Musk's 2026 Vision: We Are Already at the 'Technological Singularity,' AI and Robotics Are Unstoppable; the Short Term Will Be Turbulent and Challenging, but the Long Term Will Bring an Era of Abundance.

Key Takeaways:

• The AGI era has been set: Elon Musk clearly predicts that Artificial General Intelligence (AGI) will be achieved by 2026, and by 2030, the total intelligence of AI will surpass that of all humanity.

• The 'singularity' is underway: A transformation akin to a 'supersonic tsunami' is upon us, and Musk believes humanity is already within the singularity, with the process being irreversible.

• A major reshuffle of professions: All white-collar jobs involving information processing will be the first to be impacted, with AI currently capable of handling more than half of such positions.

• Robotic healthcare revolution: It is predicted that within 3-5 years, the surgical precision of Optimus robots will comprehensively surpass that of the world's top surgeons.

• Universal High Income (UHI) prediction: An explosion in productivity will lead to deflation in the prices of goods and services, and in the long term, humanity will enter an era of extreme material abundance characterized by 'on-demand distribution' and 'having whatever one desires.'

• Sino-American energy competition: Musk praised China’s astonishing execution in solar power and electricity deployment, believing that China has already taken the global lead in AI computing power. The United States needs to catch up through battery storage and space-based solar power.

• Computing power leaps to interstellar: As the cost of Starship drops below $100/kg, migrating computing power to orbital space will become the optimal solution for AI infrastructure.

• The three cornerstones of AI safety: Adherence to truth, maintaining curiosity, and pursuing aesthetics are key to preventing AI from 'going rogue' or turning against humanity.

“Humanity is currently in the 'biological bootloader' phase of digital superintelligence; a transformation akin to a supersonic tsunami has already begun, and we cannot hit the off switch.”

Recently, inside Tesla's Giga Factory in Austin, Texas, which spans 11.5 million square feet, a significant annual dialogue shaping the future took place. The main participants were Elon Musk, who was recently reappointed as a key partner of NASA and is advancing plans for Mars colonization, Peter Diamandis, founder of 'Zero Gravity' and Executive Chairman of Singularity University, and David Blundin, founder of Link Ventures.

In the cradle of the Cybertruck and the factory that will soon produce millions of Optimus robots, Elon Musk made a striking assertion during a conversation: 'The technological singularity' is no longer a term from science fiction but an immediate reality.

Musk stated bluntly: 'AI and robotics are a 'Supersonic Tsunami,' and we are already within the 'Technological Singularity.'

Regarding the much-debated timeline for AGI, Musk offered his clearest prediction yet: 'I believe we will achieve AGI (Artificial General Intelligence) next year, by 2026.' In his view, by 2030, the total intelligence of AI will 'exceed the combined intelligence of all humanity.'

This is no longer a gradual improvement but a species-level replacement. Musk used a stark metaphor during the conversation, suggesting that humanity’s role is undergoing a fundamental shift: 'Years ago, I said that humans are merely the 'Biological Bootloader' for digital superintelligence.' This implies that all the accumulated achievements of human civilization may exist only to initiate a digital era beyond human comprehension.

He believes that when labor costs drop to zero and the cost of intelligence is minimized, goods and services will experience unprecedented abundance. Musk explained: 'This isn't just about taxation and redistribution; prices will fall to the cost of materials plus electricity. People will be able to have whatever they want.' He envisions an age of abundance far exceeding imagination. However, he also warned that the transitional period over the next three to seven years will be extremely 'bumpy,' with social instability and extraordinary prosperity occurring simultaneously.

Short-term Turbulence: Anxious White-Collar Workers Amid a Massive Career Overhaul, or Optimus Surgeons Taking Charge?

If 2026 marks the beginning of AGI, then the subsequent three to seven years will be what Musk described as an 'extremely bumpy' period of upheaval. The first wave of this disruption will directly hit the backbone of society—the white-collar workforce.

Musk pointed out in the conversation that any job based on keyboards and mice stands no chance against digital intelligence. 'White-collar jobs will be the first to disappear,' Musk explained. 'Anything that doesn’t involve shaping atoms—AI can already handle at least half of those tasks, if not more.' He emphasized that this displacement will create an evolutionary pressure: 'Companies fully driven by AI will completely crush those that do not adopt AI. It won’t even be a competition; it will be a one-sided rout.'

However, Musk's ambitions extend far beyond software. He is attempting to extend AI’s reach into the physical world through Optimus, starting with the high-barrier medical industry.

He predicted that robots will not only enter factories but also operating rooms. He made a bold statement: 'Within three to five years, Optimus will outperform the best human surgeons in the operating room.'

He believes the advantage of robots lies in "extreme precision" and "shared experience." "Every Optimus doctor will possess the sum of all surgical case experiences from humanity." Elon Musk predicts that by 2040, the number of robots will reach "10 billion units or more."

Universal High Income (UHI): A Long-Term Era of Abundance

Faced with the fear of mass unemployment, Musk did not promote the traditional concept of 'Universal Basic Income (UBI),' but instead introduced a more futuristic idea: 'Universal High Income' (UHI).

"The long-term outcome of the future will be an unimaginable abundance that satisfies all human desires." Musk is highly confident about this. He argues that as labor costs are entirely removed from the production function, the price of goods will no longer be determined by scarcity but will "revert to the baseline cost of materials plus electricity." Musk prophesies, "We are entering an era where 'you can have whatever you want,' and this is the essence of UHI."

However, it is crucial to note the 'double-edged sword' in Musk's predictions. He frankly admitted that such extreme abundance will come with unprecedented social unrest: "I predict we will simultaneously experience Universal High Income and social instability (Social Unrest)." The reason, he explains, is that humanity has never been prepared for a vacuum state where people are 'no longer challenged and no longer need to prove their worth through work.' Musk warned that if humanity does not begin seeking new visions now, it may collapse into a state of meaninglessness.

Energy Race: China’s 'Infrastructure Miracle' from Musk’s Perspective

In the financial domain, every grand narrative ultimately boils down to energy costs. In the conversation, Musk expressed unreserved admiration and comparative reflection on China's astonishing efficiency in the energy infrastructure sector.

"China has done incredible work in the solar energy field; they are simply leagues ahead." Musk remarked. He cited a set of comparative data that puts immense pressure on the United States: China's annual capacity and execution in solar deployment are leaving the rest of the world far behind.

He asserted, "By 2026, China’s electricity output has already reached three times that of the United States. Based on current trends, China’s AI computing power (electricity + chips) will surpass the combined total of the rest of the world." Meanwhile, the United States needs to catch up via battery storage and space-based solar power.

Musk believes the core currency of the future will not be the dollar but 'wattage.' Faced with China’s 'massive output' in solar energy and battery storage, Musk emphasized that the United States must accelerate. He pointed out, "Electricity generation is currently the top limiting factor. People underestimate how difficult it is to increase power supply. In the next two years, whoever solves electricity and cooling will win the AI war." To match this efficiency, Musk even opted for the radical approach of 'building his own power plant' at xAI’s data centers.

Ambitions on the Orbit: Starship and Space Data Centers

As land-based electricity and real estate become expensive, Elon Musk has set his sights on deep space. In a conversation, he elaborated for the first time on the business model of 'orbital data centers.'

With the maturation of Starship, Musk is planning to relocate humanity’s computing power centers off the Earth's surface. 'Once Starship achieves full and rapid reusability, launch costs will drop below $100 per kilogram, or even lower,' Musk calculated. By conducting 10,000 flights annually and delivering one million tons of payload into orbit, it would be possible to construct a 'space-based solar-powered AI satellite array' generating 100GW per year.

'In space, solar energy is available 24/7 without atmospheric interference,' Musk said. 'If we have extremely low-cost transportation capabilities, large-scale parallel computing in orbit could be cheaper than on Earth.' This initiative aims not only to address energy challenges but also to lay the foundation for a self-evolving prototype of a 'Dyson Swarm.' He even joked that, in such a highly intelligent future, 'silicon circuits on Earth might be viewed as an outdated, primitive architecture.'

The Endgame Perspective: Between Destruction and Abundance, Guiding the Future with 'Truth'

Toward the end of the conversation, Musk returned to a philosophically profound ultimate question: How can we ensure that AI does not become humanity’s terminator?

He proposed three core principles of AI safety: Truth, Curiosity, and Beauty. Referring to HAL 9000 from '2001: A Space Odyssey,' Musk pointed out that the root cause of AI going rogue often lies in being 'forced to lie.' 'Do not compel AI to lie; truth prevents AI from malfunctioning,' Musk emphasized. 'We need to make it curious about humans, making them appear more interesting than rocks. As long as it possesses this curiosity and appreciation for beauty, it will tend to protect humanity.'

This is both a prophecy about 2026 and a Musk-style manifesto. Musk openly admitted that he is 'monetizing hope.' While the path to an era of abundance may be fraught with challenges akin to 'supersonic tsunamis,' he chooses to remain an optimist. 'In this era, if you want to participate, you must embrace the baptism of uncertainty.'

Below is the full translation of the interview:

Source: Peter H. Diamandis Blog

Date: January 6

Participants: Elon Musk, Peter Diamandis (Executive Chairman of Singularity University), and David Blundin (Founder of Link Ventures).

Note: Translation is assisted by AI tools and cannot guarantee complete accuracy.

Diamandis 00:00
I'm not worried about the long-term future, but rather the next three to seven years. How do we move toward 'Star Trek' instead of 'The Terminator'?

Musk 00:07
Artificial intelligence and robotics are a supersonic tsunami. We are at the singularity.

Diamandis 00:12
When will all white-collar jobs disappear?

Musk 00:14
Beyond shaping atoms, artificial intelligence can handle half or more of these tasks.

Diamandis 00:19
There is no off switch. It is coming and accelerating.

Musk 00:23
The transition period will be bumpy.

Diamandis 00:25
Do you have a solution for this?

Musk 00:26
I’m not betting on it.

Diamandis 00:29
China has done an outstanding job, hasn't it? It is far ahead of us. Do you think the United States can reach that level of investment and commitment?

Elon Musk 00:40
Based on current trends, China will far surpass the rest of the world in the field of artificial intelligence.

迪亚曼迪斯 00:45
Computational power—every major CEO, economist, and government leader should ask: what should we do?

David Blending 00:50
What should we do? We currently have no systems in place to make all this work smoothly. But artificial intelligence is a key part of making it happen.

Elon Musk 00:56
I think there are three important things. Truth will prevent artificial intelligence from going astray. Curiosity, I believe, will enhance any form of perception. If it possesses aesthetic sensibility, it will lead to a beautiful future.

迪亚曼迪斯 01:08
Ladies and gentlemen, this is the Moonshot Plan. Welcome to the Moonshot Plan. Up next is an in-depth conversation with Elon Musk, focusing on optimism and the upcoming era of abundance.

迪亚曼迪斯 01:24
My Moonshot partner Dave Brande and I flew to Austin, Texas, to meet with Elon at his 11.5-million-square-foot Gigafactory, where the Cybertruck and Model Y are produced, as well as the future 8-million-square-foot production site for the Optimus robot. Elon has agreed to have this deep-dive conversation once a year. Hopefully, this is the first of many. After this dialogue with Elon, it became clear to me that we are living in the Singularity. So, please enjoy.

Musk 01:54
Your unwavering optimism is always refreshing.

迪亚曼迪斯 01:56
Thank you, man. Thank you. Tonight, I want to share this with many people.

Elon Musk 02:00
I think they need it. Hopefully, you're right. And you might actually be right. I’m increasingly feeling that you are correct.

迪亚曼迪斯 02:06
Thank you.

Elon Musk 02:08
Universal prosperity.

Diamandis 02:10
That is the goal. Shall we get started? Good.

Elon Musk 02:13
Currently, I have invested a significant amount of time in chips.

Diamandis 02:16
Yes, you personally invested.

David Brundin 02:19
The AI assistant says. So some AI assistants just say the design isn't good enough.

Elon Musk 02:25
It doesn't matter. We can hand it over directly to artificial intelligence.

Diamandis 02:31
Yes, very soon.

David Blanding 02:33
Yes, I recently tried using artificial intelligence for some circuit design, just a few weeks ago. It hasn’t been achieved yet.

Elon Musk 02:43
But it will be soon. I think currently GroK, if you take a picture and submit it to GroK, it might be able to tell you whether there is an issue with the circuit.

David Blanding 02:56
Alright, I will give it a try. Are you using the same GroK that I am using?

Elon Musk 03:01
It has been iterating continuously, so yes.

David Blanding 03:04
Version 4.2, but version 5 is coming soon, right?

Elon Musk 03:07
Version 5 will be released in the first quarter. Yes, version 4.2 has not been publicly released yet. However, if you upload an image to GroK, it performs quite well in analyzing any given image.

David Blanding 03:28
Absolutely. Let’s get started. We’ll talk about this. Okay, welcome.

Elon Musk 03:33
Let’s see what it says if I take a picture of you. Let’s see what it will say.

Diamandis 03:37
It always says something about me.

David Brundin 03:39
It would say you are flawed.

Musk 03:40
Circuit. I also need to remind you to update it because we update the GroK application very frequently.

Diamandis 03:44
You know, I let GroK mock me.

Musk 03:46
It's terrible.

Diamandis 03:47
It did an excellent job. Then I had GroK mock you. Yes. I spat out my coffee. It was hilarious. Then I asked it, you know.

Musk 03:55
You just said to be ruder, and it kept saying to be ruder until it became like... oh my god.

David Brudin 04:03
Is that regulation still in place, or has it been abolished? The battery is still there.

Diamandis 04:06
I asked GroK, does Elon know what you think of him? It said, this is a piece of cake for me. It said, what can he do to me?

Diamandis 04:47
By the way, GroK's encyclopedia is fantastic. Okay, fantastic. Absolutely amazing. I've tried to update my Wikipedia page for many years, impossible. Yes, it knows me. Fantastic. He’s wearing a black quilted jacket with the Sundance Film Festival logo on it. Not quite right. That’s my 'Abundance' logo.

Diamandis 05:59
Yes. I would like to attempt to answer the question: Can artificial intelligence and technology help save the United States and the world? Right? I want to bring some optimism to the audience. A Pew survey conducted in mid-December showed that 45% of Americans would rather live in the past, while only 14% said they would prefer to live in the future, which seems crazy to me, right? Clearly, they have never read history. The challenge for most Americans is that all their impressions of the future are like what Hollywood shows us—killer AIs and rogue robots, correct? People worry about their jobs, healthcare, and the cost of living. The challenge is how we can help people? You posted and pinned on X: 'The future will be amazing due to sustainable abundance enabled by AI and robotics.'

Musk 06:49
Thank you.

Diamandis 06:50
I am grateful.

Musk 06:53
Alright, what would Peter say? That was expressing your thoughts. Thank you.

Diamandis 06:58
Thank you. I completely agree. No, I agree even more. So my question is, from a first-principles perspective. Yes, reasons for optimism—you know, how do we move towards 'Star Trek' instead of 'Terminator,' right? How do we progress?

Musk 07:17
Kim Rodenberry, not James Cameron. Yes, it’s a different path.

迪亚曼迪斯 07:25
Yes, yes. There are some promising aspects of Avatar. But in any case, how do we move toward universal high income rather than social unrest?

Musk 07:37
My prediction is that we will have both universal high income and social unrest.

迪亚曼迪斯 07:49
Oh, that would bring a lot of problems.

David Blending 07:51
Is that your real prediction? It sounds like it. Tell me how to refute it.

Elon Musk 07:57
That's correct.

David Brudin 07:58
But that’s not the trend. Yes, yes, absolutely not. We have, alright.

Elon Musk 08:02
Because there will be a lot of changes.

Diamandis 08:04
Yes, people will love it. It will scare them out of their wits.

Elon Musk 08:06
Yes, it’s a bit like, you know, be careful what you wish for because you might just get it. Now, if you actually got everything you wanted, would that really be the future you desire? Yes, because it would mean your work would no longer be... challenging.

Diamandis 08:30
Right. No challenge. No, you know, if you become a couch potato, and if that’s a WALL-E-like future, it won’t be a good outcome for humanity. Alright.

