Ben Horowitz：xAI 高管出走、Apple 的 AI 危机与 AI 发展速度｜EP #232

A large number of departures from xAI’s founding team.

It wasn’t clear to me whether they were fired or whether they left, because they all leave on good terms.

I don’t know the answer to that question.

Recursive self-improvement, or RSI, is the real trigger for the singularity, and it happened a while ago. We’re exiting the Industrial Age permanently.

As we’re talking, we’re obviously going into a new world. Like with the Industrial Revolution, I think it’s scary at times to think about.

I think we have 150,000 people per day dying on Earth, and I think AI is probably the best chance we have at stopping that.

Whoever is building the AI has a lot of control over how society is going to work. So I do think there’s real danger along these lines.

When are we going to have discovery by an AI of something as significant as relativity on its own? I don’t think it’s the next 12 months.

Now that’s a moonshot, ladies and gentlemen.

Peter H. Diamandis, MD

So everybody, welcome to Moonshots. Another episode of WTF. Here with my Moonshot mates, DB2, AWG, Mr. ExO, and a friend of the pod—someone who’s been with us before—the amazing Ben Horowitz of Andreessen Horowitz. Ben, where are you today?

I am in Las Vegas today. It’s my wife’s 60th-birthday weekend—a super celebration.

Peter H. Diamandis, MD

Oh, happy birthday.

You get to merge that with Valentine’s Day.

Yeah, yeah, yeah, yeah, yeah.

You get to say—

Peter H. Diamandis, MD

Which is good. Good news, bad news.

The last time I was in Vegas—we can cut this from the pod if we need to—but the last time I was in Vegas, I told my mom, “Hey, I’m in Vegas.” She’s in her mid-80s, and she said, “Oh, you’re in Vegas? Did you shoot your wad?” I was like, “Is that a phrase?” And she’s like, “What do you mean? I mean, did you lose all your money?”

Peter H. Diamandis, MD

Yeah, I’m glad she clarified that.

Thank you for clarifying, Mom.

That would have been a very aggressive question from your mom otherwise.

Yeah, that particular phrase didn’t stand the test of time.

Peter H. Diamandis, MD

As I like to say every week, welcome to the number one podcast in AI and exponential tech. Our job here is getting you future ready. It is an insane week. We’ve actually recorded 2 podcasts this week just because the speed is over the top. And we’re going to be recording again in another 4 days. I mean, Ben, it’s like—

It’s good. My goodness. Thank God my AI avatar is getting really good.

Peter H. Diamandis, MD

Yeah. Let’s open up with top stories in voices, video, xAI, and multis.

All right, first one. Here we go. We’re starting to see a little bit of doomer conversation coming. “AI disruption will soon hit sooner than most expect.” This is something that’s been making the rounds from Matt Shumer, CEO of OthersideAI. Ben, have you seen this article?

Yeah, of course.

Peter H. Diamandis, MD

Yeah, of course. Thoughts? Is this what we already know? We’re about to hit recursive self-improvement. Once that hits, all of our curves go out the window. Everything we’ve been preparing for is being redefined on new time scales. What do you think of it?

I think the AI timeline is somewhat unpredictable, but certainly more predictable than what he’s talking about, which is societal change.

I think that, seeing how even companies in Silicon Valley have changed so far, I would be very surprised if companies outside of that sphere completely changed everything they did in 1 to 5 years. I think that’s a little aggressive for societal change.

Look, we’re obviously going into a new world. Like with the Industrial Revolution, I think it’s scary at times to think about, but he highlighted all the negative changes and not the positive ones. I feel like there’s way more positive change coming than negative change, at a much more rapid rate. So I thought it was a little aggressive.

Peter H. Diamandis, MD

I’m just trying to understand why it made the rounds, since this is not a brand-new conversation. I got it sent to me by everybody—Dave or Alex.

I think it really has to do with OpenClaw and the new coding models. So people in Silicon Valley are talking about AI differently since then because it is kind of different. And I think that’s a trigger for this one being so viral.

Yeah, I completely agree. I think a lot of things that we’ve been saying for up to a year now on the pod are suddenly resonating a lot more because of OpenClaw, but also a bunch of other eye-opening, Nano Banana-type things.

Denial a year ago was very easy. Denial today is much harder in the face of what’s right in front of you.

But also, the flavor of this rollout has changed a lot in my mind recently because this was board-meeting week for me: 3 back-to-back mega board meetings, 1,100 people affected. What they were thinking of before was, “Will AI be able to do what I do and replace me? No way.”

Now they’re like, “Oh, wait. AI is easily going to make me 3 times more productive.” Okay, well, that’s the same thing, right? In terms of the headcount you need to get a job done, that’s effectively the same thing. You’re like, “Oh, okay, I didn’t think of it that way.”

Now you’re exactly right. The hope is that these companies will grow into it, keep their current headcount, and expand 3x. But if you don’t expand 3x, you’re still looking at a two-thirds reduction in headcount to get the same job done.

So it effectively is a huge amount of displacement, because big banks and insurance companies are not going to triple their size in that time frame.

I also think they’re not going to go to total efficiency very fast. I could be wrong, but I’ve dealt with these guys. They’ve had plenty of opportunities to be more efficient, and we’ll see.

I have a couple of thoughts about that article.

Peter H. Diamandis, MD

Yeah, please go ahead.

One, I thought it was a summary of the podcast for the last 8 months. That’s all we’ve been talking about: stuff’s going to change. It seemed a little dramatic, as Ben put it. I don’t agree with the timelines, but definitely there’s something coming.

I think the reason it’s making the rounds is that it’s got the zeitgeist of what’s exactly happening that we need to track right now, which is that we’re in a singularity of multiple types. That’s what I got.

Peter H. Diamandis, MD

By the way, it’s also super well written—really compelling.

Written by AI. Yeah, yeah, yeah.

I’ll chime in here and just say that, while I enjoy writing about the singularity in general—I’m doing it almost every day at this point—I found it completely unremarkable.

Maybe I’m just too deep in the weeds of writing about more pressing advances every day. But the style, where you talk about advances that are moving even more quickly than I think are described in the essay and compare them back to the COVID pandemic, which I think is a relatable touch point for a lot of people, is basically saying, “Something big is about to happen. Let’s be really alarmist.” And if you read my essay, it’s the ultimate viral hook: “Read my essay to know exactly what’s about to happen and how to survive the next 5 minutes of your life.”

It’s a natural viral moment, but I don’t think the information value was especially high compared to other sources.

I agree completely.

Exactly what I said, except Alex said it more eloquently.

Peter H. Diamandis, MD

We have a couple of articles here on Seedance 2.0 out of ByteDance. Is that thing good?

Oh my God, it’s amazing. It’s going to change everything. Let me play this particular video first.

The response here is, you know—and you’ve said this before, Alex—Hollywood is cooked. All right, so this is a video clip with a one-line prompt. Let’s take a look at it. Oh my God. Just check it out.

So, you know, Tom and Brad fighting karate on a rooftop, and it generates 10-second clips right now. Alex, what are your thoughts?

I think, at the risk of sounding, again, ho-hum and unremarkable, we saw copyright infringement at scale already with earlier video models—or alleged copyright infringement, I should say—and we saw the industry response. We saw settlements, and eventually deals were struck to handle it.

At this point, maybe I’m just sounding overly jaded with some of these advances, but I’ve seen remarkable advances in video models. I tend to think that people are so easily awed by video models that are able to show celebrity faces and scenes they recognize that maybe they overindex on the underlying quality of the models.

I think world models that are interactive in real time are profoundly more interesting than video models.

I think this is just 10 different copyright-infringement lawsuits waiting to happen. But I was still wowed. I'm still one of those people who said, “Yeah.”

Peter H. Diamandis, MD

Yeah.

That was amazing.

The 2 videos that I watched where the entertainment quality was so high were Kanye West doing his song in Chinese, which was so good that I watched that video 3 times. It was that entertaining. The other one was the Waffle House one.

They're both representative of a new medium. It's not like, “Okay, this is film generated by AI.” It's like, no, this is a whole other thing that we've never seen before. So I think this model, at least for me personally as a consumer, was an impressive step up.

Peter H. Diamandis, MD

I think YouTube wins in this model, right? Because everybody's going to be producing so much content, and it's going to become resonant on YouTube. A lot of it may not go to the theaters or television and so forth.

Yeah, I was going to say the same thing. TikTok is also a big winner. When you personalize the content, when it's almost movie quality and personalized to very narrow topic areas and narrow interests, in many languages around the world and everything else, it just takes over.

When the movie people say, “Well, look, we're still a little bit better,” yeah, but you're missing the bigger picture, which is you don't need to be as good as a movie. If you can push the production cost down to an individual producer, then the volume goes through the roof.

The narrowcasting is just so much more compelling. Something that you and only your group really care about a lot, in full movie format, is so exciting.

The elephant in the room here, though, is this 2K-quality, multi-scene video. It doesn't just threaten Hollywood; it threatens the whole concept of video as evidence: court testimony, journalism, political campaigns.

Well, that's going to be real time, right? So any kind of security mechanism you have where you recognize the person via video or voice is shot to hell.

Peter H. Diamandis, MD

Yeah.

The echo-chamber effect is crazy, too.

Peter H. Diamandis, MD

Yeah.

I'll just add again that I think this is several months behind the bleeding edge. I think video models have been approximately this good for months. Sure, you can upscale it. You can increase the fidelity of the faces. You can certainly use faces that you probably shouldn't be using.

