OpenAI Forum
+00:00 GMT
Sign in or Join the community to continue

Fusion Energy: The End of Fossil Fuels

Posted Nov 29, 2023 | Views 15.3K
# Innovation
# STEM
Share
speakers
avatar
Sam Altman
Co-Founder & CEO @ OpenAI

Sam Altman is the co-founder and CEO of OpenAI, the AI research and deployment company behind ChatGPT and DALL·E. Sam was president of the early-stage startup accelerator Y Combinator from 2014 to 2019. In 2015, Sam co-founded OpenAI as a nonprofit research lab with the mission to build general-purpose artificial intelligence that benefits all humanity. The company remains governed by the nonprofit and its original charter today.

+ Read More
avatar
David Kirtley
Founder & CEO @ Helion Energy

Dr. David Kirtley is founder and CEO at Helion, a privately funded fusion electricity company creating the world’s first fusion generator. He is passionate about inventing and leading teams to develop disruptive technologies to improve access to energy, reduce carbon emissions, and provide a better future for the following generations. Dr. Kirtley is a fellow of the National Science Foundation, NASA, and the Department of Defense. He is a leading authority on advanced fuel fusion generators, plasma-based in-space propulsion systems, and space reentry architectures. Dr. Kirtley has graduate degrees in Nuclear, Plasma, and Aerospace Engineering from the University of Michigan. He has over 100 publications, 60+ patents and has been pursuing advanced fusion systems since the late 1990s.

+ Read More
SUMMARY

Earlier this year, Sam Altman, CEO and Co-Founder of OpenAI and David Kirtley, CEO and Founder of Helion convened at the OpenAI office among a small group of OpenAI Forum members to discuss the future of energy. This is the recording of their discussion.

+ Read More
TRANSCRIPT

As usual, I just want to remind everyone that this is going to be recorded. We record these events so that the forum members that aren't here live, they can watch them on demand in the community.

I'm Natalie Cohn, OpenAI Forum Community Manager, and I want to start our talk today by reminding us all of OpenAI's mission. OpenAI's mission is to ensure that artificial general intelligence, AGI, by which we mean highly autonomous systems that outperform humans at most economically valuable work, benefits all of humanity.

It's hard to believe that we launched the OpenAI Forum Community just four months ago, seeding the interdisciplinary program with expert practitioners and some of the most accomplished researchers across many different domains. One of the areas we chose to focus on for the inaugural iteration of this community was environmental science, thinking about the surface area in a very broad terms and welcoming researchers, professors, founders, and technologists from an array of scientific backgrounds. The talk tonight was intended to resonate with this group, however, I suspect the topic of renewable clean electricity is top of mind for many of us, despite our fields of expertise.

Now I'd like to introduce our speakers and the reason we are here this evening. Our honored guest, Dr. David Kirtley, is founder and CEO of Helion, a privately funded fusion electricity company, creating the world's first fusion generator. He is passionate about inventing and leading teams to develop disruptive technologies to improve access to energy, reduce carbon emissions, and provide a better future for the following generations.

Dr. Kirtley is a fellow of the National Science Foundation, NASA, and the Department of Defense. He is a leading authority on advanced fuel fusion generators, plasma-based in space propulsion systems, and space re-entry architectures. Dr. Kirtley has graduate degrees in nuclear, plasma, and aerospace engineering from the University of Michigan. He has over 100 publications, more than 60 patents, and has been pursuing advanced fusion systems since the late 1990s. Thank you for joining us, Dr. Kirtley.

Thank you for having me.

Sam Altman is the co-founder and CEO of Oven AI. Sam was president of the early stage startup Accelerator Y Combinator from 2014 to 2019. In 2015, Sam co-founded Oven AI as a non-profit research lab with a mission to build general purpose artificial intelligence that benefits all of humanity. The company remains governed by the non-profit and its original charter today.

Without further delay, please welcome Dr. David Kirtley and Sam Altman for our fireside chat.

Thank you, and especially thank you to David for flying down here to do this. Look, this is, I think, one of the most important people in the world and one of the most important companies, so very excited to talk about all the great things about fusion. But I wanted to start off back when David was still a skeptic of fusion, and tell us about why and what changed.

Well, thank you for having me. I'm happy to talk about fusion, fusion energy, and what we believe the future of the world could look like. So I didn't start off as a skeptic. I started off as a believer. I went out and said, I want to change the world. I'm sure everybody here said something when they were young. And I thought the way to do that is clean, low-cost electricity. It's the way the sun works, the way the universe works, where most of the matter in the universe comes from. Let's do fusion. That sounds like a good idea. Went to school, spent a lot of years doing it. Became a world expert in antimatter, which is pretty fun stuff. And then figured out that the path that we were taking didn't have a commercial future, that it was tens of gigadollars to even test a theory, that I would retire my career never ever actually turning a wrench to build a fusion generator. And I was like, I can't do this. Forget it. I'm going to go work for the Department of Defense and NASA and take my fusion and plasma physics and antimatter, though they didn't buy the antimatter part, to those audiences and do something else for humanity.

