E129: Sam Altman plays chess with regulators, AI's "nuclear" potential, big pharma bundling & more

00:00:00.000 | I have a little surprise for my besties. Everybody's been

00:00:03.240 | getting incredible adulation. People have been getting

00:00:06.840 | incredible feedback on the podcast. It is a phenomenon as

00:00:10.080 | we know. And I thought it was time for us to do performance

00:00:14.640 | reviews of each bestie. Now I, as executive producer, I'm not

00:00:19.700 | in a position to do your performance reviews. I too need

00:00:23.100 | to have a performance review. So I thought I would let the

00:00:25.560 | audience do their performance reviews. So we went and we're

00:00:28.600 | debuting a new feature today. This very moment. It's called

00:00:31.480 | Reddit performance review. God cue some music here, some

00:00:35.720 | graphics, Reddit performance reviews. And so we'll start off

00:00:39.600 | with you, David Friedberg. This proves why you haven't been

00:00:42.720 | successful in life so far.

00:00:44.080 | You kidding me?

00:00:46.600 | I'm trying to build a successful enterprise instead of keeping

00:00:49.960 | the judgment to yourself, which is what all elite performers do.

00:00:53.120 | Yes. You turned it over to a bunch of mids on Reddit. Yes.

00:00:57.160 | Tell you, they were already doing it. They were already

00:00:59.200 | doing it. We just collected it.

00:01:00.560 | What elucidation were you going to get from Reddit? Yeah, really?

00:01:03.240 | Let's not ruin the bit. Do you think Elon does 360 performance

00:01:06.680 | reviews?

00:01:07.200 | On Reddit? On Reddit of all things? You know how many 360

00:01:11.160 | performance reviews I've done in my life? Zero, of course. And

00:01:14.680 | that's why this will be so entertaining.

00:01:16.480 | Okay, go go start off with me. Let's hear it.

00:01:18.520 | Here's how it's gonna work. You are going to be presented with

00:01:21.840 | the candid feedback that you've gotten in the last 30 days on

00:01:24.600 | Reddit. But for the first time, and you have to read it out loud

00:01:28.600 | to the class, Friedberg, you'll go first. Here's your first

00:01:32.400 | piece of candid feedback in your 360 from the Reddit in cells.

00:01:37.000 | Good Friedberg read it out loud.

00:01:38.280 | David Friedberg deserves more hate. He and the others have

00:01:42.080 | made it their mission to convince us that reforming social

00:01:45.000 | security is the only way forward to survive. He hides behind this

00:01:48.920 | nerdy a political persona, and then goes hard right out of

00:01:53.240 | nowhere. exact fear mongered about the deficit as an excuse

00:01:57.480 | to restructure entitlement programs, we would see that as

00:02:00.640 | the partisan right wing take it is when Friedberg does it. We're

00:02:05.080 | supposed to act like he has no skin in this game. He's just the

00:02:08.000 | science guy. No, he's a rich guy who would rather work your

00:02:12.440 | grandparents to death and pay an extra 5% tax.

00:02:15.320 | All right, there's your review. Very good. I think you took it

00:02:17.880 | well. And you don't have to respond now. Don't just take it

00:02:21.600 | just to counter briefly, I have highlighted multiple times. I

00:02:25.160 | think we're going to 70% tax rates. But hey, you know, right

00:02:29.040 | there on everyone's got an opinion that won't generate more

00:02:31.320 | revenue. Yeah, history's any guide? Yeah.

00:02:33.480 | Well, the audience has been waiting for this. David Sachs

00:02:36.120 | has never taken a piece of feedback and the feedback he has

00:02:38.880 | gotten he hasn't taken well. So here we go, David. Here's your

00:02:41.640 | performance. Go ahead. I just read this go with the bit. Come

00:02:45.600 | on. It's a little feedback for you. Come on. All right. And I'm

00:02:48.400 | gonna do this too. And I haven't seen mine.

00:02:50.920 | Oh, Nick pulled these. Nick police. This is not

00:02:53.200 | these are actual pieces of real feedback.

00:02:57.280 | Fire. That's it. Get out of here.

00:03:00.160 | Good sex.

00:03:02.280 | The thing about David Sachs, if he wasn't rich, everyone would

00:03:05.040 | dismiss him as being both stupid and boring. The secret to his

00:03:09.000 | wealth is just follow Peter Thiel around since college. It's

00:03:11.360 | like if turtle from Entourage pretend to be a public

00:03:13.240 | intellectual.

00:03:13.840 | Okay, so the dictator

00:03:20.880 | has never had a 360 review. He informs us and I think his staff

00:03:24.240 | you know, for all the people at social capital, you can get in

00:03:27.360 | on this by just posting to Reddit since he won't do 360

00:03:29.800 | reviews at social capital guy. Let's pull up two months.

00:03:31.960 | Feedback for the quarter.

00:03:33.320 | Is the biggest self serving leech. As long as he can make a

00:03:38.680 | dollar trade on it. He will burn anything down to the ground.

00:03:42.360 | Fuck the consequences to society or anyone else.

00:03:45.640 | This was actually the good

00:03:46.440 | positive

00:03:49.280 | Why'd you give him such a good review?

00:03:50.480 | That's the truth.

00:03:52.880 | I got to read mine. I haven't seen this. I'm bracing for

00:03:58.280 | impact here. Oh, no.

00:03:59.240 | What's the critique exactly? What is the critique from all

00:04:02.600 | these

00:04:02.960 | proof from that? I mean, that's like,

00:04:04.600 | pretty accurate. Okay. All right, here we go. I got to read

00:04:08.160 | mine. Okay. I can't wait for AI to replace Jake. Jake is the

00:04:12.440 | least skilled knowledge worker on the show. I think he has

00:04:15.840 | about three shows left before AI replaces his hosting skills.

00:04:19.040 | And ability to trick dentists into investing in high

00:04:21.720 | last part.

00:04:27.680 | I do have a lot of dentist friends in the funds. Okay, I'm

00:04:33.640 | gonna pay for a lot of kids dentist school.

00:04:35.080 | Great idea. Great idea. J. Kelton give

00:04:37.360 | one group one, this is one for the whole group. This is a group

00:04:41.120 | survey. This is our group 360 vote to rename the podcast. The

00:04:46.080 | vote is binding and the podcast we renamed

00:04:48.600 | three billionaires and J. Kel three kind of smart guys and J

00:04:51.920 | Cal three assholes and freebird.

00:04:53.840 | That's a good one.

00:04:57.720 | All right, everybody, welcome to the all in podcast. We're still

00:05:17.480 | here episode 129. All in summit 2023. General Admission sold

00:05:22.600 | out. Too many people apply for scholarships that's on pause.

00:05:26.400 | And there's a couple of VIP tickets left get them while

00:05:28.960 | they're hot. Just search for all in summit freeberg. Anything to

00:05:31.680 | add? We'll just get through the grift real quick here.

00:05:33.400 | No, it's gonna get it all right. That was it.

00:05:35.840 | Yeah, I mean, just more demand than we predicted. We look for a

00:05:39.360 | bigger venue couldn't find one. We're I think we're excited

00:05:42.040 | about Royce Hall. It's still as you pointed out two and a half

00:05:45.240 | times the size of last year. So we want to make sure it's a

00:05:47.280 | great quality event. But unfortunately, way too many

00:05:49.920 | folks want to go so we have to kind of pause ticket sales.

00:05:53.360 | What's my wine budget? $300 per person per night $1,000 per VIP

00:05:58.960 | per event. Thank you. Okay, I will handle it from here. So

00:06:03.600 | there's 750 of them. So I think you have $750,000 in wine

00:06:06.920 | budget. I just can't believe I just gave him 750 by wide by

00:06:15.600 | guys, I'm going to curate an incredible

00:06:17.800 | here today, by the way, he's

00:06:19.640 | Oh, doing some way. Yes, no, he Josh,

00:06:23.320 | some way Josh. All right, let's get to work. Let's get to work.

00:06:26.600 | Okay, the Senate had a hearing this week for AI. Sam Altman was

00:06:31.880 | there as well as Gary Marcus, a professor from NYU. That's an

00:06:36.960 | overpriced college in New York City. And Christina Montgomery,

00:06:41.320 | the chief privacy and trust officer from IBM, which had

00:06:44.120 | Watson before anybody else was in the AI business. And I think

00:06:47.520 | they deprecated it or they stopped working on it, which is

00:06:49.800 | quite paradoxical. There were a couple of very interesting

00:06:53.640 | moments. Sam claimed the US should create a separate agency

00:06:58.040 | to oversee AI, I guess he's in the Chamath camp. He wants the

00:07:01.640 | agency to issue licenses to train and use AI models, a

00:07:06.480 | little regulatory capture there, as we say in the biz. He also

00:07:09.600 | claims, and this was interesting dovetailing with Ilan's CNBC

00:07:13.960 | interview with I think Dave Farber, which is very good, that

00:07:17.440 | he owns no equity in open AI whatsoever, and was quote, doing

00:07:24.040 | it because he loves it. Any thoughts, Chamath, you did say

00:07:29.440 | that this would happen two months ago. And here we are two

00:07:32.800 | months later. And exactly what you said would happen is in the

00:07:36.080 | process of happening. regulation, licensing and

00:07:39.200 | regulatory capture, Sam went a little further than I sketched

00:07:46.120 | out a few months ago, which is that he also said that it may

00:07:49.440 | make sense for us to issue licenses for these models to

00:07:53.800 | even be compiled. And for these models to actually do the

00:07:59.000 | learning. And I thought that that was really interesting,

00:08:03.200 | because what it speaks to is a form of KYC, right? Know your

00:08:06.400 | customer. And again, when you look at markets that can be

00:08:10.800 | subject to things like fraud and manipulation, right, where you

00:08:14.040 | can have a lot of bad actors, banking is the most obvious one.

00:08:16.840 | We use things like KYC to make sure that money flows are

00:08:20.400 | happening appropriately and between parties that were the

00:08:23.640 | intention is legal. And so I think that that's actually

00:08:29.440 | probably the most important new bit of perspective that he is

00:08:33.600 | adding as somebody right in the middle of it, which is that you

00:08:38.040 | should apply to this agency to get a license to then allow you

00:08:41.040 | to compile a model. And I think that that was a really

00:08:44.920 | interesting thing. The other thing that I said, and I said

00:08:47.680 | this in my in a tweet just a couple days ago is I'm really

00:08:50.840 | surprised actually, where this is the first time in modern

00:08:54.360 | history that I can remember, where we've invented something

00:08:58.080 | we being Silicon Valley, and the people in Silicon Valley are the

00:09:02.160 | ones that are more circumspect than the folks on Wall Street or

00:09:05.960 | other areas. And if you see if you gauge the sentiment, the

00:09:10.800 | hedge funds and family offices right now are just giddy about

00:09:14.920 | AI. And it turns out if you look at the 13 Fs, they're all long

00:09:17.760 | NVIDIA and AMD. But if you actually look at the other side

00:09:22.640 | of the coin, which is the folks in Silicon Valley, that's

00:09:24.480 | actually making it the rest of us are like, hey, let's crawl

00:09:27.720 | before we walk before we run.

00:09:29.360 | Yeah, let's think about guardrails. Let's be thoughtful

00:09:31.720 | here. And so the big money people are saying, let's place

00:09:34.680 | bets. And the people building in are saying, Hey, let's be

00:09:38.120 | thoughtful sack, which is opposite to what it's always

00:09:40.640 | been, I think, right? We're like, hey, let's let's run with

00:09:43.160 | this. And Wall Street's like, prove it to me. sacks, you are a

00:09:47.960 | less regulation guy. You are free market monster. I've heard

00:09:51.040 | you've been called. You don't believe that we should license

00:09:54.680 | this. What do you think about what you're seeing here? And

00:09:57.720 | there is some cynical, cynical thoughts about what we just saw

00:10:02.320 | happen in terms of people in the lead, wanting to maintain their

00:10:06.160 | lead by creating red tape. What are your thoughts?

00:10:09.880 | Yeah, of course. I think, you know, Sam just went straight for

00:10:13.360 | the endgame here, which is regulatory capture. Normally,

00:10:16.560 | when a tech executive goes and testifies at these hearings,

00:10:19.440 | they're in the hot seat, and they get grilled. And that

00:10:21.240 | didn't happen here. Because, you know, Sam Allman basically bought

00:10:24.840 | into the narrative of these senators. And he basically

00:10:28.720 | conceded all these risks associated with with AI talked

00:10:32.200 | about how chat GPT style models, if unregulated, could increase

00:10:38.280 | online misinformation, bolster cyber criminals, even threaten

00:10:41.640 | confidence in election systems. So he basically bought into the

00:10:47.600 | senators narrative. And like you said, agree to create a new

00:10:50.840 | agency that would license models and can take licenses away. He

00:10:56.000 | said that he would create safety standards, specific tests that

00:10:58.840 | the model has to pass before it can be deployed. He says he

00:11:02.640 | would require independent audits who can say the model is or

00:11:05.720 | isn't in compliance. And by basically buying into their

00:11:10.400 | narrative and agreeing to everything they want, which is

00:11:13.880 | to create all these new regulations and a new agency, I

00:11:16.920 | think that Sam is pretty much guaranteeing that he'll be one

00:11:19.760 | of the people who gets to help shape the new agency and the

00:11:24.080 | rules they're going to operate under and what these independent

00:11:26.480 | audits are going to how they're going to determine what's in

00:11:28.680 | compliance. So he is basically putting a big boat around his

00:11:32.600 | own incumbency here. And so yes, it is a smart strategy for him.

00:11:39.600 | But the question is, do we really need any of this stuff?