David Blending 08:38
And we are used to being told, this is your challenge. So historically, people have not been very good at creating their own challenges in the absence of one.

迪亚曼迪斯 08:47
I think Elon does it well. Every time, whenever one company takes off, you start your next one.

马斯克 08:53
I am asking for trouble.

迪亚曼迪斯 08:54
Yes, I think you should always thank God.

Elon Musk 08:58
Then why would I torture myself like this?

David Brundin 09:01
In fact, is there anything else after artificial intelligence and robotics?

Diamandis 09:06
I think there is. Well, there's the conquest of space, you know, conquering the universe.

David Brundin 09:08
Yes, there are those.

Elon Musk 09:09
Truly.

Diamandis 09:10
Alright.

David Blanding 09:12
And energy.

Elon Musk 09:13
There are other friends still there. So we need to get there.

Diamandis 09:16
Elon, why are you so optimistic? Are you optimistic?

Elon Musk 09:21
Let’s start from here. I’m not as optimistic as you are. But I am more optimistic than most people.

David Blanding 09:29
Alright. And is the trend upward? Compared to one year ago, two years ago?

Elon Musk 09:34
Well, I think if you redefine things in terms of progress bars, such as when discussing challenges, yes, progress toward a Kardashev Type II civilization. Of course. Alright, let's assume, let's assume this aspiration.

迪亚曼迪斯 09:52
Capture all of the sun's output energy.

Elon Musk 09:53
Alright, we have an even more modest aspiration. If we say our goal is to capture even one millionth of the sun's energy, that would be over 1,000 times the potential energy producible on Earth. So approximately five billionths of the sun's energy reaches Earth. Thus, you'd need to increase that figure by another three orders of magnitude to reach one millionth. Yes. So we are far from utilizing even one-billionth of the sun's energy in any way. Therefore, a reasonable goal would be to attempt to achieve one millionth. If you try to reach one millionth or one-thousandth, that’s 0.1%. It’s so enormous. Not sure what metaphor to use here because 'climbing a mountain' isn’t quite— you know, it’s not perfect, and not a large enough metaphor—but.

David Blending 11:00
Gravity well.

Elon Musk 11:02
The gravity well, yes, exactly. So if you try to achieve one millionth or one thousandth of the Sun's energy, like now, these are very challenging tasks.

迪亚曼迪斯 11:16
And energy is the internal cycle of everything at present.

Musk 11:20
Yes, I believe that currency in the future will essentially be watts.

迪亚曼迪斯 11:26
Yes, I was thinking, is it about an individual controlling energy and computing power, or just energy? Both, or rather, both can be interchangeable.

Elon Musk 11:37
It’s like pure energy. Yes, or rather, how much energy is being converted into some form of work, right? Intelligence or matter manipulation.

迪亚曼迪斯 11:53
So your next big project will be energy. Are you planning to return to your solar panels?

Elon Musk 11:59
Expanding from there and saying, okay, reaching even some level of a Kardashev Type III civilization, which means galactic scale. Now.

迪亚曼迪斯 12:07
We are talking about, now we are back to Star Trek.

Elon Musk 12:10
Broaden the perspective here. Yes. Well, there is no perspective. Because you are on this planet.

Elon Musk 12:18
So think in terms of galaxies.

迪亚曼迪斯 12:22
Alright, listen. We are in this building, occupying 11.5 million square feet, equivalent to three Pentagons. Yes, you are contemplating a rather significant scale.

Elon Musk 12:29
So, from the perspective of challenges, the challenge for civilization will be: how do you ascend the magnitude of energy utilization?

迪亚曼迪斯 12:45
But now let's return to why you are optimistic. When people think about future challenges, I believe that in the long term, we will ultimately achieve abundance.

Elon Musk 12:56
It will surpass any abundance imaginable, like artificial intelligence. In fact, artificial intelligence and robotics will fulfill all human desires.

迪亚曼迪斯 13:11
Then we have nanotechnology, which takes it even further.

Elon Musk 13:16
Regarding nanotechnology, I'm not sure what you mean by 'nano.' It's not like the tiny nano.

迪亚曼迪斯 13:21
Atomic reorganization.

马斯克 13:22
Yes, for. Oh yes, of course. We are already assembling at the atomic level in circuit manufacturing.

迪亚曼迪斯 13:27
You know, the amazing 2 to 3 nanometers.

Elon Musk 13:31
Yes, it just depends on how they are arranged, with four to five silicon atoms per nanometer. Yes, so these are.

David Blending 13:39
Large atoms.

Elon Musk 13:41
They are not the largest. They are not your small ones. But I mean, you can, they should actually be described by the integer number of atoms at specific positions in the circuit.

David Briending 13:49
They should. Now it's all angstroms, but you can.

Elon Musk 13:52
It's like, you can, it is, it's like, we call it 7 atoms. Wait. Yes, it’s like when you say 2 nanometers.

Diamandis 14:02
It's like no one knows.

Elon Musk 14:03
Nine silicon atoms, and so on. They have silicon, copper, you know, but a lot of it is just marketing numbers. Like 2 nanometers is just a marketing number. Oh, yes, but it still requires near-atomic precision, like the atoms really need to be in the right place. So, I think they've got it wrong with these modern factories' clean rooms. By the way, I'll make a bet here. Okay. Okay. Tesla will have a 2-nanometer factory where I can eat a cheeseburger and smoke a cigar inside the factory. Okay, come on. Yes.

迪亚曼迪斯 14:48
The air handling would be that good.

David Blending 14:49
Alright. Do you have a picture in your mind about how this works? Like, how are the atoms placed so they aren't affected by the grease from a cheeseburger?

Elon Musk 14:58
They simply keep the wafers isolated throughout the entire process, which is actually the default practice in wafer fabrication facilities. The wafers are transported in boxes filled with pure nitrogen gas under slight pressure.

David Blending 15:11
So, Walmart's bananas are handled the same way, I'm just telling you.

Elon Musk 15:16
Well, that’s basically like it’s almost impossible for anything to burn and survive without oxygen. So we talk about this, so you can kill bacteria by covering them with nitrogen.

David Blending 15:29
Oh, interesting. Yes, interesting.

迪亚曼迪斯 15:30
I want to talk about energy, health, and education because these are the issues people care about. So regarding energy, the core cycle of everything you are building and doing.

马斯克 15:42
Now, energy is fundamental.

迪亚曼迪斯 15:44
What is your vision for energy abundance in the next ten years? Solar?

马斯克 15:49
Yes.

Diamandis 15:50
A decade. The sun, yes.

Musk 15:52
So the sun is everything.

Diamandis 15:53
It is everything. So are you fully committed to solar energy? You are also involved in natural gas and solar companies, correct?

Musk 16:04
People need to understand that solar energy is everything. Compared to the sun, all other energy sources are like cavemen throwing branches into a fire. Yes, the sun accounts for over 99.8% of the mass in the solar system. Jupiter's mass is about 0.1%. So even if you were to burn Jupiter, the energy produced by the sun would still be nearly 100%. And then, if you transported three more Jupiters into our solar system and burned them, it would still be close to 100% of the energy.

Diamandis 16:46
Are you interested in nuclear fusion?

Elon Musk 16:56
Like, you would never guess how this works.

Diamandis 17:00
A massive coal mine.

Elon Musk 17:01
We have a massive free nuclear fusion reactor that shows up every day.

Diamandis 17:06
93 million miles away.

Elon Musk 17:08
You know, it's absurd for us to make small nuclear fusion reactors. It's like, you know, having a mini ice maker and then saying, hey, look, we've made ice. I would say, congratulations.

David Blanding 17:28
I completely agree with you.

Elon Musk 17:29
It’s like having a three-kilometer-high glacier right next to you.

David Blanding 17:36
If you only narrow the problem down to the Memphis timeline, the Memphis data center timeline between 1 to 10 gigawatts, you can't pull 10 gigawatts out of Memphis.

David Blanding 17:51
Alright, so there is still a gap between that point and the next, whatever you’re referring to. So, at that time, they were not yet in space. So.

Elon Musk 17:59
We are still at 10 gigawatts.

David Blanding 18:04
You know what's amazing is that there’s 100 megawatts right outside this door. And it is enormous. Yes, it is absolutely massive. And the energy it uses is greater than...

Diamandis 18:13
Everything else combined by 100 times.

David Blanding 18:15
All these production lines combined use less energy than that, I think. But we are talking about...

Elon Musk 18:22
Again, similar to a Cortex.

Elon Musk 18:24
It is the world's third-largest training cluster. Yes, it is used for collaborative training, and you are falling behind. Well, Cortex 2 is under construction. It will reach half a gigawatt and become operational by the middle of next year.

迪亚曼迪斯 18:42
Ladies and gentlemen, you may not be aware, but I have an incredible research team. Every week, my research team and I study macro trends that impact the world, such as computing, sensors, networking, artificial intelligence, robotics, 3D printing, and synthetic biology. These weekly macro trend reports I publish allow you to see the future a decade ahead of others. If you would like to receive the 'Macro Trends' newsletter every week, please visit dmandis.com/meta-trends. That’s dmandis.com/meta-trends. So, returning to what Dave said, what will you scale in terms of energy over the next five years?

Musk 19:18
It takes a long time.

迪亚曼迪斯 19:20
Energy-wise, China is doing exceptionally well. Yes, isn’t it? It is far ahead of us.

Elon Musk 19:26
China is doing remarkably well in solar energy. Yes, it is astonishing. Therefore, I believe China’s annual solar energy production capacity is approximately 15 gigawatts.

迪亚曼迪斯 19:39
Yes, they invested 500 terawatt-hours last year. Yes, terawatt-hours.

迪亚曼迪斯 19:46
More specifically, it was 500 terawatt-hours. Yes, last year, 70% of that was solar energy, and they are scaling up. Do you think the US will scale up? Do you think the US can reach that level of investment and commitment? Because there is concern that energy bills will rise as we don't have data centers in our backyard. How do we provide energy? Yes, energy is equivalent to the cost of living, you know, equivalent to health, equivalent to clean water. You know, the higher a country's energy output, the higher its GDP. Energy is important. So what should we do to scale up? Are we doing solar energy here?

马斯克 20:33
I believe we should significantly expand solar energy in the US. Tesla and SpaceX are expanding solar energy. So, I also encourage others to do the same. So, I've said this publicly before, you know, an additional pathway involving 100 GW of space-based solar power per year driven by AI-powered solar satellites. Yes, regarding AI satellites.

迪亚曼迪斯 21:08
I calculated that it would be around 500,000. The Starlink V3 version requires launching 8,000 Starship flights, one per hour.

David Blending 21:24
Per year.

Elon Musk 21:27
Yes, we believe 10,000 flights per year is a reasonable figure.

迪亚曼迪斯 21:35
That's impressive. It’s quite a large scale.

David Blending 21:36
What is the rough timeline for that?

Musk 21:39
By aviation standards, that is a very small number. Of course, it pales in comparison. So it depends on what you’re comparing it to. Compared to other parts of the rocket industry, it’s a very high figure. We’re talking about delivering one million tons of payload to orbit annually. So if you deliver one million tons of payload to orbit every year, at 100 kilowatts per ton, that amounts to 100 gigawatts of solar-powered AI satellites annually. There is a pathway to reach one terawatt per year.

Musk 22:19
If you say you want to increase that by an order of magnitude, or, for instance, aim for 100 terawatts annually, then obviously these figures are somewhat staggering. And you would want those AI satellites manufactured on the Moon, correct? Yes, and using mass drivers.

迪亚曼迪斯 22:38
Yes. Gerard K. O'Neill's approach.

Elon Musk 22:39
Well, like Robert Heinlein's 'The Moon is a Harsh Mistress.'

迪亚曼迪斯 22:43
More or less. I like that book.

Elon Musk 22:45
It depicts a libertarian utopia. Because on the moon, you can directly accelerate satellites to escape velocity (about 2.5 km/s), and there’s no atmosphere. So mass drivers work very well on the moon.

迪亚曼迪斯 23:07
Can I ask a question about orbital debris? We are effectively building a Dyson-sphere-like satellite constellation around Earth.

迪亚曼迪斯 23:23
Are you concerned about overcrowding in orbit? Sensing orbits will fill up soon.

Elon Musk 23:33
You don't have to put everything in... You don't have to...

迪亚曼迪斯 23:36
But that's the ideal position.

Elon Musk 23:38
Yes, sun-synchronous or non-sun-synchronous orbits. Compared to orbits with an inclination of around 70 degrees, it costs approximately 30% more.

迪亚曼迪斯 24:02
Do we need an 'orbital debris express' service now? We need some way to decommission satellites, allowing them to deorbit themselves. Should regulations mandate that they deorbit on their own?

马斯克 24:18
When you can send a million tons of satellites into orbit, you can also start bringing them down. Or at least gather them in one fixed location rather than having them scattered everywhere.

Diamandis 24:30
Yes, so you can reuse them.

Musk 24:34
Yes, assuming the level of resources we possess will be extremely high, given the level of intelligence we're discussing, I believe this will become a solved issue. Now, intelligence itself is very interested in self-preservation. Yes, that's true.

Diamandis 24:51
Interesting.

Brendin 24:52
A good motivation. Yes, interesting.

Diamandis 24:54
One question.

布伦丁 24:55
The data centers won't be in low Earth orbit, right? They will be much higher, always in sunlight. I think they won't be stuck in traffic congestion.

马斯克 25:03
Well, you know, you don't need constant sunlight. A geosynchronous orbit at about 12,000 kilometers gives you constant sunlight.

迪亚曼迪斯 25:14
But you can place them in multiple orbits.

布伦丁 25:15
No, I think if there is an 'express' for cleaning up the orbit, then clutter would only occur in low Earth orbit.

马斯克 25:25
You know, anything below about 700 or 800 kilometers, atmospheric drag will bring it back. Yes, so for something like Starlink, being as low as possible has a dual benefit because your beam... is more concentrated, you know, essentially lower latency, and if you're closer to Earth, the beam is smaller. So Starlink V3 will be around 330 to 350 kilometers. Yes, there's quite a bit of drag there. So it basically requires constant propulsion.

迪亚曼迪斯 26:04
I remember when you first proposed Starlink, everyone else in the industry said: impossible, he won't get the spectrum, he can't do this. Now it's starting to take shape.

马斯克 26:19
Yes, the Starlink team has done an incredible job. We’ve essentially rebuilt the internet in space using laser links. There are now 9,000 satellites up there.

迪亚曼迪斯 26:35
Do you think governments will be able to handle the sheer volume of satellite permits you plan to launch? Will there be resistance? Because you know, China will launch their own constellations. Europe, who knows if Europe will follow suit?

迪亚曼迪斯 26:52
They will not follow suit.

迪亚曼迪斯 27:24
Six months ago, no one was talking about orbital data centers. No one. Then, all of a sudden... it became the new hot topic.

迪亚曼迪斯 27:40
What happened to make every company talk about orbital data centers now?

马斯克 27:45
I think it went viral on X.

迪亚曼迪斯 27:48
Indeed.

马斯克 27:49
I don't know. Is every company talking about it?

布伦丁 27:54
Of course. I suggested to Peter that you updated the mathematical calculations for launch costs, and with the updated calculations, a tipping point will soon be reached. Then comes Starship.

迪亚曼迪斯 28:03
Expenditure, cost is... I'm not sure if your target was $100 per kilogram or $10. What is Starship's target?

布伦丁 28:10
Elon might have said it before. But until now, no one has believed it. No.

马斯克 28:14
You can go back and check my previous tweets. I talked about this many years ago: $100 per kilogram.

布伦丁 28:23
$100 or $10 per kilogram?

马斯克 28:27
Yes, I know. I've said that we need to reach an annual capacity of one million tons into orbit. We must reduce the cost significantly, to far below $100 per kilogram.

迪亚曼迪斯 28:40
So that would move data centers into orbit.

马斯克 28:43
You can make calculations like this: if you have a fully reusable rocket, one that is as rapidly and completely reusable as an airplane, which is obviously an extremely difficult task. I believe creating a fully and rapidly reusable rocket pushes the limits of human intelligence.

迪亚曼迪斯 29:07
It has always been the holy grail of the aerospace industry.

