America’s national labs have always been at the forefront of scientific developments and breakthroughs, so it should come as no surprise that artificial intelligence is an integral part of their scientific process and American national security.
Scripps News national security correspondent Liz Landers caught up with Dr. Kimberly Budil, director of Lawrence Livermore National Lab, at the AI+ Expo in Washington, D.C.
Landers visited the lab last spring and spoke with Budil about how the lab ensures the effectiveness of the U.S. nuclear stockpile.
Below is a transcript of their conversation at the expo:
Liz Landers: Dr. Budil, your lab has reached ignition, we understand again. Tell me about that experiment, what happened, how much energy was output, and how did it happen?
Dr. Kimberly Budil: So, we just had our eighth experiment that exceeded the fusion ignition threshold. This time, we achieved 8.6 million joules of fusion energy for just over two million joules of laser energy input. So that's a gain of four. So, a really significant increase from where we started. So, our first experiment was a gain of about one and a half. So steadily improving. It's a combination of a few things. Part of it is the design of the target, the little capsule that holds the fusion fuel, the deuterium and tritium. We made some small changes to that. And part of it is how we operate the laser, so how we precisely deliver all that energy to get exactly the right shape. As always, we're learning with each experiment. So, there is always just a soupcon of luck that gets added, a little bit of alchemy at the end where all these things, which we can't necessarily control perfectly, come together in just the right way. So, it was very exciting for the team.
Liz Landers: You mentioned this is now the eighth time that your lab has achieved this. How much closer together are these experiments happening? Because I know one of the things we talked about was you have to recreate the target and everything because it blows up during the experiment. So, the fact that these are happening more regularly now, what does that signal?
Dr. Kimberly Budil: So that signals that we're really starting to get control of the process. So, we've learned more about what's required. So, how precise do the targets have to be? How should the laser energy be delivered? What's the right material composition to give us the best possibility for success? And so, we can do only a few of these very highest energy shots per year because they do cause damage in the laser system, so we have to meter those out carefully. So, we've also learned how to adapt the design to slightly lower laser energy, and that helps us do more experiments. So, we're on a pretty good cadence of every three to four weeks having another ignition potential shot. Still, each one is a learning experience, so they're not all successful, but every one of them makes our models a little better and then makes the chances of success the next time a little higher.
Liz Landers: How do these experiments contribute to national security here in the United States?
Dr. Kimberly Budil: So, one of the key questions that's relevant to stewarding our nation's nuclear deterrent is understanding how the weapons operate in the very extreme conditions that are generated in a nuclear detonation. Now we did many nuclear tests during the testing era, but that ended in 1992. So, the way we study those conditions in detail today is through very precise experiments on facilities like the National Ignition Facility (NIF). Where we can create these very intense fusion ignition conditions and then use lots of very exquisite diagnostics to really peer inside and understand how the physics is changing. So we use that then to inform our big computer models that can tell us something about how the stockpile is changing. We also use it to do things like qualify new materials that have to exist in these very harsh environments. We can do experiments where we subject materials directly to conditions that they might experience and make decisions about replacements. So, for example, we've made replacements for materials that are easily manufacturable or that are more environmentally damaging than we would like to work with today, really from a worker safety perspective. And so it's allowed us to make different design choices, qualify new materials, and really underwrite our confidence in the safety, security, and effectiveness of the stockpile.
Liz Landers: Are there other nations that have reached ignition?
Dr. Kimberly Budil: No, no one else has done — I love saying that. We're number one, and only at this point. Many people are seeking to do this. It's a very hard experiment. I think, you know, one reason why people pursue this science and technology is that it's also a great way to test your people, their skills, their ability to design complex systems and their judgment. You know, how do people take these very complicated models that we create and internalize all that data and then make informed judgments about it? So many nations are pursuing, but nobody currently is close as best we can tell.
Liz Landers: We're here at this AI Expo, and I know that you're part of a commission that is studying the scaling of fusion energy. And this commission, I was reading sort of the midway report of it, it frames this as almost a modern-day arms race, in particular with China, for really the United States' national security and global energy dominance. Why is fusion energy a national security imperative vis-à-vis China?
Dr. Kimberly Budil: So, China is pursuing many of the same technologies. There's been a lot of open-source reporting on their efforts to build large lasers and tokamak systems. They've been a big participant in the fusion community for a long time, so we know of their interest in this science and technology. You know, the U.S. spent many decades building a very significant science and technology lead in this area. We now have the critical technologies in hand that would enable a fusion energy future. There really is an imperative to make sure that we keep this industry here where we've made this massive investment. So that's one. Two, they pursue it for national security, for the same reasons we pursue these technologies for national security. So, having this lead really is an important signal to the world of our capabilities and our commitment to, for example, returning to nuclear testing. I think the last piece is that in order to really build out this fusion energy ecosystem, some nation will have to bring together the fundamental science, this science at scale that we're doing at the NIF, and the remaining science and engineering challenges that need to be solved, and create what will be a trillion dollar industry for the world. I mean, fusion has many attributes that make it much more attractive than fission from a national security standpoint, least of which you don't need special nuclear material for your reactor. So, this could be a world-changing type of energy, you know, clean baseload energy at scale without many of the safety and security considerations that go along with fission. So, while we're deploying fission reactors today because it really is a great energy source and it's today's technology, the U.S. could lead a transformation of the world energy system if we really lean in on fusion.
