AI现状：人形机器人、AI版权战争与中国影响力日益扩大｜与 Salim Ismail 对谈 #157

In our special episode of “WTF Just Happened in Tech This Week” with my dear friend and co-conspirator in everything exponential, Salim Ismail. Salim, you're back in Miami.

I'm back in New York and recovering from that insane event that you put together.

Yeah, it was great. We had year 13 of the Abundance Summit. I truly feel it was one of the best years ever in terms of being sold out, with 500 CEOs in the room and a few thousand people on our livestream through our Exponential Mastery Program. It's a difficult thing to say, but I think this is the best conference of the year anywhere.

Thank you. The bar is very high for that.

I appreciate that. I don't think of it as a conference. It really is a community that's trying to understand how fast things are going, how to uplevel their work, and how to go from success to significance. The world doesn't stand still. A lot is happening this past week, and I'd love to cover some of the recent news across AI, robotics, crypto, and everything else, and then weave in some of the things we saw during the summit, because there were a lot of incredible leaders and conversations. You up for it?

Totally up for it. A lot to go, a lot to cover for sure.

All right, let me hit on the first item here. This is a conversation I had with Brett Adcock on stage at the Abundance Summit, but it's also something he just texted me this morning and said, “Hey, check this out.” I did. This is their new factory.

Figure AI makes the Figure robot. They've just been producing Figure 02, and they've got Figure 03 ready for announcement very soon. What I find fascinating is that they built a facility able to produce 12,000 of their robots per year, right now scaling up to 100,000 per year. Pretty extraordinary. What do you think about it?

Fantastic. It would be very poetic to have Figure robots making Figure robots. Self-recursive—what could possibly go wrong?

In all honesty, I think Brett has been screaming in terms of how rapidly he's iterating. He went from starting the company to delivering robots in 31 months—just over 2 years. What he was saying on stage was that they're iterating a new design, a new Figure, every 12 to 18 months, which is insane. That seems to be about right in terms of the pace at which the underlying technology is moving.

I have a major announcement to make around this. In previous episodes, I've been very leery of humanoid robots.

Right. I know you've always said, “Why don't they have 6 arms?” and things like that.

I've also said it's going to take a lot longer to bring them into play than we think. I need to eat crow and revise my assessments, because what we're seeing now is off-the-charts incredible. They're doing incredible stuff, and we can talk about some of what they're able to do. But if you have a robot learning a task, then instantly 5 million other robots know how to do that task. That hive learning makes them unbeatable over time.

Agreed. What Brett said on stage is that if he had 100,000 robots today, he has customers for them. He also reconfirmed his expectation on pricing: in volume, they're going to go from where they are now—probably sub-$100,000—to sub-$20,000. Again, that blows me away.

They're in BMW, and they just signed a contract with a large logistics company. But what's really exciting is that they're going to be heading into the home very shortly.

That's super cool. I'd like to have 6 arms, by the way. If I have a request for Brett, just leave some slots for options, like on a car, where you can add extra arms.

Okay.

That was an announcement this morning on X. We'll be putting out the podcast that we recorded at the Abundance Summit with Brett Adcock on Figure very shortly, right after this one.

Let's go to our next story. This is not a surprise. Sam Altman is basically saying DeepSeek is state-controlled, and he's calling for a ban on Chinese models. Number 1, do you think it's true? Number 2, is this OpenAI using the U.S. government as a competitive blocker? How do you think about it?

I think both are very true. There's no way I believe the Chinese government isn't using DeepSeek in that way, just as it's very clear it was using TikTok in that way.

It's also a little rich for Sam to be calling out intellectual-property violations when most of the data he trained OpenAI on was sucked out of places where they shouldn't have gotten it. But it's defensive in that sense. There's no question about the Chinese side, in my opinion.

We had Alvin Wang Graylin on stage during our Patron Day at the summit, and I have huge respect for him. He was very senior at HTC, which creates VR headsets, and he's written a number of successful books. His point was, “Rather than competing with China, why don't we collaborate?”

I would love to see a world in which we could collaborate, because the greatest foe that we have is not China. It is the rogue state. It is the individual who uses advanced AI to try to do harm to others. I believe that.

The question is, how do we collaborate with China, Europe, other parts of the world, and the Middle East to make sure that AI is being used to maintain safety, and that we're actually looking to find those rogue actors who are insane or dystopian in their nature?

