Mark. Welcome to our first C-graph. How do you see the generative AI AD meta progressing today? And how are you applying generative AI to enhance your operations or introduce new features that you're offering? Well, I think we'll soon see a lot of the content that you see on Instagram today move into a zone where it's recommended to you, whether you follow them or not, from different things around the world that match your interests. I think in the future, a lot of content will also be created with these tools. Some of it will be creators using the tools to create new content. I think some of it will eventually be content that's created on the fly for you or, or, that's composited together from different things. So I dream that one day Facebook or Instagram will be a single AI model that's integrating different content types and systems to achieve different objectives in different time frames. Some of it will just show you interesting content that you want to see today. But some of it will help you build your network in the long term. People that you know and accounts that you want to follow, these multimodal models are really good at recognizing patterns and weak signals and things like that.And one of the things that people have always known is that AI is deeply ingrained in your company. You've been building out the GP U infrastructure for a long time that runs these large recommendation systems. Now, the actual deployment into GP US has been a little bit delayed, almost like you're trying to be nice. I get it. So tell me guys about Creator AI and AI Studio that enables that. Yes. So, actually, this is something that we, you know, we talked about a little bit, but we're rolling it out more broadly today. Uh, a lot of our vision is that we don't think there's going to be just one AI model. I mean, this is like some of the other companies in the industry are building like one central agent. Yeah, we're going to have an AI assistant that you can all use. But a lot of our vision is that we want to enable all of the people that use our product to essentially create an agent for themselves. So whether it's the millions of creators on our platform or the hundreds of millions of small businesses, we ultimately want to be able to take all of that content and quickly spin up a business agent to interact with customers and do sales and customer support and things like that.What we're starting to roll out now is what we call the AI Studio, which is basically a set of tools that will eventually allow every creator to build an AI version of themselves that will be like an agent or an assistant that the community can interact with. There's an underlying problem here, which is there aren't enough hours in the day. If you're a creator, you want to engage with your community more. But you're time-constrained, and similarly, the community wants to engage with you. But that's hard. I mean, you only have a limited amount of time to do that. So the next best thing is to basically allow people to create these artifacts, which are like agents, but you train your material to express yourself in the way that you want it to. I think that's a very creative endeavor, almost like a piece of art or content that you put out into the world. And this is obviously not something that's going to engage with the creators themselves. But I think that's going to be another interesting way to have an agent do that, just like a creator would put content out into a social system. One of the interesting use cases that we're seeing is people using these agents for support. Well, this was a bit surprising to me, but one of the best use cases for meta AI is to use it to role-play difficult social situations that people are going to be in. So whether it's a professional situation, whether you want to ask your boss how to get a promotion or a raise, or you're fighting with a friend, or you're having a difficult situation with your girlfriend, how would this conversation go, and you can basically role-play in a completely judgment-free zone to see how the conversation goes and get feedback on it. But a lot of people don't want to interact with the same kind of agent that everyone else is using, like met AI or Chai PT, they want to build their own. That's why LMA is really important. We've built this concept of an AI factory, or an AI foundry, to help a lot of people who have a desire to build AI. Owning your AI is really important, because when you put it into your flywheel, your data flywheel, it encodes the organizational knowledge of your company and puts it into your AI. So you can't have your AI flywheel, your data flywheel, your experience flywheel somewhere else.So open source enables that, but nobody really knows how to turn this whole thing into AI. So we created something called the AI Foundry. We give them the tools, we give them the expertise, the technology. We can help them turn this whole thing into an AI service. And then once that's done, they take it and they own it. The output of that is what we call A Nim, this NI, Neuro Micro NVIDIA inference, microservice. They can download it and run it anywhere. I like to include on-premise. And we have a whole ecosystem of partners, from O Ems, to G SI S, Accenture, who can run Nim. We've been training and working together to create Llama-based Nims and pipelines. And now we're helping companies all over the world make this happen. It's really exciting. This all started with open sourcing Llama. Rayan Met Alas, your vision of bringing AI into the virtual world is really interesting. Tell us about that. Yeah, well, there's a lot to unpack there. Uh, so the segment anything model that you're talking about, we're actually just launching it. The next version will be launching here at Siggraph Segment Anything 2. And it works now, it's faster, it works. Uh, so let's get started. Uh, now it works on video. I think these are cows from my farm in Kawaii. Uh, by the way, these are called Delicious Marks. Go ahead. Next time, Mark comes over to my house and we make Philly cheesesteaks together. Next time, bring something that you made. I was more of a sous chef, but this can create some fun effects and it will open up a lot of more serious applications across the industry. I mean, scientists are using this tool to study coral reefs and natural habitats and landscape evolution and things like that. But the thing is, this can be done in zero shots on video and you can interact with it and tell it what you want to track. So, this, this is pretty cool research. I think that ultimately we will have a series of different potential eyewear products with different levels of technology at different price points. Based on what we're seeing right now with Ray-Ban Meadows, I think that display list AI glasses in the $300 price range will be a very big product that will eventually be owned by tens of millions, hundreds of millions of people. So you're going to have a super interactive AI that speaks to you visually, has visual language, understands real-time translation that I just showed you, you can speak to it in one language and hear it in another. The display will be great of course, but it will add a little weight to the glasses and it will be more expensive. So I think there will be a lot of people who want a full holographic display, but there will also be a lot of people who will eventually want something like very thin glasses. So when Zuckerberg called it the H 100, you all know when the H 100 data center will be. I think we're approaching 600,000 units, and they're saying, “We're a good customer.” This is the beginning of the Jensen Q&A on ZIG Graph. Okay, everyone, this is Mark Zuckerberg.
