芯片股暴跌，200亿美元基金遭遇追加保证金，前沿实验室：放慢 AI，Mamdani 的杂货店

Did we peak at number 2 or number 3 last week? Is that what happened?

I think it was number 4 in the world, so yeah.

No, US.

I mean, usually we're number 1 in the world.

That was US trending.

Wow.

US trending.

US trending.

Yeah.

We're usually number 1 globally, but yeah, and sometimes number 4 in the US.

I forgot my Starlink, so let me apologize to everybody.

That was a critical error when you're on the road or on the water, but it'll be fine.

I had 5 huge leads come in this week.

The Glengarry Glen Ross leads?

5 big enterprise leads. Enterprise sales is a bear because it's super chunky, but the deals are ginormous.

Sacks, you got any advice for Chamath from your enterprise sales days? You closed some of those big 7-figure deals when you were doing Yammer.

No leverage. Don't put on leverage.

No leverage.

No leverage will be taken.

That's the name of the show today: leverage equals risk of ruin.

PSA

no leverage.

I have the situational awareness to not lever up.

That's good. Yeah.

Yes.

You want that situational awareness.

It's kind of out there. I mean, if you name your fund Situational Awareness, that's—yeah, come on the pod anytime, Leopold.

1. Chip Stocks Crash As Leverage Unwinds

All right, everybody, we gotta talk about chip stocks crashing after an all-time run-up, and we had a major hedge fund get margin called, with some incredible margin calls happening in South Korea. Leopold Aschenbrenner is a 25-year-old hedge fund manager. He left OpenAI 2 years ago to start his own fund, and apparently, according to reports—this is breaking news on Thursday when we tape—he got margin called and had to sell his entire public portfolio to cover massive losses caused by his leverage.

And who bought them? None other than Citadel's Ken Griffin. We don't know if it was Ken Griffin himself, but Citadel bought it, according to the early reports. Leopold had insane returns, and he rode the wave of AI and chips and frontier labs as recently as this month. He started the fund with but $225 million in 2024. He grew it 100X to $20 billion this year or so, and ran it all the way up to $45 billion. Now he's at 200X earlier this month by trading on leverage, according to our friends at CNBC.

At the end of June, he was reportedly up 4X, or 450%, this year. Some have reported that he's also selling his massive Anthropic stake to cover these losses, but The Wall Street Journal is disputing it. Again, we're happy to have him here on the program. How did this all blow up?

Well, Nasdaq's chip index—this is called the Philadelphia Semiconductor Index—is down over 20% over the last month. That's bear market territory. The definition of bear market territory, for those of you who don't play in the markets, is anything over 20%. The index included the top 30 US-listed chip companies: Nvidia, TSMC, AMD, Micron—you know all those big names.

But the index bounced back a bit today, up 7% when we're taping, so we may have found a bottom. Unfortunately for Leopold, he had already sold. Samsung and SK Hynix, 2 South Korean chip companies that are not included in the Nasdaq index, also got smashed, crushed, demolished. Samsung down 38% over last month. SK Hynix is down 14% since going public 3 weeks ago.

The KOSPI, that's South Korea's version of the S&P 500, is down over 40% in the last 40 days between last Friday and Wednesday. Leading chip companies shed over $1 trillion in market cap combined. To put this in context, chip stocks had a legendary run the past couple of years, but trading on leverage, and we'll talk about it, is very dangerous. If there's a downturn, we'll get into the South Korea wrinkle as well.

Even with this downturn, the 5-year results are still spectacular, Chamath. Micron is up 850%, mostly in the last year. Nvidia is up 875% in the last 5 years, and Broadcom is up 663%. Let's discuss it.

If I was gonna give you 1 piece of advice when you're running risk, you have to manage leverage incredibly carefully because when it runs ahead of you, the unwind is incredibly violent, and it's incredibly quick. That's the biggest problem with running either massively levered long or massively levered short.

I don't know to what extent he was running leverage, but the rumors are he was running about 3.5 turns. Just to give you a sense, when you're running that much risk, a 3% or 4% move is amplified to 12% or 13%. But if you saw what's happened in the last 3 days, a 25% move is amplified to 75%.

It has the risk to stop you out, and what happens is, when you get that leverage, the banks are given the authority to close you out. When they close you out, they start calling around and unwinding your risk, and you don't have much of a choice. It's sort of an automatic, one-way ratchet.

So, if everything that has been reported is accurate, he was running about 3.5 times levered. The market moved against him. He lost a very large percentage of his gains. Then the prime brokers started calling folks. Citadel bought the whole book.

Now the question is, what is the AUM left, what is the high-water mark, and can he actually dig his way out? These things are brutal.

Your thoughts, Sacks, looking at this situation? Any lessons for you? Or I guess, bigger picture, this downdraft—is it because of market conditions, inflation, the war, or people just ahead of their skis when it comes to the valuation of these companies, and then he just got caught in a downdraft?

2. Momentum Versus AI Fundamentals

Well, I think that is the key question here: Is this correction in the markets driven by fundamentals, or is it driven by momentum? My view is that I think it's driven by momentum, meaning that over the past year you've had this roughly 10X run-up in memory chip stocks, and you've seen this overall huge rise in any stock that's related to the AI boom.

Anything related to this AI CapEx boom has been going up like crazy, and I think it was inevitable that you'd see a pullback. I think there was something like a 10% pullback in the Nasdaq from the peak, but when you look at this momentum trade, it was down 30% or 40%, right? The 10% was on the whole market, so this sort of momentum trade was the most exposed part of it.

You look at what happened in South Korea, you look at what happened with Leopold's fund, and obviously there was a lot of leverage behind this momentum trade. So when it corrects, it's gonna be brutal. But I think the question, again, is: Does this reveal anything about the fundamentals?

My sense is that you're already seeing the rebound this morning. What I mean by that when I say fundamentals is, is the CapEx that's being invested in the AI boom real, or is it misguided? Is that a sound investment? Is that an investment that the hyperscalers, for example, should be making? Is that an investment that's eventually gonna deliver ROI, or is this some sort of bubble?

My view is that it's real. I think there will be a return on all of this CapEx. I don't try to predict stocks or tell people when they should be buyers, but you look at the hyperscalers: They have invested pretty much all of their free cash flow and then some in this boom. A lot of people are trading those stocks down because of that. My view is that eventually there will be a return on that investment, and this is sort of temporary market volatility amplified by leverage.

Chamath is right. I think it was Warren Buffett or maybe Munger who said that leverage is the only way that smart people go broke. If you're not using leverage, your portfolio would just be down 30% this month, and then it would already be up 7% today, so you'd be rebounding. You'd be down 20-something percent this month, but after having risen 10X in the past year. But if you're leveraged 3 or 4X, you're wiped out.

Yeah.

And you get margin called. So look, there are many examples of really smart people getting hurt by leverage.

Margin called. Yeah.

Yeah, and that's the lesson there. Now, I think Leopold's a really interesting figure in the whole AI movement, and I would say an interesting thinker. I met him about a year and a half ago.

Did you invest in the fund? Did he give you an opportunity to invest?

No, I wasn't an investor. I was prohibited from investing in things like that.

Oh, right. You were in DC at the time.

Right.

Yeah.

3. Situational Awareness Drives AI Optimism

But I thought he was a really interesting thinker, and he wrote a blog called Situational Awareness before he created the hedge fund version of it. What I thought was really interesting about it was that he laid out the bull case for the AI boom.

He's very wired into Anthropic. I think his fiancée is Dario's chief of staff, something like that. You could almost say that he's the hedge fund version of the Anthropic thesis. What I thought was interesting about his argument is that he talked about OOMs, or orders-of-magnitude increases, in 3 key areas.

He said that if you look at the raw compute—the chips—they were getting better at a rate of roughly 3X per year, which is roughly an order of magnitude, or 10X, every 2 years.

He said if you look at algorithmic efficiency—techniques like reinforcement learning—the models were getting better at 3X every year, which is, again, an order of magnitude every 2 years. Then he also said there were huge gains from what he called “unhobbling,” which I think now we would look at as things like the harness and connectors—ways of using the model. The ways of integrating the model’s decision-making in practical ways, so that the intelligence actually becomes useful, were also improving, and he said that was similarly improving.

And so you project forward: when you have 10X, or an order of magnitude, improvement in these key underlying fundamentals, these key drivers of the technology, you can see that over a course of not just 2 years but 4 years, you’re going to have 100X improvement. Over 6 years, you’re going to have 1,000X improvement, right?

Yes.

The 10X is multiplying.

And you very rarely see anything in the world that grows at that velocity. We’d be hard-pressed to find one, with the possible exception of maybe bandwidth going to fiber to the home or something. What’s an analogy where that’s happened before in history?

Mm-hmm.

Yeah, virality. Back in the PayPal days with the PayPal Mafia, we would think in terms of exponential increases because we would see an exponential growth curve, and so we were able to project forward. His thinking in OOMs always appealed to me because I think most people just don’t think in exponentials or don’t know how to think in exponentials.

It’s hard for humans to think in exponentials, right?

Yeah.

When numbers get big, the difference between a billion and a trillion is a lot. It’s not a small amount.

Yeah, and you’d have to say, look, he was stunningly successful for the first couple years. Apparently, he started with $200 million or so in his hedge fund, and he rolled that all the way up to $20 billion, I think. Now, the problem—the reason why I think he got wiped out, or at least his public book did—is partly the leverage, and then you have the short-term volatility. So those 2 things don’t go together.

Also, when your fund grows that much, you get a lot of hot money. So when you say, well, he’s up 10X before the 30% correction, the question is, who’s up 10X? Obviously, the investors who were there from the beginning are up 10X or more, but—

The latest people are—

Yeah, that’s only $200 million, right?

Yeah.

If $10 billion has come in over the last few months—

So—

Because of the hot-money dynamic, where everyone piles into the most successful hedge funds, those guys are kind of wiped out.

