World's First Trillionaire, Anthropic Fable Banned, The New Oligarchs, Iran Peace Deal

29:04It could have been 50. That democratization needs to continue because the real, back to Freeberg's points about being a wage slave or an equity holder and trying to make that transition. well people being able to open up robin hood and just buy one or two shares and vote with their dollars for 150 bucks that they want this future as opposed to other versions of the future that's what gives them that empowerment that's what gives them that agency well what if they could do that earlier and i've beaten this drum on this podcast and my other podcast for a decade on this but people knew spacex they were watching these rockets launch for a long time and they would have placed that bet 10 years ago when the company was worth$100 billion or$10 billion or$500 billion, but they weren't allowed to because we have a corrupt system in our government that says rich people, the top 4 % or 5 % of this country, are smart enough to buy private company stocks, and the 95 % are too stupid to.

30:01And it's antiquated, and it has to change, and you should be able to buy whatever you want. If you want to buy crypto, you can. If you want to buy a private company, you can. If you want to play cards with your friends and play bomb pots and watch Dave Freeberg's rage quit because he keeps losing at bomb pots when we play, that's Freeberg's right. Let people buy private company stock. It's crazy that we still have this rule keeping poor people poor. Thanks for coming to my TED Talk. Can I take a shot at answering Jamal's question? Why are there so many people who are influential, who are resentful of what Elon's done or what other capitalists do?

30:39I think it was Joseph Schumpeter, who explained this, that basically you've got this intelligentsia in a society like ours where they don't make things. They don't make the machines that make things. They just make words. And, you know, I guess ideas and most of those ideas are just wrong, right? They're just made up and they become resentful of the fact that there are all these people who are able to create, not just stuff, but the machines that make the stuff, right? And that But in our society, that is the thing that creates the greatest reward financially. Why? Because it actually aligns with what's in the greatest value for society, whereas they produce a bunch of hot air.

31:21So it's natural for that class of intelligentsia to become resentful and hostile towards the people who are able to amass this great wealth. I think that the great lie is that there are two sides to the society that is the rich and the poor. and the great truth is that there are two sides that are the makers and the takers the lie is that the rich are unfairly rich and the poor are unfairly poor and therefore the poor must take from the rich but the truth is that it's the takers that tell you that lie that the real truth is that artists plumbers electricians woodworkers computer scientists people that build people that make from all walks of life, all income levels, all wealth brackets are the makers.

32:11And the takers are what Sachs calls this intelligentsia, the analysts, the espousers, the armchair mechanics, the critics, the commentators, the politicians, they are the takers. They are the people that watch the rest of society make stuff, build stuff, specifically doing things that create value for other people in society. That's what a maker is. Tell my kids this lesson all the time. I say, what did you guys make today? They tell me something they made. And I'm like, did someone else value it? Or did you make it just for yourself? At the end of the day, if you made something and someone else valued it, you were a maker.

32:48That was an amazing achievement. That is a great day. Whether you make a piece of art, whether you build a house, whether you write a piece of software, whether you build a business, whatever you make, if you're a maker and someone else values it, they become your customer in some way, they become your partner in some way, you've done something valuable. Those are the true engines of progress for humanity. And the takers are the ones that tell the lie that it's rich versus poor, because what they want to do is rip apart the makers. And they want to tell everyone, you got to be on one side or the other.

33:16And they use that lie to get everyone to line up against each other and to give themselves ultimately the control to form the great American Politburo, which is what they're trying to create. That is what's underway right now. And that is the fundamental great truth and great lie that we're kind of fighting against at this moment. All right. This broke shortly after we taped our episode last week. We tape on Thursdays. We drop on Fridays, as fans of the show know. Well, the U.S. government has pulled the plug on Anthropics' latest model. You've heard us talk about that as Mythos. They released it as a controlled version called Fable 5.

33:53So there was Mythos. That was the one they held for 30 days while they were doing security, cybersecurity testing. Then they released it as fable and a bit of a guardrailed version of it. Commerce Secretary Howard Lutnick told Anthropic to restrict that model to U.S. citizens last week, last Friday, in fact. Anthropic couldn't implement those restrictions or wouldn't in that granular of a way, so they just decided to do something more conservative and shut it down for everybody. This, of course, created a lot of hand-wringing in the industry, rug-pulling. oh my god we lost this great model and what does that mean if i build my models on top of anthropic can i trust them yada yada and this was just days after the model was launched on june 9th how this all started seems to be that amazon ceo andy jassy who owns a large amazon is a an early investor in anthropic and they own a large portion of it he had told the administration uh that there was a security vulnerability in fable five that they were able to bypass to jailbreak the guard rails that anthropic put on it.

34:53Dario said the jailbreak wasn't serious. There seems to be some finger pointing it back and forth. And Dario said it was narrow. Separately, Semaphore, a publication, like an email publication reported that the White House acted partially over suspicion that a China-linked group had access to mythos, and they weren't supposed to. Wired reported that this group was South Korea's SK Telecom. Now you're saying, hey, how does South Korea have anything to do with China? Well, they have a longstanding, or there are longstanding allegations that SK has a relationship with China. The White House previously ordered Anthropic to revoke SK Telecom's access to Mythos.

35:30They're kind of like the Verizon 70 % market share for telecom in South Korea. According to the Washington Post, Anthropic did not initially disclose it had given Mythos to SK Telecom. And this happened before all the export controls, quote, badly damaged officials' confidence in the company's ability to safeguard sensitive technology. Anthropic says the White House never raised the Chinese access in the fable jailbreak conversations. There's a lot of finger pointing and perhaps nobody is closer to this except the principals and you, David Sachs. So tell us what is going on here. Is this politics and political between Anthropic and the administration?

