 Okay, welcome back to theCUBE's coverage here on location at Las Vegas for AWS re-invent 2023. I'm John Furrier, your host, Dave Vellantez here. He's with the analysts, he's getting more content there. We have two great guests coming on theCUBE here to break down the generative AI wave from a technology platform perspective, infrastructure, and from an entrepreneurial perspective, as well as a venture council. We've got two great guests. We've got the entrepreneur and a VC in the same segment, two CUBE alumni's, John Chouro's partner at Madrona and Luis says CEO of OctoML, both CUBE alumni. It's great to see you guys. Thanks for coming on. Great to see you. Great to see you, John. So OctoUXM was on, John. So I should have said that you know a lot about AWS being a leader and building products there. Now at Madrona, you're funding builders. As we got a founder here, been on many times. So we're in a dynamic where you're already in market with some funding. It's a pretty good funding. It's raised, what, 800 million plus? Yeah, a little bit more. So you've got some good funding. So you're in market, but now a new wave is coming in, of the entrepreneurs. And then, you know, this re-invent really speaks volumes to this next wave of GenAI. So you've been preparing for this. Got the picks and shovels. You've been funding it. What's the current state right now? Because you have people in the market. Is there a line where, if you're not in it a certain way or you're coming in, is there a table stakes to build successfully? What's the current appetite for that entrepreneurial readiness? If you had to peg that, Luis. What would you say? Wow, there's just so much. First of all, it's changing so fast, right? So the new capabilities coming from models happening almost on a daily basis, you know, the platforms evolving fast, right? So I think there's, you know, pretty clear trends now that folks want to build prototype quickly and see what works and then, you know, scale that out and put in production. We're seeing clear trends in, you know, no single model rules them all. You know, people bringing with a collection of models, say model cocktails that together form something special, right, so distance would be a popular term here. Model cocktails. And, you know, like seeing open source models, like showing incredible capabilities and, you know, Amazon, you know, announcing Bad Rock, you know, supporting Lama 2 as one of their offerings. It's a great testament that, hey, open source models have capabilities that people want, right? And then the other thing I thought was interesting here that, you know, clear infrastructure complexity, great infrastructure, like, you know, talking about the work horses there at the bottom, like with the GPUs and, you know, accelerators underneath, we saw several announcements this event, right? So, Jensen on stage, you know, talking about the new instance types, but also DGX Cloud being offered at AWS, you know, Adam talking about Trainium 2, and which is all fantastic because bringing more resource for folks to use and build awesome applications, but also more infrastructure complexity to manage, right? So, I'm a- More horsepower, more complexity with the AI workloads. Exactly. And the- For a believer that we need to manage, they make it very easy for folks to build with. I think that's one of the- I want to get into the ease of use, John. We were talking about last night about ease of use and this waiver on, but you mentioned speed. You guys have just announced in one month two major things. You did image and text products just in one month. You shipped. Yeah, we did. And we've seen people already building with it, which is really, really rewarding and awesome, right? So we have, you know, the fastest implementations of stable diffusion and Lama 2 and Mistral offerings on the market. You know, I'm not saying this just to brag, it's that latency matters here because it's a better user experience and also making it faster makes it cheaper. But more importantly, you know, making this for applications in production. Not for, you know, of course, we want people to experiment on the platform, but our end user here are folks putting in production and showing that he works and stays up and you can run an enterprise application with it. It's been a couple months since you've been in theCUBE and I know lots of change. What's your current business model now? I mean, then you've got more tech, you've got more capabilities. What are you guys doing now? What is OptoML's main focus from a product customer perspective? Absolutely, yeah. So people can run, tune and scale models and build generative applications with them, right? So it's generative AI infrastructure for application builders. And the business model is simple. You pay for what you use, right? You run a model, image. If you're an image gen, you're going to pay per image. If you're in a language model, you're going to pay per token produced, right? So very, very clear, transparent pricing and then you can use- And the person that's buying or using the service are who? Platform engineers? Are they application developers or are they business? Application developers that are not machine learning engineers, right? So there's a folks that again, abstracting complexity, right? So