AIBRASIL - Eduardo Saverin brazuca socio do Facebook entusiasta da IA fala do futuro

I'll go back to English now. Everyone was passionate about technology and the impact technology is going to bring upon the world. I can talk about today the early days of Facebook, now known as Meta, and how we got to where we got. I can talk about AI, which is basically endemic today. It's pervasive and it's one of the key paradigm-shifting technologies that's coming to bear. And I can of course talk to the heart of your question also about the concept of global. But to talk about global, I can first tell a story of myself, a young college student in Cambridge, Massachusetts near Boston, trying to empty a very small bank account and instead of going to my classes, working on entrepreneurial ventures. In this case, trying to develop a way for me to be visible to my college community, initially because I had missed the photo-taking deadline for the physical Facebook. This journey brought upon Facebook, eventually Meta, but we thought at day one we were building a product to make our lives better at our specific college, eventually at the colleges in the countries near us. But what happened in the early days of Facebook is that it became very quickly a global company, and I would say arguably initially accidentally, where India, Brazil, and a bunch of countries became massive in usage, Indonesia as well, and many of these countries became larger even before anyone in our company had visited the country. So that's why I called it accidentally global. I think today it's important for entrepreneurs to think about global by design from day one. And that's because we have an incredible ability to scale on top of other platforms that are built, whether it's old-school mobile platforms to what's happening now in AI and a lot of other forms of distribution, partnerships via global enterprises. So growing global is critical and important whether you are starting a company in the US, Singapore, São Paulo, Brazil, or effectively anywhere else in the world. It is important to think global. Frankly, even entrepreneurs in China, when I first moved to Asia quite a while back, I went to China and most entrepreneurs back then would think about the market locally being so big that they didn't need to think about globalizing. Today, people anywhere in the world, including in China, think about how to expand beyond their country. So one point that I wanted to make is global and thinking global from day one is critically important. But for my story as to why I'm based in Singapore, it is much easier for me to describe it to you. For me, beyond my incredible passion for technology, family is central and key to who I am. And I live by very simple words: wife happy, life happy. And my wife was raised in that region of the world. So that brought me to the region initially, and then I saw the massive opportunities coming to bear given the population growth, the youngness of the population especially in South and Southeast Asia. So it's an incredible opportunity to also be in a region that will be a big part of the world's population long-term. But Andrea, let me send it back to you to make sure I stay on topic here. And I can talk qualified, hungry, what I would say scrappy entrepreneur type of environment. And in some ways, the technical downturn that you would see China specifically going through the last few years, if you would argue it's sort of a downturn or a slight deterioration in the technology market in particular, you know that is actually a prime environment to train entrepreneurs and talent to be highly successful in the future. Actually, a lot of the best companies in the world were built during downturns or periods of incredible volatility. Right? If you think about Facebook itself, I mean we started in 2004 right after the dot-com bust, and also we went through the 2008 period where we were still growing. And you look at many of the largest, most well-known companies in the world, Apple, WhatsApp, Microsoft, they all got built in their early innings during periods of incredible volatility or downturns in many ways. It gives you the appetite to do things right, focus on fundamentals because money is not as available. It's harder to get. You have to use it wisely. And what you see in Southeast Asia as an example of just the strength of the entrepreneurial talent coming out of North Asia is you see a lot of the top companies in this region are built by entrepreneurs that have moved south. And you see the same thing happening frankly if you look at the US and you look at some of the top companies built out of the US and you looked at the immigrant population that is a part of it. You have Facebook, myself, I was an immigrant into the US from Brazil. You had Instagram, I believe one of the founders was also Brazilian, and you have many others across the board. But from Asia and North Asia in particular, while you may be looking at the China market itself, you really need to understand the power and impact they bring to bear as being an invigoration into the entrepreneurial DNA of the world. And I'm a big believer in 1 + 1 equals 3, which means I would fail my math class. But genuinely, even when you look at AI and everything that's happening in the world, a lot of people ask me the question, who's going to win AI, China or the US? But my view is that DeepSeek comes out and puts forward an open-source model that is more effective, more efficient, and people could argue