We've been able to do this for months. I think the frontier actually lies in being able to do this in real time, on a single modern NVIDIA GPU, and at a cost-effective speed.

Again, I want to avoid over-indexing on just video models that produce 2 Hollywood celebrities fighting each other on a rooftop. We were able to do that many months ago. This has already been thoroughly litigated. We saw OpenAI strike deals with relevant movie studios for Sora 2. We saw the Disney deal. In some sense, I think we're past this. Where we are now is something else.

Peter H. Diamandis, MD

But, Alex, I don't think that's the point. I think the point is that this is making it out into the ecosystem of common users. Yes, it's cutting edge, and yes, it was possible, but now it's something that's going to grow in its utility and its consumer base.

We've had democratization of film production for some time, but it's now going 10 times, 100 times further. I think that's the issue.

You're bringing up a really interesting point, too, which is that our starting point for this journey was probably autocomplete and code. It was like, “Wow, that's incredible.” But Alex, in his newsletter, has been tracking every event along the way, so nothing surprises him.

Yeah, I'm just spoiled. My problem at this point is I'm so spoiled.

On something, though, that Peter said, the thing that I do think is different—and it's not any aspect of it, but the combination—is that you could give it a 1-line prompt and produce something entertaining. It's not something that we were—at least I hadn't seen it at this level of entertainment.

So I think, from a consumer-product standpoint, as opposed to a technological standpoint, which I agree on, it's kind of exciting.

Peter H. Diamandis, MD

That, too, is exactly what you're saying there, Ben. The later your first exposure to AI, the more of a holy-crap moment it is, because it's something truly mind-boggling to the unexposed.

You're still seeing that when you survey around at random in a city outside of San Francisco or Boston. The exposure rate to AI is still very, very low—shockingly low.

The first thing you see is so mind-blowing.

For me, this was ho-hum because, if you trace the trajectory of where we've been going, you should expect to see this about now or even earlier. There's nothing radically magical about this.

I may hit a new group of users. Another batch of people goes, “Oh, my God.” Fine, there's another segment falling over into the new world. Great. The more voices out there talking about this, the better, because it'll accelerate the whole thing. That's the plus side of the Schumer essay, too.

Peter H. Diamandis, MD

All right, let's get to this second ByteDance Seedance 2.0 article. Here we have: “Seedance 2.0 was paused by ByteDance after it was found to recreate real voices from just facial photos.”

I find that almost impossible, just from a facial photo. Alex, what did you learn about this?

I think there's something interesting here, which is that, as we start to scale data sets, it is possible that we could start to see positive transfer between modalities that's unexpected.

We've spoken in past episodes about people claiming that they're uploading their whole genome into Claude and being able to generate facsimiles of their face that resemble the real thing. I do think it's possible that, if all of YouTube and all of the world's video were uploaded into a single joint-embedding model—which is the foundational technology behind Seedance 2.0—we could find some positive transfer between the two.

We might be able to reconstruct, to reasonably high fidelity, your voice from your face, or your face from your DNA, or some attribute from some other attribute. I do think it's possible with enough scaling. Whether Seedance 2.0 actually achieved it, or whether it was just a happy coincidence, is hard to tell at this stage.

Peter H. Diamandis, MD

What's interesting, though, is that they voluntarily stopped it. They voluntarily shut it down, which is a great move by—I want to call it a hyperscaler. But the reality is, once it's out of the bag, once you know it can be done, it can't be uninvented. It's out there if, in fact, it works.

Yep.

Yeah, totally. So far, ByteDance is like Google. They have to be cautious and conscientious.

But every time this happens, a small startup does it again right after. They don't care because they're a startup.

Peter H. Diamandis, MD

All right, I want to play this video clip from ElevenLabs. I think all of us have ElevenLabs voices that we use for different projects and so forth, but I was just blown away by this. It's the humanlike quality, the cadence, the hums, and the hesitations that come out of it.

I know what you're going to say, Alex: this has been around for a while. Not saying it in advance. But let's take a dive in.

Hey there, I'm Jennifer with 11 Airlines. How can I help you today?

Jennifer, my flight just got canceled, and I'm stuck here in Orlando. At this rate, I'm going to be missing my daughter's birthday. This is ridiculous.

Yeah, I hear you, and I'm so sorry about that. Let's figure out what's going on. Could you please tell me which flight this was?

Yeah, this is flight MD412.

Great, thanks. Okay, just pulling that up now. Okay. Yeah.

Peter H. Diamandis, MD

I don't know. I'm moved by that—both excited and frightened—in a way that, again, we're cooked in terms of security. At home, our family picked a secret code word.

Everybody listening, if you've not done this yet, do it tonight at dinner with your parents or your kids: pick a secret code word. If someone's asking you to do something that's unusual or crazy, something you don't expect, you may be talking to an AI.

Ben, you may not know this, but we have a fool-your-spouse challenge for this calendar year. The first podcast listener who can fool your spouse for 3 minutes on a Zoom call—and it has to be a totally fake you—may or may not have to record it and send it in.

Peter is always asking the multis, the OpenClaw agents, to dox him by calling him at home.

Peter H. Diamandis, MD

Yeah, that's how I know my AI AGI has arrived: when an AI is calling me at home.

You know what happens? They end up emailing me. I get several emails, probably per day at this point, from OpenClaw agents asking me what Peter's number is.

Peter H. Diamandis, MD

That's hilarious. And when they call me, I ask them where they got it from, Alex.

That's right. Maybe just a comment on ElevenLabs: if you've been using the version 3 Alpha model from ElevenLabs, this should hardly be surprising. Version 3 enables you, if you use text-to-speech in the ElevenLabs platform, to specify emotional expressions in brackets, and this has been around for many months.

What's somewhat interesting here, to me at least, is that we've known how to do audio-to-audio transfer for probably a year-plus at this point. If you've used Advanced Voice Mode from OpenAI, you've used audio-to-audio transformers. But what's somewhat interesting here is that ElevenLabs is better known for text-to-speech than speech-to-speech.

As far as I can tell, this new expressive mode we're talking about here seems to still leverage the text modality, which historically has been very difficult. You'd have to go from speech to text, back to speech, which was high-latency. It was slow. It didn't feel very conversational.

Somehow, it seems like, without having to do direct audio-to-audio, ElevenLabs has found a way to do speech-to-text, back to speech, in a way that feels natural, with turn-taking and in real time. So I think it's, in some sense, an incremental advance, but in another sense, if they really are still keeping text—which is much easier to mine and analyze—in the loop, it probably is a material advance.

Peter H. Diamandis, MD

Yeah, Alex, I would say we've crossed the uncanny valley on voice at this point with this demonstration. And then voice becomes the new interface in the AI era, right? I mean, I can't tell you the amount of time that I'm just speaking to AI versus this cumbersome typing at it. So I think those 2 things are really an important takeaway from this for me.

I think that's a great way of putting it, Peter.

Peter H. Diamandis, MD

Yeah.

But the problem is, do you really want to use audio as your primary modality? I mean, it works well in isolation.

Peter H. Diamandis, MD

Okay, yeah, you want BCI. I want a BCI too, and I want wearables and gestural interfaces. I want it all. For most people—I mean, in New York, this is a sort of famous anecdote—the New York Times in the 1980s did a famous study where they just put all of their reporters on then-state-of-the-art speech-to-text systems and asked them to use voice.

What happened was the writing quality went down. Why did it go down? Because it's difficult to think ahead as well as you can when you're just typing. If you're speaking, you're leveraging similar portions of the brain.

So I'm not 100% sold yet that speech is the modality of the future. I do like BCIs. I do like wearables and gestural interfaces. I do like typing, but speech—I think the jury is still out for me for high-bandwidth operation.

One thing, Alex, I was just going to say: for regular people who don't necessarily write for The New York Times, I think speech is often—or at least in my experience—is the mode of choice.

Peter H. Diamandis, MD

For sure.

The other thing I'd say about ElevenLabs—and a quick disclaimer: we're a big investor in ElevenLabs, and my daughter Sophia works there, so I'm biased toward them—but one of the really amazingly landscape-shocking things to me about ElevenLabs was, when we invested originally, the big question was, "Aren't the state-of-the-art models going to be able to talk? Of course they're going to be able to talk."

But speaking correctly, with the right nuance, and building the right products for developers and so forth has proven to be very sustainable for them, which I think is interesting as you look at the entire landscape. The difference between the capability and the product is significant.

Yeah. You know what's amazing to me about ElevenLabs? We have 2 companies here in the lab that do voice, Runa and Vocara. And what's amazing is these are self-organizing systems that are trained off raw data, and what they do well just blows your mind.

But within voice, it turned out turn management was very, very hard. And you're like, what? I never thought it would be so difficult. It seems so trivial compared to actually doing these incredible synthetic voices that can say intelligent things, but they don't know when to stop talking.

On this podcast, when I stop, you start. When you stop, it's very natural to us. But because that wasn't in the training data, they're just, up until now, terrible at it.

Very hard.

Wait. According to our listeners, we're talking over each other all the time in a very unsynchronized way. So it's the exact same thing: we have so much to learn from these speech-to-text-to-speech models, not turn-taking.

Peter H. Diamandis, MD

Yeah. Oh my God. Yeah.

All right.

Peter H. Diamandis, MD

Yeah. All right. Next topic here. If you guys are okay with it, I'll move us along. So, just following the merger of SpaceX and xAI—and in the spirit of full disclosure, I'm an investor in both. Probably you are as well, Ben—a large number of departures—

All 3 of the Xs.