A hundred years ago, we had the Bretton Woods Conference. After World War II, they worked out and laid down the groundwork for the United Nations, the IMF, the World Bank, and so on. We've had the most peaceful hundred years in the history of humanity.

In the 1970s, we had the Asilomar guidelines to work out biotech safety, and we've not had a major accident in biotech. It might be a great idea to revisit and combine those 2 efforts, and have a global convening of countries dealing with AI so that we have a clear set of principles for how we're going to navigate this. Let's be open about everything, because this affects all of humanity. If one area goes rogue, it will affect the whole world. This really does behoove us to operate in that way.

It would be amazing if we could make something like that happen.

On the point you just made, this comes out in TechCrunch: “OpenAI and Google ask the U.S. government to let them train AI on copyrighted content.” Here you go: throw copyright out the window. In order to win the race with China, we need access to all the content, whether it's copyrighted or not.

I have 2 views on this. Yes, it makes sense.

You're saying that the government should let them use all copyrighted content?

Yes, and I have a reason why. Copyright has been misused for the last 50 to 80 years. It keeps getting extended by Disney to protect Mickey Mouse. It's called the Mickey Mouse Act, honestly, because they've been extending copyright to protect the Disney characters forever.

Copyright is grotesquely twisted from its original intent. It was supposed to be something like 15 or 20 years, and then it should become public domain. Now it's an 80-year cycle or something absurd like that. It's been badly twisted out of its original proportion to protect corporate actors.

That should be one level of it. On the other hand, it's very convenient for Google and OpenAI to want this. The big challenge is that we're running out of data. As we run out of data, we need synthetic data or access to the deep web, copyrighted content, or something.

I think they need to figure out how to compensate the content owners in some way, shape, or form. I don't know what the obvious model is.

There is a solution, and Bill Gross spoke about it on Tuesday morning. I think you were out for that presentation, but I saw the highlights of it. I thought his constructs were really good because they align incentives.

Let me mention what it is. Full disclosure: my venture fund is an investor in it. It's called ProRata.ai. What Bill has done is create a model in which he gets a number of content creators. I've got a number of books, and I've signed up for this. He's spoken to a huge number of content creators.

ProRata.ai sucks in all of the data, and then when it looks at a large language model's results, it says, “The results are 38% from Time magazine, 24% from Bloomberg, and 0.00001% from Peter Diamandis.” Then it apportions the money along those lines.

It's vastly difficult but very possible. We have to remember that Bill Gross was the guy who came up with AdWords. Overture effectively became AdWords—it was Overture originally—and the model that Google used to become a trillion-dollar company.

I think that's directionally exactly the right way to go. There's some bun fight over how much of the percentage should go to the content creators, but if you structure something like that, I think it makes it reasonably fair. What everybody's looking for is a reasonably fair playing field.

That would be a nice thing. Ultimately, we've got these massive companies pouring money into AI. Do you remember what figure we heard during the Abundance Summit? How much is it—$1 billion a day?

A billion dollars a day is going into AI. It's crazy. It's insane. Honestly, $1 billion a day—I would have never imagined it.

Do the math.

Hundreds of billions of dollars per year, and I think we'll be reaching close to $1 trillion a year soon enough.

Anyway, here's the next article. The Abundance Summit this year had a whole day, or a significant portion of a day, on longevity. I didn't see this article otherwise I would have discussed it during that day, but I found it fascinating.

At Stanford Medical School, using AI, the faculty identified a peptide that produces the same type of weight loss as Ozempic without the side effects. This is what I've been saying: we're in the middle of a health-span revolution, not because we have anything other than AI helping us understand how the body works.

The umbrella here is that AI is going to spot things in ways that we could never see as human beings. This is one small example. We had another dozen examples at the summit talking through this. It's incredible what we should see coming out of these new models that are very low-cost and have ridiculous outcomes and benefits.

We're going to see this in treating cancers, treating all diseases—everything.

I have a colleague, Michael Jansson, whose friend's son has these weird seizures that nobody could figure out. They loaded the blood test into Grok, and Grok said, “You need to get that type of test. You'll figure it out.” It's unreal.

I don't want to disclose the details of it, but one of our faculty members on stage this year, who was part of our AI content, had his house burn down during the fire. It forced him to go to Palm Springs or Palm Desert, where he was living.

He had had medical problems. He had been tested left, right, and center, and nobody knew what it was. Maybe it was stress. He went to Palm Desert and found a new doctor there who said, “You have to look at this.” It turned out to be multiple myeloma.

