Can AI become a responsible technology, Amit Pradhan?

I like to ask sort of an uncomfortable question to audience who's often who owns you. You know, it's a provocative question. It has a visceral response, you know, in the physical world, the US fought its current longest, largest war to ensure that no one has to answer that question in America anymore. And yet, after all of the progress of the civil war and rights around that, you know, the answer to the question who owns you is about seven companies. And in the last 30 years of this attention economy, as Charlie mentioned, the product is free because you are the product, you know, what we lost was data. And why that's so much more significant now is that data is training models that have relative cognitive representation of whose data it was. So suddenly, there is a version of Andrea not in any physical form, but there is a version of Felix is a version of Andrea, there's a version of Charlie and it's interacting with businesses, it's interacting in the form of ads, it's interacting in the form of incentives. Facebook is selling the identity graph, Google sells search, Amazon sells and upsells in cross sales product based upon what it learns and knows about us. And so whether we like it or not, there are versions of us that are creating phenomenal amount of value for the businesses that have leveraged that. And in many ways, if we don't have control over it, if we don't have agency over it, and then the ultimate thing in this world of currency that we live in, if we don't have economic rights related to it, then the answer to the question who owns you is like Al Gore's inconvenience truth and it ain't you. Welcome to Charlie's Angus, it's episode number three and I'm privileged and excited to welcome Amit Burdan to this session, welcome Amit. Thank you Charlie, much appreciated. Yeah, and we're also here, hi Andrea, co-host of the show. We're also here with Felix, our producer and today's kind of a special treat, the Silicon Valley Special Edition, one could say, where are you guys calling in from Charlie Amit? Well, I'm calling in from Big Sur, which is about two, two and a half hours south of Silicon Valley. And I'm in San Francisco, so in Silicon Valley, in the heart. So yeah, we're here to talk AI today and how this technology can become a responsible technology and really excited to dive into this and before we start our conversation, we're really grateful to have you Amit and I'll quickly introduce you. So I mean, you're a seasoned startup founder, impact investor and sort of on multiple boards and you just told us that you studied computer science in Sydney, but originally you're from India and while still at university, you founded a company and that kind of led you to Silicon Valley and you didn't leave, you kind of stayed for, came back multiple times and have been an early builder and investor in the space of decentralized technology, like blockchain AI and also co-founded the Silicon Valley Blockchain Society. And now you've been building a new startup for a few years, rainfall, which we'll be talking about later in our conversation or your building applications around self-sorrowing AI. Did I kind of get that right? In a nutshell, absolutely. Okay. Well, yeah, thank you for taking the time and I've been meaning to ask, like, how did you guys meet? Was it in Silicon Valley? Let me maybe start out, I meet and then you can chime in because it's such an important story for me when I left in 2001 when I left Silicon Valley, I meet just came in in 2000. So we missed each other for a couple of years there, but I was gone for maybe 15 years because Silicon Valley was getting too commercial, too big and too impact for me and that's when we moved to Big Sur to really live down here while having access to Silicon Valley. And then I ran into Amit about seven, eight years ago, I think. And I told him what I was all about and he told me, wow, this is interesting. You know, the valley changed a little bit because I was implying that well, everybody's just in it for the money and blockchain was happening and I was starting to happen. And Amit invited me to give a keynote to one of his salons and the topic that I chose was the intersection between deep technology, like blockchain and AI, consciousness and impact. And what I found so inspiring is that Amit not only supported that positioning, but he was right in it with me, which was highly unusual for that time to combine deep tech with consciousness. And so he invited me to do that. And indeed, when I went there and gave the keynote about this, about 80 people were in the room visibly more diversity than I was used to 15 years before and very interesting discussions. So I want to thank you again, Amit for inviting me to that and that was stands out when I think about our relationship as the beginning of a really deep friendship as well. Well likewise, it's been such a privilege. It's been such a delight, but I think the really important thing was for us, there was these transitions that we were trying to make in Silicon Valley culture, decentralization brought about a lot of that. And it was very easy to find technology and get caught up around it. And in this case, it was blockchain and some of the benefits it could bring. And I think in so many ways, finding that not start to come and contextualize it, give it framework, give it give it more meaning was so powerful because that's what Charlie was able to do to come into this environment and this incredible group of people. You know, we had this amazing network of investors, builders, influencers, policy makers, you know, the whole set of bandwidth, we had this, you know, Charlie alluded to it, you know, we wanted to bring people from diverse backgrounds, men, women, people from different parts of the world. And the idea was create more access around what was going to be a very important way to think about, you know, using technology for the future. And I was so lucky to start building this friendship with Charlie and then have him come and keynote and you've probably experienced this before, but when Charlie is done with the room, everyone's resonating at a very different frequency, you know, and suddenly it's not just the symbols are energy and frequency symbols, not just dollar symbols, right? And that was really powerful in a room that was otherwise full of capital. And that's how our journey and our friendship began and Charlie is by far one of my favorite humans, not just friends. Well, we shared that and thank you for painting this picture. I can totally see this room now and still I feel like from afar, I've never been to the valley myself, but from afar, it seems like most rooms don't work that way. The builder culture has been changing a bit over the past few years, but I'm really interested and want to hear from you guys because Charlie, you've gone to the valley back in the late 80s and left around 2000 and I mean, that's when you stepped in. What would you say, how did the builder culture evolve? What kind of people were working alongside you, Charlie? What motivated those people to come to the valley and build technology? Well, a great, great question. I'm a great example of people who came to the valley in the 80s and 90s who had a technology background fairly advanced in some topics that seemed interesting to challenge the existing systems to go beyond what's possible now. And Silicon Valley was the North Star for me as a computer scientist to get there and do that. Many of my friends who came into the valley at that point in time were similarly motivated by making a difference, by creating something cool that nobody is would be able to build. And at that point in time, we were non-political politics did not play a role, Silicon Valley was not big enough to warrant that and us Silicon Valley nights were not interested in getting on the radar screen of Washington. We tried to stay away from that and whenever they came and visited, they said, "Wow, you guys do business weird." It seemed like insider information trading all the time and we said, "Well, that's how we do it. We influence each other." And that's how we build great things and move from one place to the other in order for the search of making our contribution, our level of repair, like Steve Jobs said, in the fabric of humanity with technology. And we were progressive in nature, all the engineers and most everybody, and had a beautiful can-do attitude, which I think is not missing, but in a stretch of the imagination right now, but also in not just working for money. So if we're keen, that was a great side effect of doing something cool. It was not the primary motive. And I think it will be able to speak to that better in my impression right here wrongly is that that changed a little bit in the last 20 years since I've not been there, not exclusively, but certainly to a certain extent. And that's why it was so refreshing for me to be able to access entrepreneurs that are not motivated by making money only, but really motivated by making a difference when I meet invited me seven years ago to go back to Silicon Valley. You know, just to add to that, you know, the transitions have not been linear and they've been juxtaposed by two very diametric opposites. So, you know, on one hand, there were movements that were starting to enable people to build beyond just the technology as the gating factor. And you see that now with AI a lot more, a lot of the new builders