The Exchanges

Q And you went back to open AI upon graduation, right?

A I went back because I, I couldn't start any company after my PhD, um, either because the idea was too grand that nobody would want to come work with me or, uh, I had visa problems to start it all by myself. So I felt the best thing to do was like find a job and like continue to like keep exploring and learning more. But around the time in like maybe February or March in, like news started spreading that there were companies like Jasper and copy.ai that started making more money than even OpenAI at the time. And that was amazing. Okay. Like people are building products. This is real. And, uh, GitHub Copilot Uh, when, when, when they moved away from the waitlist to the paid version, they just had like hundreds of thousands of people paying from the first day. And so that all made it clear that, um, stuff that we were all thinking was just like, you know, new ideas and research was more about like strong execution, building teams and like shipping products. So I really wanted to be part of that too, and reached out to two prominent investors in Silicon Valley, uh, Elad Gill and Nat Friedman. And, uh, they both were willing to invest and people were like, you know, these two guys are, you know, backing you. You should just do it. Even if it fails, like you learn a lot. It's like, uh, getting MBA and getting paid to do it rather than paying Stanford or Harvard.

AI assessment note: “I went back because I, I couldn't start any company after my PhD”

Q So obviously we're going to talk, um, you know, all about you, but a quick word on, on OpenAI and, and, and your experience, uh, the second time you, you sort of came back after your internship, presumably there was no longer the 35, uh, uh, person nonprofit company. Well, what was, what was it like working at OpenAI, uh, in like 21 and 22?

A It was, it was cool. Like, you know, um, I think at that time, GPT three was already there. GPT 3.5 was being developed and, and, and, uh, GPT four was not even there. The biggest hit at the time was GitHub copilot and Dolly two. Both of them are really cool. And it was not clear Like, whether, like, you know, for a successful product, you needed the largest compute put on it. That was not clear at that time. Because, like, Jasper and, and, uh, you know, Copilot were all, like, tiny models making a lot of money. Similarly, like, Stable Diffusion and Midjourney were all, like, pretty competitive with Dolly too, and actually probably making more money than Dolly too. So, that part was unclear. So I, I wasn't very, like, um, Not just me, like most people were not sure, like, what is in it for GPT 3.5 or four. But, um, so obviously I was wrong, you know, like chat GPT and GPT four are like the reason why opening is so formidable today. So that nobody predicted and I didn't predict it either.

AI assessment note: “I think at that time, GPT three was already there. GPT 3.5 was being developed”

Q market with this? I mean, obviously, um, and we'll go back into some technical stuff, but like while we're on the, you know, topic of those various things, because in particular, like you, you have this consumer, uh, you know, product and therefore motion, but then the API is more of a developer kind of like motion, right? So how do you think about, um, succeeding on, on both fronts?

A Yeah, right now we've tied both together, so the only way to get access to our APIs is UB on our pro plan, and then, um, you get the, the co-pilot unlimited experience, that's what it's like, um, a developer access to our APIs with a certain restricted rate limit. Um, and those who want higher rate limits can come to us separately, and like on an ad hoc basis we, um, extend it for them. We have to come up with a usage based pricing, obviously, and we have to like build a whole API team for this if you, and that's sort of what we think we should be doing. We are first trying to understand like, you know, what is the market here? Like what is people's incentive to come and use these models in, in a world where GPT 3.5 turbo and GPT four already exist. And these are amazing models. Um, and I think we can add a different value through a RAG, uh, retrieval augmented generation APIs, which are still not out yet, but we are working on getting it out. And in that world, I think we are committed to helping people like, you know, deploy using our, our models and APIs, um, for their applications, if they're interested. Right. So you want to start small. Like, I think the best things is like start small. Um, Allow people who want like, you know, really tiny models we deploy onto their product. Um, we are a good fit for people who are trying to build consumer facing applications because tha…

AI assessment note: “right now we've tied both together, so the only way to get access to our APIs”

Q Is that because of the citations you think?

A Citations and the fact that it's like, ah, very accurate, very fast and reliable UX. So we didn't actually take on Google. Like we never said, Hey guys, I'm the Google competitor. Uh, you know, I'm gonna, I'm gonna kill the monopoly or something. Instead we said, Hey, like, this is a pretty useful tool. Like, you know, you guys are all spending a lot of time browsing and doing your research. Um, why don't you come and use this tool? It'll help you. And people started using it more as a research companion. And even though that's like a tiny fraction of the search market, you, you, you have cornered it first. Now we cornered it, you, you got like few millions of users. You start being useful to them in other ways. And then they're going to spread it word of mouth to their friends and friends of friends and like you slowly grow. That's the idea.

AI assessment note: “Citations and the fact that it's like, ah, very accurate, very fast”

Q just executing at, um, incredible speed, so it's been, uh, wonderful to, um, to see. Um, you know, maybe, maybe to close before I open to, to questions, sort of, like, random, um, kind of notes based on some of your tweets or things I've, like, I've seen you, uh, talk about. Um, thoughts on open source, the importance of open source in the ecosystem, and how to support it?

A Yeah, I, I think open source models are, you know, like, definitely gonna, like, work out in the long run. Today, it looks like, you know, they're, like, lagging behind GPT-IV. Uh, but that's also because the amount of compute thrown at them is, like, way lower. Um, it's also important for the world. Like, otherwise, like, you know, if the person having the best model controls the prices, Uh, we'll end up in a situation like how, um, you know, like say, NVIDIA, for example, controls the GPU computer and like, nobody can tell them what, what, you know, you got a price H-one or something. We're all like, so that's, that shouldn't happen for, if you want to access these large language models to build your applications, like you shouldn't have only like one single choice or like two choices, especially when you don't control the price. So that's what open source models enable. They enable like democratization and like they enable the fact that there's going to be a free market here. Um, and, uh, we are supporting it in a way where we are not the ones building these open source models because that requires a lot more capital. Uh, but we want to, like, make sure anybody can access them easily. So we are democratizing the access to them through our APIs, through our labs, and, like, you can come and play with it. You don't need to go to ChatGPT at some point, like, you know, these mod…

AI assessment note: “we are democratizing the access to them through our APIs, through our labs”

Q And still on the obvious, you know, question of the competitive advantage and, you know, the, the whole Google thing, like I read somewhere so far, you've been using other people's indexes and crawlers. Um, what, what was that?

A There was a being initially, or yeah, we start, we started with being, I think like, um, we are not, uh, fully independent of, you know, other people's indexes and you're building your own, right? We are, we are building our own index. In fact, that, that is like, The most ambitious project that we are working on. Um, everyone's going behind training GPT-IV. Uh, we are going to, we believe that building your own index is even harder than, uh, trying to compete with GPT-IV because it's not a problem that's just solved with money. If there is a problem, I mean, I'm not saying GPT-IV, uh, yeah, like capable, like, Capability matching is just solid money. You still need amazing talent and, you know, research teams to get it done, but there are at least two or three teams in the world that can potentially get there. Building an index is like really hard because it needs you to have a good team, some amount of capital, not necessarily a lot, and an actual product that people use. Without people using a product, there's no way to build an index. Um, that's why most of the companies that were started earlier that tried to take on Google never really, like, you know, succeeded because they never got the usage.

AI assessment note: “we started with being... We are, we are building our own index.”