as to the foundation and how much they spent on it. It just makes it such that companies in the US and Brazil and other parts of the world can now leverage the best-in-class learnings from those models to innovate even more quickly. But maybe I got distracted from your original question. Initially, it was a testament to improving ROI of capex investments. Right? You can distill the entire message here. And at the forefront, if you were to believe the argument that a lot less went into training the models to get a model of a decent quality not far from some of the top premier models in the US, the argument here was that the productivity per dollar spent improved in AI. And that's both from a training models standpoint but also from the perspective of using the model, the inference layer. Why that's critical is applying that message to almost every other type of business world. If the ROI improves, the use case and applications increase. So in some ways, the first argument was, oh, this should drive down value of Nvidia, this should lower capex investment needs. The truth is if ROI is better, you should initially invest more in capex because you're going to get a higher return out of it. And what that ultimately means is that you could start leveraging AI for a lot more things in the world where the value that you're receiving from using AI may have been smaller. Really, when I think about AI and what its ultimate business model is, it's less about creating a fundamental LLM model and then charging per API call or per use. That will go the way of SMS, where eventually people are not going to be paying a dollar per SMS in the mobile industry. The use case is ultimately applications to the real world. If you think about how AI started, it started as effectively a feature, a button, a chat interface, what I would call as AI 1.0, where it basically piqued people's interests. But if you think about what AI really is going to, it's embedding into individuals' context, which is knowing everything about an individual, knowing everything about a company or about a sub-person within a company, and creating incredible productivity and scaling productivity initially by lowering the spend of human beings on things that are repetitive and less creative. And through time, making engineers 10x, engineers 100x, making business development people more effective, and making effectively everyone more effective. Eventually, we'll have also the physical joining with the software layer, which are humanoid robots, and we can talk about what that would also do. But the most important thing to flag back to your initial question about LLM models, DeepSeek, OpenAI of this world, is even if you take the top model today and you go even to lower than what the top model today can do, I believe using an 80/20 rule that the majority of productivity gains in businesses and enterprises could be achieved with a proper application layer which would then workflow integrated into that business. So we don't need significantly more advanced premier models to get you there. How do you make money out of this? If you look at Meta as an example, the first use case is improving advertising targeting and advertising efficiency. So that's going straight into their core model. Second, it's improving feed recommendations and algorithms to improve the usage and engagement rates, again going to the core business model. So actually you see a lot of examples of companies that have scaled business-wise AI to massive scale, Meta being one of them. But Raj, I'll let you speak. I think we don't need better infrastructure and we don't need better models than what we have today, even though to be clear, models and model quality is going to accelerate very, very quickly. What we need is to apply existing models and frankly even smaller versions of existing models to very precise use cases one at a time in businesses, for personal and individuals. So let me give you first a broad universal example that everyone would connect to. If I gave you two choices. Choice number one, when you're looking for something, i.e., you're about to affect a search online, there's huge businesses that have been built on the back of this, i.e., Google being a prime example of it. Option number one, you search. Let's say in this example, you're searching for a particular product, and you receive a list of 50 different pages and you have to go click each page to see which is the right page, read what is in each of the pages, and then come to a conclusion of what product you may or may not want to buy. Then you have to click to then buy and then operate and function. Option number two, you can have an AI model, as an example, Perplexity, which is one of the companies that we invested in, search exactly as Google did, take all those results, read all those pages, and then give you a condensed summary of all the products and then rank them based on previous queries that you've given, given your context of what you've asked in the past. Which one would you choose? So that's a very simple, my assumption is I can't see the audience, that you would choose the second option as a more effective, efficient approach. That is very feasible today. Now with AI, really you're going to get the next layer of this. It's not going to just say here's a summary, but there will now be an agent that could effectuate a purchase on your behalf because do you really need a human to then go into Amazon or whatever is your e-commerce site of choice