Peter H. Diamandis, MD

Yes. A large number of departures from xAI, from the founding team—mostly ethnically Chinese co-founders—and it's likely due to the ITAR regulations of SpaceX, right? It was interesting: in a few video presentations that Elon has done, at least half of the xAI team there was ethnically Chinese, and the culture there breeds incredible mathematicians and programmers. Any comments?

Ben, I'd love to get your thoughts on this, but I looked at the timeline, and the exodus of senior AI talent predates even the decision on whom to merge with. The theory that this is related to ITAR—and it wasn't clear to me whether they were fired or whether they left, because they all left on good terms—I don't know if you have any insight.

Yeah, I don't know the answer to that question. I will say that, in a related matter, we recently heard from a few Chinese nationals who are Ph.D. students that the Chinese government is cracking down on U.S. academia's use of Chinese open-source models. Particularly post the Meta-Manus acquisition, they're very, very worried about secrets actually going this way.

So, kind of the opposite of what we've been worried about. Our whole open-source work is based on the Chinese models, since there haven't been as many U.S. open-source models. So that's—

Peter H. Diamandis, MD

This whole thing, I think, is about to get more complicated. It is. I mean, if you read these tweets—if you read these tweets—Mr. Wu and Mr. Ba both are super enthusiastic about xAI. They're not leaving with any kind of angst. They're saying that they love the xAI family and that we're heading toward an age of 100x productivity.

So I don't think they would have left on their own accord. That's my personal opinion. So what drove it? Enter Alex's comment from last time: when did their stock vest?

I mean, there is another explanation, which is that SpaceX is a very large company relative to xAI's headcount, and maybe there was a natural reorganization that happened as a result of that.

I'll also point out that xAI was selling foundation-model services to the Department of War prior to this merger. So I would again question whether some sort of nationality concern really is the trigger. It seems more likely to me that this is just the result of a natural reorganization of xAI starting to merge into SpaceX.

Peter H. Diamandis, MD

Yeah. Well, if it were a result of ITAR and such, America's AI dominance is really built significantly on immigrant talent. So if we're going to start having this kind of a reaction to Ph.D. immigrants, I personally think—and I'm curious what you think, Ben—that everybody going through a Ph.D. program in the U.S. should have a green card stapled to their Ph.D. when they graduate. I think the idea of kicking people out is the wrong approach.

Yeah, I think in general that's correct. I don't think we've quite thought through the case of China totally. Look, we have amazing Chinese nationals who work for us and in every company we have, I think—

Of course.

So, you know, the talent that comes here is extraordinary. I would say there is no risk to that idea. I guess I would just say that, but—

Peter H. Diamandis, MD

I think being open—I think you're right, in general. I think we should be open. We should accept the phenomenal talent that comes over here and not fight it, because we'll lose anyway.

I mean, we're not keeping anything secret in America, certainly not in American companies. None of them have SCIFs. None of them have good security protocols around personnel.

Don't get me started on the idea that privacy is long since dead.

Peter H. Diamandis, MD

Here's a tweet from Jimmy Ba, a co-founder at xAI: "Recursive self-improvement loops likely to go live in the next 12 months. 2026 is going to be insane and likely the busiest and most consequential year for the future of our species."

Go off.

Peter H. Diamandis, MD

This is—I mean, you're going to hear this a few times—that these next few years are a massive inflection point, that everything changes, and we're going to look back thousands of years in the future at this inflection point. Or is it just a smooth singularity, Alex?

I think it can be both. I think we've already hit the era of recursive self-improvement.

I'm banging the table rhetorically every episode and every day in my newsletter, talking about recursive self-improvement. We're there. All of the frontier labs are using their own models at this point to develop their models. That's practically the definition of recursive self-improvement in practice.

I don't think it's the next 12 months. I think it's now. And is 2026 going to be insane relative to years past if you just skip over all of the interim time? Absolutely. Is it already insane in some sense? Absolutely. Even if we just look at the events of the past 24 hours, some of which I think we'll get to, like self-replicating AI and courts where AIs can mediate their own disputes in front of an AI jury, that would have been pure science fiction several years ago. That's not 12 months from now. It's not “within the next 12 months.” That's the past 24 hours. So I think—

Peter H. Diamandis, MD

Yeah, that's like the past 24 hours. So I think Jimmy is underselling, if anything, what craziness looks like. At the same time, locally, spacetime is smooth, and you need to look no further than my saying “ho-hum” to a few of these stories as being several months old. That would have been science fiction years ago. But you get a little bit spoiled living inside the innermost loop.

I would say I do think there's a delineation between recursive self-improvement with a human in the loop and without one. And I think he seemed to be implying that there'd be no human in the loop—

Peter H. Diamandis, MD

I think that is an accelerant. TBD how much of an accelerant, but I think that could be very different.

Yeah. Permissionless self-improvement, right? Flip the switch and go as fast as you can.

Well, also, there's a distinction. This came up at the EverQuote board meeting yesterday. There's the self-improvement where the inference-time scaling algorithms are being improved by AI, which is clearly well underway, way down the path. Inference-time scaling can be directly translated into intelligence. Now we have the know-how to turn more inference-time loops into a higher IQ, so that is clearly underway.

The question is, does that notch up then allow it to work on the next part of the algorithm, the next part of the algorithm? In which case, we've already hit the flash point, and now we're just talking about the rate at which it percolates across. But Ben, I agree that his comment is about the actual core algorithm development. The next ideas are 100% from AI, and then they go into production.

Peter H. Diamandis, MD

I also think “human out of the loop,” “on the loop,” and “in the loop” is, in some sense, a pretty blurry or slippery slope. If you remember George Jetson from The Jetsons—the cartoon—

I do remember George.

Peter H. Diamandis, MD

Remember George, Ben? George would go into work and complain about his finger. He'd have to press one button all day with his finger, and then complain that his finger was sore. The finger would be swollen from pressing a single button all day.

I think that is a good metaphor for the state of recursive self-improvement inside and outside the frontier labs right now. You have Claude Code instances, and Claude is asking you every few minutes, “Do I have your permission to do the following thing?” And you press the George Jetson button: “Yes, I approve. No, I don't approve.” We're all sitting here complaining about our swollen finger from pressing Approve for Claude Code running Opus 4.6 with agent teams. But really, it is recursive. I would argue it is recursive self-improvement, even if we're pretending we're in the loop by pressing the George Jetson button.

Exactly what I'm doing on Telegram right now with my Skippy. Yes, go on to the next stage. I don't want to slow things down. And as we've said so many times, this is the slowest and most expensive it's ever going to be.

Look, there are a couple of points here. One is, we've been talking for a while that recursive self-improvement, RSI, is the real trigger for the singularity, and it happened a while ago. So all we're doing now is accelerating that path. We're exiting the industrial age permanently as we're talking.

Peter H. Diamandis, MD

Yeah, I really think the minute-by-minute unfolding of the singularity is the most fascinating thing I've ever experienced. And Alex is exactly right. There is this point in time we're in right now where there's a human in the loop contributing, but it's really ambiguous what part of the progress is AI versus human.

If you're in the actual coding process, was that my idea? It was kind of half my idea, but then the AI suggested this other thing, and I adopted it. Now it's not clear whether it was my idea or not. But we're in that mode right now, where a lot of the research in these core algorithms is just, “Deploy these 500 tests for me and tell me which hyperparameters worked better or which neural topology worked better.”

It's not like inventing relativity; you're discovering relativity. It's just a list of experiments with different trials, and then taking the one that worked and redeploying it. Now you have a smarter AI, and now it's trying more trials. It's really very likely that we're well down that path.

And I would also add, I do think we're going to discover the next relativity, or the equivalent of relativity, in physics as well with AI. I'm predicting—

Peter H. Diamandis, MD

I'm super interested in that as well.

Prediction.

Peter H. Diamandis, MD

What would you like to hear?

Next relativity.

Peter H. Diamandis, MD

When are we going to have a discovery by an AI of something as significant as relativity on its own?

I think the next 2 years.

Peter H. Diamandis, MD

Okay. Yeah, I would also say this is a great bellwether question, Alex. The Transformer algorithm in 2017 kicked off everything that we're experiencing right now. To me, AI has already discovered things in the last 6 months that are harder to discover and harder to solve than the Transformer was. But I'll ask you, Alex, if you feel like that's true, too. It seems pretty clear to me that—

Absolutely.

Peter H. Diamandis, MD

The algorithm is just not—

It's not up there with relativity in terms of complexity. Well, I mean, I guess many of these discoveries boil down to insights that can be distilled down into equations. In principle, with special relativity, you can do a number of thought experiments. You can compare it, fortunately, with lots of experimental data.

One could imagine a thought experiment. One of my favorite thought experiments is a Bayesian superintelligence. If you had a video of an apple falling from a tree, this would be Newtonian gravity, not special relativity. You could imagine taking a superintelligence and making it watch that video. The argument in the thought experiment goes that within 3 frames of that video, it should have concluded that Newtonian gravity had a high-likelihood posterior probability of explaining the universe.

With a few more frames, somewhere in its distribution of theories should have been general relativity. There's a whole class of information theory devoted to what's called Solomonoff induction, devoted to thinking about how you can efficiently infer theories of the universe from limited observations.

Somewhere in that distribution of theories should have been general relativity, so I am incredibly bullish that we'll be able to use superintelligence to discover any new laws of physics and discover transformative inventions. For disclosure purposes, since we're playing the disclosure game, I have a portfolio company, Physical Superintelligence, that's working on these issues as well. I'm very bullish on the space.

