Musk's latest interview: AI has entered a self-evolutionary cycle, and humans are disappearing from the "loop".

Written by: Long Yue

Source: Wall Street News

On March 11 local time, in a public dialogue at a technology summit called "Abundance Summit", Tesla and xAI founder Elon Musk discussed AI progress, the mass production schedule of Optimus 3, and the post-singularity economic landscape.

In this interview, he made several clear judgments: AI has entered a self-improvement phase, and humanoid robots are about to enter mass production. While the economy after the AI ​​"singularity" is difficult to predict, he is betting on deflation and that "money will no longer matter."

When asked about the timeline for SpaceX's data center construction, Musk declined to elaborate, citing that SpaceX was in a "quiet period."

"Optimus Prime 3" is about to enter production: mass production may begin next year.

Regarding humanoid robots, Musk revealed that Tesla's "Optimus 3" is nearing completion .

He said, "We are completing the final stage of Optimus Prime 3, which is likely to be the most advanced robot in the world at present, and no other product has come close to its level."

According to him, Tesla plans to:

Musk emphasized that robot production also follows the common S-curve growth pattern in manufacturing: slow at the beginning, followed by rapid expansion.

He said, "Manufacturing output is usually an S-curve, starting slowly and then climbing rapidly."

Meanwhile, Tesla is also designing a new robotic production factory . He revealed that the design of this factory is significantly different from that of traditional factories, and the future goal is to continuously update robot versions, " possibly launching a new robot design every year ."

AI has entered a phase of "recursive self-improvement": AI trains AI, reducing the role of humans.

Musk's assessment of the pace of AI development is equally radical.

When asked whether AI has entered the stage of "recursive self-improvement," he said, "Actually, this has been happening for some time now."

He explained that the current development of large models has formed a cycle:

He said, "Humans are playing a smaller and smaller role in the loop. Each generation of models helps build the next generation of models."

He predicts that this process may soon achieve a higher degree of automation: "Fully automated self-improvement may appear by the end of this year, and at the latest by next year."

In his view, AI breakthroughs have entered an accelerated phase: "The current situation is that I see an AI breakthrough before I go to bed at night, and then I wake up in the morning to find another breakthrough."

After the AI ​​singularity: unpredictable, but he's betting on deflation, universal income growth, and "money no longer matters."

When discussing whether systems can keep up with the impact of AI and robots, Musk used the "singularity" as a metaphor: "The singularity is called the singularity because it is very difficult to predict what will happen."

He mentioned that Grok's logo is "the halo around the black hole," and said, "It's hard to know what happens inside the singularity, but it will be very interesting."

In terms of macro-level judgment, he gave a clearly defined optimistic range: he believed that there were "a series of possible outcomes, not all of which would be good", but "it is very likely to be good", with a probability of "80% or even higher".

Barring extreme exogenous shocks, he said he was “quite comfortable” with his assessment of economic growth: “In the absence of a third world war… I think a tenfold increase in the size of the economy within 10 years is a fairly prudent prediction.”

He attributes the inflation/deflation logic directly to a surge in supply: "We will have universal income, which is basically giving money to people." The reason is that "the output of goods and services will far exceed the money supply," thus creating deflation: "Deflation is the ratio of output to money supply... If the growth rate of goods and services exceeds the money supply, you will experience deflation."

Looking further ahead, he believes the importance of money will decline: "Money will become less important at some point in the future."

He even offered a pricing method for a "non-human economy": "I think future AI will not use human currency; it will only care about energy and quality—power and tonnage."

Employment and "Robots Manufacturing Robots": Instead of layoffs, companies are expanding recruitment, resulting in "exaggeratedly high" output per person.

When asked when robots will be able to participate in manufacturing on a large scale and replace humans, Musk emphasized that the reality is still that there are "a lot of people": Tesla has a total of "about 150,000 employees," of which "about two-thirds are in some form in the factory"; the number of people involved in its supply chain "may be between 1 million and 2 million".

But his assessment is that efficiency will leap forward dramatically: "We don't plan to lay off or reduce staff. On the contrary, we will increase the number of employees." The real change will be in unit output: "The output of each person at Tesla will become nutty high."

This echoes his description of the path of technological progress: whether it is AI or manufacturing, "it is often an S-curve, or a series of overlapping S-curves: slow at first, exponential growth, to a platform, and then another breakthrough that starts the next curve."

Full translation of the interview:

Host Peter H. Diamandis: Viewers, as you can see, I am still working hard to turn "hope" into reality.

Musk, you look fantastic.

I feel great about Diamandis.

Did Musk use any anti-aging serums or similar methods?

Diamandis is our "Longevity Express," and we're moving towards that goal. You're on this path too. I think during our last conversation, you started to accept the idea of ​​extending lifespan.