are not technical and you don't need to be right and that's, you know, one of the brilliant positives of where we are today. But at the same time, what decentralization was starting to do, starting with that, you know, iconic white paper that's Satoshi wrote it combined two themes at the same time. It combined the theme that here was technology that was able to take power away from kind of a centralized force, which is what we had, you know, lived with all of that time, everything we built was like that. And at the same time, people could start trying to see how we could fulfill these promises of wanting to decentralize access to wealth and power and access and all of these different, you know, functions that were deeply, deeply skewed to, you know, to a handful. And these two forces kind of came to it at the same time. And again, like I was saying earlier, this has been the journey over the past 25 years where you see a lot of builders and actors who come from the wider building they want to make access easier. They want to decentralize the opportunities around yield and investments and what have you and you can see, you know, all of those different themes. And at the same time, I think equally Silicon Valley has been rushing head first into this idea of look what I built. And then the next iteration of that is, oh, and I built it this way. So you talk to the average Silicon Valley founder. And by Silicon Valley, I'm actually referencing a broader culture, not just the location, because that's kind of what you're seeing around the world in these builder economies. And, you know, there's a lot of like you walk in and say, you know, like even the question of like, who are you? Is often encountered with I'm on it and I'm building this and I build it using the latest in what have you, right? And if that is your starting point as opposed to your why and your why is often if done correctly driven by the for whom? The what you build and the how are solving for that, whereas much of what we see coming out of Silicon Valley. And this has been kind of the challenge of that non-linear progress is it's been driven by the what in the how. And so if you take two Stanford engineers or Cambridge engineers and you know, or coming out of universities in Waterloo and they come out with this idea that we have access to the most amazing new models, Frontier models. And we want to, you know, take away human dependence on radiology, human dependence and law, you take your pick of domain, there will be only one outcome, which is that they will succeed. And so if so much of the vision is, how do we replace the human? Well, only one outcome will happen, you will succeed. And so this is the struggle that on one hand we have seen Silicon Valley give birth to so much of this decentralized, you know, how do we create deeper realms of access and opportunity. And equally the hyper centralization of power driven by no philosophical underpinning other than, oh my God, I raised at a billion, I exited at 50 billion, that's all that matters. Everything, you know, is irrelevant. It's a zero-sum game and we are living in certain politics of zero-sum game right now. It creates an incredible amount of friction between what can be and unfortunately what is. And maybe to move to the AI conversation, how would you say the Silicon Valley has changed since the advent of AI? Well, you asked about, you know, the early days and I'm it is going to chuckle about the early days because he was not in the valley and yet, you know, we were using neural networking technology and mutation algorithms to build prediction models. You know that on one level are still valid, but you know, Moore's law is a powerful law and the CPUs that we worked with about now 25 years ago were about 30,000 times slower. And so you can imagine that the data sets that we used to train the models for prediction were vastly different and yet similar on the small scale. And indeed, we were searching for business models that would enable us to make money with this technology. And you know, the only business model that we could come up with was that we monetized the customers, the buyers in supermarkets. So buyers in supermarkets at the checkout point used our software in order to get advertisements and we analyzed what they bought and analyzed in the back of they should have bought based on where they come from and what we know about them and then printed out advertisements in real time or sent them an email to do that. And frankly, you know, I mean that no clue what would happen to this model. I thought it was a cool model. I had no clue that this model of making us the product that scale would actually jeopardize the future of humanity and democracy. That just shows you how naive, you know, technologists, including myself were an hour to a certain extent and have no clue about what could happen at scale if this really goes out to all of that. And so while, you know, I am really not proud about that model and what we tried out interestingly enough, you know, it was still cool technology. And so it fit what I wanted to do and it we didn't have the ethical and more implications. We didn't understand that yet and we'll have a conversation about what that means for AI moving forward. But that's kind of where we left it at and I was quite impressed when the CEO after I left data mine was able to say this company to epiphany for about 400 million at the time. Remember 25 years ago and then I left AI until I meet invited me back to talk about it. You know, and it's it's, you know, one of the great ways to think about this transition if where we are today is actually think about it in human terms. You know, the early days of AI are very similar to, you know, a baby being born, right. So if you think about a baby that's just been born, it doesn't have any knowledge, but you don't need to teach it to rape. You don't need to teach it to cry. You don't need to teach it to have its bio functions, right. That's kind of it's almost like it's the pre training that it comes with, but it really doesn't know much else, right. Then you can imagine this, you know, baby is like a one year old and that one year old baby is observing you. It doesn't know that you are it's parent, but it knows that you provide for it without understanding what that means and it's observing. And a lot of that was that transition from the old days of AI, which was just like hard-coded you kind of, you know, just when the first supercomputer, you know, played chess against Gary Casparov, you know, it literally had all the moves. You know, it was hard-coded with all the moves and it just used those permutations and combinations. And then you know, this child is observing the child is being told things, you know, you've got a little, I don't know baby rhino and you're like, you know, playing with it and you're like, this is a rhino. It's an animal like that is, you know, you're categorizing it. That is how a lot of the training happens for AI. And then you can imagine this child is, you know, five, six years old. That's kind of where we are today. You know, you take this child to her first traffic intersection. And you hold the child's hand and you say, you know, you can't cross. You're waiting for that walking figure to go green. And when it goes green, that's when you can cross. And by the way, still wait for a few seconds to see if any cyclists are, you know, running through. And you probably need to do this only a couple of times. And then this child is able to adapt that anywhere in the world. It's not like they're like, well, I only know how to cross streets in America. I don't know how to cross streets in Canada, right. And that's that adaptive capability, the reasoning models, the cognitive capability that's coming into AI now. And you can see how so much of this mimicry of, you know, what it's like to be in a human starting from just pre-programmed capability to this evolving human. And you can see why some humans evolve with greater access intelligence context because they are, you know, they go to better schools. They have access to more information. It's the same thing with AI, right. Like the more data that is training it and to Charlie's point earlier, this is one of the issues with that attention economy where you're the product is it's being trained, of course, on, you know, all of our data, but it does make it this powerful engine that it is now. An infant writes better emails than I do or seems to at least and like many of us I have adapted AI workflows into my everyday life and it may be. much easier to commandean tasks of my hand, but there's always a bitter aftertaste because I feel like that like five big companies own me own my work on my intellectual property because of that I was wondering I meet from an insider perspective would you say that's true? It is entirely true you know I like to ask sort of an uncomfortable question to audiences often who owns you you know it's a provocative question it has a visceral response you know in the physical world the US fought its current longest largest war to ensure that no one has to answer that question in America anymore and yet after all of the progress of the civil war and rights around that you know the answer to the question who owns you to what you mentioned earlier Andreas is about seven companies and in the last 30 years of this attention economy I strongly mentioned the product is free because you are the product you know what we lost