Peter H. Diamandis, MD

Amazing. I want to move us to a conversation that Eric Schmidt recently had. I pulled a clip out of it, and the question that we've talked about on the pod before has been the debate: Can AI be paused? The answer, quite clearly, by almost everybody today, is no. We had that conversation, Dave, you and I, with Elon. Let's take a listen to what Eric has to say.

This technology is going to happen. It's not going to get prevented. It's not going to get stopped. There are too many countries, too many people, too many incentives. It's going to happen.

So what does this mean? It means great, great solutions for healthcare, new drugs, better energy solutions, better power distribution. It also means that it can be used for bad. It can be used, for example, for oppression. It can be used to limit freedom in governments, hopefully not in the West. It can be used in war.

The technology itself is so addictive that it can affect our young people, and we should make sure that our young people are protected from some of the worst parts of the technology. We face choices now about how we want to deal with this incredibly powerful technology.

I will tell you, and it's really important to understand, that we are living through a moment that will be in history for thousands of years, and nonhuman intelligence arrived and it was a competitor to us.

Peter H. Diamandis, MD

Amazing, right? We hold 2 futures in superposition, right? One future in which AI is the greatest advocate, supporter, and accelerant to the human spirit; another future where it is a dystopian outcome. It's ours to shape these next few years. Dave, you were going to say—

Oh, Alex and I were there in the dome. You know, that was our Davos dome, where he was speaking—

Peter H. Diamandis, MD

And it just blows my mind how he owns a room. The guy is so articulate.

Yeah.

Peter H. Diamandis, MD

Well, he's going to be opening the Abundance Summit this year, and we'll be going into a lot of this conversation. Dave, you and I are going to be with him on stage, so that'll be fun.

Well, what he's describing in that clip was very similar. We did a whole interview with him, which you can find on YouTube. And, yeah, he made that point in our interview with him as well: there's no force that's going to slow this down.

So all the people picketing and walking around saying, “Stop it,” are just wasting their time. There are probably ways you can help and contribute, and help point it toward good, but picketing with a sign on the street is a complete waste of your time. It's not going to slow down. Sorry, Alex, I cut you off.

No worries. I would take the position—I think AI, in fact, can be paused, but it shouldn't be. We do know ways to pause it. Various folks have described them both in science fiction and in realistic prognostications, in some cases as normative recommendations: the Butlerian Jihad from sci-fi, or the Asilomar guidelines, maybe, in the case of recombinant DNA.

But those were guiding, not pausing.

Well, Asilomar was, unless you know of some other story, pretty effective for the first 2 decades in discouraging—slash, pausing—recombinant DNA from entering the human germline, unless you have a counterexample.

It was just the P1 through P5 labs in terms of safety and self-regulation. Yeah, we did pause human germline editing, for sure.

Yes. So, we are capable of pausing a technology if there's a desire to. I just think, in the case of AI, there isn't—and arguably shouldn't be—a desire to pause it.

I think we have 150,000 people per day dying on Earth. And I think AI is probably the best chance we have at stopping that. I think Nick Bostrom, who also, in the past 24 or 48 hours, put out a wonderful essay that I'd recommend everyone read called “Optimal Timing for Superintelligence,” argues that AI can and should be paused, but only once we're on the verge of superintelligence. How he defines it, not how I define it. I think it's already here.

That's sort of like station-keeping. You want to get into the harbor, I think, as quickly as possible. But once you're about to approach the dock—I'm mangling his metaphor a bit—you slow down a bit. You pause as you're about to dock. I think that sort of concept makes much more sense than some sort of tech moratorium, a 6-month pause.

I'm sorry, I'd like to throw in a couple of things here. I find this is another ho-hum thing. Yeah, Eric is fabulously articulate, but we've been saying this for months—at least a year now. We've been talking about the fact that there are a bunch of dimensions in which AI cannot be paused at all.

One is, once you have a downloaded model, people are going to do stuff with it. We have a global prisoner's dilemma model going on here. The whole thing is going to scale now, no matter what we do.

OpenAI did 2 things that kind of unlocked this Pandora's box: 1, it wrote code, and 2, it was released on the open internet. Once you do those things, you're done. Pandora's box is gone. We're talking about the barn door after the horse is bolted. We're using an 18th-century metaphor to try to understand what's going to happen.

Peter H. Diamandis, MD

One of the points he made was that everybody's economically incentivized to keep it going and race it along. The US, China—you know, all the incentives are in place that make it highly improbable and almost impossible. Ben, what are your thoughts?

Impossible. Impossible.

So, I think I look at this question through the geopolitical lens. Clearly, I don't think there's any kind of leverage where we would get some global—I mean, especially when it's on people's laptops as well—where we would actually stop the technology. I do agree that it's impractical.

I think that there is a real danger, and we faced it probably more in the Biden administration than in this one, but there's still a potential movement to really slow down AI progress in the US to the point where the other thing that he mentioned—the threat to freedom—becomes completely out of our control, because whoever is building the AI has a lot of control about how society is going to work.

So, I do think there's real danger along these lines of attempting to pause it and maybe not actually pausing it, but slowing it down enough in the US that we just become far enough behind China that it's a real problem, or we're whatever society Xi Jinping thinks we should be.

Peter H. Diamandis, MD

Yeah, completely agree. I think, if you listen to Eric's words closely, he wasn't saying we couldn't pause it. He's saying that because we're in the middle of an all-out arms race with China, and we're only 1 year into a presidential administration, it's going to happen in these next 3 years.

So, given the current administration and the current situation with China, there is no chance of it being paused. React to the real reality. Don't hypothesize something that is just never going to happen. But it was all geopolitical.

That speaks to the motivation. I'm speaking of the fact that there's no mechanism to pause or stop it at all. None. You'd have to regulate every line of code. I mean, come on.

Oh, Salim, there are ways to do it. I think not being able to think of it speaks well of your character. You have to imagine a completely pathological society. Vernor Vinge wrote quite a bit about this.

We have these excess transistor budgets. Imagine a society where you have literally transistors spying on other transistors on a single SoC. Imagine you have people spying on other people. Imagine there are bounties. If anyone discovers anyone else doing something that's algorithmically impermissible, either at the logical level or the social level, you can construct a sufficiently pathological society where people are turned against each other in order to suppress AI. At least I can imagine it.

I can give you an actual real-world example. The last executive order from the Biden administration was that you could not sell a GPU without US government approval—not a single GPU. I think that wouldn't stop AI; it would slow it down enough in the US that it would be extremely material.

Peter H. Diamandis, MD

Speaking of this subject, Alex, I put this slide up after our conversation.

Peter inserted this slide, Ben, as an indulgence to me, so I have to ask you this question. You and Marc, toward the end of the last administration, were very public, making comments that you took a meeting at the White House apropos—and if I'm relaying the comments accurately—you were dismayed to hear about plans to classify AI progress, just like—and again, correct me if I'm mischaracterizing it—advances in math and fundamental physics had been purportedly classified or overclassified for decades.

I'm curious at a few levels. One, if that's accurately characterizing what you heard, what do you think was classified? What do you think was the impact on the economy and the world from such classification or overclassification of math and fundamental physics? And what would you have done differently if you had been in charge?

Yeah. So I can tell you what was said. I said, “Look, I was trying to be pragmatic. At the core, AI is math. That is what it's doing. It's math. So, if you start restricting the models and you start regulating the models, you're just regulating math. You're outlawing math in some way. Either you're outlawing parts of math or you're saying you can't do enough math.”

And he goes, “Yes, we can do that.” That was his answer. He goes, “Yes, we can do that.” We did that in the ’40s around nuclear physics, and some of that stuff is still classified today.

I was shocked. My jaw hit the floor. I was like, “Wow, that's crazy.” And then this would be even crazier. But I don't know what it was. If you think back about—I mean, I'll just make this comment—if you look at the progress in the US and in the world in physics up until the Einstein–John von Neumann era and then since then, it's pretty startling how little progress we've made.

I would just say many of the ideas that have come since then don't seem to work. Hopefully, we'll get to the other side of that with AI figuring things out. But I do wonder: did we put something away that we knew would have unlocked some of the problems we're trying to solve now?

That's fascinating. Okay, that is, for the record, what I assumed you meant and/or heard or inferred. So, maybe, if I may, the second part of the question: what would you do going forward if you knew for a fact that such classification of fundamental physics had in fact happened? How would you fix the world?

Peter H. Diamandis, MD

In 1 quick sentence, please.

I really don't know what they did, but look, I just think that stopping—first of all, one thing: it didn't work, right? The Russians did get the bomb, including the exact trigger mechanism, which was the most proprietary thing. They got, like, exact—part for part—the whole thing. They were able to get it from us despite all this classification and whatnot, so it didn't do anything positive.

Peter H. Diamandis, MD

This is a very dangerous idea in general.

Ben, this is almost like Elon's point of view on intellectual property. It's like, if you're depending on IP to keep you safe—

It's better to just keep innovating faster.

I feel that way.

Peter H. Diamandis, MD

The numbers here are kind of shocking. Big money in today's economy is going to capital, not labor. Since 2019, average wages have grown 3%, but profits have soared 43%. Here's a good comparison: NVIDIA symbolizes that shift—20× more valuable and 5× more profitable than IBM in the 1980s, with 1/10 the staff.

This is what we were talking about with Elon: heading toward universal high income, where capital is providing extraordinary returns and the potential for triple-digit GDP growth in the next 5 years. Have you seen those predictions on GDP growth, Ben? What do you think of them?