Musk is, to some extent. I don't know if we want everyone to live forever, but I think extending "healthy lifespan" rather than going through a long aging process and constantly drooling sounds like a good idea. We want to avoid that.

First of all, congratulations to SpaceX and xAI on their collaboration. This is a fantastic move that will power the world's first "Dyson Cloud." I'm curious about your timeline for launching these data centers? How much bandwidth do you anticipate in the first year? Please let us know how quickly you're progressing.

Musk's SpaceX is currently in a quiet period, and I cannot disclose information that might cause trouble.

Diamandis Okay, then let's not talk about that. I understand, but I'm really looking forward to this speed.

This Monday, we spoke here with Eric Schmidt and an executive from another hyperscale computing (cloud services) company. I won't name names, but I'm curious, where do you think we are in terms of AI's "recursive self-improvement" right now? Have we reached it? Do you think Grok is currently undergoing recursive self-improvement? How is it being done? What is the timeline for Artificial General Intelligence (AGI) and Artificial Superintelligence (ASI)? Please give us a general overview.

Musk, I think we've been in a recursive self-improvement phase for a while now. Do you mean fully automated recursive self-improvement without human intervention?

Diamandis Yes, I was referring to AI software.

In Musk's recursive self-improvement process, human involvement is indeed gradually decreasing. Each new generation of models is built upon the previous generation. This has largely happened, but full automation is not yet achieved. Perhaps by the end of this year, at the latest by next year.

Diamandis, do you think there will be a "hard takeoff" (mutational development) of intelligence by then?

Musk, we are currently in the midst of a hard takeoff.

Diamandis Okay.

Musk is here now.

Diamandis Yes.

At this stage, Musk would say, "When I go to sleep at night, AI achieves a major breakthrough; when I wake up, another breakthrough has occurred."

Diamandis, indeed.

Musk admitted that it was difficult to keep up with the pace. It was indeed a bit dizzying.

Diamandis, I think many of the dazzling breakthroughs were brought about by you.

Musk's Grok is currently developing very well. It's top-notch in some metrics. For example, it's best at predicting things, which is arguably the best measure of intelligence. The new version of Grok is excellent.

We are currently lagging behind in programming capabilities. The reason I was late is because I just attended an all-staff meeting on programming, where we outlined all the tasks that need to be completed to catch up with and surpass our competitors in this area. I believe we can do it. I think we should achieve this goal around the middle of this year.

Furthermore, I believe that people do not fully understand how vast the intelligence of the future will be, or how much it will surpass human intelligence to the point that it is impossible to fully comprehend.

Imagine this scenario: suppose we utilize a million times more energy than all the electricity currently consumed on Earth, that's still only about one millionth of the sun's energy output. Essentially, if you expanded the US economy a million times, it would still consume only a fraction of the energy generated by solar power. If we truly measure it on a solar scale, even if you expanded the current economy and electricity consumption by about a million times, you would still only be utilizing about one millionth of solar energy.

But what would an economy or intelligent entity using a million times more electricity than the entire current civilization look like? What would it think? What would it do? That would be a truly grand vision. The challenge we face is that even vaguely understanding that level of intelligence is difficult. But it's safe to say that it could solve every problem you can imagine.

Diamandis Yes, it's like a long journey. You might find that a bit ridiculous, but I really admire that relentless optimism.

Musk, I see you're turning "hope" into money. It's interesting that you've taken the phrase "turning hope into money" to heart.

Diamandis, that's all Grok's doing. That was marketing advice Grok gave me when you were vehemently criticizing me.

Musk, right? So you're monetizing hope. But you were also monetizing pain before that.

Diamandis, that's for sure.

Musk believes that when artificial intelligence and robots increase economic output by several orders of magnitude, it will be to a degree that we simply cannot imagine.

Diamandis We are likely to become one of the few intelligent people on this planet in a very short time, then a very small minority, and finally an extremely small minority.

Musk, yes. Not just on Earth, but throughout the entire solar system. Because if you only develop intelligence on Earth, the best outcome, the energy that intelligence can utilize, is only about one billionth of the energy generated by solar power. If you limit yourself to Earth, that's your best outcome.

Diamandis refers to the portion of energy that we can intercept, right?

Musk: Yes. Earth receives only a fraction of the energy from the sun, which is the vast majority of the energy available in the universe that we can access. So, in reality, the intelligence of the entire solar system is orders of magnitude greater than that of Earth.

Diamandis Elon, may I ask you a question? How far into the future can you see? How many years into the future can you make a reasonable prediction right now?

Musk has difficulty accurately predicting future paths. Many things tend to follow an S-curve, or a series of S-curves. Development starts slowly, then grows exponentially, then enters a linear region, and finally becomes logarithmic (growth slows down).