00:11:42.120 | And you know, what you heard at the hearing is that just like

00:11:45.720 | with just about every other tech issue, the centers on the

00:11:48.720 | judiciary committee didn't exhibit any real understanding

00:11:52.240 | of the technology. And so they all generally talked about their

00:11:55.400 | own hobby horses. So you know, you heard from Senator

00:11:58.400 | Blackburn, she wants to protect songwriters. Holly wants to stop

00:12:02.000 | anti conservative bias. Klobuchar was touting a couple

00:12:05.120 | of bills that have her name on them once called the JCP a

00:12:09.200 | journalism competition preservation Bernie Sanders want

00:12:11.480 | to do he wants to protect the 1% of the one Durbin hate section

00:12:15.240 | 230. That was the hobby horse he was riding and then Senator

00:12:17.560 | Blumenthal was obsessed that someone had published deep fakes

00:12:20.600 | of himself. So you know, all of these different senators had

00:12:24.680 | different theories of harm that they were promoting. And they

00:12:28.160 | were all basically hammers looking for a nail. You know,

00:12:30.920 | they all wanted to regulate this thing. And they didn't really

00:12:34.560 | pay much of any attention to the ways that existing laws could

00:12:37.880 | already be used. Absolutely. To stop any of these things. If you

00:12:41.920 | commit a crime with AI, there are plenty of criminal laws,

00:12:44.920 | every single thing they talked about could be handled through

00:12:47.000 | existing law, if they're not doing harms at all. Yeah, but

00:12:50.440 | they want to jump right to creating a new agency and

00:12:52.760 | regulations. And Sam, I think did the, you know, expedient

00:12:58.000 | thing here, which is basically buy into it in order to was a

00:13:01.640 | chance to be an insider. If this was a chess game, Sam got to the

00:13:04.680 | mid game, he traded all the pieces and went right to the end

00:13:07.080 | game. Let's just try to checkmate here. I've got the

00:13:09.600 | lead. I got the 10 billion from Microsoft. Everybody else get a

00:13:12.960 | license and try to catch up Friedberg. We have Chamath pro

00:13:16.760 | regulation licensing, I think, or just being pretty thoughtful

00:13:19.760 | about it. There you got sacks being typically a free market

00:13:24.240 | monster, let the laws be what they are. But these senators are

00:13:26.800 | going to do regulatory capture. Where do you as a sultan of

00:13:29.720 | science stand on this very important issue?

00:13:32.240 | I think there's a more important kind of broader set of trends

00:13:39.640 | that are worth noting and that the folks doing these hearings

00:13:42.600 | and having these conversations are aware of, which implies why

00:13:46.640 | they might be saying the things that they're saying. That's not

00:13:49.280 | necessarily about regulatory capture. And that is that a lot

00:13:51.960 | of these models can be developed and generated to be much smaller.

00:13:56.680 | We're seeing, you know, models that can effectively run on an

00:13:59.560 | iPhone, we're seeing a number of open source models that are

00:14:02.160 | being published. Now there's a group called mosaic ml. Last

00:14:05.440 | week, they published what looks like a pretty good quality model

00:14:08.920 | that has, you know, a very large kind of token input, which means

00:14:13.240 | you can do a lot with it. And that model can be downloaded and

00:14:17.440 | used by anyone for free, you know, really good open source

00:14:20.600 | license that they've provided on that model. And that's really

00:14:22.840 | just the tip of the iceberg on what's going on, which is that

00:14:25.760 | these models are very quickly becoming ubiquitous,

00:14:28.000 | commoditized, small, and effectively are able to move to

00:14:31.760 | and be run on the edge of the network. As a result, that means

00:14:35.400 | that it's very hard to see who's using what models how behind the

00:14:39.520 | products and tools that they're building. And so if that's the

00:14:43.400 | trend, then it becomes very hard for a regulatory agency to go in

00:14:48.600 | and audit every server, or every computer or every computer on a

00:14:52.680 | network and say, What model are you running? Is that an approved

00:14:55.080 | model? Is that not an approved model? It's almost like having a

00:14:58.360 | regulatory agency that has to go in and audit and assess whether

00:15:04.120 | a Linux upgrade or some sort of, you know, open source platform

00:15:07.920 | that's that's being run on some server is appropriately vetted

00:15:10.920 | and checked. And so it's almost like a fool's errand. And so if

00:15:16.200 | I'm running one of these companies, and I'm trying to,

00:15:18.560 | you know, get Congress off my butt and get all these

00:15:21.200 | regulators off my butt, I'm going to say, go ahead and

00:15:22.880 | regulate us. Because the truth is, there really isn't a great

00:15:26.480 | or easy path or ability to do that. And there certainly won't

00:15:29.080 | be in five or 10 years. Once these models all move on to the

00:15:33.040 | edge of the network, and they're all being turned around all the

00:15:35.600 | time every day. And there's a great evolution underway. So I

00:15:40.000 | actually take a point of view that it's not just that this is

00:15:44.000 | necessarily bad. And there's cronyism going on. I think that

00:15:47.200 | the point of view is just that this is going to be a near

00:15:50.520 | impossible task to try and track and approve LLM and audit servers

00:15:56.760 | that are running LLM and audit apps and audit what's behind the

00:16:00.720 | tools that everyday people are using. And I wish everyone the

00:16:04.400 | best of luck in trying to do so. But that's kind of the joke of

00:16:07.440 | the whole thing is like, let's go ahead and paddle these

00:16:09.440 | Congress people on the shoulder and say you got it. Right?

00:16:12.360 | There you have it, folks. Wrong answer. shamatha sacks. Right

00:16:16.000 | answer. Friedberg. If you were to look at hugging face, if you

00:16:18.880 | don't know what that is, it's a basically an open source

00:16:21.480 | repository of all of the LLM. The cat is out of the bag, the

00:16:25.720 | horses have left the barn. If you look at what I'm showing on

00:16:29.840 | the screen here, this is the open LLM leaderboard kind of

00:16:33.000 | buried on hugging face. If you haven't been to hugging face,

00:16:34.920 | this is where developers show their work, they share their

00:16:37.600 | work. And they kind of compete with each other in a social

00:16:40.480 | network showing all their contributions. And what they do

00:16:43.680 | here is and this is super fascinating. They have a series

00:16:46.760 | of tests that will take an LLM language model and they will

00:16:51.600 | have it do science questions that would be for grade school.

00:16:55.320 | They'll do a test of mathematics, US history,

00:16:58.520 | computer science, etc.

00:16:59.720 | There's a jeopardy test to I don't know if it's on here, but

00:17:02.520 | the jeopardy test is really good. It's like straight up

00:17:04.240 | jeopardy trivia and see if it can answer the questions. Yeah,

00:17:06.480 | which actually, Friedberg was actually his high school

00:17:09.920 | jeopardy championship three years in a row. But anyway, on

00:17:13.440 | this leaderboard, you can see the language models are out

00:17:18.960 | pacing what open AI did. I'm sorry, closed AI is what I call

00:17:23.480 | it now because they're closed source, closed AI and Bard have

00:17:27.520 | admitted that internal person at Bard said the language models

00:17:30.360 | here are now outpacing what they're able to do with much

00:17:33.280 | more resources. Many hands makes for light work, the open source

00:17:36.760 | models are going to fit on your phone or the latest, you know,

00:17:40.120 | Apple silicon. So I think the cat's out of the bag. I don't

00:17:43.240 | know how they

00:17:44.080 | pull something incompatible about that. What Friedberg just

00:17:46.520 | said with what I said, in fact, freeberg's point bolsters my

00:17:49.200 | point, it's highly impractical to regulate open source software

00:17:54.200 | in this way. Also, when you look at that list of things that

00:17:56.720 | people are doing on hugging face, there's nothing nefarious

00:17:59.160 | about it.

00:18:00.160 | Yeah, and all the harms that were described are already

00:18:02.280 | illegal and can be prosecuted. Exactly. Some special agency,

00:18:05.800 | you know, giving it seal of approval. Again, this is going

00:18:08.800 | to replace permissionless innovation, which is what has

00:18:11.960 | defined the software industry and especially open source with

00:18:15.240 | the need to develop some connection or relationship and

00:18:18.280 | lobbying in Washington to go get your project approved. And

00:18:22.480 | there's no really good reason for this, except for the fact

00:18:24.600 | that the senators on the judiciary committee and all of

00:18:28.080 | Washington really wants more control. So they can get more

00:18:31.280 | donations.

00:18:32.000 | Saks have a question. Do you think that creating the DMV and

00:18:36.320 | requiring a driver's license limits the ability for people to

00:18:39.840 | learn how to drive

00:18:40.720 | the DMV is like the classic example of how government

00:18:44.040 | doesn't work. I don't know why you'd want to make that your

00:18:45.960 | example. I mean, people have to spend all day waiting in line.

00:18:49.440 | He sends people to it. He's got a VIP person waiting in line to

00:18:53.240 | get your photo taken. It's insane. I mean, everyone has a

00:18:55.920 | miserable experience with it. No, but it's highly relevant

00:18:58.360 | because you're right. If you create an agency where people

00:19:01.920 | have to go get their permission, it's a licensing scheme. You're

00:19:05.200 | gonna be waiting in some line of untold length. It won't be like

00:19:09.000 | a physical line at the DMV building, it's gonna be a

00:19:11.560 | virtual line where you're in some queue, where there's

00:19:13.720 | probably gonna be some overwork regulator who doesn't even know

00:19:16.640 | how they're supposed to approve your project. They're just gonna

00:19:19.080 | be trying to cover their ass because if the project ends up

00:19:21.600 | being something nefarious, then they get blamed for it. So

00:19:26.400 | that's what's gonna end up happening.

00:19:27.600 | Let me also highlight something that I think is maybe I think

00:19:30.560 | the gut a little bit, maybe a little bit misunderstood. But,

00:19:34.160 | you know, an AI model is an algorithm. So it's a it's a

00:19:38.760 | piece of software that takes data in and spits data out. And

00:19:43.240 | you know, we have algorithms that are written by humans, we

00:19:45.440 | have algorithms that have been, you know, written by machines,

00:19:48.840 | these are machine learn models, which is what a lot of what

00:19:51.640 | people are calling AI today is effectively an extension of an

00:19:54.720 | out of. And so the idea that a particular algorithm is

00:20:00.400 | differentiated from another algorithm is also what makes

00:20:04.040 | this very difficult, because these are algorithms that are

00:20:06.240 | embedded and sit within products and applications that an end

00:20:10.320 | user and end customer ultimately uses. And I just sent you guys a

00:20:14.240 | link to the you know, the EU has been working towards passing

00:20:18.200 | this AI. Here we go. There are a couple of weeks ahead of these

00:20:22.120 | conversations in the US. But I mean, as you read through this

00:20:26.080 | AI act, and the proposal that it's that it's put forth, it

00:20:29.960 | almost becomes the kind of thing that you say, I just don't know

00:20:34.280 | if these folks really understand how the technology works,

00:20:37.640 | because it's almost as if they're going to audit and have,

00:20:40.720 | you know, an assessment of the risk level of every software

00:20:45.080 | application out there. And that the tooling and the necessary

00:20:48.960 | infrastructure to be able to do that just makes no sense. In the

00:20:52.200 | context of open source software in the context of an open

00:20:54.680 | internet, in the context of how quickly software and applications

00:20:59.240 | and tools evolve, and you make tweaks to an algorithm, and you

00:21:01.520 | got to resubmit it for authorization,

00:21:03.720 | or you can be sure their number one job Friedberg is going to

00:21:07.680 | be to protect jobs. So anything there that in any way infringes

00:21:12.640 | on somebody's ability to be employed in a position, whether

00:21:15.760 | it's an artist or a writer, or a developer, they're going to say,

00:21:19.640 | you can't use these tools, or they're going to try to throttle

00:21:21.880 | them to try to protect jobs, because that's their number one

00:21:24.320 | job. All three of you. Do you guys think that this was Sam's

00:21:27.480 | way of pulling up the ladder behind him? Of course, 100% just

00:21:31.160 | like, Oh, absolutely. It's and it's because no, you can, you

00:21:34.240 | can prove it. He made open AI closed AI by making it not open

00:21:38.640 | source. If you're Sam, you're smart enough to know how quickly

00:21:42.760 | the models are commoditizing and how many different models there

00:21:47.120 | are, that, you know, can provide similar degrees of

00:21:50.040 | functionality, as you just pointed out, J. Cal. So I don't

00:21:52.240 | think it's about trying to lock in your model. I think it's

00:21:54.960 | about recognizing the impracticality of creating some

00:21:57.920 | regulatory regime around model auditing. And so you're very so

00:22:02.840 | that in that in that world, in that scenario, where you have

00:22:05.280 | that vision, you have that foresight, do you go to Congress

00:22:07.880 | and tell them that they're dumb to regulate AI? Or do you go to

00:22:10.400 | Congress and you say, Great, you should regulate AI? Knowing that

00:22:13.520 | it's like, hey, yeah, you should go ahead and stop the sun from

00:22:15.760 | shining. You know, like, it's just Yeah, so basically, he's

00:22:18.640 | telling them to do that. Because he knows they can't. Therefore,

00:22:22.520 | he gets all the points, all the joy points, all the social

00:22:26.200 | credit, I don't wanna say virtual signaling, but he gets

00:22:28.480 | all the credit relationship credit with Washington for

00:22:31.720 | saying what they want to hear reflecting back to them. Even

00:22:34.840 | though he knows they can't compete with Facebook's open

00:22:37.400 | model, which is number one.

