Inside Nemotron & NVIDIA’s AI Lab | Bryan Catanzaro

29:40We've been doing it for a while. I think our models have gotten dramatically more useful in the past year, which is a reflection of two things. Primarily, one is that the whole company has come together. So there are many different teams around NVIDIA that now understand how important this is to NVIDIA's future. And so there's dramatically more people and better ideas that are going into Nemotron. And then number two, along with that, we've been able to scale the compute resources that go into it. Obviously, it's very important to have good computing infrastructure to build AI. We've recently increased our investment substantially because we believe that this is really, really key to our company's future.

30:22Fascinating. But just to continue the thought, I think it's really important that everybody knows that we've been doing this for a long time. We are increasing our investments substantially. And NVIDIA is a company that follows through. You know, we followed through over 10 plus years with CUDA, and we're doing that with Nemotron now. That's really helpful because I think the broader world is just starting to catch up to the fact that there is a very substantial open source frontier AI research effort that's been happening. So it's really interesting to hear that, you know, there's been this progression and now there's this family of models that we're going to talk about in a second.

31:00Another important moment seems to be the creation just in March, three months ago, of the Nemotron Coalition. Do you want to explain briefly what that is? So Nemo Tron exists to help support the ecosystem. And we were thinking, well, this is a different kind of AI project than other projects around the industry, right? Because we're not actually trying to dominate in any way. We're just trying to support. We're not trying to control the way that AI is being integrated into all these companies. We're just trying to make sure there's good AI. But we thought, well, maybe if we worked with people while we develop it, then it's going to be more useful for them.

31:46It'll be easier to integrate because we will consider what they need from the beginning. And, you know, Nemotron has always been collaborative. I was telling you that, you know, long, long time ago, our first big model that we trained, we did with Microsoft. Right. It was a joint effort where NVIDIA and Microsoft researchers worked side by side to build that. And that ended up, I think, helping both NVIDIA and Microsoft. I think we both learned a lot from that experience. And so because Nemotron is not trying to compete with other companies, but rather support, because we're going to be putting it out there openly anyway, why not collaborate before the thing is built, rather than Nemotron being a project that NVIDIA does all on its own and then posts on the Internet and says, hey, why don't you try this?

32:33We think it might be good. Why don't we make sure that it's good for the partners that are interested by working with them before Nemotron is even created and incorporating any sort of feedback, evaluations, environments, benchmarks, or any other kinds of technology that other people want to bring. It turns out that the entire ecosystem, there's a lot of companies that really want open models to succeed. And so they They have a self-interest. They have their own vested self-interest in making sure that open technologies are excellent. And so why not work with them and let them contribute however they'd like to making NemoTron better?

33:14So that's the idea of the NemoTron Coalition. It is not an exclusive coalition. We're not trying to be the only model out there. All the companies that we work with are free to continue doing the work however makes sense to them. And yet, you know, these companies want to work with us because they want to make sure that open technologies for AI keep developing quickly and that they have a chance to influence how that happens. Great. What's the current state of the Neumotron family? You got Nano, you got Super, you got Ultra. What do those models do and what are the use cases for them? So Nano is a 30 billion total, 3 billion active parameter model.

33:55Super is 120 and 12 and Ultra is 550 and 55. They're designed really to fit, it's kind of small, medium and large deployment scenarios. Nano can be really capable for things that don't require nearly as much knowledge or reasoning, but obviously for the most capable model, you go for Ultra. Super, in a lot of ways, is our most popular model because it represents kind of a great balance between cost and intelligence. So we kind of like having this small, medium and large approach to building a family just because our customers seem to respond to that pretty well. But, you know, the most important thing from NVIDIA's point of view that people are doing with LLMs is agents, right, is building agentic workflows.

34:54Having an agent working on your behalf, solving problems for you night and day is such an exciting way of approaching the problems that we have to solve. And it's our dream to make Nemotron amazing for that purpose. That's our goal. to double click on this at a hell of all nemotron families focused on urgentic reasoning with a particular focus on making it efficient is that is that the right headline that's right yeah um nemotron has always been um uh speed first approach to building models because nvidia is an accelerated computing company as i was saying we're trying to think through what is the problem here computationally from first principles and um you know nemotron 3 family has a lot of things in it that we're really proud of.