36:13Is this a misunderstanding? Or is this, you know, the future of what's to come as these models become more powerful? Well, let me give you an update based on what I learned. This would have been last Friday and Saturday. So by that point, I got a readout from various folks at the White House about what happened around the time that that letter was sent to Anthropic. And I then summarized what I learned in that long post on X. And no one told me what to say. I just listened to these various officials explain themselves. I also heard from, let's say, some of the private companies who were involved in this.

36:50And all I asked was, listen, I think this context would be helpful to provide. They said, sure, you can publish it. So I posted my summary of what was happening. And whether you agree with the decisions of those officials or not, I do think that my post provides a window into exactly what they were thinking at that time. And what this comes down to is a couple of things. So first of all, like you said, Dario came to Washington a few months ago. This was back in April and basically said that he had created a cyber weapon called Mythos. And he spiked the cortisol level, got everyone really worried.

37:24And there was some truth to it in terms of the sense that this model had advanced cyber capabilities. But he got everybody very focused on this problem. So mission accomplished. They then had this pilot program for Mythos, this trusted partners program. And as the Washington Post reported, Anthropic expanded the program to something like 50 companies or more without consulting the White House. And according to the Washington Post, I don't have any firsthand knowledge of this, but I don't have any reason to question the Washington Post's reporting on this. They did share Mythos with a company that the White House believes it should not have been shared with.

38:06So think about that. They're sharing these capabilities that Dario himself has said is a cyber weapon with parties that the White House and really we're talking about the national security officials believe might have connections to China. That's SK Telecom, according to these. I'm not going to say who it was. I know that's been reported. I don't have any firsthand knowledge of that. OK, I'm just reporting what was in The Washington Post. But you have to understand that that's the context. So first of all, you got Anthropic effectively conducting its own foreign policy here, expanding the mythos preview to groups that the White House didn't have any knowledge of.

38:42And then once they found out, the White House, you know, this is really the role of what the White House wanted is, look, we have classified information that you don't have. We know that there are certain groups that shouldn't be getting this. and Anthropic expanded the pilot without consulting, which is, again, sort of bewildering because the whole point of what they were supposedly doing by trying to get this pre-approval regime was to consult with the government. So really, that was the predicate for this. And apparently, again, this is according to the Washington Post, that's when conversations started about whether a model like Mythos needed to be export controlled.

39:18Then you have the launch of Fable, which is basically mythos with guardrails. Again, if those guardrails fail, then you have a cyber weapon being released to the public. And it wasn't the White House who came to the conclusion that the guardrails failed. It was private companies who were testing Fable, testing those guardrails. And as has been publicly reported, one of them was Anthropics largest shareholder and cloud partner. And so, well, again, I'm not going to confirm, but I think we all can read the publications. Okay. Yeah. So again, normally I wouldn't mention names at all, but this is all widely reported.

39:58Okay. In any event, they, I think, tried to convey their findings to Anthropic. I think there was some sort of breakdown in communication. I don't really understand, but they felt the need to escalate it to the White House. And it's almost like a whistleblower, okay? Because they have no reason to lie about this. I mean, again, they are Anthropic's deepest partner. Yeah, I mean, they're partners. So that's even reason to support Anthropic. Obviously, Saks, they would have told Anthropic first. But the fact that Andy Jassy got involved in this at all means somebody on the team must have said, hey, on the security team at Amazon who has to host these models said, hey, this is something that could be a liability for us, Amazon.

40:42and could be a national security threat. That's why Andy Jassy made that call because somebody put that into a Slack room somewhere or documented it. That could be - They don't do things like that based on a Slack room, okay? They had teams - Well, somebody in the security team reported it, right? Amazon is part of Mythos Preview and they are also testing Fable. Remember, they've got the biggest cloud to protect. We're talking about AWS. They've got teams banging on this and making sure the guardrails are safe. And as has been publicly reported, they came to the conclusion that there was a jailbreak and that this was a serious security problem.

41:17So they then escalate to the White House. And again, and this is where my summary takes over based on me hearing firsthand from these officials. They tried to call Anthropic. They tried to call Dario. And there were a number of conversations. And the TLDR of it is Dario refused, meaning refused their request to take down Fable until this jailbreak could be fixed. And again, the White House was really just, I think, bewildered and confused by that because, again, it's so off-brand for Anthropic. They claim to be the AI safety company. And yet here, the White House is reacting to credible information about a major security threat from Anthropic's own partner.

41:59And instead of basically saying, yes, we'll react on this right away, they get the Heisman. I mean, just imagine this call. Apparently, again, reportedly, the Treasury Secretary of the United States called Dario personally. I mean, why wouldn't this be a five-minute call? Why wouldn't Dario just say, yes, sir, we take security more seriously than anybody else. We'll fix this problem. Thank you for taking the time to call us out of your busy day. We're going to handle it right now. Click. And instead of doing that, what the administration heard, and I'm not saying maybe Dario didn't mean to communicate this, But he sort of pedantically argued that this jailbreak wasn't really a serious issue.

42:36And we know that to be true because you don't have to take the administration's word for it. Dario then published a blog post that basically conveyed the idea that, well, not all jailbreaks are that bad. And trying to differentiate between good jailbreaks and bad jailbreaks. Minor ones, major ones. Minor ones, major ones, whatever. You know, again, very off-brand for them. I didn't know until now that that was the case. Remember, this is all downstream of the fact that Dario primes all of these officials to view Mythos as a cyberweapon and to view, therefore, Fable as a cyberweapon if the guardrails failed.

