All-In Podcast E233 — Fable 5 Backlash, Nationalizing AI, Inflation, California Elections

[0:00] All right, everybody. Welcome back to the number one podcast in the world. It's your favorite podcast. It's your podcasters's favorite podcast. It's the podcast that everybody likes to rip off for 50 episodes and then they quit. But the All-In podcast is not quitting. We're doubling down with the original quartet. Everybody's here for a big big week, lots of debates, and we got to start with Anthropic.

[4:15] Do I want to have single point of failure in the supply chain. The next thing you know, there's one big box that's going to deliver the model. And what about open source? You know, do we just have this one provider that's going to be the monopolistic provider of all this intelligence? That's a concern.

[8:30] upon the model providers in a way that we have not really seen before. This is sort of unprecedented. And so the question is, are we going to use that? Is the government going to use that as a way to start to take over or take equity stakes in these companies? That's the debate.

[12:46] overreach here. So someone showed that if you try to train a better model using Claude, they were basically downgrading you. But they were not telling you that. And so that's why people are mad. It's not just the 30 days, it's the fact that they were downgrading people.

[17:00] that regulation to open source which is actually quite different. I think the open source model is going to be unbelievably important. And so I worry about overregulation there. But on the closed models, I think there's a strong case for some guardrails.

[21:14] psychosis with these like 5,000 word blog posts. Like you just write a 5,000 word blog post and everyone's like, oh my God, this is a real problem. Let's go nuke the company. Like I don't know if that's the right approach. I think Anthropic deserves a lot of credit for being thoughtful and transparent.

[25:30] the tools. If we go back to the original thesis, the price is 2x Opus 4.8. So per token, it's twice as much. But if it's actually that much better, you should be using fewer tokens. So you know, the question is, is the net effect the same, better, or worse? And so far we just don't have enough evidence to know.

[29:42] are the smarter thing would be to basically have a Strategic Equity Stake where the government is not operating these companies, but they have a passive equity stake. And so if these companies become very valuable, the government shares in that upside. But the operating decisions are still made by the private sector.

[33:58] sort of experiment, which every lab in Silicon Valley has been running for the last 18 months, is that the government doesn't actually have the capability to do frontier research and development. They just don't. They can do incremental. They can do procurement. But they can't do frontier.

[38:14] doing this deal with Saudi Arabia. The PIF took a stake in Magic Leap, in Uber, in Tesla at one point. And these are passive equity stakes. And it didn't mean that Saudi Arabia was running those companies. And I think that's a model that we could potentially adopt here.

[42:28] supposed to maximize the value to shareholders. They're not supposed to maximize national security. And so I think we need to separate the two. And the way to do that is to take a passive equity stake where the government can have some level of input on national security issues, but the operating decisions are still made by the private sector.

[46:42] the Lighthorse AI deals. I think it was called Project Maven where they basically just took all the Google engineers and they tried to build this thing. They spent like $200 million and they ended up with nothing. And so I just don't think the government has the capability to do this.

[50:58] post on the pause was to justify the creation of these new agencies. And so if you actually had a pause, you'd have to fire all of these people that are working on AI safety. And so the pause was always a bit of a Trojan horse for the safety community to get more power.

[55:12] you know, the Core CPI came in at 3.2% which is a tenth lower than expected. And the headline was 3.0% which was a tenth lower. So that's good news. But the core services ex-shelter came in at 4.5% which is really the stickiest part of the inflation print. And so that's not great news.

[59:28] published now all the interviews. Yeah. So like, but the bottom line is core services ex-shelter is the Fed's most closely watched measure. And if that's 4.5%, it's really hard for them to cut rates. They would have to see that come down meaningfully before they could start cutting.

[1:03:42] got a 3-handle on inflation. I mean, we are within striking distance of a 3-handle. The next print could be 2.9, 2.8. I mean, we are really close to a 2-handle. And so I think the Fed is going to be very cautious. They don't want to cut too early and have to come back and raise rates.

[1:07:56] inflation in this country, the core services ex-shelter, the super core inflation is the stickiest part. And if you're looking at the wage data, wages are up 4.7% year over year. So if you have wage inflation at 4.7% and core services at 4.5%, you don't have a disinflationary process. You have a stagflationary process.

[1:12:10] California elections. There's a bill that's moving through the California legislature called SB 1407. And basically what it does is it allows for same-day voter registration. It allows 16 and 17 year olds to be auto-registered. And it allows voters to change their vote on election day.

[1:16:26] can register to vote. There is no proof of residency. There is no proof of citizenship. You can just walk in and say I'm John Smith and I live at 123 Main Street. And they will give you a ballot. And you can vote. And that's the current state of affairs in California.

[1:20:40] the DMV is now issuing voter registration cards. So if you go to the DMV and you say, hey, I just moved here, I need to update my address, they will give you a voter registration card. And so the DMV is now in the business of issuing voter IDs. And this is a huge vulnerability.

[1:24:56] fraud. It's this it's designed to be vulnerable. And so when you have a system that's designed to be vulnerable, you're going to have fraud. And so the question is, are we going to fix it? Or are we going to make it worse? And SB 1407 makes it worse.

[1:29:10] the worst possible time to be making it easier to cheat. And the reason is that California has 54 electoral votes. It's the most electoral votes of any state. And in 2028, if the election is close, California could be the deciding state. And so we need to make sure that the election in California is secure.

[1:33:25] anything as we wrap here on this? I think the thing that I would say is that we are in a very unique moment in time. We've got the AI revolution happening. We've got inflation still elevated. We've got elections coming up. And all of these things are intersecting at the same time. And so we need to be thoughtful about how we approach all of them.

[1:41:55] I'm going all in.