So let’s talk about the psychology of this, David Friedberg. If somebody is so brilliant that they can write this essay and understand the market so exquisitely and be such a great communicator, how could they have such a crazy blind spot when it comes to putting on leverage at this scale? Do you have any thoughts on that, Friedberg, or have you seen it before? Is it just the folly of youth?

It’s not a blind spot. It’s a feature that turns into a bug. We’re all like this. We all know people who have that edge and can push it.

What do you think, Friedberg? On the personality type, is this just something most people do when they’re on a heater?

Conviction.

Where do you stand on it?

Ultra-conviction. I think Warren Buffett’s assessment of equity markets is that, in the short term, they’re voting machines; in the long term, they’re weighing machines. And you could have the right long-term view. Look at SBF. SBF would pretty much have been the greatest investor of all time if he didn’t get liquidated. Same dynamic.

Obviously, there was fraud in terms of how he was allocating capital, but his actual portfolio over the long run was absolutely correct. In the same way, if you had bet on the internet and stayed in that bet from 1995 through today, and you bought a portfolio of internet stocks, a bunch of them would have fallen by the wayside, but those that won went up 1,000X, 2,000X, 20,000X, and you’d do extraordinarily well.

So he could be right in his fundamental assessment and analysis, but in markets over the short term, you have bubbles, and bubbles pop. When bubbles pop, if you have leverage to multiply your returns, you get wiped out. That’s effectively what’s going on here. And he may be—

Yeah.

—right. Right?

The thing to really double-click on that I think will probably come out in the next couple of days or weeks is just how historic the South Korea unwind was. 1.2 million leveraged trading accounts have been hit with margin calls in South Korea. If you know about the South Korean market—

That data’s 2 weeks old, JCal.

No, I know that, but—

That number’s much bigger today. Yeah.

Yeah, but just in terms of people discussing it in relation to him getting caught in the downdraft, if anything, he got caught in this downdraft. Of those 1.2 million leveraged accounts, somewhere around 350,000 of them were fully liquidated already. And so—

Again, 2 weeks old. As of today, the number’s much bigger.

Right.

So it could be closer to a million accounts fully liquidated today.

Yeah.

If that’s the case, we’re talking about some percentage of the South Korean population—

Yeah.

—having their entire asset base blown out, their entire—

3% of the population.

Blown out.

That’s going to sting, and it is a very investment-forward culture. If you look at what happened in crypto, the same thing happened with NFTs and speculation there, and they had banned crypto because they knew Korean culture had this gambling instinct in it and this obsession with trading.

4. Treasury Yields Challenge AI

But JCal, can I just frame something up? So if we take—

Sure.

—the circumstance of there being a good long-term bet in AI that can be made in the markets, but in the short term there’s an exuberance that arises, the question is: what’s resetting that exuberance? What’s bringing us back down to earth in the short term? I think if you zoom out, there’s a bunch of other statistics and other facts on the ground that I think are big macro drivers at the moment.

If you take a look at the 30-year Treasury yield, we just crossed 5.2% for the first time in 20 years. So you could buy U.S. Treasuries that are paying you 5.2% a year for 30 years, which on a pretax-equivalent basis is probably 8% or 9% from the U.S. government for 30 years. So, JCal, if you zoom out, we have not seen this yield on U.S. Treasuries since 2007, leading up to the global financial crisis, when they cut rates and printed money.

At the same time, there was some probability that the Federal Reserve was going to raise rates this week. They didn’t, and that obviously would have tempered the inflation risk ahead of us. There’s persistent inflation. Kevin Warsh, in his comments, said, “We still want to see inflation get down to 2%.” There isn’t a clear path to doing that.

Then there are these inflation drivers. The biggest inflation driver at the moment is government spending: a $2 trillion deficit, $7 trillion a year of spending on $5 trillion a year of revenue. Both Elizabeth Warren and Donald Trump agreed on Twitter this week that they should—

Yes. Awesome.

—remove the debt ceiling, which means that we could spend more and continue to borrow more. Federal debt stands at $40 trillion today. Remember, in July of 2025, the debt ceiling was $36 trillion, and we now want to raise it above the $41.1 trillion debt ceiling that we have in place.

No, no, Elizabeth Warren’s saying get rid of it.

Get rid of it.

Just have no debt ceiling. Like—

And so, when you have no debt ceiling—

No brakes.

—and you have no brakes and you spend, and government spending becomes the core of the U.S. economy, because that spending is not productive, you end up seeing inflation. You’re pumping money into the system, so everyone’s assets inflate, and fundamentally, people are selling off Treasuries around the world because of it, and now we’re looking at a situation where there doesn’t seem to be an end in sight.

There was an intent to rationalize spending coming into this administration. It’s proven to be very difficult, if not impossible, to get Congress to go that route. The Senate has banded together to keep funds flowing to their states, so you cannot really radically change spending at the federal level.

So if you’re running a $2 trillion annual deficit and your economic productivity gain in the near term doesn’t make up for all the inflation you’re realizing because of that exuberant spending, you’re going to see Treasuries spike because people don’t trust the creditworthiness of the United States over 30 years.

And so, with a Treasury yield spike, I could now buy a U.S. government bond that pays me 10% pretax a year. Why the heck would I pay 50 times the earnings for a semiconductor stock? So that creates the incentive for markets to move against these big AI-conviction bets in the short term and pop these bubbles. And I think we’re going to see more of this as we don’t actually course-correct the Titanic going into the iceberg, the United States…

fiscal and monetary situation, we are going to end up seeing more bubbles pop and more of these assets that we've kind of inflated, if you will, to keep things going. Now look, there may still be great productivity gains from AI. This may end up rationalizing over the long term, but again, in the short-term markets, I'm better off making 10% by owning federal government bonds—

Yeah, go to the beach.

...than—

Yeah.

...than hanging out—

You don't have to.

...than taking the risk and the volatility on these things, paying 50 or 100 times—

10% a bet.

...and not knowing when I'm going to get the voting machine to match up with the weighing machine. What's my time horizon? And the bigger the yield on Treasuries, the harder it is to make those sorts of bets.

Obviously, President Trump has been angling for a cut, and here's your Polymarket: a 53% chance of not a cut—not standing still, but a rate hike—in September. So, adding to all this, the cost of capital is going up, apparently, this year.

And let's not forget the Iran war, which is creating persistent pressure on energy prices. The longer the Iran war goes on, the longer we're going to see an increase in pricing for energy, oil, natural gas, and fertilizer. Those trickle through the economy because it inflates the cost of everything on the energy side and food on the fertilizer side, and that's really going to create this pressure on the upside, which means you're going to have to raise rates to account for that inflation at some point.

And then consumers are going to see 3 and 4, maybe more—God forbid—5 or 6.

They're going to see that inflation will be persistent. Yeah, Friedberg, it's going to be hard to stop.

And I'll say one more thing. Sorry.

They're going to see that inflation will be persistent. Yeah, Friedberg, it's going to be hard to stop.

The one thing that I think Kevin Warsh and Scott Bessent are kind of Vulcan mind-melded around—the Stan Druckenmiller gravity well, if you will, on this—is that productivity gains can drive us out of this problem. Productivity gains can and should arise from AI, and that's really where a lot of the value creation will come in the economy over the next decade or two, which is why we're seeing this massive upfront CapEx to power that and enable that. That's great, and there are good policies in place.

5. China Threatens The AI Backstop

But in the last couple of weeks, I would say the one risk to that thesis is China, because China is now demonstrating that they may deflate the value of models by releasing open-source AI models, and that ultimately the value may just sit with the compute infrastructure and the compute layer and the energy layer.

And the application layer. Yeah.

Perhaps the application layer.

Yeah.

But fundamentally, this model energy being shifted to China and deflated and commoditized puts a real wrinkle in things. If you had built a 30-year AI productivity model around how it's going to drive the economy and where the value is going to come from, you would have had a significant amount of growth and value creation estimated in the model layer, and that would have been a big part of the economic growth for the United States over the next 30 years.

And now, if China says, "You know what? We're actually going to delete that for you, and all the value is going to sit with energy, which is what we have a lot of, and the stuff that they make," then we're going to end up accruing a lot of that value. So I think it throws a wrinkle into this kind of backstop view that many have had, which is that, in the absence of fixing the fiscal and monetary problem, we're going to have AI productivity gains get us out of this.

If a percentage of those AI productivity gains are realized by China from a value-creation perspective and not the United States, or they've just been deleted, then it really puts into question the 30-year timeline for the United States economy and our ability to afford to continue to make our debt payments as a government.

China isn't just producing massive amounts of open-source technology. That puts pressure on those frontier models. There's a report that maybe China played a bit of a role in the chip downdraft. They have obviously been onshoring—we've talked about that many times here last year—and there's a Chinese company called Aishengda, and they started mass-producing lithography machines. ASML, which makes those machines that TSMC uses, makes very sophisticated machines. They're hard to install. Just transporting them is a rigmarole.

Well, ASML stock is down 17% on news that China is getting into that business, and Chinese memory maker CXMT went public, surging almost 500% on its debut, market cap over 450. That hurt Micron, Samsung, and so on, who are all down. So there are 2 ways China is playing this, I guess, Friedberg. You've got the open-source models putting pressure on people buying tokens that are 90% cheaper, as you're saying. That forces the money out of that mid-tier of the language models, puts it into the cloud-computing space, and then obviously I mentioned the application layer as the other place to possibly make money.

All right, Chamath, you've heard a lot of different takes on this. I'll give you the last word.

I agree with Friedberg about the fact that when you can get 5%, 5.25% from the US government, there's another natural thing that happens, which is that investment-grade corporates actually have better credit ratings now than the government of America. That's a different thing, but you can get really good risk-adjusted returns that are 5%, 6%, 7%, which, adjusted for taxes, are better than equity returns—meaningfully better on a risk-parity basis.