Met my team in 2008 up in Washington and looked at some of the old ideas, some of the really earliest ideas of fusion, but with modern electronics and said, man, we might have a path here. Let's go try to build some stuff that probably won't work. I got some government grants, tried some things out, and some of those approaches worked. And one of them was a small system, back in the day we named them after beer, IPA was the most successful, that did fusion, made fusion reactions, got to thermonuclear fusion conditions, and did it for a million and a half bucks. And we were like, wow, you can do that and outperform all these billion dollar programs. Let's go figure out if we can wrap a business around this and go forward. And then figure out how to run a business, which turns out was a totally other thing that was also very hard. And since then we've built six fusion machines that do fusion, setting world records every time, and believe that we're now building our seventh generation, we are building our seventh generation machine, which we believe will make electricity from fusion for the first time. And so what changed was I measured the fusion reactions myself.

And why do you, like, this is quite interesting to me, like, you know, fusion had this moment where everyone was excited about it, then everyone was like, ah, it's, you know, billions of dollars in like decades, and it still might not work. And then one of the things I've always admired about you is you just do it and build it and see what happens. And also you do it like quickly and cost effectively. And this is a spirit that I think is very important to doing anything hard, but it's been totally missing from fusion and the world in general. Like, why and what have you done? How did you sort of build this different culture? Like, you all built, I think on 30 or $40 million, like the first sort of large scale fusion system, way outperforming systems that had cost billions or tens of billions.

It was a priori, it was in the beginning of the company, we said, our goal, it kind of comes back to that old skeptic of if I built something or designed a machine or did a simulation that then would take 40 years to make, who cares? Like, that can happen, but we weren't excited about that. I'm like, want to turn a wrench and bleed on the machine or whatever, like sweat and blood actually matters to me and building these things. And so every time we go to a decision point, left or right, we think about, okay, when is it going to get built on this path? What is it going to do? And you take the physics decisions, the engineering decisions, and it always comes back to that timeline of what can you build quickly? Okay, well, that one's too hard for now. It's going to require more money, more proof points. What can I do now? I'm going to go build a thing. I'm going to move on from beer and go to coffee, and I'm going to make tall and grande and Venti and Trenta, and each time prove something that's not electricity yet, but it's some key thing, whether it's a key piece of physics or a key piece of engineering or showing we could recover energy from the system for the first time, something like that, and just make those steps.

So the seventh generation system being built right now that'll run next year, what will that system do?

So the goal of Polaris, we ran out of Starbucks after Trenta, so went to Starz. The goal of Polaris is that it will, for the first time, take electricity, put it into a fusion reaction, and then extract electricity back out of that fusion reaction, and more electricity than you put into that reaction, and that's the goal, enough to power a light bulb for the first time from fusion. It won't be enough to power the grid yet. We still have work to do after that, but it's that first real proof point that you can do it.

So this is one of these huge if true statements, right? Next year, there will be net gain electricity from fusion. To run a light bulb, not the grid, and probably not very reliably, probably machine breaks, who knows what else. We're going to try to make it work. Of course. But this is an amazing thing. It reminds me of at OpenAI when we kind of had early versions of GPT working, and the rest of the world just didn't believe, and we're like, no, no, we can have this thing. This is a big deal. What's it like inside the company right now? What's the vibe?

Yeah, you go through these cycles where you're designing machines.

And then you're building it, and then you're testing it and you're taking it apart. And so we actually try to do all of those in parallel, as crazy as it is for hardware, but this is not unusual to probably many in this audience, that you are like designing, doing research on a system in the future, while designing the thing you're building, and then actually building hardware, and then like learning data from the last iteration of this, all at the same time.

And so yeah, we have the sixth generation machine, actually we're pulling down the racking that all the capacitors were in, like just today we're pulling down the racking, while we're testing the fuel injector, the plasma injector for the seventh generation system, and the full framework is being constructed now. And so there's all this sort of frenetic activity, all driven by what is the schedule, how can we get this built faster at every time, and then the decision points, and we're still making them actually, weirdly enough, of like, okay, that's a better connector, that's the connector we want to use, but it's three more months, let's use the shittier connector, and let's get it built now. And then go buy the big connector, the better connector, and three months later we can do an upgrade. So there's a lot of that, I think there's a lot of excitement around it right now.

We announced in March of this year, the first PPA from Fusion, Power Purchase Agreement, where Microsoft is committed to buying power from a power plant we'll build in 2028. And so there's a lot of excitement there, and a lot of frankly nervousness of like, okay, great, now where's the supply chain for all that stuff, great, where are we going to locate it? So we're going through those kinds of decision processes at the company now, and so there's interesting churn, but I think also a lot of excitement.

Before we get to all of that, can you talk about how, like very basic level, like how the Helion system works, and why you made choices you have, like why Deuterium-Helium 3 and not something else? Why the electricity recovery system?

So in Fusion, your goal, the same way the sun works, is you take a gas and you heat it, and you heat it so hot that the actual atoms combine, they actually slam into each other fast enough, at high enough velocity that they combine into a heavier element. That heavier element has something that's called a mass deficit, but it's hot. And it's a charged particle that now, you have this gas that's running it, we use units of KEV, which is on the order of 100 million degrees, but the thing that comes out of it is 10 billion degrees. And that's the key. Can you do that? But it requires hundreds of millions of degrees and thousands of atmospheres.

So what we do for our systems is we take a magnetic field, can't hold material at those temperatures, so we generate, we have an electric coil, we put electricity into it, confine a plasma, there's 100 million degree fuel, and then we squeeze it as fast as possible. Fusion happens, the pressure increases, and then as it expands back out, we extract that electricity back out.