35:41For example, Nemotron Ultra and Super were pre-trained using 4-bit arithmetic. We pre-trained those in MVFP4, which is a not trivial thing to do, to invent the algorithm so that your model can converge to an excellent result. Using such coarse arithmetic required a lot of invention. Really proud of that. Do you want to explain maybe for people what 4-bit is versus 16-bit, for example? You know, actually, there was a fantastic post I saw on Hacker News yesterday where somebody let you upload a picture and then it would basically posterize it, basically reduce the colors to fit different number formats, including NVFP4 and MXFP8 and some of the other formats that are out there.

36:24And so you could kind of swipe around and look what it does to the colors of a picture. And, you know, it's really quite dramatic. Four bits is not a lot of bits, right? That's only 16 values. Now, of course, these are all what are called block scaled formats. So groups of numbers also come with an 8-bit scaling factor. And the specifics of this can get rather complicated. So maybe they're not quite as important. But the reason why we want to do this is because, first of all, we have dramatically higher throughput for these formats in our GPUs, specifically on Blackwell Ultra. And secondly, we know that it's going to save an enormous amount of energy.

37:10One way to think about the computational problem of AI is that we are going to be running at the limit, whatever the limit is. It could be an economic limit, like we only have so many billion to buy servers with. It could be a power limit. We only have so many gigawatts that we can afford to train a model with. Whatever the limit is, we're going to be running at that limit. Every organization is... Why? Because the value of intelligence is so high that people are going to invest because they know that they're going to get return. The value of intelligence is enormous. So if you accept as the truth that we're going to be running at the limit, then what that means is that the way to get more intelligence is to be more efficient.

37:59We can't get more intelligence by applying more force if we're already at the limit. We have to be more thoughtful about how we use what we have. And, you know, 4-bit number formats are dramatically cheaper to move around. They take up less space in memory. They take up less picojoules when you move them from the memory or even on the chip, around the chip, much less energy when you compute on them. And so that's really driving the investment in 4-bit formats. And I think these days, 4-bit formats for deployment are very well established. It's pretty straightforward these days to make a good quantized 4-bit checkpoint that you can deploy and that gets you a lot of inference cost and speed advantages.

38:50But using 4-bit formats for pre-training, that's quite a bit more challenging because you have this numeric solver that's optimizing the weights and it can be quite sensitive. So if you don't treat the numbers right, your model can diverge. And instead of actually getting a model done through pre-training, you end up with, you know, basically just that run diverged, which is, you know, always scary. So it took a lot of invention for us to be able to pre-train Nemotron in 4-bit. We're really proud of that. Okay, great. All right. So as we get into slightly more technical things, the architecture of NemoTron is hybrid.

39:36Is that right? So it's a combination of transformer and Mamba state space, which is a slightly more exotic form of architecture. Walk us through that. Yeah, you know, we published a paper in 2024 that showed that you actually get a smarter model by combining state space models with transformers. And we actually did a sweep of, you know, how much of the model should be full attention and how much of it should be a state space model in order to get the lowest perplexity, basically the best language model that you could get. And we found that you actually want it to be mostly a states-based model with a little bit of attention.

40:18And kind of the intuition behind that is that the states-based models seem to be better at kind of this intuitive kind of impressionistic understanding of a sequence. Because they're kind of summarizing the entire sequence into a constant space. That's how they work. So instead of having the ability to look at the entire sequence randomly, they summarize everything at every step into a constant cache or little scratch pad that they're working on. And that constraint seems to actually make them smarter at some tasks that involve global understanding. On the other hand, the advantage of full attention is that it can pick out very specific bits of information and look at those exactly.

41:04It doesn't lose anything. There's no lossy compression going on. and you can actually see the whole thing. And so we found that using both of these together was actually better than using either one on their own. And that is independent of the speed benefit. That is just the model is smarter. And since we published that, I think a lot of other labs have also found this to be true. A lot of models these days are being built with hybrid SSM approaches, For example, QN has done that. Kimi is using what they call Kimi linear attention these days. So it's become, I think, quite widely adopted to use some sort of state space model in conjunction with full attention for the base architecture.

41:55Now, it also has some speed benefits because the amount of memory that you need to hold that state space cache is actually constant with respect to your sequence length, which then means that generally you can fit much higher batches on the GPU when you're training and doing inference because the memory requirement is lower. And it keeps the GPU fuller and busier and therefore, you know, provides some pretty important efficiency benefits as well. So the models are also based on an MOE, mixture of experts, architecture. Walk us through that and maybe remind people what MOE is in the first place.