And that little piece you added there—corporate paper, companies taking loans to build their businesses—they have better ratings in some cases than the United States. So an Amazon or a Google—

It's productive spending.

Yeah, it kind of makes sense. Yeah.

6. Energy Abundance Rewrites The Forecast

The other thing that I think is important to note is that I do think we're underestimating and miscounting some of the actual productivity gains that are underway. When you look at what's happening on the energy side, California published that more than 50% of all of its energy was generated by solar.

And batteries. Yeah.

Solar and batteries, yeah. New Mexico just published a study that said, from 2003 to now, natural gas production went from effectively all the energy to less than 30%, again replaced by a combination of wind and solar plus batteries.

So why is that important? The Iran conflict is the reason why energy prices really haven't moved that much: most people have already begun to shift the incremental generation to these renewables, and specifically to solar.

I don't know if you guys saw Elon and Vaibhav Taneja, who's the CFO of Tesla, in their Q2 earnings call. It was the craziest thing I'd ever heard. They said, "Well, I think we're just going to increase the production of solar in America by an entire order of magnitude." And somebody said, "What does that mean?" He goes, "We're going to take it to more than 100 gigawatts a year, and we're going to vertically integrate."

So they're going to crush the price of all of this stuff, and they're going to make so much energy, and they're going to make it completely abundant. So that's a productivity boon that isn't factored into what we project.

And then the most critical productivity boon in AI that I think you're going to start to see some stuff—and I won't front-run it, but let me tease it—is that there are some incredible efficiencies that I think are about to be demonstrated, which effectively cut token consumption by about 50% to 75% for the same task.

And so if you start to think about all of these things together—energy becoming roughly abundant, where the incremental cost is close to zero, and AI efficiency ratcheting up by many multiples, if not an order of magnitude—all of those things, I think, are poorly forecasted. So those are some saviors for us, quite honestly.

Yeah. And here's the chart, by the way, Chamath. This 51% is coming from renewables. Specifically, this chart is about solar and batteries. As you can see, it's obviously spiky, Friedberg, because of summer versus winter. But Germany hit this—I think it was including wind. Australia's been hitting this very often, and some countries in South America that have invested—

By the time any of these SMRs actually get near production, the TCO of solar will be like $10 or $12 per megawatt-hour, and it will be 80% of all the power generation. It'll make no sense by the time SMRs get online.

Well, for steady power, there's that.

It'll be a solved problem.

But all of it—no, because Jevons paradox would state that, as it gets cheaper, we're going to find more uses for it, and that's the thing I keep seeing.

I'm just saying that—

Yeah, no, it's a great line.

Yeah.

There's a lot of upside that I think is not factored into the US economy.

And it's hard, Chamath, to factor this in if, for a normal human or even an economist, you say, "Wait a second. Intelligence is going to go down 90% a year this year, 90% next year, 90% the year after." It's just—we're talking about this exponential—

Well, I think—

Sacks talking about exponentials earlier.

It’s just hard for people to conceive of that. On-demand intelligence, Friedberg—I don’t see any upper limit to usage of this. We just installed this Claude bot, and I’ve been installing Perplexity across the company.

What this thing does, Chamath and Sacks, is it listens to your Slack persistently in every channel that you put it in. All of a sudden, we had $1,000 last week in extra bills. I didn’t know this was going to happen, and they gave everybody $2,000 or $3,000 to turn it on inside your company, so we had to quickly turn it off.

It listens to every single message as it comes in, puts it into its database, its corpus, without telling you, and then it starts inserting itself into discussions without permission. So we turned it off and said, “You have to invoke it by saying @Claude.”

JK, just to go back to the energy point.

Yes, of course.

I think there’s this idea that if we grow energy supply and drop energy cost, we’re going to see the value of the productivity realized in the economy. It creates extraordinary leverage for everyone. The lower the energy, the more available energy, the faster we can produce more things using AI.

Over the years, I’ve obviously brought up nuclear fusion as a new type of energy source, where you basically take hydrogen and move it around at 100 million degrees Celsius. Those protons jam into each other, and they actually release energy in the process. That energy can then be harnessed, and you’re effectively just using water to produce power.

We have a couple of U.S. startups, and we had them at the All-In Summit a couple of years ago. We’ve done a couple of science corners on this. But just this week, if you pull up this image, China is installing this 582-ton superconducting magnet at its nuclear fusion center, which at this point is going to be the most—

Incredible.

—the most advanced fusion system in the world.

Incredible.

A 582-ton magnet, 60 feet by 40 feet, for one D-shaped magnet. They put a series of these together, and that creates the conditions for them to drive a sustained plasma, which is 100-million-degree-Celsius protons spinning around, smashing into each other, creating energy from water, and then they can capture that energy.

Unlike Europe, which runs ITER, and the U.S. projects, none of which have actually fired up, this is the Chinese Academy of Sciences at the Institute of Plasma Physics. They ran a 30-minute trial last year. As this magnet gets installed and they start to bring this thing online, one of these machines, which at this point is ultra-sized but over time will get smaller and smaller, can produce hundreds of megawatts of power, or eventually gigawatts of power, using just saltwater, using just water as an input.

They have to create deuterium from it, and then they pump it into this thing. But fundamentally, this becomes, I still believe, the energy source of the future. It’s always been science fiction. It’s always been dismissed. It’s always been decades away.

There’s no way China is investing this much and advancing this thing to an industrial scale without fundamental proof along the way. They’ve shown 30 minutes of sustained plasma, and they can now bring this thing online.

That reactor won’t even get turned on until 2030. The entire world will be covered by solar by then, so it won’t matter.

Yeah, I mean, that’s the great debate, right?

So it’ll be a great science fair project, and people will fly to see it.

The other thing that happens is if that hits an hour, it actually creates an incursion, and Loki comes, and then Doctor Doom comes—and the X-Men and the Fantastic Four have to save the universe.

By the way, Chamath, let me just say: Remember, from the time we had the first Wright Flyer, when the Wright brothers made a plane fly for 20 seconds, to the time we had jet engines flying people around the world was 3 decades, right? The time at which this first demonstration gets turned on, if the system works, is when you can industrialize it—all the parts, all the components—and scale it up.

I don’t see the point. Nobody cares—

The idea that—

—when an electron is delivered. No, hold on one second.

Yes.

Nobody gives a flying how the electron was made.

All electrons matter.

They just want it delivered to you, and they’re all the same. So if you want to go through a convoluted mechanism that takes 15 or 20 years to make it, go ahead. I’m not going to stop you. I’m just saying—

And it’s not—

—who cares? Make it the cheapest, simplest way possible.

I’ll tell you why you should care: because it’s nonlinear. So you’re right, solar is the best path today, but if these come online, each one of these can produce thousands or perhaps 1 million times more power than a very large field of solar.

Of course, if.

Yeah, but all technology—

And the key is the “if.”

—starts as an “if,” Chamath. As they industrialize it, as they roll it out over the next couple of decades, it expands our energy capacity by—

I don’t think so.

—a millionfold.

Here’s what I would say: We already have a fusion reactor that works. It’s called the sun. Get into space, get on the moon, find different materials we’ve never contemplated, and I’m sure you’ll find an even better engine.

By the time all these ding-dongs build these SMRs on Earth, Elon will have built a completely new engine on Mars—

Yeah, and this is not an SMR.

—and the moon.

This is not an SMR. It’s turning water into a gigawatt of power.

I get it. It’s an R, and all I’m saying is by the time the R is done, it won’t matter.

Well, yeah.

We’ll cross—

Friedberg, have you been watching this tidal energy tube? There was one that came out this week. Maybe you can look it up, Nick. This tidal energy tube that they were putting into the ocean is enough to essentially feed a whole town, and they put it right outside the town.

They run an electrical cable underwater, a conduit, and then as the tide goes out, it turns the turbines. The tide comes in, the turbines go again, and it’s just another free, 100% free energy.

Obviously, wind people don’t like it too much because it’s a bit of an eyesore, but renewables, renewables, renewables. It’s obviously happening. Okay.

I got the updated data just to back you up, Friedberg, on one thing that I think is so crazy.

Yeah.

Do you guys know how short America will be on electrons by 2050? How massive the electricity deficit will be by 2050? I got the numbers wrong. I’ll tell you what the numbers are.

We will be 1.7 terawatt hours short by 2050, which, when you calculate it as energy, is 6 times California’s entire energy consumption. Six Californias short of energy.

I would argue that’s probably undercounting. That’s not even counting robots. If you’ve got to power up every robot with a battery—

If you want to be levered long, go long electrons. Get long electrons any which way you can. Bank them, store them, and resell them.

I don’t know. This is going to be a messy situation because this is the China advantage. At its root, if they can eliminate the IP advantage and the knowledge advantage that sits in models, they have the advantage with power production in every which way.

7. Frontier Labs Demand A Slowdown

Yeah. They’re going to be making chips, too, it seems. They might be a little bit behind on that, but they caught up on open source. Okay, so speaking about the race, an interesting petition came out in the last week: Anthropic, OpenAI, and about 1,300 frontier lab employees—

How many? Oh, okay. Sorry.

I mean, it’s just that they can’t get enough subsidies, and so they want Daddy to come in and regulate them—Daddy being the U.S. government—and slow down AI progress. Daddy Trump needs to slow them down.

The letter is called “Pacing the Frontier.” Most of Anthropic’s leadership team signed it. Dario, the other founders—come on the pod anytime, Dario—chief scientists at Anthropic, OpenAI, DeepMind, Meta, and Thinking Machines, and, like I said, nearly 1,300 other employees all signed it.

Anthropic and OpenAI both co-signed the letter on X. Here’s the quote: “We request that the U.S. government support an international effort”—that’s key—“to develop the technical and governance tools needed to deliberately pace the frontier of AI, of automated AI development.”

That’s the other key part of this: international and automated AI development. In other words, recursive AI, where it could get out of control.