Most people try to take that heat and boil water with it, run a steam turbine, because we know how to do that, it's dumb simple. But we actually think that dumb simple is also expensive. And so the analogy I like is regenerative braking in your electric car. You hit the gas, electricity goes from your energy storage, from your battery into motion, you hit the brakes, the energy that you didn't use goes back into the batteries, which that means the car can use that for the next gas cycle, you can go further with less batteries, the car can get lighter, guess what, a lighter car can also go further. And that whole analogy keeps, that bootstrapping analogy works for fusion too. And so we've been able to build systems that just do fusion, and they're a lot smaller than everybody else, because we can do that, we can recover that energy. So you accelerate two plasmas together, they merge, you squeeze, they push back, and then that just drives electrons on a wire. So there's some other stuff. So it does look like that old XKCD of like, insert magic here.

So the, I don't know why I did a magic wand. I did a magic wand for that one. No, the fundamental physics, this is Maxwell's equations, this is db dt, you have a changing magnetic field, you can, it generates an electric field, that electric field drives a current. And so the analogy that I use there, it's actually the same physics that's in the armature of a motor. In an electric motor, you run current through that motor, that actually induces electric field, and you use that to actually, in that case, generate a force that spins the wheel, spins the axle. For this, that force takes the fuel and actually squeezes it, actually compresses the fuel. And then it works, ideal gas law. Ideal gas law, you compress volume of a fuel, volume of a gas, it increases intensity and temperature until something happens. And for our case, we go up high enough pressure and high enough temperature that fusion happens.

How long does that compression and expansion cycle go for?

Yeah, there's all these crazy timescales in fusion that are sometimes hard to imagine. We puff in the gas, like inject the fuel in one millisecond, one thousandth of a second. But then we ionize it, we heat it up to 10 million degrees in six microseconds, six millionths of a second. And then we slam them together in 20 microseconds. And then we do that final squeeze for another fraction of a millisecond, so another thousandth of a second. And then it's all over. So the whole thing is two and a half milliseconds. And then you exhaust the fuel, which takes its sweet time to leave. And then you pump out any remaining unburned fuel, or unfused fuel, and then puff in the fresh fuel and repeat it again. We showed we could do those electronics once a second, and even faster than that sometimes. Though things start to get hot, so you've got to worry when you go much faster than once a second.

Could you talk about, if this all works, what do you ...

Well, first, before we get there, I'd talk a little bit about the company's goals for cost of energy out of such a system, and the scale that this could be manufactured at. So when we formed the company, and when we said, we're going to do fusion, there was never a case where we said, we want to make a design trade, where the cost doesn't allow you to deploy these, doesn't allow you to change the world. And so it's, again, the idea is that we have to build it quickly. And cost goes down as you build things quickly, to an extent, but that was always the goal. And so we focused on that, again, when we were designing these systems. So we look at them now, that we believe we could get to systems, just using the cost of the materials we buy today, that you can build them for less than a dollar a watt. And what that means is a million dollars a megawatt. For people that are not familiar, that's about the cost of natural gases right now. A traditional nuclear reactor is something like, I don't know, five times that cost. You maybe know that number, but maybe like five times that cost. And what that means is if your capital costs can be that low, then you can now generate electricity at a cent a kilowatt hour, or less. And a cent a kilowatt hour is wild. That now, I mean, everybody knows what they pay at home. Residential power is quite a bit cheaper than that. But that enables all kinds of cool stuff that we don't really think about with electricity.

Any thoughts about, any favorite ideas of what those could be for electricity? It's not cheap. You can't say desalination, because that's always what I use.

Oh, that was my starter one. The thing that I think about when I think about the number of a cent a kilowatt hour, what does that mean? Obviously, that's like in Washington, where it's like $0.12 a kilowatt hour. So you're like, OK, great. You'd save a lot of money. But what it actually does is, and I think everybody here, or many people here would probably agree with this, is that you can reduce the cost of access, or the cost of anything that you're supplying. It opens up whole new avenues. And so at a cent a kilowatt hour, I'm going to say a desalination is cute. Because after $0.02 a kilowatt hour, you can now get fresh water from seawater. And what you can actually do is some other cool stuff, where there's things like molybdenum in the water. You can start to extract that stuff, things that we fight wars over. You can now get out of all water. Really interesting things happen.

But I point to other big things, you know, you can actually start to pull carbon out of the air. That costs, you know, the numbers I've seen for that is if you could have reliable always-on power at less than five cents a kilowatt hour, you can start effectively pulling carbon out of the air. And then I don't know that we know all the things that you could do with that.

The other thing I point to which is crazy is fossil fuels, is when you pull the lever at a gas station, the energy that comes out of that is extraordinary. Something like four megawatts of energy per nozzle is coming out of that gas station. Massive amounts of energy. And so in order to compete with that, really to compete with that, I think you have to be this low cost.

What does it take to do this at scale? So let's say this all works, you know, demo next year is great. The engineering challenges aren't that big. So you figure out how to make these like that are just very reliable. How many of these systems will the world need and what does it take sort of from a global supply chain to get there?

Yeah, I would say I love this question, especially in this audience, when you start thinking about like the impact of what revolutionary technology really can do. And of course, coming from Sam, I get that question like this all the time. Our goal, our stated goal, I went out, there was a fusion meeting at the White House a year and a half ago, and we got up and said, like, okay, we want to make a dent on fossil fuels. What does that actually take? And you can run that number. And it's really scary. We have something like 4000 gigawatts installed in the world of fossil fuels. 4000 gigawatts of generating capacity. Generating capacity operating now in fossil fuels. And we've been putting in lots of renewable power and you see these plants of these wind farms that are 50 megawatts, so 0.005 gigawatts, sorry, 0.05 gigawatts. Don't mess the zero up there.

