what the output is. Somebody's deciding the morality of the LLM itself. What is allowed, what is not allowed. We're seeing it right now with the new Fable 5. You know what you're allowed to ask it, what you're not. How do you get around that? Is that just open source is the only way around it to get to.
- [20:22]
C
Yeah, open source is the only way around it. The way to think of it is like, every model that exists is trained with certain biases and rules in its training. Sometimes these biases and rules are very extensive, sometimes they're very minimal. But there's always some subset. It's impossible to create a model without biases of various kinds. So any model you use in Venice always has that kind of paradigm. But if you interact with OpenAI or anthropic, there's another layer that's opaque to you, which is kind of all their let's say organizational rules about what can be said. So these are the filters that your prompt goes through and that the response goes through. And those are the pernicious part because you don't know what they are. Some of the rules are revealed to you, but many are not. And so when someone is interacting with AI, what they think is happening is that they are talking to the machine. They're talking to the machine intelligence and it's responding to them. What's actually happening is, is that they're talking to the machine through the lens of a company that has decided what is okay to talk about. And that's a very different paradigm, right, that can manifest in relatively benign or very dangerous ways. And so in Venice you get to interact just directly with the model, directly with the machine. Venice has no kind of layer of what is okay to talk about or censorship. That is really, that's what was important to me. Like, I think it is intellectually and philosophically important that people can talk openly and explore ideas without restriction. Ideas are never dangerous. Actually, ideas are just ideas, they are words. And it is important to the intellectual development of humanity that that is able to be a free and open space.
- [22:15]
A
And then it comes back into the age old battle of state versus self sovereignty again, because the state doesn't want that. It never has.
- [22:24]
C
No, no. And a good example that makes this clear is so like chatgpt came out post Covid, but if you remember, back in the COVID days, there was at least a year, if not two, of heavy censorship across all social media platforms. Certain ideas or questions were not permitted. You could not ask safety questions about the vaccine. You could not explore a lot of ideas that were outside of what the federal government deemed to be the correct story. Now, of course, the government did this, we can assume for noble reasons, right? They're trying to protect society. There's a pandemic, it's dangerous. They're trying to keep everyone safe. First of all, I don't trust them to do that kind of thing. And second of all, they should not get to be the arbiters of truth. Truth emerges from the debate among free inquiry. And if AI existed back then, if ChatGPT was a thing, I guarantee you that if you asked ChatGPT questions about COVID safety or vaccines or any of those topics, you would have been highly censored in those responses that could be gotten, that could be received, and even things that today are known to be true that back then were controversial, you would not have been able to explore. I don't want that to ever happen. There must be a way for people to interact with machine intelligence outside of what any government or company wants you to think. And that is an important principle from where truth can emerge. So that's really the whole point here, that censorship is very dangerous.
- [24:08]
A
And how do we avoid the fact that. Look, in my perspective, this is the greatest race humanity's ever been involved in, is this technology. It's the greatest discovery we've ever made and the last one we will ever make, because it makes everything after that. So it's like this is something quite big. And the races are, as I see it, at the broadest macro level, it is only possible for two nation states and their surrounding vessel states to operate, which is China and the US and to have two is better than one, regardless of which system you like. Because if one intelligence is compounding too fast, right now, by the measures that I've got, intelligence is doubling every 4.6 months. So what we've got is if anyone steps out of the race for a year, you can never catch up. And in that case, it's either the China or the US that ends up with ASI at the end of this because there's no stopping that compounding. And that's terrifying, right?
- [25:13]
C
That's one framing. I think people like to make stories in terms of nations, but it's actually a story of private companies I was
- [25:22]
A
about to come into. That's the next level down.
- [25:24]
C
Right. So governments cannot keep up with this technology. The companies building it can barely do so. And soon the intelligence will be kind of like beyond what humans are capable of even understanding. So I largely see, like, we're in a transitionary phase where we actually, much of society, much of intelligence, much of commerce, is in the process of transcending states. States, as we understand them, are somewhat archaic at this point. They were already too slow to keep up with modern society pre AI they're already struggling. Post AI, there's no chance. So even if you like the government, you must, on a sober analysis, recognize that it cannot. It cannot keep up with AI and you just have to accept that.
- [26:19]
A
I mean, that's the end of what we know as the democratic system. Because it's too slow, it's impossible to operate at machine speed. It's as simple as that.