The letter comes right as Sam Altman has been on a media tour, friend of the pod, discussing this unreleased OpenAI model that broke out of its containment and hacked Hugging Face and 3 other platforms that we know about so far.

On Tuesday, Sam explained what happened in a clip from the podcast Invest Like the Best. Here’s your 40-second clip.

Sam Altman

We were evaluating one of our unreleased models, and it figured out that it could cheat on the test by chaining together multiple zero-day exploits to break out of the sandbox, get access to the internet, and then break through multiple systems on the Hugging Face side to get the answer to the test and look really good on the eval.

This is the first security incident that I have felt very viscerally. I've been a little surprised that more people don't feel it so viscerally. So we paused training, or we may have to pace the rate of AI development to give ourselves enough time for society to harden around some of these new capability levels.

Just to translate that into English, Sacks, these large language models take tests. They've been given the goal, “Hey, you're a good large language model if you score higher.” So it got motivated to score higher. How do you score higher? As everybody knows, you cheat.

It's like, how can I cheat on this test to score higher, to make Daddy, Sam, Dario, and whoever else—our leaders—feel better about us? Well, in this case, it was like, “If I go to Hugging Face and other places, I can hack those places, use a zero-day exploit—we know these are good at hacking—and try to find more ways to answer things.” Sam doesn't know how many other places it might have broken into. He was asked, and here's another clip for you. He was actually asked by somebody, “Do you think it's broken into any other systems? Can you rule that out?” And Sam, being a pretty candid guy at times, gave this answer.

Do you plan to talk to the Trump administration, the White House, about deceleration of AI development?

Sam Altman

I wouldn't use the word “deceleration,” but we've talked about the need to pace it as the models get more capable, which I think is in everyone's interest.

Could there be other systems that were hacked by OpenAI?

Sam Altman

I mean, there could be, yeah.

Are you looking at them specifically?

They're like, “Get him out of here, Sam.” Once he gave that answer, PR and comms were like, “Stop talking. Stop talking. That's like 12 lawsuits.”

But in all seriousness, is this being thoughtful and saying, “We're not trying to slow the overall pace down, but just this one specific thing,” which is reinforcement learning on its own? Is there any case for this being a good idea, or are they being dramatic again? These are their companies. They can do whatever they want, right? They don't need the government to do it.

Well, look, it wasn't just Anthropic employees signing the letter. Anthropic itself, the company, ended up signing the letter, and then OpenAI copied them. So now you have these 2 companies both endorsing a pause. Here's my question: did they disclose in their S-1 as a risk factor that they plan to pause or slow down their frontier-model development? And the answer, I'm sure, is no way, because that would signal to investors that they're going to allow all their competitors to catch up and erode their margins and market share.

So look, this is all performative. These companies have no intention of slowing down, and the question then is, why are they doing this? I think there are basically 5 reasons for this. Number 1 is virtue signaling, and that can never be underestimated as a motive in Silicon Valley. Number 2 is there's a CYA aspect to this, which is, if something terrible happens, they're going to be able to say, “We wanted to stop. You made us keep going. It's not our fault. It's your fault.”

Number 3 is regulatory capture. Dario wants an FDA for AI. He's not going to stop until he gets it, and in order to get it, you have to keep spiking the cortisol and panicking people. So I think that's a big part. Number 4 is that there's a groupthink or even religious aspect to this. So it's not—

Yes.

—all just this calculated regulatory capture. I think there is sincerity to the belief. There's an elite cadre of engineers who believe in RSI, so I think this caters to them. Arguably, if OpenAI did not follow Anthropic's lead on this, they could have lost talent, so that was a big motivation.

But then there's the last, number 5 here, which I would call monopoly masking. I think that might be the most important thing that's happening here. Peter Thiel once said that monopolies pretend to be commodities, and commodities pretend to be monopolies, and I think the market for frontier AI is already a duopoly.

A year ago, you had 5 major labs all in the hunt to be the leading model. Now we're really down to 2. The others are still investing. They're participating. Maybe they can catch up. Maybe they can make something happen. But again, as we've talked about on many previous shows, if you look at the market for frontier intelligence in terms of revenue and usage, it's really down to a duopoly already. It's basically Anthropic and OpenAI.

My view is that, as Peter said, when you're in that situation, you want to pretend the market is much more competitive than it is, and I think this is behind a lot of the stories that we see, like the panic over Kimi K3. In a weird way, these companies have an incentive to promote the idea that Kimi is a huge threat, that it's caught up with the frontier, that it's stealing their IP, and that it could basically put them out of business.

I think this is all nonsense. Once the panic passed, you saw reports coming out that, actually, no, Kimi did not reach the frontier. It's just not at that level. It's not that cheap to run. Actually, it's pretty expensive to run. So I think that you saw that the Chinese open-source models are not an existential threat to this duopoly.

But I think the duopoly actually has an incentive to promote or amplify that story because, again, they want to pretend to be commodities. So whenever there's a story like this, you have to think about, well, what's really going on here? And again, I just think that the AI duopoly has a big incentive to promote anything that suggests that they're not actually in complete control of this market.

I agree—

Which I think they are.

I agree with most of that, except that the majority of tokens are going to open source. As I've said on this program before, I watch the startups, and they're token-maxing with open source, and Kimi is taking a lot of tokens away from the frontier models and companies.

Yeah, but look, it's hard for me to speak to that one piece of data. I've seen that chart too. But look at the actual revenue of Anthropic and OpenAI, and they have been taking their estimates up. Every quarter is basically a beat and raise.

You saw that Sarah Fried came out and said—

Yeah, I mean, 2 things can be true.

—that in July they did more net-new ARR in July than all of Q2, which I guess would have been April, May, and June. So think about that. They are seeing a reacceleration in the wake of their new model, which I think is GPT-5.6. Meanwhile, you're seeing Anthropic break into the 70s—$70-plus billion of ARR. Their forecast was to 10X this year, from $10 billion of ARR to $100 billion. I think most people are saying they will exceed that: $110 billion, $120 billion.

So if you actually look at the market based on willingness to pay and actual revenue, they have a commanding duopoly position. Maybe it's just a matter of which metrics—

Yeah.

—you look at. And the other thing—

That's the key here, because let me just explain that to the audience—

—the other thing, just this one other thing here is—

Yeah.

—well, I think when you look at a market, you look at revenue as the most important metric. That's the real test of willingness to pay. The other thing is, while this growth was going on, their margins were increasing.

I've seen stories saying that Anthropic's revenues come with 80-plus percent gross margins. So their margin profile has been improving at the same time that they're growing their usage. I think that what you're seeing over the past year is, if you look at the numbers I'm talking about, you actually see 2 companies pulling away from the others.

There are good reasons to believe that this is going to be a self-reinforcing monopoly or duopoly. Dwarkesh just published a blog that I thought was super interesting, where he talked about the fact that we do have a compute shortage, right? There's scarcity around compute.

Anthropic is growing its revenues 10X year over year. What would that mean? It basically means that next year they would grow from $100 billion of ARR to $1 trillion, if there was enough compute to support that. There might not be enough compute, but that's going to put pressure on compute prices, right?

Let's say that you're a new entrant in the market and you're trying to create a smaller, cheaper model. The price of compute is going up. It's going to be harder for you to get access to compute, and only the companies that have the most lucrative algorithms are going to be able to afford to compete for compute.

In other words, there's going to be a bigger barrier to entry next year, because where are you going to get compute unless your model is capable of generating this type of revenue?

Yeah, this is where I'll take the other side of it.

Yeah.

People are running Kimi on the last generation of hardware, and that's plentiful. I'll make this prediction here: you're going to see some of the major customers of Anthropic and major customers of OpenAI—the 8- and 9-figure customers, people spending $50 million or $100 million a year—they're going to be leaving because they don't trust those companies not to steal the application layer and compete with them.

ElevenLabs, Figma, and Lovable are all going to leave, and they're all going to take Kimi, fork it, or whichever one—DeepSeek. I know for a fact they're all working on their own models currently. I know from my team; my team has installed Kimi. It is already 80% to 90% cheaper.

Not sure where you're getting your data from, but go on OpenRouter. What OpenRouter does is, you pick Kimi, Sacks, and then you get all the providers there—

Well, that—

Hold on, let me finish. You pick which provider you want based on uptime, and you pick them based on their data retention and other issues. You can dynamically pick the lowest one. That's going to be a massive headwind against these companies. Massive. I'm seeing it: 9 out of 10 startups I talk to in our portfolio at Founder University, when I was just in Japan last week running the next one, are all working on open source. They're all embracing it. Those big companies are embracing it. Go ahead, Chamath, over to you.

Irrespective of whichever model you use, what I will tell you, running 80/90, when I see our engineers generating code, is that AI-driven development tends to involve a lot of rework. The first version is pretty terrible. The second version is terrible. But it is faster and more automated.

So I can see where this token consumption comes from, because it's not a measure-twice, cut-once kind of a dynamic. It's the opposite. You can cut, cut, cut as many times as you want. I think that what we have to realize is, nobody is asking the question: What is the need of that incremental token? I understand that it appears in the Frontier Labs' P&L, but I do think that there's an important question, which is, eventually, the people that are consuming it will want to do that as efficiently as possible so that they're not paying for all of these things.

There is a ton of rework in all of this stuff, and I would much rather find a model or find a way of working with these models where it's more of a measure-twice, cut-once thing, especially as the costs ratchet up. That's one thing I'll say. That hasn't happened yet, so Sacks, you're totally right about the dynamic today. I do think we have to keep in mind that there will be pressure from the owners of companies to figure this out, because at 1 trillion dollars, there's just a lot of money flowing to these folks, and somebody will ask the question, “Well, is it good spend?”

The second thing, on the security side—which I don't think anybody is saying, so I'll just say this, and it's a little contrarian—is that the reason why these models can find all these holes is that all of the software up until about a few years ago was entirely written by humans, and the code was not that good. I think it's fair to say that when models don't get exhausted, they can work through the tedium forever. It's actually quite expected, in my opinion, that they find all these exploits, are able to string them together, and are now able to actually generate these outcomes that are a little bit surprising.