43:11And they already had problems with the Mythos preview being overextended to parties that they didn't approve of. So, look, I think maybe there's room in here to give Dario some grace. Maybe he didn't interpret this correctly, what have you. But this was the way that the administration perceived the situation that they were in as of whatever that date was a week ago, last Thursday. And they reacted by sending the export control letter because they wanted Fable to come down until the guardrails were put in place. Now we're in a situation where Anthropic is negotiating with the government to try and fix the situation.

43:51Look, once you're in one of these situations, it's always harder to extricate yourself than to avoid getting in them in the first place because the trust is broken. The trust is now gone. And look, I hope this does get resolved as quickly as possible because I don't view what the administration did as a policymaking exercise. Some people are saying that this export control letter is now a new type of policy. Now the government has to approve every model release. I don't think that was the intention. I think this was a reaction by the administration to a credible report of a national security threat, and they felt that they had no alternative but to react immediately with the tools that they had available to themselves.

44:32So I hope that this problem gets fixed, and then we have a serious deliberative policy conversation about what the right hoops should be for models that are cyber weapons and reviewing those models which yeah let's pause there because we're going to get into model reviews and what's the best way to do that chamath what's your take on this situation uh in terms of maybe a little bit of i don't know your handicapping of the situation with the relationship with amazon i think that's kind of an interesting angle or if you have a different angle you want to take I think the leaders of the frontier labs leave a lot to be desired.

45:11Okay. I think what we're seeing is a consistent pattern of evasiveness and immaturity. And I think that that does a huge disservice to the entire movement of AI. Maybe to connect the dots between this and what Freeberg said, I couldn't agree more with what he said. The key to a vibrant life is rooted in economic mobility, and I think AI is the grand leveler. It is the thing that can enable everyone to have unique amounts of economic mobility because they are unencumbered to figure out what their upper bound is. And against that backdrop, which is an incredibly positive sum view of the world, but I think is accurate.

46:00We have to live in this constant doomerism, hype cycle, naivety, and I think it holds us back. How does it hold us back? Tactically, number one, it creates mistrust. I think that Silicon Valley was already decaying in the prestige that it held in American society. When I first came here in 2001, it was a place of misfits that were good intentioned, that were doing the right thing, that didn't take themselves too seriously, that in success wasn't ostentatious, that kind of kept their head down. We built important things. Then we veered away from that and we started building less important things.

46:45and now we're at a point where we've potentially started to rebuild important things again but we have this veneer of negativity and mistrust that are created in large part because we just cannot get our shit together and the leaders of the frontier labs are public enemy number one and so this entire episode is yet another proof point that you just can't trust these guys that's number one. Number two, I think what it creates, which I think is bad, but what it creates is an incredible opportunity for the hyperscalers. And the very simple opportunity is to convince governments all around the world, not just America, that they should be the gatekeeper.

47:33A, you can't trust these guys. B, these models are all over the place. C, let us be the ones that provisions them to the world, we will wrap it in KYC. I've been now talking about KYC for a while, right? Who are these customers? Do they have identification? Why are they allowed to run these models? What are they prompting? Let's keep them so that there's an audit trail. All of these things are going to become issues. The Frontier Lab folks made it an issue because of how they've handled all of this up until now. And what does that create? Now that creates an oligopoly for AI, the most powerful, economically leveling instrument we've ever seen, in the hands of maybe a handful of hyperscalers, who, by the way, would make an incredibly compelling argument, and they would be right.

48:23And the only counterfactual to it would be, well, trust us, guys, it should actually be much more open in a far more distributed environment. Can you imagine the cost and the complexity if you asked the neoscaler to build the same robust KYC or the same VPC infrastructure that Amazon and Microsoft and Google have spent decades investing trillions of dollars in, it's an impossibility, Jason. So you can take all of those data centers off the map. You can take all of the Neoscaler market off the map. All of this was preventable. So instead of a diverse, robust, open ecosystem, giving a tool that is the fundamental unlock for humans.

49:05We are now going to debate gatekeeping and duopoly versus oligopoly. They have created a total mess. And it's a shame. So instead of people just to summarize that instead of people being able to use the model they want from the provider they want, you're going to have to put in your driver's license, your tax ID, your social security number, whatever it is to Amazon Web Services or if you have to put in your driver's license to buy fertilizer, you're sure we'll have to put in your driver's license to use Fable 5, which is the precursor to know how to use fertilizer to make the bomb. Yeah, as we talked about last week.

49:40Sorry, last thing. Total own goal. And then what does it do? We in Silicon Valley, yet again, look like people that have just optimized for our own selves. We take ourselves seriously. We take nobody else seriously. We're not willing to play ball. It's unserious. It's immature. And it does not help society at large okay free bird there's been some talk about political tribalism here and i'll let you take this one anthropic the only company that doesn't seem to show up when president trump has ceos at the white house backed by reid hoffman who funded the lawfare against the white house gene uh carroll lawsuit this lawsuit that lawsuit there's major beef between the left and the right and anthropics in the middle of it is this the white house treating anthropic differently and do they have it out for anthropic and anthropic is basically handing him the ability to take these kind of actions in other words is this political instead of pragmatic and practical in your mind well i mean i think that there there's no regulatory framework today.