And these plants we're putting in 100 megawatt facilities, we're even putting in one gigawatt nuclear power plants that just came online in Georgia. And that's still one four thousandth of the content of what you need. So, I think that one gigawatt nuclear plant, which was the first new one we built in like 45 years, something like that, it's like $13 billion in a decade. And I think two contractors, they got fired halfway through. And a lot of politics on top of that to make it happen. So delivering even one gigawatt new is like a tremendous achievement. It's tough.

Yeah. And but if that's your goal and you want to do it quickly, then you know you need to replace those faster than they're aging. And so we set the goal of can we build systems, design systems that could be built to a gigawatt of power a day? One of those nuclear power plants a day. And it sets some interesting engineering things right off the bat that you immediately see, which is that you can't do what they did. You can't go and break ground and start pouring concrete and build big cooling towers. None of that stuff works. All that boutique one-off site stuff all is slow, all is big. And so you have to build things that look like generators. You have to build gigafactories that aren't producing a car at everybody's house where you, oh, I've never used this analogy, where you ship in the technicians and the assemblers to your house to build a car at your house. Imagine what that would cost versus building the gigafactory that's mass producing those cars on those dedicated robots.

And so that's what we do is we build those systems where our systems aren't one giant magnetic coil. It's actually 100, current one is 156 smaller coils that then we mass produce. Most of them are similar. There are variations in them, but different flavors of them and you mass produce those. They're powered by capacitors and semiconductors, each of which is a module. And we have hundreds of those modules that we mass produce per machine. And then you can put those on assembly lines. We're getting our first conveyor belt by the end of this year, which is going to be exciting to start building these on a real production line at scale. And then you just put them on a truck and install them wherever you need to.

The thing that comes off of that assembly line at the end, or the two of them a day, how physically big is each thing all in, like the actual generator plus all the capacitors and switches to support it and whatever else that needs to go somewhere?

So one of the rules of this is if you're going to ship it somewhere, then you need to be able to put it on a truck. And even more than that, it turns out there's a train tunnel you have to go through. Wherever you are, there's always train tunnels, which is why rockets actually are a certain size, is to fit through the train tunnels. And why when you build a rocket in place, it becomes a totally different exercise as we're seeing in Texas right now. And so that sets the size. And we believe that size, maximum size of one of these things is on the order of 500 megawatts of scale. And that's to fit in that train tunnel, even though the one we're building now is 50 megawatts. It's smaller than that. And so you're making two of these a day, and each of those can fit on a truck. That's great.

But the electronics don't fit on a truck. We actually have multiple trucks of these electronics. How many total trucks per generator?

Oh, I don't, I haven't, I don't know that I know that number right offhand. The actual generator was split across two?

The actual generator itself with none of the other parts should fit on one large truck. But you'll end up with, it's probably, probably a dozen or dozens. I don't know. I shouldn't know that number offhand. For the whole thing, with all the capacitors?

For the whole thing, all the capacitors all packed in tightly. And then this goes somewhere, there's a building, they put it underwater or whatever, power it up, and it just goes. Yep. And probably you have multiple of these on one site, because there's some cost effectiveness of like buying the land and having a security guard on the land and all those kinds of things. So you want to do more of that. But that's our goal.

And then you get to answer questions like, okay, great, how many, how many semiconductor switches are you going to be using? How much percentage of copper of the world are you going to start using? And the supply chain folks do not, do not enjoy it when I talk about a gigawatt a day, but we were putting those plans together now to figure out how to do that. And the answer is, is gigafactories, is you don't just have one factory. You have a dozen factories throughout the world where those materials are, and you mass producing them that way.

It's probably one of these weird things, which is as the supply chain, people talk about it one way, but in kind of whatever copper there is in the world, but as the cheap electricity comes online, it changes the definition of how much copper is available in the world. Because you can extract it and mine it in new ways. I think with cheap energy, there's a version of this that happens to us with AI as we think about designing these systems. I bet there will be a bootstrapping effect where the supply chain changes as you keep injecting energy into the world.

So Sam, what would be your answer to that question then? Like, okay, not desalination, you've already used that, but what does that low cost enable? Cheaper copper manufacturing? A lot of flops.

Ah, okay. A lot in one small place with this little network. A lot of other great things too, but I'm keenly interested in that one.

Okay. That's fair. No, I really think that energy and intelligence are kind of the two fundamental inputs to everything else that we want in the world. And if we can drive the cost of those two things down, yeah, sure, there's some other limiting reagents, but that kind of goes a super long way. With any of these things where the impact is so vast, you kind of feel a little bit silly to start saying things, so you go for the obvious, easy to imagine ones like desalination.

I think a world of hugely low cost and extremely abundant energy is probably just super better than we can imagine. I think you've seen this throughout history. At every major energy price transition, quality of life is just...

shoots up right after that happens. And I think this will be another one. And I would say, we see that in the world now. You can point to different geographic locations with cost of energy. And there's a direct straight line correlation of quality of life, which you have to use asterisks on how you measure that, and then cost of electricity in literally cents per kilowatt hour and the availability of it. And it makes sense, especially when it's, is there power on at night to study, right? Massive quality of life changes.