these are developers that now they have done some experimentation with AI. They know what they want to build and they want to put models in production. And they want to build their models, their own models. You want to choose models. Customers want choice. Like this was a common thing here. So like when we were to choose the models that do not be dependent on any specific model and choose models that have the capabilities that you need and have the right performance. They're going to be building. They're going to be developing. This is that new layer where the feeding frenzy is going to come in. They want to get into bedrock and test the long tail of power law of models and see what happens. They'll experiment, maybe discover something and then deploy it. Deploy in scale and then putting production with production quality service. Like the means of time, performance, reliability. All you entrepreneurs out there, build some stuff and then we go to John for some cash. If you're an enterprise, show your boss, get funding. Okay, John, entrepreneurs get some experimentation. Now they want cash. They come to you, you're a VC. Not just cash, he's full of insights too, John. He's full of insights. He's an added value investor. So I didn't mean that as a compliment, you are. Okay, now I want to get funding. The pace is fast. How do you decide what's good when you look at a deal because I've seen some products out there that look really good on paper, but are just a hackathon. We can project that looks good, but is there enough differentiation there or are valuations too high? So there's a lot of action happening. What's the filter now? It's a speed game, it feels like. What's the funding look like for an entrepreneur out there? So I think there is now so much tooling, so much infrastructure, so many models available to founders now. And what could be a better environment to build in? And the question is, how can you make that good for you, not bad for you? It's by using it. It's by using OctoML to get faster inference and better access to silicon than anyone otherwise could. It's by using AWS to get infinite elastic compute and things like zero ETL to pipe your data around. And all of this is really hinting at what we've said all along are the two key things that a great company needs to have to build a great product on top of technology like this. It's really powerful data. It's really powerful talent who is well-leveraged and it's better understanding of your customers than anyone else. And when we have these infrastructure, what that can do is it can actually get you more efficiency out of the talent that you do have, which is precious, which is rare and special. The technology is now making it easier to take the data that you may have, that you do have, and actually apply that. And the customer insight, that's the thing you got to start with. I think that's a very key point because I think, if you say, okay, heavy lifting, I got OctoML as a service, transparent price, and you know what you get, so you can watch your budget. If you don't want to overdrive, if it goes viral, then you got to get to funding first and then build that back. But once you have that transparent part, you can focus on the product. The customer is the business logic. So you go to the customer and identify, that's an opportunity for an entrepreneur to actually compete and unsee the incumbent. If they can go to a customer, say, I've identified a workflow that I could get in there and leverage end to end with data, potentially is an entry. So again, this brings up the question of, can the small guy come into a market and take territory? I think the answer is absolutely yes. We see it again and again and again. I think you're about to say something, but we see it again and again and again that if you can understand your customers better than other people do, and you can make the trends around you of speed and undifferentiated heavy lifting and faster performance and better generalizability and reasoning capabilities of these models, if you can do all that and understand your customers better, you can absolutely beat Goliath. Yeah, absolutely. I just want to add, the only thing I have to add here is that there's so many pieces ready to be composed into something really compelling and magical here that definitely understanding what your customer needs, but also showing what's possible with this new technology for customers that would be a pleasantly surprise and do that very quickly, right? So it's just extremely exciting to see how much resources available to do that today. We've taught in the past, obviously you've been a professor at the University of Washington, great school, great football program this year, congratulations, but also computer science programs, phenomenal. It feels like we're at an intersection, a flash point of all the things that have been kind of hanging around the table that we want to have happen, smart homes, cars, like stuff that's been supposedly supposed to happen over the past 20 years, it reminds me of the web, you know, the dot-com bubble, it burst, but all that stuff happened. We get food online, the stuff that was supposed to happen didn't, but it happened over time. It was a slower boil. Now with the speed game, we might see a flash