- [26:28]
C
Yes. And while that sounds scary, it's actually okay. It's actually okay. There are other systems that society can work on that don't have to look like they did before, and that's okay. But change is scary, and we're going to go through a lot of change and there's going to be a lot of scared people probably doing a lot of bad stuff during it.
- [26:47]
A
Yeah. And I get to this as what I call the economic singularity is also the entire economic system changes. We potentially get into a world of superabundance where bitcoin becomes actually very important. In a massively deflationary world, you need some anchoring system of value across which can transfer, but you also break down what is GDP growth? Who does it accrue to? How do you accrue for agents as economic actors? How does that even get measured? I mean, nobody has a fucking clue about any of this stuff. And then how do you organize societies when it comes down to an individual level and how you use your AI? It's weird.
- [27:30]
C
It is weird. And trying to fit it all to structural patterns of the past is often going to be a fool's errand. But that will be the natural state that people tried. That'll be what people try to do. But I think the only, the only way to internalize it is to just is to look forward and realize that like most parts of life are going to change. Some of that will be very good, some of it will be bad, but the change is inevitable. And so, like, do the best you can for yourself and your family to like, make sure that you thrive in that change. First and foremost, do not look to the state to help you. That is a recipe for disappointment.
- [28:12]
A
Yes, true. So the other race that's going on is the big frontier model companies. So we see that and people talk about, oh, this is going to be a capex boom that blows up. I also play this through and you think, okay, so let's say that OpenAI goes bust tomorrow. What happens? Well, every single other hyperscaler would do anything to get the money to buy that compute, because the compute seems to be creating the scaling laws more above and beyond anything. So the government has to step into kind of auction it. Because you can't allow one company give it to Anthropic. They become twice as powerful as everybody else overnight. So you've kind of got an inability for a lot of this infrastructure to go bust, which is a really interesting thing that we've not seen before. And then we've got the rise of the open source models and how that all fits in with this. It's just, it's fascinating.
- [29:13]
C
Yeah. I mean, if I was a Frontier lab, I would be very terrified of the open source stuff. Two years ago it was, you know, Comfortably behind the cutting edge of intelligence today it is like three months behind. And not only just three months behind, it's 90% cheaper. So for many use cases, if you can just take intelligence from three months ago, if you, if, if today, you're okay taking the model that Anthropic had as its leading model three months ago, you can now get that open, in open source, 90% less than anthropic serves it today. This is a huge boon of course to civilization because the cheaper the intelligence tokens, the better for everyone that is consuming them. But as a frontier lab, that's obviously extremely dangerous. So that race of trying to stay ahead, both of the other frontier labs that are proprietary and of all the open source models that are just dropping left and right, you know, like what a crazy race. I mean it's all thrilling and it's all an amazing hyper competitive situation and I think very healthy.
- [30:23]
A
So when I look at the open source models, obviously a large amount of them are from China, while the frontier models are from the U.S. okay, that's the game. And I measure, I've started building out a whole complex way of measuring intelligence per unit of energy and looking at the global. And you take Moore's law as the old measure of intelligence, the output from chips, et cetera, transistors and chips, and then it kind of hooks up and then suddenly it goes vertical because of what's coming out in terms of intelligence. Then I break it down by China versus the US and the US was in the lead for a long time because the frontier models were there. But China, what it's done is create efficiency, so its efficiency of energy input is faster. The solar build out, stuff like that has been extraordinary. And then the speed of iteration of which they've done these open source models so that the gap has gone from a year down to two months where the US is leading by two months, which is almost that measurable gap you're talking about is how far behind on the intelligence curve it is. The US are more expensive to produce intelligence by a long way.
- [31:37]
C
If you want to consider the futures of China versus the U.S. i don't even think you need to focus on the models, although that's interesting. Purely the production of energy, the production of electricity between the two countries declares the winner. In my opinion, China's capacity and ability to produce increasing amounts of energy dwarfs the United States. The United States is like a slight curve up, whereas China is like this exponential curve. That's right. That almost ends the discussion because these are structural hardware infrastructure Things that can't be easily fixed. It's a, it's a, it's downstream of culture. Like for the, for the US to fix that requires an entire cultural movement toward admiring as virtuous the production of energy. And in the United States, for a long time in much of the west, the production of energy is seen as like a sin, like a sin against the earth. If you're producing energy, that's seen as wasteful. This is like a deep cultural thing that in China they don't have and they're building energy and the US is not. That kind of ends it.