But I think you're right, like resources become, aluminum resources are driven by the electricity costs. Solar panel prices are driven by the electricity costs. If you can drive down that cost of electricity, you can actually put more solar panels on more roofs.

Can you talk a little bit about safety and how people should think about that, or how people will think about that, or what the attacks from like big oil and fusion are gonna be, proliferation, whatever you want?

Yep. So there's a couple of points to that. One, we have to be really thoughtful. Fusion is an atomic process, and so we want to be really upfront and not do some of the things that nuclear power did in the past, but we want to talk about some of the challenges that in fusion, it does make neutrons and it does make x-rays, and you have to deal with those. You have to actually protect against those, and that's really important. And so we actually had some big wins this year in that the Nuclear Regulatory Commission voted unanimously that how fusion will be regulated very clearly. They looked at the safety case and said, if we know what to do, we're gonna regulate this like a hospital, like a particle accelerator in a hospital. And that's very good because that's how all of our systems have already been regulated, and we've been regulated and licensed for five years to do this. And so those risks are really well understood.

The proliferation one is a sort of more interesting case, from my point of view, in that there was always this concern that if you can do fusion, you have the opportunity. Wait, before we go on, I think we should touch on some more. So I think, and I apologize if everybody in the room knows all of this, but it might be useful for the video. You know, in fission, you have this potential runaway chain reaction. You have this meltdown. You have long-lived radioactive waste. You have very high-energy neutrons hitting other materials and making new sorts of bad things. Why is fusion different?

Yep, so there's a couple things in fusion. One, fundamentally, what you're doing is taking isotopes of hydrogen, and bacterium found in all water, and you're fusing it into helium. And so your fundamental inputs and outputs are vastly different than uranium and plutonium and those kinds of things. But the actual sort of major challenge of fission is that when you build nuclear reactors, you put all the fuel in the core all at once. You put two years' worth of energy in the core all at once. In fusion, we puff in every single time. Just enough energy, just enough fuel to be able to get energy and electricity out. And worst-case scenario, that's all the energy you have available. And you just turn off the input fuel, and you're done. In a nuclear system, in a fission system, you have all that uranium sitting there just waiting to do something. And to me, that's like the fundamental engineering difference, is the quantity of energy available to do something, good or bad. And so that means we have valves. We have to puff in fuel. That sucks, but that means all the other, but you have only that amount of limited energy.

What about neutrons coming out of the reaction?

And so there are neutrons that come out of that reaction. In fusion, those neutrons are very well understood and very well correlated, very well diagnosed, and you can capture those. And if you're careful about how you do it, so you use water, or you use plastic, or you use concrete, you can actually absorb those neutrons and store those without having, or I'm sorry, absorb those neutrons and shield from those neutrons without a lot of activated material or reactive waste or any of those kinds of things. You still get some. Like one of the byproducts of this is actually an isotope called tritium, just another type of hydrogen. And that tritium is a radioactive gas. And so we actually have large programs of how do we store that tritium. Nice thing for us is tritium decays over time and it decays into helium-3 in 12 years. And that's a nice, that's a safe isotope that actually we use as the fuel. And so in the end, we burn that too.

What if I breathe in a bunch of tritium?

But if you breathe in a bunch of tritium, that's happened in the past. Tritium is one of those things.

Not to you.

What if I drink a bunch of tritium in water?

I should know the actual timeline. So you actually, you purge water very quickly. And same thing if you breathe hydrogen or tritium, the challenges that might combine with oxygen and form water, the water goes into your system. Over 12 years, that water decays. But you purge the water much faster than that. So the actual damage to your system is pretty minor. You still didn't do it. The joke is like, it's safe, but don't put the soap in your eye. It's still, you should not do it. The challenge comes with tritium is if it gets into the environment and you're drinking it every day. That's the danger. That's the thing you worry about. And so we spend a lot of time testing and making sure we're safe on these things. Again, kind of the interesting analogy is because you're only putting that little bit of fuel in the system. In the worst case scenario, that little bit of fuel is all that is that escapes. And so that's what we focus on. And then making sure we're doing it really safely all the time and measuring.

You know what? So I only have a few minutes before we open up to other questions. So I'm gonna skip the proliferation part.

What have you learned about building a culture of innovation?

A couple of things. And I actually point to Y Combinator for part of this in that we, before we spun off Helion, we had another small business and that small business did government grants, wrote research projects, like demonstrated a lot of key things. Say again? That's when we met. That's when we met. And you showed up with the physics, the fusion textbooks and said like, tell me about how your system works. Never forget that.