This is roughly what I've seen in breakthroughs in artificial intelligence. For example, after a breakthrough, it follows an S-shaped curve, seemingly growing indefinitely, but then the gains decrease logarithmically until the next breakthrough occurs. Therefore, progress in artificial intelligence is essentially a series of overlapping and interconnected S-shaped curves.

There was a time in Diamandis when you could perhaps predict what would happen ten or twenty years from now. What are your thoughts now?

Musk, what I'm about to say might sound crazy.

Diamandis, it's okay. Because you've always been a great audience for our bold predictions.

Musk: Yes, I would say that within 10 years, the economy will be 10 times larger than it is now.

Diamandis: Yes. You did say that in five years or so, GDP will grow by triple digits and the economy will be ten times its current size. But given your predictive abilities…

Musk thinks that a tenfold increase in about 10 years is actually a rather conservative prediction, unless something like World War III happens and disrupts these plans. But assuming there is no World War III and current trends continue, I would say the economy will grow tenfold in 10 years.

I like the term Diamandis. Could you give us an example?

Musk plans to establish a human base on the moon.

Diamandis Yes. And we will have humans...

Musk lands on Mars.

Diamandis We will also build a mass projector on the moon.

Musk, I think so. I believe we'll have a mass projector on the moon within 10 years.

Diamandis is amazing. Gerald K. O'Neill's space vision is coming true.

At this year's Abundance Summit, we had four robots on stage together. I'm really looking forward to Optimus. I'm curious about the timeline for Optimus 3, especially when I can buy one or two? When do you expect Optimus to be available for commercial sale? Or will you be using a leasing model?

Musk: We are currently in the final stages of completing Optimus Prime 3. It will be the most advanced robot in the world, unmatched by any other. In fact, I haven't seen any robot demonstration as impressive as Optimus Prime 3. Frankly, perhaps such robots exist, or they are kept secret, but I haven't seen them yet. Of course, I must make sure that what I'm saying is appropriately public.

Diamandis We are live streaming this conversation on the X platform.

Musk, well, that's already very public.

Diamandis Yes.

Musk: I think we'll start producing Optimus Prime 3 this summer, but initial production will be very slow. Production numbers will follow a classic S-curve over time, reaching mass production sometime next year. After Optimus Prime 4, we'll accelerate design iterations. Next year, I might try releasing a new robot design every year and improving it annually.

Diamandis: When Dave Bundy and I toured the Gigafactory, it was an extraordinary experience. Tesla's factory is 11.5 million square feet. And then you said you were going to build a 9.5 million square foot factory there for Optimus Prime, which is just amazing.

Musk's area figure is roughly like this.

Diamandis has to be this way.

Musk's plan would be quite impressive. It would be a completely new factory design, unlike any other.

Diamandis: How far are we from "robots manufacturing robots"? You've already automated most aspects of your megafactories, with humans playing only a small role. Will robots take over the current human roles in manufacturing?

Musk: We still have a lot of human employees involved in manufacturing. Those directly employed by Tesla, those on the front lines manufacturing products, or those in the factories, either manufacturing products or managing the manufacturing process. There are about 100,000 of them. So we have a lot of employees. Tesla's total workforce is about 150,000, and on average, two-thirds of them work in factories in some form. We probably have another one to two million people working with our suppliers.

Therefore, a very large number of people are involved. We anticipate that Tesla's output per employee will become extremely high. Therefore, we have no plans for layoffs; in fact, we will increase our workforce. However, human output per employee will become ridiculously high.

Diamandis reached an unbelievable level.

When you joined us on the podcast, we discussed the concept of "sustainable abundance." You emphasized that we are entering an era of "Universal High Income (UHI)," which goes beyond the scope of "Universal Basic Income (UBI)." I'd like to know your thoughts on how we can achieve this goal? Do you have any further ideas?

Furthermore, we previously discussed the potential period of social unrest, roughly two to five years. During this time, until we achieve "demonstration" and deflation, and ultimately reach a stage of universal high income (UHI), we may need to issue many relief checks similar to those during the pandemic. Do you have any further thoughts on this? People really need that kind of hope and vision.

Musk wants to be clear: I don't think we should be complacent. We do need to be cautious because the future outcomes are varied, and not all outcomes will be positive. But for now, I agree with you that the future is very likely to be positive. There's probably an 80% chance, or even more, that it will be.

I do believe we will achieve universal income. Essentially, this means directly giving money to people. Because the output of goods and services will far exceed the money supply, deflation will actually occur. Deflation is the ratio of the output of goods and services to the money supply. Therefore, if the growth rate of goods and services exceeds the growth rate of the money supply (which I predict will happen), then deflation will occur.

Diamandis: Yes. Many people will start new companies and compete with each other, thus driving down prices, which will increase variability faster and exacerbate deflation.

Musk essentially envisions artificial intelligence and robots creating so many products and providing so many services that humans will practically have nothing left to do manually. For humans, there will always be a moment when desires are fulfilled.