00:22:38.720 | Yeah, there is historical precedent interesting for

00:22:41.680 | companies that are facing congressional scrutiny, to go to

00:22:45.000 | Congress and say, go ahead and regulate us as a way of

00:22:47.800 | preemptively. Yeah. And I think that it doesn't necessarily

00:22:50.960 | mean you're going to get regulated. But it's a way of

00:22:53.440 | kind of creating some relief and getting everyone to take a

00:22:55.680 | breather and a sigh of relief and be like, okay, the

00:22:57.360 | industry is with us, you know, they're

00:22:58.560 | guard. What do you think of the strategy? Saks the guard as

00:23:02.080 | strategy he's pulling here?

00:23:03.280 | Let's guard as

00:23:04.400 | I think that's in chess when you are going to take the queen.

00:23:07.280 | What's like, anyway, what do you think of his chess?

00:23:10.200 | That's not a strategy in chess. So I think it is a chess move.

00:23:16.560 | Nonetheless,

00:23:17.800 | is he pulling up the ladder sacks or no,

00:23:19.640 | I don't think that's his number one goal. But I think it is the

00:23:23.720 | result. And so I think the the goal here is I think he's got

00:23:28.680 | to pass in front of him when you go to testify like this, you can

00:23:31.920 | either resist, and they will put you in the hot seat and just

00:23:35.600 | grill you for a few hours, or you can sort of concede and you

00:23:39.720 | buy into their narrative. And then you kind of get through the

00:23:43.520 | hearing without being grilled. And so I think on that level,

00:23:46.680 | it's preferable just to kind of play ball. And then the other

00:23:50.720 | thing is that by playing ball, you get to be part of the

00:23:53.360 | insiders club that's going to shape these regulations. And

00:23:56.040 | that will I wouldn't say it's a ladder coming up, I think it's

00:23:58.320 | more of a moat, where because it's not that the ladder comes

00:24:02.280 | up and nobody else can get in. But the regulations are gonna be

00:24:05.320 | a pretty big moat around major incumbents who know they qualify

00:24:10.120 | for this because we're gonna write these standards. So at the

00:24:13.320 | end of the day, if you're someone in Sam's shoes, you're

00:24:15.940 | like, why resist and make myself a target, or I'll just buy into

00:24:21.280 | the narrative and help shape the regulations. And it's good for

00:24:23.720 | my business.

00:24:24.200 | I like the analysis, gentlemen, this is a perfect analysis. Let

00:24:27.240 | me ask you a question tomorrow. What is the commercial incentive

00:24:31.240 | from your point of view, to ask for regulation and to be

00:24:35.080 | pro regulation, your pro regulation? Can you just

00:24:37.240 | highlight for me at least what you think? You know, the

00:24:40.840 | commercial reason is to do that? You know, how do you benefit

00:24:43.920 | from that? Like, not you personally, but generally, like,

00:24:46.280 | where does benefit arise?

00:24:48.120 | I think that certain people in a sphere of influence, and I would

00:24:52.560 | put us in that category, have to have the intellectual capacity

00:24:56.840 | to see beyond ourselves and ask what's for the greater good. I

00:25:02.720 | think Buffett is right. Two weeks ago, he equated AI to

00:25:08.320 | nuclear weapons, which is an incredibly powerful technology

00:25:11.960 | whose genie you can't put back in the bottle, whose 99.9% of use

00:25:17.600 | cases are generally quite societally positive, but the

00:25:22.480 | point 1% of use cases destroys humanity. And so I think you

00:25:27.400 | guys are unbelievably naive on this topic, and you're letting

00:25:30.880 | your ideology fight your common sense. The reality is that there

00:25:37.480 | are probably 95 billion trillion use cases that are incredibly

00:25:42.720 | positive. But the 1000 negative use cases are so destructive.

00:25:47.200 | And they're equally possible. And the reason they're equally

00:25:50.120 | possible. And this is where I think there's a lot of

00:25:51.960 | intellectual dishonesty here is, we don't even know how

00:25:54.600 | transformers work. The best thing that happened when

00:25:57.840 | Facebook open source llama was also that somebody stealthily

00:26:01.680 | released all the model weights. Yeah. Okay, so

00:26:05.480 | I don't think that a little bit for a neophyte what we're

00:26:08.080 | talking about. So there's the model and there's the weights,

00:26:11.120 | think about it as it's a solution to a problem. The

00:26:13.600 | solution looks like a polynomial equation. Okay, let's take a

00:26:17.480 | very simple one. Let's take Pythagorean theorem, you know, x

00:26:20.440 | squared plus y squared equals z squared. Okay. So if you want to

00:26:23.920 | solve an answer to a problem, you have these weights, you have

00:26:27.280 | these variables, and you have these weights associated with

00:26:30.040 | it, the slope of a line y equals mx plus b, okay, what a computer

00:26:34.320 | does with AI is it figures out what the variables are. And it

00:26:38.280 | figures out what the weights are the answer to identifying images

00:26:43.080 | flawlessly turns out to be 2x plus seven, where x equals this

00:26:48.400 | thing. Now take that example and multiply it by 500 billion

00:26:53.560 | parameters, and 500 billion weights. And that is what an AI

00:26:58.640 | model essentially gives us to it as an answer to a question. So

00:27:03.120 | even when Facebook released llama, what they essentially

00:27:05.880 | gave us was the equation, but not the weights. And then what

00:27:10.320 | this guy did, I think it was an intern, apparently, or somebody,

00:27:12.880 | he just left the weights, so that we immediately knew what

00:27:16.880 | the structure of the equation look like. So that's what we're

00:27:20.920 | basically solving against. But we don't know how these things

00:27:23.760 | work. We don't really know how transformers work. And so this

00:27:26.400 | is my point, when I think you guys are right about the

00:27:30.440 | overwhelming majority of the use cases, but there will be people

00:27:34.280 | who can nefariously create havoc and chaos. And I think you got

00:27:38.800 | to slow the whole ship down to prevent those few folks from

00:27:43.400 | ruining it for everybody.

00:27:44.280 | Let me just talk with you. I haven't had a chance to chime in

00:27:48.400 | on my position. So I'd like to just make my mind. Nobody cares.

00:27:51.760 | Okay, well, I do. I think actually, I split the

00:27:55.200 | difference here a little bit. I don't think it needs to be an

00:27:56.800 | agency and licensing. I do think we have to have a commission. And

00:27:59.800 | we do need to have people being thoughtful about those 1000 use

00:28:02.560 | cases chamath because they are going to cause societal harm or

00:28:07.320 | things that we cannot anticipate. And then number two,

00:28:10.160 | for the neophyte with the 1600 rating on chess.com sacks,

00:28:14.440 | Gardez, an announcement to the opponent that their queen is

00:28:17.840 | under direct attack similar to the announcement of check the

00:28:19.940 | warning was customary until the early 20th century. So since you

00:28:23.000 | do not know the history of check, now you've learned

00:28:25.200 | something early 20th century. Okay, well, since I've only

00:28:27.720 | played chess in the 20th and 21st centuries, I'm unaware of

00:28:30.640 | that. Jekyll and Francis Brown's got the girl they play man if he

00:28:36.040 | got the go ahead free bird in the context of what we're

00:28:39.200 | talking about that models are becoming smaller and can be run

00:28:43.320 | on the edge. And there's obviously hundreds and thousands

00:28:45.560 | of variants of these open source models that have, you know, good

00:28:49.320 | effect, and perhaps compete with some of these, these models that

00:28:53.880 | you're mentioning that are closed source. How do you

00:28:56.760 | regulate that? How do you and then they sit behind an

00:28:59.260 | application? I sit behind tooling? Yeah.

00:29:01.520 | I think in order for you to be able to compile that model to

00:29:05.240 | generate that initial instantiation, you're still

00:29:08.240 | running it in a cluster of 1000s of GPUs.

00:29:11.240 | But let's say we're past that.

00:29:13.080 | You can't be past that. We're not past that yet. Okay, we

00:29:16.040 | don't have 5 million models, we don't have all kinds of things

00:29:20.280 | that solve all kinds of problems. We don't have an open

00:29:22.680 | source available simulation of every single molecule in the

00:29:25.880 | world, including all the toxic materials that could destroy

00:29:28.200 | humans. We don't have that yet. So before that is created, and

00:29:32.080 | shrunk down to an iPhone, I think we need to put some stage

00:29:35.240 | gates up to slow people down.

00:29:36.680 | What do you mean by those stage gates? Yeah,

00:29:38.080 | I think you need some form of KYC. I think before you're

00:29:41.040 | allowed to run a massive cluster to generate the model that then

00:29:45.440 | you try to shrink, you need to be able to show people that

00:29:48.600 | you're not trying to do something absolutely chaotic, or

00:29:51.280 | have it creating that I don't think

00:29:53.360 | that that could be as simple as putting your driver's license

00:29:55.920 | and your social security number that you're working on an

00:29:58.560 | instance in a cloud, right? It could be you're putting your

00:30:01.520 | name on your work,

00:30:02.280 | it becomes slightly more nuanced than that. It's like, I think

00:30:05.520 | that J Cal, that's probably the simplest thing for AWS GCP and

00:30:08.880 | as you're to do, which is that if you want to run over a

00:30:12.720 | certain number of GPU clusters, you need to put in that

00:30:15.200 | information. I think you also need to put in your tax ID

00:30:17.680 | number. So I think if you want to run a real high scale model,

00:30:20.280 | that's still going to run you 10s or hundreds of millions of

00:30:23.440 | dollars. I do think there aren't that many people running those

00:30:26.720 | things. And I do think it's easy to police those and say, what

00:30:29.640 | are you trying to do here?

00:30:30.480 | So let me just push back on that. Because mosaic ml publish

00:30:33.920 | this model that is, let me I can pull up the performance chart

00:30:38.040 | or Nick, maybe you can just find it on their website real quick

00:30:40.080 | at the new model they published to mock bait, trained this model

00:30:43.560 | on open source data that's publicly available. And they

00:30:49.240 | spent $200,000 on a cluster run to build this model and look at

00:30:53.160 | how it performs compared to some of the top models that are

00:30:55.880 | closed source.

00:30:56.640 | Just say it for the people who are listening.

00:30:58.360 | Yeah. So for people that are listening, basically, this

00:31:00.360 | model is called mpt seven B. That's the name of the AI model,

00:31:05.880 | the LLM model that was generated by this group called mosaic ml.

00:31:09.600 | And they spent $200,000 creating this model from scratch. And the

00:31:15.000 | data that they trained it on is all listed here. It's all

00:31:17.520 | publicly available data that you can just download off the

00:31:19.720 | internet, then they score how well it performs on its results

00:31:23.160 | against other big models out there like llama seven B.

00:31:26.600 | Yeah, but he I know, but I don't exactly know what the actual

00:31:31.080 | problems they're trying to ask it to compare.

00:31:33.640 | Right. So the but the point is that this model theoretically

00:31:36.760 | could then be applied to a different data set once it's

00:31:39.720 | been, you know, built. I don't think so. I just want to use

00:31:43.320 | your point earlier about toxic chemistry, because models were

00:31:46.600 | generated, and then other data was then used to fine tune those

00:31:50.520 | models and deliver an output.

00:31:52.120 | Hold on a second, like those answers were to specific kinds

00:31:55.240 | of questions. If you wanted to all of a sudden ask totally

00:31:58.640 | orthogonal thing of that model, that model would fail, you would

00:32:01.960 | have to go back and you'd have to retrain it. That training does

00:32:05.280 | cost some amount of money. So if you said to me, h math, I could

00:32:09.200 | build you a model trained on the universe of every single

00:32:12.760 | molecule in the world. And I could actually give you

00:32:15.400 | something that could generate the toxic list of all the

00:32:17.600 | molecules and how to make it for $200,000. I would be really

00:32:21.600 | scared. I don't think that that's possible today. So I

00:32:24.680 | don't understand these actual tests. But I don't think it's

00:32:27.360 | true that you could take this model and these model weights,

00:32:30.520 | apply it to a different set of data and get useful answers.

00:32:33.240 | But let's let's assume for a minute that you can, in fact,

00:32:36.400 | take $200,000 worth of

00:32:38.280 | I don't want to assume it. Here's my point. I want to tell

00:32:41.560 | you what's happening right now, which is that's not possible. So

00:32:44.600 | we should stop so that then I don't have to have this argument

00:32:47.800 | with you in a year from now, which is like, hey, some jack

00:32:50.880 | jerk off just created this model. Now the cat's out of the

00:32:53.560 | bag. So let's not do it. Yeah. And, and then what's going to

00:32:56.520 | happen is like some chaotic seeking organization is going to

00:32:59.960 | print one of these materials and release it into the wild to

00:33:02.640 | prove it. But here's the point for the audience. We are at a

00:33:04.960 | moment in time where this is moving very quickly. And you

00:33:07.480 | have very intelligent people here who are very knowledgeable.