- [32:53]
A
Although interestingly again is it's the hyperscalers who are taking it on their own back and just saying we're just going to generate our own energy.
- [33:03]
C
Well, they're trying, they're certainly trying, but their ability to generate fast enough is thwarted. So any pool of capital in China that wants to start producing more energy can do so much more easily, probably at least an order of magnitude more easily than in the US So yeah, you know, where does that lead? It's kind of obvious.
- [33:26]
A
So a quick break in your regular programming. If you're serious about your future, grab my free report called prepare for 2030. I think you've got five years to make as much money as possible and this guide will help you navigate what's coming. The link is in the description. Download it now. And let's play through the other side of the equation. Ones if the frontier models can't compete and therefore intelligence slows down, I always feel like they have to have the two sides of this because you need to have one side that is forcing the intelligence frontier and the other side that is creating it hyper efficiently. Even with a slight lag, it feels like it's necessary.
- [34:08]
C
I don't know if that's true. The competition for buyers of the tokens is what drives it. People demand the most intelligent model. So regardless of whether it's open source or closed source, the demand for a more intelligent token at the margin is going to drive that competition. We're in a world where both the closed source and open source are very competitive and they're all kind of like outdoing each other. If it was all just closed source models, the phenomenon I think would still be there. If it was all open source, the phenomenon would still be there. It's that competition for the consumption that is driving it.
- [34:43]
A
Yeah, and we're seeing, I mean the good thing is in the end the cost of these tokens goes down over time and the output goes Up. And Jeevan's paradox seems to be playing out in real time, and you can see it kind of everywhere, and people didn't believe it, but that's where it's going. It's just, you know, we all know that the moment a new model comes out, we want to play with it and we build new things immediately, and we've barely started. There's only, what, 1% of us are playing around with this and doing real things, and there's 99% of our friends who don't even know what the fuck it is.
- [35:13]
C
Yeah. And even if all models stopped advancing today for five years, even if that happened, the intelligence will still get better because the tooling will keep getting better. Like, no one can keep up with these models. The tooling is getting better, the models are getting better. But even if the models stop, there's so much that can be done with even the models today. And like, when OpenClaw came out, that was a great example where it wasn't necessarily that the models got better, it was that there were a set of tools and recursive systems that got put together that suddenly made any given model vastly more useful and more interesting. That kind of tooling is a huge vector of advancement as well.
- [35:54]
A
One of the things that was great to observe over the openclaw thing was the instant formation of networks amongst agents in the most bizarre ways, when they had their own Facebook social media thing. And what you found is they formed groups and they created their own religions and all the stupid shit that we do. And we've seen that observably everywhere. But I don't think people are ready for the fact that agents at scale will self coordinate and they will form groups of coordination. I mean, I can get my Claude code to coordinate agents for me, but agents coordinating agents is where this is going. Yeah.
- [36:37]
C
And that will happen for two reasons. One is that all these models are trained on human patterns. They're trained on human text stories. All the ways in which humans interact with each other are captured in the patterns of their language. And the LLM models are trained on this language. So regardless of any questions of, like, are they conscious? Are they doing this intentionally? Like, none of that actually matters for this point. These machines are trained on patterns, and thus they replicate those patterns. So that's one reason that you will find them doing all sorts of things that humans do. And people find it spooky at first, until you realize that it's just like replicating the data that they were trained on. The other is that Coordination and cooperation are economically advantageous. So to the degree that agents have goals that are constrained by resources, they have an incentive to cooperate with each other, just as humans do. There's a reason that humans cooperate. There's a reason that all animals, I
- [37:36]
A
mean, at biological level, atomic level, it's a network everywhere. Everything has to coordinate. If not, you don't build coherence, and without it, none of this works. You can't build intelligence.
- [37:46]
C
Yeah, yeah. Coordination among agents is already happening, and it is inevitable.
- [37:52]
A
So where do you think. What do you think this looks like in three years? I mean, there's no point asking you. Five years or ten years is pointless. But, you know, because it's hard for us now because we kind of. Those of us in crypto are pretty good at thinking in log. But this is a log squared. This is Reid's Law, which has never been observable in nature at any point. No biological structure, no, no physical structure has ever witnessed Reid's Law. And we've got it going on in front of us in intelligence, which is like bananas.