42:42So mixture of experts is a form of sparsity. The idea is, wow, you want to train a model on the entire Internet. You want it to remember absolutely everything about the history of everything. But when you're answering a particular question, does it seem reasonable that it needs to actually think about the entire universe in order to answer that question? Actually, no. It seems like it's quite sparse, right? It seems like we're using a language model to explore a very tiny space of ideas in order to answer a question or solve a problem. We want the model to be able to draw from the entire universe.

43:16We want to train it so that it understands everything that it possibly can. But when it's actually running, it doesn't really need to see all of that information. There's been a variety of approaches to sparsity that try to take advantage of this property, but mixture of experts has been the most successful. And the way that it works is that the neural network has what's called a router that is learned that is going to decide to send activations to a subset of the experts for every token that's flowing through every layer of the model. It's going to be making choices about which fraction of the model is going to actually get to interact with this token as we try to understand it, build up representations of the problem, and then generate the next token that we're going to output.

43:59So it's a little bit like if I have a company with 550 employees, but 55 of them are in engineering, engineering i want the 55 employees who are specialists to come to my meeting about engineering and not the rest of the company that's right yeah or you can think about it as a library like if you go into a library to do research you don't read all of the books in the library like your first job is to figure out which books do you need to look at in order to find the answer to your question and so um so that's kind of the the idea behind moes now moes have fascinating implications for the systems that we built.

44:37So with Blackwell, for example, NVIDIA went all in on MOEs. That's why we built NBL 72, which allows up to 72 of our GPUs to read and write each other's memory at very high speeds, very low latency. Now, why is that important? It's because as you put a token through the stack of layers, at every layer, you have a router that's routing that token somewhere else. Why don't you partition your experts so that the experts are not sitting, every expert on every GPU, but you have a subset of the experts assigned to each GPU. And then you're routing the tokens between the GPUs very dynamically as you push the token through the network.

45:14Now, this is impossible to predict in advance where the tokens need to go because it's very specific to that particular token for that particular model. And so that's why we built NVL72. And that's why Blackwell is so amazing for inference for today's AI models is because we thought deeply about a mixture of experts when we were building it. And this is speaking to Nematron's first job. If we hadn't been working on understanding AI, we wouldn't have been able to build Blackwell properly. And that has translated directly into increased deployment of Blackwell, which we're very excited about. Is what you just described called latent MOE, or is that a different concept?

45:59Latent MOE is a specific innovation that we have in Nemotron 3 family. And what it does is actually reduces the amount of communication that has to be sent through NVLink during MOE computations by basically down projecting it. So, you know, every token produces a vector. And the idea is like, we're going to take that vector and learn a way to compress it and then send that compressed thing through the network. And then we're going to uncompress it at the other end. And as a result, we save on network bandwidth, and we also get four times the number of experts for the same inference cost. So you could think about it as like, you know, our library of books got four times bigger, and we get to, you know, read four times more books at the same inference cost because of this particular innovation.

46:48Is MOE in general becoming the default architecture for Frontier AI? Yeah, I believe MOEs have been the default in Frontier AI for a long time. They're just a really good combination of inference cost and intelligence. Great, great. But they have drawbacks as well. They take a lot more memory. If you have a very small amount of memory, a dense model is going to be smarter. And they also, they tend to work best either if you're running at batch size one, so you're running basically a single job, or you're running a huge data center with like infinite queries coming in in the middle they can be a little bit tricky another important characteristic of numatron 3 ultra is a 1 million token context the the long context window how important is that in the overall mix and what does it enable the model to do the longer the context length the more challenging problems we can solve with a language model that allows us to do things like append all sorts of information to a query, which could be a code base.

47:54It could be instructions. In the long term, I'm hoping that I have my own personal LLM that's able to read all of my emails and help me answer questions about that. The more information that we can attach to a particular query, the more useful the model can be. Now it can get more and more expensive to reason over large amounts of input data. And so that's one of the reasons why there's usually a limit on how big the context length can be. But with NemoTron 3, we tried to push it as far as we could go. We think a million tokens is a lot of tokens, and you can do a lot of things with that. Prism is particularly helpful in sort of multi-step agentic workflows.

48:41And there's this whole separate discussion around context compaction to make sure that the model doesn't get lost in too many tokens. So like, how do you all think about this? A hundred percent. I mean, compaction, that's a thing if you're using an agentic workflow you deal with all the time. And compaction tends to work pretty well because language models are pretty good at identifying the most relevant things and summarizing, and you're basically trying to summarize your context when you compact it. So compaction is not a bad approach. I think having models that can just natively reason about larger amounts of data is just inherently more useful.