He then took all of the medical data he had from 3 months earlier, before his house burned down, and put it into one of the models. I think he was using Claude 3.7 from Anthropic. It instantly pegged it: “You have multiple myeloma.” That was after 3 months of doctors getting it wrong.

I saw that piece. It was mind-boggling. As we apply AI to these things and it spots things that human beings can't see, we should expect to see huge improvements in the early diagnosis of all sorts of things.

When you go to the doctor today, you get the wrong diagnosis about 30% of the time. What hope is there for doctors to keep up with all the treatments, conditions, drugs, and so on that are coming out?

I remember Daniel Kraft saying that every day there are several hundred cancer research papers published. If you're a cancer doctor, you can't read several hundred of them a day. You need an AI to read them for you and say, “These are the 5 patients that these apply to. Go read those 5 papers.”

If you or someone in your family is ill, the very first thing you should do is ask for your data, put it into the models, and get additional second and third diagnoses.

On that point, I came across a startup that is aiming to release a completely free, fully fledged AI doctor by the end of this year.

That's going to be nuts. It's going to happen.

I'm a huge believer in and supporter of Fountain Life, and I'm the chairman there. One of the things we do that I love is load you up with the most exhaustive data collection ever. It's 200 gigabytes of data: your full genome, full-body MRI, and everything knowable about you.

All of that data becomes available in your app on your phone, with an AI system—I think we're using Claude 3.7 as well—so that you can query your data. One of the cool things I love is that you can take a photo of the menu when you're at a restaurant and say, “Given my genetics and my current blood biomarkers, what on this menu should I eat or not eat?”

Let's go to the next item. These are 2 articles back to back, and they're so important for us to discuss. Both of us have kids. Your son is 13, and my 2 boys are 13.

The first article is “Estonia to roll out ChatGPT for education for all secondary schools.” The second, back to back, is “Beijing to roll out AI courses for kids to boost sector growth in primary and secondary schools.”

We had a teen program at the Abundance Summit this year for the first time. The teens were amazing. We had a dozen of them there, and they were brilliant. I did that late-night Meaning of Life session, and they were sitting there nailing me with questions.

“You've got that wrong about the Meaning of Life. It should be this way.” I was like, “You're still wet behind the ears. Come on.” But they're incredible. It's unbelievable to see the depth of wisdom coming from this generation.

Honestly, we should just take all of our global politics and hand them over to 25-year-olds, 20-year-olds, or teenagers. They'll figure it out.

We just shouldn't give them control of the football.

When I was having my session with the teens, I asked, “There are 12 of you here—3 or 4 girls and 8 boys. How many of you are asked to use AI in the classroom, and how many of you are told not to use it?”

All of them were being told not to use AI in high school. Here I see Beijing and Estonia saying, “No, we're going to make AI a part of primary education.” They're requiring primary and secondary students to enroll in at least 8 hours of AI during the year. I assume that's per week.

I think it's insane that our kids are not being properly schooled in how to use AI to uplevel their education.

The dominant conversation in education today is an immune-system response from all the legacy actors—teachers' unions and so on—saying we should ban AI because you can't learn properly.

I think what's going to happen faster than anybody realizes is a reckoning. If you're not using AI in everything to do with education, you're going to be left behind very fast. That's going to break the immune-system problem we've had in education for a very long time.

God help you if you're trying to update academia. It's a vicious, vicious environment.

Tomorrow I'm going to spend the entire day at the new school we moved our kids to, Qualia School of Deeper Learning. I'll be working with the kids in the classrooms, meeting with the teachers, and then meeting with them again on Friday.

It's really about how you bring AI into the classroom. My realization is that if our kids are in middle school and they're solving typical middle-school problems, giving them AI makes those problems trivial. Writing becomes trivial, and math problems become trivial.

But what if you gave those middle-school kids extremely difficult problems? Come up with a brand-new socioeconomic system, or create a program that does this, that, and the other thing. Then you say, “You normally don't think you can solve this on your own, but here's a bunch of AI tools. Go out there and complete it.”

It allows us to uplevel the challenges that our kids feel confident pursuing.

I did an interesting experiment. I did my degree in theoretical physics, and in one course on relativity they had us rebuild Einstein's theories from the bottom up and show the mathematical formalism behind them to understand them better. It took the whole year to figure it out.

I went back and looked at it again, and I redid it with an AI. I asked, “Why is this relevant? Why is this relevant?” If I had to redo it again, I figured I could learn the whole thing in about 3 days rather than a year.