So, going back to my earlier example: if our economy were a million times larger than the US economy, all of humanity's desires would have been satisfied long ago. Even if our economy were a thousand times larger, you might already have met every imaginable material need.

Diamandis, do you think the value of money will decline significantly? Will we enter a post-capitalist era?

Musk: Yes, I think at some point in the future, money will no longer have any real meaning.

Diamandis, so just as you become...

Musk's future society might resemble something like the one described in Iain Banks' science fiction series *Culture*. I believe that future artificial intelligence won't use human currency; it will only care about energy and mass—that is, power and tonnage.

Diamandis, that sounds a bit ironic, doesn't it? Just when you're about to become a multi-trillion-dollar billionaire, your money starts to lose its value.

Musk is about the same. All these wealth figures actually only represent my equity stake in the companies I've founded. This money isn't in a bank account. Strictly speaking, I only own a portion of these companies. These companies are doing many useful things, and as they create value, their valuations continue to grow. The equity stakes translate into my total wealth, making the number appear very high.

Diamandis once interviewed me, asking what my motivation was, what drives me. I replied that Elon's motivation is problem-solving. He strives to make the world a better place by tackling the biggest problems time and time again. If others could solve these problems, he wouldn't have had to step in, but the problem is, no one else is doing it. So, I just want to express my gratitude to you.

Musk was not polite.

Diamandis, I'm curious, do you think democratic systems and modern institutions can keep up with this impending "supersonic tsunami"? Will they be swept away by the tide of time? Will they collapse completely? How should we deal with it?

Musk called this a "singularity" for a reason. It's very difficult to predict what will happen inside a singularity. The hallmark of Grok is the singularity.

I really like Diamandis. By the way, the logo behind you is beautiful and very elegant.

Musk, thank you. That halo symbolizes the mass and light around the black hole falling into it. It's hard to know what will happen inside the singularity, but it will be very interesting. I'm confident that the future will be very exciting. Frankly, I think artificial intelligence and robotics are also our only way to solve the budget deficit and avoid national bankruptcy. You've had some influence on my views, and I've decided to become more optimistic now. We really should be more optimistic.

Diamandis, thank you, buddy.

I wasn't an optimist before, Musk; I probably just dwelled on the negative too much.

Diamandis's combination of optimism and realism is ultimately beneficial.

Musk is absolutely right. You can't be complacent, and you can't just blindly assume everything will go smoothly; you have to work hard to make things move in the right direction. I mean, something amazing is definitely going to happen in the future. If we have highly dexterous and incredibly intelligent bionic robots, it means everyone on Earth will have access to better healthcare than the richest person in the world today. By the way, everyone says I'm the richest person, but I think those monarchs are actually much richer than me.

For example, I had to undergo three neck surgeries because the first two were botched. And my back still hurts a bit. I was wondering if artificial intelligence could help solve back pain. If it could, that would be a huge victory. And I think it definitely can. Back pain is truly awful. The torment of back pain leads to poor sleep quality, which in turn makes people irritable.

This morning, David Sinclair joined us at Diamandis. He's conducting clinical human trials of epigenetic reprogramming. A recently published paper suggests that this therapy can repair joints. Therefore, back pain might be one of the conditions it can alleviate.

Musk's approach would be truly remarkable. Of course, if it were simply about solving back pain, the average human well-being would be greatly improved. Because with back pain, it's not a question of whether you'll get it, but rather whether it will happen sooner or later. There are indeed some flaws in the design of the human body.

Diamandis, I've been wanting to invite you to Fountain Life Medical Center in Dallas. We'll help you. Please come whenever you have time.

Musk, what kind of equipment do you have there? I understand you have MRI and CT scans, but what do you do with the scan results?

Diamandis, I'd be happy to send you the service list; I'll send it to you privately. You're so generous. Next up on stage with another great "moonshot" entrepreneur, Ben Lamm. He runs Colossal, a company that "revives extinct species," and is researching the resurrection of mammoths and 15 other species. I heard you want a miniature mammoth, is that true?

Musk: Yes, I think having a miniature mammoth as a pet would be incredibly cool. It would be absolutely epic. Just imagining those furry, adorable little guys running around and bellowing is exhilarating.

Diamandis, okay, I'll put in a good word for you. That's amazing.

Musk: Can anyone build a real-life Jurassic Park? If it were built, even with the risk of death, I would definitely go see it; it would be absolutely awesome.

Diamandis believes that if anyone can do this, it must be Ben Ram and his Colossal company. He's creating life through engineering. Recently, someone asked him if he could build a Pikachu, and he said maybe.

Musk: Yes, Jurassic World or something, that would definitely be amazing.

Diamandis Okay, I'll ask him. Elon, thank you so much for coming here to share with us. Thank you, my friend. Let's give Elon Musk a round of applause! (Background voice: Nothing can stop us)