50:53And so antagonizing the government to create a regulatory framework, and then feeling antagonized when they regulate you just seems a little bit weird. And Anthropic seems to be the only one doing that. So I don't know if it's a political motivation as much as it is like some strange set of behaviors from Anthropic that that's motivating this perception. look i think at the end of the day we're all losers if the government is going to be picking winners and losers in ai and we're all losers if we all have to be restricted on accessing these tools at the end of the day like i've said before you know an entire ai model can fit on a usb drive and you just need to have good hardware to run it and the value of that model is time bound because at some point a better model is going to come around.

51:49So my point being, I think we're all losers if there's one stack that one company owns the whole thing, and we all have to put in our ID and the government decides who's the winner and who's the loser. None of us are going to be winners in that sense. That's why I think that there's currently a natural market force for two things. One is more open source. And second is a fragmentation of the AI stack. If you go back to the 1960s, IBM was the predominant provider of mainframe computers. And I don't know if you guys remember this or know this, but like, at the time, IBM did the whole stack, they actually made the chips.

52:28Yeah. So all of the, you know, the I think it was solid logic technology chips that IBM created, were were made in house, they made the hardware, they may be operating system, they also made the software. And this was the case through the whole mainframe technology era. that IBM was the monopoly, they had the whole thing. And then ultimately, the government had to intervene to get them to disaggregate the software layer. And that's what actually started the independent software vendor era, which is when the software industry really started, was after the government said, hey, we shouldn't have one company that has the whole stack in computing.

53:05Then we get to the 80s. And that's when everything disaggregated, then IBM started buying the Intel chip for the personal computer, they had their own OS, but ultimately, Microsoft came along and made a better OS. And then this entire proliferation of the software stack with independent vendors took off. And I think that that ultimately is what allowed this industry to progress the productivity boom that arose from both mainframe computing and desktop PCs was because of this disaggregation that happened. I think the same is likely going to have to happen with these AI companies. And, you know, I'm not sure that what we're looking at today as being kind of some oligopoly of three or four businesses is really where we're going to stand.

53:44I think the equivalent on the hardware side is there's going to be multiple chip vendors. There's also going to be multiple clouds. There's also going to be multiple models that run on the clouds. People are also going to run models locally, they're going to use different software factories, they're not just going to use Anthropic, they're going to use tools like Salesforce and, you know, 8090 and other things, they're going to run different applications on top of it that are going to be purpose built or custom built. So I think there's very likely kind of a diffusion that's happening right now.

54:10If it doesn't happen, it's going to be pretty nasty. But I just think the market forces are so strong for a diffusion of each layer in the stack, finding its way to the cheapest, the more available, the more accessible partner vendors that we're very likely going to see these kind of, you know, handful of oligopolies kind of break down a little bit. Yeah, I that's my that's my take on this. I think it's I'm not sure there's going to be a political capture that's going to happen. It just seems very unlikely. I'm going to steal man for the good of the pod. And also a little bit of my personal belief here.

54:44I think if we were to take Anthropics name out of their series of behaviors in how they handled the release of Mythos, let's just say we put Brock in there, OpenAI or Gemini. Gemini is a good one, Google. They're not totally in either camp. It's a company that's been around for a long time. So Gemini comes out with a model. When they test it, it's incredible at hacking and cyber warfare. and they say, you know what? This is incredible. It's super powerful. We're going to limit it to 50 partners. And those 50 partners include Microsoft, et cetera. They happen to have SK Telecom in there. They make a mistake.

55:19They put them in there. And we looked at them and we said, hey, holding the model for 30 days while they test it, we'd say, thumbs up, good move. And then one of their partners, Amazon, says, hey, we were able to jailbreak this. And they say, okay, we're going to put a couple of guardrails on this when we put it out, or let's just say second phase after they give it to their 50 partners in control beta, they said, we're going to open it to the public, but we're going to watch every single thing you do. I think, David, you said this is a privacy issue. It's pretty standard, actually, for when you release something like this to monitor it for the first 30 days.

55:52If you don't want to have people looking at the results, then don't join the beta. Keep using 4.8. Don't use the newest one. And so you'd say thumbs up to that as well. Hey, they're giving it only to logged in customers. And if somebody does something dangerous, like they try to get it to do nuclear weapons, or last week when I asked it about fertilizer regulation, it dropped me out of the model, right? We would give a thumbs up to that. And then if we were asked, hey, by the government to take it down for anybody outside the United States, and we said, you know what, we'll take it down for everybody, probably give a thumbs up to that too, if Gemini did it.

56:27So there is a definite backdrop of beef between Anthropic and this administration. And if you look at the Reid Hoffman connection, you look at Dario, who refuses to participate. The people who go work at Anthropic are the people who hate Trump, who hate the Trump administration. They're opting into that. If you look at OpenAI and Grok, those are the people who are largely pro-Trump, including Greg Brockman and his wife, who gave$25 million to Trump. So there's a political underpinning here. In fact, you may not be able to get into this. I understand that. But that's part of what's going on here.

56:58I think Anthropic actually did a really good job of rolling out this model and being conservative about it. I think their comms are the problem. Every time Dario talks, it's too much talking and there's too much fear-mongering. So somebody's got to fix the comms over there to your point, Chamath, because the whole industry is going to need to be regulated. And I think the industry needs to regulate themselves. That's the key to this. We need to have a set of tests that Google, Microsoft, Amazon all agree to, Elon, hey, these are the things we should test, and they should self-certify each model before asking the government, which doesn't understand the models, to certify them.

