E174: Inflation stays hot, AI disclosure bill, Drone warfare, defense startups & more

00:00:00.000 | All right, everybody, welcome back to the All In Pod, episode 174, almost up to 175. I believe

00:00:06.560 | there are a lot of folks getting together next week at the All In meetups around the world to

00:00:10.960 | watch the show. Unfortunately, J. Cal had an oral incident this week. He will not be joining us.

00:00:18.480 | I guess he cracked a tooth and couldn't make it. So he is sitting this one out,

00:00:25.600 | very last minute decision. Specifically, cracked a tooth, eating a bison rib at the Salt Lake in

00:00:30.960 | Austin, which is a true story. Freebird, I think we found an image of J. Cal at the Salt Lake

00:00:36.720 | cracking his tooth on a rib. Let your winners ride. Rain Man, David Sattler.

00:00:49.360 | And instead, we open sourced it to the fans and they've just gone crazy.

00:00:57.520 | All right, so we are going to announce the All In Summit 2024 back to LA.

00:01:03.280 | We had such a great time in LA. It was a great location for everyone. We really appreciated

00:01:08.560 | the facilities at UCLA. This year, we're upgrading the experience once again for everyone.

00:01:14.080 | So the event's going to be September 8th through 10th in LA. If you haven't seen it,

00:01:18.480 | please watch the recap video from last year. We put it up on YouTube.

00:01:21.920 | And then there's all the interviews from last year are also on YouTube. We're going to have

00:01:26.880 | another amazing lineup of speakers this year and have a much more kind of upgraded, elevated

00:01:31.680 | experience for everyone. Applications are open today. So go to summit.allinpodcast.co

00:01:38.000 | to put an application in to join us this year. As you know, we had way more applications last

00:01:43.120 | year than spots. And this year, we're going to have all sorts of upgrades. We're going to have

00:01:47.520 | transportation to and from all the events for everyone, a concierge booking service to help you

00:01:51.520 | book your rooms and get you set up in LA for the trip. We're doing a full day event on the Sunday,

00:01:57.680 | September 8th, followed by two days of content, Dine About Town on the first night, two blowout

00:02:02.800 | parties. It's going to be awesome. So we're really excited about it. We've got a much bigger budget

00:02:07.360 | this year. And we're going to try and really make the experience awesome, while also delivering the

00:02:12.800 | great content. So really excited to do this again in LA. I appreciate everyone's support of that.

00:02:18.400 | We had a lot of debate internally about where to take it. And we settled on going with what we know

00:02:25.520 | and trying to make it a little bit better this year. So anything else you guys want to say about

00:02:29.600 | All In Summit?

00:02:30.400 | And there's only one kind of ticket this year, right?

00:02:32.960 | That's right. So we're not doing like regular ticket and VIP. We're just doing one ticket

00:02:37.200 | for everyone. And everyone's going to get this kind of elevated experience. Should be cool.

00:02:41.920 | Saks, just show up. Would be awesome. Hope to see you there.

00:02:45.760 | I want to invite a couple of speakers.

00:02:47.200 | Yeah, you've got some good speakers, I saw.

00:02:49.600 | Yeah.

00:02:49.760 | It's gonna be great.

00:02:50.560 | When are we going to start announcing the speakers, Preberg?

00:02:53.040 | We will do that in a couple weeks. We already have a bunch booked. And it's gonna be awesome.

00:02:59.680 | Oh, that's really exciting. That's awesome.

00:03:01.840 | Yeah, I know CP, you wanted to take it somewhere else. I mean,

00:03:04.800 | we were in debate on Vegas, maybe going to some other country.

00:03:09.840 | But you went to another country. Where'd you go this week?

00:03:12.080 | I was in Paris, France.

00:03:13.360 | What were you doing there?

00:03:15.200 | There's like a important AI conference. And Jonathan Ross, the founder of Grok and I kicked

00:03:22.560 | it off with like a fireside chat for like 45 minutes at the beginning just to get everybody

00:03:27.040 | amped up and excited.

00:03:28.720 | Cool.

00:03:31.120 | Yeah, it was really good.

00:03:31.840 | Anything interesting come out of it?

00:03:35.360 | I think the soundbite that made an impact on me was from Jonathan. So I think he did a very good

00:03:42.160 | job of explaining where NVIDIA is really strong, which is really in learning, and where NVIDIA is

00:03:50.320 | trying to build the business, which is an inference. But the problem is the features

00:03:55.680 | almost compete with each other between learning and inference. And I think that that was good

00:04:00.960 | clarification. And then the second thing is he said that if he deploys his roadmap, he'll have

00:04:05.600 | 50% of the available inference compute by the end of next year.

00:04:09.680 | Wow.

00:04:10.880 | Based on all the needs of like open AI and meta and everybody else, which is a...

00:04:14.720 | Wait, but that's a smaller physical footprint. Is that right? And it's just

00:04:17.920 | because they can do higher token per second?

00:04:19.680 | Less power, higher tokens per second across all the major models. So it would be meaningfully

00:04:27.120 | less power and meaningfully cheaper. So that was a pretty amazing soundbite. But his whole

00:04:33.600 | explanation was actually really good. So I mean, if you listen to the spaces we did,

00:04:37.120 | I thought it was good. I thought this was even better. So I'm trying to get a copy of the

00:04:41.120 | talk because it was taped so that we can kind of post it. At a minimum, anybody who's interested

00:04:48.880 | in studying NVIDIA should listen to what he has to say, because I think the nuances between us

00:04:54.320 | and them are really profound. And they were pretty stark the way that he described them.

00:04:57.600 | So it was cool. It was great to be in Paris, frankly.

00:04:59.760 | How does NVIDIA respond to this movement or this notion that there is an entirely

00:05:04.720 | different chip category focused purely on inference? Do they have inference focused

00:05:08.960 | chips in development or a different roadmap that they're going to have to kick up now?

00:05:12.800 | I mean, I think it's very hard for them. It is a structural innovator's dilemma. And why is that?

00:05:18.240 | Because they've made some architectural decisions around things like HBM,

00:05:21.920 | which is high bandwidth memory, or specific kinds of extremely high throughput optical cables.

00:05:27.840 | And why these things matter is that they've gone and bought up the world supply,

00:05:31.920 | and they've basically bearhugged these markets as a structural part of their design. Now,

00:05:38.400 | that makes sense, and it's very legal. The problem is that if you architect around it and go through

00:05:44.880 | a totally orthogonal design process, we're not subject to any of those supply chain considerations.

00:05:51.040 | And I think he explained that really well. I didn't understand it in nearly as much detail

00:05:55.360 | as I did leaving. So I think that's an important takeaway. It's also really important for all the

00:05:59.760 | hundreds of chip startups and for these venture investors who are ripping money into these

00:06:03.760 | companies to know these differences. Because if you are investing in a company that is subject to

00:06:09.840 | HBM in their design decisions, or these specific optical cables, or a whole bunch of other things,

00:06:15.040 | you're going to be in a really difficult spot because NVIDIA has bought them all.

00:06:20.320 | So I thought that that was really interesting. The other thing that NVIDIA does brilliantly is

00:06:24.240 | what Intel did in the '80s and '90s when they wanted to choke the market for CPU competition,

00:06:31.120 | which is that they were able to define a metric that everybody started to pay attention to,

00:06:36.400 | which for them was clock speed. And if you remember, Intel would push Moore's Law,

00:06:41.040 | and they would push transistor density and clock speed as the measurement of value.

00:06:47.200 | Now, for a lot of consumers, none of us knew any better. And so every time we saw a new chip,

00:06:54.080 | we would go through the Pentium upgrade cycle because we thought that's what we were supposed

00:06:57.680 | to do. It turns out that clock speed doesn't really factor into speed the way that you think

00:07:03.440 | about it and experience it as an end user using software. And it's the same in AI. And so we

00:07:10.160 | talked about that as well. So it's an incredibly important company, NVIDIA is. I think they're

00:07:15.520 | performing brilliantly. They're running many of the same plays that Intel did to try to create

00:07:20.720 | total dominance. But there's some very important nuances around power and around supply chain

00:07:28.000 | decisions and technology decisions that, frankly, some people would think are unsustainable.

00:07:34.000 | And the more I understand them, I would be in that camp as well. So I thought it was a very

00:07:39.440 | important conversation and hopefully we can post it. Yeah, it's cool. And by the way, it's great

00:07:45.520 | to be in Paris, too. What do you think of Paris? You know, I have a new appreciation for what's

00:07:49.920 | going on in these major cities. I think that all major cities are crumbling. I don't and I don't

00:07:54.960 | mean to be alarmist or very dramatic about that. It's just that I spent a lot of time with

00:07:59.840 | entrepreneurs that are in Paris. And they struggle with the same things that the people in San

00:08:05.840 | Francisco do. And then I had a couple of conversations with some folks that are entrepreneurs

00:08:10.000 | in London who were there. And they suffer from the same things. Lots of petty crime, lots of garbage,

00:08:17.040 | lots of vandalism, lots of drugs. And so no place is perfect. It turns out everybody is struggling

00:08:23.760 | from a lot of experimentation gone wrong and how these cities run. That was that was actually my

00:08:29.280 | the biggest takeaway outside of the actual AI conference itself. But that's a Western City

00:08:34.640 | problem. You're not I mean, have you visited many cities in China or other parts of Asia?

00:08:39.040 | And is that like a Western problem? I have not been to China in six years. I've been to Hong Kong,

00:08:46.720 | but I've not been into China for six years. So I don't know. I would say that India has its own

00:08:53.440 | vagaries. So those issues are not the same. I've been to Singapore a couple times. My kids were

00:08:59.920 | just in Japan for spring break. So I saw a bunch of the pictures, those cities look relatively

00:09:04.240 | the same, very orderly, well run clean. But yeah, you're right. These Western cities are a bit

00:09:10.320 | chaotic. Yeah, sacks. Is this like a Western social liberal manifestation? How cities are

00:09:16.960 | being challenged? Generally? Yes. Yes. Like Malay has, I think Malay said this, right? He said,

00:09:21.920 | like, all these Western social liberal beliefs are having lifestyle impact and urban centers.

00:09:29.040 | Well, it's hard for me to speak to what's happening in Europe, but we know what's happening

00:09:32.400 | in the United States. All the biggest cities are blue, and they're run by a certain ideology. And

00:09:39.440 | in those cities, we've had now a decade plus of decriminalization, deprosecution, decarceration.