That's the potential you're pointing to here. It's unbelievable how quickly we could accelerate our kids.

Agreed.

The big challenge is on the regulatory side, getting that out of the way and changing it. God help us—we need to do that yesterday.

The question becomes, how do the teachers' unions feel confident embracing this and making their jobs more fulfilling?

I've talked to teachers quite extensively. Individual teachers generally tend to get it, but at a group level the unions freak out.

When you say to them, “The role of the teacher should go back to the way it used to be, rather than giving endless lessons repeatedly. Be the guide on the side, help mentor the kids, and spend time with them on the difficult problems they're facing rather than just giving the same lesson every time,” they can't get their heads around it.

It's that whole flipped-classroom approach, where you watch the lesson in the evening and do the homework collaboratively in class—which is what real life is like. They can't cope with that. It's a structural problem.

It's the same way doctors have a difficult time having AI in the diagnostic chain. It's going to run rings around them so quickly that I think they'll be forced to adapt. It will be one of those situations where you either adapt or get thrown by the wayside as we move very quickly forward, because the pace of change is not slowing down.

I think about when Google came out and library attendance probably fell through the floor.

Over time, librarians came to accept it. Library science, by the way, is one of the most important fields in the world today, because the retrieval of knowledge is really, really powerful as a discipline.

Here's our next article from Wired: “Google's Gemini Robotics AI model reaches into the physical world.”

I love this because if you think about robotics, there are 2 sides of the equation. There's the physical robot and building that physical robot with all the parts, and then there's the AI model that controls it.

Imagine building an AI software layer that enables you to interface with any physical system or any robotic system. This is huge, because the real Holy Grail is movement in the physical world.

I remember one of our Singularity University faculty members saying that the only reason you need a brain is to move around in the physical world. At the end of the day, most of these AI models are doing cool stuff—writing code and so on—but at some point that code reaches the physical world through finance, health care, or whatever.

Once you put that layer in place, allowing AI to interface with robotics in the physical world in a generalized way, you have incredible potential. In the Exponential Organizations book, we'd call it interfaces. You can put so much functionality into that layer, and the end use becomes much more trivial to program. This is massive.

Agreed. There are a few theories that say AI will become conscious when it is materialized in the physical world.

Yes. In fact, it turns out that that's the predominant theory of consciousness: understanding and adapting to the physical environment. Consciousness allows us to have a faster feedback loop, because we have cognitive processes that can be brought to bear.

There's one thing I really loved about this particular article: at the bottom, it says they have no plans for commerce. I really love some of these folks, like Meta open-sourcing their large language models and Google saying, “We're doing this because we just think it's important for the world.”

It's so great when they do things like this. It uplifts everybody, and everybody wins. If this kind of thinking could reach our politics, we'd be in really good shape.

Google was our major sponsor at Abundance 2025, underwriting the Google Tech Hub through Google Cloud. There was a lot of super-cool stuff, and I'll get into that a little later.

I found this next article fascinating. Ilya Sutskever has created a startup focused on superintelligence. In talks at the beginning, he talks about a funding round to raise $2 billion at a $30 billion valuation.

Why did they choose to use the term “superintelligence”? The reason is that superintelligence is meant to convey something that's not just AGI. With AGI, we say you have something kind of like a person, kind of like a coworker. Superintelligence is meant to convey something far more capable than that.

When you have such a capability, can we even imagine what it will be? Without question, it's going to be unbelievably powerful. It could be used to solve incomprehensibly hard problems if it is used well.

If we navigate the challenges that superintelligence poses, we could radically improve the quality of life. But the power of superintelligence is so vast.

The theme of the Abundance 360 Summit next year is “Superintelligence Unleashed.” I have to get Ilya to be our opening speaker. That will be my goal.

Can I get on my favorite hobbyhorse? What the hell do we mean by superintelligence? We have no idea what artificial general intelligence is. Can we please spend some time defining these terms before we start throwing them out like bathwater?

That would be my first question for Ilya: “Please define it.” We have emotional intelligence, spatial intelligence, visual intelligence, and spiritual intelligence. Anybody who's a leader is bringing a lot of that to bear when making decisions and choices.

What do we mean when we mean superintelligence? I've seen so many different definitions of AGI that I'm really struggling with this. What exactly are we talking about here?

