What Comes After ChatGPT? The Mother of ImageNet Predicts The Future

Fei-Fei Li is a Stanford professor, co-director of Stanford Institute for Human-Centered Artificial Intelligence, and co-founder of World Labs. She created ImageNet, the dataset that sparked the deep learning revolution. Justin Johnson is her former PhD student, ex-professor at Michigan, ex-Meta researcher, and now co-founder of World Labs. Together, they just launched Marble—the first model that generates explorable 3D worlds from text or images. In this episode Fei-Fei and Justin explore why spatial intelligence is fundamentally different from language, what's missing from current world models (hint: physics), and the architectural insight that transformers are actually set models, not sequence models.

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Speaker 00

  I think the whole history of deep learning is, in some sense, the history of scaling up compute.

• 0:03Fei-Fei Li?

  When I graduated from grad school, I really thought the rest of my entire career would be towards solving that single problem, which is

• 0:14Fei-Fei Li?

  A lot of AI as a field, as a discipline, is inspired by human intelligence. We thought we were the first people doing it. It turned out that was also simultaneously doing it.

• 0:26Speaker 00

  So Marble, like basically, one way of looking at it, it's the system, it's a generative model of 3D worlds. Right? So you can input things like text or image or multiple images, and it will generate for you a 3D world that kind of matches those inputs. So while Marble is simultaneously a world model that is building towards this vision of spatial intelligence, it was also very intentionally designed to be a thing that people could find useful today. And we're seeing, starting to see emerging use cases in gaming, in VFX, in film, where I think there's a lot of really interesting stuff that Marble can do today as a product. And then also set a foundation for the, for the, for the grand world models that we want to build going into the future.

• 1:06Speaker 02

  Fei-Fei Li is a Stanford professor, the co-director of Stanford Institute for Human-Centered Artificial Intelligence, and co-founder of World Labs. She created ImageNet, the dataset that sparked the deep learning revolution. Justin Johnson is her former PhD student, ex-professor at Michigan, ex-Meta researcher, and now co-founder of World Labs. Together, they just launched Marble, the first model that generates explorable 3D worlds from text or images. In this episode, Fei-Fei and Justin explore why spatial intelligence is fundamentally different from language, what's missing from current world models, hint, physics, and the architectural insight that transformers are actually set models, not sequence models.

• 1:50Speaker 03

  Hey, everyone. Welcome to the Late in Space Podcast. This is Alessio, founder of Kernel Labs, and I'm joined by Sweets, editor of Late in Space. And we are so excited to be in the studio with Fei-Fei and Justin of World Labs. Welcome.

• 2:02Fei-Fei Li?

  We're excited, too.

• 2:03Speaker 03

  I nearly said Marble.

• 2:05Speaker 00

  Yeah, thanks for having us.

• 2:06Speaker 03

  I think there's a lot of interest in world models, and you've done, you've done a little bit of publicity around spatial intelligence and all that. I guess maybe one of the parts of the story that is a rare opportunity for you to tell is how you two came together to start building World Labs.

• 2:21Fei-Fei Li?

  That's very easy because Justin was my former student.

• 2:24Speaker 00

  Yeah.

• 2:24Fei-Fei Li?

  So Justin came to my, you know, in my, the other hat I wear is a professor of computer science at Stanford. Justin joined my lab when, which year?

• 2:35Speaker 00

  Twenty twelve. Actually, the semester that I, the quarter that I joined your lab was the same quarter that AlexNet came out.

• 2:41Fei-Fei Li?

  Yeah, yeah. So Justin is my first.

• 2:44Speaker 04

  Were you involved in the whole announcement drama?

• 2:46Speaker 00

  No, no, not at all. But I was sort of watching all the ImageNet excitement around AlexNet at that quarter.

• 2:52Fei-Fei Li?

  So he was my, one of my very best students. And then he went on to have a very successful early career as a professor in Michigan, University of Michigan and Arbor and Meta. And then when we, I think around, you know, more than two years ago, for sure, I think both independently, both of us have been looking at the development of the large models and thinking about what's beyond language models. And this idea of building world models, spatial intelligence, really was natural for us.

• 3:29Fei-Fei Li?

  So we started talking and decided that we should just put all the eggs in one basket and focus on solving this problem and started World Apps together.

• 3:40Speaker 00

  Yeah, pretty much. I mean, like, after that, seeing that kind of ImageNet era during my PhD, I had the sense that the next sort of decade of computer vision was going to be about getting, getting AI out of the, out of the data center and out into the world. So a lot of my interests post PhD kind of shifted into 3D vision, a little bit more into computer graphics, more into generative modeling. And I was, I thought I was kind of drifting away from my advisor post PhD, but then when we reunited a couple of years later, it turned out she was thinking of very similar things.

• 4:09Speaker 04

  So if you think about AlexNet, the core pieces of it were obviously ImageNet. It was the move to GPUs and neural networks. How do you think about? The AlexNet equivalent model for world models. In a way, it's an idea that has been out there, right? There's been, you know, Yann LeCun is maybe like the most, the biggest proponent, most prominent of it. What have you seen in the last two years that you were like, hey, now's the time to do this? And what are maybe the things fundamentally that you want to build as far as data? And kind of like maybe different types of algorithms or approaches to compute to make world models really come to life?

• 4:43Speaker 00

  Yeah, I think one is just there is a lot more data and compute generally available. I think the whole history of deep learning is in some sense, the history of scaling up compute. And if you think about, you know, AlexNet required this jump from CPUs to GPUs. But even from AlexNet to today, we're getting about a thousand times more performance per card than we had in AlexNet days. And now it's common to train, to train models, not just on one GPU, but on hundreds or thousands or tens of thousands or even more. So the amount of compute that we can marshal today on, on a single model is, is, you know, about a million fold more than we could have even at the start of my PhD. So I think language was one of the really, really interesting things that started to work quite well the last couple of years. But as we think about moving towards visual data and spatial data and world data, you just need to process a lot more.

• 5:26Speaker 00

  And I think that's going to be a good way to soak up this, this new compute that's coming online more and more.

• 5:31Speaker 04

  Does the model of having a public challenge still work? Or should it be centralized with inside of a lab?

• 5:38Fei-Fei Li?

  I think open science still is important, you know? AI obviously compared to the image that AlexNet time has really evolved, right? That was such a niche computer science discipline. Now it's just like civilizational technology. But I'll give you an example, right? Recently, my Stanford lab just announced a open data set and benchmark called Behavior, which is for benchmarking robotic learning. In simulated environments. And that is a very clear effort in still keeping up this open science model of doing

• 6:22Fei-Fei Li?

  things, especially as in academia. But I think it's important to recognize the ecosystem is a mixture, right? I think a lot of the very focused work in industry, some of them are more seeing the daylight in the form of a product rather than an open challenge per se.

• 6:44Speaker 03

  Yeah, and that's just a matter of the like funding in the business model, like you have to see some ROI from it.

• 6:50Fei-Fei Li?

  I think it's just a matter of the diversity of the ecosystem, right? Even during the so-called Alexa ImageNet time, I mean, there were closed models, there were proprietary models, there were open models, you know, or you think about iOS versus Android, right? There are different business model. I wouldn't say it's just a matter of funding per se, it's just how the market is, they're different place.

• 7:18Speaker 03

  Yeah.