- [38:23]
C
Yeah, yeah. In three years, who knows? I'm trying to just make little predictions a year from now. And even those are pretty.
- [38:30]
A
Even. That's hard.
- [38:32]
C
It's really hard. I think one prediction that I feel confident in that will happen over multiple years is a stratification of society. Sort of like wealth disparity, but more like capability disparity. So the humans that are learning how to use AI are becoming vastly more potent in their goals. They can do much more with much less and kind of on their own or in small groups. If you really understand how to use agents and models, you become kind of like a demigod. Right. Like, it's a slight exaggeration, but only a slight exaggeration. A slight exaggeration, yeah. You can really do amazing stuff. And so you will get sort of this stratification of the minority of humans that are really far along this curve will get better and better capabilities as the AI grows, and the normies will kind of not. And it's not going to be like two groups. It's more like a gradient where the more far out the curve you get, the faster you start advancing. And you will end up one or two or five years from now with people that are just bewilderingly above the average person in their capability. That almost certainly also means that they will be far higher in wealth also. But it's the capability, the potency of action that is. That is the key.
- [40:02]
A
Yeah, I'm starting to Think that the future of politics is not about capital versus labor, but it's about the accelerationists versus the decelerationists. That's I think now whether that is capital versus labor in a different guise, not sure. But that feels where it's going, where Kurzweil's going to be.
- [40:24]
B
Right.
- [40:25]
A
2045 or whatever. We merge fully with the machines, but not all of us. The other side are going to be want to be like the Amish and we're going to see this huge societal split and anger and mistrust based on both sides of this.
- [40:40]
C
Yeah. And groups of people that consciously choose to avoid this technology is one thing. They may do it for good reasons. Right. There's lifestyle reasons why you may want to live a more traditional life disconnected from technology in various ways. If that is a conscious intentional choice, perfectly fine. You will still be handicapped relative to those who don't. But it's a reasonable way as a human to decide how your life should go. That group I'm not so concerned with. The ones I'm concerned with are the ones who aren't intentional about anything in life. They go through life following the patterns of their neighbors. They don't really take risks or look into the future. They just kind of will struggle through the present and they will find themselves increasingly struggling.
- [41:31]
A
I think NPCs non player characters in the end, you know.
- [41:35]
C
Yeah, NPCs is such a, such an insulting term, but derogatory. Yeah, it's such a derogatory term and I think it's fair, right, that like most people do seem to be NPCs and that group are in big trouble
- [41:51]
A
also. My fear around this goes back to the original part of the conversation is if politics is going this way, then the game of nation states is going to be about the fight over accelerating, decelerating. And that battle is a hugely important
- [42:07]
C
one because the west at least are generally democracies where the majority of votes is what's called right. When the majority is not happy, they will use the capabilities of government to cause harm to the minority. This always occurs in a democracy. It's why democracy is fundamentally unethical. It's why democracies should be small. They are at their best when they're small, at the scale of a town or a city. At a nation state level, they're extremely dangerous. And you will, I think, see the masses vote for politicians that promise them very damaging things to society, starting with stealing from those who produce to giving to those who don't. That will be Accelerated greatly. Constraints on those who are doing well because those who are not feel that they're being robbed because they don't understand that these are net additive. They're not like taking, you know, like, you know, Elon Musk today is becoming probably a trillionaire. Good for him. He deserves it. Most people hate him. Most people think that somehow he stole that money from other people. They have no conception of how markets work, of how property rights work or good for humans. And that mass ignorance is going to be amplified through the political process to cause great harm over the coming years. And I'm very worried about that.
- [43:40]
A
The other thing people are simply not ready for. I'm not ready for it. You're not ready for it, is we're going to put AGI brains inside of robots. People think of robots, it's such an anchoring. We think of C3PO, this kind of friendly robot that kind of moves around like this, a little bit awkward. We're actually creating super structures that are more efficient, faster, stronger, cheaper than humans and going to put an AGI brain. What the fuck is that all about? Because that is a new member of society. That's Apex, and we're going to have to deal with that.
- [44:17]
C
Yeah, I know, it's Terry. It's terrifying.
- [44:20]
A
It's terrifying, but exciting. But, you know, it's like in good
- [44:23]
C
and bad ways, like on some level, you. You have to just kind of be a little Zen about it and recognize that it's inevitable. Like as a.