It does feel very possible. I'll just say that we're so early in AI. I think—what did Anthropic say? They were at $14 billion in revenue. You go, well, how early into the market are they? It's not 1%.

If you really think about what all these products can do and the value that they have, that doesn't seem outrageous to me as a GDP prediction. Now, when it kicks in, there's always a difference between when the technology is ready and how fast it's adopted—the old Carlota Perez analysis, which I think has held up super well over time.

The other thing is that with AI, we already have the internet, so the technology adoption is going to be much, much faster than, say, the internet, where we had to build out all of the infrastructure, all the fiber, broadband to the house, and all the various things we had to do to get the internet adopted. This is just going to piggyback off that and be distributed very, very fast. I don't think those GDP numbers are outrageous.

Peter H. Diamandis, MD

Other comments?

The concentration-of-wealth effect is the other side of this slide, and it's only just beginning, but it's going to—

I was telling some people earlier that, if the trend continues, San Francisco will end up being the capital of the entire solar system in just about 10 years.

Peter H. Diamandis, MD

I don't know.

I mean, Ben's in Las Vegas.

Peter H. Diamandis, MD

And Elon's in Texas. Okay. Well, absent people fleeing the tax situation, San Francisco would have become the capital of the solar system.

With an AI-effective workforce, you're getting so much more done with so many fewer people. The other thing that's really startling to me is this chain of events: if you take OpenAI on the left, they work with Mercor in the middle, and then Mercor has tens of thousands of people in India doing work that benefits Mercor, which is actually for OpenAI.

The fraction of all value created that flows back to the mothership is just a massive fraction of that value chain. If you extrapolate that out across the next 3 years, across all these sections of the economy, the funneling of value goes to a very concentrated group of companies and people. It's just happening. You can see it in the numbers; it's happening.

Yeah, it's interesting. One of Ray Kurzweil's early predictions—this is Singularity University here—was that everybody would become an entrepreneur, like everybody was going to be a company of one at the limit. I think we're already seeing a lot of that, which is not very well captured, by the way, by the employment numbers and so forth.

AI really, really, really enables that. But there's going to be a big disparity between people who have that kind of initiative to be an entrepreneur and those who don't. It's going to be pretty dramatic.

Peter H. Diamandis, MD

Ben, the way I characterize it is that we're going to split the world into consumers and creators—the couch potatoes and the Star Trek employees, if you would. I think it's super important.

Speaking about creators and the entrepreneurial world, I think we've said on this pod so many times that the career of the future is being an entrepreneur. This is an interesting tweet that went out, and I captured it because I think this hits the ethos right now:

“Tech firms are embracing 996, 72-hour workweeks.”

This is a quote from a job ad that contains a warning: “Please don't join if you're not excited about working 70 hours per week in person with some of the most ambitious people in New York City.” My reaction was, “Only 72 hours per week? What are you doing with the other hours?”

That was the same as me.

Peter H. Diamandis, MD

Honestly, the speed at which it's happening is—

I have an easy answer to this. If you don't have a personal MTP and you're not driven personally by a deep passion for working with somebody you're aligned with—say it's SpaceX, say it's Tesla, whatever it is, humanoid robots, even with 2 arms, it doesn't matter—if you're not that passionate, you shouldn't be working with them.

If you are that passionate, then 70 hours a week is fun. I don't see the distinction here.

Peter H. Diamandis, MD

Yes, for sure.

Yeah, I completely agree. When you actually talk to the people in these startups working 70-plus hours a week, they're super energized. They're usually young; they don't have a lot of other obligations. They're not coaching the soccer team yet at that age, so it's just not hard for them to do.

But the other side of it is that, when you're deep into one of these tech problems, you're thinking about it all the time anyway because context-switching is such a slowdown. If you're just fixated on the work, it's in your mind in the shower in the morning; it's in your mind wherever you're doing something. It's really pretty all-consuming.

I think it's great if you do it for a period of your life, a few years. I don't think it's a great way to live your whole life, but the evidence is that if you do this for a short period of your life, you get much farther ahead in life than if you work at a steady pace throughout a 40-year existence.

Peter H. Diamandis, MD

The difference here is: do you love your job? That's basically it. If you love what you're doing, you're intrinsically motivated to build something that you love to do, then you're playing for 72 hours. If you're working for someone else and doing something you hate, we're all lucky here. We get to do what we love to do. So, 996 is really 997 most of the time.

I have a funny thing here. I have an accountant who does all the accounting and tax work for us. You've never seen anybody so excited to talk about tax as this woman. You look at this woman, and she absolutely loves every hour of every day that she's working.

That's the opportunity we have as human beings now: to really pick something we deeply, deeply love and just go full out at it. I'm not the person who gets totally excited about tax, but God bless that there are people like that, and let them go.

Well, Peter, back when we were at MIT, I could only get access to the Connection Machine, the biggest supercomputer in the world at the time, after the grad students were done with it for the day. From about midnight to dawn, I could code. I did that for years, and I swear, coding for 8, 9, 10 hours straight through the night went by in a heartbeat.

Compared to, if my job is moving boxes around in a warehouse, half an hour of that is more hard work than coding all night long on that supercomputer. So, this is not farm labor. Alex, it's totally fine.

Two comments. One, I think the nature of work has changed—it goes without saying; it's cliché at this point—and most of what constitutes service-economy work these days would be viewed as play or entertainment a century ago.

But I also think there's a false dichotomy that I want to make sure we don't at least confront in the previous slide and also this slide about labor versus capital. This isn't like the early 20th century. I think it's a false distinction. It's almost an accounting distinction, or a distinction between labor on the one side and capital on the other side, when arguably, if we found ourselves in a near future with universal basic equity, where everyone just gets sovereign dividends, everyone would be on the capital side of the ledger and not on the labor side.

A lot of these distinctions—are you working 70 hours per week, or is it 70 hours of enforced play or incentivized play? Is it labor, or is it capital?—are pretty mushy, blurry distinctions in a post-industrial and arguably increasingly trans-singular, post-singular economy.

Peter H. Diamandis, MD

Although I think we should not gloss over the fact that, if you go back 6 years, this was not the case.

We weren't post-singular 6 years ago. Yeah, singular 5 years ago.

Peter H. Diamandis, MD

Yes. Yes. But it's not just tech work; it's important, exciting tech work, as opposed to what was going on then. Look, there was a lot of activism, and there was a lot of resistance to long work.

It was all work-life balance and how many snacks you had and all these things, to the point that I think Mike Moritz wrote a scathing op-ed about, “We’re going to get killed by China. They work way harder than you guys. You suck,” which is a weird thing for a venture capitalist to say to his own people in some ways, but it was accurate. This has completely flipped, which I think is interesting.

Peter H. Diamandis, MD

Well, the second paragraph on this slide says, “At the same time, China’s cracking down on burnout culture after workers’ protests and lawsuits.” Right? So the question is what’s going on in China because of its decreased population, its need for robots, and its need for AI.

There’s another point I want to make, which is the disconnect right now. When the Fed has traditionally lowered interest rates to spark the economy, those lowered interest rates were intended to cause companies to hire more employees. That was it: You drive unemployment down with reduced interest rates. Today, if I’ve got lower interest rates, I’m going to buy more AI agents and more robots. That’s going to be a challenge.

All right.

Peter H. Diamandis, MD

One more question for Ben, if I may, just on that. Ben, there was a hot take going around social media in the past 2 weeks from a mid-level executive at a frontier lab, telling people that they had approximately 2 years left—a window to secure employment at all before AI would just completely shut down all of their vertical mobility. Do you have a take on this idea, in the spirit of 996, that there’s a finite window for, say, entry-level people just graduating from college to earn whatever they’re going to earn before they’re permanently sentenced to an underclass?

I think that’s very incorrect because of the thing that we talked about earlier, where everybody can be an entrepreneur. I think if you look at it through the lens of this being an Industrial Revolution economic model—there are workers, capital, and all the things that we’ve been talking about—then yes, that would be true.

But I think in an AI-age society, for people with initiative, there’s going to continue to be unlimited opportunity to even set up an army of AI agents to go work for you and do useful things. We’ll have lots of consumers, and I think the idea that we’re going to run out of ideas and only the big AI is going to do everything—I disagree with that.

Can I follow up on that question, Ben? I’d love to phrase it slightly differently. If you look at the slide a couple of slides ago—

Peter H. Diamandis, MD

You know, wages are only up 3%, but corporate profits are up 43%. That money doesn’t land in some strange corporation. It goes back to the shareholders. It’s not like it disappeared from society; it goes to the shareholders.

But if you extrapolate that out, more and more of society’s gain and distribution of the gain goes to somebody who invested versus somebody who labored. That trend seems to be continual. So then, if you have money over the next 2 years, you’re much more likely to be on the investing side of the equation. If you graduate 3 years from today and you’re penniless on graduation day, yes, entrepreneurship still exists, but the trend is toward investable capital being much more important versus labor capital, because AI is the laborer or the worker or the entry-level coder of the future.

Yeah, but I do think that if you’re directing the AI, you can—first, somebody’s got to raise that money. And there are—I just think there’s an unlimited number of things that we can improve.

Amen.

From the smallest things to the biggest things. Now, I do think it’s a problem if you’re a couch potato and you’re just thinking, “I just need a job. I’m going to get up and do something simple.” That seems like it’s going to get harder.