57:36The industry should have an industry certification like they do for countless other things. I've talked about the MPAA and the video game industry. We should just self-certify. It's the simplest thing in the world to do, and then we could release the models ourselves without the government getting involved. Anybody else have any thoughts on this before we go on to our next topic? Well, let me just respond to this accusation. You're saying that this is somehow politically motivated. There's just, that's just not true. Look, there are Anthropik aligned influencers who are trying to spread the story that all of this is basically a beef between me and Anthropik.

58:10They're trying to say that somehow I did this, which obviously is untrue. I don't have the power to do it. I'm not a day-to-day employee of the White House. And I only got the readout of this once the letter was sent. So, you know, I wasn't involved. Secretary of War Pete Hegseth, he wasn't involved. People were trying to make it sound like this was a beef with the Department of War. Just not true. This happened as a result of the things I mentioned, which is they got a report of a jailbreak. Dario had the chance to fix it. He basically said no. Maybe that's not the way he thought he was communicating, but that's what the administration heard.

58:43Dario doubled down that blog post and had the prior history of expanding the Mythos preview to at least one company that the administration felt was a national security threat. That's the reason it happened. It has nothing to do with the people. And by the way, it was the people in the administration who were most sympathetic to Dario and Anthropic who wanted to reset this relationship after the Department of War. And then Dario basically blew it with those people. So, look, I mean, you can just make stuff up if you want. Oh, no, no, I'm not making it up. Just, you know, I'll take offense to that, Sax, because it's your guy, Pete Hegseth, who is doing tweets like this.

59:22Three months ago, Department of Work kicked Anthropic out of the building forever. Every passing day proves why this was the right move. I'm not making anything up here. There's serious beef between these folks. If you're saying that Pete Hegseth has nothing to do with this, read the tweet. That was after the letter was sent. Check the timestamp. I get it. No, no, it's totally fine. there is beef between Anthropic and the administration. To say there's not is just silly. I didn't say there wasn't a prior history. I'm just saying that that wasn't the reason why this happened. This was not a dispute between Anthropic and Department of War.

59:52This was not a dispute between Anthropic and me. We were not the parties who were involved. I don't think you're involved in it. I didn't say that. But neither was Department of War. Now, what Hegs' tweet reflects is he feels vindicated. There's a sense of vindication that, hey, we had a really difficult time with these guys, when we did a contract negotiation, people may have thought that we were the difficult ones. In fact, we said they were difficult and communicated in a strange way. And now everybody else can see it too. So I think what Hegseth is saying there is like, hey, we told you so.

1:00:26But he did not make the decision here. This had nothing to do with it. And look at the timestamp of that. It was after all this whole series of events went down. If I'm a shareholder in Anthropic, I would have told Daria to show up every time Trump had the CEOs over there. And I would have told somebody in that management team to drop a$10,$25 million check into some ballroom or some America pack, just like Greg did at OpenAI. He really didn't need to do that. All he needed to do is that when a cabinet secretary calls you and says that we have a credible national security threat, report to us by one of your own most trusted partners, that he take it seriously.

1:01:08I mean, he took it down. How more seriously can you take it than taking the whole thing down? He didn't take it down until after they got the letter, which was the whole point. Yeah. Anyway. Can I jump in here? Please jump in. Yeah. I had Claude read Machines of Loving Grace and AI Policy on the Exponential. And then I asked it to do... What is that? Those are his two long form essays. Dario's. Dario's. Just to be clear for the audience. Yeah. And I asked it to do a psychological analysis. And well, no, I'm not joking. I mean, I think it's important, you know, and use these as insights into his personality and give me a psychological analysis of him based on these as well as based on the current facts on the ground with mythos be truthful don't protect simply because you are an anthropic model and then i asked it to refine through the lens of eac movement and tell me if there's any psychological leanings do you want me to read to you movement just so people know is the accelerator movement to accelerate society and technology as opposed to decelerate uh okay i think people understand yeah let's can i read this okay two paragraphs delusions of grandeur no well let me i'm not saying it to make fun of him look he's built an incredible business the shareholders are are doing phenomenally well i think the revenue run rate of this business continues to accelerate all of it's incredible but i want to understand this person i am paying millions of dollars and so you know i just want to understand here we go okay he distrusts other labs positioning them as racing recklessly.

1:02:44He distrusts authoritarian states because he says they will weaponize AI. He distrusts markets to distribute the gains. He distrusts institutions to move fast enough. And post-mythos distrusts the government to wield power, quote, transparently and fairly. That's a very long list of untrustworthy actors. The list of trustworthy ones is conspicuously short, and it has a suspicious tendency to resolve towards people who reason the way I do operating under the rules I help design. When your safety framework requires that someone hold the keys and your analysis keeps concluding the other key holders can't be trusted, you've built a machine that outputs me no matter what you feed it.

1:03:21EAC would call that a god complex. The more precise term is epistemic exceptionalism. Not that I am superior, but more my reasoning is the load-bearing one, and the failure of others to reach my conclusions is evidence of their corruption or slowness, not of my error. That's the genuinely concerning pattern, and it's worth naming plainly because it's the engine of the dangerous version. Once you believe your judgment is the uncorruptible reference point, every conflict gets recoded. Government overreach isn't powerful actors have competing interests and I lost. It becomes they failed to be transparent and fair.

1:03:56Losing the Pentagon fight isn't a negotiation he misplayed. It's evidence the system runs on leverage instead of reason. The mythos collision is we shipped a dangerous capability and a hostile state reacted predictably. It's a misunderstanding. The word his company actually used. Notice that misunderstanding presumes that if everyone simply understood correctly, they'd agree with him. That word choice under stress is the closest thing in the public record to the tell. It's not megalomania. It's the quieter, more defensible feeling conviction that disagreement is downstream of error. I thought that was really interesting.