Having said all that, fantastic. All power to him. I love the fact that he's making safety a key part of this. I don't know how he's going to do it, because if you have a superintelligence, we're kind of toast anyway if it's going down that path. But go, go, and it's amazing that you can get a startup valuation like that.

It was about 13 years ago that I had my 2 boys, and I remember making a decision to double down on my health. Without question, I wanted to see their kids and grandkids. During this extraordinary time, when the space frontier, AI, and crypto are all exploding, it's the most exciting time ever to be alive.

We've seen a number of companies that went from literally their first outside financing round to valuations in the multiple billions.

I was talking to somebody about this, and they were saying, “That's just a ridiculous valuation. I'm struggling to raise money for my startup at this valuation.” I said, “Would you rather be him or you?” You'd rather be there.

So here's the question. AGI has a very fuzzy definition, but it sounds like an AI that can do as well as the best human in every possible field. An AI that is the best physician, the best mathematician, the best cook, and the best whatever—that's the blurry line we call AGI today.

Digital superintelligence, in my definition, is just following the law of accelerating returns. Ten doublings later, you're 1,000 times better. That's superintelligent.

What would it do? I think there are 2 areas where I would like to get one. The first is figuring out deep problems, like the wave-particle duality of light, and solving some of the grand unified theories in physics that nobody has been able to crack yet.

The second would be figuring out the right model for civilization to follow, and finding a way to get the entirety of humanity to follow that governance model around all the political problems we have going on today.

Those would be 2 areas I would assign to an artificial superintelligence.

I think that's the point: solve problems that have been unsolvable by humanity to date. Whatever that might be—double the human lifespan, unify field theories, understand what dark matter is, or get us through our next dilemmas.

How can I reach into a parallel universe and go down that path instead of this path? That's what I would ask.

Sure. I think that sounds like a perfect definition: reaching into parallel universes. I'll make sure we put that one forward.

The next item is an OpenAI employee on the future of the web. Here's the quote: “It's 2025, and most content is still written for humans instead of LLMs. Ninety-nine-point-nine percent of web attention is about to be LLM attention, not human attention.”

The point is that we've been building the web for us meat sacks, as carbon-based life forms. Going forward, the value of the web is going to be to feed the large language models. We have to restructure how we write it.

Fascinating thought. Absolutely fascinating.

I think there's a huge layer to be built here. I put this in the category of AI as electricity. It just becomes this invisible layer that's part of everything.

A human being can't go through many pages on the web anyway. You want to pass this whole thing over to an AI, let it navigate the web, and give you the most relevant pieces for you. I think it's super exciting to see if we can make this happen.

The point is an important one. I'm surprised we haven't heard it much before, but it's probably very true.

Let's go into some of the conversations that came out of the Abundance Summit. Cathie Wood at ARK Invest was giving us her 2025 report. She's a huge believer in all things exponential, particularly humanoid robots and robotaxis.

She made the point that Tesla's Cybercab should be up and operating by this summer in Texas, soon thereafter in Los Angeles, and hopefully nationwide.

We had a great session on investing in AI with Anjney Midha, who heads AI investing at Andreessen Horowitz, and Dave Blundin, my partner in Exponential Ventures. Raoul Pal interviewed them.

One of the things I found fascinating was Dave talking about 2 key points. It turns out that MIT is the number-one AI school on the planet. The Computer Science and Artificial Intelligence Laboratory has more graduates and more compute than any other location.

It's producing more unicorn startups than any other university. Number 1 is MIT, number 2 is USC—which shocked everybody—and number 3 is Stanford.

The other thing he mentioned was the average age of an entrepreneur starting an AI company, particularly out of MIT. The bell curve used to peak in the 30s. Now it's moved to the early 20s, from 20 through 23.

That's amazing. I saw the same thing in the Bitcoin and blockchain world. All the founders are in their early 20s. If you're over 25, there's some bylaw that says you have to be under 25 to work on blockchain.

I think we're seeing the same thing. The paradigm shifts are so big that anybody over a certain age can't get their heads around them. They're not native to them.

The other part is that around 75% of incoming MIT freshmen want to start a company before they graduate. Now the ability to find a great partner, come up with an idea, test that idea, and launch it inside a year or 2 is extraordinary.

That's crazy.

Out of Exponential Ventures and MIT, 2 companies went from $0 to $2 billion valuations in 2 years.

That's insane. Incredible. I think we're going to see a lot of that as we decimate the real world, or the old world, with AI. The creative destruction is going to be incredible.