was data and why that's so much more significant now is that data is training models that have relative cognitive representation of whose data it was so suddenly there is a version of Andrea not in any physical form but there is a version of Felix is a version of Andrea there's a version of Charlie and it's interacting with with businesses it's interacting in the form of ads it's interacting in the form of of incentives and you know I often say that you know if you think about this journey all of these large companies that have effectively ingested you know all of our data for what has been a few decades now you know they're not in the business of selling data the Facebook Cambridge Analytica did happen and that was you know obviously very irresponsible and unfortunate but Facebook is selling the identity graph Google sells search Amazon sells and upsells and cross sells product based upon what it learns and knows about us and so whether we like it or not there are versions of us that are creating phenomenal amount of value for the businesses that have leveraged that and in many ways if we don't have control over it if we don't have agency over it and then the ultimate thing in this world of currency that we live in if we don't have economic rights related to it then the answer to the question who owns you is like Al Gore's inconvenient truth and it ain't you and I mean my data interacting in this way does it have to be a bad thing because it also offers some upsides I get the right advertisements I I get exactly the advertisement for the yoga mat that was interested in and I can buy it right away it makes my life very easy does it have to be a bad thing or to pose a question differently what would be a responsible usage of this data that would benefit me in more than just one way look like so you know convenience is a drug more powerful than most things and it is amazing how we have been drawn into this you know convenience economy and convenience is not always or oftentimes even at any time a bad thing the American economy has been built on convenience we got fast food because we wanted to make it convenient then of course you start looking at the health outcomes from that fast food you start looking at family dynamics because we don't sit at the dining table and connect anymore you know that came from this aspect of convenience so yes there are a lot of aspects of what big tech has been able to do with our data that has made our lives incredibly convenient and you know Charlie referenced this earlier of like you know we call it the unintended consequences of technology you cut and you know one of the big unintended consequences of technology has been even the right intentions the you know the road to hell is paved with good intentions you know you look at something like Facebook Facebook was born to connect the world and I would make the argument that Facebook has created more divisions than any product platform technology in in the history of the world and today your feed looks the way it looks because it creates the most most reaction from you right and that goes back to that attention economy so is personalization a good thing of course you know is convenience a good thing who would argue that right but to the second part of your question there are ways to do it the right way the problem with those ways is that they are not popular because they don't in no power to those who do it and that power has been rewarded with the most incredible financial return we have ever seen in the history of the world and like Charlie alluded to earlier an industry that was completely apolitical didn't understand or know what lobbying meant even though it made so much more money than the gun industry or their oil and gas industry it never lobbied it had no interest and that changed really dramatically and you can see over the past ten years the most influential voices around today's you know political leaders are the tech moguls and you know a lot of it has been around this idea that we can draw parallels right now to rare minerals where getting ready to fight a whole bunch of wars for who controls rare earth right and of course the Chinese have done this incredible job with being in Africa and building up the infrastructure and owning it and you can see that parallel to data you know you've probably heard their phrase data is the new oil it's that new coin infrastructure and I say the world of AI it's not the new oil it's the new nuclear fusion because what you have is not a one-to-one mapping you give me this data I give you this or I create X value it's exponential when it starts becoming AI right because it can do all of these incredible things that can unpack you know powerful you know you want to generate your documents and it's lovely it's fantastic for that but you know at what price has this convenience and this innovation come to us and if that price is our sovereignty our self- sovereignty our ability to be able to have a say if not even fully controlled what happens with our data because you know there's personally identifiable information embedded in that then you know we have to start questioning is that price worth it and to Charlie's point you know if the product is free you're the product and there are implications to that and I mean you know the leaders the newer older leaders who have let this transition from not being seen by Washington to lobby extensively for being able to make more money for their investors that has been really disappointing to me on a personal level because I found that when I talked about my career in Silicon Valley I was proud of it that I made a contribution and we you know it was indirectly and directed benefiting so many people and now we have the captains of the industry not just lobbying but openly participating in what in many parts of the world would be considered corruption you know and they're just really driven by how the financial system is rewarding you know it's investors and that's it and on top of that you know what pains me the most is that all of this technology actually right now is helping the old economy system to cement itself it's helping to cement the hierarchies of the control structures the surveillance structures of the old economy if that whole apparatus of Silicon Valley with its imagination and innovation would be repargeted at these bigger issues of how do we design a regenerative economy and a humanity that can thrive in it with the support of AI I think that would be a vision that I would jump on board immediately no agreed entirely and I think that that goes back a little bit to you know they not having a why not having a North Star not having a purposes to you know why you do what you do and you know the unfortunate truth is Charlie people like you are the extreme minority and you know obviously that's something that you know you and I have talked about for seven eight years you know how do we change that and it's been a slow process but you know just a quick example of that and you know we've seen it in the last you know year and a half in our lives and we've been like a bold but the way it's worked you know when this administration came one the election and you know if you recall meta was was not on board you know it was they were not part of the inauguration they were not part of you know what we're seeing now is the you know the power structure of Mar-a-Lago and you know at the time meta across its platforms had a massive not-class action but action from from the EU finds approaching about five billion and basically you know the current leaders you know sat with the Zuck and you know basically explained to him that you pay that five billion fine that's not a five billion cost meta can of course handle that that's about 50 billion in market cap when it gets announced right and it goes back to your point of like if all you're driven by a shareholder value and investment. to returns, then you look at that and go, oh my goodness, we can't accept that. So what's the solution? Well, the solution is get on board, right? Kiss the ring and watch what happens. And they did that in two days later, you know, a visit to Brussels, ensured that this leadership was able to explain how an attack on Meta and Apple and Google constituted national security issues for America, since they are, you know, too big to fail their strategic interests of the country. And everything went away magically. Well, the resultant effect of that was a 350 billion increase in market cap for the same platform. So, you know, guess what? We can talk about, you know, writing wrong thing, but, you know, here was a leader who came back and showed his shareholders in investors that he had found a way to increase market cap dramatic in. And it's only, you know, gone on because the regulations have loosened in. And, you know, they have so much influence. And so suddenly, when you start looking at ROI, you think, wait a minute, that's the best money I've spent in a long time. And that goes back down to what Charlie, you were saying earlier, like if you don't, you know, have that culture, if you don't have that philosophy, if you're not driven by purpose, you know, if you're idea of how you define the lens with which you view the world is, in the case of Google, do no evil on a napkin. And like literally already, do no evil. That's your benchmark. And then of course, you know, you never live up to it. Then we get these challenges. How can we change that? I mean, I