to then execute that purchase? Same thing in travel, same thing if you're trying to sign up for a new credit card account. You could argue this effectuation from search that is inefficient, you have to go read everything, you have to click the links, to it being summarized and then contextualized to your past queries and more importantly through time contextualized to you as an individual, even bigger context layer. That's a simple example of an app layer that can be built today and will happen right now. But if you now go to companies and enterprises, there is an incredible amount of productivity and efficiency that can be brought to bear leveraging AI applications. I don't know Raj if you want to flag one example in the drug discovery space. This is a category that we all care about very deeply because this is about having new medicines to treat illnesses to help improve both the quality and longevity of life in the world, something that fundamentally matters. But this is an industry where historically the amount of money, the amount of time that it has taken to develop new drugs and the innovative nature of said drugs have decreased through time and have become more expensive and taken more time to execute. But today you have a huge array of companies including multiple that we backed that are leveraging AI models to accelerate drug development. But Raj, I'll let you speak. It does, but once you combine AGI, quantum, and humanoid robots, which we are big investors in, and I think every day about the world they're going to grow around, and to me it's critically important. We can talk about regulation. And we can talk about what entrepreneurs building in the space should think internally to also optimize societal outcomes beyond and above also business outcomes. But for me it's critically important that as humans we think of, we emphasize our mental health, our sense of well-being, our sense of internal security, that we preserve humanity, right? Our empathy, our creativity, our values. And we really need to ensure that AI enhances rather than diminishes what it means to be human. And really what I mean by this is we really have to be strong enough to see AI, which will not just live in software, eventually will live in physical undulations that could look like robots today and will evolve through time, for them being collaborators versus competitors. That's a very critical thing especially as Raj discussed we start getting to AGI and materially more advanced modelings here. It's very, very, very important. I do think this technology will bring upon structural change upon society. There's a role for government to play but this technology is evolving so quickly that it is hard for governments, quote unquote, collectively globally to coalesce and drive regulations that quickly and that agile. So it's important for the entrepreneurs themselves and for every user of AI, for every company that buys AI, for everyone in the world to be thoughtful and make sure that AI is both balancing your personal or business outcomes with broader societal needs as well. I think the only way this works is if all of us take part in both containing the negative elements of it and allowing it to flourish, and more importantly that we love our kids, love ourselves, understand both the benefits of being a human and never attempt to compare ourselves to software and robots. I know there was an era where chess players would play against the computer. If we go to that route, we're not going in the right direction where one is trying to beat the other, collaborate and win together. Practical example that is not that far out from a technology standpoint, but maybe it's further out from an adoption standpoint. If you look at self-driving cars, I believe AI will help accelerate the development of said technology and actually if you look at one of the leading causes of death in many developed and developing nations in the world, it would be car accidents because humans make mistakes. We have two eyes. We don't have 20 eyes like some other AI cars, autonomous cars have here. And you have a massive workforce that works in the taxi industry, in the driver industry. And interestingly, some of the very companies that have enabled and created a more agile job opportunity in that space like Uber as an example, which our chairman helped seed as Raj discussed before, they themselves are also working on some of this technology. So if you think about that just as an example of one industry, what happens to the drivers? On one hand you have an improvement in lowering causes of death, causes of people being incarcerated for driving and drinking and things like that. On the other hand you have a huge loss of jobs and how do you train those people to do something else and what do they do? These are very challenging questions. When I speak to people in the industry they talk about the cars still need to be maintained in the first iteration so they could become people that help maintain it. They could own a car and then send the car into a network and own a part of the profits. There's a lot of different ways that one could evolve, but arguably if the same example starts happening across a large array of industries, there's a very important question to be answered. I've always thought that one of the critical important things potentially would be for entrepreneurs and companies to balance profit optimization with societal benefits. But that's one example of how this could be looked at.