49:23So, of course, we want to push the boundary on that as well. Great. Can you talk about the multi-token prediction, which is also very interesting? If you're running at a low batch size, which is when you are trying to get the most interactivity if you're in a data center. So you want the model to respond as quickly as possible, and it's okay for it to be more expensive. Your cost per token might be higher, but you want the result as quickly as possible. Or if you're running locally, so you might be running a batch size one just because you're the only person using it. It turns out that the GPU has extra execution capabilities that are just lying there unused.

50:03The bulk of the work when you're running in these scenarios is actually fetching the weights from memory. And then you push the token past those weights and then you fetch more weights from memory. But it turns out if you push two tokens or even five tokens through those same weights, it would cost basically the same amount of time because the expensive thing is not doing the math to push the token through the weights. The expensive thing is just reading all of those weights from memory, all those parameters they have to come in. And so the idea with multi-token prediction is to take advantage of this by having the model predict multiple tokens at once.

50:40Let's say that the model predicts five tokens. We know the first token is correct. The next four tokens may or may not be correct. So then what we do is on the next pass, we take those four tokens and we stick them into the model and then run it through. And at the end, we check, you know, the model then predicts another set of tokens, right? Then we check where the extra tokens we predict last time correct. If so, then we just accept them and then we get like a 4x speed up. And if they were incorrect, then we only accept the ones that were correct and then proceed from there. So the benefit of this is it doesn't degrade accuracy at all because you're using the model to double check.

51:25So all this speculation is going to get checked during the next token that you run through the model. So it doesn't degrade your accuracy at all to turn on multi-token prediction, but it can give you a speed up and it's probabilistic depending on the acceptance rate of your predictor. So if your predictor is more accurate, the acceptance rate goes higher, you get a higher speed up. So with our recent Neotron models, we're pretty proud of our acceptance rates, but we're always trying to make them better, always trying to improve that acceptance rate. This is a really good example of accelerated computing.

52:01With multi-token prediction, the speed that you get is a function of the accuracy of your model. The more accurate your model is, the faster the inference is, the cheaper the inference is, the more accurate it is. That's not usually how it works, but in this case, that's how it works. And what that implies is that if we're trying as NVIDIA, as a company, to provide meaningful acceleration to the world's most important computational workloads. This has to be an important part of how we think about it. You know, if there's a 3x cost reduction or speed improvement for inference, which is the most important computational workload of 2026, if that's on the table, and it depends on the accuracy of the multi-token prediction network, then that's something that NVIDIA needs to understand very deeply, because it's going to affect our business directly.

52:47Fascinating. To continue on the tour, multi-teacher distillation. We talked about distillation a little bit up front. What does that mean in the context of Nemotron 3? So with Nemotron 3 Ultra, we did post-training using something called multi-domain on-policy distillation. And what that entails is that, you know, we have many different aspects of the model we want to improve. For example, For example, science understanding is different from math theorem proving, which is different from coding, which is different from agent harness interactions, right? With NemoTron 3, I think we had about 10 or 15 of these teachers.

53:32So the idea is that you take these teacher models and you push them as far as you can go on some specific domain, so you just don't worry about making good at everything, just make it really, really smart at this one domain. Then you have a collection of these models and you want to create one model that learns to be good at everything. And we do that using a specific reinforcement learning technique that a lot of labs these days use called MOPD. And the good thing about this is that because the teachers are supervising, they can give really dense rewards to the student model. Basically, every token is getting supervised.

54:08And so the student can learn really quickly and then become, you know, almost as good as all of the teachers at all of the things. So one benefit of this is that it really helps the team work together better. You know, if you don't have a technique like this and you have, let's say, 500 people working to try to make a model better and one team's like, well, I'm trying to make it better at this thing. And then another team's like, I'm trying to make it better at that thing. There can be a tug of war where it's like, well, who wins? You know, and if you have to make a choice like, oh, I'm going to make them, I'm going to choose to prioritize this one over that one.

54:45Then you make the other team feel like their work doesn't matter. You know, it's just really hard. One of the challenges of building AI in 2026 is that you have to figure out how to get the people to work together, even though you're only building one thing at the end of the day. And so this particular technology has been really instrumental in helping more people work together to make Nemotron stronger. Fascinating. So it's just as much a technology question as a human organization question. Exactly. Okay, fantastic. Let's put a pin in this and get back to this in a second because it's a fascinating topic.