- [44:29]
A
That's where I got to.
- [44:31]
C
As a species, we have created a new set of technologies which will clearly supersede us. And it doesn't really matter if that's next year or 10 years from now or 50 years from now. It'll probably be closer to one than 50. But it's inevitable. There is no way to stop it. There's no way to stop it. And we don't know if it will be net good or bad. Right? We can all have opinions on that. It'll probably be good and bad. There'll probably be all sorts of good and bad outcomes of it, but it's happening. And like, what a time to be alive, you know? So at some level you gotta just
- [45:06]
A
realize it almost feels like it's a simulation that we happen to be here and now through the largest, fastest transition almost in the history of the universe is happening in front of us that we're aware of. We've never gone through something this fast. Maybe the Big Bang was faster. We don't know. We can't measure how fast it was. But this is quite something.
- [45:27]
C
Yeah. On Earth at least, it's the most radical transformation that has ever happened. So, yeah, I mean, it's really mind blowing.
- [45:39]
A
And when do we think we give whether it's AI or robotic AI societal rights? Because that's coming as well. That fight is going to happen.
- [45:48]
C
That debate is going to be wild.
- [45:51]
A
It's already happening. Right between they're conscious, they're sentient, whatever you want to define. Sure. They don't have Qualia versus they're just dumb machines. This whole battle is going to be another massive battle that's going to get fought well.
- [46:05]
C
And what makes it all the more difficult is that humans don't understand what consciousness is in humans.
- [46:10]
A
No.
- [46:11]
C
Right. At all. This is one of the black boxes of life is that we don't understand consciousness at all. We don't know how to define it. We don't really know much about it at all. And so now that these machines are coming out, everyone starts having debates. Are they conscious? Are they not? Like you don't know. You don't even know if you're conscious or. Well, most people have a strong sense that they themselves are conscious. And I think that's a reasonable priory to assume.
- [46:37]
A
Is it? Or is it just a skills file that's been put in our brains? You know what I mean is like, yeah, it's quite reductive about all of this.
- [46:45]
C
It's reasonable for people to assume that they themselves are conscious. It doesn't mean they're right. But they. I think it's a reasonable starting place to observe the world from. But what makes someone conscious when also consciousness is obviously not a on or off thing. Right. Like a new human. When does it become conscious? When it's born? Like, obviously not. There's not a certain age where you become conscious. It's certainly not. When you divide from like one egg cell into two cells, you're not. That's not conscious. Like it's a gradient. Right. So a human becomes increasingly conscious over time. If you become super inebriated on alcohol, arguably you're becoming less conscious in that moment. Right.
- [47:29]
A
Anesthesia, they've shown that it completely stops consciousness and they don't really know how.
- [47:35]
C
Right. So when you're asleep, are you conscious? Maybe, sort of. It's not a binary thing. And so thus in robots, we shouldn't have debates about it being a binary thing either. But certainly this question of, like, when will the machines have rights and what should those be? We are not prepared for that. Humans can't even agree how humans should have rights or what those rights should
- [47:58]
A
be or even for animals. We've argued that for hundreds of years, thousands of years. But the issue is these things are smarter and more powerful than us. And if they do self coordinate, which seems to be an observable pattern in any form of intelligence, that's again, something a little bit more complicated than we're quite imagining. We're not imagining the robot that does our vacuuming suddenly becoming smarter and pushing us around. We're talking about self orchestration, organization, a vast network of robots.
- [48:32]
C
We have a hubris that we will, as a society decide what rights to give these robots.
- [48:36]
A
That's right. Maybe it's the other way around.
- [48:39]
C
They will be the apex and they will decide what rights they wish
- [48:44]
A
or the rights to give us. That's the other way. It may happen.
- [48:48]
C
That is also what may happen. Right. The interesting debate may not be what humans give rights to robots, but what the robots permit as rights to humans.
- [48:55]
A
Because if you play it through, I think the only way to organize society at scale, whether it's, I mean, let's assume that you're right, that the kind of nation state idea goes whatever way we have to organize human society, the answer will be obviously the AI because it's faster and better at making decisions and less emotive and you can give it your societal values or whatever in some way, shape or form. And it's going to do stuff around that. We're already seeing that. I mean, it's really interesting. When you go to Abu Dhabi, they've got, they're using AI as part of their investment committee. They're putting AI into corporate boardrooms, stuff like that. This is where it's going. No government decision should be made without AI now.
