저자: ElonMusk · 외국인
요약 보기
Musk는 규제 완화와 스타트업 지원이 기술 실용화의 핵심이라고 주장. AI가 향후 1~2년 내 소프트웨어 영역에서 인간을 압도하고 전세계 경제를 20~30% 성장시킬 것으로 예측. 휴머노이드 로봇이 10년 내 10억 대 이상 보급될 것이라 언급. AI 데이터센터 확산으로 2027년 전력 부족(15GW) 예상, 자체 발전소 보유 필요성 강조. 핵융합 등 에너지 기술은 민관 협력과 기존 규제 적용이 중요하다는 취지의 발언이 이어짐.
다듬은 글
nuclear reactor in space by 2028. We are going to deliver the first elements of a lunar base to the Moon by 2030. Um, and that's just our space portfolio. As part of our quantum efforts, where we want to be world leaders, we are going to create the first scientifically significant fault-tolerant quantum computer by 2028, by the end of this term.
So this is a national effort that I think only governments can bring together: a broad ecosystem of academia, the private sector, and government to mobilize resources . So, as you think about rebuilding your scientific ecosystem, I would be happy to discuss these issues with you in more detail, because I believe they are critical to the future of all of our countries.
So, I think we can take a break for a few minutes, because our guest should be joining any minute . Thank you. One thing I wanted to point out while we wait for our guest—I think some of us have been attending these meetings in various amazing places around the world for several years now. And I think one thing that I was very pleased with as this process progressed was that we were all able to reach a consensus on the Carolinian Principles of Emerging Technologies.
I think we are a diverse group of countries with many priorities, and we are all very committed to achieving the best outcomes for our people. And I think that the fact that we were able to agree on a set of principles that will ensure that we lead in the latest technologies is very inspiring, and it's something that we should all be proud of .
I think your teams did an incredible job working with our negotiating team to make this happen, and those of you who have done it know that achieving this outcome is not always easy. So it means a lot that we were able to do this, and I believe that this will be a lasting and important document for all G20 economies.
Um, I can't overstate how important we believe technological progress and innovation are to economic growth. Um, and when we think about the biggest drivers of economic growth in all of our economies over the last few years, it becomes clear that technology is the biggest potential driver of success.
And if we can create systems that allow these technologies to evolve and ultimately thrive, I think we all benefit from that. So, again, I think we should all be proud of this, and I just wanted to thank all of you for your great work in this direction. Perfectly. So, our first speaker is Elon Musk.
As CEO of Tesla and SpaceX, Elon has proven that overcoming the challenges of physics and engineering can be easier than fighting bureaucracy. And yet, I believe his dedication to both of these aspects continues to inspire people around the world. So thank you, Elon, for joining us. We have just concluded our opening session, where we made it clear that during this presidency we see economic growth as one of the most important drivers for all of us, and we believe that technology is a big part of that.
You have been at the forefront of this process for quite some time, and one of the key questions we would like to discuss with you concerns your experience implementing innovations in different countries around the world. In your opinion, what distinguishes countries where innovators successfully turn breakthroughs into ready-made technologies from those where progress is slowed?
After all, I think there are many lessons about what governments should do and what they perhaps shouldn't do . Well, yes. I think there's been a lot of talk about this , and part of it is, frankly, quite simple : you have to have an environment that's relatively free of regulation. This means that new things should be allowed by default, not disallowed.
For example, in the EU we see that the level of regulation is extremely high, and everything is usually prohibited by default. This hinders the progress of new technologies. This slows him down. Ultimately, it doesn't stop it completely, but it slows it down considerably. And of course, you need venture capitalists and an environment that supports new companies.
You can imagine new companies as little saplings in a forest. So most countries tend to give too much support to large trees in the forest and not enough to small seedlings. But large trees do not need such support. These small seedlings are what startups are. Therefore, the system should generally be geared towards supporting small trees, not large ones.
But this rarely happens. Because large companies usually have access to the country's leadership, while small startups do not. So you really need to promote the development of young companies and take active measures to do so . And, as I said, make everything allowed by default, not disallowed. One issue that many countries face here is the issue of implementation .
I think there is a general recognition that the introduction of new technologies can be very beneficial for economic growth, regardless of the size or type of country. Do you know what you think about implementation? What should countries do to encourage implementation? Does it come down to the same regulatory frameworks, or how do you view implementation as a key driver of growth?
Yes, I think it's worth being open to trying new technologies rather than getting stuck in the past. Ahem. Of course, new technologies need some support, and I believe they should be embraced. We will see—and are already seeing—significant productivity gains thanks to artificial intelligence. And we're going to see a very dramatic increase in productivity thanks to robotics.