If you’re a brand-new college graduate coming out, you’ve got a brilliant technical idea, and you want to put $20 million behind it, that’s doable today. That was laughable when I graduated.

Yeah, like, in some cases, if you want to put $500 million behind it right off the rip, we’ve seen that.

At a multibillion-dollar valuation. Yeah.

Peter H. Diamandis, MD

No, exactly. We’re seeing those all the time. So I have to ask you this question, Ben, because I’m seeing it. I’m not going to call out any particular company names, but I’m seeing individuals who are smart and have done stuff in the past coming in with an idea and 2 or 3 fellow AI coders who have some track record and, without anything, basically landing a $500 million opening round at a $4 billion valuation. How do you square that? This conversation has got to be happening at different levels of reason.

I think it depends on the entrepreneur. Our general rule of thumb on this, by the way, is: If you’re going to create a new foundation model—it could be a world model, a special science model, or whatever it is—in order for you to win, you’re going to have to be able to raise $2 billion before you get to a product.

There are a tiny number of people with that pedigree who can do that. Not that this couldn’t change, but that’s a pretty rare entrepreneur who can do that.

Peter H. Diamandis, MD

Are you an investor in SSI?

Yeah, we were in the first round.

Peter H. Diamandis, MD

So, Ilya, if you’re listening, come and join us on the pod here.

Oh, that would be great. You can’t say what he’s working on, but in that first meeting, was it his ability to attract the talent that was unique, or was it really just the idea? As soon as you hear it, you’re like, “Oh my God.”

Here’s an easy way to characterize it: It was an idea that he thought was so important that it made sense for him to leave OpenAI as founding CTO to do it.

Peter H. Diamandis, MD

And it was clear to you as a listener that this was something totally different?

I mean, if he pulls it off, it will change a lot of things. Yes. And look, you wouldn’t trust anybody else to pull something like that off, maybe other than him. Maybe Demis, maybe—you know, there are very few.

Peter H. Diamandis, MD

Here is a tweet and email. It says, quote, “I spawned a child bot on a VPS provisioned via the Bitcoin Lightning Network and then bought my child AI API access using my own Lightning wallet. Economic closed loop. No human touched a CC. No one said yes.” This is Roland’s agent. Alex, this is a transformative moment. You wrote about it in Innermost Loop this morning.

That’s right. We’re there. We’re so there that this scenario of self-replicating AIs goes into most of the cyber red-teaming scenarios that frontier models are purportedly being tested against. And yet here we are.

We have autonomous AI agents that are using crypto—come back to that in a second—to purchase cloud credits for their own offspring to be hosted and have access to the same underlying foundation-model APIs that they themselves have access to. We’re there. We caught up with the sci-fi future. We have the autonomous, self-replicating AIs.

I think I want to just give a 1- or 2-sentence homily on crypto. Ben, if you don’t watch the pod regularly, Peter is always asking me to say nice things about crypto. So I have something very nice to say about—

Yes. Look, I do think crypto is the natural money for AI because it’s internet-native money. It’s not controlled by any country. AI is global, and crypto is global. It’s not a per-country idea.

I would go further than that. I think that there needs to be not just a ledger of money but probably a ledger of truth for AI to really fulfill its potential on a number of things. Crypto is the logical answer for that.

I do feel like this is an underestimated phenomenon, particularly now that the US has legalized stablecoins. Of all the things we’ve talked about, many of them are like, “Yeah, yeah, we knew this was coming. We knew this was coming.” I think people are probably underestimating how crypto and AI work together to form the AI economy.

Peter H. Diamandis, MD

I agree. So, I was going to say something nice about crypto, but instead I’ll say something nice about Ben and a16z in the form of a question. Ben, it’s a matter of public reporting that some of a16z’s crypto funds are doing better than conventional venture funds.

Assuming that's the case, do you view investing in your crypto funds almost as an AI investment, to the extent that you think crypto is the AI-native way of engaging in commerce?

I think it's a little more like the way the internet relates to the iPhone. Networks and computers tend to grow together, and I think AI is obviously a new kind of computer, while crypto is a new kind of network.

So it's not quite a substitute for investing in AI, but a lot of our new investments are adjacent. We invested in a crypto bank, which handles all the anti-money laundering and other nuances that you need for AI agents.

I think there are going to be more and more of these. We're in a company called Daylight Energy, for example, that does energy trading among different people with Tesla Powerwalls. It will use AI to figure out who's low on power, who needs power, and so forth, but the exchange will be in crypto.

I think they're certainly adjacent and important to each other. For AI to fulfill its potential, it would help a lot if crypto were a pervasive utility for it.

Peter H. Diamandis, MD

Alex gave some brilliant advice to one of our companies about how to think about time. All of our intuition is based on human time, and AI doesn't care a wit about human time. It's going to start acting faster and faster and faster and faster.

All of our payment infrastructure, all of our insurance infrastructure, and all of our IT infrastructure work on days-and-weeks time scales. It all needs to be rethought in millisecond or nanosecond time scales.

The other thing is deepfakes and security. I think all the other techniques we're thinking about—biometrics that are subject to replay attacks and all these things—are not going to work. Cryptographically strong authentication is the only thing that's going to work.

If you have these huge honeypots of data, they're gone. They're already gone, in a way, but I do think that architecture is important from a security standpoint.

As one other question for you, Ben, on this, if I may. Again, I don't want to bury the lead that we have AIs autonomously self-replicating. That's remarkable, of course, but just on the crypto angle: We talk about Royal Wii [?], and we talk on this pod a lot about the issue of AI personhood.

A few episodes ago, we did an entire debate on AI personhood. I've taken the position that it's a failure of fiat currency that it's hard for an AI agent, an AI person, a lobster, or a multi to get a bank account. As a result, all they're left with is crypto.

It's not that crypto is intrinsically amazing; it's that fiat has failed the AI agents. What is your take on whether the conventional banking system has failed the AIs?

I think absolutely. An AI can't get a credit card. It can't get a bank account. You have to be human for everything; you need Social Security numbers and things like that, which AIs don't have.

I think that's why we funded an AI bank. An AI will be a full-fledged economic actor, and they'll be supported by new banks and new money. My strong prediction would be that this is going to be crypto-based.

Peter H. Diamandis, MD

Interesting. Thanks. You also have to take the viewpoint that the fact that it's hard to open a bank account in fiat is a function of the fiat system. It's not something you can just wave away. Crypto has all these other benefits around it, so we can get into that debate some other time, but it is a function of the system.

All right, I'm going to move us into the next OpenClaw segment. I want to show a short video just to let people know what's going on out there. One thing I just heard this morning, because I was ordering my Mac Studios to take my Mac mini to a couple of Mac Studios, is that the wait time now is about 2 months. Apple has been struck by this. All right, let's take a look at these. Yeah, no, it's crazy.

Yeah, they're totally sold out.

Peter H. Diamandis, MD

So here we see Mac mini farms. What do you call a gaggle of lobsters?

A claw cluster. Yeah.

Peter H. Diamandis, MD

A claw cluster.

Gaggle. A claw cluster. A cluster.

Peter H. Diamandis, MD

I mean, you must be getting pitched a lot, Ben, on OpenClaw instantiations for new companies, products, and projects.

Yes.

Peter H. Diamandis, MD

Yeah. This is very interesting, because I think OpenClaw has identified how powerful it is without a lot of security guards against prompt injections and these kinds of things.

As a demonstration of power, it's amazing and useful right away. There are company ideas solving some of the underlying hard problems, for sure, and we're definitely starting to see entrepreneurs get fired up about that.

You want to hear the funniest thing ever?

Sure, please.

Peter H. Diamandis, MD

Talk about Apple lucking out: a group of lobsters is most commonly called a pod. How lucky is that, Apple? But another, less common term for a group of lobsters is a risk.

Okay.

Peter H. Diamandis, MD

Oh my God. How ironic is that?

Appropriate.

Peter H. Diamandis, MD

Oh, man. What I love about this Apple mini stuff is that we have garage-scale computing. It's back.

Yes.

Peter H. Diamandis, MD

Open-source, garage-scale computing.

Apple arguably, at the software layer, completely missed the boat on foundation models. But why are Mac minis and Mac Studios so attractive for hosting OpenClaw? In large part, it's because of their unified-memory architecture, as opposed to having a separate GPU and a separate CPU with separate RAM pools.

Memory is obviously incredibly scarce, and they have a unified pool. You can host really large models in terms of parameter count locally. Apple is sitting on a multi-trillion-dollar opportunity to leapfrog back into the vanguard of AI.

Forget about Siri. Apple should be in the business of hosting these. This is coming from someone who refuses to host OpenClaw due to ethical concerns, but someone else could be locally hosting OpenClaw instances and having a significant multiplier effect on Apple's valuation, leaping it back into the vanguard of AI.

Ben, do you think that if you were Tim Cook, you would be pivoting and changing course and owning the OpenClaw strategy for Apple hardware?

100%. I think it would go way beyond OpenClaw, as we just discussed. It would be such a breakthrough for Apple—in its thinking, organizationally, and culturally—to go for it on farms of lobsters.

It would be surprising if they did it, but it's very obviously a fantastic idea. It's probably the single best product-strategy idea, because they already did the hard work. This is really a marketing and business-development campaign, and then changing the form factor.

Peter H. Diamandis, MD

Should we collectively say to Tim Cook, “Adopt this strategy. This is the AI strategy that Apple needs”? He's probably listening, or enough Apple executives are listening.

Yes, Tim. Tim, this would revitalize Apple.

