Kyle Corbitt: prediction

Transcript context

…Construct your way into it that way, yeah. But even there, there's always going to be a gap between the simulation and reality, and you're going to have to ground it. Yeah, now you asked how quickly we'll see that happening, where we have potentially automated labs. And that's a great question, which I don't have I'm not sure about. I think on our current trajectory, let me bound my answer. So I think if model progress stopped today, we would still, where models didn't get any smarter, we just had similar capability levels, but we can keep RLing them. I think the rollout to the physical world would be very slow, probably just because there's a lot of constraints there and the data efficiency is going to be low, so the ROI is going to be relatively low and likely Frontier Labs is going to be very concentrated on automating everything digital first and then eventually potentially, there's these long tail, these physical things which are annoying to work with. And so I could imagine in that world we're maybe 15 plus years away before we see that being something that's a substantial part of the physical economy. But if we're on this recursive self-improvement thing and we're moving super fast, it's pretty soon, and arguably we're already there, it's the sort of physical stuff becomes the bottleneck and it becomes the most important thing to fix next. And we're now in a world where our The GP growth rate is going to be going fast. The labs are going to have effectively not unlimited resources, but extremely large amounts of resources. And it's like, hey, if we've got to figure out some new material science property so we can design the next generation of chips, yeah, sure, we can put $100 billion into building the automated lab that gets us the data we need to do that. And so, yeah, on that trajectory, which I think is more likely the trajectory we're on, then maybe we're two or three years away from this showing up in a major way would be my guess. One other possible Galaxy brain take I've had over time is it seems like some of these things favor Elon Corp in that they collectively seem to have a differentiated flow of hard engineering problems that they are solving on a continual basis in like relatively clean environments with their obsession with like removing best part is no part and so on and so forth. Do you think that this future you're describing plays especially to their strengths?…

Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.