And we had to transition. We had to transition from looking at interesting science, which is really important to be done. And we were doing that. And I'm glad that like so much of the world still does that. But taking that and saying like, okay, but now we're gonna take this direction, this science that we learned, and we're gonna apply it. We're gonna start applying it and we're gonna start innovating the engineering around that. And we had to learn, like there's new science all the time. There's material science and all kinds of science you learn as you're doing, especially fusion, but you're doing anything. And so we needed to actually like reinforce that in the culture and actually not tell people you can't publish, but say like, wouldn't it be way more exciting to go build that thing instead? Let's put the money to building it instead of spending the time on publishing it and researching it. And so building that culture and then sticking to it, sticking to the culture of iterating quickly and innovating and making sure you're never stagnating, you're never getting too bureaucratic. So today we have both the systems we're actually constructing now. We have separate research programs and advanced concepts programs at Helion. We have external scientists that we fund to do new science, to like really investigate other potential things we'll do at national labs and at universities. And we find other people outside of the company to do that too. And so, but it comes from the top of continuing to push, like how fast can we iterate? How can we, the hardest thing, actually, this reminds me, the hardest thing we ever have to do at Helion is turn the machine off. You have this working fusion machine that's doing fusion every night, you're doing fusion. And you have to say like, I know it's working guys, but tomorrow we're not gonna come in. Tomorrow we're gonna turn the key because we gotta focus our resources on the next one. And the little revolt that happens right then is one of the hardest management things to deal with. But it's to reiterate, like we did that last time on the last machine and look what we learned. Let's do it again on this machine and next, guess what we're gonna learn next. Okay, questions? This has been great. Thank you so much for doing this. And Sam, you're really good at scale. So I'm listening to you talking about shipping two of these out every single day. Before Robert Busser passed away, he explained his frustration with the Department of Energy.

For the longest time, they were only doing Takamak. And he did funding from the Navy, and he could have spent more than $20 million or something. And the joke has been that fusion's always 20 years away, but we've only been doing Takamak when it comes to government research. Why is that? Why was the DOE just so set on that being the way versus trying a bunch of different things like this?

I'm gonna dodge your politics question. I'm gonna answer a slightly different one. We looked at the way the evolution of the fusion technology evolved. And you can point to actually work done, classified work that was later unclassified in 1958, that actually is the predecessor to what we do at Helion. And the idea of taking a fusion fuel, and don't hold onto it with big magnets like a Takamak does, but just in a pulsed way, compress it, squeeze it to those core conditions. And then the idea of directly recovering electricity, that's not new either. The earliest references to that were actually Ed Teller in the 1960s. And there's really good work done in 1978 modeling it. But the kind of crazy thing is, you can do the math and you know that in 1958, they needed the same math I use, which is they needed 10s or even 100 megaamps of electrical current that switches on in less than 10 microseconds in six microseconds. That was before the transistor was out of the laboratory. And so you could do the math and say like, wow, if we could generate a magnetic field that did this, this is how you do fusion. And then if you could do it efficiently, this is how you recover energy. And so the technology didn't exist. And so you had to do things slower. And slower for fusion means bigger. There's heat loads to the wall, so you have to go bigger. And then you go down an iterative path based off the technology you had at the time. And it just sets you to a scale. It sets you to a design that is a tokamak or something like it. And you're there. But unfortunately, I think the catch is that you missed. And maybe there's other good, maybe you can think of good analogies in other engineering and physics where this applies. But eventually you got to a scale of that technology then was so slow moving that you can't actually build the thing anymore. And so you then go down this slow moving path where you can't get to electricity anymore, where you could have stopped, looked back at the drawing board and said, oh, but fiber optics exists now. The transistor exists now. Oh man, gigawatt scale pulse switching exists now. And if you were to go back and revisit it, you would go down a different path today. And so that's what I would say, is you go down a technology path with the technology you have at the time. And maybe you just get trapped. For 70 years, you get into a local minimum and you actually need to step out and look at, oh man.

It's actually a huge problem. I mean, everyone grows up in some sort of system with whatever like educational received wisdom is pushed on them. And so if something like didn't work decades ago, you're still taught that it's just impossible. And very few people think about, all right, well like what has changed in the toolkit? What might be different now? What can we do that was just like, that seemed completely impossible, or if not that implausible because it does take like billions of dollars and decades that now might be possible. And so I think like the area that I've consistently seen value left in plain sight is when something wasn't possible when the people who are like now the teachers or the experts were like coming up. And that is just totally left out of what gets even considered. And it's dismissed by real experts because they can't even let themselves think about how the playing pieces have changed.

When I was in my academic era, post-magnetic fusion wasn't a thing that was taught. Like it was in all the textbooks in the 1950s and 60s and 70s, but it doesn't exist now. Like it wasn't in my fusion engineering class. When I was in my academic era, the one thing that I learned as like a undergrad study in AI was that, you know, a lot of promising approaches, but it's very challenging. But the one thing, if you wanted to have a failed career, the one guaranteed thing you shouldn't study is neural networks.

Really? It wasn't that long ago. Like it's amazing how quickly things can change.

Yeah. So one of the problems with like Tuck-Mack reactors is that you can't contain the neutrons that come out. They damage the reactor. So like the question I have, how long do you expect your reactors to run before they're restored? Or do you have some solution to this? But like, I mean, how long will they last?

Yeah, so there's two parts. So the Microsoft PPA that we have is to build a power plant for Microsoft that lasts for 15 years. And that's the goal of that system. We obviously want them to last longer. If you're capital cost driven, the longer it runs, the lower the cost of electricity is overall. But that's what we designed to right now. That's the lifetime we designed to. There's two parts to that. One part to that is that in Tuck-Mack system, again, if your goal was to do it big and steady, what you end up having to do is put the magnetic field in the middle of the plasma, the core part. And that's the part that's hard to shield. That's the part that's hard to insulate. That's the part that gets damaged. And so it's just a fundamental geometry of essentially you have a donut that's emitting heat and the part in the center of the donut gets a lot of heat. And so we design these systems and these pulse systems are all open. There's nothing in the middle. There's no core magnet. All, everything's on the outside where it's actually easy to shield. The other part to it is the fuel. So we didn't talk about this too much, but traditional, if you can use a deuterium tritium fuel, which generates, which is the easiest to do fusion with, but all the energy comes out in neutrons. In a deuterium helium three fuel, only about 5% of the energy comes out in the neutrons. Five to 10, depending on how you count, but like 5% of the energy comes out in neutrons. And mostly lower energy neutrons. Neutrons are two and a half MeV instead of 14 MeV. So they actually damage things a lot less. And so we can get to those lifetimes. And they damage them in different ways. So they sort of like, they weaken the structure, but you can then reheat treat them and actually they get strong again versus transmute them into other materials. But you can't, I don't know how to fix that. You can't, maybe you can. I don't know how, but we haven't figured that out yet.

