The Best of 2025 (So Far) with Sarah Guo and Elad Gil

Dr. Fei-Fei Li: I think from a neural and cognitive science point of view that spatial intelligence is a really hard problem that evolution has to solve for animals. And what's really interesting is I think animals have solved it to an extent, but not fully solved it. It's one of the hardest problem because what is the problem animal has to solve? Animals have to evolve the capability of collecting lights in something, which we call eyes mostly. And then with that collection of eyes, it has to reconstruct a 3D world in their mind somehow so that they can navigate. and they can do things, and of course, they can interact. For humans, we're the most capable animal in terms of manipulation. We can do a lot of things. And all this is spatial intelligence. To me, that's just rooted in our intelligence. What is interesting is it's not a fully solved problem, even in animals. We, for example, for humans, right? If I ask you to close your eyes right now and draw out or build a 3D model of the environment around you, it's not that easy. We don't have that much capability to generate extremely complicated 3D model till we get trained. You know, there are some of us, whether they're architects or designers or just people with a lot of training and a lot of talent, and that's a hard thing to do. And imagine you do it at your fingertip much more easily and allow much more fluid interactivity and editability. That would just be a whole different world for people, no pun intended.

Brendan Foody: I think displacement in a lot of roles is going to happen very quickly, and it's going to be very painful. and a large political problem. Like I think we're going to have a big populist movement around this and all the displacement that's going to happen. But one of the most important problems in the economy is figuring out how to respond to that, right? Like how do we figure out what everyone who's working in customer support or recruiting should be doing in a few years? How do we reallocate wealth once we have, once we approach super intelligence, especially if the value and gains of that are more of a power law distribution. And so I spend a lot of time thinking about like how that's going to play out. And I think it's really at the heart of it. What do you think happens eventually? X percent of people get displaced from like color work. What do you think they do? I think there's going to be a lot more in the physical world. What does the physical world mean? Well, it could be everything ranging from people that are creating robotics data to people that are waiters at restaurants or are just like therapists because people want like human interaction. I think that automation in the physical world is going to happen a lot slower than what's happening in the digital world, just because of so many of the self-reinforcing gains and a lot of self-improvement that can happen in the virtual world, but not physical one.

Noubar Afeyan: The motivation for Flagship stems from what I was doing before, which was that I started a company in 1987 when 24-year-old immigrants didn't start companies in this country, but instead it was kind of like former Merck senior executives or IBM senior executives were the only ones who were entrusted with the massive amounts of venture capital, namely two, $3 million per round used to go into venture capital. So this was very early days. And I had the kind of chance, opportunity to start a company right out of my graduate school and ended up raising quite a bit of venture money and eventually kind of went down a path of entrepreneurship. Along the way, one of the things that interested me was why it is that kind of the entrepreneurial process was supposed to be random, improvisational, kind of idiosyncratic, almost emotional, gamey, All of those things I kind of thought was a bit of a put-off when it comes to actually doing things in a serious, professional way. And I kind of used to go around in the very early 90s saying, why isn't entrepreneurship a profession. And if it was going to be a profession, how could it be a profession? Because it's like supposed to fail most of the time and once in a while you win and then you celebrate the win. And what I mean is like it's random. But not only random, but there's like winners and losers and keeping score. I don't know, it's maybe the wrong word, but I just mean like people even call gamification in the software space. There is a version of this, like I don't mind being playful because if you're overly serious, sometimes you miss things, but it can't just all be played. We take hard-earned money, we deploy it to do things that are damn near impossible. Once in a while, we reduce them to practice so they become not only possible but valuable. And yet, People treat it like, oh, well, it didn't work. There's 20 different things we tried. One of them worked. That, I don't know, as an engineer by background, as a scientist, I just thought that what we do, especially, listen, in healthcare, especially in climate, especially in agriculture, food security, you can't think of this as shots on goal and this. Now, you've got to say, hey, we can get better at this.

Brandon McKenzie and Eric Mitchell: I can give maybe very concrete cases for like the visual reasoning side of things. There's a lot of cases where, and back to also the model being able to estimate its own uncertainty, you'll give it some kind of question about an image and the model will very transparently tell you, I shouldn't have thought like, I don't know, I can't really see the thing you're talking about very well. Or like it almost knows like that its vision is not very good. But what's kind of magical is when you give it access to a tool, it's like, okay, well, I got to figure something out. Let's see if I can manipulate the image or crop around here or something like this. And what that means is that it's much more productive use of tokens as it's doing that. And so your test time scaling slope goes from something like this to something much deeper. And we've seen exactly that. The test time scaling slopes for without tool use and with tool use for visual reasoning specifically are very noticeably different. Like for like writing code for something like there are a lot of things that an LLM could try to figure out on its own, but would require a lot of attempts and self verification that you could write a very simple program to do in like a verifiable and, you know, much faster way. So I do some research on this company and use this type of valuation model to tell me what the valuation should be. You could have the model try to crank through that and fit those coefficients or whatever in its context, or you could literally just have it write the code to just do it the right way and just know what the actual answer is. And so, yeah, I think part of this is you can just allocate compute a lot more efficiently because you can defer stuff that the model doesn't have comparative advantage to doing to a tool that is really well suited to doing anything.