mean, do we need different metrics to hold the accountable or does it need a different financing structure altogether like where do you see opportunities or leverage points to make AI something that can be or become a force for good instead of what it is right now? Like where do you see the actual leverage points to change this? I'm sure Charlie will have a lot of amazing inputs around this, but just at that, you know, as a quick answer, though, you know, I'm a big believer in so much of the industry and then the markets move around heroes rather than vision or category or, you know, even trends. So to kind of explain that point with an example, it'd be very easy to make the argument that electric mobility didn't give us Tesla. Tesla gave us electric mobility, right? Even though technically electric mobility game well before Tesla did, right? Gendered of AI didn't give us chat GPT, chat GPT gave us gendered of AI, right? And so oftentimes, it's the hero. I mean, you can think of like young kids, you know, that got into F1 because Michael Schumacher or, you know, Cristiano Ronaldo, you know, like the heroes, you know, drive the trends and the trends is what, you know, investors then jump on and then you start if it works, it starts to create this, this flywheel. And I think that if you look back at the last 10 years, which has been this evolution of the AI big 10, if you may, not just the traditional, you know, mag, mag seven. We've constantly been driven by stories of the heroes that have emanated from there. It used to be chat GPT, so it was Sam Altman. You know, then we, of course, we had Tesla before that and so it was Elon and then it was, you know, anthropic and now we've got Dario and, you know, the list goes on. And if you look at the subsequent evolution of the category, it's come after that hero has come to the surface. So like a billion people interacting with gendered of AI was not the case before chat GPT, right? Even though we had some AI products out there. And so until such time as we don't produce heroes that can show what the alternative looks like, that the economics work, that the broader return on value, not just investment exists. And so people can resonate with that. You know, they're going to follow whatever fields like, you know, like water flows, you know, to wherever the opportunities are. And today, look, if you invested in open AI, you invested in anthropic. If you invested in any of these frontier models as the AI example and you went in six, seven years ago, your returns, your multipliers are nothing the world has ever seen before, right? If you went into XAI or SpaceX, your returns are nothing the world has seen before. And what that inspires is multiple to put money in there. Are we moving towards a bubble because of that? I mean, Charlie, you and I mean, you as well, because you were in the valley around 2000. And you experienced the dot com bubble firsthand. Do you believe that we're moving in a similar direction again? Or what are some of the signs that you're reading? Yeah, maybe, maybe starting off here because I did manage in Silicon Valley during that bubble. And it was very painful. The 2001 technology bubble where the internet euphoria was just creating so much finair that the whole venture community and even, you know, people who don't invest normally started investing in it and putting that bubble on steroids until it deflated and came down very painfully for many investors and technologists. To me, you know, until maybe a few months ago, I really felt a lot of similarities between that bubble and what I perceived as the valuations, the insane valuations of AI were similar. And I could not imagine a scenario where all of these companies could grow into these valuations. And yet, you know, if there are a couple of winners, and I mean, then I talked about this and I mean, you can share your opinion on that in a second. But if you invest in the winner and there is a winner, then maybe that's justified, but everybody is, we lose their money stays. But still, it's still a bubble. To me, the consequences are really how many ripple effects are going to be happening if that bubble adjusted itself or burst. And the big fear that I have is, you know, the first iteration of financing of these bubble companies was done with free cash flow, meaning that it's not, it was, it was not borrowed money. It was actually money that they had on the balance sheet. And therefore, it didn't have any big effect on the financial market. But the way that data centers are being financed right now is a whole different ballgame. And when that bubble, if and when that bubble burst, then there's a lot more writing on that. And I do believe that I still believe here in the bubble, but I don't know how big it is. Nobody knows. Nobody knows when it's going to burst, but it's going to be painful for many people. And I hope that the implications for the financial system itself would be contained. Even so, on a different level, I think the financial system needs to break down anyway before we go into the regenerative economy, but that's a different argument. That's a different angle. Different angle, yes. I mean, what, what did you say? Yeah, and I would say, you know, in a very simple sound bite, we are probably in an AI valuation bubble. We are nowhere close to an AI bubble, right? These two things are very different. To Charlie's point, I think there are no categories that have been funded today in the AI space that will fail. There will be many companies in those categories that will fail. And that you can look back at the history of Silicon Valley and you've seen this throughout that time, right? And investors that are coming in at these very large valuations, mostly the Silicon Valley investors are not investing in one company in the category, right? So they invest across the category. If you look at the cap table of Anthropic and the cap table of OpenAI and the cap table of a few of the other frontier models, you will see very common names, a lot of those big funds and across all of those names. And so I would make the argument that fundamentally a company like OpenAI has the highest possibility of failing its valuation at failing its hype, because its economics don't work and something would have to change dramatically. They lose 86 cents on every dollar of inference that they deliver. Now, how many industries can that work with, right? You know, it's one thing to lose 14 cents, but if you lose 86 cents, you know, I meet the economics and in impact investing. If the unit model doesn't work, you don't go to scale, but they don't really do the opposite. They do the opposite. They're pretending like scale is the way to fix the unit economics. And it's like, you know, building on a foundation of little cards put together, right? But at the same time, there's some fundamentals that are changing. If you look at the cost of inference, compared to chat GPT 3.5, which is like one of the first main ones to where we are today, has reduced by a factor of 280X. Now, that is an exponential collapse in the cost of delivering inference. I think you have to explain inference maybe for the non-tech audience. Sure, yeah, not absolutely. So if you think about what these LLMs deliver, they are delivering, they're inferring what you've asked and giving you output, right? These are probabilistic systems. And so they're all known as basically inference engines and what they're delivering is inference. They're inferring what you've asked and then they're generating based on their probabilistic models what the outcome should be. We'll talk about this later. There's obviously a lot of issues with that. The positives are that the cost of inference is coming down. We're seeing a massive investment in data centers, good and bad, which will effectively reduce the cost of inference further. We're starting to see competitive models, especially from China that are open weight or open source, and that's dropping the cost of inference even more. So, eventually, we'll get, inference will be a commodity. We don't even think about how we consume the internet. We will AI will be like electricity and it will be powering all of our different businesses and our lives on the consumer side. But a lot of these companies like OpenAI have fundamental challenges that will get better but cannot be solved, not with the model and the structures they're built up with. But the facts are that 88% of global organizations have started to use AI. That's a very high penetration. At the same time, that they have actually implemented AI that works is less than 6%. So, if you think about that, we get caught up. It feels like AI is everywhere. It feels like AI is consumed our lives. And yes, that's because we ask the most inane, mundane questions to our favorite whether it's cloud or chat GPT or whatever, but that doesn't mean that it has become incredibly pervasive. 