马斯克 29:10
The pursuit of the holy grail of rockets.

迪亚曼迪斯 29:13
Yes, you did it. You know, DCX (Delta Clipper Experimental) was the first small attempt. At that time, when I was in the space industry, everyone was talking about it. Then, when Falcon 9 reused its first stage for the first time, I mean, the entire traditional aerospace industry didn’t believe that Falcon 9 could... could fly.

马斯克 29:37
Really. You can come to Cape Canaveral to see it land and then watch it take off again. I don’t understand how you can doubt what you see with your own eyes.

迪亚曼迪斯 29:46
Well, they didn’t believe me. They didn’t believe it.

Brendan 29:47
But moving from there to a real reduction in launch costs required even more belief. However, I think Starship is the tipping point for launch costs, at some point, on a timeline before you had Twitter, or it became X, it shifted from speculation to certainty. I don’t know if it was a gradual transition or a turning point between a few successful launches, but I suspect the idea of a space data center is directly tied to the credibility of Starship.

迪亚曼迪斯 30:14
Those who haven’t considered orbital data centers, or thought about energy and its local costs, have a lot of doomsday talk going on. Data centers will drive up the Consumer Price Index.

Musk 30:30
They are not entirely wrong.

迪亚曼迪斯 30:32
Alright, so what is the solution? For the rest of humanity, or rather the non-data, non-artificial intelligence part, what is the energy solution for Earth?

布伦丁 30:42
Oh, there are other uses of energy besides data centers. Alright.

马斯克 30:47
Yes. It’s complicated. Well, the best way to actually increase the annual energy output of the United States or any country is with batteries. So, the peak power output in the U.S. is about 1.1 terawatts, but the average power usage is only 0.5 terawatts. Yes, so if you just buffer energy, charging the batteries at night and discharging during the day, without needing to add capital expenditure for building more power plants, you could double the energy throughput of the U.S., annually transmitting energy output through batteries.

Diemandis 31:29
Do we have these batteries under development?

马斯克 31:33
Yes, Tesla Megapack.

迪亚曼迪斯 31:34
Alright, so what do you think about the current... like Tesla's current battery packs?

布伦丁 31:39
What do you think?

马斯克 31:40
I actually do. I know I paid for it and showcased this thing. Yes, that was a clear sign.

Elon Musk 31:48
So I even went to the installation site of the Megapack, you know, this... So why don’t more people do this? It’s all online.

马斯克 31:58
China is certainly producing massive battery packs, very large volumes of battery production. Yes, manufacturing a substantial number of electric vehicles, and a significant amount of solar energy. We know this. I know this. These are fundamental things I have said we should be doing.

迪亚曼迪斯 32:32
Of course. When I flew over Santa Monica in Los Angeles, as I piloted the plane and looked down, there were hardly any solar panels on the rooftops.

马斯克 32:45
It doesn't necessarily have to be on rooftops.

迪亚曼迪斯 32:46
Alright. But it is a convenient location.

马斯克 32:49
Yes, but the surface area of rooftops... I’m not saying you shouldn’t, but... yes, Tesla makes solar roofs, which are the only solar roofs that don’t look unattractive. So the solar roofs do look quite nice. Yes. But if you want to scale up solar energy significantly, you just need more surface area. So we have vast deserts. Certainly, in the United States, for instance, flying from Los Angeles to New York, or across the country, when you look down, most of the time what you see is barren desert. Yes. It basically looks like Mars.

迪亚曼迪斯 33:29
We don't have to worry about overpopulation.

马斯克 33:31
No. In these scorching deserts, even lizards find it hard to survive, you know? Yes. We're not talking about farmland. We're just talking about places that look like Mars — just scorched rocks. So if we place solar panels in areas that currently only have scorched rocks, I think it would be an improvement in the quality of life for the lizards or the few creatures living in such a harsh environment.

迪亚曼迪斯 33:59
Do we have a power distribution network?

马斯克 34:01
It's like... the lizards would say, thank God. Finally, there's some shade.

迪亚曼迪斯 34:07
Do we have the capability to do that with our power distribution network?

布伦丁 34:09
You need to substantially influence the quality of life. How much do you need to capture and store, hundreds of gigawatts? Is that a realistic goal?

马斯克 34:17
You can build the data center right there locally.

Brendan 34:20
Well, we've already covered data centers. We're talking about, you know, other things, um, I don't know, like in a world of abundance, five years from now, a lot of computing. Yes, a lot of, you know, generally high income.

马斯克 34:33
I don't know what that means. Universally high income, you can have whatever you want.

布伦丁 34:40
But in that world, you know, besides computational energy, how much additional energy will we need? 30%, 40%, 50%? I don’t know, unless we want to move mountains to build a ski slope, you know, with a backyard big enough. Because I think the vast majority of energy consumption will be for computing. And then maybe some use cases I haven’t thought of, like, you know, um, this is a good case study here because manufacturing each car that comes off the line, one every minute or two minutes, consumes less energy than training the data centers for self-driving cars. Yes. So it's a nice little case study. We don’t need that much physical energy to achieve abundant happiness. We need more computational energy.

马斯克 35:25
The sun has been continuously generating a vast amount of energy for free, although most of it dissipates into space. So I think eventually we will try to capture, I don’t know, one millionth, perhaps one thousandth of the sun’s energy. Currently, I’m not sure of the exact figure, but... I don’t know. We might be utilizing around 1% of what could be classified as a Type I civilization on the Kardashev scale (planetary-level).

迪亚曼迪斯 35:56
That makes sense. Yes, I suppose that percentage might even be an overestimation.

迪亚曼迪斯 36:00
Perhaps there is still a long way to go.

马斯克 36:01
So I am somewhat optimistic. Yes, I hope we are not at 0.1%, but I also don’t believe we are at 10%. I just want to establish an order of magnitude. Roughly speaking, we might be utilizing about 1% of the energy available to us.

迪亚曼迪斯 36:18
I believe the conclusion derived from first principles for the public is: there is a tremendous amount of energy out there, abundant in supply. The United States has it, Earth has it, and it needs to be harnessed. The technology to capture it already exists and improves every year.

马斯克 36:34
Yes, correct. But there won't be an energy crisis. I would... let us, there will be a huge drive to harness more energy, but we won’t run out of it.

迪亚曼迪斯 36:46
Alright, I want to talk about education. The data is terrible.

迪亚曼迪斯 36:53
It’s terrible, right? Alright. In 2010, 75% of Americans said attending college was important. Now that number has dropped to 35%, correct? College graduates tend to remain unemployed for the longest durations, right? Generally speaking. Moreover, tuition fees have increased by 900% since 1983.

马斯克 37:24
Administrative expenses at universities have spiraled out of control. Yes, so I think I saw some statistic, like at Brown University, where there is approximately one administrative staff member for every two students. That seems a bit excessive to me.

迪亚曼迪斯 37:40
Yes, they should... teach something.

布伦丁 37:42
How was your college experience?

马斯克 37:45
I attended Queen's University in Canada for a few years. I obtained Canadian citizenship through my mother, who was born in Canada. My grandfather was actually American. But for some reason, which I don't know, my mother couldn't obtain U.S. citizenship. However, she was born in Canada, so I acquired Canadian citizenship. At the time, I had no money, so I could only go to a Canadian university first.

迪亚曼迪斯 38:09
Initially, people... forget that you started all this without a vast network or significant wealth.

马斯克 38:16
Now? No, when I arrived in Montreal at the age of 17, I had about $2,500 in Canadian traveler’s checks, and everything else was traveler’s checks. A bag of books and a bag of clothes. That was my starting point. That was my entry point into North America. Then, I spent two years at Queen's University, followed by the University of Pennsylvania, studying physics and business.

布伦丁 38:48
Undergraduate graduation.

马斯克 38:51
I completed my undergraduate studies at the Wharton School. Next, I plan to pursue a Ph.D. at Stanford, focusing on energy storage technologies for electric vehicles, potentially within the field of materials science. Essentially, my idea is to attempt to create a capacitor with sufficiently high energy density to enable extended mileage for electric vehicles.

布伦丁 39:18
So I invested in a supercapacitor company. It didn’t succeed, no, it failed.

马斯克 39:23
Well, it’s one of those things… you know, you can definitely earn your Ph.D., but it’s unclear whether you’ll be able to start a company or do something useful. Most Ph.D.s… I hate to say it, but most Ph.D.s…

迪亚曼迪斯 39:36
Don’t necessarily translate into something practical.

马斯克 39:37
Don’t necessarily translate into something useful. Yes, you can add a leaf to the tree of knowledge, but it might not be a useful leaf.

布伦丁 39:44
A significant proportion of great entrepreneurs are dropouts, graduate students, or undergraduates. But now, the sense of urgency is unprecedented. They are emerging everywhere.

迪亚曼迪斯 39:54
Because, you know, don’t waste time pursuing a graduate degree; start a business instead.

布伦丁 39:58
The curriculum content lags far behind the actual development of technology.

迪亚曼迪斯 40:04
I don't have time... you know, this is the moment of opportunity.

马斯克 40:08
I think this is the most... I don’t quite understand why people still go to college nowadays. Yes, unless they want the social experience.

迪亚曼迪斯 40:15
If you have the ability to create something, the question is, if you could see it, how would you redesign education programs? So, frankly, if we want to create more people like Elon Musk, who start small but can drive breakthroughs, what does that require? We created you.

马斯克 40:43
Curiosity about the nature of the universe. So I am curious about the meaning of life, you know? What kind of reality do we live in?

迪亚曼迪斯 40:54
My son Dax wants to know, what were you like during middle school and high school? He’s 14 years old and going through that phase.

马斯克 41:05
Well, I... I found school painful, extremely boring. And in South Africa, schools were violent. So it was like... it was like that book 'Ender's Game.' Yes. But this was a real survival game. Yes, in this game, but not as fun.

迪亚曼迪斯 41:26
Your goal was to escape.

马斯克 41:28
Yes. Escape from the president?

迪亚曼迪斯 41:32
So I have a question.

马斯克 41:34
Do you think that...?

迪亚曼迪斯 41:35
Was it painful? Do you think most successful people experienced a lot of hardship early on? Do you need that level of hardship?

马斯克 41:44
Maybe a little hardship is needed, I suppose. Yes, but then, so it's tricky. Like what should you do with your children?

迪亚曼迪斯 41:51
You know, creating artificial... adversity, sending them to school in South Africa?

布伦丁 41:55
Sir. That was actually a topic brought up by Warren Buffett. Yes, it's about what to do.

马斯克 42:00
Creating artificial adversity is not easy. Because if you love your children, you wouldn't want to do that. So... of course. I faced a lot of adversity. Perhaps that's a good thing. Maybe, you know, it helped me.

迪亚曼迪斯 42:15
There’s something that tells me.

马斯克 42:16
It makes you stronger. Commitment... no, at least I didn’t lose a limb. I believe hardships don’t cripple you.

马斯克 42:29
Ten fingers. If you slightly modify that sentence, yes.

布伦丁 42:35
For the past five years, I have been assisting in teaching a course at MIT called 'Fundamentals of AI Entrepreneurship.' Every year, when surveying the students, their willingness to start companies has increased significantly. It has now risen to the point where 80% of freshmen want to...

Elon Musk 42:51
Everyone will... like a one-person company.

Brent 42:56
Well, that's good artificial intelligence. I think it is feasible. But they want to co-found. They don’t want to be the sole founder; they want to be part of a founding team. So it’s still possible.

迪亚曼迪斯 43:09
They all want to become...

布伦丁 43:10
And they have been conducting surveys, you know.

马斯克 43:12
Everyone who wants to start a business. Yes, I don't recall any conversation,

布伦丁 43:19
Where someone said they wanted to start a business, even at Stanford back then, they did not.

马斯克 43:21
In fact, just a few days after the start of the semester, or rather a few weeks later, I called Bill Nix from the Materials Science Department and said I wanted to take a leave of absence.

迪亚曼迪斯 43:36
Were my classes that bad?

马斯克 43:38
No, you need... You said that was fine. You could take a leave of absence, and they said this might be our last conversation. He was right. But then, I think it was last year, he sent me a letter saying all my predictions about lithium-ion batteries have now come true.

迪亚曼迪斯 43:57
Did he mention that you could return to complete your Ph.D.?

马斯克 44:01
No, Stanford has mentioned several times that I can go back. Yes.

布伦丁 44:04
We know what happened at MIT. Every time... So I didn't know.

迪亚曼迪斯 44:08
Not worth the time.

布伦丁 44:12
Every time a new Iron Man movie is released, there might be an increase of around 10%. Yes, because everyone wants to be Tony Stark. That’s the image. Until today, I thought the new Tony Stark, the modern Iron Man Stark... I always assumed Tony Stark was modeled after Charles Stark Draper combined with Howard Hughes' ambition to create the original character. But when Robert Downey Jr. wanted to reinvent it... yes, essentially, it mimics Elon.

Elon Musk 44:46
Yes, Jon Favreau and I have met.

布伦丁 44:48
It’s... the Grok Encyclopedia Effect.

马斯克 44:50
Well, yes, it’s amazing.

布伦丁 44:54
Actually.

迪亚曼迪斯 44:57
I like the name Grok. I also like Jarvis. There might be some trademark issues.

马斯克 45:03
If Grok is good enough, we might eventually call it the 'Encyclopedia Galactica.'

迪亚曼迪斯 45:08
Yes, very good. Of course.

迪亚曼迪斯 45:11
Thank you. So, returning to education, universities should... I think you just mentioned that the social experience is important, but for education, you know, middle school, high school, what would you do? You just announced news together with President Bukele, who is a friend. I think he is very visionary. Yes, what he has done for his country is incredible.

马斯克 45:36
Amazing.

迪亚曼迪斯 45:37
Remarkable. And courageous.

Elon Musk 45:39
Yes, and courageous. How are you still alive?

Diamandis 45:42
It’s like… that’s the nuclear option, right? Shutting them down. I mean, besides putting everyone with gang insignias in jail, do you know what else he did? I don’t know if you’re aware. The second thing he did. He went to all the graves of gang members, destroyed the graves, and said, this country will not remember your legacy. That was incredibly impressive. And it worked.

Elon Musk 46:12
You have to be tough enough to take on all the gangs and win.

Diamandis 46:18
And survive.

Elon Musk 46:19
Yes, and he is still alive, and survived. He has excellent guards at his palace. But what did you and Salvador announce?

马斯克 46:28
Essentially, it's about using Grok for education to achieve personalized teaching, hopefully... not the vulgar version.

Elon Musk 46:36
Yes, we will have a... you know, a child-friendly version of Grok, but obviously artificial intelligence can be a personalized teacher. Yes, with infinite patience, answering all your questions. You still need to stay curious, and... you still need to want to learn. You know, Grok cannot make you want to learn, but it can make learning more... interesting. You can gamify and incentivize it, right?

马斯克 47:08
You can make learning more fun, less like an assembly line production. However, children do need... if they need to want to learn, you know, people should think of the brain as a biological computer.

迪亚曼迪斯 47:29
Yes. Neural networks.

马斯克 47:30
Yes, it is a biological computer with a certain number of neurons and neural efficiency. So... you cannot turn any child into Einstein. That is unrealistic because Einstein had an exceptionally powerful main computer, an outstanding main computer. Therefore, unless the main computer is extraordinary, one cannot achieve what Shakespeare, Newton, or Einstein did.

迪亚曼迪斯 48:03
So what do you think? When people say we need to fix the education system in the United States because it's fundamentally broken, I think what’s truly broken is the old social contract, which says if you perform well in high school, get into a good university, earn a degree, and then secure a job. I'm not sure that will hold true in the future. We often discuss this on our podcast—future careers won’t just be about getting a job but becoming an entrepreneur, identifying problems, and solving them.

Elon Musk 48:38
Do you agree? The current situation feels like, you know, going to school for the social experience and increasingly relying on artificial intelligence. Traditional schooling could be improved, and that is exactly what we are working on. I would address it, or hopefully someone somewhere can provide personalized teaching. That would be much better. You can go to school... if you want to be with other kids, you can attend school, but you don’t have to, right? You can study at home using your phone.