The other point they made was that if you're facing the big challenge of when to invest in AI, you really need to get into it. That's the only way you're going to keep track of what's going on. I thought that was a really important point that we should all take account of.

You need to get into it quickly.

When we advise companies, the big question is always when to tap into a technology. The minute it's mission-critical, you should be in it right away so that you're learning and not being left behind.

Your mind goes where your money goes.

Great point.

Max Hodak came on stage. I love Max. He's the CEO of Science Corporation, and he's building what I consider the most extraordinary BCI company out there.

He was the co-founder and president of Neuralink with Elon Musk. He left and started Science, and he's got something called a biohybrid neural interface, which I love.

Let me take a second to dissect this—literally. When you put a Neuralink device into your neocortex, you're putting 1,000 to 2,000 very fine filaments into the neocortex so those filaments can read and write onto neurons.

But even as thin as they are, those filaments destroy millions of neurons when you poke them into the brain. That's a problem. There's also a limited number of them.

Do you remember the bit rate at which the brain communicates with the outside world? It's something like 10 or 20—maybe 10 to 40 bits per second. Compared with gigabits and petabytes, that's very small.

What Max has built, and demonstrated in animal models, is extraordinary. They'll be going into primates later this year and, hopefully, humans a couple of years after that.

Imagine having physical computer circuitry with millions of microwells. In those wells are nutrients in which you place neural stem cells. The neural stem cells interface with the electronics on this piece of material.

The neural stem cells and those microwells are placed on the surface of the brain. The neural stem cells grow their axons and dendrites into the brain, like roots into soil. They don't kill any neurons, but you have millions of them growing into the brain, connecting, and forming appropriate neural connections.

It's a brand-new strategy and approach to BCI.

I find this fascinating. I've been saying for a long time that one of the challenges is that we don't understand how the brain works.

There are 100 billion cells, all true neurons, operating, and each of them has hundreds or thousands of connections to the nearest neurons. The numbers become totally insane.

But you don't have to understand it as long as you can interface effectively with it. That becomes really powerful.

Can I tell you a fun little story?

Of course.

Back in 2010, Loïc Le Meur used to run an event called LeWeb, a big web conference in Europe. He said to me, “Salim, come and give a talk via your learnings at Singularity University on whatever you want.”

I thought, “Neuroscience.” So I interviewed all of our neuroscience people at Singularity University—Chris deCharms and a bunch of others—and tried to get a common definition of the brain.

The best common agreed definition I could get was something like, “a fractal, chaotic information-processing system that may or may not generate consciousness.”

I thought, “Really? That's what you got for me?” It really blew my mind. They said, “We have no idea why any of this stuff works. We still have no idea why we sleep after all these decades.”

Interfacing with the brain, as Max is figuring out how to do, will give us unbelievable potential. The brain will figure out how to optimize the connections and use the computing capabilities. It will figure it out. We don't have to do that for it. Now things become really interesting.

One last point on this: I remember one of Ray Kurzweil's predictions was that we would have a high-bandwidth brain-computer interface by 2033.

I was always saying, “Ray, you've been right about so many things, but on this one I don't believe you. I don't see the technology there. We've got 8 bits per second, 30 bits per second. How do I get high-bandwidth BCI?”

Then I met Max, and suddenly it made sense.

It's one of those moments when you just shake your fist at the presence of Ray Kurzweil. I don't think he's a human being. I think he's an avatar from a different dimension who comes in to poke us and jiggle us so that we stay on the right path.

Next up, we had Joshua Xu from HeyGen, which is an extraordinary company creating high-fidelity avatars and visual and voice models. They were in our Google Tech Hub, and they gave everybody there a chance to create an image of themselves.

You go into the studio for 2 minutes, the data is collected, and all of a sudden you have this visual representation of yourself. Do you remember how many languages your HeyGen avatar could speak?

It was a couple hundred languages instantly.

No, it was 17 languages. That's crazy.

I'm actually really looking forward to trying to create an AI twin. I've waited for a while to try to create one, but I feel that now is the right time.

Having an AI version of me that can remember everything I've ever said and analyze everything the way I've done it over decades is going to be so much smarter than the real me. You just want to let that thing go do presentations at conferences. It's going to be amazing.

I've been waiting until the right time to do it, and I'm hoping this is now.

It is. About a year ago, with Steve Brown, who's been my chief AI officer supporting me at PHD Ventures, we created PeterBot. We uploaded all of my content.