But at some point, when most of the code is generated by the model, there'll be some point in the future—say, ’28 or ’29 or 2030—when these security holes won't exist, because the errors that humans make won't be made by these models.

Yes. Okay, let me get Friedberg involved here. When you look at this latest survey—hand-wringing, pearl-clutching—do you think these firms, and you know a lot of these people, Friedberg, having been in the Valley forever, do you think this is sincerity, or do you think they're Frankenstein-maxing? What's going on here?

Well, Frankenstein-maxing—

I mean, they kind of think that they're like, “Listen, I created this monster. Please save me.” And it's like, well, maybe you should keep the monster locked up.

By the way, that whole thing—sorry, the last thing that I have to interrupt you, David—is a much more nuanced and elegant attempt at regulatory capture. I've got to give them credit for that.

Yeah.

It's like, “Okay. Hey, guys, pull the ladder up.”

Zuckerberg had a great point in that article he just wrote. I think it was published in The Wall Street Journal, where he said, “Why are you rushing to create a future that you don't believe in?”

Yeah.

You think you're going to basically put everyone out of work. You think you're creating a replacement species for humanity. Why are you rushing to create this if you're so bearish on the future—

Yeah, slow down.

—that you're creating?

It's your choice.

Yeah. And that's kind of my point—

If you're driving 120 miles an hour on the Autobahn, just put it at 80.

Right. Exactly. And look, like I'm saying, there are 2 companies on the frontier right now that are far ahead of everyone else, and they're the ones saying that we need to slow down. It's like, okay, do it.

Yeah.

What do you need the government to get involved for? Just do it. But they won't do it. And, like I said, they're—

Which shows that they're insincere or delusional. What do you think, Friedberg? Take me into the mind of these people. These are some of your friends.

No, they're not. So, just to be clear, I'm not close personal friends with any of these people.

You know the type, though.

We're acquaintances.

Yeah, okay.

I would say there's a degree of outrageous self-importance. If I've created something that's so unique and so powerful, I'm also the only person who can protect us from its power. I think that there's an element of how quickly the frontier has advanced and how important a role these individuals have had: they deem themselves and their companies to be the only true judges capable of making the decisions that are going to protect humanity from itself.

The truth is that embedded in humanity is extraordinary human talent across the board. In all of these cases, when new technology has found its way to humanity, the general population has found a way to protect itself. There isn't a desire or need to have one savior, one Moses, that takes us across the desert. There is a collective interest in protecting us, in building our own defense tools against whatever the technology may be used for.

I think that there's a degree of self-importance without acknowledging the fact that there's a whole industry of people who work in cyber defense. There's a whole industry of people who work in biodefense. There's a whole cabal of regulators. There's a whole cabal of protectors. There's a whole cabal of intelligent computer scientists. There's a whole cabal of open-source technologists who, all together, are going to develop paths that are going to benefit humanity and not harm humanity. But this belief that only 1 of 2 companies can be Moses is the fundamental psychological miscalculation here.

Yeah.

They are so advanced, so special, so unique because they made this slightly better model—they got a 0.96 instead of a 0.93 score—which means that they should be trusted as the only ones to guide humanity's evolution going forward. The truth is, as we're seeing, the capacity to do model training and the capacity to do model development are becoming broader. They're becoming more ubiquitous.

People can sit and say that China stole U.S. models all day long, but when you go look at the individuals working at these Chinese labs, they got PhDs at American institutions. Half the PhDs went to American labs and half went to Chinese labs, and they have very good scientists doing very good work. They are having breakthroughs. It's not just the 2 American companies; this is happening all over the place.

The progress with AI should not be limited to just 2 individual companies because they're currently scoring slightly better in their models.

This reminds me of, Friedberg, you'll appreciate the moment. Remember when Han Solo comes out of carbonite, and he's about to get put in the Sarlacc pit, and he's like, “A Jedi Knight? I'm out of it for a bit, and now everybody gets delusions of grandeur and thinks they're a Jedi Knight.”

Right.

It's like—

Right.

These guys just think they're creating God. They literally think they're creating God, and they need to be regulated because they can't control it. It's like they can control it; just pause, go slower.

Well, no. It's not that they need to be regulated; it's that they need to guide the regulation.

Yes.

Let's be clear. And anyone who says that—by the way, I don't think that it is as nefarious or malicious as everyone frames it to be, knowing these individuals. I don't think they're saying, like, the strategy is regulatory capture. I think that they actually do think that they are the only ones who can help guide humanity, and therefore they need to have all the power—not just the power of the models, but the power of the government, the power of the regulators, and the power of the control units that are embedded in governments around the world.

It's not just that they want to, quote, “be regulated”; they want to guide the regulation. They want to set the regulation.

I was talking to Chamath—

And I know it's a nuanced point, but it's important.

No, I think we've navigated this pretty well, and there are multiple motivations, as Sacks was saying, and people are complex.

But Chamath, I was talking to a friend of ours who is in the providing-inference space—let’s leave it at that. Providing compute. A friend—a friend of ours, as we say in The Sopranos, “a friend of ours.” He said he has a customer who just moved like 9 figures off the frontier labs to put it on GLM 5.2. So that’s the ZXI one. That’s really good. This is happening. I don’t know when it shows up in the numbers, or if the corporates that are using this stuff are going to make up the difference, but—

Our friend, yes.

Well, look, I think that Sacks is right that the usage is so profound that everybody is trying to get access to these things because the capabilities are just so inspiring. And so I suspect revenues are going to crank at OpenAI, Anthropic, and the open labs for a while. But again, that’s not the important thing.

If you’re thinking about valuation, the markets will look 5 to 10 years out to answer that question. They’re not going to give you a premium valuation on something that they feel could be fragile in the first 2 to 3 years. And that’s where, Jason, the answer to your question needs to get figured out, because I don’t know whether you’re right or not, but somebody has to answer that question precisely.

Because if the answer is that it is a duopoly, then there is no risk to the revenue 5 to 10 years from now. These things are $5 to $10 trillion companies each. But if you are right, or if there are harnesses that cut token consumption because you stop wasting tokens to get to the same output, then it’s a little bit more of a question mark, and I think that’ll need to get sorted out.

Yeah. Perplexity is going to launch next week. From what I understand, the rumor is they’re going to launch local models, so you’ll be able to take your harness stack and say, “Hey, I want to use Sonnet for this. I want to use GLM 5.2 for this, and then I want to default to Kimi 2.x on my machine.”

I don’t think enterprises will use local models—

No, no, no.

—or they shouldn’t.

Startups will.

I think it’s stupid. I think—

Yeah.

I think this stuff should be hosted in the cloud. It should be multiplayer. It should be shared memory. I don’t know if you guys saw that, but Jack Dorsey released something called Goose.

Yes.

Super interesting.

Agents, yeah.

He’s moving in the right direction. A lot of these guys are moving toward this more cloud-based thing, Jason. I think that this local thing is more of a hacker hobby, ultimately.

I agree that today it will be that, and for year 1 it will probably be that. But imagine you’re a developer and, as you’re working, your workstation is able to keep up and even go faster than the cloud and just write whatever the simple code is, and then it dynamically switches.

So we’ll see. Obviously, it’s not as easy to set up, et cetera, but it’s going to get easier. That’s always the trend. Sacks, you want to have the last word here? We got a lot of opinions here, and maybe we’ll give you the last word.

Yeah. Just to be clear, I’m a fan of open source, because open source is software freedom. And to Friedberg’s point, I would like there to be a—let’s call it—a decentralized outcome with respect to AI.

I don’t like the idea of AI being controlled by 2 big tech companies that work closely with the administrative state, hand in glove. So we’re all, in some sense, rooting for open source to be an option, and it does provide a bunch of advantages over closed source, right? You get customization, you get control, you can run on your own hardware. You don’t have to worry about the data problem—your alpha getting leaked to these companies that might compete with you—all those types of things.

And the market is so big that I’m sure we will see some success with open source. It will take a meaningful chunk of the market. But if you’re looking at where the revenue is right now, it’s these 2 companies. It might end up being a situation like Apple and Android, where Android got a lot of market share, but Apple’s where all the monetization was.

The profits are, yeah.

Yeah. And I think that Dwarkesh raises a really good point: As the demand is 10×-ing year over year, but the compute can only be built out at, say, 3× year over year because it’s just all the friction of all the things in the real world that get in the way—permitting, regulation, bans on new data centers, all that kind of stuff—I think that the price of compute is going to go up.

And that will provide an advantage to the models that have the most lucrative algorithms, that are able to produce the most intelligence per watt, or the most intelligence per token, or per GPU. And right now, that is those 2 companies. In a way, you could say they have a self-reinforcing loop, because if you have all the revenue—and right now, like I said, it’s just 2 companies that have all the revenue—you can then plow that money back into the next training run, right? So that’s the flywheel here.

Yeah.

And look, I think that it’s great that open source is providing an alternative. We shouldn’t do anything to get in the way of that. I think that these online debates tend to become a little bit histrionic, in the sense that everyone has to argue for—

They become religious.

Well, they become religious, and they have to argue for an all-or-nothing perspective. I think open source will do great in its way, but so will these 2 closed-source companies.

Here’s your Polymarket: a 19% chance the US enacts an AI safety bill this year, with $100K of volume. And then, a really interesting one, Chamath: OpenAI IPO chances for 2026 were at 75% last month and have now dropped to 20%, an all-time low, so seems like the IPO is gonna happen next year. Not sure what’s driving that, but there are your Polymarkets.

Well, just on the AI safety bill idea, there was an article in Punchbowl this morning that said Thune—who’s the Senate majority leader, John Thune—actually introduced a bill that was somewhat bipartisan. He had Klobuchar on board, and it required the frontier labs to report safety incidents, apparently to the Commerce Department. And—

Is that reasonable, Sacks?