55:16In terms of the post-training that you just alluded to, one of the exciting things that you all did in the context of Nemotron is also to publish the data, the training data. Does that include per industry data for specific reinforcement learning tasks? Yes. That's the beauty of a conversation like this today, where you guys can actually talk about those things. So where does one get the data from for post-training, reinforcement learning focused efforts? Obviously, one of the key questions in the world today is that LLMs or AI systems have become great at coding and great at math. The next big question is, can they become great at law and consulting and then all sorts of different domains?

56:12And part of the black box of closed models is how people go about doing all of this, where do they get the data from? To the extent that you can talk about all of this, I'd be very curious about how you guys have gone about it. It's not an easy question to answer because it is quite complex, but I would say we rely on a number of things. One is that we do purchase data from companies that are building data sets that you can purchase. And to the extent that we have the rights to redistribute or to open up that data, we do as part of our NemoTron data effort. With NemoTron, we are trying to be maximally open with the data that we release because our goal is to support the ecosystem.

57:01Our goal is not to be the only model out there. And we love it when we hear of other models around the industry that are using our data sets to make their AI stronger, because that means we're succeeding in our job to keep the ecosystem thriving and growing. Now, we also are big believers in synthetic data generation. we use an enormous amount of compute, running language models on our own systems to create synthetic data that then helps our models be better at solving problems in specific domains. And we release a lot of that data as well. Now, it's, of course, not very straightforward to do this.

57:44Like, you know, AI is always garbage in, garbage out. So you have to work really hard to make sure that any synthetic data that you create is actually adding value that's actually helping the model generalize and solve problems more intelligently. But those are the primary ways that we go about building our data sets. Since we're talking about post-training and RL in different domains, just so curious to get your thoughts on where we go from here in terms of generalization. So just to build on what I was saying a second ago, like the industry seems to be marching from coding and math, which are domains with verifiable rewards to different industries.

58:24Do you think that this is where things are going and that the AI industry as a whole is going to be able to cover those next few domains as efficiently as coding or math? Coding is really special because it's a very intellectual exercise that created a lot of economic value, which then meant that we had an enormous amount of tokens that we could learn from, as well as tooling that allows us to verify whether our models are actually solving problems. So coding is always gonna have a special place in our heart and something that I think AI is gonna continue to get much better at because we have this special relationship with it.

59:13With regards to other domains, I think what I'm excited about about has to do with significantly more diverse environments for AI to learn in during reinforcement learning. I believe that reinforcement learning is such a general form of teaching an AI how to solve problems. We're just getting started at figuring out how to apply that. And I think as As our environments get more sophisticated, the AI then learns more understanding of the problems that it's trying to solve as well as the implications of the actions that it can take. Then it becomes much better at actually solving those problems.

1:00:02When I look at the environments that we're using today, they're still fairly simple, all things considered. And I think that's going to become significantly more complex and diverse over the next few years. All right. So you mentioned making 500 people work together. And I said that we would get back to it because it's so interesting. So just taking a step back, tell us about the research organization at NVIDIA. How is it structured? How does it all work? Well, NVIDIA is not structured according to an org chart. We have one, but it's not actually the best way of understanding how we work. My team, for example, is not part of the official NVIDIA research team.

1:00:48My team is actually part of the organization that builds the GPU. And my team is not the only team building Nemotron. There's probably 10 teams around the company that have significant involvement in building Nemotron. in different parts of the company, in enterprise software, in our AI software division, the part of NVIDIA that actually designs the GPU also significantly is involved in building Nemotron. So there's so many different teams that have to work together. We always like to say that the mission is the boss rather than the organization. But what that implies is that people have to figure out how to work together, which is challenging in the sense that humans are naturally tribal creatures.

1:01:43And it's not natural for us to be friendly with people we don't know very well or trust co-workers that we don't have success working with in the past. And, you know, actually the name NemoTron reflects that. We had the Nemo team, which was building software for AI, and the Megatron team, which was building primarily focused on systems research for building large language models. And, you know, we decided to work together and then start calling our projects Nemo Tron reflecting, you know, sort of the collaboration between these teams. Since then, Nemo Tron has dramatically expanded. There's so many more teams that are part of the effort.

1:02:28And it's really important that we have structured it in this open way inside of NVIDIA. You know, we are inviting volunteers from around the company to come help build NVIDIA's AI. We think it's very important to the future of the company. And, you know, as that vision continues to develop, more and more people want to join. That's fantastic. We're really excited about that. And it means that we then have to figure out how to organize the work so that everybody has a chance to contribute and feel heard and feel like their ideas are, you know, fairly evaluated on the path towards impact. We have a formal process for doing that.