• 7:19Speaker 04

  But do you feel like you could redo ImageNet today with the commercial pressure that some of these labs has? I mean, to me, that's like the biggest question, right? It's like, what can you open versus what should you keep inside? Like, you know, if I put myself in your shoes, right? It's like, you raise a lot of money, you're building all this. If you had the best dataset for this, what incentives do you really have to publish it? And it feels like the people at the labs are getting more and more pulled, and the PhD programs are getting pulled earlier and earlier into these labs. So I'm curious if you think there's like an issue right now with like how much money is at stake and how much pressure it puts on like the more academia, open research space? Or if you feel like that's not really a concern?

• 8:01Fei-Fei Li?

  I do have concerns about, less about the pressure, it's more about the resourcing and the imbalanced resourcing of academia. This is a little bit of a different conversation from World Labs, you know? I have been the past few years advocating for a resourcing, a healthy ecosystem, you know, as the founding director, co-director of Stanford's Institute for Human-Centered AI of Stanford High. I've been, you know, working with policymakers about resourcing public sector and academic AI work, right? We work with the first Trump administration on this bill called National AI Research Resource,

• 8:46Fei-Fei Li?

  NER, bill, which is scoping out a national AI compute cloud as well as data repository. And I also think that open source, open data sets continue to be important part of the ecosystem. Like I said, right now, in my Stanford lab, we are doing the open data set, open benchmark on robotic learning called behavior, and many of my colleagues are still doing that. I think that's part of the ecosystem. I think what the industry is doing, some startups are doing, are running fast with models creating products is also a good thing.

• 9:28Fei-Fei Li?

  For example, when Justin was a PhD student with me, none of the computer vision programs worked that well, right? We could write beautiful papers. Justin has some beautiful.

• 9:40Speaker 00

  Actually, even before grad school, like I wanted to do computer vision and I reached out to a team at Google and like wanted to potentially go and try to do computer vision like out of undergrad. And they told me like, what are you talking about? Like, you can't do that. Like, go do a PhD first and come back. What was the motivation that got you so interested? I had done some computer vision research during my undergrad with actually Fei Fei's PhD advisor.

• 10:01Fei-Fei Li?

  There's a lineage.

• 10:02Speaker 00

  There's a lineage here. So I had done some computer vision even as an undergrad, and I thought it was really cool and I wanted to keep doing it. So then I was sort of faced with this sort of industry academia choice even coming out of undergrad that I think a lot of people in the research community are facing now.

• 10:17Speaker 00

  But to your question, I think like the role of academia, especially in AI, has shifted quite a lot in the last decade. And it's not a bad thing. It's a sense of, it's because the technology has grown and emerged, right? Like, five or 10 years ago, you really could train state-of-the-art models in the lab, even with just a couple GPUs. But, you know, because that technology was so successful and scaled up so much, then you can't train state-of-the-art. Models with a couple of GPUs anymore. And that's not a bad thing. It's a good thing. It means the technology actually worked. But that means the the expectations around what we should be doing as academics shifts a little bit. And it shouldn't be about trying to train the biggest model and scaling up the biggest thing. It should be about trying wacky ideas and new ideas and crazy ideas, most of which won't work. And I think there's a lot to be done there.

• 11:01Speaker 00

  If anything, I'm worried that too many people in academia are hyper focused on this notion of trying to pretend like we can train the biggest models or or treating it as almost a vocational training program to then graduate and go to a big lab and be able to play with all the GPUs. I think there's just so much crazy stuff you can do around like new algorithms, new architectures, like new systems that, you know, there's a lot you can do as as one person.

• 11:24Fei-Fei Li?

  And also just academia has a role to play in understanding the theoretical underpinning of these large models. We still know so little about this. Or extend to the interdisciplinary, you know, just because wacky ideas. There is a lot of basic science ideas. There is a lot of blue sky problems. So I agree, I don't think the problem is open versus close, productization versus open sourcing. I think the problem right now is that academia by itself is severely under-resourced. So that, you know, the the researchers and the students do not have enough resources to try these these ideas.

• 12:08Speaker 03

  Yeah, just for people to nerds snipe, what's a wacky idea that comes to mind when you talk about wacky ideas?

• 12:13Speaker 00

  Oh, like I had I had this idea that I kept pitching to my students at Michigan, which is that I really like hardware and I really like like new kinds of hardware coming online. And in some sense, the the emergence of the neural networks that you use today and transformers are really based around matrix multiplication, because matrix multiplication fits really well with GPUs. But if we think about how GPUs are going to scale, how how hardware is likely to scale in the future, I don't think the current system that we have, like the GPU, like hardware design, is going to scale infinitely. And that we start to see that even now, that like the unit of compute is not not the single device anymore. It's this whole cluster of devices. So if you imagine.

• 12:49Speaker 03

  A node.

• 12:49Speaker 00

  Yeah, it's a node or a whole. Cluster, but the way we talk about neural networks is still as if they are a monolithic thing that could be coded like in one GPU in PyTorch. But then in practice, they could distribute over thousands of devices. So are there like just as transformers are based around MatMall and MatMall is sort of the primitive that works really well on GPUs. As you imagine hardware scaling out, are there other primitives that make more sense for large scale distributed systems that we could build our neural networks on? And I think it's possible that there could be drastically different architectures that fit with the next generation or like the hardware that's going to come 10 or 20 years down the line. And we could start imagining that today.

• 13:24Speaker 03

  It's really hard to make those kinds of bets because there's also the concept that the hardware lottery where let's just say Nvidia has won and we should just scale that out in infinity and write software to patch up any gaps we have in the mix, right?

• 13:38Speaker 00

  I mean, yes, yes and no. Like if you look at, if you look at the numbers, like even going from Hopper to Blackwell, like the performance per watt is about the same.

• 13:45Speaker 03

  Yes.

• 13:45Speaker 00

  They mostly make the number of transistors go up and they make the chip size go up and they make the power usage go up. But even from Hopper to Blackwell, we're kind of already seeing like a scaling limit in terms of what is the, what is the performance per watt that we can get. So I think, I think there is room to do something new and I don't know exactly what it is and I don't think you can get it done like in a three month cycle as a startup. But I think that's the kind of idea that if you sit down and sit with for a couple of years, like maybe you could come up with some breakthroughs. And I think that's the kind of long range stuff that is a perfect match for academia.

• 14:14Speaker 03

  Coming back to the little bit of background in history, we had this sort of research note on the scene storytelling work that you did or newer image captioning that you did with Andre. And I just wanted to hear you guys tell that story about, you know, you were like sort of embarking on that for your PhD and Fei Fei, you like having that reaction that you had.

• 14:33Fei-Fei Li?

  Yeah, so I think that line of work started between me and Andre and then Justin joined, right? So Andre started his PhD, he and I were looking at what is beyond image that object recognition. And at that time, We, you know, convolutional neural network was, uh, has proven some power in image that tasks. So, so ConvNet is a great way to represent images. In the meantime, I think in the language that space, a early sequential model is called LSTM, was also being experimented. So Andre and I were just talking about, this has been the long term dream of mine.

• 15:18Fei-Fei Li?

  I thought it would take a hundred years to, to solve, which is telling the story of images. When I graduated from grad school, I really thought the rest of my entire career would be towards solving that single problem, which is given a picture or given a scene, tell the story in natural language. But things evolved so fast. When Andre and I, uh, when Andre started, we're like, maybe combining the representation of convolutional neural network as well as the, the, the language sequential model of LSTM, we, we might be able to learn uh, through training to match a, um, caption with images. So that's when we started that line of work.