point with this market, with this event showing us the GNI that this might come together, the smart home, the edge, the smart car, the stuff that was kind of crawling along that might actually go faster now. It feels like we're at a point where there's enough compute, models as a service. You know, if I can offer a hot take, I would say that we haven't seen anything yet because, you know, go back a click or tune history and the App Store launches in 2008 or so, mid-2008, iOS 2.0. And we had a lot of apps, a lot of them were flashlight apps, a lot of them were simple games or direct ports from the web. And we did get apps that were native to mobile, that were special, that showed us what was possible, but not for two years. And I would assert that you could have built Uber or Instagram in 2008, but it's not just a matter of the technical building blocks. The technical building blocks have to be there and then lightning has to strike. And so what we now see is that the infrastructure is there and we're waiting for a lightning strike. It feels static in here, doesn't it? Like the lightning's gonna hit. It does feel static, right? It does feel static, right? It does feel static, right? A lot of the craps are done. What is the lightning? Is it timing? Is it just a moment in time? What does the lightning strike look like? Is it, you know, what do you see that? I didn't know what it's coming to. It's just showing it's possible to automate things that only humans could do before. Like think about, you know, the velocity we're seeing applications that can do, like automating, producing a summary of a long email thread, you don't have to read the whole thing. This is actually working, showing meaningful progress. We actually needed people to do that meaningful in the past, right? Or automating, generating graphical content out of text. I mean, this is something that was just in its infancy like 18 months ago, and now it's not only showing that it's possible. It's actually in production. A lot of really compelling use cases. This happening so fast that I'm with John that we haven't seen anything yet, but at the same time, what we've seen already is really, really, really cool. So that's why. Two years at the App Store, maybe two months in AI. But again, this brings up the shrinking of the acceleration. Or is it at the pace of human speed at this point? Yeah, I think we're at the pace of human speed. Take the examples I just gave before. Uber, the lightning strike was Travis Calcanic, can't get a cab in Paris. And Airbnb, those guys needed, you know, they were trying to, they needed, they had an air mattress. And so for all the pace of innovation, there is, we're waiting for human inspiration. And there's a lot of people working on it, but there's some things that actually you can't rush. Yeah, there's some table stakes. I think the bedrock thing, choice and open always wins. We talked about that a little bit last night. You guys had a founder dinner, Madrona, the great founder dinner. And it was interesting to hear the perspective and the demographics. I think we were the oldest ones in there. The young guns. I mean, if you're under, if you're an entrepreneur today and you're not excited about this environment, then maybe you shouldn't be in tech. Cause like, this is probably the best ripe opportunity. But no one knows what's coming. They know it's going to be big. That's why I was asking about what, what is good look like and how do you differentiate? You know, I mean, think the web or App Store, Flashlight app or a webpage. Do you get funding for that? Well, some people did. That was the dot com bubble. But I think now we're more pragmatic about that. But you know, people are asking me, where's the white space? Where's the opportunity? And so I just, I mean, I just think it's everywhere. Literally, literally every type of application and many business and your human processes that are not even applications is in scope for this kind of technology. You could probably contrive one that's not. And so I should say almost literally everything, but I can't contrive one. And so what that means is anything that, you know, founders ask, where should I focus? Where's the opportunity to focus? And I think that's the wrong question. I think the question is what customers, what problem do you know better than anybody else on this earth and have more passion for than anyone else on this earth? That is the thing. I totally agree with that. I think that's great advice. And also the point about the customer, the access to customers now are a lot easier path or to get data about a customer to say, okay, I can validate the problem that I know about. And I know what the answer looks like. Let's get going and then use AI to help you get in there and then get back and get standing up some solutions and get the iteration, the latency of the answers that you need. So there's latency questions two ways. It's like latency of speed, speed of packets and speed of creative creation. I know latency, I have confusion. So I think about latency as in how long do you wait and then velocities, how fast it's, you know, things are evolving, right? Yeah, that's a great point. All right, so what's your take of the show here? Let's take a step back. Give me the summary of your perspective