Liz Landers: One of--I think I heard this--I was walking up kind of late to your panel, but I think one of the questions, or there was a part of the conversation, about returning to explosive nuclear testing. Is that something that's being discussed right now amongst directors of national labs?
Dr. Kimberly Budil: So we have, for the 30 years of the Stockpile Stewardship Program, not seen any technical need to return to nuclear testing. We're obligated by law every year to do an annual assessment of the stockpile and weigh in on that question, and from my perspective, our science tools have gotten so good; we're actually making ourselves more robust against a need from a technical perspective to return to nuclear testing. So that for me is a real signal about our commitment to global stability. The U.S. is in a great position to continue down this path without a return to that way of operating.
Liz Landers: So you would not recommend a return to that?
Dr. Kimberly Budil: No.
Liz Landers: If an adversary like China were to start doing nuclear testing, or Russia, let's say, I mean, we know they're in the midst of a war that they're fighting with Ukraine, do you think that would change the calculus?
Dr. Kimberly Budil: So, there are certainly political questions that would have to be answered, and that's always a consideration in this realm. I mean, it's very possible we would get directed to perform a nuclear test for a reason that's not being driven by us or by technological considerations. And so, we also maintain a test readiness posture so that if directed we could execute a nuclear test on a salient time scale.
Liz Landers: How fast?
Dr. Kimberly Budil: So typically, two-year time frame.
Liz Landers: Wow, that would take a while.
Dr. Kimberly Budil: With many caveats.
Liz Landers: Yeah.
Dr. Kimberly Budil: I mean, it's quite amazing to think about what it took to do a nuclear test and to diagnose something that is such an extreme environment. But the U.S. did many, many nuclear tests, so we're in a great position now, and actually it's an advantage to us to be in this place where there's a taboo against testing. I think that really is a good thing for the world.
Liz Landers: The panel that you were on earlier, this expo that we're at, is talking about AI. How does artificial intelligence drive the mission of your lab?
Dr. Kimberly Budil: So, AI is starting to inform almost everything that we do. So of course, there are a whole host of operational and business considerations for how to use AI to make our system smarter, our infrastructure better, to make everything much more efficient. But on the science front, I think there are really three areas. One is how we use data. So, AI is data hungry, we have lots and lots of data. And rather than taking all that data, having humans sift through it, and maybe distill out a few numbers, we can now use that data at scale and derive insights and improve our understanding of physics much faster than we could before. It's really remarkable. The second is how we do experimental science. So you can use AI to control your experimental system, and AI can make decisions about what the next experiment should be because it can analyze the data on the fly. So the rate at which we learn is going to change. We'll be able to gather much more data and again, virtuous cycle back to the first point. But the third area I'm really excited about, particularly for national security, is the way we develop technology. So today it's sort of a serial process. You have a problem or a challenge, you think about a technology solution, you design it, you engineer it, you make it. With tools like additive manufacturing, AI tools, design optimization, on-machine control, and other AI in the loop computing, you can start that cycle at any point. So, you can look at your manufacturing technology with additive manufacturing, and use that to inform the design space, and maybe come up with technology solutions you didn't even know existed. So this very virtuous cycle of how we can think about technology, engineering, and production, I think it's really going to change dramatically in the next decade.
Liz Landers: Would your lab be able to reach fusion ignition without artificial intelligence?
Dr. Kimberly Budil: Not in the way that we see we need to make things like fusion energy or our national security missions work. So, we need be able to move much more quickly to ingest all the data that we're taking and improve our physics models. And so, we would probably get to the place we want to be, just on a time scale that would make it much harder to be relevant to what's going to be happening in the fusion industry writ large.
Liz Landers: You made a comment on stage that sort of struck me. You said that basically we have been doing AI for a long, long time at our labs. El Capitan is the largest computer in the world. And you said that you all will be a bellwether. What do you mean by that?
Dr. Kimberly Budil: That means that as we explore what these technologies are capable of doing, we understand the many national security missions and challenges they could intersect with. So I have experts in biosecurity, chemical weapons, nuclear weapons, in space, cyber, directed energy, in all these technology areas. And they can help, as these tools develop, begin to understand how an adversary might put them to use. Rather than regulating up front, for me, that idea of being part of the process means we can get the benefits of these tools to the fullest extent and only put in guardrails where they're really needed, where we see that there's an outcome that really would be bad for the public at large, then we can work with the companies to help them rein in their models in those areas. So, I think having people who think national security day in and day out means we tend to, I don't know, if you tell me about a new technology, in my mind, I'm going to be thinking about how it could go wrong. And I think that's a useful tool in this environment where things are moving so fast.