• 16:02Fei-Fei Li?

  And I, I don't, it was 2014 or 2015?

• 16:06Speaker 00

  It was a CBPR 2015 was the captioning paper.

• 16:08Fei-Fei Li?

  So it was our first paper that, uh, Andre got it to work that was, you know, given an image. The image is represented with ConvNet. The language model is the LSTM model. And then we combine it and it's able to generate one sentence. And that was one of the first time. It was pretty, I, I think I wrote it in my book. Uh, we thought we were the first people doing it. It turned out that Google at that time was also simultaneously doing it. And a reporter, it was John Markov from, uh, uh, New York Times was breaking the Google story. But he by accident heard about us. And then he realized that we really independently

• 16:52Fei-Fei Li?

  got there together at the same time. So he wrote the story of both the Google research as well as Andre and my research. But after that, I think Justin was already in the lab at that time.

• 17:04Speaker 00

  Yeah, yeah, I remember the group meeting where Andre was presenting some of those results and explaining this new thing called LSTMs and RNNs that I had never heard of before. And I thought like, wow, this is really amazing stuff. I want to, I want to work on that. So then he had the paper at CBPR 2015 on the first image captioning results. Then after that, we started working together and we did it. First, we did a paper actually just on language modeling.

• 17:26Fei-Fei Li?

  Yep.

• 17:26Speaker 00

  Back in 2015, ICL 2015.

• 17:30Fei-Fei Li?

  Yeah.

• 17:30Speaker 00

  Yeah, I should have stuck with language modeling. That turned out to be pretty lucrative in retrospect. But we did this language modeling paper together, me and Andre, in 2015, where it was like really cool. We trained these little RNN language models that could, you know, spit out a couple of sentences at a time and poke at them and try to understand what the neurons inside the neural net.

• 17:48Fei-Fei Li?

  Yeah, I remember you guys were doing analysis on the different like memory and.

• 17:53Speaker 00

  Yeah, yeah, it was really, it was really cool. Even at that time, we had these results where you could like look inside the LSTM and say like, oh, this thing is reading code. So one of the, like, one of the data sets that we, that we trained on for this one was the Linux source code, right? Because the whole, the whole, the whole thing is, you know, open source and you could just download this. So we trained an RNN on this, on this data set. And then as the network is trying to predict the tokens there, then, you know, try to correlate the kinds of predictions that it's making with the kind of internal structures in the RNN. And there, we were able to find some correlations between, oh, like this unit in this layer of the LSTM fires when there's an open paren and then like turns off when there's a closed paren and try to do some empirical stuff like that to figure it out. So that was pretty cool. And that was just like, that was kind of

• 18:35Speaker 00

  like cutting out the CNN from this language modeling part and just looking at the language models in isolation.

• 18:40Fei-Fei Li?

  But then we wanted to extend the image captioning work. And I remember at that time, we even have a sense of space because we feel like captioning does not capture different parts of the image.

• 18:52Speaker 00

  Right.

• 18:53Fei-Fei Li?

  So I was talking to Justin and Andre about. Can you go what we what we end up calling dense captioning, which is, you know, describe the scene in greater details, especially different parts of the scene. So that's.

• 19:07Speaker 00

  Yeah. And so then then we built this system. And it was me and Andre and Fei Fei on a paper the following year, CVPR, so in 2016, where we built this system that did dense captioning. So you input a single image and then it would draw boxes around all the interesting stuff in the image and then write a short snippet about each of them. It's like, oh, it's a green water bottle on the table. It's a person wearing a black shirt. And this was a really complicated neural network because that was built on a lot of advancements that had been made in object detection around that time, which was a major topic in computer vision for a long time. And then it was actually like one joint neural network that was both, you know, learning to look at individual images. Because it actually had like then three different representations inside this network. One was the representation of the whole image to kind of get the gestalt of what's going on. Then it would propose individual regions that it wants to focus on

• 19:51Speaker 00

  and then look at, you know, represent each region independently. And then, once you look at the region, then you need to spit out text for each region. So that was a pretty complicated neural network architecture. This was all pre PyTorch. And does it do it in one pass? Yeah, yeah. So it was a single forward pass that did all of it.

• 20:03Fei-Fei Li?

  Not only it was doing it in one pass, you also optimize inference. You're doing it on a webcam, I remember.

• 20:10Speaker 00

  Yeah, yeah. Yeah, so I had built this like crazy real time demo where I had the network running like on a server at Stanford and then a web front end that would stream from a webcam and then like send the image back to the server. The server would run the model and stream the predictions back. So I was just like walking around the lab with this laptop that would just like show people this like this network running in real time. Identification and labeling as well.

• 20:33Fei-Fei Li?

  Yeah.

• 20:33Speaker 00

  Yeah, yeah, yeah.

• 20:34Fei-Fei Li?

  It was it was it was pretty impressive because most of my graduate students would be satisfied if they can publish the paper, right? They packaged the the research, put it in a paper, but just the wind does step further. He's like, I want to do this real time web demo.

• 20:50Speaker 00

  Well, actually, I don't know if I told you this story, but then we had a there was a conference that year in Santiago at ICCV. It was ICCV 15. And then like I had a paper at that conference for something different, but I had my my laptop. I was like walking around the conference with my laptop, showing everybody this like real time captioning demo. And the model was running on a server in California. So it was like actually able to stream like all the way from California down to Santiago. Well, Lindsay does. It was terrible. It was like one FPS. But the fact that it worked at all was pretty was pretty amazing.

• 21:20Speaker 03

  I was gonna briefly quip that, you know, maybe vision and language modeling are not that different. You know, DeepSeek CR recently tried the crazy thing of let's language, let's model text from pixels and just like trained on that. And it might be the future. I don't know. I don't know if you guys have any takes on whether language is actually necessary at all.

• 21:40Fei-Fei Li?

  I just wrote a whole manifesto of space technologies.

• 21:43Speaker 03

  This is my segue into this. Yes.

• 21:46Fei-Fei Li?

  I think they are different. I do think the architecture of these generative models will share a lot of shareable components, but I think the deeply 3D, 4D spatial world has a level of structure that is fundamentally different from a purely generative signal that is one dimensional.

• 22:13Speaker 00

  Yeah, I think there's something to be said for pixel maximalism, right? Like there's this notion that language is this different thing, but we see language with our eyes and our eyes are just like, you know, basically pixels, right? Like we've got sort of biological pixels in the back of our eyes that are processing these things. And, you know, we see text and we think of it as this discrete thing, but that really only exists in our minds. Like the physical manifestation of text and language in our world are, you know, physical objects that are printed on things in the world and we see it with our eyes.

• 22:41Fei-Fei Li?

  Well, you can also think it's sound, but even sound, even sound you can translate into a chorale, yeah, a choralegram, which is a 2D signal.

• 22:49Speaker 00

  Right. And then like you actually lose something if you translate to this like purely tokenized representations that we use in LLMs, right? Like you lose the font, you lose, you lose the line breaks, you lose sort of the 2D arrangement on the page. And for a lot of cases, for a lot of things, maybe that doesn't matter, but for some things it does. And I think pixels are this sort of more, more lossless representation of what's going on in the world. And in some ways, a more general, general representation that more matches what, what we, what we humans see as we, as we navigate the world. So, so like there's an efficiency argument to be made, like maybe it's not super efficient to like, you know, render your text to an image and then feed that to a vision model.