1:03:09We have an internal website where people share ideas and then those ideas are assigned to one of 25 different leads that are, you know, over various parts of building Nemotron. They interact with those ideas. Some of those ideas get further developed. Some of those ideas get deferred until, you know, the next time we go around building a new model. But we're trying to build Nemotron in an open and inclusive way so that we can really come together as a company to build it. I think organizations that figure out how to collaborate to build AI succeed. Organizations that struggle with control over who owns the AI tend to waste a lot of effort.

1:03:54And so NVIDIA's success and Nematron's success, I think, is directly proportional to our ability to collaborate. It's something that I care deeply about. Fantastic. But you mentioned earlier that despite the fact that you work at the number one undisputed leader in GPUs, you all as a research organization don't have all the GPUs that you would want in the world. So like how does the allocation of GPUs and computes happen? Is that based on how promising an idea is or early success? Do you give GPUs, withdraw GPUs based on success? It's a really complicated question and it's obviously a difficult problem for everyone in the industry to figure out how to allocate their compute.

1:04:44Inside NemoTron, we have a budget for NemoTron and inside NemoTron we allocate compute based on what we think the needs of the project are. We have a hierarchy, so we have a set of programs and inside of each program we have have a set of projects and each of them put forward their requests. And then, you know, we have a two-week cycle where we review requests and we review the budget and then we make decisions in kind of a hierarchical way. And then, you know, compute gets decided that way. Now, having said that, this is something that I think we can still do better at. It's hard when we're making decisions about compute allocation because every researcher is convinced that their idea could change the world if it just got a thousand times more GPUs attached to it, right?

1:05:39And they might be right. It might actually be true. And yet we're running at the limit. We don't have a thousand X more GPUs for every idea that we have. We have to operate within the limits that we have. And so it is a challenging process. we try to incorporate as many people's perspectives into that as possible so that it's as much as possible a shared sense of understanding, maybe not agreement. So there may be times when one project feels like it really deserved more GPUs because the impact of that would have been so high, but it didn't get it. We hope in that circumstance that they have an understanding of why some other project did get more GPUs and why that was considered more of a priority during this particular allocation round for the company so that people can at least understand that there's a reason for the allocations that we have.

1:06:35Having said that, this process is always improving. There's always more work to be done to make this more transparent and more fair. And then, of course, my number one is just to get more GPUs so that we can also fund more things because I would like to do that too. How do you balance useful research with great exploratory research? My belief is that research needs to be bootstrapped. Research is a chicken and egg problem. So it is always the case that every researcher believes if I just had a lot more resources my idea would change the world. Actually, it's important that researchers feel that way, because if you didn't feel that way, you wouldn't have the conviction that's required to go do something crazy and new, right?

1:07:21So you have to believe. And so of course, you start with that belief. But then how do you translate that belief into something that other people can understand, right? That other people are willing to invest in? This is what I call the chicken and egg problem, right? Because once your research idea is obviously good and impactful, it's easy to get resources, but how do you get it to be obviously good and impactful without those resources, right? So the way you solve chicken and egg problems is by bootstrapping. This is an iterative problem solving approach where you do something small, you get some sort of signal about this is a good idea, and you tell people about that.

1:08:00And then you ask for just a little bit more. And if people saw like, oh yeah, that experiment turned out pretty well. That's pretty intriguing. We should probably do a little bit more there. Then you're on track, right? And that over time, iterate a lot, iterate quickly, iterate many times. You can bootstrap to finding significant resources for your idea and also usually attracting more people to come along with it on the way because they have a chance to see that this idea is going to change the world and then they want to be part of it. Is that how the moonshots at NVIDIA got started as well over the years, whether that's in AI or otherwise?

1:08:41So it was bottoms up, somebody coming up with a good idea versus Jensen saying this is what we need to do? Well, you know, Jensen has lots of good ideas, too. And so the company is very responsive to his ideas. And that's that's important as well. But Jensen very explicitly says all the time, this is a company of volunteers. You know, each of us is here because we choose to. We could we could be doing something else in our lives, but we choose to be here. And so, you know, we tend to make decisions, especially for early stage research, it tends to be very bottoms up because, you know, it's sort of an invitation, like bring your best ideas.

1:09:23Let's figure out, you know, what are all of our best ideas and then we'll take a step from there. Now, do we sometimes have top down ideas that are important for the company strategy? Of course, you know, of course. um uh nvfp4 pre-training is one of those you know so we decided as leadership of the company we're going to really invest in in fp4 hardware um now it's time to go invent some optimization algorithms that succeed in using it and um so uh so we told the team we didn't say to the team you have to work on nvfp4 pre-training what we said is there's an opportunity we're making a big investment and if we can figure this out it will be significant for our company and then we We let the people who are interested in that work on that.