00:33:10.160 | Talking about the degree to which this is going to manifest

00:33:14.480 | itself, not if it will manifest, you are absolutely 100% certain

00:33:18.440 | freeberg that somebody will do something very bad in terms of

00:33:22.200 | the chemical example as but one, it we're only determining here

00:33:26.120 | what level of hardware and what year that will happen to Martha

00:33:29.080 | saying, we know it's going to happen, whether it's two or 10,

00:33:32.400 | or, you know, five years, let's be thoughtful about it. And I

00:33:36.400 | think, you know, this discussion we're having here, I think is

00:33:39.040 | on a spectrum. It's this is a unique moment where the most

00:33:45.080 | knowledgeable people across every single political spectrum

00:33:49.680 | persuasion for profit, nonprofit, Democrat, Republican,

00:33:53.280 | right, Elon and Sam, I'll just use those as the two canonical

00:33:56.320 | examples to demonstrate our pro regulation, and then the further

00:34:00.600 | and further you get away, the less technically astute you are,

00:34:03.880 | the more anti regulation and like pro market you are. And all

00:34:07.360 | I'm saying is, I think that should also be noted that that's

00:34:10.600 | a unique moment that the only other time that that's happened

00:34:13.080 | was around nuclear weapons. And you know, that's when I actually

00:34:18.120 | think, I think it's politically incorrect. Right now, I think

00:34:23.520 | because of what you're saying, just give me a second, I think

00:34:25.320 | because of what you're saying, everyone on the left and the

00:34:27.640 | right, it's it's become popular to be pro regulation on AI and

00:34:31.560 | say that AI is going to doom the world. And it's unpopular. And

00:34:34.760 | I'm saying,

00:34:35.240 | no, I'm saying,

00:34:36.960 | I've explained my point of view on Sam, Elon's different, but I

00:34:40.040 | think like it, I think it's become politically incorrect to

00:34:43.080 | stand up and say, you know what, this is a transformative

00:34:45.480 | technology for humanity. I don't think that there's a real path

00:34:48.400 | to regulation. I think that there are totally there are laws

00:34:51.160 | that are in place that can protect us in other ways, with

00:34:53.560 | respect to privacy, with respect to fraud, with respect to

00:34:56.640 | biological warfare, and all the other things that we should

00:34:58.800 | worry about.

00:34:59.360 | Elon has said pretty clearly, he doesn't give a shit about what

00:35:02.120 | it does to make money or not. He cares about what he thinks. So

00:35:05.200 | all I'm saying is that's a guy that's not trying to be

00:35:06.920 | politically correct.

00:35:07.880 | Elon has a very specific concern, which is a GI, he's

00:35:11.800 | concerned that we're on a path to a digital super intelligence

00:35:14.400 | singularity. And if we create the wrong kind of artificial

00:35:19.040 | general intelligence that decides that it doesn't like

00:35:21.720 | humans, that is a real risk to the human species. That's the

00:35:26.200 | concern he's expressed. But that's not what the hearing was

00:35:30.800 | really about. And it's not what any of these regulatory

00:35:33.400 | proposals are about. The reality is none of these centers know

00:35:37.880 | what to do about that even the industry doesn't know what to

00:35:40.440 | do about the long term risk of creating an AGI.

00:35:43.920 | Nobody knows.

00:35:44.760 | Nobody really. And so and so I actually I disagree with this

00:35:47.960 | idea that tomorrow, the early earlier you said that there's 1000

00:35:51.840 | use cases here that could destroy the human species. I

00:35:54.680 | think there's only one, there's only one species level risk,

00:35:58.560 | which is AGI. But that's a long term risk. We don't know what to

00:36:01.480 | do about it yet. I agree, we should have conversations. What

00:36:04.680 | we're talking about today is whether we create some new

00:36:07.360 | licensure regime in Washington, so that politically connected

00:36:11.400 | insiders get to control and shape the software industry. And

00:36:15.360 | that's a disaster. Let me give you another detail on this. So

00:36:18.080 | in one of the chat groups, I'm in, there was somebody who just

00:36:23.360 | got back from Washington, I won't say who they are. It's not

00:36:26.480 | someone who's famous outside the industry, but they're kind of

00:36:28.280 | like a tech leader. And what they said is they just got back

00:36:32.560 | from Capitol Hill and the White House. And I guess there's like

00:36:36.040 | a White House summit on AI. You guys know about that? Yeah. So

00:36:41.840 | what this person said is that the White House meeting was

00:36:46.000 | super depressing. Some smart people were there to be sure,

00:36:50.520 | but the White House and VPS teams were rapidly negative, no

00:36:54.520 | real concern for the opportunity, or economic impact

00:36:57.600 | just super negative. Of course, basically, the mentality was

00:37:00.640 | that tech is bad. We hate social media. This is the hot new

00:37:04.120 | thing. We have to stop it. Of course, that basically is their

00:37:07.000 | attitude. They don't understand the technology White House.

00:37:09.480 | Yeah, the Y and the VP specifically, because she's the

00:37:11.960 | now the AI is our to put Kamala Harris in charge of this makes

00:37:15.280 | no sense. I mean, does she have any background in this? Like it

00:37:18.400 | just shows like a complete utter lack of awareness. Where's the

00:37:21.960 | Megan Smith, or somebody like a CTO to be put in charge of

00:37:25.520 | this? Remember, Megan Smith was CTO under, I guess Obama, like,

00:37:29.160 | you need somebody with a little more depth of experience here,

00:37:31.640 | like hopefully,

00:37:32.280 | then you guys then saying you're you're pro regulation,

00:37:35.200 | depending on who's in charge.

00:37:36.920 | Well, I'm pro thoughtfulness. I'm pro thoughtfulness. I'm

00:37:39.800 | illustrating that really, this whole new agency that's being

00:37:43.320 | discussed is just based on vibes. You're not down with the

00:37:45.800 | vibes.

00:37:46.120 | The vibes

00:37:46.840 | are the vibes, the vibe is that a bunch of people in Washington

00:37:51.200 | don't understand technology, and they're afraid of it. So

00:37:54.040 | anything you're afraid of, you're gonna want to control

00:37:56.040 | these are socialists, David, they're socialists. They hate

00:37:59.240 | progress. They are scared to death that jobs are going to

00:38:02.040 | collapse. They're socialists, they, they're union leaders.

00:38:05.520 | This is their worst nightmare. Because the actual truth of

00:38:08.560 | this technology is 30% more efficiency, and it's very

00:38:11.960 | mundane. This is the truth here. I think that representatives

00:38:15.680 | have 30% more efficiency means Google, Facebook, and many other

00:38:20.360 | companies, finance, education, they do not add staff every

00:38:24.360 | year, they just get 30% more efficient every year. And then

00:38:27.320 | we see unemployment go way up, and Americans are going to have

00:38:30.120 | to take service jobs. And white collar jobs are going to be

00:38:33.080 | refined to like a very elite few people who actually do work in

00:38:37.720 | the world. There is absolutely a lot of new companies. If humans

00:38:41.960 | can become if knowledge workers can become 30% more productive,

00:38:44.800 | there'll be a lot of new companies. And the biggest

00:38:47.120 | shortage our economy is coders, right? And we're going to have

00:38:49.800 | an unlimited number of them. Now. They're all going to go.

00:38:51.560 | Yeah, I don't know if it's unlimited. But yes, it's a good

00:38:53.560 | thing. If you give them superpowers. We've talked about

00:38:55.280 | this before. Yeah. So I think it's too soon to be concluding

00:38:59.000 | that we need to stop job displacement that hasn't even

00:39:01.880 | occurred yet. I'm not saying it's actually going to happen. I

00:39:04.360 | do agree, there'll be more startups. I'm seeing it

00:39:06.000 | already. I just think that's what they fear. That's their

00:39:09.320 | fear is and that's the fear of the EU, the EU is going to be

00:39:12.560 | protectionist unionist, protect pro workers, which is fine,

00:39:15.760 | they're gonna be affected because these are not blue

00:39:17.640 | collar jobs we're talking about. These are knowledge. There's

00:39:19.680 | white collar unions that all the media companies created unions

00:39:22.600 | and look at them. They're all going circling all these media

00:39:24.800 | companies are circling the training going. But that's on

00:39:27.200 | the margins. I mean, that's not just trying to start tech

00:39:29.360 | unions. Sure, they're trying to start them. But when we think of

00:39:31.600 | unionized workers, you're thinking about factory workers,

00:39:34.120 | and these people are not affected.

00:39:36.200 | Okay, listen, this has been an incredible debate. This is why

00:39:39.120 | you tune into the pod. A lot of things can be true at the same

00:39:42.080 | time. I really think the analogy of the the atom bomb is really

00:39:44.800 | interesting, because what you want is scared about with

00:39:47.880 | general artificial intelligence is nuclear holocaust, the whole

00:39:51.000 | planet blows up. Between those two things are things like

00:39:54.600 | Nagasaki and Hiroshima or a dirty bomb and many other

00:39:58.560 | possibilities with nuclear power, you know, Fukushima, you

00:40:02.960 | know, etc. So let's

00:40:04.040 | send no Hiroshima.

00:40:05.320 | Not yet. Right. And you know, the question is, is a three mile

00:40:09.880 | island is a Fukushima is a Nagasaki are those things

00:40:13.920 | probable. And I think we are all looking at this saying there

00:40:17.080 | will be something bad that will happen. There will be the

00:40:19.360 | equivalent. All right.

00:40:20.760 | It strings together and these large language models string

00:40:24.480 | together words in really interesting ways. And they give

00:40:26.560 | computers the ability to have a natural language interface that

00:40:29.160 | is so far from AGI. I think it's a component. Hold on, I think

00:40:34.640 | it's a obviously the ability to understand language and

00:40:37.400 | communicate in a natural way is a component of a future AGI. But

00:40:41.840 | by itself, these are models for stringing together

00:40:44.640 | auto GPT, where these things go out and pursue things without

00:40:49.080 | any interference, I would be the first one to say that if you

00:40:51.840 | wanted to scope models to be able to just do human language

00:40:56.920 | back and forth on the broad open internet, you know, there's

00:41:02.320 | probably a form David where these chat GPT products can

00:41:06.280 | exist. I don't I think that those are quite benign. I agree

00:41:09.080 | with you. But I think what Jason is saying is that every week

00:41:12.400 | you're you're taking a leap forward. And already with auto

00:41:15.760 | GPT is you're talking about code that runs in the background

00:41:18.240 | without supervision. It's not a human interface that's like,

00:41:21.240 | hey, show me how to color my cookies green for St. Patty's

00:41:25.440 | Day. It's a plan my trip to Italy. Yeah, yeah, it's not

00:41:28.320 | doing that. So I just think that there's there's a place well

00:41:31.240 | beyond what you're talking about. And I think you're

00:41:33.000 | minimizing the problem a little bit by just kind of saying the

00:41:35.920 | whole class of AI is just chat GPT and asking, kid asking it to

00:41:40.160 | help it with its home.

00:41:41.040 | This example, I hate to say it out loud. But somebody could say

00:41:44.240 | here is the history of financial crimes that were committed and

00:41:48.280 | other hacks, please with their own model on their own server,

00:41:51.840 | say, please come up with other ideas for hacks, be as creative

00:41:55.360 | as possible and steal as much money as possible and put that

00:41:57.960 | in an auto GPT, David, and study all hacks that occur in the

00:42:01.800 | history of hacking. And it could just create super chaos around

00:42:05.160 | the world.

00:42:05.640 | And you can be sure history is going to regret buying into this

00:42:08.320 | narrative because the members of the Judiciary Committee are

00:42:11.080 | doing the same playbook they ran back in 2016. After that

00:42:14.320 | election, they ran all these hearings on disinformation,

00:42:16.920 | claiming that social networks been used to hack the election.

00:42:19.400 | It was all a phony narrative. Hold on that. What happened?

00:42:22.400 | They got tech companies to buy into Biden, they Russians hacked

00:42:25.800 | Hunter right after that. Stop, they got they got all these tech

00:42:28.920 | CEOs to buy into that phony narrative. Why? Because it's a

00:42:32.240 | lot easier for the tech CEOs to agree and tell the senators what

00:42:35.560 | they want to hear to get them off their backs. And then what

00:42:38.200 | does that lead to a whole censorship industrial complex.

00:42:41.320 | So we're going to do the same thing here. We're going to buy

00:42:44.240 | into these phony narratives to get the senators off our backs.

00:42:47.960 | And that's going to create this giant AI industrial complex

00:42:52.080 | that's going to slow down real innovation and be a burden on

00:42:54.880 | entrepreneurs.

00:42:55.440 | Okay, lightning round lightning round. We got to move on three

00:42:57.600 | more topics I want to hit. Let's keep going.

00:42:59.200 | If honored to become an evil more evil. Yeah. What is it

00:43:02.920 | called an evil comic book character? A supervillain, a

00:43:07.080 | supervillain. I would you want to be an even more loathsome

00:43:10.520 | supervillain continue, I would take every single virus patch

00:43:15.360 | that's been developed and publicized, learn on them and

00:43:18.800 | then find the next zero day exploit on a whole bunch of

00:43:21.160 | stuff. I mean, it's like even publish that idea. Please don't

00:43:23.400 | I mean, I'm worried about publishing that idea. That's

00:43:25.880 | not an intellectual leap. I mean, you have to be an hour.

00:43:29.280 | Obvious. Okay, let's move on. Another great debate. Elon hired

00:43:33.680 | a CEO for Twitter. Linda yakarino. I'm hoping pronouncing

00:43:37.840 | that correct was the head of ad sales at NBC Universal. She's a

00:43:40.440 | legend in the advertising business. She worked at Turner

00:43:43.760 | for 15 years before that she is a workaholic is what she says

00:43:47.840 | she's going to take over everything but product and CTO

00:43:51.240 | Elon's gonna stick with that. She seems to be very moderate

00:43:57.040 | and she follows people on the left or right people are

00:43:59.400 | starting the character assassination and trying to

00:44:01.640 | figure out her politics. She was involved with the WF World

00:44:06.800 | Economic Forum, which anybody in business basically does. But

00:44:10.400 | your take sacks on this choice for CEO and what this means just

00:44:16.360 | broadly for the next six months because we're sitting here at

00:44:20.160 | six months almost exactly since you on took over obviously you

00:44:23.040 | and I were involved in month one but but not much after that.

00:44:25.480 | What do you think the next six months holds and what do you

00:44:27.480 | think her role is going to be obviously some precedent this

00:44:29.840 | for this with you know, when at SpaceX?