62% of these organizations are already experimenting with agents. So, they're already at the next era of AI, not the direct, large language model, your chat GPT or whatever, but the agents powered by them that live on top. Right? So, there's a huge opportunity and that's why when people say, "Oh, AI is in a bubble," I'm like, "Well, that is not an educated comment to make." We are in a company valuation bubble. I don't think we're in a category valuation bubble. And some companies absolutely don't deserve to have the valuations that people have put money in. But a lot of investors in Silicon Valley and around are putting money into the experimentation around the category. And they're okay with that particular company failing or those economics failing because of what they believe that they can learn from that for the entire category. So, from a pure AI perspective, super optimistic. You know, since we're talking about numbers, at the same time that that cost of inference is dropping for the frontier models by that massive amount, the frontier model transparency score. How did you train? How is it coming up with this generative outcome? What is its reasoning transparency? Has dropped by a factor of about about 18%. So, a transparency score that was at 58, which is higher and now it's dropped down to 40. So, again, it goes back to like our earlier conversation, Charlie, was mentioning as well. Like, we've achieved more capability, but at what cost? And because so much of our conversation at a mainstream level, hasn't cared about that cost. It goes back to that point of convenience is an amazing drug. You know, we've ignored those things. And I know we'll talk about a lot of the other negative implications of AI and what responsibility AI means. But, you know, these are kind of the issues facing the industry right now. And I think there is a genuine belief. And this is where that Silicon Valley attitude tends to be, which is like, okay, great. We'll solve those things. We'll make in Friendster cheap. Infrastructure will become even cheaper. Chips will be made in their gazillions. We won't be only reliant on Nvidia's current production. And, you know, we'll find new materials to make even faster processes. So, why do we have to worry? Let this model continue. We'll continue making money hand over fist. And, you know, billions are not cool anymore. It needs to be trillions. But for what, right? It for what? To me, it's important for all of us to understand how the current wave of AI actually works. And that will give us a clue what it cannot do right now. And the current wave, as exemplified by large language models, not by all the other AI that we can also talk about, is really language-based. And so, anything that's beyond language, the emotional side of being human, the human side of being human, is not taken into account. It cannot be taken into account because of the technology limitations. And so, then AI does pattern recognition of the data sets that it inhales. It makes predictions based on very sophisticated analysis and algorithms and statistics. But notice where it gets its intelligence from data sets. And as long as the data sets represent the biases and the lack of understanding of humanity's future, AI amplifies these biases and plays them back to a second mirror. I was like really shaken almost by a work that Vanessa, who wrote hospicing modernity, did in co-writing a booklet about AI that was called, what was it called now, you know, there? Well, put it in the show notes. Yeah, show notes. In any case, you know, we ask AI the questions of how can you become more ethical, not understanding that AI is as ethical as the data set that it inhales and not more or not less. And right now, the data sets that they inhale are actually the biases of the mostly the white people in the global north. And the white people of the global north have created the existing economic system, which is currently tanking the plant. And so, unless we change the data sets, we cannot expect AI to come up with other things because they cannot make things up that they don't inhale. But how can we change that, Charlie? How, what's this emotion? That's why I explained, you know, that the alternative for this crop of AI is other data sets that actually encompass encapsulate a new awareness, a new ethical way of being, and the way to do that, Andrea, is that we need to change ourselves so that the data sets that we create in body who we are on that level. At that point in time, AI will be totally in alignment with who we are and what we want to do. And but that's a heavy lift. You know, last time I asked the research boys and they're all men of meta and Google and Facebook. If they have any responsibility to create data sets that actually matter for the future of humanity, they all three of them said no. We just care about having enough investor capital to buy billions of CPUs such that we can run our models better. They learn better from the existing data sets and now that we're running out of data because all data they have inhaled, they create synthetic data sets and once you start training the models on your own, the yes, if I might say so, then the models actually think. And so, you know, the current crop of AI has significant problems that itself it cannot solve, but we, as humanity, need to solve ourselves. Yeah, I cannot agree more. And I think that it's kind of this loop, right? And there is a movement now to like, you can choose not to have your prompts, become additional training data when you use it came from regulation. We are seeing improvements from that perspective. The problem is it is resulting in more hallucinations. It's resulting in these, it's the term used for when AI makes things up, right? Make things up because it doesn't have the right data reference base. And you mentioned something much earlier in this conversation, Charlie, which is very important. The elder knowledge that has existed in cultures for such a long time. If that is not priority in terms of what AI is trained on, well, guess what? Like we're starting to see to your point, Charlie, with global North perspectives being the truth that AI understands in beliefs, it's now not just how it's being used. It's codified for the future. You've got to remember these frontier models are the foundation layer for how we're going to live our lives in the future. They're going to power, you know, all systems we interact with, without even realizing it, our banking systems, our retail systems, when you go and bear to shop, when you, everything we do is going to be powered by these frontier models. So kind of like law, kind of like a codified law that we live under. Absolutely. It's even more than that, Andrea, because it's normalizing our thinking. So if everybody starts using the same inference engines, if I may use that term, and they spit out similar things because they're all programmed similarly, well, then that influences how we think. And the most destructive consequence of this is that it drops humanity of the opportunity to imagine a different future for humanity, because we're all driven into an algorithmic outcome that nobody decides anymore. And we together don't have the awareness to fight that back. And so we go back into individual, you know, behavior, but that is the huge risk on the human level that I see with AI if we don't change it. So does it kind of make the corridor that we can walk down narrower? Is that what you mean? Yes. If you think about the perception of that, we think we have free will, right? And to a certain extent, depending on where you come from, that might be true or not. But I certainly think that I have free will to decide some things that do. But as we all are influenced by algorithmic guidance, if you want to use that word, I would say coercion, right? Then we rob our saves of things that we don't see anymore. And that is the problem. We rob our saves of the infinite possibilities that human evolution has given us as humanity by delegating that to algorithms. And that is very dangerous. That's exactly true. So how can we move towards a wiser and a way that Charlie used stated and more creative AI? Maybe I see this with one more thought and then I'll make please take over because it's following my logic of what I just said. So if I think about what AI should and could do for us, independent of how it's programmed right now, I believe that we need to fundamentally change our economic structure from an extractive one to a regenerative one. And AI can certainly help do that. It is brilliant in analyzing backwards looking, of course, you know, what's going on. It's brilliant in inferring between different systemic issues, between agriculture, development, climate change, building out traffic, you know, mobility. It's getting so complex that AI can be hugely beneficial for us if we put in the right data sets to analyze what's happening or be making progress in moving towards regenerative words or not. And so I'm very, very positive that if we muster the will to use AI in that way and create the right data sets that in a few years, AI will be a tremendous hope, you know, moving, moving us in the same direction. The second issue that I see for humanity that I wonder what AI can do for us is the elevated consciousness and awareness that I said is necessary because we can't really force legislation and regulations on everybody, including AI, to create a different future. And there, I mean, you know, we talked about this before. Some people in the AI were think that AI itself might develop the level of consciousness necessary to guide us that way. I totally don't buy that, especially not with the existing design of what's going on. I just meet the case that as a matter of fact, it cannot possibly happen with the current AI technology. But still, I wonder how can AI help our consciousness, the human consciousness, do developments such a way that we can take responsibility for our own future again. And you're much closer to the AI chiefs right now. There is maybe