1:04:34I mean, yeah. Well, I think it's like super accurate. I mean, look, there's no question that Anthropik's whole mantra, and actually Ben Thompson pointed this out in one of his pieces, is that they believe that AI is super dangerous and only they are virtuous enough to basically control the negatives of it. I mean, that is basically their view on it. They're the Jedi. They're the Jedi. They believe that they will bring balance to the universe and that they are the ones who can adjudicate all of these rulings. Yes, they believe they're superior to everyone else, which is why they are a magnet for talent.

1:05:07Now, what I would say is I don't think that they have a particular problem with the Trump administration per se. What I would say is they got spoiled by the Biden administration because they had totally captured the Biden administration. And the proof of that is the fact that the leaders of the Biden administration's AI policy, not just one of them, but all of them went to go work in Anthropoc like the minute. the Biden administration was over. I'm talking about officials who were in charge of AI for the NSC, the ones who were in charge of the new AI Safety Institute, which was something that Anthropik basically set up under the Biden administration.

1:05:46The person who is my predecessor effectively as AI czar, they all went to go work at Anthropik. So there's no politics going on here, except it's incredibly tribal. Okay. No, I just think that they got used to the fact they, I mean, they were incredibly organized for a startup and prioritize government affairs from an early stage, maybe before any other AI company. And I think Dario went to Washington very early on and convinced key officials in the Biden administration of this idea that AI was extremely dangerous and needed to be controlled by the government in some sort of like merger between the government and then a very small number of hand-selected companies, two or three of them.

1:06:27It was that message that Marc Andreessen heard, which is what caused, remember this, when he went to go meet with the government and they said, don't waste your time investing in different AI companies. We're going to anoint the winners. There's only going to be a couple of them. We're going to create a cartel. I think the way that Anthropic describes it is that they describe competition between AI companies as a dangerous race condition, meaning they all just start racing towards AGI and therefore safety falls by the wayside. And they want to centralize and control. They see competition as a nefarious force.

1:07:00So they want centralization. They effectively want to create a cartel, and that is their view of AI safety. Now, you can see how they could see it in their language as being virtuous. But if you see it from the point of view of what's really happening, which is they're trying to create a monopoly or duopoly or cartel, it's incredibly self-serving. And that is the danger. And I do think that a fundamental philosophical question is whether you think competition is a good force or a bad force. In my opinion, competition is a good force. It's what protects consumers. It gives consumers choice. It also brings out the best in competitors and prevents regulatory capture.

1:07:36It balances the power too, right? You don't have one person who you're dependent on. Yes. And it has a greater chance of leading to decentralization. I think centralization is the greatest threat of AI, right? You don't want it to become a totalitarian force. So there's a fundamental ideological difference here. But look, you have to say Dario was incredibly successful at resetting the conversation with Mythos. You go back to, I think it was April, and they had just been kicked out of Department of War because of that unsuccessful contract negotiation. And then all of a sudden, they come to Washington and declare they've invented this cyber weapon.

1:08:12And they basically freaked everyone out, and they got everyone to basically go along with their idea of some sort of government approval regime. And it's only because I think of their own, I don't know, you could say hubris. Hubris or delusions. That they misplayed it. Yeah, their delusions or hubris, and they misplayed it. And it ended up getting kicked out under their own nascent pre-approval regime. Hey, Freiberg, just one last question for you. If we do tip over into superintelligence, and it becomes like cataclysmic for humanity, and we look back on this conversation, what would have the right thing to do been for the government?

1:08:48Again, there's this important force that needs to continue to maintain its strength, which is to avoid the natural instinct to control this stuff. This happened with the desktop computing revolution and the mainframe computing revolution. Do you guys know there were articles published in Newsweek and New York Times in 1961 that talked about the demise of the workforce because of the computer. He said, we've now automated everything. Humans are irrelevant. No one will have a job. The workforce will go away. It was an existential threat in the media at that time. And what happened? The mainframe computer drove productivity.

1:09:32Every individual that worked was able to do 100 times more work. Then the same thing happened with the desktop computer. And that also enabled economic mobility and it enabled productivity gains enabled new industries. And everyone said when that happened, this is the end of the workforce, no one's going to be able to get a job anymore, everyone's getting replaced, it's over, we've got to retrain our entire workforce. In fact, the federal government did this deal in the early 80s, they spent hundreds of millions of dollars in workforce training programs that literally did nothing. And it was all about Oh my god, we got to get our workforce ready to transition to the desktop computer industry.

1:10:06You don't ended up happening, people figured out. It wasn't that difficult to turn on a Windows PC, right? And companies figured out how to help employees navigate and transition to higher levels of productivity by taking advantage of this new technology and these new tools. The same thing is going to happen here. So this whole notion of AGI wipes out everything. And this is the end conclusion. There's a deep arrogance to technologists, a deep arrogance, and we know it. It's what drives all of us. But that arrogance always leads to these existential conclusions This will cure all cancer. This will solve all disease.

1:10:40This will eradicate all jobs. None of us. And by the way, no one's immune to this arrogance. From Elon to Sam to Dario to all of us. We all assume that this time is different and everything's about to change completely. And you know what? It doesn't. We're all part of this long-term continuum of technological and productivity improvements driven by human ingenuity, which unlocks human potential and gives humans more capacity to do more. And it's awesome. And there's no job loss and narrative. the end of the world. And that is driving us all into an insane frenzy. Everyone needs to chill the f*** out and just let technology do what it does.