You know, things like the Tesla self-driving car are going to be a huge advantage or already are . For users , and I think humanoid robotics will be an incredible change. I think to give you an idea of the scale, I think AI will probably increase the global economy by 20–30%. This is my rough estimate.
That is, within 20–30 trillion per year. And AI will be able to do everything digitally, everything that doesn't require manual manipulation of atoms. Probably by the end of next year. So, as I'm sure people know, AI is already incredibly good at working with software. Um, and it's getting to the point where AI won't just be good at programming, it'll reach what I call the Stockfish level.
So, Stockfish is a chess program that can easily beat the best chess players in the world. In fact, you can now run Stockfish on your phone and beat Magnus Carlsen at chess. So, my prediction is that it will happen sometime next year. Software , AI programs will become so advanced that they will reach the level of Stockfish , which means that humans will no longer be able to compete with AI in writing code.
AI will simply crush all humans in the software industry. And I think it's going to be extremely efficient, maybe even at the level of Stockfish, but definitely extremely proficient in all kinds of engineering and anything digital in literally 12 to 18 months. I am one of the Let me add something else, I apologize : this will lead to a 20-30 percent growth in the global economy.
So, we are talking about a fairly significant level of well-being. Not just from digital AI, but also from robotics, from humanoid robots, or, in fact, from general- purpose AI. I think we will see a multiple growth of the global economy, meaning it could increase by 10 times or more. These are simply stunning numbers.
Yes, I was just about to say that. One of the statistics we've given here is that in the four years since ChatGPT launched, I think over a billion people in the world are using AI. It spread very quickly throughout the country and the world. I think the question for you is that many people consider the next stage to be applied or physical AI.
How do you see the future of robotics? How quickly will it be implemented? I remember during the first Trump administration we were talking about automated factories, and now it's been 10 years. How fast do you think these changes are happening in the world of physical AI? Yes. Anything physical always takes longer than anything digital, because when you solve a problem in the digital world, it's just software code that can be easily copied to other computers.
When it comes to the physical world, you have to create a whole huge supply chain. You have to move many, many atoms. Um, all supply chains are pretty much global at this point. Therefore, it takes more time. However, um, when you think about humanoid robotics, I think the right approach is to understand that the usefulness of a humanoid robot, a general- purpose robot, is roughly equal to the quality of the digital AI software, multiplied by the power of the AI chip at work, and multiplied by the quality of the electromechanical dexterity, especially the hands.
Um, now all three of these components are improving exponentially. Um, and the usefulness of a robot is the product of those three things. Then, when you create robots, the robots will, um...um, the robots will start making robots themselves. So you get a recursive effect. So it happens very, very slowly at first, but then it grows at an explosive rate.
Um, so if you say, like, in 10 years, I would say there will be well over a billion. Oho. Um, and the productivity of each robot will probably be five times higher than the productivity of a human. This means that the productivity of humanoid robots in 10 years , and I consider this a conservative estimate, by the way.
Hmm, that's something I'd be willing to bet serious money on. Um, that in 10 years there will be at least a billion robots and that these robots will have at least five times the productivity of a human. This means that a billion humanoid robots will be more productive than all humans combined. Oho.
Um, let's change the subject a bit and move on to the issue facing the United States today. Um, you know, data centers have become a big political issue in the last, um, 6-8 months here in the United States. And I think there's a general understanding that in order to drive and power the AI revolution that's coming, we need electricity and the computing power of data centers to train and output all of this AI.
Um, you know, how do you look at this particular problem? Um, what stage are we at in terms of how much we've already built compared to how much we need? And as government leaders here think about how to prepare their economies and build the right energy and digital infrastructure, how should they think about what the computing and electricity needs will be in the next few years?
Well, there is actually a pretty serious energy crisis. So this is actually what it is , if you, say, just follow the topic of AI on Platform X, which is where, by the way, almost all the discussion about AI happens. I think you'll get a very good idea of where things are going. This is how I get the news, and it's incredibly effective.
Everyone who has any significance in AI posts in X. So I would advise you to just go to the AI topic on X and you will understand all these things and get a daily report of what is happening. And at this point, there is a consensus that there will be a significant electricity shortage next year. So, this is not the distant future .
It is expected, I think , that the consensus forecast among analysts who closely follow the AI field is that there will be at least a 15 gigawatt power shortage for AI chips in 2027. This is perhaps an obvious thing to expect, as the pace of AI chip production is increasing incredibly rapidly. They grow by approximately 40-50% per year.