• 23:24Speaker 03

  That's exactly what DeepSeek did.

• 23:26Speaker 00

  Yeah.

• 23:26Speaker 03

  It was like, kind of worked.

• 23:29Speaker 04

  I think this ties into the whole world model. Like, one of my favorite papers that I saw this year was about inductive bias to world models. So it was a Harvard paper where they fed a lot of orbital patterns into an LLM. And then they asked the LLM to predict the orbit of a planet around the sun. And the model generated looked good, but then if you asked it to draw the force vectors, it would be all wacky, you know, it wouldn't actually follow it. So how do you think about what's embedded into the data that you get? And we can talk about maybe tokenizing for 3D world models. Like, what are the dimensions of information? There's the visual, but like, how much of the underlying hidden forces, so to speak, you need to extract out of this data?

• 24:14Speaker 04

  And like, what are some of the challenges there?

• 24:16Speaker 00

  Yeah, I think there's different ways you could approach that problem. One is like, you could try to be explicit about it and say like, oh, I want to, you know, measure all the forces and feed those as training data to your model, right? Then you could like, sort of run a traditional physics simulation and, you know, then know all the forces in the scene and then use those as training data to train a model that's now going to hopefully predict those. Or you could hope that something emerges more latently, right? That you kind of train on something end to end and then on a more general problem. And then hope that somewhere, something in the internals of the model must learn to model something like physics in order to make the proper predictions. And those are kind of the two big paradigms that we have more generally.

• 24:53Fei-Fei Li?

  But there's no indication that those latent modeling will get you to a causal law of space and dynamics, right? That's where today's deep learning and human intelligence actually start to bifurcate. Because fundamentally, the deep learning is still fitting patterns.

• 25:13Speaker 00

  There you sort of get philosophical. And you say that like, we're trying to fit patterns too, but maybe we're trying to fit a more broad array of patterns, like with a longer time horizon, a different reward function. But like, basically, the paper you mentioned is sort of, you know, that problem, that it learns to fit the specific patterns of orbits, but then it doesn't actually generalize in the way that you'd like. It doesn't have a sort of causal model of gravity.

• 25:34Speaker 04

  Right, because even in marble, you know, I was trying it and it generates this beautiful sceneries and there's like arches in them. But does the model actually understand how, you know, the arch is actually, you know, drawing on the center kind of like stone and like, you know, the actual physical structure of it? And the other question is like, does it matter? That it does understand it, as long as it always renders something that would fit the physical model that we imagine.

• 26:00Fei-Fei Li?

  If you use the word "understand" the way you understand, I'm pretty sure the model doesn't understand it. The model is learning from the data, learning from the pattern. Yeah, does it matter, especially for the use cases? For it, it's a good question, right? Like, for now, I don't think it matters because it renders out what you need, assuming it's perfect.

• 26:25Speaker 00

  Yeah, I mean, it depends on the use case. Like, if the use case is, I want to generate sort of a backdrop for virtual film or production or something like that, all you need is something that looks plausible. And in that case, probably it doesn't matter. But if you're gonna use this to like, you know, if you're an architect and you're gonna use this to design a building that you're then gonna go build in the real world, then yeah, it does matter that you model the forces correctly because you don't want the thing to break when to actually actually build it.

• 26:46Fei-Fei Li?

  But even there, right? Like, even if your model has the semantics in it, let's say, I still don't think the understanding of the signal or the or the the output on the model part and the understanding on the human part is a different word. But this gets, again, philosophical.

• 27:07Speaker 00

  Yeah, I mean, there's this trick with understanding, right? Like, these models are a very different kind of intelligence than human intelligence. And human intelligence is interesting because, you know, you know, I think that I understand things because I can introspect my own thought process to some extent. And then I believe that my thought process probably works similar to other people's so that when I observe someone else's behavior, then I infer that their internal mental state is probably similar to my own internal mental state that I've observed. And therefore, I know that I understand things. So there, I assume that you understand something. But these models are sort of like this this alien form of intelligence, where they can do really interesting things, they can exhibit really interesting behavior. But whatever kind of internal, the equivalent of internal cognition or internal self-reflection that they have, if it exists at all, is totally different from what we do.

• 27:52Fei-Fei Li?

  It doesn't have the self-awareness.

• 27:53Speaker 00

  Right. But what that means is that when when we observe seemingly interesting or intelligent behavior out of these systems, we can't necessarily infer other things about them because their their model of the world and the way they think is so different from us.

• 28:06Speaker 04

  So would you need two different models to do the visual one and the architectural generation, you think? Eventually, like, there's not anything fundamental about the approach that you've taken on the model building? It's more about scaling the model and the capabilities of it? Or like, is there something about?

• 28:23Speaker 03

  Being very visual that prohibits you from actually learning the physics behind this, so to speak, so that you could trust it to generate a CAD design that then is actually going to work in the real world.

• 28:36Fei-Fei Li?

  I think this is a matter of scaling data and bettering model. I don't think there's anything fundamental that separates these two.

• 28:44Speaker 00

  Yeah, I would like it to be one model, but I think like the big problem in deep learning in some sense is how do you get emergent capabilities beyond your training data? Are you going to get something that understands the forces while it wasn't trained to predict the forces, but it's going to learn them implicitly internally? And I think a lot of what we've seen in other large models is that a lot of this emergent behavior does happen at scale. And will that transfer to other modalities and other use cases and other tasks? I hope so, but that'll be a process that we need to play out over time and see.

• 29:11Speaker 03

  Is there a temptation to rely on physics engines that already exist out there that are, you know, basically the gaming industry has saved you a lot of this work? Or do we have to reinvent things for some fundamental mismatch?

• 29:25Speaker 00

  I think that's sort of like climbing the ladder of technology, right? Like in some sense, the reason that you want to build these things at all is because maybe traditional physics engines don't work in some situations. If a physics engine was perfect, we would have sort of no need to build models because the problem would have already been solved. So in some sense, the reason why we want to do this is because classical physics engines don't solve problems in the generality that we want. But that doesn't mean we need to throw them away and start everything from scratch, right? We can use traditional physics engines to generate data that we then train our models on. And then you're sort of distilling the physics engine into the weights of the neural network that you're training.

• 29:57Speaker 03

  I think that's a lot of what, if you compare the work of other labs, people are speculating that, you know, Sora had a little bit of that, Genie 3 had a bit of that. And Genie 3 is explicitly like a video game, like you have controls to walk around in. And I always think like, it's really funny how the things that we invent for fun actually does eventually make it into serious work.

• 30:21Fei-Fei Li?

  Yeah, the whole AI revolution started by graphics chips, partially.

• 30:28Speaker 03

  Misusing the GPU for generating a lot of triangles into generating a lot of everything else, basically. We touched on marble a little bit. I think you guys chose marble as I kind of feel like you're sort of a little bit coming out of stealth moment, if you can call it that.

• 30:41Fei-Fei Li?

  Yeah.

• 30:42Speaker 03

  Maybe we could get a concise explanation from you on what people should take away, because everyone here can try marble. But I don't think they might be able to link it to the differences between what your vision is versus other, I guess, generative worlds that we have seen from other labs.

• 30:58Fei-Fei Li?