1:11:14To your point, when the desktop computer... These companies need to stop doom trolling. They need to stop doom trolling. Do you see that New York Times op-ed? Look, this is all downstream. This is all downstream of the fact that these AI CEOs have been scaring the daylights out of the public with respect to AI. It was totally unnecessary. And by the way, J.K., I just want to say one other thing, which is I am not happy about this result. I'm not happy about the fact that Anthropic is in this predicament. Why? Because I don't like the precedent. You know, I hope this does not become precedential.

1:11:45I don't want policy being made out of these types of emergencies. I trust it won't be. I hope this won't become the basis for policy. I hope it's a one-off. You think they should regulate themselves, as I've been saying? If you go back to April, here's the thing that's really crazy, okay? I just saw that Dario did this interview with Bloomberg, you know, Emily Chang. And he said that by not releasing Mythos in April, the company suffered massively commercially. You know, he's acting like he's doing this great and noble thing for humanity, this great sacrifice. The truth is that if he had released Mythos with no guardrails in April, he would have exposed the company to massive legal exposure because thousands of companies would have gotten hacked.

1:12:26And it would have been very easy to say that was a foreseeable consequence of releasing Mythos with no safety features, no guardrails. Look, right now, Meta is being sued. They just lost a multimillion-dollar verdict because a user accused them of body shaming. Meta and Google are both being deluged with product liability lawsuits right now. So you're telling me that it was this great sacrifice for humanity that you didn't release Mythos when it was patently unsafe back in April? Of course, you had to include the safety features. Of course, you would have expose yourself to liability that could have put you out of business.

1:13:00That is just good business practice. That is corporate responsibility. You shouldn't deserve this giant pat on the back, and you didn't need to go to Washington and freak everybody out like this. I mean, you should have quietly done it. You should have quietly gone to the NSA and said, listen. They're going to get kneecapped by the hyperscalers. It's a clear lane for them to say, let us be your trusted partners. And by the way, the hyperscalers, if you look at their balance sheets, the hyperscalers are out over their ski tips. They have trillions of dollars of on-balance sheet and off-balance sheet exposure to AI.

1:13:31The best thing that they can do to underwrite all of that money is to kneecap the Frontier Labs and play gatekeeper, charge a toll, take a tax. It is business 101. They're going to say they're the adult supervision. And by the way, as far as regulatory capture 101, yes. That's not regulatory capture. Yeah. Well, they're creating a market opportunity for themselves because these companies have doom trolled their way into it that they've doom trolled their way into it it's a public mass distrusting us a policeman that's what i'm saying captures when you surreptitiously in the dead of night swoop into washington and introduce some bill as a christmas tree ornament in a christmas tree bill this is literally the exact opposite with the same outcome i i agree with you jason which is you bumble your way into getting knee recapped by everybody around you.

1:14:17No, no. I guess that analogy would be they literally walk themselves into prison and hand the hyperscalers the keys and we're like, please lock us up. They closed the door. They closed the door themselves. They locked it. They threw the keys out the window. Yes, we're locking ourselves up. All right. The Iran war might be ending. After 110 days, President Trump has announced for the 37th time that the war has ended and that the initial agreement on june 15th will be codified a lot of details here a lot of unknowns conflict began obviously on february 28th formal signing happens friday june 19th in geneva deal was mediated by pakistan extends the ceasefire for 60 days includes a lebanon ceasefire u.s gets the strait of hormuz reopening which it was open before the war obviously Iran commits to not developing nuclear weapons.

1:15:08Iran agrees to destroy its stockpile of enriched uranium under IAEA supervision and that they will freeze its nuclear program at current levels for 60 days. Iran gets all sanctions lifted and$300 billion in reconstruction. So we spent$100, $200 billion blowing the place up, and we and our partners in the Gulf are going to spend$300 billion, half a trillion. We're not on the hook for a dime, J. Cal. Okay, I was just saying, there's some people in the Gulf who are going to be involved in this, and we'll get into it. Iran's going to get access to their frozen assets, and the U.S. is going to remove its forces from the region after the final deal.

1:15:52A lot of stuff not defined yet will be deferred, like Israel's sign-on to this deal, and a bunch of spicy comments happening there. And Iran's nuclear enrichment going forward, that's not resolved. Iran's ballistic missile program, that's kind of important. That's also not yet resolved. I'll stop here. Obviously, there's a lot in play. Many people said the reason to vote for Trump was because he would never start wars. Started a couple. What's your take on where we're at today, Sachs? Well, I think this deal is a tremendous achievement for the president. I think that it was very difficult to get.

1:16:28I think the Iranians are not easy to negotiate with. Let's review this deal. Number one, it reopens the Strait of Hormuz. That means all the oil will flow and all the other vital materials that are necessary for the global economy. So that's number one. Number two, we have a commitment from them not to pursue a nuclear program and to collect all the nuclear materials. Number three, there's a ceasefire of the war on all the different fronts. Four, this will not cost us a dime. There is a reconstruction fund, but we're not paying for it. I don't see why that's a problem for us. Who's paying for it?

1:17:00I guess it's something that the Iranians will do in combination with the Gulf states, but the US is not on the hook for that money. We're not paying a dime. And then lastly, this could lead to a rapprochement with Iran, which I think could be a good thing because I don't think we need to be at war with that country forever. And look, what is the alternative to this deal? I'm seeing a lot of people, it's easy to take potshots and criticize the deal. But what is the alternative to this? I'm hearing neocons who basically want us to put ground troops in and try and effectuate a regime change in Iran.