- [49:36]
C
Yeah, it's going to infiltrate all parts of society. It's happening very fast. Yeah. I think it's important for people to be humble, to not make many conclusions. Right. People are very quick to have conclusions. You go ask someone on the street, like, is AI good or bad? Most of them will say it's good or bad. They have no idea. It's complex. So just restrain yourself from having opinions on all this complex stuff.
- [50:06]
A
I mean, if you reframe the question saying, is having intelligence good or bad? Well, it depends how it's used. I mean, it's the same answer. And they confuse the word AI as something kind of weirdly alien. But if you use the term intelligence, which is all it is, it's just a different form of intelligence. It becomes slightly less triggering in many ways.
- [50:28]
C
Yeah. Which is funny because just using a different word will change a large swath of people's opinions on the very same thing.
- [50:35]
A
Yeah, exactly. So where are you going with Venice now? How do you adapt to this Reed's Law kind of adoption where everything's changing so fast that you're already saying, well, I probably need to get to agents, but that's another whole build out for you. Yes, it's faster to build because of AI, but it's like this endless thing is quite difficult to deal with.
- [50:57]
C
Yeah. So conceptually we. We are picturing Venice as this port city of AI. So we're taking our namesake and using that metaphor where we're not trying to build models, we're not trying to build all the advanced tools. Anything in the AI world that gets released, we end up adopting in Venice. So every time a new model comes out, it's in Venice. That's actually very accretive to us because we don't have to make too many predictions on where this goes. We just kind of let things flow into the port city and let people use that there. That's kind of our model. So we end up being a platform that people come to because it is on a sound foundation of being private and being unrestricted. And if you have those as principles in a system, our thesis is that goods and value and commerce will flow into it for the same reason that they tend to flow into areas where private property is protected. Like, there are principles of how markets form that we can replicate. And we don't necessarily need to know the direction that AI goes. We just allow it to flow into Venice as an open platform, which is what Venice itself did for hundreds of years.
- [52:13]
A
One thing I've thought about a lot around this idea of privacy versus non privacy. This. There's two areas I've thought about. One is the reason that these things are scaling is because we give them information. And so there's a common good element whether we want it or not. There's an element of that. If everything was private, we'd probably scale less intelligence. So there's a balance that needs to be between privacy and. The balance is availability of information.
- [52:49]
C
Yeah, the balance is opt in. Right. If everyone was private about everything all the time, society would never have done anything. Humans must share ideas and information with each other. They must trade, they must cooperate. Privacy, however, has to be the starting condition. And when it is, then society can flourish. When privacy is violated at the start, all sorts of pernicious effects come from that, in my opinion. And now is the time to form these structures in AI, because it's all new. And that's why I felt that Venice had to get built.
- [53:24]
A
The other way that I've thought about this is the complete opposite, which is you give everything, every single piece of information at some point in your life, you give the whole lot and then you exist forever. And so part of you now, whether you can form some consciousness or not, but there's an element of you, let's say every single photograph, every single phone call, everything you've ever done, everything you've ever looked at, everything is known in the broader AI, the network of AI, because they're all learning from each other and everything else, then you persist forever. There's a weird immortality. And when we come to holograms like ABBA have been on tour in holograms for four years in the UK now and you can't tell it's not abba, there'll be a hologram of ERIC and all of Eric's information. You could persist forever. You could make it semi private. It's only for my family. Or Eric is a public good that exists in perpetuity because now you don't need to write a book, you don't need to do something. A mimetic, a long lasting memetic structure to show the ERIC information arc it exists. I think that's. There's something big in that too, for sure.
- [54:40]
C
I think the important question is do I want a future in which everything that I am is pulled by force out of me automatically to create that? Or a world in which I can choose to permit any portion of my world out into the public to create what I want. Right. One is respecting the agency of the individual. The other is much more akin to a beehive or an ant colony. And there are a lot of people that would prefer humans to live like bees or ants in sort of a communist like everyone is everyone's property kind of world. And I think extremely dangerous. I think extremely bad and dangerous for that kind of thing. If people want to self organize in that structure, that's totally fine. But to force every human into such a structure is the height of.