  So marble is a glimpse into our model, right? We are a model, spatial intelligence model company. We believe spatial intelligence is the next frontier. In order to make spatially intelligent models, the model has to be very powerful in terms of its ability to, you know, understand, reason, generate in very multimodal fashion of worlds, as well as allow the level of interactivity that we eventually hope to be as, you know, complex as how humans can interact with the world. So that's the grand vision of spatial intelligence, as well as the kind of world models we see.

• 31:38Fei-Fei Li?

  Marble is the first glimpse into that. It's the first part of that journey. It's the first in-class model in the world that generates 3D worlds in this level of fidelity that is in the hands of the public. It's the starting point, right? We actually wrote this tech blog. Justin spent a lot of time writing that tech blog. I don't know if you had time to browse it. I mean, Justin really broke it down into what are the inputs? We can multimodal inputs of marble, what are the kind of editability, which is, you know, allows user to be interactive with the model, and what are the kind of outputs we can have?

• 32:20Speaker 00

  Yeah, so Marble, like, basically, one way of looking at it, it's a system, it's a generative model of 3D worlds, right? So you can input things like text, or image, or multiple images, and it will generate for you a 3D world that kind of matches those inputs. And it's also interactive, in the sense that you can interactively edit scenes. Like, I could generate this scene, and then say, "I don't like the water bottle, make it blue instead." Like, take out the table, like, change these microphones around. And then you can generate new worlds based on these interactive edits, and export in a variety of formats. And with Marble, we were actually trying to do sort of two things simultaneously, and I think we managed to pull off the balance pretty well. One is actually build a model that goes towards the grand vision of spatial intelligence. And models need to be able to understand lots of different kinds of inputs, need to

• 33:03Speaker 00

  be able to model worlds in a lot of situations, need to be able to model counterfactuals of how they could change over time. So we wanted to start to build models that have these capabilities, and Marble today does already have hints of all of these. But at the same time, we're a company, we're a business, we were really trying not to have this be a science project, but also build a product that would be useful to people in the real world today. So while Marble is simultaneously a world model that is building towards this vision of spatial intelligence, it was also very intentionally designed to be a thing that people could find useful today. And we're starting to see emerging use cases in gaming, in VFX, in film, where I think there's a lot of really interesting stuff that Marble can do today as a product, and

• 33:45Speaker 00

  then also set a foundation for the grand world models that we want to build going into the future.

• 33:50Speaker 03

  Yeah, I noticed one tool that was very interesting, is you can record your scene inside.

• 33:54Speaker 00

  Yes.

• 33:55Fei-Fei Li?

  It's very important. The ability to record means a very precise control of camera placement. And in order to have precise camera placement, it means you have to have a sense of 3D space. Otherwise, you don't know how to orient your camera, right, and how to move your camera. So that is a natural consequence of this kind of model, and this is why, this is just one of the examples.

• 34:21Speaker 03

  Yeah, I find when I play with video generative models, I'm having to learn the language of being a director, because I have to move them, like, pan, you know, like...

• 34:31Fei-Fei Li?

  But even there, you cannot say "pan 63 degrees to the north," right? You just don't have that control. Whereas in Marble, you have precise control in terms of placing your camera.

• 34:43Speaker 04

  Yeah, I think that's one of the first things.

• 34:46Speaker 03

  People need to understand this like it's not you're not generating frame by frame, which is like what a lot of the other models are.

• 34:52Fei-Fei Li?

  Yeah.

• 34:52Speaker 03

  What are, you know, people understand that LLM generates one token. What are like the atomic units? There's kind of like, you know, the meshes, there's like the splats, the voxels. There's a lot of pieces in a 3D world. What should be the mental model that people have of like your generations?

• 35:07Speaker 00

  Yeah, I think there's like what exists today and what could exist in the future. So what exists today is the model natively outputs splats. So Gaussian splats are these like, you know, each one is a tiny, tiny particle that's semi-transparent, has a position orientation in 3D space. And the scene is built up from a large number of these Gaussian splats. And Gaussian splats are really cool because you can render them in real time really efficiently. So you can render on your iPhone, render, render everything. And that's, that's how we get that sort of precise camera control because the splats can be rendered real time on just on pretty much any client side device that we want. So for a lot of the scenes that we're generating today, that kind of atomic unit is that individual splat. But I don't think that's fundamental. I could imagine other approaches in the future that would be interesting. So there like there are other approaches that even we've worked on at World Labs,

• 35:52Speaker 00

  like our recent RTFM model that does generate frames one at a time. And there the atomic unit is generating frames one at a time as the user interacts with the system. Or you could imagine other architectures in the future where the atomic unit is a token, where that token now represents, you know, some chunk of the 3D world. And I think there's a lot of different architectures that we can experiment with here over time.

• 36:12Speaker 03

  I do want to press on, double click on this a little bit. My version of what Alessio was going to say was like, what is the fundamental data structure of a world model? Because exactly like you said, like it's either Gaussian splat or it's like the frame or what have you. You also in the previous statements, focus a lot on the physics and the forces, which is something over time, which is loosely. I don't see that in marble. I presume it's not there yet. Maybe if there was like a marble two, you would have movement. Or is there a modification to Gaussian splats that makes sense? Or would it be something completely different?

• 36:45Speaker 00

  Yeah, I think there's a couple of modifications that make sense. And there's actually a lot of interesting ways to integrate things here, which is another nice place of working in this space. Then there's actually been a lot of research work on this. Like when you talk about wacky ideas, like there's actually been a lot of really interesting academic work on different ways to imbue physics.

• 36:59Fei-Fei Li?

  Because there's also the wacky ideas in this.

• 37:03Speaker 00

  All right, but then it's like Gaussian splats are themselves little particles. There's been a lot of approaches where you basically attach physical properties to those splats. And say that each one has a mass, or like maybe you treat each one as being coupled with some kind of virtual spring to nearby neighbors. And now you can start to do sort of physics simulation on top of splats. So one kind of avenue for adding physics or dynamics or interaction to these things would be to, you know, predict physical properties associated with each of your splat particles. And then simulate those downstream, either using classical physics or something learned. Or, you know, the kind of the beauty of working in 3D is things compose and you can inject logic in different places. So one way is sort of like we're generating a 3D scene. We're going to predict 3D properties of everything in the scene. Then we use a classical physics engine to simulate the interaction.

• 37:48Speaker 00

  Or you could do something where like, as a result of a user action, the model is now going to regenerate the entire scene in splats or some other representation. And that could potentially be a lot more general, because then you're not bound to whatever sort of, you know, physical properties you know how to model already. But that's also a lot more computationally demanding, because then you'd need to regenerate the whole scene in response to user actions. But I think this is, this was a really interesting area for future work and for adding on to, into potential Marvel 2, as you say.

• 38:18Fei-Fei Li?

  Yeah, there is opportunity for dynamics, right?

• 38:22Speaker 04

  What's the state of like splats density, I guess? Like, do we, can we render enough to have very high resolution when we zoom in? Are we limited by like the amount that you can generate? The amount that we can render, like, how are these gonna get super high fidelity, so to speak?

• 38:37Speaker 00

  You have some limitations, but depending on your target use case. So like, one of the, one of the big constraints that we have on our scenes is we wanted things to render cleanly on mobile. And we wanted things to render cleanly in VR headsets. So those are, those devices have a lot less compute than you're used, than you have in a lot of other situations. And like, if you want to get a splat file to render at high resolution, high like 30 to 60 FPS on like an iPhone from four years ago, then you are a bit limited in like the number of splats that you can handle. But if you're allowed to like work on a recent, like even this year's iPhone, or like a recent MacBook, or even if you have a local GPU, or if you don't need, if you don't need that 60 FPS, 1080p, like, then you can relax the constraints and get away with more splats and that lets you get high resolution in your scenes.