00:09:46.960 | And the impacts of this are obvious. We also, at least in San Francisco, we have subsidies

00:09:53.760 | for people who want to live on the streets and do drugs all day. I mean, it's not prohibited to do

00:09:59.920 | that. So all of our public spaces have been taken over by people who are happy to take the whatever

00:10:04.960 | it is, $700, $800 subsidy and do drugs all day. We also had this referendum in California that

00:10:12.640 | essentially lowered a whole bunch of felonies to misdemeanors. Basically, if you steal less than

00:10:18.960 | $950, that's considered a misdemeanor now instead of a felony. And so there's been a huge increase

00:10:25.840 | in crime because of that. You know, shoplifting just doesn't get prosecuted anymore. And

00:10:30.400 | misdemeanors don't get prosecuted anymore. We've had all these SOARs, DAs who don't want to

00:10:33.520 | prosecute misdemeanors. So I think if voters had understood that lowering a lot of these crimes

00:10:41.920 | from felonies to misdemeanors meant that they wouldn't get prosecuted at all, I don't think

00:10:45.200 | they would have done it. But that's kind of where we are. But are we seeing a rebound from that

00:10:48.800 | now, Sax? Is democracy working? Because I know that in San Francisco, we had an election recently,

00:10:53.680 | and in the ballots, it seemed to indicate that voters were sick and fed up with the living

00:11:01.120 | situation in the city, and they're putting different people in office and have an intention

00:11:06.400 | of changing the city charter and changing the laws in the city so that we can reverse this trend.

00:11:13.200 | Is that not where things are? I mean, like, sitting where you sit, like, don't you see

00:11:18.320 | San Francisco kind of trying to right itself? And as a result, doesn't democracy work over time?

00:11:24.000 | Yeah, it's just that we live in a one-party state, and that one party has been captured by ideologues,

00:11:30.000 | and so it's very hard to change. It is true that in San Francisco, there was an election recently,

00:11:35.040 | and the more radical supervisors were defeated. We have, I think, finally a good DA in San Francisco.

00:11:43.520 | But these radicals have had a long time to entrench their power, and the party machinery

00:11:48.960 | helps keep them entrenched. So I think you can argue that we've bottomed out because the public

00:11:54.080 | has recognized the problem, but it's still very hard to fix, and it's going to take a while.

00:11:57.840 | I think for the first time in a long time, I'm more optimistic than not, given the people I've

00:12:03.040 | seen that tolerated the conditions historically have reversed course over the last year and have

00:12:08.880 | said, "We've got to act, and we've got to do our kind of civic process."

00:12:13.120 | Well, time will tell. You have to remember, when New York City was a chaotic mess in the '70s and

00:12:18.160 | '80s, it took two different mayors, and it took almost a decade for that to get cleaned up, but it

00:12:23.600 | took the strength of leadership. First, David Dinkins set the framework, and then Giuliani really

00:12:29.280 | executed Rudy Giuliani. And between the two of them, and I think what people call the broken

00:12:34.480 | windows theory, they hired a lot more policemen, and they just really cleaned up all of these really

00:12:40.240 | bad areas, and then it had this knock-on effect. The problem is, when I'm staying at a hotel in

00:12:47.600 | Paris, or when I was speaking to some of my friends who live in London, it's the same thing.

00:12:52.400 | It's like the amount of knife crime is so egregious, as an example, that they tell you,

00:12:58.000 | "Don't wear a watch," et cetera. I mean, I don't wear watches anymore when I go out.

00:13:02.320 | But what's crazy is they will tell you that you'll get stabbed over a phone,

00:13:06.000 | where it's like, "You don't need to stab me for my iPhone here. Just take the iPhone."

00:13:10.800 | So there's a level of lawlessness and petty crime that's affecting tourists

00:13:14.960 | that, to me, was very surprising. And so even when I was walking around Paris, I was like,

00:13:21.600 | "Should I really feel scared here?" And my reaction was, "No, I've never felt scared in

00:13:28.720 | Paris." But then what's ringing in the back of my ears are these warnings from people that have

00:13:34.240 | lived there for 10 and 20 years, who've had this stuff happen to them. That was really surprising

00:13:38.480 | to me. So that stuff in San Francisco, as far as I can tell, or any major city in the West,

00:13:42.320 | hasn't really been fixed, because you haven't had a shift in leadership yet, where they've

00:13:46.560 | really said, "Okay, we're going to double down on policemen and policing, and we're going to

00:13:52.480 | amp up the punishment." That hasn't happened. Well, I was just in Vegas this week,

00:13:57.040 | and it doesn't feel... How was your trip? Give us the trip report.

00:13:59.600 | Yeah, I went to the Google Next Conference. But I will say, first off, no crime in Vegas.

00:14:04.480 | Vegas is Vegas. It's a very different environment. But yeah, I went to the Google Next Conference.

00:14:11.280 | I ended up going out because at O'Halo, we did a partnership with GCP. So we're moving all of our

00:14:18.000 | infrastructure and doing a lot of model development on GCP. And it was great. I think there was 30,000

00:14:23.920 | people at this event, which was insane. Have you guys been to any of these, like AWS Reinvent, or

00:14:28.720 | the Google Next, or any of these?

00:14:31.520 | No, I have not.

00:14:33.120 | What I didn't realize prior to going to this was how much of an ecosystem there is. They

00:14:38.080 | take over the whole convention center at Mandalay Bay. And there's hundreds of companies that have

00:14:44.240 | built services or build services on GCP that are there, pitching CIOs and CTOs of other companies.

00:14:50.480 | And so it's this real network effect type of business model that I hadn't fully appreciated

00:14:55.280 | just how like, deep the network is and how entrenched it is. I also got to go to this

00:15:00.960 | thing where I spent a lot of time with CIOs and CTOs at other companies, big companies.

00:15:05.120 | And a big takeaway for me that I also didn't realize was how most of them are not on a single

00:15:11.680 | cloud provider. Almost everyone I spoke with is on multiple clouds. So they're all on like AWS

00:15:19.600 | and Azure and GCP. And they use different aspects of each cloud. And there's a lot of interoperability

00:15:27.200 | between the clouds now, which I had previously my naive concept of this, because I'm obviously not

00:15:32.160 | deeply in the space was that a company picks a cloud and they move everything over to that one

00:15:37.200 | service provider. But it turns out that they split, and then they allocate dollars based on

00:15:42.240 | the services and the pricing. So it was pretty clear that everyone is, it's like a race to the

00:15:48.640 | bottom on pricing for basic commoditizing services in the cloud, like data storage, right? And then

00:15:55.520 | whoever has the better higher level services, whoever has access to better models or has a

00:15:59.440 | lower price on running models, or on running things like inference. That's where the money

00:16:05.280 | is made. And the gross margin on that is quite different. And I think that's where a company

00:16:09.760 | like Google probably has a much bigger advantage because of the uniqueness and the quality of the

00:16:16.320 | models that they've developed. But yeah, I just didn't I didn't really appreciate like how much

00:16:20.240 | there was also. So my other big takeaway was the cloud is a wide open competitive market right now.

00:16:25.680 | Like there is no AWS lock in in this market. There's no Azure lock in. There's no GCP lock

00:16:30.720 | in like you guys. Yeah. The lock in comes from the developer services that they provide. So

00:16:36.240 | both Amazon and Google, and I'm assuming Azure as well, will give you access to incredibly talented

00:16:42.400 | engineers who will help you build production code. Yeah. And the ecosystem on how so like a lot of

00:16:47.760 | the clouds are competing for the consultants who actually do the work for the big enterprises.

00:16:54.400 | So Capgemini, Deloitte, PwC, they have these massive booths and facilities, and then go and

00:17:00.480 | build tools for companies is also a key part of the business strategy. And yeah, I mean,

00:17:05.600 | it's such a competitive market. I did not appreciate just like how much of an open game

00:17:11.360 | that still was. Sure. It was a great, great couple days in Vegas. And I got to see some

00:17:16.320 | of our friends. We ducked in and saw Helmuth in the Poker Go studio for a minute. Some big

00:17:21.920 | tournaments going on at the ARIA. It was fun. We had a good time. Okay, let's move on. So,

00:17:26.160 | you know, on the docket today, we were going to talk about CPI. The March CPI numbers came in

00:17:31.680 | from the Bureau of Labor Statistics yesterday, higher than forecast at 3.5% year over year.

00:17:38.080 | This is up from 3.2% year over year inflation reported by the Bureau of Labor Statistics in

00:17:43.920 | February, higher than expected for three straight months now. Let's pull up this chart. And as we

00:17:48.880 | all know, the Fed did 11 rate hikes in 2022 and 23. And the intention was that we would see

00:17:54.960 | inflation come down to their target of 2% and they would start to lower rates. And obviously,

00:18:00.560 | we're not seeing that inflation is remaining hot. Things are getting more and more expensive.

00:18:07.440 | I pulled together some of the numbers on year over year price increase. Housing is up 5.7%

00:18:14.400 | year over year. Transportation costs are up 10.7% year over year. And the highest car insurance is

00:18:21.600 | up 22.2% year over year. So the expectation has been that the market will do rate cuts.

00:18:29.360 | Larry Summers came out and said you have to take seriously the possibility that the next rate move

00:18:33.360 | will be upwards rather than downwards. I guess, Saks, let me let me ask you first for your

00:18:40.160 | commentary on how this plays into the election cycle. I know Biden was expecting a rate cut or

00:18:46.000 | several rate cuts going into the election, which usually helps the case of the incumbent. Maybe you

00:18:49.760 | can comment on where the election cycle might be influenced by this inflation report.

00:18:56.160 | Well, I think it's definitely bad news for Biden. I was expecting a Fed put this year.

00:19:00.480 | And really, the whole market was. The market was expecting three rate cuts this year,

00:19:04.160 | which would have been really good news for Biden. And I think this latest inflation

00:19:09.280 | print was kind of a dagger to the heart of that expectation. And I think the Wall Street Journal

00:19:14.480 | put it best. It said here, Wednesday's report had been hotly anticipated because Fed leaders

00:19:21.280 | had been willing to play down stronger than anticipated inflation readings in January and

00:19:27.200 | February as reflecting potential seasonal quirks. But a third straight month of above expectations

00:19:34.720 | inflation data erodes that story and could lead Fed officials to postpone anticipated rate cuts

00:19:41.040 | until July or later. So it's not just the fact that this inflation print was higher than expected.

00:19:46.400 | It was also higher than expected in January and February. But people were willing to kind of

00:19:51.040 | overlook that, saying, well, maybe it was just kind of a quirky reading. But now we've had three

00:19:56.640 | straight months, it's pretty clear that the narrative that we had going into this year,

00:20:00.880 | which was that inflation was on its way down, that it peaked at 9% as the official rate, I think,

00:20:07.600 | in 2022, and then it was going down every month through all of 2023. And I think the expectation

00:20:14.480 | going into '24 was it would keep going down and we'd get these rate cuts. Well,

00:20:18.000 | after three straight months of inflation being more persistent and stickier than people expected,

00:20:24.480 | I think that narrative is basically dead. So I think the new narrative now is it's going to be

00:20:30.800 | rates are going to be higher longer. And I think Larry Summers is reflecting that view. Larry's

00:20:36.080 | going further. He's not just saying that rates are going to be, you know, at the current rates

00:20:40.560 | for longer, he's saying they might actually increase. And that's a risk. Yeah, last summer,

00:20:45.760 | he said, and this was last summer. So almost a year ago, he was saying, the market's got it

00:20:50.000 | totally wrong. We're actually going to need much higher rates than the market's anticipating for

00:20:53.840 | much longer, as well, than the market is anticipating. And so did Mario Draghi,

00:20:58.400 | they both said the same thing. No one paid attention to him, everyone ignored it and

00:21:02.160 | assumed that this was going to be a quick rebound to normalization. It's clear that,

00:21:06.400 | you know, once again, Larry has proven himself to be fairly prescient.