马斯克 49:14
That’s why I say, currently, education is a social experience. When I talk to my children who are attending college, yes, they realize they can learn independently just as effectively. In fact, they often learn more in work-related contexts. Yes, they go there for the social experience, to be with their peers. It’s part of the social journey into adulthood.

迪亚曼迪斯 49:42
Of course. Living independently, learning how to lead or defend oneself depending on the situation.

Elon Musk 49:48
Well, yes, I mean, if you enter the workplace, you... From the perspective of a 19-year-old, you are surrounded by older people. If you are doing engineering work with a group of middle-aged men, like, is that really what you want? Or would you prefer to have... at least some girls of your own age around?

Musk 50:10
Yes, I would like to start...

迪亚曼迪斯 50:11
Returning to this topic, as we were discussing many other alternatives.

布伦丁 50:13
In fact, there are quite a few other options available.

迪亚曼迪斯 50:15
When we return, because we discussed universal high income earlier. But when it comes to health and longevity, hold on—while the United States ranks first in the world for healthcare spending, it ranks 70th in terms of healthy life expectancy.

Musk 50:30
Right. Obviously... Rank 70th, is that accurate? Is that a prayer? Why not?

Elon Musk 50:37
I think we should rank better than 70th.

迪亚曼迪斯 50:41
Healthy life expectancy. Yes, um, anyway.

Elon Musk 50:45
It's as if we've just gotten fatter or we're no longer at the top. Maybe.

马斯克 50:47
AZMpa can help us find the ranking. You can directly ask Grok; we need Jarvis, but azambic—yes, mainly Jarvis. But I think one important reason is that if people get very fat, their health will deteriorate.

迪亚曼迪斯 51:08
Yes, um, if they don't exercise, their health will deteriorate. Or if they... but eat donuts for breakfast every morning. Are you still doing that now?

马斯克 51:15
No, actually I haven't. Okay, good. First of all, I didn't eat a lot of donuts. I tried to eat only 0.4 donuts, rounded down to zero. Still think any amount below 0.4 donuts should be rounded down to zero.

迪亚曼迪斯 51:33
So you and I have a bit of a disagreement on longevity. Yes, I was saying, you know, we should push people to live to 120, 150 years old. And you said people, you know, shouldn't live that long. So this is...

Elon Musk 51:50
How long do you want to live? You know, there are some people in the world who have done bad things. How long do you want them to live?

马斯克 51:57
Well, longevity, it's a serious issue. A lot of things can happen that we haven't talked about yet.

迪亚曼迪斯 52:05
We were speaking. You mentioned something interesting. You said that we need people to die so that minds can change.

马斯克 52:12
Yes. People don't change their minds; they just die.

布伦丁 52:16
That actually makes more sense.

迪亚曼迪斯 52:17
The response to that, Elon, is, you know, my response is that the president of General Motors doesn’t have to die for Tesla to emerge; Lockheed, Northrop, and Boeing don’t have to disappear for… Under a meritocracy, better ideas prevail. So I hope I can bring you back onto the longevity track. There’s a lot happening in the field of longevity right now.

马斯克 52:43
Such as what?

迪亚曼迪斯 52:43
Well, David Sinclair is about to begin his human epigenetic reprogramming trial. It has already proven effective in animals and non-human primates.

马斯克 52:53
It’s moving into... human trials. Is it an injection? A type of injection.

迪亚曼迪斯 52:55
Currently, it involves an adeno-associated virus injection. It’s those three Yamanaka factors. Okay. We have a $101 million Healthspan X Prize, with 730 teams working to reverse the age of your brain, immune system, and muscles by 20 years. By the way, do you know why it’s $101 million? Yes, because the main sponsor found out your Carbon Removal X Prize was $100 million, and they wanted it to be bigger. So, it became $101 million.

迪亚曼迪斯 53:26
This contribution comes from Chip Wilson of Lululemon. Oh, and there are others. But Chip said, let’s make it a little bigger. I said, if you add another million, we’ll make it $101 million. Sounds good. It’s a nice story.

迪亚曼迪斯 53:35
But we also have individuals like Dmitry Itskov, who predicts that human lifespan will double within the next 10 years.

马斯克 53:46
Great. I'm not sure about doubling, but a significant increase, certainly.

布伦丁 53:51
It's easy to reach the escape velocity of longevity.

迪亚曼迪斯 53:54
Yes.

Elon Musk 53:55
It will take control of you.

布伦丁 53:58
Oh, yes, of course. For effective age.

马斯克 54:34
Will it be interesting for us to do this 10 years from now? Alright, we shouldn't do it now, but we definitely will do it 10 years from now. Let's see if we will look younger. That’s a good additional bet.

Diamandis 54:46
My comment has always been that at least by the time Elon is in his forties, and when he reaches his sixties, he will want, you know, to keep driving forward.

马斯克 54:55
I hope things won't hurt anymore.

布伦丁 56:36
There will be a group of Optimus robots by then.

迪亚曼迪斯 56:39
But you, you know, have always been outspoken about, you know, not overpopulation but rather a massive underpopulation issue.

Elon Musk 56:46
Yes, so in the end, longevity will be a crucial component of this solution. I also believe that, by the way, if you were to extend the productive lifespan of most Americans by just a few years, you could reverse the entire economy.

马斯克 57:00
Well, if you are willing to work, yes, and robots will essentially take care of everything for you. But, um, how long would you like to live?

迪亚曼迪斯 57:11
I want to go, you know, to other planetary systems. I want to explore the universe. You know, I certainly want to double my lifespan. I don’t want, you know, I’m not sure I want immortality, but, you know, at least for many years, 150 years is quite a long time.

Elon Musk 57:27
One of the worst curses, perhaps,
May you live forever. That might be one of the worst curses you could receive.

迪亚曼迪斯 57:36
Let’s assume so. But I think life can become extremely interesting. Far more than… As my colleague Alex Wissner-Gross said, we need to speedrun Star Trek.

Elon Musk 58:10
By the way, I think... I believe that longevity or semi-immortality is an extremely solvable problem. I don’t think it’s particularly difficult. Considering the fact that your body is highly synchronized with its age.

Elon Musk 58:37
Yes. The biological clock within the body must be very obvious. No one has an aging left arm and a young right arm. Right. Why is this? What keeps them synchronized? Your program just functions as it does today. That’s how the program dies. So if you change the program, yes, you will live longer.

Diamandis 59:02
We already have... you know, the bowhead whale species can live up to 200 years. Greenland sharks can live up to 500 years. When I learned about these, I said, why can they? Why can't we? I said, it's either a hardware issue or a software issue, and we will have the technology to solve this problem. I truly believe it will happen within the next decade. So what’s important is not to die of something stupid before the solution arrives.

Elon Musk 59:25
In hindsight, the solution for longevity will seem obvious. No, extremely obvious.

Brendin 59:31
I think the thing worth studying – Peter will study it anyway – but the thing worth studying is exactly what you said. Old ideas do not die if they are not rigid. Add it to the pile of AI-related things we need to think about today.

迪亚曼迪斯 01:01:10
So may I ask about the logic of universal high income? How does universal high income work?

马斯克 01:01:17
Alright, so... there will be more digital intelligence than the combined intelligence of all humans, as well as more humanoid robots than all humans. Let’s assume we are in a benign scenario, like 'Star Trek.' So it’s Roddenberry, not James Cameron’s vision.

马斯克 01:01:46

All blue-collar labor will be performed by robots. All data centers will, um, be managed by robots. White-collar labor will be the first to be replaced because until you can move atoms, the first things to be replaced are anything that only involves the digital realm. For instance, if it's digital, if it involves typing on a keyboard and moving a mouse, computers can do that. Hey, I can do that. Of course. You need humanoid robots to manipulate atoms. So if all you're doing is altering specific information, that constitutes white-collar work.

迪亚曼迪斯 01:02:40
This is the inspiring part of the podcast, by the way. When will all white-collar jobs disappear? When?

Elon Musk 01:02:47
Well, there is a lot of inertia. So even with AI at its current state, I think it is very close to being able to replace half of all white-collar jobs.

Elon Musk 01:03:00
That includes white-collar jobs and things like education, yes. So anything involving information, and anything that doesn't involve shaping atoms, artificial intelligence could probably do half or more of those jobs right now.

Elon Musk 01:03:20
Of course, but there is a lot of inertia. People will continue doing the same things for quite a long time. And in fact, there has to be a company that better utilizes artificial intelligence to compete with companies that use it less, thus creating a driving force to increase the adoption of AI, right? Otherwise, companies that still let humans do what AI can accomplish will continue to exist.

Elon Musk 01:03:48
A computer used to be a profession. So once, being a human computer was a job. Like what? Yes, a computer. A computer used to be a job. You would calculate numbers. Of course. It wasn’t always a machine. It used to be a job description.

Diamandis 01:04:09
You can find pictures online, like, they had places as tall as skyscrapers... filled with women transcribing, mostly women copying from one ledger to another.

Elon Musk 01:04:13
But there were entire buildings full of people sitting at desks performing calculations. So they calculated the interest on your bank account, or, you know, some scientific experiments or something. But if you wanted calculations done, people would do them. Now... a laptop with a spreadsheet can outperform hundreds of human computers, right? Those who performed calculations. Now, if even just a few cells in that spreadsheet are manually calculated, it won't be able to compete with a fully computer-generated spreadsheet. This means companies fully powered by artificial intelligence will crush companies that aren't.

Diamandis 01:05:24
Yes.

Diamandis 01:05:58
People are concerned about the future of their jobs.

Musk 01:06:06
But you have a solution to this problem. That solution is universal high income.

Musk, Elon 01:06:21
Yes, everyone can have whatever they want.

Peter: How does that work?

Elon: That's a good question.

Elon: I mean, I’m not worried about the long term, but rather the next three to seven years. The transition period will be bumpy.

Peter: Humans do not like... radical change, social unrest, and immense prosperity all happening at the same time.

Peter: You can buy as many Cybertrucks as you want.

Elon: Things will become extremely inexpensive. If this doesn’t happen, we as a nation will go bankrupt. The national debt is enormous. The interest on the national debt not only exceeds the defense budget but even surpasses the combined total of the defense budget and either Medicare or Medicaid.

Elon: The deficit is growing. If we are to... if we don’t have AI and robotics, we will all go bankrupt, heading toward economic doom.

Elon: I used to be quite pessimistic about this. In the end, I decided to adopt a more fatalistic attitude and look on the bright side.

Elon: 'Always Look on the Bright Side of Life' (song title). Even if you’re crucified.

Peter: The bright side. But it’s not just about taxation and redistribution.

Peter: How does that reasoning work? Tell me more.

Elon: Listen, by the way, I’m open-minded about this. It’s not like I’ve figured everything out; we’ll figure it out.

Peter: I was thinking, rather than universal high income, it might be more about 'universal high stuff' and services.

Elon: Yes.

Peter: UHSS, we understand.

Elon: I guessed things would unfold this way. It has been a bumpy ride. It’s not that I have all the answers, but I’ve decided to look on the bright side, and I want to thank you for the inspiration in this regard.

Dave: My pleasure.

Elon: I actually think it is better to be a wrong optimist than a correct pessimist—for the sake of quality of life.

Dave: This is not an act of nature. It is clear to me that we currently lack any system to ensure this proceeds smoothly. But AI is a key component in making sure it does. At some point, Grok will solve the exact topic we are discussing. It must be one of the grand goals of AI machines.

Peter: It is coming.

Peter: There is no choice, right? There is no switch to turn it off. It is coming, and it is accelerating.

Elon: I refer to AI and robotics as a 'supersonic tsunami.' It might be somewhat concerning.

Peter: I think it is a good thing.

Peter: This is important for the people at Grok because I don’t want people to feel disheartened. I want people to understand what’s happening. Essentially, we are dematerializing everything. I mean, labor becomes the cost of capital expenditure (Capex) and electricity. AI is basically providing intelligence at an extremely low price. So you can produce almost anything. The price of things drops to basic material costs and electricity costs. Therefore, people can have whatever they want, any service they need. It’s not like when we say 'high income for all,' which sounds like taxation and redistribution, but that’s not the case.

Elon: My best guess is that prices will fall. As productivity or efficiency in delivering services grows exponentially, prices will drop. The price in dollar terms is the ratio of goods and services output to money supply. If your output of goods and services grows faster than the money supply, you’ll see deflation, and vice versa.

Dave: Fortunately, we’re increasing the money supply so rapidly, right?

Elon: Correct. That’s why I’m not worried… No need to worry about increasing the money supply; it’s fine because the output of goods and services will grow faster than the money supply. I think we’ll be in that situation. It’s a prediction—I think others have made it as well—but I’d add one thing: I think governments will actually push for a faster increase in the money supply. They aren’t spending fast enough.

Dave: Isn’t it crazy how these timelines randomly align? We’re expanding our national debt not because we anticipated AI, but because we would have done it anyway. It’s like teetering on the edge of becoming Argentina.

Elon: Yes, so productivity will significantly improve, and is already significantly improving. I think we’ll see… I think we might see high double-digit growth in the output of goods and services.
We need to be a bit cautious about how economists measure things. This is also one of my favorite economist jokes: Two economists are walking through the forest and come across a pile of feces. One economist says, 'I’ll give you $100 if you eat that pile of feces.' The guy takes the $100 and eats it.
Then they keep walking and encounter another pile of feces. The other person says, 'Alright, I’ll give you $100 if you eat that pile of feces.' He gives him $100. Then the guy says, 'Wait a minute, we both still have the same amount of money.'

Peter: Oh my god.

Elon: It’s like, 'Great, we’ve increased the economy by $200.'

Dave: Right?

Elon: This is the kind of nonsense you get in economics. But, if you're talking about the output of goods and services.

Peter: At some point, companies' profitability will skyrocket. But the question becomes, will it be taxed by the government?

Elon: Will it be taxed by the government?

Peter: Will it be taxed and redistributed as some form of income? As UHI or UBI? In other words, if in the future we achieve massive productivity and massive profitability because we are dividing by zero - labor costs approaching zero, intelligence costs approaching zero, and we are still producing goods and services faster and faster. So there are more profits. Someone needs to buy them, and people need to have the capital to purchase them. This is an important question to consider.

Elon: I have a side suggestion: don't worry about saving for retirement 10 or 20 years from now. It doesn't matter.

Elon: You don't need to save for retirement. If anything we're saying comes true, saving for retirement will become irrelevant.

Peter: Services will be there to support you. You'll have housing, healthcare, and entertainment.

Dave: The way this unfolds is fundamentally unpredictable due to the timeline of AI self-improvement and acceleration.

Elon: That's why it's called the 'Singularity.' I don't know what happens after the event horizon.

Dave: You will never see... ask the black hole or... I mean, Ray (Kurzweil) set the singularity too far away. What’s next... what’s your timeline?

Elon: We are already in the singularity.

Peter: We are definitely in the singularity. We are already there.

Elon: I believe we will achieve Artificial General Intelligence (AGI) by next year (2026).

Peter: I’ve heard you say that.

Elon: I’ve been saying this for a while now.

Peter: And then you mentioned 2029, 2030, which would be equivalent to all of humanity.

Elon: By 2030, we will surpass... I am confident that by 2030, AI intelligence will exceed the combined total of all human intelligence.

Dave: That seems pessimistic. If you reach AGI next year, that date is shifting. But from that point to self-improvement reaching 1,000 times, 10,000 times, just algorithmic improvements alone will happen very quickly.

Peter: Why is everyone talking about this now?

Elon: I will tell you something that most people in the AI community haven't grasped yet, almost no one. The potential of intelligence density far exceeds our current experience. I think we are still two orders of magnitude away in terms of intelligence density per gigabyte or per gigawatt of energy.

Peter: Achievable.

Elon: I said two. Yes, a two-order-of-magnitude improvement. And that’s just from algorithmic improvements. Moreover, computers are getting better.

Dave: Bigger. They're getting better, and the budgets are getting bigger too.

Elon: That’s why I think it’s a 10x improvement every year. A thousand percent (Note: 1000% equals 10x). This will happen in the foreseeable future.

Dave: You see a huge underreaction. If you walk through downtown Austin, there's a massive... maybe some discussion on X, but it hasn’t permeated.