• 39:21Speaker 03

  One use case I was expecting but didn't hear from you was embodied use cases.

• 39:26Fei-Fei Li?

  Mm-hmm.

• 39:27Speaker 03

  You're just focusing on virtual for now?

• 39:29Fei-Fei Li?

  If you go to World Labs homepage, there is a particular page called Marble Labs. There, we showcase different use cases. And we actually organize them in more visual effect use cases or gaming use cases, as well as simulation use cases. And in that, we actually show this is a technology that can help a lot in robotic training, right? This goes back to what I was talking about earlier. In speaking of data starvation, robotic training really lack data. You know, high fidelity, real world data is absolutely very critical, but you're just not going to get a ton of that. Of course, the other extreme is just purely internet video data,

• 40:14Fei-Fei Li?

  but then you lack a lot of the controllability that you want to train your embodied agents with. So simulation and synthetic data is actually a very important middle ground for that. I've been working in this space for many years. And one of the biggest pain point is where do you get this synthetic simulated data? You have to curate assets and build these, compose these complex situations. And in robotics, you want a lot of different states. You want the embodied agent to interact in the synthetic environment. Marble actually is a really potential for helping to generate these synthetic simulated worlds

• 40:58Fei-Fei Li?

  for embodied agent training.

• 41:01Speaker 03

  Obviously, that's on the homepage. It'll be there. I was like trying to make the link to, as you said, like you also have to build like a business model. The market for robotics obviously is very huge. Maybe you don't need that, or maybe we need to build up and solve the virtual worlds first before we go to embodied. And obviously, I think that's a big step in zone.

• 41:20Fei-Fei Li?

  That is to be decided. I do think that...

• 41:24Speaker 03

  Because everyone else is going straight there.

• 41:27Speaker 04

  Right?

• 41:27Fei-Fei Li?

  Not everyone else, but there is a, there is an excitement, I would say. But, you know, I think the world is big enough to, to have different.

• 41:37Speaker 04

  Approaches.

• 41:38Fei-Fei Li?

  Yeah, approaches.

• 41:39Speaker 00

  Yeah, I mean, and we always view this as a pretty horizontal technology that should be able to touch a lot of different industries over time. And, you know, marble is a little bit more focused on creative industries for now. But I think the, the technology that powers it should be applicable to a lot of different things over time. And robotics is one that, you know, is maybe gonna happen sooner than later.

• 41:56Fei-Fei Li?

  Also, design, right, is very adjacent to creative.

• 41:59Speaker 00

  Oh, yeah, definitely. Like, I, I think it's like.

• 42:01Speaker 04

  The architecture stuff?

• 42:02Fei-Fei Li?

  Yes.

• 42:02Speaker 00

  Okay. Yeah, I mean, I, I was joking online. I posted this, uh, this video on Slack of like, oh, who wants to use marble to, to plan your next kitchen remodel? It actually works great for this already. Just like, take two images of your kitchen, like reconstructed in marble, and then use the editing features to see what would that space look like if you change the countertops, or change the floors, or change the cabinets? And this is something that's, you know, we didn't necessarily build anything specific for this use case. But because it's a, it's a powerful horizontal technology, you kind of get these emergent use cases that, that just fall out of the model.

• 42:30Fei-Fei Li?

  We have early beta users using a, um, um, API key that is already building for interior design, uh, use case.

• 42:40Speaker 04

  I just did my garage. I should have known about this.

• 42:42Fei-Fei Li?

  I know. Next time you remodel, we can.

• 42:46Speaker 04

  Exactly.

• 42:47Fei-Fei Li?

  Be of help.

• 42:48Speaker 04

  Well, kitchen is next, I'm sure.

• 42:49Fei-Fei Li?

  Yeah.

• 42:50Speaker 04

  Yeah, I'm curious about the whole spatial intelligence space. I think we should take more into that. One, how do you define it? And like, what are like the gaps between traditional intelligence that people might think about a L L M S? When, you know, Dario says, we have a data center full of Einsteins. That's like traditional intelligence. It's not spatial intelligence. What is required to be spatially intelligent?

• 43:14Fei-Fei Li?

  First of all, I don't understand that sentence, a data center full of Einsteins.

• 43:19Speaker 04

  Ooh.

• 43:21Fei-Fei Li?

  I, I just don't understand that. No, no, no, it's not a.

• 43:24Speaker 04

  It's an analogy. It's an analogy.

• 43:26Fei-Fei Li?

  Well, so. A lot of AI as a field, as a discipline, is inspired by human intelligence, right? Because we are the most intelligent animal we know in the universe for now. And if you look at human intelligence, it's very multi-intelligent, right? There is a psychologist, I think his name is Howard Gardner, in the 1960s, actually literally called multiple intelligence to describe human intelligence. And there is linguistic intelligence, there is spatial intelligence, there is a logical intelligence and emotional intelligence. So for me, when I think about spatial intelligence, I see it as complementary to language intelligence.

• 44:10Fei-Fei Li?

  So I personally would not say it's spatial versus traditional, because I don't know, tradition means, what does that mean? I do think spatial is complementary to linguistic. And how do we define spatial intelligence? It's the capability that allows you to reason, understand, move and interact in space. And I use this example of the deduction of DNA structure, right? And of course, I'm simplifying this story, but a lot of that had to do with the spatial reasoning of the molecules and the chemical bonds in the 3D space

• 44:54Fei-Fei Li?

  to eventually conjecture a double helix. And that ability that humans, or Francis Crick and Watson, had done, it is very, very hard to reduce that process into pure language. And that's a pinnacle of a civilizational moment. But every day, right? I'm here trying to grasp a mug. This whole process of seeing the mug, seeing the context where it is, seeing my own hand, opening of my hand that geometrically would match the mug, and touching the right affordance points, all this is deeply, deeply spatial. It's very hard.

• 45:38Fei-Fei Li?

  I'm trying to use language to narrate it. But on the other hand, that narrated language itself cannot get you to pick up a mug.

• 45:47Speaker 03

  Yeah, bandwidth constrained.

• 45:48Fei-Fei Li?

  Yes.

• 45:49Speaker 03

  I did some math recently on like, if you just spoke all day, every day for 34 hours a day, how many tokens do you generate? At the average speaking rate of like 150 words per minute, it roughly rounds out to about 215,000 tokens per day. And like, your world that you live in is so much higher bandwidth than that.

• 46:10Speaker 04

  Well, I think that is true, but if I think about Sir Isaac Newton, right? It's like, you have things like gravity at the time that have not been formalized in language that people inherently spatially understand that things fall, right? But then it's helpful to formalize that in some way. Or like, you know, all these different rules that we use language to like, really capture something that empirically and spatially you can also understand, but it's easier to like describe in a way. So I'm curious, like, the interplay of like spatial and like linguistic intelligence, which is like, okay, you need to understand. Some rules are easier to write in language for than the spatial intelligence to understand. But you cannot, you know, you cannot write, put your hand like this and put it down this amount.

• 46:55Speaker 04

  So I'm always curious about how you leverage each other together.