00:21:09.360 | Larry Summers has, for the most part, been spot on from this on this whole inflation

00:21:15.760 | question going all the way back to 2021. Remember, 100% in Q1 of 2021, Biden's first quarter in

00:21:22.240 | office, the big legislative push was for the $2 trillion COVID relief bill is the so called

00:21:28.640 | American Rescue Plan. And Larry Summers said that it was it risked inflation, it was an inflationary

00:21:34.160 | bill. It was unnecessary. Yeah, we didn't need it. The economy was already coming back, and we

00:21:38.560 | didn't need it. And Biden was risking inflation. But of course, inflation was only at 2% at that

00:21:43.600 | point. So the administration kind of poopooed Larry and said, you know, this is Larry Summers being

00:21:49.520 | Larry or whatever. And sure enough, we were at 5% inflation by that summer. And you have to wonder

00:21:56.240 | if Biden had listened to that advice. Would he be in a different position right now going into the

00:22:01.920 | election? I think probably he would have been. And you got to wonder for what I mean, what did

00:22:06.160 | that $2 trillion accomplish? I mean, COVID was winding down. I mean, these were the last bit of

00:22:12.800 | stimmy checks and payments to these pharma companies. And I mean, it was basically a grab

00:22:18.480 | bag and it was passed on straight party lines. And it was just totally unnecessary. I mean,

00:22:23.280 | the economy certainly didn't need it. And so here we are. And I think that a lot of people,

00:22:29.280 | including the markets thought that Biden was kind of out of the woods that this year,

00:22:35.280 | we'd see the final leg of inflation going back to normal, but that's not going to happen.

00:22:40.400 | And I think, you know, going beyond the political ramifications, I think there could be

00:22:44.640 | several other knock-on effects that we should talk about. I mean, one is for the consumer.

00:22:50.640 | This means that the cost of borrowing is going to be higher for longer. That makes it harder

00:22:56.320 | to buy a house. Mortgage payments are higher. That also means if there's fewer home sale

00:23:01.920 | transactions, that means the price of housing could come down. So there could be a correction

00:23:06.480 | in that market. Second, if you want to buy a car, your car payments higher. And if you have

00:23:11.040 | loans, your personal interest is going to be higher. And this is why I think consumers feel

00:23:16.240 | like they're worse off than the economic data would otherwise reflect. And Larry Summers actually

00:23:21.440 | had a tweet storm about this about a month ago, where he calculated that if you included the cost

00:23:26.320 | of borrowing in inflation, that inflation was much higher than people thought and that it actually

00:23:32.320 | peaked not at 9%, but at 18%. So Larry had an excellent tweet storm on that. And he said that

00:23:38.400 | the cost of borrowing, which used to be calculated in inflation but is not anymore, was the reason

00:23:44.400 | why consumer sentiment about the economy was depressed. He said that that accounted for about

00:23:49.600 | 70% of it. So people should be feeling better off because GDP growth is good and unemployment is

00:23:56.960 | low, but they're not because inflation is so high. And if you include cost of borrowing,

00:24:01.440 | the inflation is even higher. So I think this is really bad news for Biden. But it's a story

00:24:07.120 | that's been going on now for three years. It's not just this one inflation print.

00:24:11.920 | Right. It's funny. Yesterday, Biden said, during a press conference with Japanese Prime Minister

00:24:18.960 | Kishida, that he's sticking with his prediction of a rate cut before years end, but there might

00:24:24.240 | be a delay, which I think is obviously, you know, based on what the data showing and what

00:24:30.480 | folks that seem to know and have been pretty good predictors are saying is unlikely.

00:24:35.040 | We were supposed to get one in April. We were supposed to get one in June.

00:24:38.560 | Oh, yeah. Remember, the Fed Open Market Committee in December

00:24:41.520 | stated that they expected three rate cuts this year. And I think the market, in some cases,

00:24:46.640 | were predicting as high as four or five. Right.

00:24:49.520 | And now we may not have any and we may actually see a rate hike if you were to follow.

00:24:54.480 | Well, if we see a rate hike before the election, I think Biden is toast.

00:24:58.800 | Yeah. But I mean, I think Summers gave that maybe

00:25:01.040 | a 15 to 20% chance, which is it's still not the most likely scenario.

00:25:05.600 | But it's possible. But it's possible. And no one was counting on that in December.

00:25:10.000 | Chamath, if you're a Fed governor, and obviously, the Federal Reserve Board of Governors sets the

00:25:16.640 | rates, they meet and they make these decisions. How influenced are they politically? How influenced

00:25:22.880 | are they by this election cycle? And, you know, historically, this is meant to be a fairly

00:25:26.960 | independent board. But there's been a lot of consternation over the last few years that this

00:25:32.560 | board acts like almost a political apparatus, particularly in election years.

00:25:36.560 | You mean when they're not plagiarizing?

00:25:39.200 | Well, tell us what you're talking about.

00:25:42.400 | It turns out that before I answer your question, that was a joke. Nick, you can please show the

00:25:51.760 | tweet. But there was an article that came out that said one of the Fed governors has apparently

00:25:57.200 | massively plagiarized most of the work that got her to the position of being Fed governor.

00:26:03.280 | Yeah.

00:26:03.840 | So the same thing that happened to the president of Harvard. So TBD, what happens to that Fed

00:26:08.000 | governor? But it seems, you know, my observation there was just more that it's like the preferred

00:26:12.880 | method of defenestration now of academics and the apparatchik, right? Before, if you had

00:26:19.440 | right leaning views or centrist views, you get run out of town and get fired. Now, all of those folks

00:26:25.040 | are using the search engines to basically figure out really quickly that you plagiarized a large

00:26:31.120 | quantity of your academic work, and that your scholarship is is a little bit more speculative,

00:26:38.160 | which then brings into question, why do you get the right to help set the policy of the most

00:26:43.040 | important economy in the world? Well, anyways, so that that was that joke. But so I think that

00:26:47.680 | you're asking the absolute right question. I think that the Fed has become increasingly politicized.

00:26:55.280 | And I do think that there are actions that foreign governments take to influence the election,

00:27:04.240 | but I do not think it's what most people think. When I say that most people's ideas goes to like

00:27:10.080 | some darkroom conspiracy theory and misinformation. And I think that's a little bit naive.

00:27:16.640 | I think the more obvious way in which other governments can impact the US election is

00:27:23.040 | exactly how SAC said, which is they know that if inflation ticks high, and interest rates are up,

00:27:28.880 | and consumers are under pressure and meaningfully displeased with the economy, they'll vote out the

00:27:32.880 | sitting administration. And so what can other governments do? They can do what they're doing.

00:27:39.040 | So Nick, I love this like, explanation and pre charts, Nick chart number one, please.

00:27:45.120 | What is this? So this is a core driver of inflation, which is the price of energy.

00:27:51.760 | And what you see here is that since the beginning of the year, oil prices have gone up,

00:27:58.800 | which ultimately will translate to into a higher cost of doing business, right? Running your

00:28:03.680 | factory, keeping the lights on, keeping your home warm, getting from point A to point B,

00:28:09.760 | all of these things go up, which ultimately get reflected in prices, which is what inflation is.

00:28:14.400 | And so that's a very bad trend. Obviously, we had a lot of really good stuff happen

00:28:19.520 | through the course of last year, but it's largely reversed itself. So it's important to double click

00:28:25.360 | into this and figure out, well, who's driving this? Is it just a random act? Or is it a systemic set

00:28:31.600 | of decisions? And it turns out it's systemic. So if you go to the next chart, what this shows you

00:28:37.040 | is what the US has been doing. So the United States has clearly created an incentive to flood the

00:28:44.160 | economy with oil, which the laws of supply and demand would say prices should go down.

00:28:50.800 | And by bringing prices further down, they would have further brought down inflation,

00:28:56.080 | and they would have pulled forward these rate cuts that we had been expecting from the Fed.

00:29:00.720 | The problem is that the rest of the world actually said, no, hold on. And that's what you can see on

00:29:06.160 | the next chart. So for every time we were trying to increase production, all of these other countries

00:29:12.240 | have been cutting production. And this is the change in the OPEC plus crude production targets

00:29:17.280 | of 2024. So what you see here is for every barrel of oil, the US would pump out of the ground,

00:29:23.920 | more than that one barrel of oil would get actually restricted by all of these countries

00:29:27.840 | on this list, which is just to show you that these countries have a very specific view

00:29:32.080 | about how they want that energy market as and that's just one market, but it's an input a key

00:29:38.160 | input into inflation and rates to work. So you're saying that the OPEC has a price per barrel price

00:29:45.680 | target, and they're able to manipulate production to maintain that price target. So while the US is

00:29:50.160 | trying to lower the price per barrel, these other countries are are able to have a stronger

00:29:54.560 | influence and balance out to get when you when you make these cuts like this prices will go up

00:29:59.520 | because OPEC plus is meaningfully bigger, the sum of all the parts of OPEC plus is meaningfully

00:30:04.320 | bigger than just the United States on its own. And so the United States may amp up a few 100,000

00:30:10.080 | barrels a day. But if OPEC cuts by a few million barrels a day, you see what you saw in that first

00:30:16.640 | chart, which is the price will go up. And that's largely why prices have gone up. So I think why

00:30:21.440 | this happens is because when they see the United States, trying to manage inflation and manage

00:30:27.920 | global rates, right, because us rates are not just for the United States, they're the, they're the

00:30:32.240 | trendsetter for everything else. I think that in part, there's a decision that this rate philosophy,

00:30:37.760 | the economic philosophy, the political philosophy, if they wanted to support that, what they would do

00:30:43.760 | is actually also keep prices where they were or increase production. By reigning in production,

00:30:49.120 | what you're essentially doing is voting against that entire apparatus. Right? So OPEC, and all of

00:30:55.120 | those countries, and those decisions are not made by an oil minister, let's be honest, they're made

00:30:58.320 | by the leaders of those countries, whose decision then gets reflected by the minister of petroleum

00:31:04.080 | or oil in all of these countries, is a vote against Biden, and the United States government

00:31:09.520 | and the politicization potentially of the Fed. And then the last comment is just that when you

00:31:13.680 | look at the reverse repo rate, which is a fundamental measure of liquidity, what you

00:31:17.840 | also see is liquidity very quickly getting sucked out of the system, which is also generally a

00:31:23.840 | system health check that the Fed uses to make sure that the system is operating as required. So

00:31:28.480 | in totality, what I see is that there is a structural disillusionment with the current

00:31:38.000 | administration, and where people can make decisions in their own best interests,

00:31:44.480 | versus the collective interests of the United States in a broader framework that the US may

00:31:49.280 | represent are voting against it. And so this is why the idea of a persistent inflation rate

00:31:55.520 | is a lot more credible than it was six months ago. And I think, sorry, just the last point,

00:32:01.920 | and I think it's just what Zack says, which means that we're probably now on balance 50/50 between

00:32:07.040 | a hike and a cut. And I think if you don't see this thing change in the next three months,

00:32:11.120 | you're going to see 75/25 for a hike of at least 25 basis points. And I do think David's right. I

00:32:18.080 | don't see how the sitting administration in any government will be able to withstand the onslaught

00:32:24.720 | of electoral displeasure, if rates are four and a half, five and 6% going into November.