00:44:33.120 | Listen, I think this choice makes sense on this level

00:44:37.400 | Twitter's business model is advertising. Elon does not like

00:44:40.960 | selling advertising. She's really good at selling

00:44:43.200 | advertising. So he's chosen a CEO to work with who's highly

00:44:46.640 | complimentary to him in their skill sets and interests. And I

00:44:52.000 | think that makes sense. I think there's a lot of logic in that

00:44:54.640 | what Elon likes doing is the technology and product side of

00:44:58.240 | the business. He actually doesn't really like the let's

00:45:03.880 | call it the standard business chores. And especially related

00:45:07.600 | to like we said advertising and he loves to offer that stuff.

00:45:10.720 | Advertisers make is his personal nightmare. Right? So I think the

00:45:15.440 | choice makes sense on that level. Now. Instantly, you're

00:45:18.040 | right, her hiring led to attacks from both the left and the right

00:45:22.360 | the right, you know, pointed out her views on COVID and vaccines

00:45:27.160 | and her work with the WF. And on the left, I mean, the attack is

00:45:31.560 | that she's following lives at Tick Tock, which you're just not

00:45:33.840 | allowed to do apparently,

00:45:35.040 | a follow is not an endorsement.

00:45:36.920 | Well, if you're just following lives to tick tock, they want to

00:45:39.320 | say you're some crazy right winger now.

00:45:41.000 | Well, she also follows David Sachs. So that does mean that

00:45:44.280 | she's pretty that that is a signal. But the truth is, if

00:45:47.280 | Sachs correct me if I'm wrong here, or trim off, maybe I'll

00:45:49.360 | send it to you. If you pick somebody that both sides

00:45:51.800 | dislike or trying to take apart, you probably picked the right

00:45:54.320 | person. Yeah.

00:45:54.920 | Here's what I think. Okay, go ahead. We're not going to know

00:45:59.320 | how good she is for six to nine months. But here's what I took a

00:46:03.200 | lot of joy out of. Here's a guy who gets attacked for all kinds

00:46:10.440 | of things now, right? He's an anti Semite, apparently. And then

00:46:13.800 | he had to be like, I'm very pro Semite. He's a guy that all of

00:46:16.720 | a sudden people think is a conspiracy theorist. He's a guy

00:46:19.200 | that people think is now on the raging right, all these things

00:46:23.080 | that are just like inaccuracies, basically firebombs thrown by

00:46:27.280 | the left. But here's what I think is the most interesting

00:46:30.440 | thing for a guy that theoretically is supposed to be a

00:46:33.560 | troll and everything else. He has a female chairman at Tesla,

00:46:37.480 | a female CEO at Twitter, and a female president at SpaceX. Of

00:46:42.560 | course, it's a great insight. It's the same insight. I think a

00:46:46.240 | lot of these virtue signaling lunatics on the left virtue

00:46:49.160 | signaling mids. They're all nice. And you know, like, you

00:46:53.120 | have Elizabeth Warren and Bernie Sanders, giving, you know, the

00:46:58.920 | CEO of Starbucks a hard time when he doubled the pay of the

00:47:04.480 | minimum wage, gave them a free bird. You love mid right? That's

00:47:08.440 | a great term, isn't it? These fucking meds and I paid for the

00:47:13.240 | college tuition. What gives you the right at Starbucks to pay

00:47:17.280 | for college tuition and double the minimum wage? That's so so

00:47:20.680 | mad. I don't know why it's so funny to me.

00:47:23.000 | Isn't it so great? Like, you can picture them when I say that

00:47:26.520 | these are these mids feverishly typing on their keyboards, their

00:47:30.080 | virtue signaling nonsense, sex wrapping it up.

00:47:32.480 | Yeah, look, like you said, Elon has worked extremely well with

00:47:36.480 | when shot well, who's the president of SpaceX for a long

00:47:38.680 | time. And I think that relationship shows the way to

00:47:44.240 | make it work here at Twitter, which is they have a very

00:47:46.280 | commoner skill set. I think my understanding is that when

00:47:48.640 | focuses on the business side and the sales side of the operation,

00:47:52.280 | Elon focuses on product and technology. She lets Elon be

00:47:55.280 | Elon. I think if Linda tries to reign Elon in, tell him not to

00:48:00.200 | tweet, or tries to meddle in the free speech aspects of the

00:48:04.040 | business, which is the whole reason he bought Twitter, which

00:48:06.720 | is the beginning and end of it. Yeah, that's right. That's when

00:48:09.400 | it will fall apart. So my advice would be let Elon be Elon. You

00:48:13.760 | know, he bought this company to make it a free speech platform.

00:48:16.640 | Don't mess with that. And I think it could work great. And

00:48:20.680 | a free speech platform. It is when you are saying anything

00:48:23.600 | about COVID. And I really don't even want to say it here,

00:48:26.000 | because I don't want our to even say the word COVID or vaccine

00:48:29.240 | means that this could get tagged by YouTube and be D. You know,

00:48:33.400 | algorithm, the algorithm could could could D. I don't know what

00:48:37.840 | they call it deprecate this and when we don't show up and people

00:48:40.360 | don't see us because we just said the word COVID. I mean, the

00:48:42.640 | censorship built into these algorithms is absurd. But

00:48:45.480 | speaking of absurd, Lena Khan, who has been the least effective

00:48:50.160 | FTC chair, I think, started out pretty promising with some

00:48:54.660 | interesting ideas. She's now moved to block a major pharma

00:48:57.640 | deal. In December, Amgen agreed to acquire Dublin based horizon

00:49:02.980 | therapeutic for 27.8 billion. This is the largest pharma deal

00:49:06.040 | announced in 2022. FTC has filed a lawsuit in federal court

00:49:08.920 | seeking an injunction that would prevent the deal from closing.

00:49:11.600 | The reasoning is the deal would allow Amgen to entrench the

00:49:14.860 | monopoly positions of horizons I and gout drugs. The agency said

00:49:19.680 | that those treatments don't face any competition today and that

00:49:22.840 | Amgen would have a strong incentive to prevent any

00:49:25.840 | potential rivals from introducing similar drugs.

00:49:28.880 | Chamath the pharmaceutical industry is a little bit

00:49:32.080 | different than the tech

00:49:32.800 | here insanity.

00:49:34.080 | Explain why and then sacks will go to you on the gout stuff.

00:49:37.680 | Well, I know that personally impacts you. Go ahead.

00:49:39.520 | I think that this is a little like scientifically illiterate,

00:49:44.040 | to be honest. Unpack the thing is that you want drugs that can

00:49:48.600 | get to market quickly. But at the same time, you want drugs to

00:49:51.600 | be safe, and you want drugs to be effective. And I think that

00:49:55.480 | the FDA has a pretty reasonable process. And one of the direct

00:50:00.360 | byproducts of that process is that if you have a large

00:50:03.040 | indication that you're going after, say diabetes, you have to

00:50:06.540 | do an enormous amount of work, it has to be run on effectively

00:50:10.440 | 1000s of people, you have to stratify it by age, you have to

00:50:13.700 | stratify it by gender, you have to stratify by race, you have to

00:50:16.960 | do it across different geographies, right? The bar is

00:50:19.720 | high. But the reason the bar is high is that if you do get

00:50:22.580 | approval, these all of a sudden become these blockbuster 1020

00:50:26.660 | $30 billion drugs, okay. And they improve people's lives, and

00:50:31.040 | they allow people to live, etc, etc. What has happened in the

00:50:35.200 | last 10 or 15 years, because of Wall Street's influence inside

00:50:39.480 | of the pharma companies, is that what pharma has done a very good

00:50:45.440 | job of doing is actually pushing off a lot of this very risky R&D

00:50:51.320 | to young early stage biotech companies. And they typically do

00:50:56.080 | the first part of the work, they get through a phase one, they

00:50:59.720 | even may be able to sometimes go and start a phase two trial to a

00:51:03.640 | trial. And then they typically can get sold to pharma. And

00:51:09.080 | these are like multi billion dollar transactions. And the

00:51:12.920 | reason is that the private markets just don't have the

00:51:15.160 | money to support the risk for these companies to be able to do

00:51:19.040 | all the way through a phase three clinical trial because it

00:51:21.260 | would cost on some cases 5678 billion dollars. You've never

00:51:26.080 | heard of a tech company raising that much money except in a few

00:51:28.520 | rare cases. In far in biotech, it just doesn't happen. So you

00:51:33.080 | need the M&A machine to be able to incentivize these young

00:51:37.200 | companies to even get started in the first place. Otherwise, what

00:51:40.800 | literally happens is you have a whole host of diseases that just

00:51:44.520 | stagnate. Okay. And instead, what happens is a younger

00:51:48.920 | company can only raise money to go after smaller diseases, which

00:51:53.000 | have smaller populations, smaller revenue potential,

00:51:56.040 | smaller costs, because the trial infrastructure is just less. So

00:52:00.200 | if you don't want industry to be in this negative loop, where you

00:52:04.080 | only work on the small diseases, and you actually go and tackle

00:52:06.720 | the big ones, you need to allow these kinds of transactions

00:52:10.160 | happen. The last thing I'll say is that even when these big

00:52:12.720 | transactions happen, half the time they turn out to still not

00:52:16.480 | work, there is still huge risk. So don't get caught up in the

00:52:20.120 | dollar size. You have to understand the phase it's in.

00:52:22.800 | And the best example of this is the biggest outcome in in

00:52:26.120 | biotech private investing in Silicon Valley was this thing

00:52:29.160 | called stem centric. And that thing was a $10 billion dud,

00:52:32.200 | right. But it allowed all these other companies to get started

00:52:35.800 | after some centric Scott bought for 10 billion.

00:52:37.880 | Freiburg, I want to get your take on this, especially in

00:52:41.640 | light of maybe something people don't understand, which is that

00:52:44.520 | the amount of time you get to actually exclusively monetize a

00:52:48.680 | drug, because my understanding you correct me if I'm wrong, you

00:52:52.520 | get a 20 year patent, it's from the date you file it, but then

00:52:55.320 | you're working towards getting this drug approved by the FDA.

00:52:57.840 | So by the time the FDA approves the drug, this 20 year patent

00:53:00.920 | window, how many years do you actually have exclusively to

00:53:03.880 | monetize that drug? And then your wider thoughts on this?

00:53:07.000 | FTA? Yeah, I'm not going to answer that question right now.

00:53:10.000 | Because I do want to kind of push back on the point I'm

00:53:12.960 | generally pretty negative on a lot of the comments Lena

00:53:16.240 | contemade and her positioning. And obviously, as you guys know,

00:53:18.760 | we've talked about it on the show. But I read the FTC filing

00:53:22.480 | written federal court. And if you read the filing, let me just

00:53:28.040 | start the company that am gens trying to buy it's called

00:53:29.800 | horizon therapeutics, which is a company that's doing about 4

00:53:33.000 | billion in revenue a year about a billion to a billion and a

00:53:35.360 | half in EBITDA. So it's a it's a business that's got a portfolio

00:53:39.200 | of orphan drugs, meaning drugs that treat orphan conditions

00:53:42.240 | that aren't very big blockbusters in the in the

00:53:44.320 | pharmaceutical drug context. And so it's a it's a nice portfolio

00:53:48.480 | of cash generating drugs. And Jen, buying the business gives

00:53:53.320 | them real revenue really, but and helps bolster a portfolio

00:53:57.920 | that you know, is aging. And I think that's a big part of the

00:54:01.240 | strategic driver for am Jen to make this massive $28 billion

00:54:04.920 | acquisition, the FTC is claim in the filing, which I actually

00:54:10.200 | read and I was like, this is actually a pretty good claim is

00:54:13.840 | that the way that am Jen sets the prices for their

00:54:16.760 | pharmaceutical drugs is they go to the insurance companies, the

00:54:19.720 | payers and the health systems, and they negotiate drug pricing.

00:54:23.160 | And they often do bulk multi product deals. So they'll say,

00:54:27.760 | hey, we'll give you access to the product at this price point,

00:54:31.320 | but we need you to pay this price point for this product. And

00:54:33.800 | over time, that drives price inflation, it drives costs up.

00:54:37.160 | And it also makes it difficult for new competitors to emerge.

00:54:39.680 | Because they tell the insurance company, you have to pick our

00:54:42.400 | drug over other drugs in order to get this discounted price. And

00:54:46.520 | so it's a big part of their negotiating strategy that they

00:54:48.520 | do with insurance companies. So the FTC is claim is that by

00:54:52.080 | giving am Jen this large portfolio of drugs that they're

00:54:55.240 | buying from horizon, it's going to give them more negotiating

00:54:58.040 | leverage and the ability to do more of this drug blocking that

00:55:00.880 | they do with insurance companies and other payers in the drug

00:55:03.920 | system. So they're trying to prevent pharmaceutical drug

00:55:06.760 | price inflation, and they're trying to increase competition

00:55:09.280 | in their lawsuit. So I felt like it was a fairly kind of

00:55:12.040 | compelling case. I'm no lawyer on antitrust and monopolistic

00:55:16.600 | practices and the Sherman Act. But this was not sorry, let me

00:55:18.680 | just say this was not an early stage biotech risky deal that

00:55:22.160 | they're trying to block. No, it is a mature company with

00:55:25.080 | 4 billion in revenue and a billion and a half in EBITDA. I

00:55:27.280 | understand. And I read it too. But two comments. It is because

00:55:31.440 | the people that traffic in these stocks are the same ones that

00:55:34.200 | fund these early stage biotech companies. And I talked to a

00:55:37.160 | bunch of them. And they're like, if these guys block this kind of

00:55:39.680 | deal, we're going to get out of this game entirely. So just from

00:55:43.560 | the horse's mouth, what I'm telling you is you're going to

00:55:45.320 | see a Paul come over the early stage venture financing

00:55:48.640 | landscape, because a lot of these guys that are crossover

00:55:51.280 | investors that own a lot of these public biotech stocks that

00:55:53.880 | also fund the private stocks will change their risk posture

00:55:57.480 | if they can't make money. That's just the nature of capitalism.