1:10:09And as a result, we succeeded. So it is a balance of bottoms up and tops down. But it always has this bootstrapping feeling. Even with something like NVFP4, where there's a significant strategic top-down component, the actual technical solution, which is very intricate and complex and has a lot of moving parts, that came from the researchers themselves. And, you know, that's my belief is that research always comes from the researchers themselves. You can't tell research exactly how to go solve a problem because then it wouldn't be research. It would be engineering. But in a world of AI where the most important problems we have to solve all have this research component, there needs to be freedom for researchers to innovate if we're going to make progress.

1:10:53listening to everything you're saying i'm struck by how entrepreneurial the culture at nvidia still seems to be so like i'm so very large companies i'm i'm sure there's all sorts of politics and you mentioned the tribal instincts like i'm sure all of this is happening but um especially given you know how long the company's been around a phenomenal success the fact that people have been making a lot of money internally it still seems to be very entrepreneurial bottoms up driven and maybe meritocratic. Is that the right takeaway? Yeah, I mean, one thing that's very unusual about NVIDIA is the tenure of its leadership.

1:11:33Jensen Huang has been running the company for 33 years, but he's not alone. There are a lot of other very senior leaders in the company who have been there for three decades or longer, including my boss. And these people remember what it feels like to work at a very small NVIDIA. And they know what it feels like to work at a very large NVIDIA. They have a shared sense of ownership for the company. NVIDIA is a place we often say, no one fails alone. And the point of that, that's just a statement of fact, right? You work at a company, it's one company. You all succeed together, you all fail together.

1:12:13You work in accelerated computing. Accelerated computing is the composition of thousands of technologies. If any of them fail to deliver acceleration, the value is destroyed. It doesn't matter whether the chip is great if the compiler sucks. At the end of the day, the thing that you're selling is time and capability to researchers that are trying to build the future of AI. And if they don't get that, it doesn't matter whether it was the, you know, the transistor or the math unit or the compiler or the library or the networking or anything else along the way that failed to live up to its expectations.

1:12:46the whole thing in composition fails, the whole value is destroyed. And so we have a deep understanding of that culturally at NVIDIA. And it is something that motivates the way that we work together. Maybe to close the conversation, I'd love to zoom out, get your take from the perspective of somebody who's like as deep into all of this as it gets about where things may be going. So like who knows in a few years, but I don't know, in the next year or two, maybe there's some visibility. I read somewhere that you're not necessarily a big singularity kind of person. Is that fair? True. And why is that?

1:13:29Well, I think that intelligence is just so incredibly multifaceted. You know, I always think about this question, like, if a company were to be looking for its next CEO, would it find the next CEO by looking for somebody who won the International Math Olympiad? Probably not, right? Even though it's incredible for people, I could never even compete in any way at the International Math Olympiad. Those people are amazing, right? They have just incredible brilliance. That's not the right kind of brilliance to run a company. If we look, for example, at other aspects of our culture that are really important.

1:14:11For example, musicians. What kind of intelligence does it take to become a hit musician? Don't assume that it's all luck. It's not. These people are working hard and they're very smart in ways that I might not understand with my PhD, right? I might not have that kind of intelligence. And so when I think about intelligence, I think it's just so multifaceted and so contextual. You know, it really depends on the situation. It's not just about raw intelligence. Raw intelligence is kind of like the horsepower of an engine, but an engine running without wheels doesn't go anywhere, right? So intelligence, the impact of intelligence has a lot to do with the context that the intelligence has put in the harness, the platform.

1:14:56And so when I think about that, I think, you know, So the singularity is, although it's an attractive idea, I think that it's really a wrongheaded idea because it doesn't really take into account these other factors. So I believe that artificial intelligence is going to continue to develop at a rapid pace. It's going to unlock significant capabilities for people in every aspect of our world economy, people doing every kind of work. I'm very excited about the opportunities that it's going to bring. I am also a little bit concerned with how we're going to manage the transition. So I do think that transitions are hard for humans in general, like we're conservative generally.

1:15:44And, you know, there is going to be a lot of change. This is a profound change in the way that we think, in the way that we work, the way that we learn. um uh ultimately i have faith in our ability as humans to figure it out um you know we've done it in the past this is how this is who we are um we we build tools uh we build external organs that help us solve problems you know we we have an external stomach we call it kitchen it creates enormous value for us we can eat things that we couldn't eat without a kitchen right now we're creating an external brain. You know, the implications of the external stomach were pretty profound for us as a species.