But electricity available outside China is growing at only 10-20% per year. So it's obvious that what grows faster will eventually surpass what grows slower. And so, I would even say that at this point there are power issues until next year, which is why Google, Anthropic and many other companies are actually renting computing power from SpaceX.
Because we managed to launch AI better than others, but that's only thanks to building our own power plants. This is the only way we were able to do it. China currently has a huge amount of electricity, but due to the ban on the export of graphics processors, it is impossible to build data centers with the latest chips in China.
Therefore, it is actually worth considering what kind of electricity growth there is outside of China. And this is currently a significant deficit compared to the production of AI chips. So this creates an opportunity, I think, for countries around the world, if they're interested in AI data centers , to build power generation facilities and offer them to AI companies.
And in return, of course, these AI data centers will be taxed and have to pay reasonable fees and so on. But it really creates opportunities for many countries. Certainly. Well, thank you very much for your time. It means a lot that you were able to join us, and we really appreciate it. Thank you very much.
Always please. Thank you. Thank you. Perfectly. We'll now hear from David Sachs, former White House advisor on AI and my co-chair of the President's Council of Advisors on Science and Technology. He is a successful investor, and I think David understands well the importance of regulatory clarity and certainty for the development of new industries based on the latest technologies.
So I think he'll be joining us in a second. Here he is. Hello, Michael. Surprisingly. Good to see you, David. How are you doing? Good. Glad to be here. Thank you for inviting me. We are very glad to have you with us. It seems like you and I have been working for some time to ensure that the United States continues to be a center of great innovation and technological discovery.
I guess my first question to you is to set a little bit of context: the main theme that the President is pushing at the G20 is a return to growth, and how all of us, the G20 economies, can think about growth as our main focus. And specifically for this meeting—how technology can drive this growth. So my first question to you is: when you think about the policy conditions that need to be created to allow innovators to innovate, invest, experiment and bring new technologies to market, how should all the ministers here think about it?
What types of regulatory conditions should we introduce in our countries? Well, I think, first of all, we have a real AI boom in the United States. I mean, the AI boom is here, and I think it's a very positive phenomenon that we should nurture, not hold back. Some quick numbers: This year, the US is investing about $800 billion in capital expenditures to build infrastructure for AI and computing power, and that amount is expected to reach $1.4 trillion next year.
And every time I see these numbers, they keep growing . We have truly never seen infrastructure development on this scale. You have to go all the way back to the construction of the railroads in the 1800s to see anything equivalent. This is actually even bigger than laying fiber optic networks for the internet.
This is a new industrial revolution. And I believe that if we use it properly, this is an opportunity for the United States to reindustrialize, modernize the energy grid, and bring manufacturing back to cutting-edge industries, which has been a priority for President Trump. But obviously, I think it's also something that all the other countries in the world can benefit from if they take on it.
Yes. I think one of the issues that our Carolina principles cover is thinking about the policy environment that we can create to support innovation, and I think we have a pretty diverse mix of G20 members who are looking for unique approaches to regulation that can ultimately support their citizens.
When you think about how we , as technology ministers, should approach the regulatory framework. Are there any general lessons about what the government should be doing more or less of, or what is your general advice or opinion on broad technology regulation? Well, I think when we talk about regulating AI, it's worth first understanding that we are not without regulation in this area.
You often hear demands: we must regulate AI, we must regulate AI. Listen, there are already a lot of laws that apply to AI. So if you're talking about harms like fraud, discrimination, privacy violations, defective products, copyright infringement , sharing images without consent, and many other types of harms.
They are already illegal under current laws in the United States, and I'm sure there are similar types of laws in other countries. So we believe that there are already many laws that apply to AI, and if we are going to develop it, we should build on the laws and regulatory bodies that already exist.
For example, greater transparency, disclosure , audits—those things might be a good idea. We also have antitrust laws to prevent anticompetitive behavior by monopolies or duopolies. I think it's too early to say that this is the current situation at this stage of market development, but we know that technology markets sometimes have a tendency to monopolize or turn into duopolies.
And if that happens, we again have tools to prevent anti-competitive behavior. So, I think there is already a whole array of laws that deal with AI. This, this, this is one approach that we would really oppose and that we think would be a disaster for us in the United States—this is it. The so- called FDA for AI, where you model a new approval process for AI models on the model of what the FDA has for drugs, or what the FAA has in the US for airplanes.