00:56:00.360 | The second thing is Lena Khan did something really good about

00:56:03.680 | what you're talking about this week, actually, which is she

00:56:06.280 | actually went after the PBMs. And if you really care about

00:56:09.600 | drug inflation, and you follow the dollars, the real culprits

00:56:12.800 | around this are the pharmacy benefit managers. And she

00:56:16.160 | actually launched a big investigation into them. But

00:56:18.960 | this is what speaks to the two different approaches. It seems

00:56:21.720 | that unfortunately, for the FTC, every merger just gets contested

00:56:26.080 | for the sake of it being contested. Because I think that

00:56:29.280 | if you wanted to actually stop price inflation, there are

00:56:31.920 | totally different mechanisms. Because why didn't you just sue

00:56:35.280 | all the PBMs? Well, there's no merger to be done. But you can

00:56:39.000 | investigate, and then you could regulate. And I think that

00:56:42.040 | that's probably a more effective way. And the fact that she

00:56:44.640 | targeted the PBM says that somebody in there actually

00:56:47.240 | understands where the price inflation is coming from. But I

00:56:50.520 | don't think something like an Amgen horizon, because what I

00:56:53.560 | think will happen is all the folks will then just basically

00:56:56.640 | say, Well, man, if if, if these kinds of things can't get

00:56:59.720 | bought, then why am I funding these other younger things? Yeah,

00:57:02.040 | but you know,

00:57:02.440 | yeah, we're just not seeing a lot of the younger stuff get

00:57:05.440 | get blocked. I don't think we've seen any attempts at blocking

00:57:09.920 | speculative portfolio acquisitions or speculative

00:57:12.360 | company acquisitions. So

00:57:13.560 | I think these guys are getting caught up in the dollar number.

00:57:15.800 | You know, yeah. So I think the problem is they see 28 billion,

00:57:19.160 | they're like, we need to stop it. You know, it's amazing. I'll

00:57:21.760 | just wrap on this, because it's a good discussion. But I think

00:57:23.840 | we got to keep moving here is I took the PDF that you shared.

00:57:27.320 | And I put it into chat GPT now. Oh, and you don't need to upload

00:57:32.960 | the PDF anymore. You could just say summarize this and put the

00:57:35.600 | link. And it did it instantly.

00:57:38.080 | Are using the browsing plugin or

00:57:41.240 | I just use it. No, it's not the browser plugin. I just did. This

00:57:44.400 | is the game. But 3.5 model, I just gave it a link and it

00:57:47.240 | pulled the link in the GPT 3.5 model does know we could do

00:57:51.360 | that. That's new. They must add a browsing in the background.

00:57:54.960 | Yeah, or just pulling it. They did today, by the way. Whoa.

00:57:58.560 | They did it today. Yeah. Well, remember last week, we said that

00:58:02.000 | they had to build browsing into the actual product like bard,

00:58:05.080 | right? Otherwise, it can move. And I got to say, closed AI

00:58:09.080 | closed AI is on the top of their game. The app is available in

00:58:11.760 | the apps are now right as of today. No, they they had a test

00:58:15.080 | app. I was on the test. No, no, no, they just launched the app.

00:58:17.560 | Today. Oh, that's game over, man. If this thing is an app

00:58:20.520 | form, that's going to 10 x the number of users and it's going

00:58:22.880 | to 10 x the amount of usage. The same thing for bard, we should

00:58:25.880 | compare the two but bard bard is pretty good as well. Yeah. Yeah.

00:58:29.680 | Well, bring you out to actually compare its summary with chat

00:58:33.320 | GPT summary. Tell us which one's better. Right? This is some

00:58:37.720 | interesting news here. You know, we speaking of platform shifts,

00:58:42.160 | do I get to give my view on the Lena con thing?

00:58:44.200 | Oh, well, yes. But first, I don't want to read this was

00:58:49.160 | getting a little, it was getting a little personal here, David.

00:58:51.920 | And I didn't I didn't want to trigger you. I know you've been

00:58:53.720 | struggling with the gout. Because of your lifestyle

00:58:56.920 | choices, the alcohol, the foie gras everything, but now you've

00:59:00.840 | lost a lot of way to give you a lot of credit. Tell us what do

00:59:03.360 | you think about the bundling we're seeing here? Because it

00:59:05.200 | does seem Microsoft s with the operating system. Yeah, it's

00:59:07.400 | very similar. And what I said in the context of tech is that we

00:59:10.480 | should focus on the anti competitive tactics and stop

00:59:12.880 | those rather than blocking all mergers. And I think the same

00:59:16.360 | thing is happening now in the pharma space. If bundling is

00:59:19.120 | the problem focus on bundling. The problem when you just block

00:59:22.360 | M&A is that you deny early investors, one of the biggest

00:59:27.520 | ways that they can make a positive outcome. And what's the

00:59:30.200 | downstream effect of that? Yeah, exactly. Look, it is hard enough

00:59:33.800 | to make money as either a pharma investor or as a VC, that

00:59:38.000 | there's only two good outcomes, right? There's IPOs, there's

00:59:40.240 | M&A, everything else basically goes, everything else is a zero

00:59:43.400 | goes bankrupt. So if you take M&A off the table, you really

00:59:48.800 | suppress the already challenged returns of venture capital.

00:59:52.720 | Yeah, well said well said.

00:59:53.840 | And you're right, you mentioned earlier that we were willing to

00:59:56.520 | give Lena Khan a chance. We thought that some of our ideas

00:59:59.520 | were really interesting, because I think there are these huge

01:00:02.000 | tech companies that do need to be regulated these big tech

01:00:05.880 | monopolies, basically that you have the mobile operating system

01:00:08.720 | duopoly with Apple and Google, you've got Amazon, you've got

01:00:11.600 | Microsoft, and there is a huge risk of those companies

01:00:14.760 | preferring their own applications over downstream

01:00:17.800 | applications or using these bundling tactics. Yes. If you

01:00:22.520 | don't put some limits around that, that creates I think, an

01:00:25.280 | unhealthy tech ecosystem.

01:00:26.960 | This is the insight. And I think it's exactly correct sex. For

01:00:30.800 | Lena Khan, and if she listens to the pod, hey, Lena, you want

01:00:33.560 | to go after tactics, not acquisitions. So if somebody

01:00:37.920 | buys something, and they lower prices and increases consumer

01:00:41.520 | choice, that's great. If it encourages more people to

01:00:44.840 | invest more money into innovation, that's great. But if

01:00:48.600 | the tactics are we're going to bundle these drugs together to

01:00:51.160 | keep some number of them artificially high or reduced

01:00:54.000 | choice, or if we're going to bundle features into the you

01:00:56.880 | know, suite of products, and we do anti competitive stuff, you

01:00:59.920 | have to look at the tactics on the field are people cheating?

01:01:03.280 | And are they using the monopoly power to force you to use their

01:01:06.560 | app store? Just make Apple have a second app store. That's all

01:01:10.880 | we're asking you to do. There should be an app store on iOS

01:01:15.320 | that doesn't charge any fees or charges 1% fees. Break the

01:01:20.680 | monopoly on the app store.

01:01:21.800 | Saks is so right. Perfectly said, she actually did issue

01:01:25.040 | compulsory orders to the PBMs. So to your point, Saks, the FTC

01:01:30.400 | has been worried that what freeberg said has been

01:01:32.840 | happening. But the real sort of middleman manipulator in this

01:01:36.240 | market are the pharmacy benefit managers. And so this week, she

01:01:39.400 | actually issued compulsory orders to the PBMs and said,

01:01:41.880 | Turn over all your business records to me, I'm going to look

01:01:44.240 | into them. That makes a ton of sense. But then on the same end,

01:01:48.320 | it's like you see merger and you're like, No, it can't

01:01:50.280 | happen. It just doesn't speak to a knowledge of the market.

01:01:53.200 | We should have Lena on with the Halina. I know you listen to the

01:01:56.040 | pod. I've heard the back channel just come on the pod. I should

01:01:58.200 | be a good guest, right? We would have a good conversation. I

01:02:00.120 | think. Yeah. Open invite Nikki Halley is coming on the pod. By

01:02:03.640 | the way, you have homework to do for the summit, which is to

01:02:06.120 | see if you can get Donald Trump to come on to the summit. Okay,

01:02:09.240 | huge. Love J. Cal. Love all in great. Okay.

01:02:13.400 | Cute. Okay, not as good as apprentice but class. Okay,

01:02:18.880 | great.

01:02:19.480 | As your your mannerisms are unbelievable. How did you

01:02:23.240 | practice it? I did I did a little bit only because I like

01:02:25.560 | to show people and trigger them. I'm going to really dial in my

01:02:28.680 | Trump in the coming weeks. All right, here we go. Apple's long

01:02:31.520 | anticipated a our headset that stands for augmented reality,

01:02:35.680 | which means VR, you can't see the real world, you're just in a

01:02:40.760 | virtual world AR lets you put digital assets on the real

01:02:44.600 | world. So you can see what's happening in the real world. But

01:02:46.600 | you can put graphics all around that's expected to be revealed.

01:02:49.800 | As early as June, the projected cost is going to be around

01:02:55.360 | $3,000. It won't ship into the fall. This is a break from

01:02:59.080 | Apple's typical way of releasing products, which is to wait till

01:03:03.040 | it's perfect and to wait until all consumers can afford it.

01:03:05.800 | This is a different approach. They're going to give this out

01:03:07.640 | to developers early. And Tim Cook is supposedly pushing this

01:03:11.760 | there was another group of people inside of Apple who did

01:03:14.240 | not want to release a Chema. But there is some sort of external

01:03:18.600 | battery pack, it seems like a bit of a Franken product

01:03:21.440 | Frankenstein kind of project here. That, you know, perhaps

01:03:26.120 | Steve Jobs wouldn't have wanted to release but he needs to get

01:03:28.480 | it out. I think because Oculus is making so much progress to

01:03:31.840 | killer app supposedly is a FaceTime like live chat

01:03:35.320 | experience that seems interesting, but they look like

01:03:38.000 | ski goggles. Your thoughts Chamath on this as the next

01:03:42.200 | compute platform if they can get it, you know to work would you

01:03:46.200 | wear these? Would they have to be Prada? What's the story? No,

01:03:49.920 | no,

01:03:50.320 | this seemed like a weird conversation because none of us

01:03:52.240 | fucking know none of us have seen this product and none of us

01:03:54.280 | have used it. So like, this is just lucky friend of the pot

01:03:57.160 | poma. Lucky says it's incredible.

01:03:58.640 | So what that's just like commenting on one guy's five

01:04:01.320 | letter five word tweet.

01:04:02.400 | Let's Palmer knows I mean, Palmer invented Oculus.

01:04:05.920 | Great. What are we talking about? We have nothing to say

01:04:07.760 | to that.

01:04:08.040 | I'll a form of really good question here. Do you believe

01:04:11.800 | this is going to be a meaningful compute platform in the coming

01:04:14.720 | years because Apple is so good at product? How do we know till

01:04:17.680 | we see it? We got to see it. I think on Facebook, of course,

01:04:20.800 | they're of course they're good at product. Let's let's see the

01:04:22.920 | product though. Like,

01:04:23.760 | all right, fine. Saks, what are your thoughts?

01:04:25.480 | I think it's a good thing that they're launching this. Like you

01:04:27.880 | said, it is a deviation for what they've normally done. They

01:04:30.200 | normally don't release a product unless they believe the entire

01:04:33.520 | world can use it. So their approach has been only to

01:04:37.080 | release mass mass market products and have a very small

01:04:40.200 | portfolio of those products. But when those products work,

01:04:43.240 | they're, you know, billion user home runs. This obviously can't

01:04:46.920 | be at a $3,000 price point. And it also seems like it's a little

01:04:49.960 | bit of a early prototype where the batteries are like in a

01:04:53.000 | fanny pack around your waist. And there's a ways to go around

01:04:56.440 | this, but I give them credit for launching what is probably going

01:05:00.800 | to be more of an early prototype so they can start iterating on

01:05:03.440 | it. I mean, the reality is the Apple watch the first version

01:05:05.960 | kind of sucked and first five versions. Yeah, now they're on

01:05:09.320 | one that's pretty good, I think. So look, I think this is a cool

01:05:11.840 | new platform. They get knocked on for not innovating enough. I

01:05:15.160 | think good, let them try something new. I think this will

01:05:18.120 | be good for meta to have some competition. Yeah, it's great.

01:05:21.720 | Having two major players in the race, maybe it actually speeds

01:05:25.160 | up the innovation. I mean, we get some work.

01:05:27.240 | She wants any of that here. I mean, I think they should have

01:05:30.680 | done something in cars. I agree with David, like they do in

01:05:35.800 | cars. Well, if you were going to talk about the car, what would

01:05:38.120 | it be? Tell me what you think would be the right approach you

01:05:40.760 | were going to do the Facebook phone that could have changed

01:05:42.880 | the entire destiny of Facebook, they should have bought Tesla

01:05:45.000 | when they could have the chance for four or $5 billion. They

01:05:47.200 | could have bought it for 10 billion 20 billion. And when it

01:05:49.920 | got to 5060 that it got out of reach. What do you think the car

01:05:53.880 | should know they could have bought it at 100 billion. It

01:05:55.400 | could have bought 100 billion Tim Cook famously wouldn't take

01:05:57.600 | the meeting. Elon said it he wouldn't he wouldn't meet was

01:05:59.760 | our bizarre.

01:06:01.000 | Maybe they missed an opportunity there. But I do think the end

01:06:03.680 | game with the AR headset or glasses, right? Yes. Where you

01:06:07.640 | get the screens and you get the Terminator mode. And in is that

01:06:11.920 | what is that?