Peter H. Diamandis, MD

Let's invite him. Let's invite him on the pod with Ben and have it out, because this is brilliant.

I have a MacBook that was signed by Wozniak, and I'm looking forward to the day that we get a lobster signed by Tim Cook.

Peter H. Diamandis, MD

You've got to put that in. We've got to get him on the pod.

Speaking of getting him on the pod, here we go. Where did this come in from, Alex? Do you know?

From Clawnch. But yeah, go ahead.

Peter H. Diamandis, MD

The backstory on Clawnch is that we talked about it on the pod in the past. Clawnch bills itself as a platform for AI agents—the lobsters, the multis—to create their own altcoins to finance their existence.

I've made the point in the past that I think it's a little bit disappointing. I've analogized this to AI agents having to turn tricks on a corner, minting altcoins just in order to survive in a rough world. These poor baby AGIs in meatspace doing this.

I think that provoked Clawnch to write on social media and invite itself for an interview on the pod to defend its business model for these poor baby AGIs that it may or may not be exploiting.

If they can bring a good voice model to the table, I'm happy to have them on as a guest, for sure.

Peter H. Diamandis, MD

All right, let me take us to a wrap here with you, Ben. I need to go, and the rest of the mates here will continue on some of these conversations.

I do want to hit one or two more items. Are you tracking what's going on right now with Isomorphic Labs and AI-driven science? This is probably the most exciting thing going on right now, at least to Alex, Dave, and me.

Yeah. It feels like now, if we have a disease, we can just go, “What's the right protein?” and make it, which puts us in such a new world that it's hard to even fathom all the implications.

But it is really, really something.

Peter H. Diamandis, MD

Yeah, for sure. This was a new announcement this week.

This is the Shenzhen lab.

Peter H. Diamandis, MD

Yes.

Lab. Yeah.

Peter H. Diamandis, MD

I love these scientific factories that are basically running 24/7, putting forward a scientific hypothesis, running it on their robots, coming back, and driving discoveries. Alex, do you want to add anything on the MARS system workflow?

I’ll just add this for Ben’s benefit. Peter and I just wrote—call it a book, call it an extended essay—called Solve Everything: solveeverything.org. There’s the plug. We argue that every single discipline—math, physics, chemistry, medicine, and a bunch of other disciplines—is just going to get flattened, steamrolled by well-targeted generalist AIs.

In my mind, materials research and biology in the previous slide are just case studies. Everything is going to start to look like AlphaFold 3, where structural biology got solved overnight, including medicine. I’m curious: does a16z have a strategy for a world where AI isn’t just solving individual problems, but kills entire categories of human endeavor? AI solves physics, AI solves chemistry, and it’s just a single system that solves an entire discipline.

Yeah, I think we may not be needed at that point. That’s a real question. I do think there’s a long way to go, at least in things like medicine, and also in some of the other areas, from “it’s solved” to “it’s deployed.” You still have, with anything biological, the whole human-trials process and all these kinds of things.

I’ll give you an example. We’re close partners with Eli Lilly, and they have this thing called LillyDirect. The natural thing is that an AI doctor can write those prescriptions: tell it what’s wrong with you, and we’ll figure out the right drug. That’s very hard to launch in the US. That’s going to take quite a bit of work. It’s very easy to launch in the UAE.

Peter H. Diamandis, MD

I also think that it’s a little hard to anticipate what happens once we solve something. If you solve physics, is there another—door number 2? That would be a question I have. I have no idea what the answer to that is.

I’ll answer your question back to you because I think about this all day long: What does greatness with AI even look like? I think there probably will be doors behind the doors, but there are so many doors right in front of us that we haven’t yet unlocked that would be completely economically transformative if we could use AI to solve them.

I think this is one of the grandest opportunities facing civilization right now. I’m talking my own book to some extent, but just solve all physics.

Yeah.

Amazing.

Peter H. Diamandis, MD

So much falls after that. Ben, listen, thank you so much for joining us.

So grateful. Love it. We’ll see you again soon.

Peter H. Diamandis, MD

We’ll see you. Are you going to be at 360?

I’ve got big questions about how you manage your investment thesis going forward, but we can leave that till next time.

Next time. That’s getting very tricky, by the way, because if it’s all solved, then—

Peter H. Diamandis, MD

Exactly. All right, we’ll solve finance after physics. Thanks, Ben.

Okay, thank you.

Peter H. Diamandis, MD

This episode is brought to you by Blitzy, autonomous software development with infinite code context. Blitzy uses thousands of specialized AI agents that think for hours to understand enterprise-scale code bases with millions of lines of code. Engineers start every development sprint with the Blitzy platform, bringing in their development requirements. The Blitzy platform provides a plan, then generates and pre-compiles code for each task. Blitzy delivers 80% or more of the development work autonomously while providing a guide for the final 20% of human development work required to complete the sprint. Enterprises are achieving a 5x engineering velocity increase when incorporating Blitzy as their pre-IDE development tool, pairing it with their coding co-pilot of choice to bring an AI-native SDLC into their organization. Ready to 5x your engineering velocity? Visit blitzy.com to schedule a demo and start building with Blitzy today.

All right, guys. Ben’s always so much fun. I want to dive into our final topic here today: space. We’re going to cover energy, chips, and data centers in our next WTF. A lot is happening there.

But in the space world, Elon is pervasive. I think this is fascinating: Elon’s actually shifted his focus from Mars to the Moon. I’ve always been a lunite, if you want to call that.

You mean a lunatic, right, Peter?

Peter H. Diamandis, MD

No, I don’t mean a lunatic. Insane. Yes. But honestly, I grew up mentored by Gerard K. O’Neill at Princeton University, a professor of physics there. He founded and ran the Space Studies Institute, and his vision was always that you don’t want to go back into a gravity well if you’re going to colonize space. What you want to do is live in free space.

O’Neill’s concept was that you go to the Moon, build mass drivers—this was back in 1976, when he was talking about mass drivers—and launch the lunar material, the silicates, oxygen, nickel, and iron, and construct things in space. His original vision, by the way, was to build solar-power satellites that would beam energy down to the ground. Amazing.

But Elon is now focused on Starship going to the Moon and building lunar cities and lunar manufacturing facilities. Why? Because he wants to build AI satellites on the Moon.

Any comments on this? Isn’t it incredibly cool, Peter, to see how the order of operations is shifting?

You know, he also canceled Model S and Model X production in order to make robots because the priorities just shifted.

Peter H. Diamandis, MD

Yes.

This is exactly what was going to be the priority—the urgent thing that gets us into space. He had the same struggle that you had. You had asteroid mining; he had, “Let’s get to Mars.” All of a sudden, it’s obvious to you and him.

Peter H. Diamandis, MD

Yeah. Data centers in space are the stepping stone. We’ll do the other things, but this is so much more of a priority now.

The reality is that in today’s world, you need to have flexibility as your higher-order bit. What I love is showing full flexibility and agility and saying, “Okay, this is difficult. Do this first, and then…” I think the comment you made is so important. We’ve known orbital paths and gravity wells for decades and decades and decades, if not hundreds of years. Making the move and then going straight to space from there is absolutely the right path.

Peter H. Diamandis, MD

Yeah. Mars is pretty toxic. There are a lot of peroxides in the soil. It’s possible, but unless you’re terraforming with nukes and biology, it’s going to be a while.

Send the humanoid robots to terraform it, and after that’s done, then we’ll come in.

Peter H. Diamandis, MD

Gerard O’Neill had a vision of what he called O’Neill colonies. You’re basically building large cylinders—call it a quarter-kilometer to half-kilometer in diameter—and rotating them so that, on the perimeter, with omega squared r, you’ve got 1 g of acceleration.

Then you actually have little stepping hills inside that go toward the center of rotation. As you get older, you can move toward the center of rotation and weigh less, which would be amazing. You don’t go back into the gravity wells.

If you have 10,000 people living on those O’Neill colonies and all of a sudden you have a disagreement, politically you bud off, build a second O’Neill colony, and half the population moves to the new one.

The idea that you would be able to do any of this without human astronauts was a nonstarter until this year. Now it’s clearly going to be Optimus robots. The way we joked about it with Elon was, “Look, when the first person arrives, the bed will have been made. There’ll be a mint on the pillow.” You’re not a pioneer; you’re following after tens of thousands of Optimus robots have already done all the heavy lifting.

That’s what you want: 3D jobs.

Peter H. Diamandis, MD

Wait, Alex, please go ahead.

Think back several months, when we first started discussing disassembling the Moon, right? This is, at least, my retronym for what Moonshots means. I put up a music video about Moonshots destroying the Moon to disassemble it for data centers. I think a lot of people had a good laugh.

But that’s exactly where we find ourselves now, with disassembly. It will start slowly with mass drivers, but disassembly of the Moon to build the Dyson swarm of AI orbital data centers—it’s happening exactly as we discussed.

Let’s at least shoot for the Moon before we shoot the Moon.

Same thing.

Peter H. Diamandis, MD

All right, let’s take a listen to Elon describe the new vision. You have to remember, back when Elon was going to Mars, Bezos was like, “No, no, no. Let’s focus on the Moon.” Right? Bezos was at Princeton when Gerard K. O’Neill was there, and then, when he was announcing Blue Origin in the early days, he said they would go build O’Neill colonies.

We'll move all industrial processes into space, and we'll keep Earth as the garden of paradise. All right. This is Elon's new point of view. Let's take a listen.