• 46:59Speaker 00

  I mean, if anything, like the example of Newton, like Newton only thinks to write down those laws because he's had a lot of embodied experience in the world.

• 47:05Speaker 04

  Right, yeah, exactly.

• 47:06Speaker 00

  And actually, it's useful to distinguish between the theory building that you're mentioning versus like the embodied, like the daily experience of being embedded in the three dimensional world. Right? So to me, spatial intelligence is sort of encapsulating that embodied experience of being there in 3D space, moving through it, seeing it, actioning it. And as Fei-Fei said, you can narrate those things, but it's a very lossy channel. It's just like the notion of, you know, being in the world and doing things in it is a very different modality from trying to describe it. But because we as humans are animals who have evolved interacting in space all the time, like, we don't even think that that's a hard thing, right? And then we sort of naturally leap to language and then theory building as mechanisms to abstract above that sort of native spatial understanding.

• 47:50Speaker 00

  And in some sense, LLMs have just like jumped all the way to those highest forms of abstracted reasoning, which is very interesting and very useful. But spatial intelligence is almost like opening up that black box again and saying, maybe we've lost something by going straight to that fully abstracted form of language and reasoning and communication.

• 48:05Fei-Fei Li?

  You know, it's funny, as a vision scientist, right? I always find that vision is underappreciated because it's effortless for humans. You open your eyes as a baby, you start to see your world.

• 48:18Speaker 03

  We're somehow born with it.

• 48:19Fei-Fei Li?

  Right, we're almost born with it. But you have to put effort in learning language, including learning how to write, how to do grammar, how to express, and that makes it feel hard. Whereas something that nature spent way more time actually optimizing, which is perception and spatial intelligence, is underappreciated by humans.

• 48:41Speaker 03

  Is there proof that we are born with it? You said, you said almost born. So it sounds like we actually do learn after we're born.

• 48:48Fei-Fei Li?

  When we are born, our visual acuity is less and our perceptual ability does increase. But we are, most humans are born with the ability to see. And most humans are born with the ability to link perception with motor movements, right? I mean, the motor movement itself is, takes a while to refine. But, and then animals are incredible, right? Like, I was just in Africa earlier this summer. These little animals, they're born and within minutes, they have to get going. And otherwise, the, you know, the lions will get them. And in nature, you know, it took 540 million years to optimize perception and spatial intelligence.

• 49:33Fei-Fei Li?

  And language, the most generous estimation of language development is probably half a million years.

• 49:41Speaker 03

  Wow.

• 49:41Fei-Fei Li?

  Yeah.

• 49:42Speaker 03

  That's longer than I would have gonna say.

• 49:44Fei-Fei Li?

  I'm being very generous.

• 49:45Speaker 03

  Yeah. Yeah, no, I was, you know, sort of going through your book. And I was really realizing that one of the interesting links to something that we covered on the podcast is language model benchmarks and how Winogrand actually put in all these sort of physical impossibilities that require spatial intelligence, right? Like, A is on top of B, therefore, A cannot fall through B is. Obvious to us, but to a language model, it could happen. I don't know. Maybe it's like a part of the, you know, the next token prediction.

• 50:13Speaker 00

  And that's kind of what I mean about like unwrapping this abstraction. Right? Like if your whole model of the world is just like saying sequences of words after each other, it's really kind of hard to like, why not?

• 50:23Speaker 03

  It's actually unfair.

• 50:24Speaker 00

  Right. But then the reason it's obvious to us is because we are internally mapping it back to some three dimensional representation of the world that we're familiar with.

• 50:31Speaker 03

  This, the question is, I guess, like, how hard is it? You know, how long is it going to take us to distill from like, I, I, I use the word distill. I don't know if you agree with that, to distill from your world models into a language model? Cause we do want our models to have spatial intelligence, right? Like, and do we have to throw the language model out completely in order to do that? Or?

• 50:50Fei-Fei Li?

  No.

• 50:50Speaker 03

  No, right?

• 50:51Speaker 00

  Yeah, I don't think so.

• 50:52Fei-Fei Li?

  I think they're multimodal. I mean, even our model, Marble today, takes language as an input.

• 50:56Speaker 03

  Right.

• 50:57Fei-Fei Li?

  Right? So it's deeply multimodal. And I think in many use cases, these models will work together. Maybe one day we'll have a universal model.

• 51:06Speaker 00

  I mean, even if you do, like, there, there's sort of a pragmatic thing where people use language and people want to interact with systems using language. Even pragmatically, it's useful to build systems and build products and build models that let people talk to them. So I don't see that going away. I think there's a sort of intellectual curiosity of saying how like intellectually, how much could you build a model that, that only uses vision or only uses spatial intelligence? I don't know that that would be practically useful, but I think it'd be an interesting intellectual or academic exercise to see how far you could push that.

• 51:36Speaker 04

  I think, I mean, not to bring them back to physics, but I'm curious, like, if you had a highly precise world model and you didn't give it any notion of like our current understanding of the standard model of physics, how much of it it will be able to come up with and like recreate from scratch? And what level of like language understanding it would need? Because we have so many notations that like we kind of use that like we create it, but like maybe we'll come up with a very different model of it and still be accurate. And I wonder how much. We're kind of limited, but I, you know how people say humans always need to be like humans because the world is built for humans? And in a way, it's like the way we build language constraints, some of the outputs that we can get from these other modalities as well. So I'm super excited to follow your work.

• 52:20Speaker 00

  Yeah, I mean, like, there's another, I mean, you actually don't even need to be doing AI to answer that question. You could discover aliens and see what kind of physics they have, right? And they might have a totally different.

• 52:28Speaker 04

  Well, Faye said, we are so far the smartest animal in the universe.

• 52:32Speaker 00

  So what are you, so I mean, but that is a really interesting question, right? Like, is our knowledge of the universe and our understanding of physics, is it constrained in some way by our own cognition or by the path dependence of our own technological evolution? And one way to sort of, and like, do an experiment, like, you almost want to do an experiment and say, like, if we were to rerun human civilization again, would we come up with the same physics in the same order? And I don't think that's a very practical, practical experiment to run.

• 52:55Fei-Fei Li?

  You know, one experiment I wonder if people could run is that we have plenty of astrophysical data now on the planet or celestial body movements. Just feed the data into a model and see if Newtonian law emerges.

• 53:12Speaker 00

  My, my guess is it probably, probably won't.

• 53:14Fei-Fei Li?

  That's my guess, it's not. The abstraction level of Newtonian law is at a different level from what these language L L Ms represents.

• 53:25Speaker 00

  Yeah.

• 53:26Fei-Fei Li?

  So I wouldn't be surprised that, given enough celestial movement data, an L L M would actually predict pretty accurate movement trajectories. Let's say I invent a planet surrounding a, a star and given enough data, my, my model would tell you, you know, on day one where it is, day two where it is. I wouldn't be surprised. But F equals M A or, or, you know, action equals reaction, that's just a whole different abstraction level. That's beyond just the today's L L M.

• 54:01Speaker 04

  Okay, what, what model would you need to not have it be a geocentric?

• 54:06Speaker 03

  Model, because if I'm training just on visual data, it makes sense that you think the sun rotates around the earth, right? But obviously, that's not the case. So how would it learn that? Like, I'm curious about all these like, you know, forces that we talk about. It's like, sometimes maybe you don't need them because as long as it looks right, it's right. But like, as you make the jump to like, trying to use these models to do more high-level tasks, how much can we rely on them?