00:32:32.000 | Timoth, if that is, I think everything you said makes sense going into this

00:32:37.040 | election rates are high and and or going higher. What is the Biden administration do?

00:32:43.840 | The fact is, they are very likely going to come up with a strategy to try and win votes that will

00:32:51.280 | likely involve some sort of stimulatory process would be my guess. And that stimulatory process

00:32:57.120 | may look something like more student loan forgiveness, more debt forgiveness, Kamala

00:33:01.760 | Harris. She had a whole speech about debt. And here's here's the irony, the actions that the

00:33:08.480 | administration is likely to take to support themselves in the election cycle are likely

00:33:14.640 | going to further fuel inflation. And that's the biggest concern and worry I have is that there's

00:33:20.080 | almost this inevitable set of behaviors that are coming down the pipe here over the next few months

00:33:24.480 | because of the inflation report that are actually going to make things worse, not better.

00:33:28.720 | I mean, to your point, Biden just announced a new student loan forgiveness plan. Remember,

00:33:32.560 | the Supreme Court invalidated his last one. So now he's trying again. And it would potentially

00:33:37.440 | give loan forgiveness to something like 30 million borrowers. So yeah, they're going to try and buy

00:33:42.080 | votes. But I think it's already very late in the game to be trying to do that. I think that's what

00:33:46.640 | they've been doing over the past few years. I mean, Biden has been a huge spender. You know,

00:33:51.760 | he passed the COVID relief bill, he passed the Inflation Reduction Act, which was a ridiculous

00:33:56.800 | name for was it 750 billion of, of more spending. And there was the CHIPS Act, which was well,

00:34:04.880 | yeah, and there's the CHIPS Act and the infrastructure bill. And then remember,

00:34:08.400 | he wanted four and a half trillion of build back better on top of that. And he wanted even more,

00:34:13.040 | he didn't get any event. I think that this is a classic case of Biden spent the last three years

00:34:19.280 | making his bed and now he's gonna have to sleep in it. And if in fact, inflation is persistent,

00:34:25.520 | sticky and not going away, and he doesn't get rate cuts, again, this may cost him the election.

00:34:29.600 | But, but look, I want to move beyond the political aspect of this and just talk about some more of

00:34:33.680 | these knock on effects, because we talked about the impact on the consumer. There's also an impact

00:34:39.680 | on investors, there's an impact on the government, and there's an impact on on the banks. So let's

00:34:46.240 | just deal with the banks for a second. We've talked about on the show how commercial real

00:34:50.720 | estate developers are hanging on by their fingernails right now. You know, they, they

00:34:55.840 | basically borrowed huge amounts of money to buy commercial real estate, when rates were really low,

00:35:01.920 | and they did so at high valuations. Now there's been a huge correction on that market, and they

00:35:05.680 | cannot pay back those loans. And they can't get refinance. So what's happening? Well, they're

00:35:11.280 | working with the banks to restructure those loans to tack on the interest they can't pay now,

00:35:16.560 | to the end of the loan. And then the bank will hope and pretend that they'll be able to get

00:35:20.960 | paid back one day. It's called extend and pretend. Well, the extended pretend strategy could work if

00:35:26.720 | you believe that rate cuts are coming really soon, because you just need to hold on until the rate

00:35:31.280 | cuts are here, then you can get refinanced at lower rates. But if rates are going to be higher

00:35:35.840 | longer, that strategy is not going to work. And you're going to see more and more real estate

00:35:39.360 | sponsors throwing in the towel. And banks are going to end up foreclosing on these properties,

00:35:44.240 | and you're going to see more and more fire sales. There's a really dramatic example just the other

00:35:48.160 | day in St. Louis, a building that was bought a few years ago for $200 million just fire

00:35:53.280 | sailed at $2 million. I mean, 99% reduction. This is like dot com crash level reduction.

00:35:58.800 | That's for the equity, right? Yeah.

00:36:00.320 | No, this is like the price of the deal. I mean, I assume the equity was completely wiped out. Yeah,

00:36:06.080 | I guess they sold it for three and a half million.

00:36:07.920 | Are you sure that's not the equity? That's got to be the equity. There's no way that's the-

00:36:11.200 | No, it's just the price of the building.

00:36:13.680 | Okay.

00:36:14.560 | I don't know. I mean-

00:36:15.120 | With the debt. It's got to have the debt attached to it for that, yeah.

00:36:17.520 | I mean, either way, it's a total wipe out for the equity holders.

00:36:20.240 | Yeah, yeah.

00:36:21.040 | But it's also going to be a wipe out for the bank that made that loan originally,

00:36:25.200 | or that bought that loan when it was repackaged into CMBS or whatever it was.

00:36:29.520 | So I would expect that if rates stay higher longer, that's going to create tremendous

00:36:35.040 | stress on commercial real estate, and therefore on the regional banking system that made all these

00:36:40.080 | loans to commercial real estate developers. So that's knock-on effect number one. Knock-on effect

00:36:46.000 | number two is the government's cost of borrowing. The 10-year rate's already risen to 4.5% now.

00:36:51.280 | And if rates stay higher longer, as more and more of the government debt rolls,

00:36:55.600 | it's going to end up getting refinanced at higher rates. And the need to refinance

00:36:59.440 | all of this government debt is going to continue to put pressure on the bond markets.

00:37:02.800 | And I think this is why Larry is fundamentally pretty bearish on rates going down is the

00:37:08.240 | government's financing costs are so huge, and they continue to grow. We're adding something

00:37:13.040 | like a trillion dollars of new debt every 100 days. There've been some reports of this.

00:37:17.680 | Yeah, 7.6 trillion is getting refinanced of old debt in the next 12 months.

00:37:23.120 | Right.

00:37:24.160 | And that 7.6 is sitting around 2%, and now it's going to get bumped up to close to 5%,

00:37:31.520 | which adds an incremental $210 billion a year of debt service expense just on the debt that's

00:37:39.040 | getting refinanced in the next 12 months, not to mention the deficit, not to mention,

00:37:43.200 | et cetera, et cetera, et cetera.

00:37:44.640 | Right. So the government was extending and pretending as well. I mean, they were hoping

00:37:49.040 | that rates would go down fast enough that when all of this massive government debt rolls and

00:37:54.240 | needs to be refinanced, it'd be done at lower rates. So the interest costs wouldn't swamp us.

00:37:58.880 | We're already, I think, at over a trillion dollars now of annual interest expense for

00:38:03.200 | the federal government.

00:38:03.760 | We are. Yeah, and it's going to rise to 1.6 trillion. So it's going to be 60%

00:38:08.880 | greater than our defense budget, just servicing the old debt.

00:38:12.000 | We're basically in a classic debt spiral where we're borrowing. Our annual deficit

00:38:16.480 | is we're at 2 trillion plus now. So we're not paying back the principal on the debt.

00:38:22.240 | We're not even paying back the interest on the debt. We're borrowing money to pay off

00:38:26.880 | all the interest and then some on top of that. So that's a huge problem for the federal government.

00:38:32.400 | And then I think the last piece of this is investors. If you pull up this Fred chart

00:38:37.680 | on the Fed funds rate over the very, very long term. So what this chart shows,

00:38:43.840 | if you just zoom in on this, is the Fed funds rate basically over the last, I don't know,

00:38:49.120 | like 75 years. And you can see that it peaked in the early 1980s when Volcker decided to

00:38:56.560 | break the back of inflation by jacking up interest rates to like 18%. And it was very painful. We had

00:39:02.960 | a pretty severe recession and I think it was '81 or '82. But then the economy came roaring back by

00:39:09.520 | '83 and Reagan got elected in '84. In any event, that began a period of declining and low interest

00:39:16.720 | rates for investors, again, starting in the mid '80s. And it continued all the way through,

00:39:24.640 | I mean, especially through I think the post-GFC period and then this COVID period where we had

00:39:30.480 | zero interest rates for a long period of time. And you can see the last spike we've had was when the

00:39:35.440 | Fed decided to jack up rates to 5% in response to inflation. So the fact that we've got this

00:39:40.880 | persistent sticky inflation suggests that this whole, you could call it really almost a 40-year

00:39:47.600 | period of declining and low interest rates may be over. And when interest rates are going down,

00:39:53.040 | it creates a huge tailwind to the stock market and the bond market because the discount rate is lower,

00:40:00.880 | so stocks and bonds are going to be more valuable. Well, if we're in a long-term environment that's

00:40:06.960 | more inflationary and rates have to stay higher longer, that's not great news for investors.

00:40:12.720 | It hurts the stock market and it creates in the private investing world like VC and real estate,

00:40:18.960 | it creates a much higher hurdle rate. So it's harder for private investors to justify

00:40:25.360 | fundraising. So I think it's too early to say that this 40-year trend is over, but I think

00:40:32.080 | it could be. And the reason for that is just the huge pressure of government borrowing,

00:40:38.960 | crowding out private investing and this sort of ever-present need to keep issuing more and more

00:40:46.320 | debt at higher and higher rates to make it more attractive is going to create, I think,

00:40:50.400 | very bad conditions for investors. So Chamath, let me ask you if you concur,

00:40:56.720 | given that today where we're sitting as of recording right now, towards the end of market

00:41:03.200 | on Thursday, April 11, the NASDAQ is sitting at an all-time high, and the NASDAQ is up 1.6% today.

00:41:10.560 | Following this inflation report yesterday, what's the disconnect in the market on why

00:41:16.560 | investors are buying equities, given that we're likely going to have persistent higher rates,

00:41:23.520 | and as a result, multiples should compress, particularly in growth stocks,

00:41:27.200 | and why the NASDAQ is kind of firing up? I think it's important to look at what

00:41:34.240 | stocks are actually lifting the NASDAQ, just like what stocks are lifting the S&P 500. And

00:41:39.840 | we've seen this. There's a massive dispersion. There's like seven or eight companies that matter

00:41:43.840 | and 500-odd companies that don't matter much. And so the reason why things go up in an environment

00:41:50.560 | like that is, irrespective of the economy, money managers are paid to deploy money.