of re-invent 2023. Well, I'm generating AI everywhere, not surprising. And, you know, seeing, you know, interesting announcements from AWS on AWS, Amazon, actually Amazon, not AWS Q, you know, was interesting. Also showing, you know, bad rock coming together. And again, like, evolution of infrastructure on the compute side is clearly clear that compute is top of mind for everyone here. Just look at the scale of compute that we're talking about. That was announced yesterday. It's just several hours of magnitude than we've seen in a while, right? So I want to bring up with people that I don't know wake up the annoying professor in me, but one thing that's, I think it's interesting for people to internalize here is that every output that Genevieve AI produces, like a word from a language model or an image from these text image models, it's like literally many hours of magnitude more than the kind of compute requirements for any application we've ever had in human history. Right, so I think it's a moment, it's a moment of like the number of calculations required to produce a meaningful output for application Genevieve AI is orders of magnitude more than a typical application like a spreadsheet sitting in front of you or a browser. So the compute involved for that is more than people can even imagine. It's a moment to be grateful that computers are really, really fast and they're getting energy efficient, you know, so. Well, that's a great point. I mean, the energy thing is the envelope now for the constraint. It constrains everything. Yeah, and John, what's your take? That's a good point about the magic of how fast that hardware and the speed of the processors and the system around it, it's like, it's not just chips, it's what's around it. Well, I think it really brings up why we're all here at re-invent because this new wave of AI requires science and math that would not be possible without the cloud, without what happened here in 2012, what happened in Seattle in 2006. That level of compute and coordinate and cross many nodes would not have been possible. And now, back to 2023, when you see what Peter DeSantis announced in the infrastructure keynote, there's the best way to believe, the best way to know for sure that Amazon believes in AI is the way Amazon is consuming AI. For example, to make the databases more scalable and more efficient. It's really inspiring and deep, deep investment that's happening. And it's meaningful. It has a direct application to what people need. Just get that lightning strike going back to the spark. You didn't get the cab in Paris, start Uber, or I want to solve this problem. I see that no one else solved. As Steve Jobs said, half the stuff is invented by people no smarter than you, right? That's the famous line from Steve Jobs. So people can start solving the problems. You don't have to be a machine learning engineer to go build apps with unprecedented compute available for you. Yeah, and I'll add one more thing so that you might not even need to be a regular, typical application developer because now there's a lot of really meaningful ways for you to actually start building experiences without writing code. I mean, natural language becoming, no, you'd be hearing this more and more, but there's something natural language programming has been a dream of computer scientists for a long time and it's finally becoming, it's finally being realized in a way that's very different than computer scientists have thought. And it just feels magical even to them, right? So people can build stuff, more people that can build stuff, more creativity that can be unlocked very quickly, more experiences, and that's where it's a wealth creation, it's a wealth creation. That's just not money, but assets, society, benefit. Love having the professor on masterclass here and I think you always love having you on. Special time, last minute we got. Give a plug for the company and when you're looking for an investment we'll start with the company. What are you going on? Who are you hiring? You look up to hire. What's going on? Put the plug in. No, we're definitely on a roll. So come talk to AI or product to run to scale your Genitive AI models. We're a Genitive AI infrastructure for application builders. We make it very easy for folks to go and build the model cocktails and solve the problems that they want and have the models they can choose. We want to hear from you anywhere up or down the AI stack if you know your customer is better than anyone else on the planet. And you're writing big fat checks or series C, A, what's the range of checks that you write? Madrona will write a check anywhere from the very first check through series C. Great, great stuff. Great firm. Congratulations, Dr. Mel, continue. We can't wait to get the update next. Always great to have you guys. Great to see you guys last night in the team. Thanks for coming on theCUBE. All right, we're bringing it down. A lot of opportunities to get funding if you're a startup and also if you're a builder, the tools are available, unprecedented resource available to get that spark of creativity to create value for your company or startup. Perfect time to be out there. Of course, theCUBE's got all the action, more coverage after this short break. Back to the studio in Palo Alto.