If I recall correctly, when like Leslie took over what became the Manhattan Project, he had the rule, which was if you ever had to fork in the road, you take both, right? You're like, don't, you don't, you don't slow down. You just take both. And granted, he had the Nazis breathing down his neck, obviously fighting it, you know, you don't have that. But I'm curious, like how much faster could you go if you had the sort of resources that we had at the time of the Manhattan Project?

Yeah, it's an interesting question. And that actually comes back to the first question, which there was a chart in 19, oh my gosh, I forget the year, 1987, around the late 80s, which said if you invested in fusion, the tokamak specifically, if you invested in tokamaks, you'd get fusion on the grid on this day. And all the smartest scientists got together and said, this is what you do. And then there was a number of how many billions or if you did the equate number, trillions you would need, and you'd have it in this decade is when you'd get it. It was funded at half of the line that said never. And there was a line, like obviously you spend more money, you get it sooner, and the short ones were Manhattan Project numbers, and it was like 20 years. And then they went further and further, and there was a flat line that said you keep funding it this way, you'll never get there, you're actually not making advancement faster than, you're losing knowledge faster than you're gaining knowledge. And it was funded about half that. And so it's not a surprise, and it wouldn't be a surprise for anybody in that era of what happened. But I'm not convinced that if you threw unlimited money at it all the time at the beginning, you'd actually.

actually get there. Because one of the things our goals of fusion is to make low-cost electricity. So it has to get to low cost. It's not simply doing fusion at all costs. It's doing it the right way so that you get to low-cost electricity. So cost has to be a forcing function in the beginning through the whole system. Does that make sense?

From my point of view, where, frankly, in the Manhattan Project, it didn't matter what it cost. It didn't matter what each of those units cost. You wouldn't need to shut down one of the reactors every time you created one, maybe. Yeah, maybe. Maybe. And we do spend extra money in that way in that right now, like I said, we are taking apart sixth generation, building seventh generation. I have my team. There's a research team and a design team of five to 10 people right now designing right now the eighth generation system and then theorizing on what you could do and the risk reduction exercises to look forward to what comes after. Because I don't want 15 years of life. I want 60 years of life. What is it going to take? What is the material work I need to do to get there? And so we have a small team that are doing that all in parallel. That's not the same thing as building six different systems all in parallel, though.

Have you been fortunate to work on a lot of projects with this kind of shape? The right amount of money to start with for the fastest speed is not infinite. There comes a time when you do want to throw infinite money, and I think that'll be when Leon is trying to ramp up manufacturing. But you can just do it faster with more money earlier. It is astonishingly bad advice most of the time, even though it feels like it should be right.

I'm really curious about how you feel about finding the right people to build out your team and how that's changed over time from the early days to where you're at now. Yeah, I mean, I think as any executive or a founder, growing a team is a thing that keeps you up at night and growing the right team. And then as you hand off where there is a responsibility and say, okay, now take, here's my baby, go run with it, that you've created that organization that can do it well. And so we focus on a couple of things. We focus on people that are always wanting to move fast, that are iterating and that have shown in their life that they can do something quickly and get across the finish line with that. People that are optimistic, that you're looking forward to actually solving the problem and you want to solve the problem, rather than it's like, well, it was me, the problem can't be solved, I'm not gonna bother, but wants to solve the problem, and that's really exciting. And then is honest, is that when you're solving really tough problems, the ability to say, actually, this doesn't work is really, really important and be able to pivot, change direction, and change the engineering, whatever you need to do to make it work. But you can't do that unless you're gonna be really up front of that. And when you're building big complex systems, even just dropping a bolt in the machine, you better be it, you better be up front and be like, I've dropped a bolt in the machine, it's gonna cost a million dollars to get it out, but if I don't get it out, it's gonna cost a lot more. And so those are some of the things that we focus on, we're actually hiring, by the way. So, and across the spectrum, the company is probably about half technicians, machinists, people that are hands-on building the machine. And then about a quarter of the company is engineers and scientists that are doing the design work and the research work around it.

David Harris, UC Berkeley. Really exciting talk, thank you so much for this. I actually really wanted to hear Sam's proliferation question, but maybe I'll put a spin on it, which is I've been thinking a lot about the risks of open sourcing complicated or powerful AI systems lately, I know that's something you've thought a lot about, would you open source what you're building at some point, or if not, how do you think about maybe the dual use and proliferation risks of what you're building?