unintentionally, maybe unintentionally, people confuse consciousness with AI. Yeah, and look, that's a much deeper conversation that I think unfortunately is not happening at the places that are of influence. But, you know, there's like two parts that you touched on here. One is, you know, what is the potential of AI? And again, it, you know, help us with, you know, finding cures for new diseases, increasing the velocity of drug discovery. We can do faster clinical trial testing. We can do amazing things with financial literacy, financial access. You know, all of these things that AI is, you know, incredibly powerful at doing autonomous systems, making roads safer. You know, there's so many things that, you know, it's easy to get caught up into some of these challenges that AI is creating. And you can't ignore the human challenge here. You know, the, the global jobs or like advanced economy jobs that are going to be exposed to AI, which, you know, like Andrea, I was telling you, the, you know, when you drive up from San Jose to San Francisco, like every 10 feet, there's a new build board of an AI startup. And every alternate build board talks about how many humans it can replace. If that is your fundamental aim in life, you will achieve it. And we're seeing it. We're seeing it being achieved. There's a, you know, 40 to 60 percent of 40 percent of global jobs, 60 percent of advanced economy jobs are already exposed to AI. And we're seeing kind of that temporary, oh, we need to hire more forward deployment engineers. We need to hire some like it feels like there's more hiring, but it's sort of like that, you know, you know, what a ship is about to, to sink its most powerful when it's preaching the surface. If you're close to it, you'll get pulled down. But after that, it's just this helpless, you know, Titanic that has gone and sunk to the bottom. And, you know, there is this feeling that, oh, maybe this whole job loss thing was, you know, exaggerated and yes, temporarily. But once the impact starts happening at enterprise scale, you know, these implications are going to be really wide and broad. But the problem I think is think about the questions that are being asked at OpenAI right now. As one example, you can apply that to any of the frontier models and the harness providers. And of course, the infrastructure providers. How do we improve our unit economics? How do we increase the valuation of our upcoming IPO? How do we get users to subscribe for a higher value? Those are the three questions. Do you think Charlie's questions of like, you know, the human state, human consciousness? What does this mean? Did the impact to society long term? You think those are the questions being asked? They're not. And so as long as the economics are being powered and driven by those who are asking questions about just that infinite loop, we don't have at scale those, those questions being answered. And, you know, we'll talk about rainfall later. But, you know, it's something that we're really focused on doing. And, you know, Charlie is a front-end center to the rainfall journey. You know, had we not cared about some of our core fundamentals of responsibility I and I know we'll go into that shortly. But if we didn't care about those things, we would have been able to raise any amount of money at this point. You know, from investors who I've had meetings with where they've set things like, if you guys pull this off, you will have more access to data than volunteer. Imagine what we can do with that, right? And from our perspective, we're like, I know this hurts us, but we're going to have to walk away, right? Because this is precisely the opposite of everything we stand for and why we build. But, to be very honest, outside of small startups that are resonating with that philosophy and mindset, the ones who are really moving the needle, both in markets as well as their technology deployments, are not asking those questions. You know, I don't think Sam stays up every night thinking about this consciousness issue, but definitely thinking about AGI and ASI, right? So AGI for the people who perhaps, you know, don't know, artificial general intelligence, the best way to think of it is when AI can be smarter than any doctor or smarter than any lawyer, right? And you have that ability to be, you know, beyond the best doctor there is. And then there's like the Holy Grail for a lot of the decade-lead, which is artificial superintelligence. And we're starting to see super intelligence labs being built when that AI is not just smarter than any doctor, it's smarter than all doctors put together, it's smarter than all lawyers put together. And you can imagine what that starts doing to that number I was saying of like human employment that will get affected. That's what keeps them up. What are the conversations that you try to have when you meet people like the Sam's of the Valley? Do you try to talk to them in a way and open their minds to these questions? Yes, in terms of the conversation, no in terms of any sort of meaningful impact to be very honest, right? And this is not a, I mean, maybe this is a poor way to frame it, but if you think about who's building the frontier foundations that modern society runs on, it's largely autistic men or men on the spectrum who think that like human connection is a weakness, empathy is a weakness, you know, Elon actually said that in a public speech, you know, this is not conjecture, that the biggest weakness of Western civilization is empathy. And if you are building the foundational models while, you know, sleeping on beanbags in the office and your most human interaction is with your full delivery person or, you know, which will become increasingly physically eyeed, but, you know, robot that shows up. Then you don't even have the context to have this conversation. You can have this conversation when, you know, you even have that context, you know, you're talking about those narrowing walls. If all you can see is the scenery inside those walls, how do you even have those conversations outside? So then when people like me bring up these conversations, quoting people like Charlie, it comes down to how much have you raised? How much penetration do you have of this responsible technology? And, you know, that goes back to my earlier point of well, I hope it's us, but it doesn't matter if heroes start to show up. That kind of have that escape velocity of doing it the right way. Guess what? It'll get people's attention and people like Sam will want to have that conversation, but right now the only conversations they're having. And like always in Silicon Valley, right? I mean, I would not count on the current elite, the current oligarchs to change that. One of the reasons I invested in new is because you are representing a different future than is represented by the current winners, I use their terminology in Silicon Valley, but I personally have witnessed three psychos in Silicon Valley where the people who thought they were the winners were not the winners anymore ten years later and I hope that that's true without going too far down that rabbit hole of losing the momentum for humanity itself before that happens. That's what I would like to maybe focus on is, well first I can share why I invested in your company, it's because it articulated a very different vision and future than the existing technologists would portray and in a way that really resonated with me and then you gave me an opportunity to cope with the vision for an impact rainfall and we can talk about this because it's still very very deep in my soul to be able to use that and we add our challenges and our tries and tribulations, but that doesn't mean that we cannot figure it out, I think we will figure it out but you know the headwinds are quite strong, the financial models are not aligned and the power structures are currently not aligned and yet everything that seems so powerful eventually is not powerful anymore and I hope that it goes faster than slower and then I want to be ready for that and ready with what you can articulate much better but I was you know maybe to see that conversation I was so intrigued by the way that you earlier years ago talked about how beautifully it would be to enable moving from the idea of universal based income to universal earned income based on the data that you generate in support for humanity right and that's a wide and big idea. Now I'm curious, I mean would you mind like giving us a quick elevator pitch so that can valley style? Yeah, no, absolutely, thank you Charlie, you know if you think about it modern society largely has been built on this idea that human lives are valued based upon physical land that we inhabit if we're lucky what we find below and we know we have you know extracted that to the hilt to the cost of humanity and this planet are stay on this planet and then what we build on top and so you can see the evolution like we had the agrarian revolution, we had the industrial revolution, we had the automation revolution and now we can argue that we have kind of this autonomous revolution and on all of those revolutions we built kind of modern capital markets, 10 layers of derivative products on top of that, right? But you can look at the 10th layer of derivative product and where it in your it's core value from was land factories, crops, oil, everything came down to kind of us being valued based on the land being inhabited. The unfortunate truth is that you know 90% of the