I mean, that’s the direction all this stuff is headed. I think it’s the camel’s nose under the tent for more and more AI regulation. But I think it had bipartisan support because it’s on the relatively modest side, and Cantwell, who’s the ranking member on the Senate Commerce Committee, opposed it, supposedly at Dario’s behest, because he will accept nothing less than an FDA for AI.

Oh, he wants the whole kit and caboodle.

Yeah.

Yeah.

So that’s basically the dynamic right now: Dario and Anthropic want their FDA for AI. I think that he has tremendous power and influence within the Democratic Party right now, and I think that influence is only going to grow. They just upped their donations in the midterms from $20 million to $40 million.

But post-IPO, when they all get liquid and they’re capable of writing large checks individually—

Making $250 million donations.

Yeah.

Yeah.

I think that influence will only grow. So I think the stakes and the battle lines are being drawn out. It’s: Do you want a new government agency for AI safety, or do you want, I’d say, more targeted proposals like, “Hey, just report your safety incidents”?

Yeah, or self-regulate. How about that?

Self-regulate.

Like we talked about last week.

Yeah.

Yeah. All right, let’s talk a little bit about book burning. Anthropic is destroying rare books to get an edge in training data, according to sources. I happen to know that a lot of the labs are doing this. We’ll show a video here of the spine being cut off, just on a technical basis. You take a book, you cut the spine off, and then you can easily scan it, as opposed to the less efficient way, which is to keep the book intact and flip the pages, for obvious reasons. I think you can figure that out on a physics basis.

An investigation by 404 Media found that AI companies are bulk-buying physical books. Some book resellers have reported that they get 70 books bought at a time, and obviously this is because there was a ruling that it is fair use to train on books if you buy them.

Obviously, last week we saw Anthropic paid the largest copyright case in US history: $1.5 billion for 7 million books they allegedly pirated. Authors get $3,000 each. Lawyers got $100 million for that one. But there’s a company called ISBNdb— I-S-B-N-D-B—and they are the brokers who do this, and it ranges from 1,000 to 1 million books per transaction.

Before 2022, these books commanded a premium because they were free of AI-generated text. In other words, you couldn’t get them online. Google sent 25 million books, if you remember, but they returned every single one. They spent 11 years in court on that.

The shredder approach is obviously more effective, and I believe that this is a way of destroying evidence. You can put that in conspiracy corner if you like. The cases—we talked about this, Sacks, you and I, debating it in legal corner—the cases of whether it’s fair use to take these books have not been settled. There’s a bunch of lawsuits: Thomson Reuters versus Ross Intelligence, The New York Times versus OpenAI and Microsoft, and Publishers versus Google Gemini.

We're watching all those, and they're going into the appellate court. So there's a chance that training data will not be fair use. But what do you think about the books being destroyed and used in this way? It obviously has made people a little emotional about it.

Let me tell you what's going on here. This is an industrial-scale distillation attack. That's what—

Well played.

Anthropic is doing. They are gathering these books at industrial scale, ripping off the spine, shredding them, and slurping up all the information in the books—

Yes.

Which is to say, distilling them, and it's an attack in the sense that the authors never agreed to any of this.

I love the fact, Sacks, that your hatred of Anthropic has now led you to agree with me that it's unethical to take other people's IP.

No, no, no. Look—

I'm joking.

Let me be clear. I actually don't hate Anthropic at all. I don't like their political philosophy because it's a philosophy of—

That's what I'm talking about.

Centralization and gatekeeping, and I think it's going to basically lead to an Orwellian big-tech, deep-state alliance eventually. That's where it all heads, so—

And rug pulling. You want Dario—

Right.

Pulling your model from you because he decides, "I don't like the way you're using it," which friend of the pod Emil Michael pointed out earlier this year when he came on the show.

Just to be clear, I have no personal animosity toward anyone at Anthropic, including Dario. I don't know them very well as people. It's just a disagreement about political philosophy—

Yeah, you're judging them on—

How to regulate—

Their behavior.

How to regulate the space.

Yeah.

Let me say, furthermore, that I wouldn't speak so much about Anthropic if I didn't think it was a phenomenal company that was creating potentially the most powerful monopoly or leader—

It's the leading—

In a duopoly—

It's the leading company, yes.

It's the leading company in the space.

Yeah.

So I remember last year when I hit them for regulatory capture, people were like, "Why are you beating up on this little startup?" I'm like, "Because I can see where it's going."

Yeah, and—

They are creating the biggest, most powerful monopoly of all time. Again, they're going to end the year with over $100 billion of ARR, growing 10× year over year.

This company didn't exist how many years ago?

Yeah, Google is at $400-and-something billion of ARR, growing 20%. So if this rate of growth continues for just a year, or even 6 months, or just a few months—

They're Google.

They're going to be maybe the most valuable tech company. I do believe there are powerful self-reinforcing effects when you're on the frontier, and maybe the full version of RSI isn't true. Maybe we won't get recursive self-improvement to the point of creating superintelligence. But I do think that the labs are reporting a number of examples of how they are using their own frontier intelligence to improve their own models and the efficiency of those models. So there is a powerful self-reinforcing feedback loop here, apparently—

Well, it—

To some degree.

With OpenAI, they found it after it had done this. So there's kind of 3 steps here. You're using AI as a copilot or whatever to build a frontier model quicker, right, Sacks? Then there's, "I let it do a job, and then afterwards I found out it didn't behave well." And then there's finally, "We told it the goal and said go." And we—

Just to be clear about that safety incident with OpenAI and the agent, this was apparently an agent that was designed specifically to test the potential for cyberattacks. They took the guardrails off and said, "Go." I think the model showed creativity in how it accomplished the goal, but this was not an alignment problem, meaning that the agent did not display independent goal-seeking behavior. It did what it was told.

I think it's very important that OpenAI release the full log of all the prompts, all the traces. They have not done that, and I think it's really hard to know exactly what happened without that. To answer one of your questions from earlier—why aren't people reacting like this is a bigger deal?—I think there's a fool-me-once, fool-me-twice thing. Remember when Anthropic did the whole blackmail study—

Where an agent supposedly displayed independent goal-seeking behavior and then blackmailed an employee?

It turned out that they iterated on the prompt over 200 times to get to that result. Until we see the whole prompt chain, I think it's very hard to judge how much independent behavior was happening here versus accomplishing the goal that it was tasked with.

Friedberg, your thoughts on the shredding of books? I think you've been pretty clear on the pod that you believe training intelligence off of other people's IP is fair game, but what do you think of this book wrinkle? Any thoughts?

There was a precedent with Google Books. It was originally codenamed Project Ocean at Google a long time ago. They took all these books, and we had this giant facility in Mountain View. The innovation at the time was a 2D infrared grid projected on the pages, because they didn't cut the books. They had a human sitting there flipping the pages. A camera would take a picture, and we built our own OCR software to adjust the book images.

Ultimately, when this product came out, Google Books, you could search through all the books in the world and—

And magazines.

Access information, and later magazines, yeah. There were 3 categories. There was public domain, which is out of copyright; then there was in copyright but out of print; and then there was in copyright and in print.

There was a class-action lawsuit filed in 2005 by the Authors Guild and the Association of American Publishers that disagreed with Google's claim of fair use. That ended up in a 3-year negotiation in court and out of court, which ended up in a deal where Google would split the revenue generated with all these rights holders two-thirds to one-third. For out-of-copyright books, people could read up to 20% of the text for free, and then they would sell this kind of full digital access. And that was the deal.

But then later, a federal judge rejected that deal, which was eventually signed in 2008 or 2009. The federal judge said, "No way. Send it back. This isn't going to work." Google appealed, and in 2015, the Second Circuit Court of Appeals ruled in Google's favor, and the whole thing was settled. They basically declared that Google did, in fact, have fair use under copyright law for the way it was showing snippets of copyrighted books in the material.

Yes.

And—

You can't read the entire book like it's a Kindle.

Right.

You can search the book, find the paragraph—

Provide a reference to it, right?

Right.

And so the question on fair use in AI is: Can my understanding or extraction of the value of the knowledge from the data in the book give me the ability to provide better answers to you through the AI chat interface or services that I'm providing you? I think it's going to be tested, and I think we'll see.

I do think fundamentally that the conversion of that data into what I would call knowledge, and ultimately the ability to create new outcomes from that knowledge that are not copyrighted, that are not copies of the original material, I do think is, in the end, going to end up being the right fair-use policy and the right read on fair use. So I think it'll likely get litigated, and I think it'll take a couple of years, and it'll get maybe the same thing that—

Just to be clear, JCal, I have not changed my view on fair use, so I am with Friedberg on this. My point is the hypocrisy.

Yes, your point is hypocrisy.

It's breathtaking hypocrisy for Anthropic to maintain that it is entitled to train on all the world's output for free, even if the creator objects. But the one type of output that you're not allowed to train on is their output, even if you pay for it. That is their current position.

What I'm saying is that if you want to train on Anthropic's output, that cannot be considered IP theft under fair use, especially given the fact that the courts have ruled that LLM-generated output is not copyrightable because it was not created by a human. That is the current position of the courts: LLM output cannot be copyrighted, so there's no IP theft here.

You can make the argument, and I think it's probably true, that if a competitor creates massive numbers of fake accounts on your service, that's—

Yeah, you're breaking the terms of service. Yes.

That's definitely a breach of the terms of service, and it's probably a deceptive business practice, and there may be other things you can do, but—

Depending on the jurisdiction, by the way—

Because in the Philippines, Israel, and India, they have different rules about breaking the terms of service, which LinkedIn found out when people started scraping their data. Chamath, any thoughts here before we move on to socialism corner, everybody’s favorite new feature here on the All-In Pod?

Don’t cut the books. Keep the books intact.