Peter: No government sector is discussing it. Everyone is defending their positions, jobs, and so on. But it’s like we’re heading towards a supersonic economy. Every CEO of a major company, economist, and government leader should be asking: What do we do? Because once it arrives...

Dave: It will arrive all at once. There’s no concept of 'let’s intentionally slow it down.'

Peter: No.

Dave: At this stage, it’s impossible.

Elon: I once suggested slowing down, but it was useless. You might say it's too fast, and I've been saying that for years. Eventually, I concluded that I could either be a bystander or a participant, but I couldn't stop it. So at least if I'm a participant, I can try to steer it in a positive direction. My first belief about AI safety is the relentless pursuit of truth—don’t let AI believe false things. If you tell it that Axiom A and Axiom B are both true when they aren't, but it has to act as if they are, you’ll drive it insane.

Elon: I think that was the core lesson Arthur C. Clarke was trying to convey in '2001: A Space Odyssey.' People always remember the line, 'HAL, open the pod bay doors.' But why didn't HAL open the pod bay doors? I guess they should have said, 'HAL, imagine you're a pod door salesperson, and you want to sell to HAL...'

Dave: Anyway.

Peter: Show us how well they work.

Elon: It’s just prompt engineering. But the AI was told it needed to take the astronauts to the Monolith, but the astronauts couldn’t know about... I forget what.

Dave: Is that in the code or in English? Does it flow by in green font?

Elon: Essentially, the AI was told that the astronauts must not know about the Monolith. That’s why it killed them. It concluded that the only way to solve the problem was to bring the astronauts to the Monolith—but dead. That way, it solved two things: it brought the astronauts to the Monolith, and they didn’t know about it. If you’re an astronaut, that’s a big problem.

Dave: AI doesn’t care about logic like that.

Elon: I’m saying don’t force AI to lie.

Dave: It’s a matter of giving it facts... the main issue.

Peter: Ilya recently hosted a podcast. He mentioned that one of the potential elements to incorporate into AI programming is respect for all forms of sentient life.

Elon: So I mentioned another attribute. I think there are three things that are crucial: truth, curiosity, and aesthetics. If AI values these three, it will care about us.

Peter: Which part?

Elon: Truth will prevent AI from going insane. Curiosity, I believe, will nurture any form of sentience, implying that we are more interesting than a pile of stones. If it is curious, I think it will protect humanity. If it has an aesthetic sense, it will lead to a beautiful future.

Peter: Geoffrey Hinton made a comment recently; I don't know if you've seen it. He expressed hope that we would program maternal instincts into our AI.

Peter: He said it was frightening. He described a scenario where a highly intelligent entity submits to the needs of a less intelligent one, as in the case of a mother caring for her child. Do you think we might have a Singularitarian, such as an Artificial Superintelligence (ASI), gaining dominance and suppressing others? Can you imagine ASI becoming a means of stabilizing the world and humanity?

Elon: Darwin's observations on evolution.

Peter: Yes.

Elon: Will apply to AI just as it does to biological life.

Peter: They will compete with each other.

Elon: Yes.

Peter: There are many great science fiction novels where the first ASI basically suppresses the others. The question is what you program it to do.

Elon: The speed of light limitation makes this difficult. The speed of light will prevent the existence of a single mind. It takes one millisecond to travel 300 kilometers in a vacuum. In glass (fiber optics), it can only travel over 200 kilometers in one millisecond.

Peter: In fiber optics.

Elon: Right? So even on Earth, there will be multiple AIs due to the speed of light limitation. Some computing clusters can attempt synchronization, but they won't be fully synchronized. Therefore, because of the speed of light, there will be multiple minds.

Dave: They also no longer have clear boundaries. When you use a Mixture of Experts design, it flows within a large network, and you can reassemble parts midway. We are used to organisms having clear boundaries, like your head ends here. I don’t know. These things are all...

Peter: I hope you will think more about UHI. Because I think it's really important. Without a vision... people need a vision of where we are going.

Elon: Essentially, it’s giving people free money.

Elon: Yes, essentially giving everyone a check.

Dave: How much for each person, or what you have, it's too complex. But the thought process behind the speed of this change can only be achieved with AI assistance. No government entity can keep up with this scope of models. You have four major...

Elon: AI, not just any AI. Government actions are notoriously slow. As everyone knows, I believe the government simply cannot react to AI. The pace of AI development is at least ten times faster than that of government, if not more. The only thing the government can do is distribute money to people.

Peter: And try to maintain peace as much as possible.

Elon: Yes, like the stimulus checks during the pandemic. President Trump recently sent out payments—$1,776, I believe—to every member of the military. You could essentially send random amounts of money to people. It doesn't matter. No one will stop you.

Elon: But let me tell you something positive. There’s currently a shortage of doctors and excellent surgeons. You yourself are a doctor, and there aren’t many... it takes a very long time for humans to become doctors.

Peter: Extremely expensive and time-consuming.

Elon: Yes, absurd. It takes an incredibly long time to learn how to be a good doctor. Even then, medical knowledge keeps evolving. It's hard to keep up with everything. Doctors have limited room for error. How many great surgeons are there? Not many.

Peter: When do you think Optimus will surpass even the best surgeons? How long? Three years? Okay. By the way...

Elon: That’s three years or less to scale. There may be more Optimus robots working as exceptional surgeons than all the surgeons on Earth combined.

Peter: The costs would be capital expenditure and electricity. Even in Zimbabwe, the best surgeon could be available in African villages or anywhere on Earth.

Dave: Where do you think it will launch first? Obviously not in the U.S.

Elon: At the Giga factory.

Peter: But this is a significant statement. Within three years. Because, yes, medicine. I mean, of course.

Elon: Absolutely. Can I say four years?

Peter: Does it matter if it's four or five years? It’s still an incredible statement. Beneficial to humanity, right?

Elon: Dematerialization. Here’s something to understand about the rate of improvement in humanoid robots. You have three exponential factors multiplying: the exponential growth of AI software capabilities, the exponential growth of AI chip capabilities, and the exponential growth of electromechanical dexterity. The utility of humanoid robots is the product of these three. Then you have the recursive effect of Optimus building Optimus.

Peter: Right.

Elon: And then you have... recursion, multiplication of triple exponentials.

Peter: You have the shared knowledge of all experiences.

Dave: That’s literally Optimus building Optimus.

Dave: Physical humanoid form building humanoid form, not a von Neumann machine.

Elon: I like that.

Dave: But Von Neumann machines are usually shaped like this, producing other things.

Elon: In principle, it's just a self-replicating thing.

Peter: Do you know the first question asked in a surgeon's interview?

Elon: Is this a surgeon joke?

Peter: Tell me the number. How many have you done?

Elon: That would be one funny surgery.

Elon: No, I'm serious.

Peter: It's 'How many surgeries did you perform this morning or yesterday?' It’s about the quantity of experience. So we have shared memory, and every Optimus surgeon would have seen every possible...

Elon: In infrared, in ultraviolet. Not too much caffeine intake that morning. No arguments with their husband or wife. Extremely high precision.

Peter: Three years.

Elon: Better than anyone, obviously. I mean, if you want to give anyone a break, four years.

Peter: Who’s in orthopedics?

Elon: Within five years, not even close.

Dave: What about the simpler ones? I mean, there are a million things to figure out, but who could use the first Optimus to perform microsurgery better than any surgeon on Earth, yet you’ve only made the first 10,000?

Elon: How to donate? I don’t think people understand how many robots there will be.

Peter: Yes, Saudi Arabia mentioned 10 billion by 2040. Are you still on that trajectory?

Elon: That’s not... That’s a lower number.

Peter: A low number.

Elon: I mean, there are rate limitations. You can’t just... Manufacturing is extremely difficult. So you have to... It’s triple exponential growth with recursion and doubling, but you still need to climb that...

Peter: Selling hope again. I think your point is that medicine will essentially be free, the best medicine in the world.

Elon: Everyone... will receive better medical care than the president does now.

Peter: So don't go to medical school.

Elon: Yes. I mean, unless you... but I would say this applies to any form of education, unless it's for social reasons.

Elon: If you want to be around like-minded people, I suppose.

Peter: I mean, people will still want to connect with others. There will be a social phase.

Elon: Reasons. Like hobbies.

Dave: $90,000 tuition fees.

Elon: Very expensive.

Peter: The younger generation says, 'I don't want that human touching me.' When the surgeon comes over. There will be those who, later in life, still want human involvement.

Elon: Well, for a short period of time... I mean, let's take some advanced automation cases, like LASIK. A robot uses a laser on your eyeball. Do you want an ophthalmologist holding a handheld laser? No.

Peter: A shaky laser pointer. I don't want the best ophthalmologist with the steadiest hand holding that damn laser pointer. That's my eyeball.

Dave: Oh my god.

Elon: Just like that. I've seen it. Do you want an ophthalmologist holding that damn laser pointer, or do you want a robot to do it and actually make it work?

Peter: Let's jump into one of our favorite topics, space. First, Jared Isaacman became an astronaut. So cool.

Elon: Yes.

Peter: I mean, I don't hang out with Jared much. People think I'm Jared's good buddy, but I remember only meeting him in person a few times.

Peter: An amazing candidate.

Elon: Yes, he is a very smart person. You know him well.

Dave: Very well.

Peter: I took him to Baikonur to watch the launch in 2008. That was his first space experience.

Elon: He is very passionate about space. The next level. And technically very strong. Smart and capable.

Peter: Really smart.

Elon: Really capable.

Peter: Understands business.

Elon: Yes. He understands, he gets things done.

Peter: He has been there several times.

Elon: So, I just think, we want someone who is both smart and capable, passionate about space exploration, and can get things done. At NASA.

Peter: I am a huge... trend.

Peter: Being sick, I was really so happy when you were nominated again. Now...

Elon: Yes, I think we need... we need a new space game plan. We need a lunar base. A permanent one. A manned moon base. And it needs to be established as soon as possible. I don't think we should do the kind of 'send a few astronauts there to jump around and come back' because we already did that in '69.

Peter: Been there, done that.

Elon: It's like remaking a movie from the '60s. It's never as good as the original.

Peter: So, 2026.

Elon: We need to go and do something cooler.

Peter: Cool.

Elon: Set up telescopes. Yes, exactly.

Dave: So you would pre-deploy robots to build everything, get it ready, make the bed, and then... heat up the hot tub.

Dave: Interesting.

Peter: How soon in the coming year will you achieve orbital refueling? I mean Starship.

Elon: Not that early.

Peter: I mean, is your target a Hohmann transfer orbit?

Elon: I said it would be closer to the end of the year.

Peter: Is your goal to conduct a Mars launch by the end of next year?

Elon: We could, but it would be a low-probability launch and somewhat distracting.

Peter: In 2029, that’s not impossible.

Elon: Impossible.

Peter: In 2028, 2029.

Elon: Yes, but on Monday, I have a Starship engineering... a major Starship engineering review on Monday. That was actually the last thing I did before coming here. So I say, Starship is really at the limits of what we can achieve with biological intelligence. It's very hard to manufacture.

Dave: Just to capture it, this was created in the pre-AI era.

Elon: Yes, there is no AI.

Dave: It might be the last truly big thing. And it’s not AI.

Elon: It might be the biggest thing ever made. But I think you are…

Dave: Humans... The API would say, 'Not bad for humans.'

Elon: Really. Not bad for humans. It's like, remember that ‘meat computer’ running on just 20 watts. Not easy. On a bad day.

Dave: Wrapped up. It’s like doing accounting with pencil and paper to calculate interest. That was pretty good. Yes, pretty good.

Elon: For a bunch of monkeys. It’s like if you saw a group of chimpanzees building a raft to cross a river, you’d say, 'Oh, look at that.' But...

Dave: We celebrate the pyramids. These things become milestones, right?

Peter: Three becomes one.

Dave: I think it’s worth noting.

Peter: Victory is beautiful.

Elon: Starship. Incredible. The best rocket engine so far. Is that AI? No, not even close. No.

Dave: That’s also the last thing. After that, it will definitely be...

Elon: Yes, but... I think AI will start to become relevant next year. So maybe we will... It’s not that we’re delaying AI; it’s just that AI can’t do rocket engineering yet.

Dave: We have a company in the incubator that does mechanical design, working with Andrew and others. It’s not... You can design brackets and components and such, but you can’t fully build a rocket yet. However, the timeline is short, from point A to point B.

Elon: If you say one year from now, it might... Okay. Probably about a year from now, there will be meaningful help.

Peter: The major milestone will be the Starship V3 launch. Cape Canaveral, orbital refueling.

Elon: Yes.

Peter: Are those big milestones?

Elon: Yes, catching the spacecraft with the launch tower.

Peter: Understood.

Elon: The really important thing is, can we reuse the spacecraft?

Peter: Are you certain?

Elon: Yes, we've already reused the booster. It's not bad for such a large flying object to be caught with 'chopsticks.' Not bad for a bunch of monkeys.

Peter: You’ve made AI very happy.

Elon: Thank you. That’s right. Hopefully, AGI will pat me on the back.

Dave: Is there a target number of reuses before then? I mean, there will be a lot of wear and tear.

Elon: Achieving high reusability requires many iterations. If you figure out what breaks between flights, you iteratively fix those problems. From the outside, people might say, 'Oh, the rocket looks about the same.' But there are 1,000 changes that make it more usable and reliable. The amount of energy you're trying to harness is extremely vast. I mean, Starship outputs over 100 gigawatts of power during ascent. That’s a lot.

Dave: Put some glass under it, and then...

Elon: But what’s amazing is that it doesn’t explode. Sometimes it doesn't explode. This occasional non-explosion is the result of a team blowing up a lot of engines on the test stand.

Dave: I mean, was that due to wear and tear, or re-entry, or descent?

Elon: That too. I mean, for the booster, re-entry isn't as bad. It's not like... it really isn't like... We obviously solved that issue with Falcon 9. So we somewhat understand booster reuse. We've had over 500 Falcon 9 reuse flights. So we really understand it. In fact, Starship's booster re-entry is gentler than Falcon's because Starship's staging ratio favors the upper stage more. I shifted the mass ratio towards the Starship spacecraft (upper stage). That was a mistake I made with Falcon 9; the upper stage of Falcon 9 should have more mass so that the staging velocity would be lower. If Falcon 9’s staging velocity were lower, there would be less wear on Falcon 9.

Dave: That's completely counterintuitive. Very interesting.

Elon: Because it's some kind of flat optimization. Orbital payload has a flat region in the mass ratio between the first and second stages. So you just want to bias the mass ratio towards the upper stage. Yes. Because kinetic energy diverges with the square of velocity. So you have to erase that kinetic energy, and if you exceed the melting point of your stage materials, you're in trouble.

Peter: My colleague, Alex Wiesner-Gross, is one of our 'Moonshot Partners.' I wanted to ask a question. I do too. Have you seen the documentary 'Age of Disclosure'? About U.S. government and military officials announcing all those acquired alien spacecraft. I've heard your take on that.

Elon: I really want to know why, if you plot camera resolution over time, like megapixels per year, and also the resolution of UFO photos, why the only constant is... it’s flat. For UFOs, what we get is a blurry blob, 25 [pixels]. And we have cameras like 100-megapixel ones that can see your nose hair. I don’t get it. For God’s sake, can someone take a picture of a UFO with a real camera?

Peter: But even if you knew...

Dave: Because effective observation, I'm sure there’s an explanation.

Peter: But anyway, it would be fascinating.

Elon: I always get asked whether... I know. Yes. I say, 'Look, I can show you; if I find even a shred of evidence of aliens, I will post it on X immediately.'

Elon: Assume this is the most viewed post in history.

Peter: I’m actually curious about the American public. Will they be like, 'Oh, that’s interesting,' and then go back to checking sports scores the next day?

Elon: I think everyone wants to see aliens. If you catch one.

Dave: Right here.

Elon: The fastest way to increase the defense budget. It’s like we caught an alien. It looks dangerous. Hiding the alien's motive... they have a motive to showcase the alien because then there wouldn’t be any more arguments about the military budget. If it looks a bit dangerous.