world's land owned by less than 5% of the world's population. So the economic benefit whether at the first layer or the completely abstracted derivative layer, all in yours back to the land owners, right, that owner economy and we're seeing that extrapolated now and yet if you think about it in today's world there is not a single footstep you or I or someone in the global South can take on that piece of land that doesn't generate a digital footprint. On location a number of steps, what we visited, what photo we took, whatever it might be, whether we're doing it with a beautiful iPhone 17 you know in the western world or someone is doing it on a $10 Android device in a small town in India. The fact is that we are generating this digital footprint everywhere we go and if you think about it, if that 1910 number felt skewed, imagine what a 99.99% to 0.0011% number feels like because 99.99% of the world's data is now controlled held monetized by 0.001% of the world's population. So imagine based on that if we could start making the shifts and say, wait a minute, I'm generating that data, why can't I share in this? And if we can start creating that same fundamental parallel and saying we valued human life based on land we inhabited and now we'll value it based on data that we all generate and then generating that data is our work, whether I walk somewhere, I'm dealing a farm somewhere or I take a photo, it's my work, right? It's work that has changed a lot, but it's still my work. And if I can earn from that work, then instead of relying on this idea of when everything goes to hell in a hand basket and AI has taken everyone's jobs, will we rely on UBI, which is someone paying taxes, a government coordinating it and then delivering it. And instead of that, we can actually focus on UEI, the universal earned income where that earning is not magical, it comes from data that we're generating data that has immense value and it's not theoretical because guess what, trillions of dollars of companies have been created on that data, right? So it's not like a shocking new model of like what data will have value, this is such a shocking idea. And if we can transpose that back to where the data originated, who created it, you start getting into these principles of self-sabrinty, self-sabrint data and as we transition into this world of AI, where the AI is like a force multiplier of that data because it unlocks so much contextual value that we can start creating this model where people can earn a daily potentially sustainable income from the use of their data. And I would make the argument that as important as that will be for the future of humanity, that actually this is an argument I like to make a lot, this benefits entrenched powers. This is not the enemy of entrenched powers. So like it's important to not have to fight them in that way because they are obviously too powerful, but eventually even going back to the earlier conversation of these AI systems and what they're trained on and if they don't get enough data, they start to hallucinate more and more is that if we can create a sustainable model of consistent, responsible data flows into big tech systems or small tech systems, the systems will get better and the better they get, the more it will unlock in terms of the next phase for AI, which is the application layer. And everyone stands to gain money from that application layer if done in the responsible way because people will be one of those line items. So when a company says, what are my constructors for how I've delivered this dollar of inference, there'll be one two cent line item which will say, Andrea, I'm extrapolating. And that just creates a much fairer, a much more sustainable, and guess who loves, who hates, UBI is capitalist because it suggests free money for someone paid for from my taxes. But UEI, a capitalist loves it to be earned and in reality, the person loves it to be earned as well. And so that's the idea behind, you know, these core principles that Charlie and I have talked about and worked on for so many years. But will someone in the global south then earn the same amounts that I would earn with my generation of data and do you see the future in that way that many people will kind of lose their jobs or their jobs will change a lot due to AI. And then this will come in to kind of level it out or what's your vision when it comes to that. Yeah. And yes, absolutely, we're going to have to find ways to make this sustainable because it's not enough to say, oh, my goodness, AI is going to create so much value. The world will be a better place for it. And it's like, yeah, but someone still has to go and buy groceries, which they were doing from that job that they had, you know, working as, you know, a paralegal or what have you. So we definitely need to find ways to solve for this. To the first part of your question, yes, not all humans are equal. Not all data generated by them is equal. But what's important in this world of AI and that's what, you know, people often miss. Yes, the four of us on this goal have way more value when it comes to, let's say, retail because we go to get coffee and we go to get a nice meal and we go and pick up, you know, shirts from wherever. And so the data that trains helps train to Charlie's first point of like ads that know where you're going and where you want to go and you want to buy shirts in any case. So may as well give you some deals around that are very relevant. And therefore your data, Charlie's data, my data, Phoenix's data has more value. But take, you know, Rashida, the refugee mom sitting in, you know, one of the camps of the Norwegian refugee council, I served on the board for six years in Lebanon, the Starbucks or, you know, the coffee shop or the restaurant care about her data, no, obviously not. But if you're AstraZeneca or Pfizer or buyer and you're using AI now to accelerate preclinical trials on the new antiviral that you're building or the, you know, the vaccine, guess where these new germs and new viruses with morph, where conditions are more suitable for them, right? So you can imagine refugee camps, you can imagine the streets of Mumbai, you can imagine, you know, this the smaller poorer communities in Sapsar in Africa. Suddenly, their data has more value than yours or mine, because we're sitting in places where, you know, it's it's all safe and clean and all of these things. And so once you start looking at it from a use case perspective, and not just from a generic data perspective, to the first part of your question, you can start to unlock the kind of value that if you were just comparing an apple to an apple, you're absolutely right. You versus Rashida, no comparison, right? But then once you start mapping it to wait a minute, there's so many of these use cases where her data is more valuable, you can start to create fairer models. I'm not saying they're equal models, but they're fairer models. So on that level, right? On our journey together, you know, the universal earned income is a moonshot. And the technology will support this, but we then have to reflect on, okay, how do we make a business model that actually works? And that's when we keyed off of what I made it just now. And that's, you know, we were exploring insights that we can generate by people volunteering anonymously and with privacy preserving technology to work with us. These impact insights can we make money with these? Can we say that Mr. Marketplace of these just like there's a marketplace for a second hand clothing or whatever? And who would pay for that? You know, weeks for that by in depth, Amit and tried to work with other networks like Joy on the enable four million moms around the world do not get monetized by Facebook, but by actually earning something and working with rainfall. And yet, you know, this was also too early because there's not that many marketplaces. And more importantly, there's not that many people in organizations who want to buy these insights. And so we have to temporarily abandon that idea too. And now we're exploring Agentec AI. And I'm really working hard with Amit and others to figure out how can AI agents be used at scale to accelerate the transition to a regenerative economy to help us realize who we are. And while we go hopefully, you know, commercial to make the whole thing viable on the financial side, but really again, use it in a way that is augmentative to the future of humanity as opposed to just accelerate consumerism only. And Amit, what I really value, you know, in our partnership on that is that we didn't get stuck in one model that temporarily doesn't work yet. I mean, if you grew up in Silicon Valley, so to speak from a business standpoint, then you're not used to giving up you adapt. And you then say, well, this didn't work. Wonderful. I've learned something. What will work? And maybe you can paint a little bit of a picture, you know, in language that's a classic. And I have to challenge myself to understand that sometimes, but that's why these interactions are so important. You know, of this revolution of agents running around for saving humanity as opposed to thanking it. Yeah, maybe to close off our conversation, I would be really interested to hear what kind of practical examples of impactful AI usage could come out of that, which applications you have to offer with rainfall. Yeah, no, absolutely. And to, you know, what Charlie mentioned earlier, so just, you know, quick context, you know, we, we built, we