The problem here is that it takes years for new drugs or airplanes to be approved. I mean, sometimes 5 years or more . Whereas in the AI industry, companies release new models every two months. It's too dynamic and fast for this type of approach. And I don't think we want to limit our innovation like that.
I think I called it...I don't know if the ministers there know what our DMV is—it's our Department of Motor Vehicles , which is known in the US for the fact that the process there takes a very long time. When you go to get your driver's license or something like that , and. I believe that any new regulatory body that has to approve AI product releases will become a kind of DMV for AI, where new models will get stuck in a long line for approval.
Just one last point on this: a lot of people wonder what the secret to Silicon Valley's success is? What made it such a jewel of the American economy? And if I had to boil it down to one principle , it would be permissionless innovation. This is really the principle that made Silicon Valley so successful: you can have a founder who dropped out of college and is building their product in a dorm room, or you take two founders who are, say, graduate students and are building their product in a garage.
You know, two trillion dollar companies were founded this way; this is Meta, as well as Google . What if these founders had to travel to Washington to get approval for their products? I'm just not sure these founders would be able to navigate this kind of bureaucratic maze to get their products approved.
So the idea is to just let the innovations develop and then regulate them later when they become more successful. I think that's really been fundamental to the success of Silicon Valley, and I wouldn't want us to lose that to some kind of pre- approval approach. Yes, definitely. I guess the last question for you is: we just heard Elon before you and asked him about the data center and energy needs that await us both in the United States and around the world to power the AI revolution.
And I know you've been thinking a lot about this issue of data centers here in the United States. President Trump also made a statement yesterday regarding data centers. I would like to hear your thoughts and advice for this group as they think about this development and the best way to provide the infrastructure we need for AI, but also do it in a way that we can engage communities.
So , what do you think about this issue of data centers both in the US and around the world? Well, this is definitely a hot topic today. I know this has become a controversial issue. Let me just make a few comments on this. So, supporting what President Trump said, here are some facts: If done right, data centers reduce electricity costs, not increase them, because AI companies will create new energy generation capacity.
This is a really key point: obviously, forcing AI companies to compete for energy or electricity with local residents and simply plug into the grid could lead to higher electricity prices. But if they are allowed to generate their own electricity, they will not only produce what they need—the so- called “ metering” mode—but they will also be able to feed the surplus into the general grid, which reduces the cost of electricity.
This also provides, one might say, financial support for the long-awaited modernization of the networks. So, this creates the necessary investment for this. So I believe that with the right approach, local communities benefit significantly from such data centers. We see the creation of hundreds of thousands of new jobs in construction and skilled fields thanks to them.
We also see property taxes decreasing in counties where data centers operate . For example, residents of Loudoun County in Virginia in the US pay an average of $6,000 less in taxes because AI companies are taking on the costs. And one final note: while the administration and President Trump have emphasized that data centers can be beneficial to communities, we are not forcing anyone to accept them if communities don't want to.
It seems to me that there are a lot of fakes circulating around this. Michael , as you know, as we've been working on our national AI vision, and then the president signed an executive order last December, he made it clear that the federal government should not limit state decisions on data centers.
This is a fundamentally local decision, we respect it and have not changed anything. Although we are convinced that this is a profitable business for local communities if they dare to take such a step. Certainly. Thank you for your time, it's very appreciated that you were able to join us today. Thank you very much, David.
Thank you, Michael. Thank you for inviting me. Thank you. And now, for our final speaker, I have the honor of introducing Bob Mumgard. As co-founder and CEO of Commonwealth Fusion Systems, the world's largest fusion energy company, Dr. Mamgard is the ideal expert to talk about the full life cycle of emerging technologies and the path of innovation from laboratory experiments to commercialization and changing the world.
So, thank you very much. Thank you very much for joining us , Dr. Mamgard. Glad to be here. Thank you very much. Well, Dr. Gracios. Um, great. Well, in the Caroline Principles, which were agreed upon by the G20 countries, we outlined how governments can accelerate the development of new technologies.
And I think your technology is unique in that just a few years ago, the private sector had almost no investment in it . How do you think the policy decisions or political environment in the US allowed you to raise capital, build, and ultimately develop fusion technology? So, you see, fusion technology does not suffer from a lack of market demand.
We know that in the future, intelligence and energy will be the drivers of the economy, which is why many are seeking advanced energy technologies; these markets are already well established. But the barrier to creating products in these markets is how to apply the basic science ? How to combine this with capital?
How do you implement this in organizations that can stay true to the science, develop it, scale it up and achieve results, and eventually create a product, worth billions of dollars. These are not consumer or software products. As Elon said, moving atoms is not the same as moving bits. And moving atoms at a temperature of 100 million degrees is completely different from simply moving them.