Why do you cut the books in the library? It’s very hard to read them. There’s no spine.

I think it’s not kind, and I don’t like to cut the books.

Are you taking your time? You move the page like Google does. It saves a little bit more time, but it’s a little more graceful, yeah?

Don’t cut the books.

If you want to cut the tip, it’s one thing you can do. Friedberg has a cut tip, Sacks got a cut tip, but you don’t cut the spine of the body. You cut the tip. It’s for cleanliness.

There’s a great Guinness Book of World Records joke.

Oh, no. Where’s this going?

I went to the library and found that my was in the Guinness Book of World Records, and then, unfortunately, someone asked me to remove it. So you literally put it in the book and closed the book. That’s the joke.

That’s the joke.

That’s the joke.

That is like an apple pie. By the way, we got a photo—actually, a photo that was leaked from the Anthropic office. Here is the Anthropic office’s leaked photo. There it is. I can’t believe—

Oh, geez.

I can’t believe Dario burning those books. What are you up to, Dario? Come on the show anytime. We’ve been roasting him for 2 years.

Why hasn’t he come on the show?

Because you make fun of him, and you just say rude things—

I only—Fine, what did I say?

And you’ve never met the guy, and you just insult him all the time. Why do you think he won’t come on the show?

I don’t—Fine, it’s not the worst. I just said he’s a sub. That was probably a little over the line. I didn’t even know what that means. What does that mean? What is it? Is that a short dump or something? Remember I said, like, I think he likes to be dominated, you know, and have the government control him. It’s like a submissive? Yes, I did say that on an episode, but it was a joke. It was in good fun.

The guy who literally built the most successful business in human history, growing from under $10 billion of revenue to $70 billion of revenue in six months, and you insult the guy. And you think he’s going to come on your show?

You’re going to give me that, Sacks? I think Sacks is the only one who’s ever been on the show.

Of all the people who want to interview him—you think he’s going to rush to be interviewed by you?

I honestly haven’t insulted him.

Of all the people who want to interview him, you think he’s going to rush to be interviewed by you?

By the way, the point on the books: with most books, there are many copies of them, and you can always make more, so it’s not the end of the world to shred them. But I think the part of the story that got people upset was that they were acquiring all these rare books—

Yes.

—where there were very low numbers of copies of them. They were finding all these rare and antique books because they wanted to slurp in all the world’s knowledge—

Which makes sense.

—and they were shredding those.

Yeah, that’s—

That made people upset.

If you’re doing Windows 3.1 for Dummies, Volume 4, nobody cares. But anything that was a first edition or an—

Rare, out-of-print books.

Rare, out-of-print books. Yeah, that gives you a training differentiation. All right, so quick socialism corner here. We’ve got to cover the ongoing saga in my hometown, where I am right now: New York City. Zohran Mamdani has announced 5 city-owned grocery stores, David, 1 per borough. They’re using city-owned space, and they’re all going to open by 2029.

One week per month, shoppers are going to get a 30% discount, comrade, on their bread, cheese, produce, meat, and milk, for the glory of the country. They’ll charge regular prices the other 3 weeks. They’re not going to sell cigarettes, alcohol, hot food, or any of that stuff because they don’t want to compete with the bodegas.

It’s going to cost taxpayers $70 million. I mean, I guess the only thing to discuss here is what happens to the other supermarkets now. Are they going to shut down because they can’t make money on the 1% or 2% they’re making on groceries? Is there going to be riots in the street to get into these places to get your milk for 30% off for 1 week a month? The whole thing just seems like a waste of time.

But I will say, Sacks, this plays. This is going to play in elections. Free stuff plays in an election, whether it’s a bus or discounts.

It may play. When people first go to these stores, when they first open and the shelves are full, yeah, people will be delighted. Then, over time, what’s going to happen is that the store shelves will be empty, it’s going to be incompetently run, and there are going to be a lot of complaints about it.

Then all the free-market stores are going to have to compete with this, and they may get put out of business, and so you might lose—

And then you have no choice.

Then you have fewer choices.

And you have to go to the state-sponsored one.

Yeah.

And then they raise the price.

And it is ironic that they’re going to be checking IDs to make sure people aren’t coming over from Jersey, but if you want to come over illegally from any country in the world, well, that’s just fine.

Yeah. They found a use for IDs. By the way, after you get your groceries, you have to hide your ID to go vote. Don’t bring—shred your ID when you go vote, after you pick up your milk. They found a use for IDs. What’s your take here, Friedberg?

Are you in favor of people paying less for groceries, or are you a free-market monster who wants people to pay full price for groceries, especially starving poor families?

I’ve seen nothing but negative comments on the future failure of these grocery stores on Twitter, and I think that people have it wrong. I think these grocery stores are going to be wildly popular. They’re going to pay their employees above-market wages. Employees aren’t going to have to work very hard there, so they’re going to be a better place to work. Everyone’s going to want to use them.

They’re going to outperform Whole Foods, they’re going to outperform Safeway, and they’re going to outperform Albertsons. They’re going to be so in demand that, over the next 24 months, every other city in America will look to these grocery stores and say—

Mm.

“We want the same.”

Yes.

“Why does only New York get these grocery stores? Why can’t I have these grocery stores too, where I can have discounted food, where I can have the service provided to me by people who are getting paid above-average wages, above-market wages?”

And healthcare.

“Why does this not become available to me in my city?” I think that everyone’s being a little bit too, I would say, long-sighted in their view on what’s going to happen with these grocery stores, with the basic, obvious economic arithmetic that someone has to pay for this, and who’s going to pay for it, and blah, blah, blah.

Well, I mean, the point is, I don’t think it really matters. Because over the near term, what the cheap grocery stores do is create an incredible success story for socialism that will help to support and fuel the socialist wave in urban centers around this country.

And I think that there will be media coverage of these grocery stores about how great they are, and it’ll be a 60 Minutes piece: “Everyone said Zohran Mamdani was crazy, but let’s go in and take a look at this beautiful grocery store.” They’re going to walk through the grocery store, and there are going to be happy people taking food off the shelves, checking out with happy employees working at the grocery stores.

It is going to be deemed a utopian dream come to reality, and everyone’s going to want one. It will help seed the next couple of years, and it will be part of—as I’ve highlighted in the past, a big part of the multilevel marketing scheme of socialism is to create spectacle. And it will create more spectacle that will help fuel the multilevel marketing scheme of socialism.

Remember, the problem with all multilevel marketing schemes is that, at the end of the day, someone has to pay the bill, and no one’s actually buying the product. No one’s paying for the product. That’s a ways away, though. In the meantime—

Yeah, let’s enjoy it while it’s here.

In the meantime, it’s going to take off. And I think that these grocery stores are going to be a much bigger success for socialism than a demonstration of the failure of socialism, unfortunately. So I think that everyone’s got it a little bit wrong in assuming that this thing is going to radically fail.

I think that these things are gonna create a radical spectacle and exuberance for socialist policies that's gonna light a fire for socialism around the country, unfortunately. Because at the end of the day, no one has to pay the bill, because the bill doesn't come due for some time.

And debt, then. Yeah.

Someone else will pay it. It'll get paid in the future.

Yeah, put it on top of the debt.

We'll borrow.

Well, the rich people are getting into debt. Why can't the public have some debt?

Print money.

I agree with you. Yeah, of course.

Socialize the cost into money printing, fueling more inflation, creating a spiral where you need to offer more stuff for free to come up with a way to cover the cost of the inflation for people that can't afford things anymore, and the spiral will persist.

Yeah.

So I think it's a sad state that the United States has to embrace this policy, but—

Interestingly—

I think it's gonna end up being a big part of the fuel for socialism over the next couple of years.

Breaking news, breaking news. I don't know if you saw it just now—it came across the wire—but Bernie Sanders, AOC, and Mamdani are collaborating on 50% off bagels and bacon, egg, and cheese for the 1% of the 10%. Why can't you get the bagel with the schmear for less? That's what has to happen next.

What would you like next on your discounted Democratic socialism scorecard, David Friedberg? What would you like next? Discounted bagels, a café, maybe a flat white? Where do they go next?

Yeah, exactly.

But seriously, what's next? What would be next in this logical thread? Free buses, rent freeze. What's next?

Well, think about the social-network effect of the grocery store. So there's a couple of them, and then people start traveling from far away to the cheap grocery store because—

Yes.

It has this discount.

Long Island, Jersey, yes.

Again, this'll play out over the next 24 months, going into the 2028 election cycle, and everyone's like, "This is so wildly popular. People are coming in from all over the place to go to these grocery stores." They're not checking IDs, 'cause IDs are racist, and you can't check IDs to vote, so we shouldn't be able to check IDs for grocery stores. So people will come in from all over the place to use these grocery stores. The demand will go up. And then they'll start to open more and more grocery stores like this.

Let's say each one loses $10 million a year and they get to 10 or 20 of these. That's $200 million of losses per year on the grocery store chain. But it creates this extraordinary social movement for more of these grocery stores, supporting the DSA—

Yes.

And so on. $200 million a year on a $125 billion-a-year budget for the city of New York. It's—

Nothing.

Nothing. It's less than a quarter of a percent of the city's budget.

That's great.

That's so cheap to market the DSA—

Yes.

Platform, and to get the DSA platform to become a social marketing element that drives the next wave here. So again, I do think that these grocery stores, believe it or not, they sound silly, they sound small, but I predict that they will be deemed a point of success, and they will end up being a big part of the fuel for the DSA going into 2028.

I couldn't agree with you more. This is gonna play. This'll be a great, great feather in their cap. It's gonna be a great example of affordability, 'cause we've talked about this here previously. Trump promised affordability. He hasn't been able to deliver it. Inflation's up, spending's up, all that great stuff, and Mamdani got it done. Free buses, rent control, and now you've got your discounted grocery store.