00:41:55.120 | And I don't know if you guys saw this, but Nick, maybe you can find it. I linked to it in my

00:41:59.840 | newsletter last Sunday, which was an article about the fact that the global amount of total

00:42:05.520 | available equities to buy is shrinking, right? So if you think about that, what that means is that

00:42:12.000 | the number of dollars is growing, right? Linearly, and in some years, it seems exponentially,

00:42:18.480 | but then the end, so then a certain percentage of that finds its way into the equity markets.

00:42:24.560 | But when the equity markets continue to shrink, and there's fewer and fewer things to buy,

00:42:28.160 | prices will still go up, irrespective of the underlying justification.

00:42:33.520 | So I think that the thing is, you can't look at these things as a sign per se. You know,

00:42:38.240 | if you look at commodities as a different example, Nick, I sent you this.

00:42:41.840 | Commodities are up, broadly speaking, 30% since the beginning of the year.

00:42:47.440 | I think the reality is the following. Countries are feeling increasingly uncomfortable

00:42:58.480 | with the decisions that the United States are making, whether that's its position to keep

00:43:04.080 | funding the war in Ukraine, whether that it's the continued instability in the Middle East,

00:43:08.640 | whether it's posturing with China, whatever it is. And so every country is now making a very

00:43:14.560 | different calculation about what they want to do. And I think how it gets reflected is that

00:43:18.320 | they're optimizing for their own economies. And what that yields is each of them trying to make

00:43:23.920 | as much money as they can in the short term. And as a country that is a net importer, we are going

00:43:30.560 | to feel the price of those decisions as inflation. And I think Sachs is right. Like, that's where we

00:43:35.760 | have to start figuring out how we rebalance these decisions from a foreign policy and economic

00:43:42.000 | perspective. Because if you don't change something here, all of these countries are basically going

00:43:45.920 | to be like, "We're on our own. We're going to optimize for ourselves." These countries are

00:43:50.480 | basically voting in a way that says we are not being led in a way that we trust. And that

00:43:55.520 | unfortunately falls at the feet of the sitting president. Yeah. Well, look, I'm hopeful,

00:44:01.120 | as we all are, that work in the tech industry, that new tools and AI and other technologies would

00:44:09.600 | have a massive driver on productivity, which net-net could help even this out. That's the one

00:44:15.600 | last great hope. Sitting here, AI obviously viewed as being highly disruptive by non-technologists

00:44:23.120 | and deeply concerning, and particularly in DC. But what is that thing that Bill Gates says,

00:44:28.800 | things are overestimated in the short term and deeply underestimated in the long term? I think

00:44:33.680 | that that's what AI is. That's right. And so AI is a bunch of, no offense, probabilities and

00:44:39.760 | statistics today. Okay? And so we talked about this last time, so we should debunk this whole

00:44:45.200 | thing of like, there's some glowing brain somewhere that's capable of doing everything.

00:44:49.440 | Doesn't mean that we shouldn't underestimate it, being capable of this in five or 10 years,

00:44:56.240 | which is probably the window in which we have some form of AGI, but today it's not that. And so the

00:45:01.360 | great leaps in productivity, Freeberg, that you're talking about are probably a three to five year

00:45:05.360 | phenomenon, not a three to five months. So what SAC says is right, which is how does the sitting

00:45:11.200 | administration buy votes today? Stimulatory methods.

00:45:14.480 | It'll have to make a ton of promises. The problem is it can't really work its way into the economy

00:45:19.360 | fast enough come the November election. And so that's where I think they're in a really tough

00:45:25.040 | spot, which means that then, and I hate to be the bearer of bad news, but this is where you have a

00:45:31.280 | lot of these wag the dog kind of scenarios that come to mind, because then the only way to really

00:45:37.040 | galvanize a populace is to distract them. Yeah. Well, look, let's talk a little bit

00:45:41.920 | more about AI, because I do think that regardless of the near term and longer term implications,

00:45:49.840 | there is a peaking of fear, peaking of interest, both different spelling of the word peak.

00:45:56.480 | Adam Schiff yesterday proposed a bill, sorry, earlier this week called the Generative AI

00:46:01.600 | Copyright Disclosure Act, in response to, I think, over 100 musicians signing a letter

00:46:08.960 | saying they're concerned that their copyrighted works are being used to train AI models.

00:46:13.920 | And Nick, if you'll pull up the text from the bill that I pulled out here real quick,

00:46:18.960 | we can take just a quick look. It's a very short bill. It's only five pages long.

00:46:24.000 | And it defines a generative AI model that's a combination of computer code and numerical

00:46:29.600 | values designed to use artificial intelligence to generate outputs in the form of text, images,

00:46:35.760 | audio, or video, and that it substantially incorporates one or more models as a system,

00:46:42.320 | and it's designed for use by consumers. And what the bill then asks is that anyone developing

00:46:48.000 | these models has to submit to a register run by the federal government, a list of all of the data

00:46:55.840 | that they use that might be copyrighted to train the model and that this becomes a precursor to

00:47:02.160 | being able to legally develop and use AI models that you actually have to register the training

00:47:07.200 | data. So I have a rant on this, which I'm going to preserve for the end given my rights as moderator

00:47:13.360 | here today. But I'd love the reaction from Sachs first on Schiff's proposed bill to moth. And we'll

00:47:19.280 | run this one around the horn pretty quick. Well, first of all, I think it's too soon for this

00:47:23.840 | legislation. I mean, the first thing that needs to happen is we need to get the question of fair

00:47:28.480 | use arbitrated by the courts. We've talked about this before. People publish lots of information

00:47:34.640 | on the internet. And I guess technically, it's copyrighted, but it's been published

00:47:39.600 | in a completely public way. And humans are obviously allowed to read it and use it to

00:47:44.880 | generate their own work. I think there's a question about whether AIs are allowed to

00:47:49.120 | learn from it as well. And I would argue that they should be to some degree under fair use.

00:47:54.400 | Maybe if you're talking about copyrighted songs, I don't know if that's different. Anyway,

00:47:59.120 | these questions need to be arbitrated. And once we know that, then we'll know whether there's a

00:48:04.720 | need for a rights clearinghouse of some kind, which may not be a bad idea. I actually got pitched

00:48:09.760 | by a founder recently to create a rights marketplace with copyright holders on the

00:48:19.040 | one side or rights holders could be musicians, could be movie studios, could be anyone who owns

00:48:24.960 | content, and then AI companies on the other that want to license that content for

00:48:29.440 | training purposes. And I actually thought it was a pretty good idea. So my point is just that

00:48:35.760 | there are entrepreneurs working on solving this problem as well. And I guess what I'm saying is,

00:48:40.320 | I'd like to see fair use figured out. And I'd like to see what the private solutions are going to be

00:48:45.520 | before we try to legislate this problem away from really a point of ignorance, because we're so

00:48:52.320 | early in the development of this market. Now, I think one of the questions you have to ask here

00:48:57.040 | is, why did Adam Schiff introduce this bill? And who is Adam Schiff? Adam Schiff is a powerful

00:49:03.040 | California congressman who is the candidate for US Senate right now. And he's expected to win.

00:49:09.840 | He's ahead of Steve Garvey, who's the Republican candidate by like 30 points.

00:49:13.040 | Schiff is, in addition to being a vicious partisan who's known for for Russiagate,

00:49:21.840 | he also is a prodigious fundraiser. And when he's not in front of the cameras, he's very effective

00:49:28.080 | in closed doors appealing to the key special interests in California that matter. And my

00:49:34.400 | guess is that he's not carrying water here for the 100 musicians who signed that letter.

00:49:38.480 | He's carrying water for the big studios, the big Hollywood studios, including music labels,

00:49:44.160 | but also the movie studios. And he's an LA based congressman. Yeah, I mean, that's who he's

00:49:49.520 | representing. Now, at the same time, my guess is that he's run this by the big tech companies,

00:49:54.800 | and he's probably got their blessing to some degree, because like Bill Gurley always points out,

00:49:59.200 | they like rules like this, because of regulatory capture, they can comply with them. But the little

00:50:04.480 | guy, the entrepreneur who's trying to create something new has a much harder time

00:50:08.480 | complying with these regulations. So this is basically a big special interest group

00:50:13.920 | sort of deal that I think he's he's putting forward.

00:50:17.840 | Do you think this becomes law?

00:50:19.440 | Not yet, but it's paving the way for something.

00:50:22.560 | And it's more like his promotional effort during his campaign effectively.

00:50:27.200 | Well, I mean,

00:50:27.920 | his move it legislatively, but

00:50:32.160 | yeah, I mean, the big LA powers that be like this, and they love this. Yeah, yeah.

00:50:38.320 | Any reaction to this bill?

00:50:43.360 | I think there was an article in The Verge that I posted.

00:50:48.080 | It turns out that OpenAI transcribed over a million hours of YouTube videos to train GPT-4.

00:50:54.000 | My takeaway when I read this was, well, if Google's not going to sue OpenAI,

00:51:00.480 | then this is a moot point. Because you're talking about one of the most valuable companies in the

00:51:08.240 | world with one of the most valuable sources of training data in the world. And so if that if

00:51:14.480 | it's true that there was a copyright violation, and Google just lets us go, what's anybody else

00:51:20.400 | supposed to do? What chance does these group of singer songwriters have? And you know, no offense

00:51:25.760 | to Adam Schiff, but he's not going to be able to do anything. So if this multi trillion dollar

00:51:32.080 | company doesn't draw a line in the sand, then none of this matters. And everybody's going to

00:51:36.640 | train on everything. I think, like, my big takeaway on all this is that there's a real

00:51:44.080 | deep misunderstanding on what these models really are, and what they're doing. The assumption,

00:51:57.120 | I think, naively, is that they take in a bunch of data, and then they spit that data back out.

00:52:06.080 | And in some cases, maybe slightly transformed. But the truth is that the models don't actually

00:52:11.680 | hold the data that they're trained on. They develop a bunch of synthesized predictors,

00:52:19.040 | that they learn what the prediction could or should be from the data. And this is similar

00:52:25.440 | to how humans operate. Can you pull up the article, Nick that I posted? This article is from last

00:52:32.720 | May. And you guys may remember this case. I can't remember if we talked about it on the show.

00:52:37.120 | But Ed Sheeran was sued by the estate of the writer of the Marvin Gaye song, Let's Get It On,

00:52:46.240 | for infringement in his song Thinking Out Loud. And he actually prevailed in court,

00:52:53.040 | where he went through with the jury, how he comes up with how he runs his creative process,

00:52:58.640 | how he came up with the song, and how he does his work. And as you know, Ed Sheeran,

00:53:04.560 | and nearly every musician or every artist listens to other people's music, looks at other people's

00:53:10.640 | art, and they synthesize that data, they synthesize that knowledge and that experience

00:53:15.280 | into their own creative process to output what in their mind is a novel creation. And I think that

00:53:24.240 | AI works in a similar vein in that it's not actually storing the images, it's not actually

00:53:29.280 | storing the music of third parties. It's learning from that process. We all like musicians learn on

00:53:36.560 | classical music, they listen to and play other pop artists music. And then they learn things that

00:53:41.600 | they like things that they don't like things that fit together well, certain intonation,

00:53:46.720 | syncopations, and that's how they develop their own tracks. And so I think that the assumption

00:53:52.720 | in AI is that it's almost guilty until proven innocent is what's going to really challenge this

00:53:58.800 | technology, getting mainstay appeal and being useful in the way that it could and should be

00:54:04.880 | that the de facto is that it's learning and training is copyright. I don't I don't buy that.