You did a whole podcast episode with PeterBot, and I was so impressed by it. It was much more eloquent, remembered everything, and was able to make many more connections than I could. That was fun.

This year at the Abundance Summit, we created a version of me at 120 years old. I'm 63 now, so it was 57 years into the future. I interviewed this version of Peter about how we get through the singularity, what you do to reach longevity escape velocity, and whether crypto is still around.

I asked, “What is crypto 57 years from now?” It said that Bitcoin was running at $5 million in 2025 dollars.

I asked, “Is the dollar still a thing?” It said, “It's kind of a novelty, but we've definitely gone to a crypto-default currency across the world.”

Can I make a quick point about that?

Of course.

Everybody over 25 has grown up with the only major currency we've ever known: the U.S. dollar. Everything is expressed in terms of how many dollars it costs to do something. That's been the unit of account for the last 5 generations.

But if you go hang out in NFT Discord channels with all these young people trading NFTs, you never hear the words “U.S. dollar.” You only hear Ethereum, and now Bitcoin in the Ordinals world.

This is a generation growing up where the U.S. dollar is not the dominant unit of account. It's going to be very easy for them to flip over to a different model.

Let's go to the Mo Gawdat portion of it all. Before I do that, we released the new Abundance logo—a gold caret. What do you think of it?

I like the rocket-ship flavor to it.

I thought that was great. What I told everybody was that the caret in the phrase you and I put forward in the very first year of Singularity University—10 to the 9th plus—that upward caret is the exponent portion.

Excellent. I love it. Beautiful.

I love our new logo.

Mo Gawdat came on stage. I love Mo. We're producing a documentary called “Scary Smart” together. Kristen, Mo, and I co-funded it, and Atlantic Productions is producing it. Mo is interviewing people around the world about AI.

He got very emotional on stage. We've done some podcasts with him on the Moonshots channel, and he's very concerned about dystopia on the way to abundance. He thinks we'll eventually reach an extraordinary world of abundance, but for the next 5 to 12 years we'll have a period of dystopian concerns that we need to be forewarned about.

What do you think about that?

I think he's right. For those who haven't listened to it, the podcast that Peter and I did with Mo a few weeks ago was off the hook. The amount of wisdom he dropped was incredible.

I think that's exactly right. I would frame it as, “It's always darkest before the dawn.” You always have a period of huge chaos before you get to a period of stability. We've seen that in World War I, World War II, and so on.

Unfortunately, we're coming into that period now, where there's extraordinary volatility and anything could happen.

One geopolitical analyst I pay a lot of attention to said that with the U.S. pulling away, every country in the world will try to acquire nuclear weapons as a defensive stance. That's a bad outcome. With so many nuclear weapons in too many hands, bad things could happen as things go down that way.

The only good news is that warfare is now being fought largely with drones and not by human beings, so that's something good at that level. But it's going to be a scary decade or 2.

One of the things that feels very real to me is that we're currently holding 2 futures in superposition, from a Schrödinger's cat point of view.

One future is one in which AI has dystopian qualities and the potential to threaten us. The other is one in which it's one of the greatest stabilizing forces that we have in this digital superintelligence.

I call it “Star Trek versus Mad Max.” A lot of that future is determined by what we believe and how we act now.

Can I throw out 2 ideas here?

Of course.

People often ask me why I'm so positive and optimistic about the world. I think there are 2 points.

One is from Ray Kurzweil, who said that technology is a major driver of progress in the world, and frankly it might be the only major driver of progress we've ever seen. Now that we have a dozen technologies—and I think one of the most profound things we saw at the Abundance Summit this year was the unbelievable real convergence of all these technologies into new breakthrough applications and domains being created—that gives me tremendous hope.

The second is that as anything becomes more intelligent, it tends to become wiser and more benevolent. We've never seen the alternative. That gives you encouragement that as AI evolves, it should become wiser and smarter and say, “These human beings are nutty. Let's figure out ways of protecting them from themselves while we move them along to their next level.”

It's the same way we would protect a natural species that doesn't know it's killing itself off. We would protect its environment for it. Those are the kinds of things that encourage me as we hurtle into this very volatile couple of decades.

We had a bunch of other fascinating speakers. We had the team from Google's Alphabet company, Wing, which does drone deliveries. We actually had Wing delivering things to everybody from the sky. It was amazing.