• 54:33Speaker 00

  I think you can need kind of a different learning paradigm, right? So like, you know, there's a bit of conflation here happening where saying, is it LLMs and language and symbols versus, you know, human theory building and human physics? And they're very different because an LLM, like, the human objective function is to understand the world and thrive in your life. And the way that you do that is by, you know, sometimes you observe data and then you think about it, and then you try to do something in the world and it doesn't match your expectations. And then you want to go and update sort of your, your, your, your understanding of the world online. And people do this all the time, constantly, like whether it's, you know, I think my keys are downstairs, so I go downstairs and I look for them and I don't see them. And oh, no, they're actually up in my bedroom. So we're like, because we're constantly interacting with the world,

• 55:17Speaker 00

  we're constantly having to build theories about what's happening in the world around us and then falsify or add evidence to those theories. And I think that that kind of process writ large and scaled up is what gives us F equals ma in Newtonian physics. And I think that's a little orthogonal to, you know, the modality of model that we're training, whether it's language or spatially.

• 55:36Speaker 03

  The way I put it is almost like, this is almost more efficient learning because you have a hypothesis of here are the different possible worlds that are granted by my available data. And then you do experiments to eliminate the worlds that are not possible. And then you resolve to the one that's right. To me, that's also how I also have theory of mind, which is like, I have a few thesis of what you're thinking, what you're thinking. And I try to create actions to resolve that or check my intuition as to what you're thinking, you know? And obviously, LLMs don't do any of these.

• 56:09Fei-Fei Li?

  A theory of mine possibly also will break into even emotional intelligence, which today's AI is really not touching at all, right?

• 56:19Speaker 03

  And we really, really need it. You know, people are starting to depend on these things probably too much. And that's a whole topic of other debate. I do have to ask, because a lot of people have said this to us, how much do we have to get rid of? You know, is sequence to sequence modeling out the window? Is attention out the window? Like, how much are we questioning everything?

• 56:40Speaker 00

  I think you stick with stuff that works, right?

• 56:43Speaker 03

  So attention is still there.

• 56:45Speaker 00

  I think attention is still there. I think there's a lot, like, you don't need to fix things that aren't broken. And like, it's, there's a lot of hard problems in the world to solve, but let's focus on one at a time. I think it is pretty interesting to think about new architectures or new paradigms or drastically different learning ways to learn. But you don't need to throw away everything just because you're working on new modalities.

• 57:05Fei-Fei Li?

  I think sequence to sequence is actually, in world models, I think we are going to see algorithm or architecture beyond sequence to sequence.

• 57:16Speaker 00

  Oh, but here, actually, I think there's a little bit of technological confusion. And transformers already solved that for us, right? Like transformers are actually not a model of sequences. A transformer is natively a model of sets. And that's very powerful. But because a lot of the transformers grew out of earlier architectures based around recurrent neural networks, and RNNs definitely do have like a built-in architectural, like they do model one-dimensional sequences.

• 57:39Speaker 03

  Okay.

• 57:40Speaker 00

  But transformers are just models of sets, and they can model a lot of, those sets could be, you know, 1D sequences. They could be other things as well.

• 57:46Speaker 03

  Do you literally mean set theory, like?

• 57:48Speaker 00

  Yeah, yeah.

• 57:49Speaker 03

  Galois, whatever.

• 57:50Speaker 00

  Yeah, yeah, yeah. So a transformer is actually not a model of a sequence of tokens. A transformer is actually a model of a set of tokens, right? The only thing that gives, that injects the order into it in the standard transformer architecture, the only thing that differentiates the order of the things is the positional embedding that you give the tokens, right? So if you choose to give a sort of 1D positional embedding, that's the only mechanism that the model has to know that it's a 1D sequence. But all the, like, operators that happen inside a transformer block are either token-wise, right? So that either you have an FFN, you have QKPI projections, like you have the per-token normalization. All of those happen independently per token. And then you have interactions between tokens through the attention mechanism. But that's also sort of, it's permutation equivariant.

• 58:31Speaker 00

  So if I permute my tokens, then the attention operator gets a permuted output in exactly the same way. So it's actually a natively, an architecture of sets of tokens. It literally a transform. Yeah.

• 58:43Speaker 00

  In a math term.

• 58:44Speaker 04

  I know we're out of time, but we just want to give you the floor for some call to action either on people that would enjoy working at World Labs. What kind of people should apply? What research people should be doing outside of World Labs that would be helpful to you or anything else on your mind?

• 58:59Fei-Fei Li?

  I do think it's a very exciting time to, to be looking beyond just language models and think about the, the boundless possibilities of spatial intelligence. So we are actually hungry for talent ranging from very deep researchers, right? Thinking about the problems like Justin just described, you know, training large models of world models. Uh, we are hungry for engineers, good engineers, building systems, you know, from training optimization to inference to a product. And we're also hungry for good business, uh, you know, um, product thinkers and go to market and, you know, business talents. So, so we are hungry for, for talent.

• 59:44Fei-Fei Li?

  We, especially now that we have, uh, uh, exposed the model, uh, to the world through Marble. I think we have a great opportunity to work with even a bigger pool of talent to solve both the, the model problem as well as, uh, deliver the best product to, uh, to the world.

• 1:00:03Speaker 00

  Yeah, I think I'm also excited for people to try Marble and do a lot of cool stuff with it. I think it has a lot of really cool capabilities, a lot of really cool features that fit together really nicely.

• 1:00:11Fei-Fei Li?

  In the car coming here, Justin and I were saying, people have not totally discovered the, okay, it's only 24 hours. Have not only discovered some of the advanced mode of editing, right? Like turn on the advanced mode, you can, like Justin said, change this color of the bottle, you know, change your floor and change the trees and...

• 1:00:31Speaker 03

  Well, I actually tried to get there, but when it says create, it just makes me create a completely different world.

• 1:00:36Fei-Fei Li?

  You need to click on the advanced mode.

• 1:00:39Speaker 03

  It's a good UI issue.

• 1:00:40Fei-Fei Li?

  We can improve on our UI UX, but remember to click.

• 1:00:43Speaker 03

  Yeah, we need to hire people who work on the product.

• 1:00:47Speaker 03

  But one thing we got that was clear from you guys are looking for is also intellectual fearlessness, which is something that I think you guys hold as a principle.

• 1:00:54Fei-Fei Li?

  Yeah, I mean, we are literally the first people who are trying this, both on the model side as well as on the product side.

• 1:01:03Speaker 04

  Thank you so much for joining us.

• 1:01:04Fei-Fei Li?

  Thank you, guys.

• 1:01:05Speaker 03

  Yeah, thanks for having us.

• 1:01:06Fei-Fei Li?

  Yeah.

• 1:01:09Speaker 02

  Thanks for listening to the A16Z podcast. If you enjoyed the episode, let us know by leaving a review at ratethispodcast.com/a16z. We've got more great conversations coming your way. See you next time.

• 1:01:22Speaker 02

  This information is for educational purposes only and is not a recommendation to buy, hold, or sell any investment or financial product. This podcast has been produced by a third party and may include paid promotional advertisements, other company references, and individuals unaffiliated with A16Z. Such advertisements, companies, and individuals are not endorsed by AH Capital Management LLC, A16Z, or any of its affiliates. Information is from sources deemed reliable on the date of publication, but A16Z does not guarantee its accuracy.