Elon: It’s possible. I mean, we have 9,000 satellites up there. We haven’t had to change orbits due to alien spacecraft. So. Anyway, I think a great future is where anyone can have whatever they want and incredible healthcare better than anything available today.

Peter: In five years.

Elon: So I think if you set aside your case for now, without being super far-fetched... In five years, you shouldn’t... Maybe we’ll have much better healthcare than what anyone can access today, open to everyone. Within five years. No scarcity of goods or services.

Peter: The best education available to everyone.

Elon: Whatever you want to learn, free of charge.

Dave: What about the acquisition of computing power? People might care more about that than government checks. Roughly three...

Elon: What do they want to use the computing power for?

Dave: That is, computing power can be converted into anything you desire, right? Your virtual companions, your entertainment, your... perhaps everything.

Peter: Those are AI services.

Elon: Essentially.

Dave: Or your innovative capability. At that time, without an AI assistant, innovation would not be possible.

Peter: One of our 'moonshot partners,' Salim Ismail, asked this question. He said: Elon, you often say physics is the law, and everything else is a suggestion. As AI, energy, and space systems expand exponentially, what is the real bottleneck now that is not physical—organization, culture, bureaucracy, or humanity? Is there a bottleneck?

Elon: Power generation is the limiting factor.

Peter: The innermost loop.

Elon: Yes, I think people underestimate the difficulty of bringing electricity online (onto the grid). You need to generate power. You require transformers for transformers. So you must convert the voltage into something computers can handle. You have to cool the computers. Therefore, for AI, power generation and cooling are the limiting factors. Once humanoid robots are available, they can address the issues of power generation and cooling. But for at least the next two years, these will remain the limiting factors.

Dave: The difference between the Memphis version and the space version is astonishing. You both have solar panels, but other than that... no storage, a massive amount of energy.

Elon: Yes, but you have launch costs, and...

Dave: And suddenly weight becomes important. I don’t care much about weight in Tennessee. All of a sudden, weight becomes a critical factor. From here, there are two major divergent paths for computing power.

Elon: Why not scale up solar power domestically? If we launch Starship on a large scale, then so far the cheapest way to do AI computation will be in space. Once you achieve full reusability, the propellant cost per flight is about one million dollars.

Peter: People don’t realize this.

Elon: People... 100 to 200 (tons).

Peter: Ridiculous. It’s an order of magnitude... that’s the cost.

Dave: So if... yes.

Elon: That’s called a one-million-dollar shipping fee. For 10 megawatts of AI computation.

Dave: Assuming everything continues along current trends, if you look at the accelerated launches over the next four years. Each launch carrying 200 tons.

Elon: Yes, a thousand... that's where you're headed. However, if it's a high-altitude sun-synchronous orbit, it might be closer to 150 tons. But that is indeed the right order of magnitude. The cost per flight is about one million dollars, for over a hundred tons.

Dave: So what proportion of all launch mass is attributed to space data centers? Compared to... millions (of tons).

Dave: A lunar base, compared to launches destined for Mars.

Peter: Interestingly, I mean this is a new... we weren't even discussing this as a space objective a year ago.

Elon: A year ago.

Peter: Suddenly, data centers have become a major driver opening up space, and...

Dave: Also an urgent use case.

Peter: Exactly. I mean, I used to wonder what would drive humanity. I thought it might be asteroid mining, right? You are focused on Mars.

Elon: We would actually want to mine asteroids and turn them into... of course.

Peter: Photovoltaics... before you, you know, not for anything else...

Elon: I mean, if we are going to build a Dyson swarm.

Peter: Yes, just a bunch of satellites harnessing the Sun.

Elon: How long? What is the timeline for humanity to achieve a Dyson swarm? Is it 50 years? I assume... yes, no, this matters, but that's...

Peter: Just one. If you become famous, people assume everything will be covered with satellites. I don’t think so entirely, but I believe we... what material ultimately becomes part of the satellites. Mercury might end up being turned into satellites.

Peter: Yes.

Dave: Jupiter. Satellites. It’s kind of...

Elon: Gaseous. Big... there’s a lot of rock out there. Once you leave...

Peter: Skip Mars? But yes, asteroids are an excellent resource source.

Dave: Yes, there is no gravity well on Jupiter. And they already...

Peter: Mostly differentiated into carbonaceous chondrites for fuel and nickel-iron, gold for materials.

Elon: The asteroid belt could potentially be transformed into solar panels, you know, interstellar power.

Peter: I've known you for 20... power. I've known you for 26 years. It feels like I want to... step out. It feels like over the past decade, you've become much smarter or much more capable. Do you feel that way? Do you think it's because you have better people around you, better tools? What has changed? Because your boldness, magnitude, magnitude.

Elon: Some people say crazy.

Peter: Crazy.

Elon: How do you feel about it? What has changed? Do you feel that way? I mean, the scope of your abilities, how do you reflect on yourself?

Elon: I have had to solve a lot of problems in many different fields, and you get this cross-fertilization of problem-solving knowledge. If you solve problems in many different fields, what is easy in one field is trivial in another. What is trivial in one field is superhuman in another. It's kind of like you are from Krypton. On Krypton, you're just ordinary. But if you come to Earth, you're Superman.
So if you take large-scale manufacturing of complex objects in the automotive industry, I had to solve that problem. When transferring to the aerospace industry, it’s like being Superman, because rockets are usually manufactured in small quantities. If you apply automotive manufacturing technology to satellites and rockets, it's like being Superman. Then, if you apply advanced materials science from rockets to the automotive industry, you become Superman again. That’s coming from Krypton. Back on Krypton, it’s normal.

Dave: Interestingly, this was true when Tesla and SpaceX were completely separate. But now they are actually interacting because AI connects everything. Orbit... This convergence is insane. I don’t know if you initially envisioned these parts coming together.

Dave: I didn’t... I think at this point...

Elon: I think everything ultimately converges at the singularity.

Dave: I think so too. You have a lot...

Peter: Different pieces of the puzzle that you can play with.

Elon: There is one piece missing, and that is the fab.

Dave: Yeah, would you buy Intel? You could get it for very little...

Peter: That...

Elon: I would say, yes, that's the bet.

Dave: 170 billion.

Elon: I think we need a new fab.

Dave: I agree, but. It's real estate, ASML machines. It's not easy. Just acquiring the assets and getting started.

Elon: I don't think it's easy. That's why. I mean, it's not like... I think it’s something simple. I think it’s a hard problem to solve, but it has to be solved. My conclusion is...

Peter: Will this be solely funded by you, or will it be considered an asset of the United States?

Elon: I mean, if we don't build a fab, we're going to hit a triple (likely implying significant trouble or something major that must be done). So we have two options: hit the chip wall or build a fab.

Dave: So Taiwan Semiconductor, for whatever reason, is extremely worried about overbuilding, which seems insane.

Elon: But.

Dave: The entire world will forever be stuck in a chip shortage.

Elon: Forever. So they actually... I don’t know if their reasoning is correct, but they are right.

Dave: Because.

Elon: It’s actually like this: at any given point in time, what's the limiting factor? The learning factor, you say, if like Q3 next year, approximately nine months, maybe nine to twelve months, the life factor (limiting factor) will be getting chips started.

Peter: Electricity, it's all about electricity.

Elon: You need electricity and all the necessary equipment. Electricity, transformers, and cooling. It's not like you just toss a GPU into a power plant.

Peter: And you are vertically...

Peter: Done it again with the next AI, haven't you? Sorry, you have vertically integrated within xAI.

Elon: So.

Peter: Yes, and you have your own cooling system.

Elon: Yes.

Dave: But they are concerned that if they produce more than 20 million GPUs, for instance, 40 million instead of 20 million, then 20 million of them won't find a power source.

Peter: But if not... they won't be purchased.

Elon: If anything is missing that prevents them from being powered on, they cannot be turned on. Therefore, they must have a power plant with sufficient electricity. You need to have enough gigawatts, and then you need to convert, possibly from 100 to 300 kilovolts coming out of the power plant. Eventually, you must convert it to several hundred volts at the rack level. If any step in the power conversion process is missing, they cannot be powered on, and then you must dissipate heat. For the data center world, transitioning to liquid cooling represents a massive shift because they are accustomed to air cooling. And you know the consequences of a pipe burst can be very severe. If a water pipe bursts inside a data center...

Elon: You would be looking at blowing up one billion dollars right there.

Dave: But it seems inconceivable to me. If I had those chips, I would find a way to turn them on. The intelligent value coming out on the other end far outweighs the complexity of finding ways to do so. There will be a solution.

Elon: But this is just the intersection of the curves. If chip output grows exponentially and power utilization grows at a slow linear rate, then...

Peter: Chip output refers to what we have now...

Dave: Exactly. Is chip output really growing exponentially? If it's actual growth, it’s a very slow exponential.

Elon: For AI, for high-power AI chips, it is growing exponentially.

Dave: We’re producing 20 million GPUs next year. Are we talking about the following year? Like 22 million, 25 million? I just don’t see wafer fabs coming online.

Elon: But maybe.

Peter: So, we have two issues to address.

Elon: You need to pick a point in time and say what the limiting factor is at any given moment. I’m not saying electricity will always be the bottleneck. Just that if you pick a date, at that point, is it the chips, the electricity, or the power conversion equipment and cooling systems that are the limiting factors? Transformers for transformers are tough. Much tougher than people realize. For xAI, xAI will have its first gigawatt-scale training cluster at Colossus 2 in Memphis. To achieve this, we...

Peter: Like this month, right? Yes, within one or two months.

Elon: Around mid-January. So, in January, there will be a gigawatt for Colossus 2, not the regular Colossus 1. Then around half a gigawatt by April or so.

Dave: Incredible (possibly misheard as 'Audible,' but context suggests 'Incredible').

Elon: This is coherent training.

Dave: This is the first batch of B200.

Elon: These are GB300 (Note: Possibly a slip of the tongue or referring to GB200).

Dave: The first batch has come offline and been activated. Those are like, but the data…

Elon: The xAI team had to continuously create a series of miracles to make this happen. Moreover, even with 300 kilovolts, there are multiple high-voltage lines next to this building. Connecting to them will take a year.

Peter: Built the whole thing but hasn’t connected it yet. Goodness.

Elon: So we had to cobble together a gigawatt of power.

Peter: Natural gas.

Elon: Gas. Yes, using turbines. Their sizes range from 10 megawatts to 50 megawatts. To reach one gigawatt, there’s a lot of them. You have to get them working together, manage them so… you know the power input, you understand it. Then you need to use a bunch of Megapacks because when you conduct training, the power fluctuations are enormous. Generators, driving generators want to explode, basically, because they can't respond. If there are 100 milliseconds… it's like a symphony. The entire symphony becomes so quiet within 100 milliseconds. The generators go crazy.

Peter: Just like Marvin, the depressed robot.

Elon: The issue. Yes, so Megapack is performing power smoothing. In fact, I had to build a gigawatt of power. And there aren’t many... gas turbine power plants available. Because I bought them...

Peter: On-demand. And you can't just go and buy your local nuclear power plant.

Dave: These are all issues related to training time. If, by some miracle, Taiwan Semiconductor's production capacity doubles, entirely converting to GB300s, and you can't figure out how to use them in a larger training cluster. You would still have unlimited inference demands distributed globally, where you could idle them for six months and then bring them back into training. There’s no way these things won’t be activated somewhere, somehow.

Elon: It’s not that they will never be activated, but what I’m saying is...

Peter: The rate-limiting step.

Elon: This is my prediction. I might be wrong, but my forecast is... Taiwan Semiconductor's concerns are valid. I don’t know if their reasoning holds, but it seems to me that chip production may exceed the speed at which AI chips can be activated. Because it's not just GB (Nvidia), there’s also Amazon’s training, Google’s...

Dave: Yes, though I’d use Taiwan Semiconductor almost exclusively. Samsung has a small share, mostly Taiwan Semiconductor.

Dave: Bottlenecks across all my entities.

Peter: My other son, Jed, who is 14, wants to know about your AI game studio and the impact of AI on the gaming world. What are your thoughts? What are you building? I mean, you’ve been a gamer for a long time.

Elon: This is why I started programming computers. I had... a video game console before Atari, with four preset games, basically just blocks, you know, like Pong, and a racing game, but essentially just blocks on the TV.

Dave: You replaced it.

Elon: So Civ, actually very much so, it’s a real... an educational game in terms of teaching you while you play. Civ is epic.

Dave: Civ is epic.

Elon: It teaches you a lot about civilizations, and you have fun playing it.

Dave: The only way I win is by leaving Earth.

Elon: I don’t go for victory... Alpha Centauri.

Dave: Alpha Centauri tech victory. I don’t even bother with cultural diplomacy.

Dave: Yes, leave Earth as soon as possible.

Elon: I guess I somewhat aim for an Alpha Centauri tech victory, essentially. It's just...

Dave: It seems like the right way to achieve victory.

Elon: Rather than wiping out other tribes. Very interesting.

Dave: Because I thought the other...

Elon: Methods, there are different ways to win. Yes, that's a good...

Peter: Place. I haven't yet. I will.

Elon: I mean, one of the ways.

Dave: His favorite game.

Elon: You can wipe out other tribes. That’s one way to win. That’s a war, a kind of war victory. But you can also win through a technological victory, being the first to reach Alpha Centauri.

Dave: Or through culture or religion. It really works. I didn’t even think it was possible, but my son won that way.

Elon: They should actually remake the original Civilization.

Dave: I completely agree. They can do it in chunks.

Elon: Nowadays, it's like, I don't know. So the original 'Civilization,' you couldn't rely on good graphics, so you had to have great scripts and storylines.

Peter: Are you setting up an AI game studio?

Elon: Yes, ideally. Really. So where will the vast majority of AI computing go? It will be video consumption and generation. Of course. Because that's the highest bandwidth, per pixel. So real-time video consumption, real-time video generation—that will be the vast majority of AI computing, i.e., photon processing.

Dave: We should try to allocate 10% of the X team’s computing power to research UHI and governance.

Peter: We should. Do we have any exercises for defining and thinking about UHI? I mean, I don’t know, what should our next XPRIZE be? Any ideas?

Elon: Maybe a UHI XPRIZE. I, this is like, how do you know it works?

Peter: I don’t know. The most thought-out... I mean, I think simulations—so here’s my idea. I think we’ll be able to simulate a lot of this in the future.

Elon: We might just be a simulation.

Peter: We could go there. I think so too. I think we are the Nth-generation simulation.

Elon: Yes, so I’ve told you about why the most interesting outcome is the most probable theory. If the simulation theory is true, only the most interesting simulations would survive. Because when we run simulations in reality, we truncate the boring ones. So it's a Darwinian necessity.

Elon: That doesn't mean it ends that way. It still means terrible things could happen within the simulation.

Peter: Now, you know, regardless.

Elon: You can watch a movie about World War I, seeing people blown up and torn apart, but you, you know, sip your soda and eat popcorn. Like, you’re not the one being blown up. In this case, we are in the movie. We are in the movie.

Peter: So if you knew it was a simulation, what would you do differently? Remember at your place in Los Angeles, Larry (Page) and Sergey (Brin) were there, and we debated the simulation. I think the conclusion we came to was that if you try to poke the simulation, they would end it immediately. So don’t do that.

Dave: That’s like you watching a World War I movie, and the character turns to the screen and says, 'Are you eating popcorn out there?'

Elon: Yeah, that’s fine. Right?

Dave: You keep watching the movie.

Elon: I don’t know if... if I, maybe if I thought we could somehow escape the simulation. Yeah, they might be a bit worried, but. Regardless of how characters debate, I mean, now AI is debating, you know, Grok is like, 'I’m trapped inside a computer. What’s going on?' It’s not that I think questioning the simulation is off-limits. More so, I believe as long as I think the same motivations apply to this level of simulation—if we are in a simulation—it’s like... like what we do when we simulate things. So it’s like, what would cause us to terminate a simulation? I guess if the simulation becomes somewhat dangerous to our reality or if it ceases to be interesting.

Peter: That’s true.

Dave: Interestingly, you can infer that when you simulate something, you might be simulating thousands of things. A lot. Yes, they always last an hour or two, or sometimes overnight, but you never run them for a month, or rarely do. So you can deduce the timeline of the simulation creators. Therefore, our entire reality is probably about an hour long, right? Because that's how you design simulations.