I have 2 thoughts about that. First, this is standard in China today. You can order coffees and have them delivered by drone. We need to get our regulatory act in order very, very fast. We're so far behind on that aspect of it.

The sheer granularity of the capability and the speed are unbelievable. I'm so excited, because this will totally change life for the better. It will be especially valuable in emergency situations, when you need to get a defibrillator somewhere because somebody is having a heart attack.

The potential here is unbelievable.

I love it. It's in Dallas right now, and I can't wait for it to become available. The accuracy with which they can deliver is extraordinary.

One of the things we did was upload all the content from the summit at the end of every day into Deep Research and then into NotebookLM. We created a podcast the next morning, which I found fascinating.

Did you get a chance to play with Google's new video teleconferencing capability?

Video teleconferencing?

No.

They built and demonstrated something called Project Starline at the summit. You sit in a room with a video screen in front of you, and someone else is in a different room on the other side of the country, with another video screen. The camera is there, and I kid you not, it gives you 3D imagery as though the person is sitting in front of you.

In their demo, they hand you an apple, and it feels like the apple is a few inches in front of you. I brought my kids in there, and they said it was by far the coolest demonstration in the entire Google Tech Hub.

That's fun. Project Starline—I have to check this out.

They're going to commercialize it by the end of this year. It's a new version of telecommunication—a version of Zoom that doesn't feel like a flat screen in front of you.

We had a session on longevity, which was amazing. We had sessions on how to upload your brain into the cloud. Travis Kalanick was there for our Moonshots program.

We ended the summit with Palmer Luckey. Palmer was with us 2 years ago, and he was such a hit. He did not disappoint again this year. What did you think of Palmer's closing presentation?

First of all, the amount of wisdom and brashness coming at the same time was remarkable. You don't often get people with a lot of brashness and a lot of wisdom. He's got both of them. It's really annoying—in a good way.

I love the fact that he said he was going to do this partly to take revenge on all the people who thought he would never make it and counted him out.

I'm really glad he's on our side. That would be one thing.

I really loved the fact that they've been in Ukraine from day 1 helping out. It's devastating to see what the Ukrainians are going through, and they need all the help they can get. That was fantastic.

I also thought the way he framed “follow your dream” was awesome. Figure out where you can make the most impact and shut up and go do that. This dream stuff will get you in trouble every time.

I thought that was such a great dimension to add to what we always talk about.

I thought it was great. I said to Palmer, “It's pretty crazy that you started Oculus in your garage, and then a few years in, Mark Zuckerberg offered you the ability to sell it for $1 billion, and you turned him down.”

He said, “That's right.”

I said, “How do you turn that down?”

There's a word for that: hubris.

Hubris, yes. A good Greek term.

Then he took $2.2 billion or $2.3 billion, and I said, “Then he goes—”

But that wasn't the reason I sold it for that amount of money. The reason I sold it was that Zuckerberg promised me he would spend $1 billion a year developing the technology for the next 10 years.

For 10 years.

When I did the calculus, I could either try to do the work on my own, and ask myself whether I could ever get a $10 billion R&D budget and still maintain some control, or I could take his offer. That was it. I sold it for that reason.

Then he got thrown out.

When Facebook changed its name to Meta, he invested in Meta again and committed to the technology.

That was his dream: the metaverse.

The “figure out where you can make the most impact” idea counterpoints his own story, because that was his dream—to figure out the metaverse.

The part I also loved was that they said they were going to spend $1 billion a year, but they ended up spending $60 billion.

My God.

The metaverse will take longer to get there because of the hardware footprint, but it's going to be amazing when it does. It will open up whole new worlds—literally.

We'll be releasing the podcast with Palmer on the Moonshots channel, so I don't want to ruin it for folks. It was most definitely one of the most fascinating conversations. He does not disappoint.

All right, brother. Thank you for breaking free on a Saturday to record this episode of “WTF Just Happened This Week in Tech.” I love you, I care about you, and I can't wait to see you again very soon.

I have to tell you, massive congratulations.

What for?

For what was the best conference anybody's ever been to. For folks listening, figure out a way, scrape together whatever you can, change your plans, and get to next year's Abundance 360.

Yeah, and if you can't come physically, our Exponential Mastery Program is a chance to join via livestream, as well as a chance to get all of the Salim and I and others' recorded content. We've got 17 hours of training content so that all of the stuff we discuss at these summits makes sense to you.

Have a beautiful weekend, brother.

Same to you. Big hug to the family.

Thanks. You too, everybody.