1:17:33This is what I'm hearing. I heard John Podoritz say this. I'm hearing other neocons say this. I mean, are you serious? Are we really going to escalate this war by putting ground troops in? Who exactly do you want to go fight? What I heard John Podoritz say is, well, we have a volunteer army. So they signed up for this. No, they didn't. They signed up to protect America, not to go charging into Iran, which is basically a mountain fortress. I've heard estimates it would take over a million troops. And even then it might not be successful. We send half a million to Iraq and Iran is three times bigger as a country and it's a mountain fortress.

1:18:07So that would be a suicide mission. That'd be insane. Nobody wants that. I'm not sending my kids to fight that war. You don't want to send your kids to fight that war. If John Podores and the neocons want to do it, let them send their kids and their family. If Reza Pallavi wants to go to Beverly Hills and muster an army from his legion of supporters, let him try and do that. I don't think they will sign up for it. I think they're living too well in Beverly Hills. Okay. Nobody wants to fight that war. That is the alternative here. Okay. So if we're not going to send in ground troops and we're not going to continue the bombing, because what's the point?

1:18:41We already hit every, we already flattened every military target above ground in that country. So what is the point of continued bombing? Then we might as well, hold on. My point is, if we're not going to send in ground troops, because it makes no sense, we're not going to continue the bombing because it makes no sense, then we might as well try what's behind door number three, which is a deal here that will try to create a peace. Remember, we're only at the MOU stage of this, but let's give peace a chance here. I don't understand these people who just want this war to go on forever. Haven't we tried that before?

1:19:13The forever wars of Iraq and Afghanistan? Are we really going to sign up for another one of those? I don't think so. So look, it's easy for people to take potshots at the president here, but I think this is by far the best of all the alternatives. Saks, the long-term issue has been the enrichment of uranium to the point that it can be used as a nuclear weapon. Is it part of the deal that all that enriched uranium that we do know there's a stockpile of gets removed from the country because if that goes i personally feel very good about let the citizens deal with what they have to deal with that's what they've said this is part of the they will not only agree not to pursue a nuclear weapon they will give up their nuclear material this is that's a big deal if all that enriched uranium comes out of the country i think it's a very big deal because even if they restarted the program even if they resourced the equipment rebuilt the intelligence to do it, etc, etc.

1:20:06It's going to take years to a decade to 15 years plus to build enough enriched uranium to build a nuclear weapon. That would kind of be the natural setback, if you will, that provides some defense for the conditions that we really care about. I didn't realize that was in the deal. I thought the deal was just that they would make a quote commitment to stopping a program. But if all that enriched uranium is coming out, that's a really big deal. Well, they still have to define it. You know, again, this is the MOU stage, isn't at the level of high-level principles. They have to define exactly how that program is going to work.

1:20:36But what they get in exchange is relief from the sanctions, and they get a more normalized relationship with the U.S. Yeah. There's an incentive there for them. If they're willing to basically give that up. You have any take here? You want to check in or not? I think the market's going to the moon. I'll just add, I'm glad if this gets consummated, and it was a huge blunder for Trump to do this. and obviously he's trying to get out of it should never done it but here we are and i hope it resolves and i hope the people of iran are free eventually you think it's worth it jake out to have gone through what we went through over the last couple of months if all of the iranian enriched uranium is gone their missiles are are depleted and they have no capacity to enrich uranium again i mean just just looking back like if you just put this all together summarily and said hey like there was this big war and the stuff that happened and all the it's a great loss of life and like suddenly this this real risk of nuclear as i've said from the beginning of this we don't have complete information so you know for a venture capitalist capital allocators entrepreneurs can speculate all they want but we don't see what trump sees right so i'll you have to be humble enough to know we don't know what information he had in all of this but the trimming of the grass the the hedging you know of the garden approach that israel had been taking all this time that was the right approach isolating dictators waiting them out is always the better approach than starting full-scale wars with them because you don't know what could happen in an escalation so simply not worth the risk much better to just send the air troops in and clean it up and we did not need to you know go collect this uh you know material obviously because we have these bunker buses and we can just slam it every x number of months and that would have been a much better approach just like we've contained north korea for all this time and you know we contain other dictators with weapons.

1:22:27North Korea hasn't new. What are you talking about? North Korea hasn't new. No, no, but we've contained them. And we actually did have this discussion with Graham Allison about, hey, why didn't we contain that one when we had a chance? That was a mistake by other presidencies to allow them to get this far. And then once they have it, it's really hard. And in this case, it was working very well. Every 18 months, we bombed the facilities. It was fine. We didn't need to start a full-scale war. and we're probably going to just wind up where we were at the start of all this. But I hope that at some point the people of Iran are free.

1:23:00That's my true hope. But people have to, we can't enforce democracy in the Middle East. We've learned that over and over and over again. The people have to take, you know, they have to lead their own revolution. We can't facilitate it. All right, listen, another amazing episode of the All In Podcast for your Sultan of Science, the dictator and the rain man yeah david sacks i am the world's greatest moderator and we'll see you next time bye bye you boys bye we'll let your winners ride rain man david sacks and it said we open source it to the fans and they've just gone crazy with it love you I'm going all in.

1:23:51Besties are gone. That's my dog taking an ocean drive away. Oh, man. My avatasher will meet me at one. We should all just get a room and just have one big huge orgy because they're all just useless. It's like this like sexual tension, but they just need to release somehow. Let your baby. Let your baby. We need to get merch. I'm going all in.

1:24:23I'm going all in.