Liz Landes: You mentioned your physical proximity to Silicon Valley, to OpenAI, Meta, Google, these big companies that are driving artificial intelligence right now. What is your working relationship like with some of these companies?
Dr. Kimberly Budil: So, we have good working partnerships with most of them. That means they sometimes come to our researchers with their newest models and give us early access so we can see what they're capable of. They're interested in our feedback. They'd like access to a lot of our data. All this scientific data is something they don't really have available to them, and they don't have the expertise to really use it. So that's a very virtuous part of this partnership. We have some unique computer tools. So, our big computer is based on AMD chips. So, NVIDIA dominates the market for AI and inference. We're really working hard to help mature the software stack so that AMD will be a peer competitor in that. And of course, our computer is a beast. 45,000 GPUs. So, this is a really mighty resource for AI. So, it's a really interesting win-win relationship. We think their tools and what they've developed is just incredibly interesting and cool and they think the science we do with them is incredibly interesting and cool, and that bringing those two together, we can help advance their tools in ways they couldn't do--we couldn't without each other.
Liz Landers: That is a fascinating perspective. Does the presence of AI in your laboratories present any national security risks?
Dr. Kimberly Budil: So, the presence of any new tool always presents some risks. I would say that the way we think about our cyber environments in particular, I don't have many additional worries on the national security front from that perspective. We protect our networks very actively and aggressively with very leading-edge tools, we work with all other elements of the government to make sure we know what we're doing in that arena and we protect access to our spaces, to our people, to our technology, to our ideas. So, I feel like we're very comfortable operating in this boundary zone between open science and our national security missions. So, we have to be careful, but in some respects, there are advantages for the companies to working in this space because we're working on special networks behind firewalls where they could be less worried actually about the security of their models and their proprietary information. So, it's an interesting dynamic.
Liz Landers: It sounds like it's really a win-win for the lab and also for these private companies to have partnerships together.
Dr. Kimberly Budil: I think so. So, Tom Mason from Los Alamos talked about this partnership we have with OpenAI on the Venado machine. That's the three national security labs working together to bring their latest greatest model into a classified environment where we can have full access to the model. Again, bring unique data which will give them information to advance the model and give them unique perspectives on how the model works in some science areas without any fear that proprietary information from their model is going to leak out of our system. So that's a really exciting new thing.
Liz Landers: Since the last time I talked to you, we've had a few more months to settle into the new administration. You recently had a visit from the Energy Secretary. How is the relationship with this new administration? Do you feel supported by Secretary Wright and do you have all the financing you need right now?
Dr. Kimberly Budil: So certainly for the national security labs, the Secretary is incredibly dedicated to our missions. He is a tech guy at heart, an engineer, and so he's really excited about the labs and the science and technology that we do. He has a long-ago connection to Livermore, so we feel like, we hope we have some special place in his heart. He won an R&D 100 Award earlier in his career with one of our teams. So that part is great. He's learning a lot about what we do and what we're capable of. He's very excited about AI, he's very excited about fusion, and we're just getting the early indications of what's going to happen in the budget process. So, it's very hard to know where we'll be six months from now. But the nuclear security missions have been very strongly supported by this administration, and I'm pretty confident that NNSA (National Nuclear Security Administration), our part of the DOE (Department of Energy), is going to be okay.
Liz Landers: Do you think that your lab may experience any kind of brain drain from some of the changes that were seeing to immigration and visas? I'm specifically asking, I guess, because of the Chinese students being targeted.
Dr. Kimberly Budil: Yeah, so we have a long history of working with international scientists from all over the world, many of whom come to the U.S., stay here for many years, become U. S. citizens and then participate in our national security missions. They're incredibly grateful for the opportunities they've had in the United States and really dedicated to their work. So, I think that is really important to acknowledge that this is not something that we're unfamiliar with and we have always understood the risks of bringing in people from other countries into our environment. We really work to manage our environment carefully. So, there are areas of science where we participate freely with other countries, China included. There are other areas where we don't. And so by managing access to our spaces, to our people, to our technology, to our research, we feel we can manage this environment. It has been a difficult transition. We don't have a very large population of Chinese researchers at our lab, but it was not zero. So it's been an adjustment. But I think we're having an ongoing conversation to help people understand, you know, we live this national security mission day in and day out, and we understand very well the risks and always seek to strike the right balance. I want the U.S. to get the benefit of the best brains in the world, and so we really do need to find a way to manage through this.