00:54:12.320 | I don't buy that. No, let's use the data.

00:54:14.880 | No, that that just because you're learning like a human, it's not copyright infringement. And

00:54:20.640 | I'm sorry, let me let me make my last statement, which is that I do think that similarity is

00:54:24.480 | measurable. And that is copyright infringement. So I would rather than focus on how you train

00:54:29.440 | the model, I would encourage legislators to find, as Saks points out, a better process

00:54:35.840 | to define what makes one thing distinct from another, run that analysis, and then use that

00:54:42.000 | analysis to determine uniqueness. And that analysis is ultimately what should determine whether

00:54:47.360 | there is copyright infringement versus the how was the model trained exercise, which is wrought

00:54:52.720 | with challenges, as you guys know, how do you determine weightings? How do you determine

00:54:56.240 | whether this was even relevant in the model? We don't even know if the models ignore the data or

00:55:00.800 | use the data in their synthesis or in their weightings. It's all very opaque. And so it's

00:55:05.680 | almost like, hey, here's a list of stuff that I read. But I may not have even paid attention to

00:55:09.760 | 95% of it. When I was reading it. We don't know. And so I'd rather focus on is there copyright

00:55:15.280 | infringement in the output? Does the output mimic an original copyrighted source? And if it does,

00:55:21.280 | then that's where the infringement should lie, not on the How did you do it?

00:55:25.520 | I think that you're creating an impossible standard. Is it 5%? The same? Is it 8%? The same

00:55:30.000 | is that 14%? Where's your threshold? Is it different than training? How's that different

00:55:34.480 | than training? If I'm JK Rowling? Am I not allowed to read other authors copyrighted books for fear

00:55:40.000 | that by reading it, I will now use their copyrighted material in my novel book? No. Okay,

00:55:45.520 | stick with that example. But we can also use the Ed Sheeran example. Nick, can you play this video?

00:55:49.520 | I could play you every song in the pop charts right now over four chords. You name any song

00:55:53.600 | and I'll play it in the four chords. So these, these are the chords. Messenger. Your friend

00:56:00.240 | because you only need the light when it's burning low. Only miss the song when it starts to snow.

00:56:07.040 | Only know your lover when you let it go. Let it be. I'm walking away Craig David.

00:56:11.040 | Greg David. I'm walking away from the troubles in my life. Last one. Let it be Beatles.

00:56:18.880 | When I find myself in times of trouble. Why is that important? Obviously, it's not illegal for

00:56:28.080 | Ed to have learned these four chords. But if he tries to produce those songs and calls them

00:56:33.040 | something else, that's clearly somebody else's IP. Similarly, I think that if you don't give

00:56:39.520 | people the right to say, yeah, of course, the letters A, B, C, D, E, F, G, all the way to Z

00:56:44.000 | are not my invention. It's the alphabet. I don't control that. I don't control words and their

00:56:48.080 | meanings. But when I put one word in front of another in front of another purposefully, and I

00:56:52.400 | create something that has value to me, I should have the right to protect that. And I think that's

00:56:57.760 | the fundamental question. Should you not have a right to protect it? Should you allow somebody

00:57:02.880 | else to take that? And in some opaque way, and undefinable way, make a system that could compete

00:57:09.600 | with that thing and have no repercussions. I think that that's what people need to decide.

00:57:14.480 | And I'm fine, by the way, one way or the other. Yeah, in the Ed Sheeran case, he was saying the

00:57:19.600 | exact copyrighted lyrics. And he was singing the exact copyrighted melody. He was just playing the

00:57:25.040 | harmony with four chords. I mean, that was kind of the point he was making. And in the case of

00:57:30.400 | all the copyrighted material, you're stating, I want to protect it from showing up in the end

00:57:35.360 | product of someone's creative process. I shouldn't care about them listening to my song in their

00:57:40.000 | creative process. Everyone always references how Oasis sounds like the Beatles, right? Those guys

00:57:44.800 | love the Beatles, they listen to the Beatles, so much of their music mimics and sounds like the

00:57:48.960 | Beatles, but the Beatles aren't going to file infringement suits against them. They were

00:57:51.680 | inspired by the Beatles, that was part of their creative process. And so I would argue that the

00:57:56.240 | output should really be where we assess this stuff. And it's going to take a lot of cases,

00:58:00.240 | you're creating an impossible standard where no any individual person will have the economic

00:58:06.000 | wherewithal to fight this. And so it's basically seeding all creativity to the board. And I don't

00:58:12.960 | get what is what is the standard of measuring how someone is training their creative process?

00:58:18.160 | Your standard is I got to go I don't think I got to go listen, I got to go you got to tell me every

00:58:22.240 | time you as an artist, listen to a song, and then I'm going to determine whether or not you are

00:58:28.320 | listening to copyright music and making your own music. That's basically what we're asking for here

00:58:32.640 | with this. That's not what I don't I'm not talking about the bill. I'm talking about this idea that

00:58:36.880 | if you train a system that can actually automate a task for you, based on some learning that's

00:58:42.800 | unique, you should be able to uniquely protect your ownership. For example, we have a different

00:58:49.120 | set of ways in order to protect processes that are unique to you, which is, you can have a patent,

00:58:53.840 | or you can have a trade secret. But if you strip all of this stuff away, where some system can in

00:58:59.440 | an automated way do anything for you, and replicate a whole bunch of work that it sits on top of to be

00:59:06.000 | able to do that without any repercussions. I think that that's a very important explicit decision.

00:59:11.040 | And what I'm saying is, you cannot expect individual people to go and fight these little

00:59:15.680 | battles against these big multi hundred billion dollar companies. If the big multi hundred billion

00:59:21.040 | dollar companies and trillion dollar companies don't fight the other multi hundred billion dollar

00:59:25.280 | and trillion dollar companies, what they're effectively saying is none of this is worth

00:59:29.120 | anything. Copyrights are worth nothing. And I'm okay with that decision. But it should be an

00:59:33.680 | explicit one. And it's not going to happen the other way where some five page law from a guy

00:59:40.240 | who's trying to get elected is not the thing that's going to make the difference. The difference is,

00:59:44.400 | if Google doesn't sue open AI, it means copyright is worthless. And by the way, that article didn't

00:59:49.840 | just point the finger at Google. It pointed the finger at meta where there were explicit decisions

00:59:54.320 | to say, Yeah, no, let's just violate the copyright. It's, it's okay. And if it's okay, then it's okay.

00:59:59.120 | Saks, you're the independent arbiter here. Any summary?

01:00:05.200 | Well, somebody's gonna litigate it. Somebody's gonna get gonna litigate it. You're right,

01:00:09.360 | it'd be surprising if Google didn't want to, given how valuable their YouTube content is,

01:00:14.320 | and a competitor, you know, slurped it up to train their model. But somebody will will litigate this,

01:00:20.240 | it could be the songwriters, who knows. And then we're gonna we're gonna get some messy arbitration

01:00:25.760 | around fair use, and it's probably gonna work its way up over the different circuit courts,

01:00:29.760 | you'll probably get different judgments. And finally, the Supreme Court will resolve it.

01:00:34.000 | And then we'll kind of know where things stand. And then and then there'll be a legislative fix.

01:00:39.120 | I don't know where it's going to end up. I mean, I know what Jay cow would say if he were here,

01:00:42.880 | I'm sure he's like chomping at the bit be part of this conversation. He would say that creators

01:00:47.520 | deserve to get paid for their AI rights. And that if you want to make money training a model

01:00:53.440 | on their content, you got to compensate them. I also I buy I buy your argument too, which is

01:00:59.440 | you're not really violating their their rights for the model to

01:01:06.880 | to read their content and analyze it, which is what it's really doing as long as it doesn't

01:01:12.800 | output a violation of their copyright. And that's what all artists do. That's what all humans do.

01:01:19.520 | The creative process is one of initially synthesis. You learn from other people's works,

01:01:25.680 | you learn from the environment around you, you learn from the world that you're exposed,

01:01:28.480 | there was compensation in those other models. There was not you don't of course,

01:01:32.800 | Oasis never paid the Beatles to listen to their music. They

01:01:36.480 | know that's not true. They listened on the radio, they may not have paid,

01:01:40.800 | but some advertiser paid for an ad. The radio station had to pay ASCAP and BMI,

01:01:45.920 | there was publishing rights, there was all kinds of other rights. So that's not true what you're

01:01:49.840 | saying. I'm not what I'm what I'm saying is it's not copyright just because somebody got paid. So

01:01:54.800 | that singer songwriter did get paid, they may not have getting paid by the listener. But in this

01:01:59.440 | example, nobody's getting paid. Look, I think there's three potential outcomes here. Okay,

01:02:03.920 | outcome number one is that Freiburg wins. And the courts decide that this is fair use. And

01:02:10.720 | these AI companies aren't in violation of copyright infringement is measured in the work of art that

01:02:15.760 | is produced, not in the process by which it is generated. Right? Okay, so that's solution number

01:02:20.960 | one. I think solution or outcome number two is that the courts throw a big wrench into it or a

01:02:28.080 | bill like schist bill, which is largely punitive, throws a big wrench into it. And I think what

01:02:33.280 | happens then is new AI companies have a really hard time accumulating training data. And so the

01:02:39.360 | incumbents who already have all the training data, they may have to pay a fine, there'll be some slap

01:02:44.080 | on the wrist. But now you've created a real moat and barrier to entry. And so secretly, they're

01:02:48.640 | going to love it. And it's going to be a big problem, I think, for the open source movement.

01:02:52.400 | So that's outcome number two. And then outcome number three is you create some sort of rights

01:02:58.800 | clearinghouse, kind of like exists in the recording industry, where if you want to cover somebody's

01:03:05.440 | song or perform somebody else's song, there's already predefined terms on which you can use

01:03:10.720 | those rights. And then there's a compensation mechanism for that. So you don't have to go

01:03:15.360 | negotiate every single deal that you want to do, you can just use the song, you can use the rights,

01:03:20.560 | and there's again, compensation for that. That would be like the compromise, I could see that

01:03:25.360 | alternative number three existing, in which case, the incumbents probably like that, because

01:03:31.520 | they again, they want there to be a barrier to entry. They want to make it hard for the guy in

01:03:36.480 | the garage, to pony up the capital necessary to create a new model. But at least that would be a

01:03:43.120 | somewhat functional system, it wouldn't be outcome number two, which would totally put the kibosh on

01:03:48.400 | the use of training data. So in any event, I don't know which one of the three it's going to be. But

01:03:52.240 | those are the three potential outcomes I see happening here. Have you guys seen you do you dio?