So, one thing that happened last year in Vienna that was very exciting, that surprised me, is we were at a conference and talking to experts, and I'd now written some papers on proliferation around fusion and the risks of it or the not risks of it. And one thing that we had is a lot of these weapons experts in the world came to us and said, please develop fusion and get it in the world as fast as you can, because if you don't, then we're gonna build enriched uranium systems throughout the world, and you will have a much different proliferation problem than we have now already, and it's going to happen unless someone can deliver clean, cheap base load power quickly. And so it was very interesting, and really eye-opening to me, because before, we've always been on a defensive of like, okay, we make a neutron, but it's the wrong kind of neutron in the wrong way. It doesn't really make any sense. It can't be really used to do what you wanna do. But this was the first time it was like, actually, maybe fusion is the solution to the problem. Also, tritium on its own does not make a bomb. Like, you know, we can talk about how bad it really is to have that sitting in a lot of countries, but I think it gets overstated significantly.

So I have a question that I think both of you have different perspectives on. I'm interested in it, and it's a little bit piggybacking off of that one. So I spent time at UC, so policy background, climate, energy side of the world. David spent some time at Pentagon, at NASA, et cetera. Yeah. The geopolitics of this is fascinating to me. So the Z machine, it's a weapons lab in Zambia, right? It does good work, and China's got two facilities, they say they're doing it over there, but they're doing something. And what does, and Sam, you know, I have been making coaching people to give the textbook showing it in Congress. So, you know, just from the geopolitics of all this, like, what does the future look like for who figures this out first? What does competition look like? Does it change the sort of energy dynamics of geopolitics? And yeah, what do you think the future looks like?

I think that's a really good question. And I don't know that we know the answer to it. And maybe that's a little scary, actually, in ways. Our goal is that, I mean, there's a world need for clean electricity in all kinds of countries, in addition to the United States. And we're seeing energy growth in the United States, electricity production growth has been actually pretty stagnant, few percent growth for decades, up until just recently. And now we're like, we're starting to see big, big growth. We're talking about, you know, tens of percentages of growth, and we're talking about doubling power output for the first time. And we're having to, it's actually a hard thing because we're having to learn how to build power plants again. And that's a tough thing. And that's only in the United States. The geopolitics of it also, right? We have all the things happening in Ukraine. We have all of those things. So you see an immediate need for it. But a couple of things I think are unique around fusion is that the fuel is deuterium. Like, you really can get it anywhere. And in fact, the United States actually doesn't have the monopoly on deuterium. Ironically, Canada does. But you can still get it almost everywhere. Might also be worth mentioning helium-3 you can get just by fusing deuterium and deuterium. So it literally all does just come from the water. So I think you've decentralized power from the source of the fuel, but you still have the generators themselves. And so our goal essentially early is that we wanna build these generators and deploy them. And deploy them at low cost, which means sell them, but sell them at low cost. Because if we have now monopolized, or sorry, don't use that word. But if we've held on to a technology that's too expensive and we're not willing to actually deploy it in the world, then we didn't do our job. We didn't do the company's goal.

And I don't know, Sam, as an investor in Helion, maybe you wanna answer that. How low of a cost do I wanna sell these for? Very low, I think. I think it'll all work out.

All right. I think we have time for one more question. Great conversation. So just to give you a segue, what does the roadmap from this research prototype to a scalable operationalization look like? What are the big challenges that you envision? And when can we have cheap energy?

Yep. Great. Kind of what I said, we were testing this first, the light bulb machine. We're designing right now our first power plant. That's the Microsoft data center power plant to come online in 2028. And we're looking for the site for that power plant. And right now, I mean, I was touring sites last week of potential sites that we would go and actually start building a facility on starting as early as next year and breaking ground anyway on the facility. That takes a long time to get through all the permits and stuff.

Timeline, we were really worried about timeline, even last year, where we didn't know how Fusion would actually be licensed and regulated. And it could have been on this 10-year nuclear timeline. And so the ability to turn that around, and I said regulated like a hospital, but what that means is that now your licensing timeline is well less than a year. And it's really well understood how to do it. It's not boutique. It's not new. And so that means we can now talk about licensing and building these much faster. So that timeline is not a concern. I'm left to how fast can I build these? And then how fast can I hook them up to the grid? Which turns out, I discovered this year, is a real problem. Really hard.

Yeah. I think what's going to happen is by, let's say by 2035, we will be capable of generating a significant fraction of the world's electricity needs. And it will be happening, and we'll be building lots of machines. But it'll be happening in weird ways because the grid is pretty corrupt and pretty bad. And people will say, just give me the electricity right at my data center. Just give me the electricity right at my desal plant. And it's going to be this very strange thing where the global infrastructure cannot keep up with the technology and the rate of production. And we're just going to find cooler things to do with it. And the grid changeover will be slower than you think it should be, and then way faster because the economic pressure will overwhelm the regulatory capture, corruption, whatever you want to call it. That's a great way to say that. Better than I did. I love it.

All right. Thank you so much for spending your evening with us.

Thank you.

+ Read More
Sign in or Join the community

Create an account

Change email
e.g. https://www.linkedin.com/in/xxx or https://xx.linkedin.com/in/xxx
I agree to OpenAI Forum’s Terms of Service, Code of Conduct and Privacy Policy.

Watch More

Deciphering the Complexity of Biological Neural Networks
Posted Feb 13, 2024 | Views 17.2K
# Life Science
# Higher Education
# AI Research
# Healthcare
AI Literacy: The Importance of Science Communicator & Policy Research Roles
Posted Aug 28, 2023 | Views 28.6K
# AI Literacy
# Career