Therefore, an extremely well-established innovation environment is needed. You must have sources of capital. You need scientific experts, but you also need the grease and glue that government can provide to bring it all together. Today, more than 50 companies operate in the fusion industry, and private investment exceeds $14 billion.
If you go back 10 years, this didn't exist at all. And of that $14 billion, the vast majority —both funds and companies—is concentrated in the United States. And CFS—we are the largest of them , having raised about $4 billion. We are building a facility outside Boston that, a generation ago, would have been considered a government facility on the scale of the Manhattan Project.
The reason this was possible is because the government was a partner, not an obstacle; Here are a few examples: first, from the regulatory side—how to license a fusion power plant? Just four years ago, this was a completely untested question. It would be easy to say, "let's figure everything out and build a new structure from scratch ," which would be very prescriptive.
This would be a mistake, as we don't know what technology it is. We don't know the problems at a detailed level. But we know enough to protect the public. So instead, we in the United States and Great Britain said: let's look at the dangers. Let's take the most relevant example that already exists—in our case, particle accelerators—and make sure that fusion fits into that framework , ensuring that regulation doesn't stop it in its tracks.
We will have enough space to launch the first products. And it gave investors the confidence to pour billions of dollars into the business, removing a hurdle that everyone had seen in cutting-edge new technologies before. Another part of this was using existing publicly funded research that had accumulated expertise and finding ways to partner between the private and public sectors.
It would be easy to say that the government should go and build the first generation of power plants, but that is a path to nowhere. These power plants will never reach the market. It's like we're throwing it over the fence, and there's no one to catch it. Instead, we can say: let's take the expertise that has been developed and create a public-private partnership, provide access to this knowledge in a way that preserves intellectual property so that companies can be created and developed on this basis.
Let's make sure that capital—a kind of help for the very first ideas—can be channeled into larger projects, which can then attract additional private capital in addition to public capital. And this gave impetus to the development of almost all American fusion companies. Yes. You know, I think that's a really great observation.
And I think we often emphasize that when these new technologies start to become commercialized. The immediate reaction of regulators should not be to create completely new mechanisms to regulate them, but instead, they should carefully consider which existing rules can be applied to a particular technology first.
And I think David emphasized that . I think Elon confirmed this to some extent as well. Perhaps the last question for you, and I think what interests all of us, is the timing. I think with fusion, that's always been a major issue in history. We had two important conversations with David and Elon about how important energy will be to the AI revolution , which is already here and will only accelerate in the coming years.
Of course, the inclusion of thermonuclear fusion in the energy balance would be one of the most remarkable achievements in human history. So, we are all eagerly awaiting commercialization. So what do you think about this? And when do you think there will be a real impact on our energy grid? Well, you know, first of all, we should avoid the fallacy that the past is a prologue to the future.
Today we are dealing with a situation in fusion and other physical technologies where the toolkit available to innovators has become much broader and deeper than even 5 years ago. So when we look at what we can do with large-scale computing, AI, advanced materials, and how we organize ourselves and who we bring into the fold.
What is their previous experience? How were they taught? This is a whole new generation of people. And when you combine this, we see that processes that previously seemed time-consuming can be significantly accelerated. Of course, this is the first time we've seen this in software. We're seeing this now in large-scale computing and AI.
Things that even 6 years ago seemed like a distant prospect , so we shouldn't fool ourselves into thinking that everything has to go slowly. Now, when you look at implementation schedules, it's essentially a question of how fast you can build. We are currently completing construction of this facility in Massachusetts, but we are preparing to begin construction of the first commercial power plant in Virginia, not far from you.
And this could allow fusion energy to be connected to the grid in the early 2030s. There are many technological issues that will have to be addressed in the process, but the toolkit is well adapted to this technology. If we plan for the longer term, say that fusion won't be here until the 50s, then we won't move faster, we'll just do it in the 50s.
So we must plan for the most aggressive pace we are capable of and use all possible tools to achieve this. And I think the United States has created a good model, and it's an example for the whole world, where we see countries systematically reducing their schedules for fusion year after year, watching the progress that's happening mostly in the United States.
No, I think it was incredible to see, and the amount of private sector investment in this industry was just staggering. And I think for those of you who have some free time in the United States, I highly recommend visiting Bob's outside of B. It's truly incredible what our private sector is creating.
So thank you very much for your time. It means a lot that you were able to join us today. Thank you. So, I think we'll let the press out, and then we can have our discussion without the press, like
댓글 0