David Luan

“Now the critical path for model improvement is, is is shifting over to this to this sort of broader sort of simulation slash synthetic data slash like RL loop sort of path. I think it's just a natural consequence of the fact that it's so expensive to just keep…”

“A model trained that way is only as good as the smartest data in the training set. Like, it cannot discover new knowledge, because its job, the way the models are trained, is to do what a human would do in that situation.”

“I kind of think that chat bots like chat GPT and stuff and agents are kind of becoming different species of technology a little bit, right? Like I think they'll be useful in very different ways and what they need to be used for super different, right?”

“Having hallucinations in chatbots and in, like, image generators is, like, a really good thing, right? Because it gives you like, a starter tool for, like, getting to, like, solve the blank page problem, right? Like, and, like, gives you, like, little bits of …”

“And I know exactly the behavior of the system that I've built will be. But the cool thing about AI is that every day you come to work and you make some tweaks to the model. And what you get on the other end is actually somewhat unpredictable. Like you kind of …”

“When we were training GPT-II, we trained GPT-II in various different sizes. And at the smallest size, the model was just, like, unable to do three-digit arithmetic. But as the models got bigger and bigger and bigger, we didn't change anything else. We just had…”

“But I think pure model scaling does not deliver solutions to reasoning.”

“I think the general capability of reasoning will need to be solved at the model provider level. And that's because What you're actually doing is you're not just using the model to reason. You are trying to improve the model's ability to reason, which means the…”

“I do think there will not be that many LLM players. I think that there will probably be, my guess is somewhere between five to seven long-term steady state LLM providers at maximum scale, just because of the costs involved.”

“I just think in five to 10 years, people are gonna use their computers by giving them high-level goals, right?”

“Like, I think co-pilots are a great incumbent strategy because it lets them morph their existing software business model to something that kind of looks the same while getting in on the AI thing.”

“Like, that's not, I don't think this is how this is going to play out. I think the way this is going to play out is, Is that what we're going to have as humans fundamentally be the drivers of these agentic systems that like that, like basically give everybody …”

“And so that's why the HCI problem, like people just aren't spending enough time thinking about like chat is obviously not it.”

“You know, when you go work with a coworker, right, you don't just you don't just talk back and forth. You, like, actually, like, share the same canvas. You'll go use a whiteboard. You'll go, like, maybe look at the same thing on the computer together, and then…”

“I think second or second or third day of my time at OpenAI Greg and Ilya pulled me in a room and were like, hey you know, the you should take over our directs and we'll go mostly do IC work.”

“The number one driver of AI progress over the next couple of years is going to be the deep co-design and co-evolution of like product and users for feedback and actual technology.”

“I think agents are just absolutely the correct long-term direction, right? You just go to find what AGI is, right? You're like, hey, like, Well, first off, actually, I don't love AGI definitions that involve human replacement because I don't think that's actua…”

“Like de novo RL is like a pretty terrible way to get there quickly. Why are we rediscovering all the knowledge about the world? Like years ago, I had a debate with a Berkeley professor as to like what will it actually take to build HCI? And his view is basical…”

“Multimodal models are becoming more of a thing, we're behavioral cloning the visual world, but really what we're just going to have is this like universal byte model, right? Where like tokens of data that have high signal come in, and then all of those pattern…”

“I think pure play foundation model companies are just gonna be pinched by how good The next couple of llamas are going to be, and the next, like, what next good open source thing, and then seeing the really big players put ridiculous amounts of compute behind …”

“One, I'm so excited for someone to train a foundation model of robots. Like, it's just, I think it's just gonna work. Like, I will die on this hill. I mean, like, again, this whole time, like, we've been on this podcast, just continually saying, you know, like…”

“Having an advertiser be the financial party responsible for paying for these services means that ultimately these AI services are beholden to that advertiser and not to the end user.”

“I now think that the correct organizational structure for building an AI organization is actually to eliminate the concept of having a separate safety team. Because by creating a safety team, you're now defining this is the unsafe team, this is the safe team. …”

“And I talked to very smart people who haven't seen these models be trained and haven't seen the objectives these models are trained on. That just automatically assume that because they've seen this sliver of intelligence from these models, that they therefore …”

“I think it's like 10 x larger than the disinformation problem, if not a hundred x larger. In a world where these models are much more easily used as a mechanism for attack than a mechanism for defense, there's gonna be this period of time over the next couple …”

“In the process of solving a particular task, if a human gives the language model one piece of feedback, like you're writing code and you forgot to import the Python OS library, then the solve rate for these problems would jump up like 20, 30%.”

“While I was there, sort of in the last year in twenty-twenty-one, I tried pretty hard to get things launched at Google. We were all kind of banging on the table, being like, come on, this is the future, and you know obviously David Luan from Adept was also in …”

“And I think the reviewers now are pretty tired of that and they're moving on, but.”

“But in most cases, customers aren't willing to pay for a product that only gets them 80% of the way. You have to kind of like specialize and focus on the problem to make sure that you get to being 100%.”