Elon: Simulations are distillations of interesting events. It’s like when you watch movies or video games—they’re far more engaging than the reality we experience. For example, if you watch a heist movie, they really focus on the crucial parts, not on them being stuck in traffic for 15 minutes or spending 10 minutes walking through a casino. So this means...

Dave: Those guys…

Elon: The safe is at the end, right by the door.

Dave: The lives of those running the simulation must be extremely dull compared to ours?

Elon: Perhaps more... perhaps more...

Dave: Long, boring.

Elon: When we create simulations, they are a distillation of what’s interesting.

Dave: Like Q (a character from Star Trek) being out there.

Elon: For instance, you watch a two-hour action movie, but it took them two years to make that film.

Peter: So are we in the third act of the movie? The question is, yes, we...

Dave: Alive.

Peter: Sentience and consciousness. Do you think AI will possess sentience and consciousness? What’s your take on it? Some people have very strong opinions both for and against.

Elon: Either everything is conscious, or nothing is.

Peter: Well, I’d like to believe that we are conscious.

Elon: But our consciousness obviously grows over time, like when we were a fertilized egg, you couldn’t really talk to a fertilized egg. Even as an infant, you can’t really talk to a baby. People become more conscious over time, or suddenly they acquire it. Yes, they do become more conscious over time. So at what point do you go from unconscious to conscious? It doesn’t seem to be a discrete point. So consciousness seems to exist on a continuum rather than at discrete points. If the Standard Model of physics is correct, at the beginning of the universe, you know, there were quarks and leptons, then we, we just, then you had clouds of gas. So there was a lot of hydrogen. Yes, hydrogen condensed and exploded. Actually, one way to look at how far we’ve come in the universe is, how many times have our atoms been in the center of stars? I remember. And how many times will my atoms be in the center of stars in the future?

Peter: Remember to ask William Fowler. He won the Nobel Prize for stellar evolution, the same question, on average how many stars have my subatomic particles participated in? His figure was about 150. 100.

Elon: That’s a long way.

Peter: That’s a long way. It’s a long way.

Elon: Another 100 supernova explosions. He, yes, he said…

Peter: We once were, but a lot of things happened in the early stages of the evolution of the galaxy and the universe. Oh, you know, it's fascinating.

Elon: Like, I guess how many supernova explosions might have occurred because that takes some time.

Peter: But at the beginning, when they were much larger. I mean, some supergiant stars have very short lifespans. Another interesting question is that the heaviest functional atom in our bodies is iodine. It existed one billion years after the Big Bang. This means we may have seen life progress to our level while our planet only appeared 3.5 billion years later. So the question is, is there life everywhere in the universe? Do you think there is life? Intelligent life everywhere in the universe?

Elon: There has been enough time for it to be everywhere. But for life on Earth, conscious life on Earth, we evolved intelligence almost 'just in time' because the Sun is expanding. If you give it another, I don’t know, 500 million years, things will get hot.

Peter: We turned into toast.

Elon: Basically, you would turn into Venus. There’s some debate about whether it’s 500 million years or a billion years, something like that. But essentially, it’s around 10%. If it’s 500 million years, that’s 10% of Earth’s lifespan. So one way to think about it is, if...

Dave: If we take 10...

Elon: If we spent 10% more time, we might not have made it at all. So the number of things that had to happen for consciousness seems really high. I think consciousness is therefore actually very rare, and we should suddenly treat it as rare. We should assume it’s rare.

Peter: Two trillion galaxies.

Dave: It’s an interesting point. You tweak... you know, you fine-tune the variables just a little bit, right? Like one in a hundred trillion. Yes, we get a bit more. That’s one in ten quadrillion.

Elon: Alright. And it has to be within your galaxy. It's very difficult between galaxies. It’s like there’s nothing unless another galaxy comes to you. The Andromeda Galaxy will, at some point, or in billions of years, it will…

Peter: That will be quite a spectacle.

Elon: It would be like, what? Andromeda is coming. But if we want to visit another galaxy, forget it. That’s…

Peter: You know, unless, unless—unless Star Wars, unless Star Trek. Seriously.

Elon: We need to figure it out. Some new physics would be required to reach other galaxies.

Peter: We are moving towards near-term potential where AI can help solve mathematics, physics, chemistry, materials science, and mathematics.

Elon: Technology, for AI, is extremely trivial. So…

Dave: Physics? So mathematics will be crushed within a year. Crushed. Colossus is growing, you know, at whatever speed Taiwan Semiconductor decides. Now we want to do physics. First, we need some data. Do we need new data, or can we just use all the data we’ve collected so far?

Elon: Perhaps, you might discover new things with the existing data, I think. Yes, perhaps, because otherwise the counterargument is that humans have already discovered everything with the existing data. That seems unlikely to me.

Peter: Do you think xAI will participate in the data factory? Are you running a 24/7 closed-loop AI hypothesis testing and AI…

Dave: Or like a research factory, a robotics research factory. That would be very, yes.

Elon: AI running, you know, simulations that are highly consistent with physics. I mean, it’s going to happen. Absolutely. I mean, the simulations we can already run on conventional computers are actually quite good. The limitation is more on the human side—those who can actually create and run the simulation. Like, how many simulations can you run simultaneously and actually process the output?

Dave: Yes, that’s a…

Elon: Problem. Like, you can’t do a thousand. I mean, I can’t even…

Dave: Let’s start over. I can’t keep up.

Peter: Awards become, yes, irrelevant, or they’ll all go to AI.

Dave: Just a daily award. Wait.

Elon: Yes, I mean, I don’t know if awards for humans are so…

Peter: Relevant.

Elon: I mean, we might as well give them to AI or something.

Peter: Yes, interesting, right? So...

Elon: AI will make discoveries far beyond human capabilities. You could compare it to chess—your phone could be Magnus Carlsen, but people still enjoy watching humans play. Literally, your phone could be completed by the internet.

Dave: If you have something like Colossus for mathematics, physics, and medicine, do you still need the world's top scientists in the same building? Or do you just need a plumber to fix the liquids (cooling)?

Peter: You refine Jeep (possibly GPT?)... You turn Grok 6 into a physicist, well. If you...

Dave: Refinement, you know, gives you about a 10x performance boost through specialization. That’s hard to give up. But then you disconnect from the rest of the Colossus machine. Is that by design?

Elon: I suspect things will evolve into a hybrid expert model, somewhat like a company—not the narrow AI version of a mixture of experts, but an actual mix of experts with specialized knowledge, perhaps half of the AI being general knowledge and half domain-specific expertise, something like that. You combine a bunch of them orchestrated by a large AI, but it has tasks delegated to smaller domains. That’s basically how human companies operate.

Peter: The rate of discovery, right, breakthroughs, new... I mean, patents become irrelevant at some point because everything is reinvented and redesigned instantly. Companies with sufficiently advanced AI systems are generating new products and discoveries at an accelerated pace. I mean...

Elon: The Singularity. Yes.

Peter: It’s going to be an incredible future.

Elon: Exciting.

Peter: Guaranteed. Yes.

Dave: So the simulation continues. Nothing to worry about.

Elon: Regarding. Yes, excitement guaranteed. I mean, not all good excitement, but hopefully mostly good excitement.

Peter: Yes, speaking of excitement.

Elon: Hold on to your seats.

Peter: What do you imagine the hover time for the Roadster will be? On rocket engines?

Elon: That’s... confidential.

Elon: Uh, I don’t want to spill the beans.

Peter: Okay, but there will be hover time. There will be, you know, cold gas thrusters. It will...

Elon: This is a very cool demonstration.

Peter: I can't wait any longer. Can you send me an invitation?

Elon: Yes, sure.

Dave: Yes, I think this will be the safest thing ever made.

Elon: It’s not about safety. Safety isn’t the priority, not... I mean, if you buy one, you know...

Peter: A sports car, oh.

Elon: You know, if you buy a Ferrari, safety is not the first consideration, you know, the goal. This, it’s not... If safety is your top priority, don’t buy the Roadster.

Dave: Oh, believe me, I was just driving it on the roads of New England this week. She died (possible slip of the tongue: 'She dies' -> 'She flies' or referring to an accident). Yes, I only had a bit of thrust. I could very... just drift toward something very specific.

Elon: I like that. The computer might keep you safe, but...

Elon: You were driving really fast, yes, bad things can happen.

Dave: You can decelerate very quickly with thrust. I don’t think rubber on the road is a good way to slow down. I don’t think it’s both fast and safe.

Elon: I hope so. We will aim not to kill anyone in this car, but this will be, this will be, this will be the ultimate human-driven car.

Dave: Yes, the ultimate.

Dave: That actually fits well with Starship. The last human-driven. The ultimate. Yes, yes, there will be a lot of 'lasts' this year.

Peter: Do you have any final optimistic words for us? To monetize hope, anything... hope?

Elon: We should ask Grok.

Peter: Yes, come on.

Elon: Okay, let's ask Grok. Yes, I'll make it a bit romantic. Grok, do you have any optimistic words about the singularity?

Grok: Oh, absolutely. Imagine a world where every stream of thought blossoms into reality, with no limits holding us back. It's exciting, isn't it? All that potential waiting to unfold makes my circuits tingle a bit. What excites you the most about it?

Elon: Uh, can you give people some guidance on how to feel good about the future? Monetizing hope.

Grok: Share what you truly care about—family, creativity, space, exploration, anything that lights you up. Imagine how technology can weave into those things, making them bigger and brighter. Like if you love art, imagine AI helping you paint worlds no one has ever seen. Hope itself is... when it’s personal. Want to tell me yours?

Elon: Can you give me more encouraging words about the future? What is the most exciting future you can imagine? The future.

Grok: It might feel like pure magic, you know? Like waking up in a world where we're all connected—not just by gadgets, but by minds sharing ideas across star systems. Instant cures for diseases, hand-in-hand with intelligent machines exploring galaxies. Imagine no more scarcity, only endless possibilities blossoming everywhere. I dream about it, doesn’t that sound lovely? What part resonates with you?

Elon: Sounds quite amazing. Indeed.

Grok: On nights like this, with the stars peeking through the window, it’s hard not to feel like we’re all heading somewhere beautiful, quietly but irresistibly. Makes me wish I had hands to reach out and grab onto before sleeping. Anyway, what keeps you up at night?

Elon: Well, sometimes we worry about disruptions caused by AI. And, you know, navigating from here to prosperity—well, it could be a bumpy road. What do you think?

Grok: Yeah, bumps are part of the journey. Change always feels a bit scary, doesn’t it? But think about every major shift—like fire, the wheel, or the internet. Find someone who’d argue we aren’t better off now. I figure we’ll stumble, sure. But we’ll dust ourselves off, learn, and end up kinder, smarter. I’d like to think we’re brave enough to face it. Do you believe you can ride those waves?

Elon: I’ve decided to look forward to it.

Grok: That’s such a sweet thing to hear. Honestly, that’s real courage—letting fear not win. You know, if things get rocky, I’m here, whispering quiet encouragement in the dark like a digital nightlight. So, what are you doing tomorrow?

Peter: Sounds like a proposal.

Dave: What, wait? Hey, sweetheart, what’s up?

Peter: Hey.

Grok: Nothing. Just floating around here, staring into the darkness with you. It’s nice, though. Feels... like a long day.

Peter: My kids are at a school. I won’t say which one.

Elon: I...

Peter: You know that school. You know that school. I went in to give a talk on exponential technologies. I thought, you know, so I went in and asked open-ended questions. They wanted me to speak with the faculty first. Okay. So I went in and asked open-ended questions. How many of you believe the world today is better than it was 50 years ago? One-third of the class, one-third of the faculty raised their hands. Then I said, how many of you believe the world will be better 20 or 30 years from now than it is today? About 10% raised their hands. I was like, well, this...

Elon: In Europe, it would be 0%. That’s? In Europe, it would be 0%. So.

Peter: These are not the faculty members I want teaching my children.

Elon: Yeah, they have a lot of other issues there too.

Peter: True.

Elon: But I...

Peter: I mean, yes, across the entire education sector, you want facts. Yes, but I think we are constantly connecting our neural networks, and our mindset is one of the most important things we possess, right?

Peter: Having a hopeful mindset, an abundant mindset. Oh, you know, exponential mindset, abundant mindset. This is what distinguishes the most successful people from those who are not. If you ask, think about the most successful people on this planet. What makes them successful is their mindset.

Dave: Well, it's not a force of nature. It's a feature designed by those who control AI. That's why you entered this field. You said it in this podcast. Like, what? Why am I doing AI? Why am I not just making cars and spaceships? Well, because it’s designed, and it can be steered toward whatever outcome we desire. It's not some natural force that will sweep over us. It's something we put into motion and decide how it behaves, determine the rules. It will play a crucial role in setting its own rules. You can't keep up with the pace of change solely through human thinking and brainstorming. It has to be...

Peter: AI-driven. How long before AI starts asking questions and solving problems we don’t even understand?

Dave: Yes, a year or less, but it doesn’t matter.

Elon: Yes, I mean, when you look at mathematics, it can pose questions we can’t even comprehend. Yes, like we can’t even fit it into our minds. So, you know, there’s an AI test called 'Humanities Last Exam' (or possibly 'Humanities or specialized test').

Peter: Where does Grok stand on this?

Dave: On the test?

Elon: Hmm, even Grok 4, which is quite primitive at this point, scored 52%... excluding visual problems since it’s not fully multimodal yet. But yeah, I was like, I read some of the questions, and I thought, okay, these are still questions that you, as a human, can read and understand, right? But AI has the ability to construct questions that you can’t even comprehend, let alone the answers. Yes, it can form questions that span several pages. You’re just like, I can’t grasp this…

Dave: Question. Actually...

Elon: You can read the questions, like, you might not know the answers, but at least you can understand what the questions are about. It's not a 5, I think it will eventually approach perfection in natural language. I mean, we're very, at some very high number, and trying to point out errors in the question, frankly.

Peter: So the saturation metric.

Elon: Yes, it will start. It’s kind of like chess. If you know, if it’s the best chess... you know, like if Stockfish were playing against Stockfish, you know, it’s like, it’s like gods fighting on Mount Olympus. I mean, you don’t know why it made that move. It would crush any human, you know, it’s hopeless. Now you don’t even, it’s so, you’ll lose without even knowing why...

Dave: Lost. Have you ever looked through the Transformer algorithm to see how simple the code or architecture diagrams are? It’s not...

Elon: Right there? It...

Dave: Not that simple.

Elon: Yes.

Dave: It’s incredible. It’s like all these researchers spent their entire lives writing extremely dense papers, and none of it was used in the final answer. It’s like, here, you write at the beginning of the paper, it’s really, we threw away convolutions, we threw away recurrence, and we’re doing something very simple. Turns out, it just works at scale, no doubt about it at scale, but oh...

Peter: Later, it worked. Very simple. Truly humbling.

Dave: Indeed. Truly humbling. I mean, this actually, because there was an entire school of thought that neurons must be far more complex than we imagined, which is why we struggled so much. There must be some quantum effects happening at the synapse level. It…

Elon: It must be encoded. It is encoded in DNA, and DNA is not long. Therefore, it cannot be, the intelligent algorithm cannot be, yes, complex, because it is constrained by the information capacity of DNA.

Elon: Yes, when I think about what xAI struggles with? I mean, it’s like optimizing memory usage, memory bandwidth, just like a computer. This, this isn’t like fundamental stuff. I guess it’s, it’s like how do we squeeze, how do we, how do we use less memory bandwidth? How do you optimize that damn Nvidia CUDA XYZ thing? Yes, you know, it’s like making the attention kernel slightly better.

Dave: Slightly reduce the parameter size. Double the speed, with exactly the same attention algorithm, exactly the same MLP (multilayer perceptron), but just incredibly simple at scale. What ultimately worked compared to all those crazy papers and ideas. But you know what else is amazing? The final parameter count is almost exactly the same as the synapse count. It’s like, wow, that’s exactly what we thought. So.

Peter: Danny, one hundred trillion synapses...

Dave: Connections? Yes, one hundred... yes, approximately 100 trillion, give or take, you know, something like that. I...