01:03:56.960 | No, it's like a new music gen AI system. It is friggin unbelievable.

01:04:07.520 | Is that the notes thing or the notes another thing?

01:04:09.200 | I think the notes another thing this one just came out, it was apparently rumored for a while

01:04:14.480 | ex DeepMind employees started you do it's been this kind of like hot stealth company.

01:04:20.160 | For a couple of months, people have been talking about it as rumored to be the best music AI.

01:04:26.880 | But here is someone that said, Dune the Broadway musical, and this is what it output.

01:04:32.240 | Go ahead.

01:04:41.920 | That's crazy.

01:04:48.800 | You can hear the layers of music, the layers of the tracks,

01:04:51.840 | the smoothness, the tempo, obviously, like there's just so many layers to this.

01:04:57.600 | Nick, can you pull up another one? Some of these are just incredible,

01:05:00.320 | because it spans genres. It spans styles. It's like so beautiful, like the way it's put together

01:05:07.040 | every every piece feels somewhat emotional. It's unique. It really doesn't feel like you're just

01:05:12.640 | listening to someone printing copyright music. funky dance pop song. Create a birthday song

01:05:24.640 | for my friend urban. Create an Irish folk tune with violin and instrumental.

01:05:34.160 | It's pretty incredible. And I think it shows the state of the art is such now that this

01:05:49.840 | is going to become a real challenging question from a legislative point of view.

01:05:54.240 | Given how far ahead these technologies have gotten, and I think that musicians,

01:05:58.480 | artists, consumers are going to start to use these tools in a really

01:06:01.520 | prolific way, given how good they are now. I heard from someone that just sat for hours,

01:06:07.440 | creating tracks that they listened to on UDO the other day. And so you basically just playing your

01:06:12.720 | own music all day long. It's really incredible. Anyway, exciting time.

01:06:16.640 | Hopefully all the lawsuits and the legislators don't get in the way of all this innovation,

01:06:20.560 | because it's moving really fast. And we should just kind of see where it goes before trying to

01:06:24.480 | clamp down. Totally. Guys, I want to talk about, as you know, this like emerging trend that we've

01:06:31.440 | been talking about offline about drones in warfare. Recently, we saw small drones controlled

01:06:36.800 | by the Houthis attack cargo vessels in the Suez Canal. It led to rather large scale supply

01:06:42.320 | disruptions, economic value destruction. We've seen videos, sex, I think you have one, we can

01:06:47.840 | pull it up and take a look of Ukrainian controlled drones flying into destroy tanks and large

01:06:53.760 | warehouses in Russian held territories, causing 10s of millions of dollars of damage. I think

01:06:58.800 | we're seeing the acceleration of the changing face of warfare technology, lots of these small

01:07:03.040 | form factor, low cost autonomous physical agents. And I think there's a bunch of implications both

01:07:09.280 | with respect to war, but also how we spend money on defense, and how Silicon Valley is involved

01:07:15.520 | in response. tax, I'd love to hear your points of view. Maybe you can highlight for us some of the

01:07:20.080 | things you've been seeing in the utilization of these new technologies and in the war, given our

01:07:25.200 | your our in house work correspondent, and you obviously do some work in Silicon Valley. Go

01:07:30.480 | ahead. I've heard a lot about the impact of drones just by watching the coverage of the Ukraine war.

01:07:36.320 | I was listening to an interview with an American mercenary who's fighting on the side of Ukraine,

01:07:42.320 | and he described how ubiquitous these drones were on the battlefield. Now he said that you literally

01:07:47.680 | can't get out of a trench to go to the bathroom because a drone will basically get you. And he

01:07:53.840 | said they're buzzing around, they sound like mosquitoes, because they're kind of just everywhere

01:07:57.840 | on the battlefield. Both Ukraine and Russia have them. And several months ago going into this year,

01:08:05.360 | as long as he said that this would be the big game changer for Ukraine, they're going to make

01:08:08.400 | a million drones. I don't think it's worked out that way. It turns out that as many drones as

01:08:12.880 | Ukrainians have been able to put on the battlefield, the Russians have even more,

01:08:16.240 | because they're able to mass produce them in factories, they've got bigger drones, better

01:08:21.280 | drones. And I think the most important variable in the war so far, with respect to drones is that

01:08:26.800 | the Russians have pretty advanced electronic warfare. And so they've been able to jam a lot

01:08:32.160 | of the Ukrainian drones, whereas the reverse has not been true for the Ukrainians. So I would say

01:08:38.560 | that drones have been a huge factor in the war. But so far, the balance there is tipping towards

01:08:45.040 | the Russians as it is in so many other areas as well. But to your larger point, there's no

01:08:50.560 | question this is the future of warfare. And you're seeing that it's creating a lot of opportunities

01:08:57.440 | for asymmetric warfare. So for example, with the hoodies, they've been firing cheap missiles and

01:09:04.560 | drones at our at our ships in the Red Sea, and we've been having to spend $2 million air defense

01:09:11.040 | missiles shooting down $2,000 drones. So if that continues, and we don't have a good response to

01:09:16.000 | this problem, it's going to really change the balance of power. I don't know if this was at

01:09:20.240 | our summit, or if I heard this later from some senior person in the military, who said that

01:09:24.640 | aircraft carriers are outdated technology, like they don't make sense anymore. And we're going

01:09:29.840 | to see a shift towards lots of small, autonomous drone like systems on the battlefield taking out

01:09:38.080 | targets. And as a result, if you gain theory this out, the necessity for defense systems against

01:09:44.720 | large amounts of swarming drone like systems. And then, you know, what do you actually have to do

01:09:51.440 | with your existing military architecture to play into this kind of new tactical system?

01:09:57.840 | And then that's going to require a massive evolution in technology. And is the United

01:10:04.400 | States prepared? Should Silicon Valley play should be, we should I maybe Nick, you can throw up this

01:10:11.840 | the first link that I sent you. But eight years ago, I came to this conclusion. And I invested

01:10:17.600 | in this company called sail drone, which is autonomous drones in the seas, our customers

01:10:22.720 | include the US Navy, the chairman of our company actually is Admiral Mike Mullen, who is a former

01:10:27.280 | chairman of the Joint Chiefs of Staff. But these drones are capable of going into some of the most

01:10:32.800 | dangerous places in the world, and collecting enormous amounts of surveillance data and

01:10:37.680 | information that otherwise takes the United States government lots of time and lots of money,

01:10:46.240 | and is pretty scattershot. And instead with these things, you can be constantly in hotspots. And if

01:10:53.840 | you go to the next one, so much so that you know, we've deployed some of these drones in the Middle

01:10:58.320 | East, and there was a point at which the Iranian Navy intercepted two of our drones and pick them

01:11:03.440 | up. And this was kind of like a global thing a few years ago. But I really believe in this trend.

01:11:09.120 | And I think that we have an enormous responsibility to be funding these things. These are really

01:11:14.800 | complicated systems to build, obviously, and they take lots of time. So these are not

01:11:18.320 | overnight successes, Friedberg, and they take lots of money, which is hard to come by as well.

01:11:22.880 | But these are absolutely the right kinds of businesses, because eventually,

01:11:26.480 | at a minimum, you're building systems that can measure and collect enormous amounts of really

01:11:31.600 | critical data, so that people can make better decisions, which hopefully is measured in saving

01:11:38.000 | life, right. And then over time, you actually get to a military that should be much cheaper,

01:11:44.720 | much safer, and has fewer people in general on our side and on our enemy sides on the front lines,

01:11:52.080 | which means fewer casualties. And I think that that's the whole value of this entire movement.

01:11:57.040 | It's cheaper, it's way better. It's much more useful. And it gives the respective armies that

01:12:04.480 | use these things or militaries rather, a fuller picture so that you can make decisions that have

01:12:08.880 | enormous consequences. Yeah. And in February, the Department of Defense on a panel in Arlington,

01:12:16.080 | Virginia, briefed reporters and industry folks on their interest in growing their partnerships in

01:12:22.640 | Silicon Valley, to access the necessary technologies that are going to rewrite the

01:12:29.360 | face of warfare. I don't know about you guys, but I've seen kind of two sides of this in Silicon

01:12:35.120 | Valley. A good number of venture capitalists and entrepreneurs and technologists think that it's

01:12:41.840 | morally incorrect to support warfare technology. And then there are a few that are going blazing

01:12:48.480 | ahead with supporting this new evolution and these technologies. Do you guys see the same?

01:12:53.360 | And is there going to be kind of this? It's a big step to weaponize these things,

01:12:57.040 | right? So Saildrone, we have a very good, big, thriving, successful business. These

01:13:03.440 | machines aren't weaponized by any stretch of the imagination. And so I think that that

01:13:08.400 | comment is more genuflecting and virtue signaling than it is real. Because in order to even be

01:13:17.840 | legitimately considered for some kind of like armed, unmanned product, you have to be deep

01:13:26.480 | inside of the bowels of the DoD and Pentagon and have been working with them for years,

01:13:31.280 | don't even be taken seriously. So the startups that are like eschewing those deals are not even

01:13:36.400 | close to those deals. And the startups that are close to them are probably being much more quiet

01:13:42.400 | and thoughtful and hush hush, because the path to do that isn't legitimate unless you've been

01:13:47.120 | in business with these folks for 10 plus years. Yeah, I think it's, it's more commentary on the

01:13:53.440 | fact that if this is where an organization, the Department of Defense that has a trillion dollar

01:13:59.920 | annual budget is going to be allocating resources, that technology is going to need to come from

01:14:06.960 | somewhere. And the investors in Silicon Valley, that are making these investments are likely going

01:14:12.720 | to outperform those who are not. Does that sound reasonable to you, Sax, that there is this shift

01:14:18.000 | underway? And that, again, those who have shied away from defense technology are, you know,

01:14:24.160 | necessarily going to be left out of a, you know, kind of a new industry that's emerging in Silicon

01:14:28.240 | Valley? Well, I guess the way I put it is that the US government spends over $800 billion every

01:14:34.240 | year on defense. And it's not clear what we're getting for that money. Because in Ukraine,

01:14:39.280 | for example, we've run out of artillery shells, we've actually we've run out of patriots, we've

01:14:42.800 | run out of javelins, we're on a stinger. So we're spending all this money, we're not getting a lot

01:14:47.040 | for it. And part of the reason is because the defense industry is consolidated down to these

01:14:51.920 | five prime defense contractors who are basically an oligopoly. And we have this cost plus procurement

01:14:59.040 | system where they just raise their prices every year and the government pays it. So I think all

01:15:05.040 | of us want the United States to have an effective defense. I mean, I want us to use our military

01:15:09.280 | power more wisely. I don't like all these stupid wars we keep getting in. But I do want the United

01:15:13.360 | States as an American to be the most powerful country. I do want us to get the best value for

01:15:20.720 | our defense dollars. And the only way that's going to change is if the defense industry gets disrupted

01:15:27.680 | by a bunch of startups doing innovative things. And there's no question that I think drones are

01:15:33.280 | the future of warfare. To your point about autonomy, I think that is where this is going next

01:15:38.160 | is that right now, the drones are typically controlled by, you know, somebody with like a VR

01:15:45.520 | headset, tele operated, operated and their FPV, this first person vision drone, and you basically

01:15:53.200 | strap some sort of explosive on to it, and then you drive it into whatever your target is. Yeah.