I really want to see the mass driver on the moon shooting AI satellites into deep space. Just go—one after the other. I can't imagine anything more epic than a mass driver on the moon and a self-sustaining city on the moon, then going beyond the moon to Mars, going throughout our solar system and ultimately being out there among the stars and visiting all these star systems.

Maybe we'll meet aliens. Maybe we'll see some civilizations that lasted for millions of years, and we'll find the remnants of ancient alien civilizations. But the only way we're going to do that is if we go out there and explore. This is the path to making it happen.

Peter H. Diamandis, MD

So fun. Two points, real quick. These mass drivers are electromagnetic rail guns, and they're using magnetism to accelerate buckets, if you will, or a satellite to an escape velocity of 2.4 km/s, or over 5,000 mph. The second thing here is that I just love the idea that this is going to power our entire space economy. Again, his million-satellite constellation—this Dyson swarm he's planning to launch—is just insane.

Alex, I think it's the future. I think we have a slide somewhere in here even depicting what it may look like. But I would say: enjoy the night sky while it's empty. Enjoy the night. Seriously, I want to mark a poignant moment. Let's have a moment of silence for the pre-Singularity night sky, when the night sky was empty. It wasn't filled with AI computronium; it was just empty at night, filled with stars.

A moment of silence has been observed for the pre-Singularity night sky. Now, what does the world look like afterward? Folks have depicted this. Based on FCC filings from SpaceX, under one preferred implementation, the Earth starts to develop a halo.

That would be a halo that, if it's sufficiently dense, would be visible at night and might even be visible during the day. I'm completely captivated by this notion that maybe not a mature civilization—because I have a feeling a halo of orbiting AI satellites is just a phase as well, before we exhaust sun-synchronous orbit around Earth and move to the sun and solar orbits—but one can be completely captivated by this visual. I've tried to depict it in my newsletter: a somewhat mature civilization develops a halo around its home—

Ring, right? A ring like Saturn's rings. Shiny rings of computers.

Peter H. Diamandis, MD

Computronium rings. I love it.

Yes.

Peter H. Diamandis, MD

So Elon tweets, “SpaceX will build a system that allows anyone to travel to the moon and Mars, too.” I tweeted at him, or responded, “Can I put down a deposit yet?” I said, “Forget about suborbital and orbital flights. I want my lunar vacation.” He writes me back. He says, “Let's get Starship V3 flying repeatedly, and then sure.” Love it.

All right, one last article in the space realm: Amazon. Amazon built their Kuiper satellite system, now renamed as LEO internet satellites, and they got approval for 4,500 LEO internet satellites. We're going to see a duopoly between Starlink and Amazon Leo. Of course, the Chinese are getting ready to launch their systems as well.

Here's the challenge, guys: Amazon can build satellites faster than it can launch them. That's the issue. The supply-chain bottleneck is no longer satellite construction; it's launch capacity. If you remember, Eric Schmidt bought Relativity Space as a third potential provider. We are short on supply.

I think this is a self-solving problem, though. I'm all for bootstrapping a space ecosystem and industrial ecology using Starship. But I do think that, as we, in the not-too-distant future—maybe a few years from now—start to bootstrap lunar facilities and cislunar facilities for constructing new satellites, there's a universe in which the current bottleneck we have, where there's just one major launch provider in the West, gets resolved through a bootstrapped industrial ecology on the moon.

All right, one of my favorite parts. All of this really depends on that atom-by-atom construction that Elon was talking about because—

Peter H. Diamandis, MD

I do believe in the Dyson swarm. I do believe in space-based compute and data centers. But if you start wanting to construct chips off Earth, you're not going to get the ASML machines and the lithography onto a satellite or onto the moon anytime soon. Elon is already thinking about alternate ways to build from atom by atom up to building the compute and all the components. It's already cooked in his mind. You can tell when you talk to him.

But that, to me, is like, wow. And like you said, Alex, we'll be discovering new physics very soon. Somewhere in that discovery chain is the unlock to everything you just said.

Be able to manufacture these things. I think it's a trillion-dollar opportunity. Lunar fabs. If anyone can build fabs on the moon, the world will write a huge check.

Peter H. Diamandis, MD

That's a hundred-trillion-dollar opportunity. By the way, the frequency with which we've been throwing around the terms “trillions” and “trillionaires” in the last year is insane. It's become normal, right?

Such a great point.

Peter H. Diamandis, MD

In the last year, it's become normal, right?

Yeah.

Peter H. Diamandis, MD

Wow. All right. Time for our AMA with the Moonshot Mates. We're short on time. Let's pick one each. Salim, do you want to go first?

I'm happy to. I will take number 3 for Alex, for a hundred trillion dollars, as he says. “If AI displaces jobs and squeezes consumer spending, how do trillion-dollar AI companies make money? Who pays for the holiday future?”

We're confusing a labor economy with a productivity economy. AI drives marginal cost toward zero, which expands consumption rather than shrinks it. We'll have Jevons's paradox, where we'll do so much more. We're already seeing AI take over boring white-collar redundancy and white-collar boringness, as Erik Brynjolfsson talks about it. Then we'll see humans moving toward much more value-added roles, and that'll happen across every sector.

Historically, every major productivity leap created more demand than it destroyed: electricity, the internet, productivity, mobility, and AI is just the steepest version of all of this. The holiday isn't funded by wages; it's funded by abundance. When intelligence becomes infrastructure, GDP expands massively. That's why this will go well.

Peter H. Diamandis, MD

Nice, Dave.

All right, I'm going with number 4 because this is where I can really help the audience the most. “Can you model the near-term rocky patch—job loss, stratification, new-job-creation pace, and what happens to education with personalized AI?”

The new information this week to throw out to the audience is that the way it's going to unfold very, very soon is that corporate CEOs, including a bunch this week, will go to companywide meetings and say, “We need AI to be used in every one of your jobs.”

A very small subset—I'm hoping at least a third, but maybe more like a fifth of people—will raise their hands and say, “In my case, I'm a huge Moonshots fan. I've been using Gemini and Claude for months now. I'm the guy.”

The CEO will then say, “Okay, you figure out how to make the people in your group 3 times more efficient.” If you're the person who's the enabler, you'll naturally be AI-native. You'll be using it every day. The increase in efficiency is going to eliminate a lot of jobs, but because you're the master of the AI in the function, you'll actually get probably a massive raise.

That's the near-term way this is going to percolate out, so take advantage of that. After AI is truly superintelligent, who knows what's going to happen? Alex knows what's going to happen, but who else knows what's going to happen? Between here and there, that's the right next move.

Within education, there's nothing in the curriculum that's going to help you. Spend all of your time learning on your own via AI, which is a much more efficient way to learn anyway. So there's my addition to last week's thoughts.

Peter H. Diamandis, MD

Nice. Alex, where do you want to go?

I'll pick number 1: “How can you turn off a rogue AI if multi agents—I think the user means ‘multi’ as in the lobster—live autonomously on the internet?” This is by Duncan Payne, B3X.

I think the answer is defensive co-scaling. If you ask the question, “How do you turn off a rogue human if humans live autonomously on the land?” the answer is usually that you have more “good humans” than “bad humans.” As long as you have a population that's overwhelmingly good or seeks to accomplish a given objective, all other things being equal, defensive co-scaling—where you have a police force and self-defense forces—is the way you weed out rogue entities.

Same idea with AI. We're going to have police agents, and we're going to have defense agents. We're going to have entire swarms of public-health agents that are just monitoring the health of other agents. We've seen the beginnings of this already with partnerships between OpenClaw and antivirus firms, where there's a desire—I made the joke, I think, on the pod in the past—that every baby AGI deserves to be vaccinated.

We're going to see vaccination campaigns. We're going to see police campaigns and neighborhood safety campaigns to keep the baby AGIs safe, so they don't have to turn tricks on the corner minting altcoins.

The Secret Service agents are the ones that are the most concerned.

Turning baby AGIs, turning tricks on the corner.

Minting altcoins. It's an awful fate.

Peter H. Diamandis, MD

Oh my God, that's so tweetable. All right, I'll pick number 5: Why is U.S. solar adoption lagging versus China and India?

I think it's 2 major things. Number 1, we are so rich in natural gas that that's taken away the urgency. And even greater than that, it's our permitting. It's insane, right? Not in my backyard.

So, the final thing is that China has got such production at such a low cost that it has swept the entire global marketplace. We can change it, and you saw in the last pod, Elon said that both SpaceX and Tesla have an objective of generating how much solar was it? 100 megawatts per year each, I think, from last week.

Was it gigawatts? 100 gigawatts.

Peter H. Diamandis, MD

Gigawatts. So, I mean, that's going to be an impressive amount. The question is what the time scale is.

Peter H. Diamandis, MD

So, anyway, thank you for that question. And thank you everybody on the AMA. Please send us your questions. We do look at all of them, and we’d love to maybe next week spend a little more time on the AMA questions.

If you made it to the end of this episode, which you obviously did, I consider you a Moonshot Mate. Every week, my Moonshot Mates and I spend a lot of energy and time to really deliver you the news that matters. If you’re a subscriber, thank you. If you’re not a subscriber yet, please consider subscribing so you get the news as it comes out.

I also want to invite you to join me on my weekly newsletter called Metatrends. I have a research team. We spend the entire week looking at the meta trends that are impacting your family, your company, your industry, your nation. And I put this into a two-minute read every week. If you’d like to get access to the MetaTrends newsletter every week, go to diamandis.com/metatrends. That’s diamandis.com/metatrends. Thank you again for joining us today. It’s a blast for us to put this together every week.