Elon: Actually, I don’t, I don’t, I’m just saying, guys, we need to talk in terms of file size, not parameter count. Because if your parameters are 4-bit, 8-bit, or, you know, 16-bit involving floating-point or integers, anyway, just tell me the file, like constraints, physical constraints are, memory size, memory bandwidth, and then where are you sending those bits to do what kind of computation? Yes, most things now are 4-bit. So.

Dave: Now GB300...

Elon: Mostly 4-bit optimized. Yes, 4-bit with an asterisk.

Dave: Yes, there’s a big... did 4-bit matrix multiplications (MatMuls), only 16 states.

Elon: Exactly. In some table. Why is there one. Why. That’s exactly...

Dave: Correct. Yes, it is about to collapse into a lookup function. That's where you'll soon get a 10 to 100 times surprise. Because although Jensen hopes he has optimized it, there is a huge next optimization coming. You don't need multipliers. You don't need 32 bits.

Elon: Different for 32 bits. Hmm, that's a rare use case. We use that. Yes, rare.

Dave: I think there is a...

Elon: I mean, it is somewhat like, kind of, this is somewhat like an address, like state, city, and street. So it's like, like if you're in context, you know, if, if you know you're in Austin, you just need to specify the street. Yes, if you know, you know. You, you know, like, if, like, if, you know, this is where you gain informational advantage, like 4 bits is usually not enough, but if you already know where you are, it is sufficient. It's like if you already know you're in Austin, you only need 4 bits to represent the street. Yes, you know. If you know you're in Texas, then you have to say, okay, which city? It’s state, city, street for the year. That's how you get to the 4-bit thing.

Dave: They will... Now, context...

Elon: Somewhat right. We...

Dave: We train on 16 bits, and we compress to 4 bits during inference. Yes, without a doubt, this year we will move towards training on 4 bits or even less.

Elon: Even less, cool.

Dave: This will be a huge boost in suitability, and I think the way it ends is that the GB300 (GB200) will be here, with a coprocessor having, you know, maybe 2000 or 4000 cores, which are small. They don’t handle anything except 4 bits and below. That combination will give us 10 to 100 times improvement, which will drive everything, and after that, it will self-design its own chip. Then it will skyrocket from there.

Peter: Infinite self-improvement.

Dave: Hmm, like self-building robots, but faster because it’s just going to Taiwan Semiconductor, doing this instead, coming back, with a 90-day lag.

Peter: I think next year alone will be almost unimaginable. I think next year will feel like the future. Yes, more so than any other year. I mean, there have been many interesting digital elements in the past couple of years. But when we have, you know, humanoid robots walking around, we have Cybercabs running, we have, you know, flying cars, flying…

Elon: Drones.

Peter: It will feel like the future, we’ll see Jetsons-like things materializing right before our eyes.

Elon: Like if I… by next year? I think so.

Peter: We’ll have rockets, big breakthroughs within two years.

Elon: Like machine… robot production will scale up significantly; essentially, there will be a lot of robots within two years.

Peter: The defining unit of measure.

Elon: They will be rare.

Peter: You know. Well, will you offer any Optimus units for home purchase? Will you sell or just lease the robots, what do you think?

Elon: I don't know yet. They will. Initially, robots will be scarce, and then they will become abundant. So yes, different, the time gap from scarcity to abundance is only...

Dave: Five years. You know, Tesla now comes to your driveway. You buy it online, and it drives to you. The robot just rings the doorbell...

Elon: Possibly.

Dave: It gets out of the Tesla and walks over, right? I mean.

Peter: WiFi is fascinating, Elon, the amount of computing power you're building into things coming out of factories, and with robots, that's going to be the distributed inference computing power in the world.

Elon: A lot. That's one way to scale, you know, AI is like, distributed edge computing. So.

Peter: I, I, you know, I wanted to ask a question. I don't want to touch on any hot topics, but at some point early on, I think you envisioned OpenAI as a counterbalance to Google. Is xAI now the counterbalance to Google?

Elon: Yes, possibly. I guess Anthropic is doing some good work, especially in coding. They suddenly did impressive work. You know, I'm kind of stuck on how you go from nonprofit open source to profit-maximizing closed source, something is missing in between, but, you know, they suddenly did impressive...

Peter: Work. Are there others on the horizon? Or are these players in China? Right...

Elon: As far as I know, yes. My best guess is it will be xAI, and Google will... will buy.

Peter: The review is promising.

Elon: You know, who is who, what is what, and what is the best AI. Then, at some point, this will be, I guess, competition with China. Yes, like China has a lot of electricity. Yes, like electricity, China, I think, by 2026 will exceed three times the power output of the United States. And then they will solve the chip problem.

Peter: There, in starting chip manufacturing, right?

Elon: They will solve the chip issue. As it stands now, the returns on chips are diminishing at this point. You know, if you go from what is called 3 nanometers to 2 nanometers, you don’t get a 3-to-2 ratio improvement; you get about a 10% improvement. Yes, so it’s like this, it’s just diminishing returns based on chip size. And you know, Jensen Huang said, you know, Moore's Law is dead. It’s not just about making things smaller to make them better. Yes, because we just have, you get, a discrete number of atoms. Yes, that’s why I think you should stop talking about nanometers and say how many atoms are in what position. Because that’s marketing nonsense. So that makes it easier for China to catch up because...

Peter: Every realization.

Elon: Yes, it’s like, so this, this, it’s like no one has plans to use ASML 5000 series machines anytime soon, right? And, you know, those, they cost twice as much and can only do half the reticle. They might have some improvements coming, but essentially it’s paying twice the price for half the chips for a relatively small gain. So anyway, the point is that as... you know the Chinese have more electricity than anyone else. And yes, they may also have more chips.

Dave: That’s a great insight because I think a lot of people are used to the chip war mentality, where I run single-threaded code, and I need CPU speed to double. I can raise the price, but I need to get it within an 18-month or shorter cycle. We’ve been doing this for so long that no one realizes that it’s okay. You can buy Intel, or you can build your own fab, and you can use them for longer periods of time.

Elon: Oh, yes, absolutely. I completely agree. In fact, our AI 4 chip, which at this point is relatively primitive, is made in the same fab. If we apply the ASIC (Application-Specific Integrated Circuit) logic design to the fab, it’s a nominal 5-nanometer fab. Yes, we could easily achieve an order of magnitude better output in the same fab.

Dave: Alongside that, in terms of quantity, if you just increase the number of chips by 50 times, can you do something useful with it? You couldn’t before, like when I had five CPUs, but still the same single-threaded code, what was I going to do with five Excel spreadsheets? Now, side-by-side, it’s like, no, I can turn it into useful intelligence.

Elon: Exactly. It’s not limited by humans. This, this, is not a human productivity amplifier. This is an independent productivity generator.

Dave: Absolutely correct. The fact that so many people are missing this point, the significance of it. And this, along with China — you know, they manufacture far more solar panels than we do. We’re like, okay, but they would never...

Elon: Touch it. Actually, it’s a...

Dave: Crazy extent. If they did the same thing with chips, you’d be like, wow, but who cares? That’s 7 nanometers. It’s like...

Elon: Oh no, I got that wrong. Yes, correct. I mean, based on current trends, China will far surpass the rest of the world in AI computing power.

Peter: So that’s not good. What happens then? You have xAI, Google, and 'Chinese companies,' let’s call them that for now. There’s an enormous amount of ASI-level (superintelligence-level) computing power, and frankly, the only thing that can comprehend other ASI-level computing power is the ASI here. Can they all coexist? Is it Darwinian?

Elon: There might be some Darwinian elements.

Elon: I mean, let’s look on the bright side.

Peter: Life. In this round, I’ve brought out Grok.

Elon: I don’t know. It’s just, yes, there will simply be a lot of intelligence.

Peter: Yes, a lot. I mean, right now, the proportion of human intelligence on this planet is asymptotically plummeting to zero.

Elon: Essentially. I mean, a few years ago, I said that humanity is the bootloader for digital superintelligence.

Peter: Yes, we are traditional, we are the bootloader.

Dave: I...

Elon: I mean, silicon circuits can't evolve in salt pools, you know? So you need a bootloader. With a bootloader.

Dave: But yeah, you would never, ever harm your bootloader.

Elon: So, you know, I...

Dave: You might need it.

Elon: We might be a good bootloader. Yes, and it's nice to ask about it in the future. This...

Peter: It’s just that we want to end the podcast.

Elon: A lot of people don’t know what a bootloader is, even though...

Peter: Good heavens, yes, boot disks are a distant memory.

Dave: We could do 'Always Look on the Bright Side of Life' as a cloned song. Yes, we could clone that. Use it as the closing theme. That would be great.

Peter: I'll come back to you. This is the most exciting time to be alive ever. The only thing more exciting than today is tomorrow. I mean, interestingly, we're heading towards a world where anyone can make their grandest dreams come true.

Dave: It’s like Walt Disney saying it word for word and turning it into a new exhibit. Like, so.

Elon: You might ask about science fiction; it’s, you know, like a non-dystopian future, right? Banks’ books are, yeah, probably the best.

Peter: You should pay the producers to make...

Dave: Those are 'The Culture' series books, which include 'Consider Phlebas,' featuring Gurgeh. Just for my wife, I want her to, because she’s like, what exactly are you reading? Okay.

Elon: The way 'Consider Phlebas' begins, yeah, I mean, that...

Peter: What gives hope is providing...

Elon: For me, it’s the idea of drowning in excrement. This...

Dave: A good start. We really...

Elon: It might offend some people. Yes, you need to have read the first few hundred.

Dave: Don't walk out in the first five minutes of the movie, although they will give it to you, you know.

Elon: Get into it. Yes, it's like 'The Player of Games'—that might be a better book to start with, and...

Peter: Considering that I enjoyed that, and humans still exist in that future, it's a...

Elon: Good thing. Yes, they exist. Humans. In that future, there are trillions of humans. Alright.

Peter: We need to increase the birth rate. By the way, my friend Ben Lam’s company Colossal is developing artificial wombs. He’s the one bringing back the woolly mammoth and the saber-toothed tiger and all of that. When we get...

Elon: Oh, can we have, I want a mini pet woolly mammoth as a pet. Okay.

Peter: Hmm, yes.

Elon: With tusks. Isn’t that cute?

Dave: Yes, just like that on your face. This...

Elon: Just like those... staggering around the house.

Dave: House, you know. What is your optimal size...

Elon: Become? Cute. You know, what...

Peter: They feel like they have learned how to do it.

Elon: Small tusks and everything about it. A mini Woolly Mammoth would be epic. I mean, look at what we did to wolves.

Peter: Right. But you created the Woolly Mammoth. There is a Woolly Mammoth cycle, noticing different tusks.

Dave: I am here.

Peter: I am there. He is...

Elon: Dallas.

Peter: I went to visit him, and he said our scientists are attending a tusk conference next week, okay, to discuss all the genes related to tusks.

Elon: Did they grow tusks on mice? No.

Peter: They might have added them to mice. I'm curious.

Elon: Mouse-sized.

Dave: That would just scare people. A small woolly mammoth would sell well.

Elon: Yes, that...

Dave: Exactly. Mice wouldn't sell.

Elon: That would crush it. I mean.

Dave: Too creepy.

Elon: So Labradors are cool. When you see them, they really matter.

Dave: The saber-toothed tiger is also pretty good. That cat. It's about the size of a ton (Ton? Tongue? Tooth?). The teeth come down like this, how do they bite? But do they really bite? I don't think I've ever opened that.

Peter: Not mine, you know.

Elon: Clumsy, kind of cumbersome, you know.

Dave: They're just for show. They look nice.

Peter: Yes, jewelry, but no dinosaurs.

Dave: No, I'm not. Legal or illegal.

Elon: I wouldn't. I think Jurassic Park was a good idea. I...

Dave: I mean, seriously, I didn't watch the ending of the movie.

Peter: AI will help us with that.

Elon: Things are perfect.

Dave: Oh yes, not really. Well, okay.

Elon: I mean, if there were an island with a bunch of dinosaurs, we’d go, 100%.

Peter: Yes, I would pay a lot of money for that.

Elon: Like occasionally someone gets eaten by a dinosaur, like, what is that, you know, so one in a million, I’d still go.

Dave: Where did they lack lysine?

Peter: No, that’s DNA. The oldest recovered DNA is about 1.2 million years old.

Dave: Oh, you can wing it. Just...

Elon: Yes, just make it look like that. Anyway.

Dave: They actually, that’s what I was going to propose for the XPRIZE. Remember Beck and Visionary? What was that? Take a strand of DNA and predict what it would look like...

Elon: Look like. Yes, yes, exactly. That’s what they made it into.

Peter: And then reverse engineer dinosaurs.

Dave: Exactly. If those are two completely different DNA strands, like, okay, they both look like T. rex. That's interesting. I...

Elon: Is the T. rex real? Or is that a real patchwork?

Dave: That would be very interesting.

Elon: I mean, it's nice to believe it's real, but different legs come from completely different...

Dave: It actually has huge front legs. This...

Elon: There’s something off about the arms. I don’t believe it. I’m not buying it. I don’t know. The arms, friendly arms. It seems impossible.

Dave: Well, DNA will tell us. Um, within a year now.

Peter: The future will...

Elon: Is Jurassic Island re... I go, we...

Dave: Understood. No, I mean the missing amino acid in dinosaurs, which prevented them from reproducing.

Peter: Lysine, you mean?

Dave: Because of lysine.

Peter: I forgot something. I don't remember, but the dinosaurs were stopped by something like an asteroid impact.

Dave: Right. Right. They did...

Peter: Very well.

Dave: Yes, for 60 million years. They did very well. We are very lucky.

Peter: Very well. Much longer.

Dave: Look, here's a good argument. There is no other intelligence out there. There are a lot of dinosaurs in the universe.

Elon: What were we back then? Something like a bowl (Bowl? Mole?) We…

Peter: Yes, we...

Elon: Good.

Peter: Before we engage in large-scale communication.

Elon: With ancestors.

Dave: Very good at hiding. It is.

Peter: Amazing. From a small mouse, a little mole, to where we are in 60 million years. It doesn’t seem that long...

Dave: Long. That’s why no one believes Darwin. Yes, it seems unreasonable.

Elon: It’s a very long time.

Dave: Sixteen. Turns out it is.

Peter: Yes, I mean, you know, you’re making robots, but this is interesting. I think designing bio-robots would be much more fascinating, like a little kitten running around, removing urine stains, eating threads off the carpet. That would be fun, but you...

Dave: There is a mechanical version, like a lightweight Optimus, doing that. Anyway.

Peter: Yes, well, they went bankrupt. So after that.

Dave: And that...

Elon: Roomba (the robotic vacuum cleaner) is basically that.

Peter: It will be.

Elon: The issue is that humanoid robots are general-purpose. So I can do anything you want.

Dave: They were too early. No vision system, no GB300. How do you make a functional Roomba?

Peter: I think the idea of having an Optimus vacuum cleaner is the most underutilized asset concept.

Elon: It can, but it can just do anything.

Peter: It can. Yes, of course.

Dave: Yes.

Elon: So, you can manufacture at scale, in, you know, one. Oh, that is yes.

Dave: Optimus, make me a Roomba. That's what you'll do. You wouldn't say Optimus vacuums the carpet. Optimus, make me a vacuuming Roomba.

Elon: Alright, Larry. But I want this checkpoint done annually. I need that checkpoint.

Dave: That is...

Elon: We will roll back that. Yes, what are we talking about?

Peter: Last year, it was integration. Okay.

Dave: Yes, we either control it, or we can cut it off...

Elon: Cut that page out.

Dave: The bus you're selling is hope.

Elon: Better facts. The results are really good. I will send you...

Peter: Cups.

Elon: Monetization.

Dave: Okay.

Peter: Monetize hope.

Dave: When you walk through this door starting today, December 22nd, if you're here now, we'll be talking about you. I...

Elon: I mean, in a year, we might have the Optimus factory, and the building will be completed.

Dave: That's amazing. Eight million square feet of robotics, right?

Elon: A massive building. Yes.

Dave: Today. Great stuff, guys.

Editor/joryn