01:15:58.960 | Those types of drones are easier to disrupt by electronic warfare, because if you can disrupt

01:16:05.680 | the signal from the tele operator, then the drone basically doesn't know what to do. And so you're

01:16:12.160 | right, the next step here is autonomous systems that can be programmed with a target and can find

01:16:20.960 | it on its own, make decisions on its own. And then they also build in some shielding against

01:16:25.200 | some of this like jamming technology or this EW electronic warfare technology.

01:16:31.040 | So that's where all these things, that's where the battlefield is of the future is headed.

01:16:35.120 | And, you know, in a weird way, if you think about humans becoming a smaller and smaller

01:16:41.040 | piece of the battlefield, and autonomous drones becoming a larger and larger piece of it,

01:16:46.160 | these wars become resource wars, and that's right, actually, technology wars.

01:16:51.200 | That's right, which which may or may not be good. I don't know.

01:16:53.600 | But by the way, so this is the point I wanted to make, if you pull up this chart,

01:16:56.800 | if that is where warfare is headed, is a large number of small autonomous systems

01:17:03.440 | that go and find a target and try and execute a mission. And, you know, in the case of Ukraine,

01:17:08.800 | saying they want to have a million, China might say, I want to have 100 million, all of these

01:17:13.840 | systems are dependent on lithium ion battery systems. Today, 79% of lithium ion battery

01:17:18.640 | production comes out of China. The US is only 6.2% of global lithium ion battery production.

01:17:24.240 | And China has talked about scaling up drone manufacturing to a level that the US simply

01:17:31.280 | cannot even contemplate in its industrial architecture today. So it seems to me that

01:17:38.160 | if that is where warfare tactically is headed, that China has a huge leg up, and is going to

01:17:44.400 | become a critical point of dependency for the United States to develop an arsenal necessary

01:17:50.720 | to be competitive in this this kind of next evolution of warfare. I will say that, like,

01:17:56.000 | if you then game theory this out, like how do you defend against these autonomous electronic systems,

01:18:00.800 | there's a technology called EMP or electromagnetic pulsing, where, as you guys know, if you run a

01:18:07.440 | very high current electric field, you can actually send out and emit a pulse that then when it hits

01:18:13.280 | electronic circuits far away, induces a high electric current in those circuits and short

01:18:18.560 | circuits them. So EMPs are a defense system that allow you to take out electronic systems. And this

01:18:26.560 | has been, you know, kind of a part of warfare, since probably the 1960s 50s, but targeted EMPs

01:18:32.800 | and targeted systems for eliminating all of these autonomous systems becomes a new defense technology

01:18:37.520 | that I know several startups in Silicon Valley that have been funded with a lot of capital

01:18:43.920 | by folks that we all know very well, that are kind of pursuing this,

01:18:48.160 | we don't have a choice. Because I think the point is that if you just make me just post this photo,

01:18:53.360 | we have an enormous human capital problem with the military, which is there's just not enough

01:18:57.520 | folks enlisting anymore. So we don't have any choice except to automate and become

01:19:02.720 | drone dependent. Just tell us these numbers. This is a chart from the Department of Defense

01:19:07.440 | that shows military enlistments looking back from about the 1970s, through today. And these are two

01:19:15.040 | lines that are just going from the upper left to the lower right. And so we're at all time lows

01:19:20.160 | with respect to the number of people that actually want to join the military. And so if we are

01:19:25.280 | supposed to, as Zach said, be this very sharp fighting force, then all of the money that we're

01:19:32.240 | spending needs to get allocated into things that can be remotely operated. We don't have a choice.

01:19:38.400 | Yeah. I mean, so Freeberg, you asked the question, how do we defend against these drones?

01:19:42.560 | Yes, there's EMP technology. If you want to use an EMP, you better make sure there's not any planes

01:19:48.480 | nearby, because it will take them out of the sky, you better make sure.

01:19:51.120 | By the way, there are now, you can target EMPs in a narrow cone, which is part of the

01:19:55.440 | technology evolution.

01:19:56.640 | Yeah, you better make sure you don't hit any of your own stuff, because you'll fire the electronics

01:19:59.920 | on those unless you EMP harden them, which by the way, will happen. I mean, if the US gets really

01:20:04.960 | good at sending out EMP pulses, then our opponents will start EMP hardening their drones. But anyway,

01:20:12.640 | yes, it's a category of defense. So is electronic warfare. The one defense investment I've made is

01:20:18.720 | actually in this idea of how do you defend against drones. And it's a startup where I

01:20:23.920 | led the seed round called Allen Control Systems, and they have a product called Bullfrog,

01:20:27.600 | which is a gun turret that uses a standard M240 machine gun. But it's got a scanner next to it,

01:20:34.800 | next to the machine gun that is searching the sky for enemy drones. And it uses computer vision to

01:20:42.560 | recognize them, and then it just shoots them down like very quickly. And it seems to me this is,

01:20:47.680 | it's maybe not the only way of doing it, but it's a really good way of doing it because

01:20:51.360 | you can mount these things to a vehicle. I mean, imagine in the future,

01:20:56.960 | that we have these autonomous drones everywhere that are basically assassination drones.

01:21:03.600 | How are you going to stop them? I mean, how do you move the President of the United States around?

01:21:06.640 | I mean, in his caravan, you're going to have to have, you know, his caravan is going to have to

01:21:11.520 | have some sort of electronic warfare system on it to basically prevent drones, but it's gonna need

01:21:19.440 | some sort of last line of defense where you can actually shoot a drone out of the sky.

01:21:23.360 | Or he stays in his basement.

01:21:25.120 | So Sax, you don't have any moral trepidation around investing in defense technology,

01:21:29.920 | just to be clear?

01:21:30.720 | Well, you know, I'd have to think twice about investing in a weapon. This is more of a defense

01:21:35.360 | technology. This is defensive in nature where you're trying to take out a drone that would

01:21:39.760 | be attacking one of our bases or installations or convoys. So this is fundamentally defensive

01:21:46.240 | in nature. And look, I want America to be the most powerful country. I'm an American, I'm patriotic.

01:21:52.800 | So I don't want our troops to be hit by enemy drones. So no, I don't have any moral problem

01:21:59.040 | with this. I do have a problem with all the stupid wars the United States gets into,

01:22:04.080 | but that's on a policy level. I feel like that's separate from the question of do we want our guys

01:22:09.600 | to be protected? Yes, we do.

01:22:11.440 | And Chamath, you don't care, you would invest in defense technology companies.

01:22:15.040 | No, I care. I think that I would probably have to think twice about making weapons. I'm not sure

01:22:19.360 | that I'm really interested in that. But like I said, eight years ago, my thought was we need to

01:22:23.600 | move to just looking at those trends, we need to move to an entire surface area of unmanned

01:22:29.040 | autonomous vehicles. And I really liked this idea of a class of machine that's there to surveil and

01:22:37.600 | collect all the important data so that you can make better decisions. And

01:22:40.640 | it's taken us eight years, but we're firmly entrenched now.

01:22:44.400 | I think the fact that both of you highlight a degree of trepidation, and you guys are like

01:22:50.160 | fairly outside of the middle bounds of how Silicon Valley would think and make decisions indicates

01:22:56.320 | like why so much of Silicon Valley has deep concerns about investing in defense technology.

01:23:00.480 | And the US, despite having the best technology, may be challenged by the fact that our Silicon

01:23:07.360 | Valley technologists and investors are held back on pursuing these evolutions because

01:23:17.200 | of a moral suasion. And I think it is something that the Department of Defense has highlighted.

01:23:22.320 | No, I want to be clear though, it's never like we've been faced with that choice. And the reality

01:23:28.000 | is like, if the choice is stay small, or scale up and be really big, you know, it's hard to

01:23:36.800 | turn away from that decision. But that choice isn't a realistic one. This is why I was saying,

01:23:41.840 | I think it's, you can come into it with the moral framing. But I do think that Sachs is right,

01:23:46.320 | which is, if you believe that America should be the most important country in the world,

01:23:51.040 | and you have the ability to help it be that, eventually, some of those decisions will

01:23:56.880 | come in contact with this kind of decision. And usually, though, it takes many, many years before

01:24:04.320 | that's realistic. And to be honest, from what I've seen, the opportunity to pursue that is

01:24:13.280 | generally so financially remunerative that everybody's like, Okay, I think we have to do

01:24:17.520 | you know, I don't have a problem with investing in a company like and roll. I mean, America needs

01:24:22.160 | defense companies in order to maintain its defense. And the problem right now that we have

01:24:26.640 | is there's these five prime companies that are antiquated and really expensive. And we need

01:24:33.200 | startup disruptors to that. So I'm not like, morally opposed to it. And in fact, I think it's

01:24:37.680 | probably a good thing. I think my concern would be, I don't really want to be part of the military

01:24:41.920 | industrial complex. I feel like it might, like, corrupt my views on things like I don't want to

01:24:48.000 | know you want war, you'd be girding for Yeah, like, I don't want to be I don't want to feel

01:24:52.400 | that's your dimension. That's Yeah, right. I don't want to feel an incentive to be like these

01:24:57.200 | generals who go on CNN and Fox News and justify every single war. That's my biggest fear as well.

01:25:02.960 | It's like, it's one thing to be investing in a bunch of companies where we're sitting around

01:25:07.520 | hoping for cheaper clicks on Facebook and Google. It's an entirely other thing to sit on the board

01:25:12.160 | of a company and hope for war. That's a horrible place to be in. On that point. That's a great

01:25:16.560 | point to wrap on, guys. This has been fun. A little more flat than it would normally be with

01:25:21.440 | Jay cow. We miss him. We wish him well, we hope his tooth recovers. And we will be back with your

01:25:28.640 | it's like that scene in Dune where he eats his tooth and releases that poison gas.

01:25:32.880 | That's right. Get better soon. J. cow. Please, if you're interested in the summit,

01:25:39.520 | summit that all in podcast.co. Click Subscribe on YouTube. I've been given notes to say that

01:25:45.440 | for the chairman dictator Chamath Palihapitiya, the rain man David Sachs, I am your Sultan of

01:25:51.440 | science day Friedberg. We will see you guys next week. We open source it to the fans and

01:26:09.040 | they've just gone crazy. We should all just get a room and just have one big huge orgy

01:26:32.720 | because they're all just useless. It's like this like sexual tension, but they just need

01:26:36.000 | to release it. I'm going all in. I'm going all in.

01:26:56.000 | [BLANK_AUDIO]

E174: Inflation stays hot, AI disclosure bill, Drone warfare, defense startups & more

Chapters