called the first self-sovereign AI platform. So instead of thinking of like a super large model that, you know, serves the whole world, which is, you know, obviously these frontier models, we focused on the individual. We basically said, you know, here's deterministic AI that represents Andrea, that represents Felix, that represents Charlie. But for it to be truly self-sovereign, it must belong to Andrea. You must have control over it. You can switch it off if you want. And that data that trains that model, that is the, what we call the aura of Andrea, that data, the raw data has to stay with you. Because if it stays with us, well, guess what? We're just, you know, a wolf in sheep's clothing, right? We've done the same thing. And, you know, yes, where we've got good intentions in all of those things. But show me a tyrant, dick data, and I'll show you a one-time freedom fighter, right? And so by architecture, you have got to prevent this from happening. Because, you know, if rainfall is one day, you know, acquired by, you know, a big company, or we have, you know, investors that, you know, at this point, it's a hundred billion dollars. And I'm not management. My team is not management. You know, how do you make sure that architecturally this doesn't change? How do we make sure that it's not just volunteer disguised in this, you know, fluffy pink, you know, costume? And so, architecturally, we don't take your data off your control. We don't have servers. We don't serve, store the data, things like that that we built into the core. What Charlie was saying earlier was, you know, an interesting part of this journey, like we were saying the road to hell is paved with good intentions. The road to happiness paved with ridiculous, you know, obstacles and course corrections. And, you know, and one of the things we had along the ways OS platforms changed how we could access data. And so if we can't access your data in the way that was intended earlier to just be able to compensate you for that data, then it really puts the whole model in question. And this goes back a little bit, sorry, to segue, but to earlier conversations of, you know, why things have not changed, you know, what's preventing these thoughtful conversations with a SAP, you know, with a Dario, with the people in and powers that be. It's because it doesn't align with their incentives. And until such time as those incentive structures can change, either because regulations have caused it, but more importantly, I think ground swell movements of people realizing, wait a minute, if I can start making money from my data, if I can have control over my AI, if I can control my privacy and we all have different spectrums of what it means to have privacy. And that's perfectly okay. But if you have that ability and that control, does that start to change the way people think about what technology they'll use? What product resonates with them versus what doesn't? You know, Tesla is a great example. You know, people who bought high-powered gasoline cars, especially in California made that big switch to Tesla, not because, you know, they think South Africans make great entrepreneurs. It's because, you know, they align with a philosophy of, you know, a cleaner, more responsible, et cetera. And that drove legacy car makers to create massive EV programs, right? Which otherwise, you know, would, you know, it's by Charlie alone have driven, you know, a Ford or a BMW or a Toyota to make an electric vehicle. Well, you know, we've seen the trajectory of that. So, you know, unfortunately, we had that big setback where without that access, you know, it really brought into question the model, but it brought into question, you know, we can adapt, we can stop doing that. But then does that still help us achieve the core mission of why rainfall? We could sell, sell out. Does that help us achieve why we're doing this in the first place? And of course, the answer is no. And so we went through a large reimagining rainfall process. And, you know, Charlie mentioned this. We realized that, you know, we're starting to see, and I quoted that number earlier, 62% of all enterprises are now moving agentic. And one of the big challenges in this agentic move is that because these AI systems are probabilistic, you know, sarcastic models, they're inferring what they have to do next. If you ask the same question to your same choice of AI in the same chat twice, you will not get the same response. So, good reason is a probabilistic model. It's generally now, hopefully it doesn't hallucinate, so the facts remain the same. So you provide coherence. And we provide that coherence, especially now in the world of agents. It's become even more critical because agents need to perform really, you know, precise functions and need to do it repeatedly. And AI does a very poor job of that. And so what we were able to do, as we said, create, we built the aura for Andrea. So here, the actor was Andrea, the human. But we can now treat an actor as an agent, as an artifact, as code that you generated in Claude code, or in lovable, or wherever, and still give it that same deterministic boundary layer so that agents can start to behave in this, you know, precise way that you expect them to behave. But what we did was we said, instead of pushing this privacy and self-sovereign AI narrative as the selling point, we baked it into the core system. Andrea, since you mentioned coherence, I want to maybe mention that we want to connect islands of coherence that are meaningful islands to make a positive contribution to humanity in the planet, to archipelagos of coherence. And then from the bottom up, we have a chance of turning things around, maintaining our agency, why things become a little bit more turbulent. And I believe that because our approach is fundamentally aligned with the purpose of humanity, will in the end win. I agree. I agree completely. And I think that the reason I'm definitely optimistic is that I feel that we are seeing a convergence of responsibility and economics. And I think responsibility with that economics and economics without responsibility gives us the system that we have. And when you start seeing that convergence, even though it's slow, even though it can be tedious, it's not at the pace that Charlie or I would like. You know, what we've realized is people, therefore organizations, have their own cadence for their journeys. And there are always little little treats and inflection moments along those journeys that can allow for tides to turn. And I feel like we are seeing more and more conversations around self sovereignty. We've seen one of the hottest areas of AI development in the last year. And this was highlighted in Davos at the weft this year is sovereign AI, meaning that AI remains in the sovereign control of a country, a state, even an entity. And yes, that is not the people yet. But you know, you don't get self sovereign without sovereign already in there. And so, you know, I am optimistic. We are seeing that conversion. Some ecosystems would move faster than others, but you know, it really comes down to can we on one hand create the right incentives? On the other hand, can we solve these problems? You know, very quickly, we're working with large enterprises now. And they have spent millions and millions of dollars on their AI implementations and their stock. Their stock because once they start taking them out into predictable use cases, they're finding that their systems are drifting. You can imagine in the world that Charlie was building software. It did precisely what it was intended if you ran it once or a million times, right? That was what it did. And there were a couple of our weird bugs. I mean, you know, that we chase down. That's true. That's true. That's true. Once you got past them, you know, in general, you managed to achieve that. And today the reason people are moving AI native compared to that software architectures because you want the contextual breadth and bandwidth of these models, right? But then now you've lost that preciseness. You've lost that consistency. And so we're able to now go in and say great, we can do this much faster. And we can make this happen. And that's driving a lot of the incentive because people are like, great, if this works, we're willing to go down this path. And they automatically get privacy and self-solvering AI and all of those pieces tied in. I think that's a nice way to wrap up this conversation. Thank you for giving us a glimpse into what you're building. It makes me very hopeful. I can sleep better now and knowing that you guys are building rainfall. I was wondering just maybe in three words, thinking of my 19-year-old brother, who was, you know, about to enter university at the job market. And this crazy time from a Silicon Valley perspective, what do you think in three words? What are the most important skills those young people need to establish for themselves to not only survive but thrive in these times? I think the most important skill in three or five words is to know who you really are. Adapt, absorb, own. I think those out three kind of core philosophical principles, I would say the next generation needs to think about as they interact with the future. Well, thank you. Thank you for this great conversation. Thank you. And always great to see you amid in any context always in the big country and Felix. Likewise, thank you, Charlie. Thank you, Andre. Thank you, Felix. Lovely to meet you all as well.