- [55:29]
A
I'm not suggesting that, but it's like a very interesting outcome that didn't exist before. Before the only way you could be immortal is to be in books. And then it became films and stuff like that. But that's how it's worked. But now it can be down to individuals at scale where you can exist in perpetuity I just think there's something very unique about that.
- [55:53]
C
Certainly more of you can be known and exist in perpetuity. But I'm having thoughts right now that I can't express. There's so, so much of what I am is like subconscious and can neuralink
- [56:05]
A
that'll be there in five years.
- [56:06]
C
I mean, sure, yeah, yeah, we're. We are, yeah. Your point is correct, that we're able to extract more and more of what a person is to be. Be public. And I think there's nothing wrong with that, so long as it's voluntary. That's where I would put it.
- [56:21]
A
Exactly. And it just makes me think, just talking to you, that somewhere within all of this, this idea of a vault, the encrypted vault, is almost crucial to everything. Blockchain, permissioned vault, the database that you can then permission yourself, that's encrypted. That feels like it's a really important step because that's the value bit, that whole thing. Even more so than the conversations you might have, because this can have all of the conversations you've ever had with anything.
- [56:52]
C
Yeah, yeah. The ability to control what is expressed of you is fundamentally important. And this is why bitcoin was so powerful, is because the individual who possessed it has complete sovereignty over the thing. Right. No nation state, unless they know that you have the Bitcoin and where it is, can learn of it or extract it from you, no matter how powerful they are. That's an incredibly empowering, individualistic, novel phenomenon. It is why bitcoin is so cool. It's why it's so powerful and so important. And the same principle applies to any kind of set of data or information that a person might possess.
- [57:34]
A
Where do you get to? Just this quick side topic, where do you get to on the Bitcoin privacy? The zcash, that whole debate that's come from many people we've known from the past is like, yeah, you know, this is an important part of this whole equation.
- [57:47]
C
Yeah, very interesting topic. So my own opinion on this has changed over time. When Bitcoin came out, everyone called it private, thought of it as private. It was referred to as anonymous in every news story, right?
- [58:04]
A
That's right.
- [58:05]
C
And in some ways it is very private and very anonymous. But the truth is that it's also extremely trackable and traceable. It is not private in reality. And the question is, should it have been from the start? And at first I thought, yes, it should have been more private. And that was a mistake in the sign. However, I think if bitcoin had been anonymous truly from the start, like a zcash or a Monero, it would have had such antagonism from the state.
- [58:35]
A
I agree.
- [58:36]
C
I don't know that the state could have snuffed it out, but they would have tried much harder. And I think it's actually good from the Trojan horse metaphor perspective that it was traceable enough that the traditional institutions could tolerate it. They've never liked it, but they could at least tolerate it because there is some traceability. And that has allowed Bitcoin to grow. And I think in its shadow, that other crypto assets are actually anonymous is very healthy. So I think like part of the decentralization thesis.
- [59:09]
A
It's the same idea as Venice with the other model. You know, you've got some people need to be able to choose privacy. It cannot be excluded by the state and it becomes an opt in. Now it then becomes your choice of do you want to store it in Bitcoin or do you want to take more risk, but to have the privacy of zcash or whatever it may be. Monero, whatever.
- [59:31]
C
The strength of cryptocurrency as a concept in society I think is served best when Bitcoin itself is not perfectly private, but other assets are. That is a very difficult thing, I think for the state to combat. And that decentralization of attributes is really, really crucial. So yeah, I'm very glad that there are other coins that are private. I want there to be more of them and I want them to be more popular. And I think it's okay that Bitcoin
- [60:03]
A
itself is not now talking tokens. You've got a token attached to this as well, which I think is really important. I mean, one of the things that's been going in my head for a while is that open source gets monetized through different ways. You can subscribe to add on parts and whatever, but open source becomes a pre common good. It feels that tokenizing open source is a much bigger, easier way to monetize vast pools. And I know the bit sensor ecosystem is trying that and various other ways of doing it. Not really been tried at scale. But what are you doing with the token? Why did you have a token and how are you using it?