01:06:12.560 | These are just glasses that

01:06:14.120 | were like fancy technology. This this size glasses is what you're

01:06:19.040 | talking about. Yeah, like a little camera built in and

01:06:22.480 | conjunction with AI, then it gets really interesting. So

01:06:26.240 | that that's the end game here. I think

01:06:27.600 | give the audience an example of what this combination of AI plus

01:06:30.720 | AR could do

01:06:31.560 | when you're walking around, it could layer on intelligence

01:06:33.960 | about the world or you meet with somebody and it can remind you

01:06:36.880 | of their name and the last time you met with them and give you a

01:06:39.240 | summary of what you talked about. What action items there

01:06:42.600 | are.

01:06:42.960 | You can be walking in a city and it could tell you it knows you

01:06:46.600 | like Peking duck, it could show you Hey, there's a Peking duck

01:06:48.800 | place over here. Some reviews have it just knows you and it's

01:06:51.080 | customized in the world.

01:06:51.960 | What about for people that do the same routine 99% of the time

01:06:55.920 | how does it can help you that

01:06:56.880 | it could tell you your steps every day could tell you

01:07:00.200 | incoming messages so you don't have to take your

01:07:01.880 | $3,000 on that. No, but you would spend people spend plus

01:07:06.400 | your people spend $400 on notifications on your wrist. Why

01:07:10.120 | do you want it on your eyes for 3000? I would love this.

01:07:12.600 | Maybe I just do like a lot of meetings or I'm at events where

01:07:16.000 | people are coming up to me and I've met him like once a year

01:07:18.280 | before. Like it would be really helpful to kind of have the

01:07:21.440 | term. Let's be honest, though, the terminated mode for you to

01:07:24.360 | be able to be present with your family and friends, but be

01:07:28.320 | playing chess with Peter Thiel on those glasses. That's your

01:07:31.000 | dream come true. You and Peter in a are playing chess all day

01:07:35.520 | long. Throw up the picture of sacks beating Peter Thiel. I

01:07:39.480 | watched the clip from the early all in episodes when we

01:07:42.400 | discussed you beating Peter Thiel. What a great moment it

01:07:44.760 | was for you. All right, listen, let's wrap up with this gallop

01:07:47.600 | survey the number of Americans who say it's a good time to

01:07:51.360 | buy a house has never been lower 21% say it's a good time to buy

01:07:55.480 | a house down 9% from the prior low of a year ago prior to

01:08:00.520 | 2022 50% or more consistently thought it was a good time to

01:08:03.320 | buy significantly significantly fewer expect local housing

01:08:07.120 | prices to increase in the year. Hey, sacks is this like a

01:08:11.360 | predictive of a bottom and pure capitulation and then that means

01:08:15.720 | maybe it is in fact a good time. How would you read this data?

01:08:17.880 | I don't see it as a bottom necessarily the way I read the

01:08:21.120 | data is that the spike in interest rates have made it very

01:08:24.680 | unaffordable to buy a house right now you've got you know,

01:08:27.040 | the mortgages are what like 7% interest rate or even slightly

01:08:31.520 | higher. So people just can't afford the same level of house

01:08:35.360 | that they did before. I mean, mortgages were at three, three

01:08:38.720 | and a half percent like a year and a half ago. Now I think

01:08:41.800 | what's kind of interesting is that even in the 1980s, the

01:08:44.840 | early 1980s, when interest rates were at like 15%, you still had

01:08:50.360 | 50% thought it was an okay time to buy a house or attractive

01:08:53.120 | time to buy a house. So for the number to be this low tells me

01:08:58.000 | that is not just about interest rates, I think consumer

01:09:00.600 | confidence is also plummeting, and people are feeling more

01:09:05.480 | insecure. So I think it's just another economic indicator that

01:09:11.160 | things are looking really shaky right now. And I'll tell you one

01:09:13.960 | of the the knock on effects of this is going to be that people

01:09:16.720 | can't move. Because in order to move, you have to sell your

01:09:20.160 | current house, and then buy a new one. And you're not going to

01:09:23.360 | want to sell your current house when prices are going down. And

01:09:26.800 | then for the new one, you're going to lose your 3% mortgage

01:09:29.760 | and have to get a new one at 7%. So you're not gonna buy

01:09:32.360 | anything like the house. So it freezes the market, it freezes

01:09:36.160 | mobility. I think over the last few years, during COVID, you saw

01:09:39.720 | tremendous movement between states, I think that's gonna

01:09:42.800 | slow down a lot now, because people just can't afford to

01:09:45.520 | trade houses. So as a result of that, I think discontent is

01:09:49.640 | going to rise. Because I think one of the ways that you create

01:09:52.320 | a pressure valve is when people are unhappy in a state, they

01:09:55.600 | just move somewhere else. Well, now they're not going to do

01:09:57.320 | that. Well, you can also move to a better opportunity for you and

01:09:59.960 | your family, whether that schools, taxes, a job,

01:10:02.680 | lifestyle. So yeah, you can you can, you're going to reduce joy

01:10:06.080 | in the country. And it also, it screws with price discovery,

01:10:09.100 | doesn't it? Your mouth if you if you don't have a fluid market

01:10:11.800 | here, then how does anybody know what their house is worth? And

01:10:14.840 | this just, again, creates more, I think, frost,

01:10:18.360 | I think Friedberg has said this a couple times, Friedberg, you

01:10:20.840 | can correct me if I'm wrong. But like the the home is like the

01:10:24.280 | disproportionate majority of most Americans wealth, right?

01:10:28.080 | It's all their wealth.

01:10:30.600 | All their wealth. Yeah. So I mean, there's that factoid.

01:10:36.960 | Yeah. And then what does that do for their attention savings or

01:10:40.200 | whatever? It's okay. You got incoming what's going on?

01:10:42.280 | They're bringing you lunch?

01:10:43.160 | No, I was looking I was looking at a mansion that's for sale.

01:10:48.200 | Like $175 million, but they just got the price to 140. So I'm

01:10:54.720 | just taking a little again. I mean,

01:10:56.160 | there's gonna be a lot of distress in the market soon. I'm

01:10:58.880 | predicting a lot of distress. Actually, can we shift to the

01:11:01.040 | commercial side for a second? I just passed away. Yeah. Sam

01:11:04.840 | Zell passed away. Today. Oh, wow. Rest in peace. Yeah. Rest in

01:11:08.280 | peace. Chicago. Yeah, crazy. But speaking of bastic, interesting

01:11:13.560 | guy. Yeah. But speaking of the real estate market, so I want to

01:11:16.080 | give an update on San Francisco, Siri, I was talking to a broker

01:11:20.640 | the other day. And so here are the stats that they gave me. So

01:11:25.720 | it was a local broker than someone from Blackstone. And

01:11:28.840 | they're fans of the pod and just came up to me, we started

01:11:30.760 | talking about what's happening in San Francisco. Shout out to

01:11:33.760 | them. Didn't take a didn't take a photo. But But in any event,

01:11:37.160 | they're they're fans of the pod. So we started talking about

01:11:39.080 | what's happening in San Francisco, real estate. So the

01:11:43.000 | SF office market is just a level set is 90 million square feet,

01:11:46.440 | they said the vacancy rate is now 35%. So that's over 30

01:11:50.000 | million square feet vacant, and vacancy still growing as leases

01:11:53.880 | end, and companies shed space because some of that space that

01:11:57.280 | they're not using is not for sublease. Everyone says what

01:12:00.200 | about AI is AI going to be the savior? The problem is that AI

01:12:03.640 | companies are only that's only about a million square feet of

01:12:07.320 | demand. So 1 million out of 30 million is going to be absorbed

01:12:11.480 | by AI. And you know, maybe that number grows over time over the

01:12:14.440 | next 510 years as we create some really big AI companies, but it's

01:12:18.440 | just not going to bail out San Francisco right now. The other

01:12:21.040 | thing is that, you know, VC backed startups are very

01:12:24.360 | demanding in terms of their tenant improvements, and

01:12:27.120 | landlords don't really have the capital right now to put that

01:12:29.440 | into the buildings. And startups just are not the kind of credit

01:12:33.320 | worthy tenants that landlords really want. So this is not

01:12:37.720 | going to bail anybody out there. They said there are a ton of

01:12:40.600 | zombie office towers, especially in Soma. And all these office

01:12:44.800 | towers are eventually gonna be owned by the banks, which you're

01:12:46.600 | gonna have to liquidate them. And then we're gonna find out

01:12:49.280 | that these loans that they made are gonna have to be written off.

01:12:52.080 | Because the collateral that they thought was blue chip, that was

01:12:56.280 | backing up those loans is not so blue chip anymore. So I think

01:12:59.800 | we've got not just a huge commercial real estate problem,

01:13:02.680 | but it's gonna be a big banking problem as basically people

01:13:06.120 | stop pretending, you know, right now, they're trying to

01:13:08.240 | restructure loans, it's called pretend and extend, you reduce

01:13:11.280 | the rate on the loan, but add term to it. But that only works

01:13:15.320 | for so long. If this keeps going, if the market keeps

01:13:18.160 | looking like this, I think you're gonna have a real

01:13:20.160 | problem. And that will be a problem in the banking system.

01:13:22.440 | Now, San Francisco is the worst of the worst. But they said that

01:13:25.000 | New York is similar. And all these other big cities with empty

01:13:28.400 | office towers are directionally.

01:13:29.920 | I'm in New York right now for the side connections conference.

01:13:32.880 | And it is packed, the city is packed, getting anywhere,

01:13:38.880 | there's gridlock, you can't walk down the street, you got to walk

01:13:41.840 | around people every restaurant, it is dynamic. And then I talked

01:13:45.600 | to people about offices. And they said, people are staying in

01:13:48.640 | their houses and their tiny little New York apartments,

01:13:51.080 | instead of going three train stops to their office, they go

01:13:53.640 | to the office one or two days a week, unless you're like JP

01:13:55.920 | Morgan, or some other places that's the drop the boom. But

01:14:00.280 | there's a lot of people still working from home, the finance

01:14:02.720 | people have all gone back media people are starting to go back.

01:14:04.880 | So there are three to five days here. And the city is booming.

01:14:09.240 | Contrast that I spent the last two weeks in San Francisco,

01:14:12.920 | walking from Soma to the Embarcadero back dead, nobody in

01:14:17.760 | the city. It's like literally a ghost town. It's a real shame.

01:14:21.760 | It's a real, real shame. I wonder if these this is the

01:14:26.840 | question I have for you, Saks. Can they cut a deal? Can they go

01:14:30.480 | to like month to month ran sub lats? You know, Lucy, Lucy, just

01:14:34.720 | give people any dollar amount to convince them to come back? Is

01:14:38.200 | there any dollar amount because I'm looking for a space for the

01:14:40.520 | incubator in San Mateo, I've been getting a ton of inbound,

01:14:43.640 | but the prices are still really high. And I'm like, how do I cut

01:14:47.240 | a deal here? Because shouldn't people be lowering the prices

01:14:51.160 | dramatically? Or just pretending or will I get a lower

01:14:55.280 | rates are definitely coming down big time, especially for space

01:14:58.520 | that sort of commodity and not that desirable. But what's

01:15:00.640 | happening is, according to the people I talked to is that the

01:15:03.520 | demand, the people who actually are looking for new space, they

01:15:06.880 | only want to be in the best areas. And they want to be in

01:15:09.400 | the newest buildings that have the best amenities. And so that

01:15:13.240 | sort of commodity office tower where there's barely anybody

01:15:16.000 | ever there, like no one wants that. So I think people would

01:15:19.360 | rather pay a higher rent. I mean, the rent will still be

01:15:22.520 | much lower, probably half the price of what it used to be, but

01:15:24.600 | they'd rather pay a little bit more for that than get like a

01:15:27.800 | zombie office tower. We can't talk about all this without

01:15:30.000 | talking about two cases. Tragically, a shoplifter, a

01:15:36.320 | criminal who was stealing from a drugstore in San Francisco, I

01:15:40.880 | got shot. And the video was released, I'm sure you've seen

01:15:44.200 | it sacks. And then here in New York, everybody's talking about

01:15:47.280 | this one instance of a marine trying to subdue a violent

01:15:53.360 | homeless person with two other people. And it's on everybody's

01:15:57.040 | minds here. And Brooke Jenkins is not prosecuting in San

01:16:01.720 | Francisco, the shooter, they look like, you know, a clean

01:16:05.640 | shoot, as they would say in the police business, and

01:16:08.360 | appropriate. And it's tragic to say it is, but the person did

01:16:11.720 | charge the security guard, the security guard did fear for

01:16:15.200 | their life and shot him. So Brooke Jenkins is not going to

01:16:18.040 | pursue anything. But in New York City, they're pursuing

01:16:21.040 | manslaughter for the person who did seem a bit excessive from

01:16:23.720 | the video, it's hard to tell what the reality is in these

01:16:25.600 | situations. Any thoughts on it, David, these two cases in two

01:16:28.880 | cities, tell us.

01:16:29.600 | Yeah, look, I mean, the only time you can get a source da

01:16:34.160 | excited about prosecuting someone is when they act in

01:16:37.080 | self defense or defensive others. I mean, this marine, I

01:16:40.440 | guess Daniel penny is his name, he was acting in defense of

01:16:44.200 | others, the person who he stopped was someone with an

01:16:48.120 | extensive criminal record, who had just recently engaged in a

01:16:54.000 | attempt at kidnapping who had punched elderly people had

01:16:58.080 | dozens of arrests. In fact, people on Reddit were talking

01:17:02.680 | about how dangerous this person was, apparently, a dozen years

01:17:06.760 | ago or so he was seen as more of like a quirky like Michael

01:17:11.720 | Jackson impersonator, street performer, street performer, but

01:17:15.560 | something happened. This is according to a Reddit post that

01:17:18.360 | I saw where something happened. And there was some sort of

01:17:21.040 | psychological break. And then since then, he's had dozens and

01:17:24.200 | dozens of crimes. And they just keep letting him loose through

01:17:29.320 | this revolving door of a justice system we have. And now look, no

01:17:32.640 | one likes to see him basically dying. And yeah, it's too bad.