1:16:23They led to agriculture, which led to organized societies, the way our cities are built. So we think about what is the implications of an external brain? Pretty profound. Nobody actually really knows. But what I do believe in is the power of humanity to solve problems and to learn and to incorporate new technologies in ways that benefit us. I also believe that the problems we face as a planet all require more intelligence, every single one of them, whether that's inequality or climate change or any of the other structural problems that I think are very worrisome that we face. the solutions to those are going to require invention and intelligence.

1:17:07And what that means for me is that the only kinds of tools that we can really create moving forward are going to be AI. Because the problems that we face are all about intelligence. And regardless of the technological approach to solving those problems, the solutions will always be called AI. and so that makes me hopeful for the future but also you know somewhat you know respectful of the challenge that it is going to bring to us as we we try to figure out how to live in a new way with this new external brain but I believe in our our ability to to learn and to change and I think ultimately this is going to make our lives better.

1:17:50Do you guys feel the AI backlash that seems to be forming internally? Is that something that you all perceive, think about? And if so, do you think it's a communication problem that our industry may have, you know, in particular, given what you just said about all the obvious potential of AI? You know, I'm always worried about the way that the public thinks about technology and interacts with it. It matters a lot. And it is definitely the case that societies that want technological advancement have more technological advancement than societies that don't want change. So I think it is actually important to think about it.

1:18:30One thing that's interesting about AI is that I believe it tends to be much more accepted when it is part of everyday life. And at that point, people stop thinking about it as AI. It's just, oh, this is the tool that I use. Like, do you care whether it's AI that's helping you route your car when you ask the map application to help you drive somewhere? Like, I mean, it is. There is actually sophisticated AI that's going into that. But you're not really thinking about that, right? You're just using a tool. And so I feel like people's acceptance of AI, you know, comes with experience, right? The more experience we have working with it, the more we learn how to work with it productively.

1:19:15I think the more comfortable we become with it. Great, Brian. So it's been a fascinating conversation. Maybe as a very last question to make sure we cover it, I want to make sure that we talk about safety. What is the state of safety currently? And where does open source and closed source sort of fit in the safety conversation today? Safety is on everybody's minds right now. You know, watching the Fable release and the way that the government interacted with that, I think, is a consequence of concerns about safety, about these models. You know, they get stronger and stronger and then they could be misused.

1:20:02And, you know, there's different approaches to thinking about safety and trying to define safety. I have maybe a slightly unorthodox opinion about this, which is that I think open technologies are generally safer because there's more sunlight. You know, when more people are thinking about the safety of a technology and evaluating it and then contributing to making it safer, I think that's inherently safer than having a small group of people being in charge of safety for everyone else. I also think with artificial intelligence, because it is really about ideas, it's really about exploring ideas in different ways, that diversity is more safe than monoculture.

1:20:51And what that means is that there's going to be different beliefs. Like diversity isn't just about like the easy stuff. Diversity is about the hard stuff. Like when people have deeply felt disagreements, they really, really totally disagree with each other. Making it possible for people to explore their ideas in a diverse way, I think, is more safe than trying to create a walled garden where, you know, certain ideas are considered safe and certain ideas are considered unsafe. And, you know, this is controversial in today's AI environment, which I think is interesting because we've had hundreds of years of tradition that speak directly to this.

1:21:36You know, in in the United States, for example, we have laws about freedom of conscious conscience and freedom of speech. And, you know, it's not because we didn't consider for thousands of years, would it have been safer if we didn't have those? Right. We tried that. We tried actually having a monoculture about like these ideas are safe to talk about. these ideas are safe to believe. And we found that to be much less safe than a pluralism where we officially don't take a position about what ideas are safe. We actually found that as much safer as a society to support diversity than it is to try to keep everybody safe top down.

1:22:19And so I believe that open technologies for AI are inherently the safest way of building AI. All right. Love it. Controversial take to close the conversation. Brian, it's been fabulous. Thank you so much. We appreciate your spending time with us today. Thanks for inviting me. Hi, it's Matt Turk again. Thanks for listening to this episode of the Matt Podcast. If you enjoyed it, we'd be very grateful if you would consider subscribing if you haven't already or leaving a positive review or comment on whichever platform you're watching this or listening to this episode from. This really helps us build a podcast and get great guests.

1:22:55Thanks and see you on the next episode.