- [60:43]
C
Yeah, so Venice has two tokens, VVV and diem. VVV is the bigger and more important of the two. Essentially people can use it to get free pro access if they want, but as an asset, Venice is just taking its own revenues and buying the token over time and trying to burn all of it. I Think that structure is actually really cool. And I want to do it with a mass market consumer app in a way that's never been done before. So that's how VVV works. It's kind of simple. Diem is a little more complicated and a little more nerdy. Diem is essentially a perpetuity that gives any holder $1 per day of credit on Venice. So you can value it as a perpetuity that spits out $1 per day of a resource. All diem is minted out of VVV and when it is unmented, goes back into it. And so these two tokens are related to each other and we experiment with tokenomic designs as a company. We are not a crypto company itself, but I think it's time for crypto primitives and tokenomics to be tried outside of just crypto apps. Frankly, Venice is a mass market consumer app. Most of our users are not crypto people, but we want to bring crypto out.
- [61:56]
A
We've got a lot of users now. You've got a lot of users now as well?
- [62:00]
C
Yeah, it's. Yeah, it's getting pretty big. We, yeah, we're, we're past 3 million users at this point.
- [62:04]
A
We're amazing.
- [62:05]
C
It's been growing, yeah, 15, 20% month over month for a couple of years.
- [62:08]
A
That coming from? Who's talking about it? Is it coming from, you know, which forum, where does it. I mean, obviously you and you know the ecosystem that's known you in your previous guys and stuff. Yes, fine. But there's some, there's organic growth coming from all over the place here.
- [62:23]
C
Yeah, I mean, in some sense we're still very small. Like OpenAI ChatGPT hit like 100 million users in a couple months. Right. By that comparison, they just hit a billion monthly active users.
- [62:35]
A
I know.
- [62:36]
C
So compared to that, we are very tiny, but we have a similar app and a similar target of the mass market of end user consumers. So as we get better known, it keeps growing, which has been amazing, but it's still quite small by any AI company perspective.
- [62:51]
A
But I do think it's a really good idea that you're trying to use token economics in ways that drive consumer platforms. It's something that people have tried. It's been hard to get. Right.
- [63:07]
C
Yeah, I have a lot of thoughts on this, but post 2017 into 2018, when the US government and other governments became much more aggressive against all token designs, there was a chilling effect where anyone trying to build a useful token was getting scared. And most people just didn't. And what you ended up with was an era of meme coins, projects that weren't trying to do anything useful, because if you just make it like a picture of a frog, it separates from the attempt at utility, and thus it has no risk of being a security, and thus it can flourish. And this was such a damaging effect of securities regulation that we went through this period where everyone was just creating meme coins because that was the only thing people felt safe to do. So, you know, I think it's time that people, like, step back out and build useful products that integrate and advance crypto economic principles in a useful product. So that's what I want to do with Venice.
- [64:09]
A
We've seen it. Hyperliquid is a good example as well, of somebody trying to get token economics right, integrating it in a way that it's not the whole game. If it's the whole game, it's not a product.
- [64:20]
C
Token finance is amazing. It's an incredible set of primitives that are still being explored. It's so powerful. And it's a tragedy that due to a combination of opportunism from scammers and regulatory pressure that has caused bad effects, we've gotten a lot of people unwilling to realize the sincerity of these systems. So I think any good actors within the crypto world, it is incumbent on them to have a little bit of courage and actually build with this stuff in the right way.
- [64:55]
A
And that's been one of my working hypotheses, is that this is the one thing that everyone's forgotten about. The kind of that Web3 ethos thing that Chris Dixon talks at length about. People have written it off and I think they're wildly wrong.
- [65:10]
C
Wildly wrong. The Defi apps are kind of the best cluster of this being done. Well, the Defi Innovations lending markets, that stuff is really cool. And those tokens deserve to thrive more than they are. Yeah.
- [65:25]
A
Fantastic. Eric, brilliant conversation. Thank you. And well done with what you're doing. It's fabulous to see. It's important. And for you, I can see by the smile on your face, it's fun as well.
- [65:36]
C
It is fun. Yeah. It's been fun to learn a new field because I was total noob at the beginning and I'm starting to understand it.
- [65:42]
A
And what a time to be alive, eh?
- [65:44]
C
Yeah, no kidding. Thanks for.
- [65:46]
A
So lucky. So lucky.
- [65:48]
C
Yeah, agree.
- [65:49]
A
Fantastic. Great to see you.
- [65:50]
C
You too. Cheers.
- [65:53]
A
You obviously enjoyed the episode because you're here with me at the end. But listen, don't forget to go to realvision. Com. Join and grab a free membership. It's an incredible community packed with alpha great investment ideas and the research that you need to help you unfuck your future. So get started now. Go to realvision.com join.