01:17:37.640 | It's horrible that that happened. Tragic. I don't know,

01:17:40.640 | though, that if you're trying to stop someone, I don't know how

01:17:44.640 | easy it is to precisely control whether you use too much force

01:17:48.640 | or not. So I think Daniel Penny has a strong case that he was

01:17:52.960 | acting in self defense and defensive others. And there

01:17:55.560 | were two other people, by the way, who are holding this

01:17:57.320 | person down. There are three of them restraining him. And what

01:18:00.680 | universally New Yorker said to me of all different backgrounds

01:18:05.080 | was this is not a race issue. The other I think one or two of

01:18:07.880 | the other people were people of color. It was not a race issue.

01:18:10.560 | And they're trying to make it into a race issue in both these

01:18:12.640 | cases. And it's this is literally what happens. It just

01:18:16.320 | having been through this in New York in the 70s and 80s when you

01:18:18.880 | do not say who's they when you say trying to make a bunch of

01:18:22.240 | protests on the street, both in San Francisco, and New York

01:18:26.240 | people protesting these as you know, justice issues. The fact

01:18:29.720 | is, if you do not, if you allow lawlessness for too long a

01:18:33.680 | period of time, you get a Bernie gets situation. And Bernie gets

01:18:38.080 | people can look it up in the 80s. I was a kid when it

01:18:40.840 | happened. But they tried to mug somebody, he had a gun, he shot

01:18:44.680 | him. And like, this is what happens if you allow lawlessness

01:18:47.400 | for extended periods of time. It's just, you're basically

01:18:51.860 | gambling. And what happened to Bernie gets he got not guilty,

01:18:55.080 | the case, he got not guilty, but I think he had an illegal gun.

01:18:58.480 | So he was guilty of that. The Bernie gets thing was really

01:19:01.680 | crazy. Because at the time, the climate in New York, and this

01:19:06.480 | 1984 shooting. There was a portion of people who I don't

01:19:12.560 | want to say they made him a hero, but they made it a see,

01:19:17.080 | this is what happens if you allow us to be assaulted

01:19:19.840 | forever, we're going to fight back at some point. That was the

01:19:22.520 | vibe in New York when I was a child. I was 1415 years old when

01:19:25.080 | this happened. He was charged with attempted murder assault.

01:19:27.640 | Jason, what was the name of that vigilante group that used to

01:19:30.400 | walk the streets to something angels that was the guardian

01:19:33.640 | angels guardian angels. So it was so bad in the 80s. And I

01:19:36.880 | actually almost signed up for the guardian angels. I went to

01:19:39.560 | their headquarters because I was practicing martial arts. And I

01:19:41.360 | thought I would check it out. And they had their office in

01:19:44.800 | Hell's Kitchen, I didn't wind up joining. But what they would do

01:19:46.880 | is they would just ride the subway, they would wear a

01:19:49.920 | certain type of hat, and wear a guardian angel shirt. And all

01:19:53.560 | they did was ride the subways, a red beret, and they would just

01:19:56.600 | ride the subways. And you felt like martial arts were you

01:20:00.080 | taking? Taekwondo? I was in Taekwondo. Yeah, this is before

01:20:03.680 | mixed martial arts, but they just rode the subways. And

01:20:06.680 | honestly, I've been on the subways with the many times you

01:20:08.800 | felt safe. And it wasn't vigilantes. They were guardian

01:20:12.640 | angels, they use that term. And many times they would do exactly

01:20:16.160 | what this Marine did, which is try to subdue somebody who was

01:20:18.440 | committing crime. I was, I had two distinct instances where

01:20:23.120 | people tried to mug me, you know, riding the subways in New

01:20:25.800 | York in the 80s. Two distinct times. And one was a group of

01:20:29.520 | people and one was one person like it was pretty scary. Both

01:20:33.360 | times I navigated it, but it was Yeah, not pleasant in the 80s in

01:20:38.360 | New York. I say one more thing about this. Danny's honey,

01:20:41.040 | Jordan Neely case. So look, at the end of the day, this is

01:20:43.240 | gonna be litigated. I don't know all the details, they're gonna

01:20:46.080 | have to litigate whether Daniel pennies use of force was was

01:20:49.160 | excessive or not. But But here's the thing is that the media has

01:20:52.760 | been falsely representing Jordan Neely by only posting 10 year

01:20:56.800 | old photos of him and leaving out crucial information. This

01:20:59.320 | was a press report. So again, this is why I mentioned the

01:21:02.560 | whole Michael Jackson impersonator thing is that the

01:21:04.640 | media keeps for training Neely as this innocent, harmless guy

01:21:07.760 | who is this like delightful Michael Jackson personator. In

01:21:10.840 | truth, he hasn't done that in more than a decade, because

01:21:14.520 | again, he had some sort of mental break. And since then,

01:21:17.840 | he's been arrested over 40 times, including for attempting

01:21:20.680 | to kidnap a seven year old child. And so the media is not

01:21:26.120 | portraying this case, I think in an accurate way. And I think as

01:21:30.280 | a result of that, it leads to pressure on the DA to prosecute

01:21:35.280 | someone who has I think, a strong self defense claim. Or

01:21:38.280 | you know, maybe the DA just wants to do this anyway. And it

01:21:41.280 | gives the DA cover to do this Soros is I mean, I know that we

01:21:45.420 | had this back and forth with us. Why is CNN being inaccurate? Do

01:21:47.880 | you think sex? They're basically cooperating with Alvin Bragg's

01:21:51.800 | interpretation of the case. They're trying to make the case

01:21:54.520 | against penny look as damning as possible.

01:21:57.120 | Why don't they just take it straight down the middle? It's a

01:21:59.200 | tragedy. We have a screwed up situation here. We got a mental

01:22:02.360 | health crisis. And it's a tragedy for everybody involved

01:22:04.920 | on the Bernie get stuff. He served eight of a 12 month

01:22:08.200 | sentence for the firearm charge. And he had a massive $43

01:22:11.800 | million civil judgment against him. In 1996. Decade later, it's

01:22:17.560 | just

01:22:17.880 | this is a little different than the gets thing because pulling

01:22:21.360 | out a gun and shooting somebody. Yeah, that's deadly intent.

01:22:24.040 | Yeah, that's that's a huge escalation. Whereas penny, he's

01:22:27.840 | a he's a trained Marine, right? He's trying to immobilize him.

01:22:30.520 | He has to believe that he's just trying to subdue. Yes, nearly.

01:22:34.800 | And so still using a chokehold to kill him. That's an

01:22:37.920 | unfortunate consequence of what happened. But he was trying to

01:22:41.360 | restrain the guy. As far as we know, right, as far as we know.

01:22:45.160 | Yeah, I mean, tragedies all around. We got to have law and

01:22:48.480 | order. I tweeted like, I don't know why we still have the post

01:22:51.400 | office, maybe we can make that like once a week, and redo all

01:22:54.680 | of that space and allow every American who's suffering from

01:22:57.640 | mental illness to check in to what used to be the post

01:23:00.280 | office. You know, maybe like once a week, and obviously give

01:23:04.120 | those people very gentle landings, but I don't think we

01:23:06.240 | need postal service more than once or twice a week. And then

01:23:08.800 | let you let's reallocate some money towards mental health in

01:23:13.320 | this country where anybody who's sick, who feels like they're

01:23:16.280 | violent or feels like they're suicidal can just go into a

01:23:19.540 | publicly provided facility and say, I'm a sick person, please

01:23:22.440 | help me this would solve a lot of problems in society. We've

01:23:25.680 | got a mental health crisis, we should provide mental health

01:23:28.160 | services to all Americans. And it's obviously easy thing for us

01:23:31.760 | to afford to do.

01:23:32.520 | And if we had done that, then this never would have happened.

01:23:34.880 | Exactly. I mean, literally, you have sacks who wants to balance

01:23:38.240 | the budget saying, Hey, this is something worth spending on. We

01:23:40.920 | can all agree on this

01:23:41.800 | compared to the impact on society. I don't think it'd be

01:23:44.160 | a huge expense.

01:23:44.960 | We would save money. We'd save money because a city like San

01:23:48.120 | Francisco could become quite livable or New York if and then

01:23:51.640 | you've got forbid these terrible school shootings, you know, if

01:23:53.880 | you avoid even one of them, it's 30 people's lives or 10 people's

01:23:57.240 | lives. So it's

01:23:58.440 | convert post offices, what we need to do is stand up scale

01:24:02.400 | shelters, and it doesn't need to be done on the most expensive

01:24:05.040 | land in a given city outside of cities. Why does this? There is

01:24:09.720 | no expectation in Europe for like Paris or London to be

01:24:13.360 | affordable or Hong Kong to be affordable. There are affordable

01:24:15.920 | places 30 minutes outside of those places where you could

01:24:18.760 | put these facilities. I just want to ask one question to

01:24:21.320 | sacks because I don't know. And I know sacks is a little bit

01:24:24.720 | deeper into this. And what is George sources motivation for

01:24:28.680 | putting in these lawless insane da is like, I understand that

01:24:34.120 | he was able to buy them. They're low cost. There's not a lot of

01:24:37.080 | money in them. Okay, I understand that that's

01:24:38.480 | tablestakes. But what is his actual motivation for causing

01:24:41.880 | chaos in cities?

01:24:42.760 | Listen, we can't know exactly what his motivation is. But what

01:24:46.000 | he did is he went into cities where he doesn't live and

01:24:50.680 | flooded the zone with money to get his preferred candidate

01:24:53.440 | elect his da. Now, the reason he did that was to change the law.

01:24:56.360 | And the way that he changed law was not through legislature

01:24:58.640 | is the way you're supposed to operate, but rather, by abusing

01:25:01.560 | prosecutorial discretion. So in other words, once he gets his

01:25:04.880 | Soros da elected, they can change the law by deciding what

01:25:08.320 | to prosecute and what not to prosecute. Right. And that's why

01:25:10.880 | there is so much lawlessness in these cities.

01:25:12.720 | But there's a better path you're saying?

01:25:14.400 | Yeah, but this is not the only way that Soros has, I'd say,

01:25:19.480 | imposed his values on cities that he doesn't even live in.

01:25:24.320 | Where does he live? I think he's a New York guy, but I'm not

01:25:26.640 | sure. But but he's gone far beyond that, obviously, in these

01:25:31.120 | elections, but also, he's done this across the world. Soros has

01:25:35.400 | this thing called the Open Society Foundation, which sounds

01:25:37.680 | like it's spreading democracy and liberal values, but in fact,

01:25:40.640 | is fomenting regime change all over the world. And he's been

01:25:45.160 | sponsoring and funding color revolutions all over the world.

01:25:47.800 | Now, if you like some of the values he's spreading, then

01:25:50.880 | maybe you think that's a good thing. But I can tell you that

01:25:54.040 | the way this is perceived by all these countries all over the

01:25:57.480 | world is it creates tremendous dissension and conflict. And

01:26:02.520 | then they look at America and they basically say, you know,

01:26:05.680 | this American billionaire is coming into our country, and

01:26:07.800 | he's funding regime change. And it makes America look bad. Now

01:26:12.400 | he's doing this, I think, with the cooperation of our State

01:26:16.000 | Department, a lot of cases, and maybe the CIA, I don't know. But

01:26:19.640 | this is why America, frankly, is hated all over the world as we

01:26:22.540 | go running around meddling in the internal affairs of all

01:26:26.320 | these countries.

01:26:27.080 | Two is this guy? Well, they're like, that was the other thing I

01:26:29.400 | heard is that he's not all there. And the people around him

01:26:31.520 | are doing these kind of things in his organization.

01:26:33.440 | I heard something similar is that as the idiot son,

01:26:35.760 | Alexander, who's really now pulling the strings. Would you

01:26:39.480 | would you allow Soros to speak at all and summit? Would you

01:26:41.560 | enter? Sure. Yeah, let's have Soros or a son and Nick and

01:26:44.260 | explain themselves if they're so proud.

01:26:45.640 | Well, apparently, there was an article that Alexander Soros has

01:26:48.640 | visited the White House like two dozen times during the Biden

01:26:51.240 | presidency. This is an extremely powerful and connected person.

01:26:55.020 | I mean, I'm sure he listens to pod. Okay, we'll see you all

01:26:57.900 | next time. This is Episode 129 of all in we'll see you in

01:27:02.160 | Episode 130. Bye bye. Love you. Bye bye.

01:27:05.060 | Let your winners ride.

01:27:08.020 | Rain Man David

01:27:10.580 | we open sources to the fans and they've just gone crazy.

01:27:17.340 | Love you.

01:27:18.760 | Queen of

01:27:19.360 | besties are

01:27:27.600 | dog taking a

01:27:30.880 | driveway.

01:27:31.560 | We should all just get a room and just have one big huge orgy

01:27:39.960 | because they're all just like this like sexual tension that

01:27:42.820 | they just need to release.

01:27:45.240 | What you're about to be your fee.

01:27:49.720 | Good. Good. We need to get murkies are

01:27:52.080 | ♪ I'm doing all in ♪

01:27:57.080 | ♪ I'm doing all in ♪

01:28:03.080 | You

E129: Sam Altman plays chess with regulators, AI's "nuclear" potential, big pharma bundling & more

Chapters