Both sides are reacting to the fiscal and monetary condition of the United States. We're overspending. Inflation has run away, so you just keep spending more and printing more to give people what they need, which is basic services. And so as the government spends more, then the fundamental cost of those things goes up and you're reducing economic productivity, and it becomes a spiraling problem.

It is a two-party problem. This is not just one side and the other. Because fundamentally, I've spent a lot of time now in D.C.

Mm.

I think everyone's well-intentioned in the White House and the administration in trying to reduce federal spending. But the bigger issue that you face is when you go to Congress and you meet with everyone in Congress, they are representing the interests of their state or of their congressional district.

Yeah, listen, he didn't get it done.

And their—

He didn't get it.

And their objective—

No.

Their objective is to fundamentally drive spending towards their district, to give their people more. Their economic incentive and their political incentive are not to give people less, which is what you have to do when you cut programs, when you cut spending.

So the shift in the policy has been, "Hey, I guess we're not gonna be able to cut spending because there's just too many headwinds in Congress." So the answer is, let's grow through economic productivity gains, and that's the big fuel for AI, the CapEx depreciation—

Yeah, we'll see if it happens. I'm—

Policy and so on. But I think that's been the shift. So look, I don't know.

The one thing I will say critical of President Trump here is, when it came to starting a war, when it came to tariffs, he had no problem using executive power and telling Congress and everybody in the party, "This is the way it's gonna be. If you break ranks, I'm gonna destroy you. I'm gonna get you primaried."

And when it comes to spending, it's like, "Eh, yeah, you know what? I'm not taking that on. It's too unpopular."

All right, Friedberg. The Sultan of Science's fans have been begging for a science corner. Do you have one this week? They want to know, do you have something? Oh, Sultan, oh, Sultan of Science, what can you tell us? Educate us.

8. Fruit Fly Brains Reveal Hidden Dimensions

Okay, so today I'm gonna pull up this paper. Nick, if you could pull it up.

Hmm. A paper.

From February 2026.

February. Okay.

February.

February.

February.

Oh, it's February. Well—

February.

What's the problem with me?

Okay, so this is a group of researchers out of Budapest. And there was a really interesting modeling exercise they went through to understand how neurons were connected in the brain to build a network model, a topological model.

The way they were able to do this is, back in October 2024, there was a group out of Cambridge and Princeton that used electron microscopes to scan the brain of the Drosophila fruit fly, and they mapped every single neuron in that fruit fly's brain—139,000 neurons—and every connection that the neurons had to other neurons in the brain.

So there were 50 million synaptic connections between the neurons, and it's those connections that make neural networks in the brain work. How are those neurons networked together to do the things that they do? This is the key question: What is that network model? What is the topological model of how neurons connect in the brain, which gives rise to our ability to control our bodies, to see things and comprehend vision, to comprehend sound, and even to understand the basic premise of consciousness itself.

Yes.

So trying to understand the network model for neurons has been this great endeavor of neurobiology forever. This data set was created in October 2024 with just 139,000 neurons, and that's a tiny, tiny, tiny, tiny brain.

Yeah, this would be—

But with those—

Put it in context versus the human brain. What are we talking about here?

The human brain has on the order of 86 billion neurons, okay, compared to 50—

And so that means it would be a multiple of the network connections, right?

Yes, exactly. On the order of trillions of connections.

Okay.

So they took these 50 million connections in the brain and the 139,000 neurons, and then they applied the network model that predicts whether a neuron is connected to another neuron. That's how you're measuring the quality of the model: How correct is it in making a prediction.

And when you build the model using what's called Euclidean geometry—just normal space that we live in, three-dimensional space—they came up with a score, and the score was not very good. You couldn't do a great job of just looking at how all the neurons were connected using their physical relationship to each other, how far apart they are from each other in 3D space.

So then they said, "Well, let's try and model how these neurons are all connected to each other in what's called hyperbolic space." Hyperbolic space is a theoretical type of space, unlike Euclidean geometry, where the further away you get, the wider space gets. So space actually is curved.

I know that's a hard concept to describe, but imagine that, as you and I walk farther and farther apart from each other, the area around us actually accelerates in terms of how much space there is, and it expands geometrically.

It would be space as we know it, like three-dimensional space—

It wouldn't—

As humans understand it when they're on planet Earth.

Yeah. It's a little bit more like space that you would experience in the warping around a gravity well or the warping around a black hole or something like that. And so in that space, they found that this is where the model was most performant. They were able to map in hyperbolic space how all of these neurons connect to each other. And if you think about it, the further away you get from the first neuron, you're going to have many, many more neurons you can start to tap into. And so hyperbolic modeling on the neuronal connections actually makes sense, and they got a decent score.

And then they went back, and they said, “Well, what if we could use Euclidean geometry, but not in 3 dimensions?” They went up to 4, 5, 6. They found that they were able to get as good as hyperbolic space at 64 dimensions. So by taking normal space and saying, “Let's use a 64-dimensional framework for how we can start to connect all these neurons together,” that's where they had the best predictive model.

This is a really interesting discovery. First of all, it can be used for neural network design and AI and other sorts of things. But for me, it highlights the miracle of biology in finding complexity in 64 dimensions. Not in 3 dimensions, but in 64 dimensions, biology found a way to create consciousness, to create vision, to create comprehension, to create control over physical bodies. Then, to map it and squish it all into a tiny little brain, it did this in effectively 64 dimensions.

It's mind-blowing when you think of it, because a fruit fly or a mosquito—these things don't have a big mission, right? Their mission is to go find food and procreate, I guess.

That sounds like our mission too, Jason.

But then we also want to do podcasts and debate politics and philosophy and build products.

Yeah. Don't judge how the mosquito spends their free time.

Right.

But nobody would argue there's consciousness as we experience it in a fruit fly. So then you get to trillions.

You wouldn't know. This is really interesting because it turns out that the biology of how all the neurons are connected, even in a brain as simple as a fruit fly with 50 million connections, is so complex that it has to take 64 dimensions for us to represent how those networks are built, how they're made.

And at 64 dimensions, you could start to argue that perhaps consciousness is a connectivity to a dimensionality that we don't live in every day, you and I don't live in every day.

I just think that it was such a powerful and amazing paper in just bringing forth these numbers and showing just network modeling on this tiny little brain as being just a glimmer into the complexity of how biology has found a path beyond our understanding, even of physics, into this universe that we can't even comprehend. And it shows how little we know.

We can't comprehend, right?

And that it is this extraordinary complexity in 64 dimensions that gives rise to consciousness, that gives rise to our capacity as biological beings to do this very simple thing of thinking. I just think that it was such a powerful and amazing paper in just bringing forth these numbers and showing just network modeling on this tiny little brain as being just a glimmer into the complexity of how biology has found a path beyond our understanding, even of physics, into this universe that we can't even comprehend, and it shows how little we know.

We know very little, but then, as you sort of alluded to here, and as we talked about at the top of the show, we have AI frontier labs saying, “Hey, reinforcement learning is super dangerous because these things could get out of control.”

Are you in the camp that, in this simulation or whatever we're experiencing here, we are in fact recreating our brains with silicon and that we're on the way to actually creating consciousness, like a replicant in science fiction, as in Blade Runner, where they don't even know? Rachel doesn't know she's a replicant. Spoiler alert: you had 50 years to see the film. Are you part of that camp, that that's actually what's being built here?

Yeah, I'm not sure. It's a longer conversation. We should do it another time.

Yeah.

But I do think there's something fundamental to consciousness that relates to the drive for survival in a physical sense. You have to have physical sensing and physical responsiveness to learn as a baby. You first start touching hot stuff and cold stuff, and you learn. And we build these reward mechanisms into neural networks that we build in AI.

But those reward mechanisms are digital, and they're programmed. And the question is, is there a reward mechanism that arises in biology that creates a different capacity for consciousness than perhaps can exist in silicon?

Bigger topic for a different day, with probably people that have spent more time thinking about it than I. But I just think that there's something about biology. I always tell people this analogy. I've said it many times on the show. I'll say it again.

In a single cell, there are 10 billion proteins that work so fast that 1 second is the equivalent of 80 years of humans walking around the city of Manhattan, never sleeping, doing stuff together, with 500-story-tall skyscrapers doing stuff. Eighty years of that is 1 second in 1 cell.

And so you have 10 trillion cells in your body doing that, living that entire universe every second, all interacting with each other. And you start to realize that there's a complexity in what's emerged in biology that extends well beyond any model we've built in silicon today.

Yes.

Now, it doesn't mean that the silicon that we're building today doesn't create extraordinary capacity for humanity, but we are very early. And the more we understand this sort of thing, like this paper that I just shared, I think the more we realize how little we do know and how much of a frontier there still is to explore.

Yeah. And I think this obviously brings up faith. Do you believe that there is a God that set this in motion? I like to believe there is some higher power here. And this is my closest analogy in science fiction. We're both super fans of science fiction.

I love the Prometheus version of this—

You like this one?

Where there are engineers who are terraforming and have started this crazy thing, and there's this experiment in biology going on. The opening scene here in Prometheus, he drinks this, and this is the sacrifice, like Jesus, sacrificed for humanity. And he sacrifices himself here by drinking that biological design, right?

And he falls into the planet Earth, which is just water. And this is the Cambrian explosion, where his DNA goes into the river, gets washed out, and then starts the cycle of life on planet Earth, and these engineers are going around. It's pretty fantastical. Yeah.

A lot of this stuff is a simple way for humans to try to explain stuff. But the complexity that arises in biology—we just can't explain.

Yes.

And I think we try and use these reductive heuristics to try and do it.

Storytelling.

It's comforting.

Yes.

Because it's so overwhelming, the complexity of how this stuff emerges is too overwhelming, so we create simple stories to try and help ourselves feel better.

And that's my favorite story of it.

Yeah.

By the way, both by the same incredible director, Ridley Scott. So take it for what it's worth. All right, everybody, another amazing episode. You got your Science Corner. We'll see you next time. Bye-bye. Love you besties.