Novara Live · · Transcript & summary by WhipScribe
It sorts out contested facts on Iran, Nvidia and Anthropic, and gives a clear map of the Iran options with the reasoning behind each.
Tap a chapter to open that moment in the transcript; the link to it is copied for you.
“The blockade is an act of war.” Trita Parsi · at 10:05 — open in the transcript →
“If you actually have a deal, that will make it much, much more difficult.” Trita Parsi · at 17:06 — open in the transcript →
The episode opens with Trump's apparent retreat from Iran war talk after the midterms. Hosts Michael Walker and Richard Haymes frame the show around stories where the facts themselves are disputed. J.D. Vance blamed the war for the Senate loss, and Trump now says he sees no reason to escalate, pointing to satellite coverage of bombed sites and Pickaxe Mountain. Oil fell below $90 a barrel by Tuesday, down from $108 in October, once traders judged a hot war unlikely in the near future.
Trita Parsi of the Quincy Institute argues the risk of war has eased but has not disappeared. His proposed Swiss cheese option has Trump walk away while the sanctions stay on the books and the naval blockade quietly winds down. Asian countries and Gulf states would then invest in Iran and entangle their economies, and a future president would be reluctant to restart the fight. Parsi calls it suboptimal, a choice driven by political cowardice rather than a real settlement, and says Israel would gain a hole in the sanctions wall that keeps its confrontation with Iran alive.
The second half turns to Nvidia. Jensen Huang has admitted an affair with Alice Wang, whom US intelligence memos describe as a Chinese intelligence asset. The hosts weigh whether the timing was a political play and cite a Hong Kong businessman who argues Wang was more likely a white glove fixer than a spy. Reporting says Commerce Secretary Howard Lutnick delayed adding Megaspeed to a blacklist, and his former fund's Nvidia stake raises a conflict question. The show closes on Anthropic's IPO and an AI R&D report that suggests recursive self-improvement may already be under way.
Trita Parsi is a Quincy Institute figure who wrote a new piece on how Trump may walk away from Iran. He joins the host for an interview segment on the show.
Trita Parsi said the situation is somewhat better than a few weeks earlier, but he has not yet seen enough data to conclude the risk of war with Iran has receded. He pointed to Trump's statements about moving away from military action, but said they were not yet backed by changes in troop readiness or asset movements to the region.
“But a broader situation in which we can say that the risk of war has really receded, I don't see yet.”Trita Parsi · 6:22
Trita Parsi argued that the US naval blockade of Iran is itself an act of war, so the two sides are constantly on the verge of a larger conflict while it stays in place. He said his proposed walkaway option would have Trump gradually ease the blockade and sanctions enforcement instead.
“The blockade is an act of war.”Trita Parsi · 10:05
Trita Parsi said a walkaway without a formal nuclear deal is not the worst outcome for Israel, but it is not good for them either. He argued that a full deal would make Israel's push for confrontation with Iran much harder to sustain.
“If you actually have a deal, that will make it much, much more difficult.”Trita Parsi · 17:06
Trita Parsi argues that Trump is unlikely to reach a proper peace deal with Iran and is more likely to choose a middle path: walking away while keeping the naval blockade and sanctions on the books. He says the option is suboptimal but attractive to Trump because it avoids the political cost of a deal, and that it would let other countries gradually invest in Iran.
Trita Parsi said the Swiss cheese option is a walkaway in which Trump stops enforcing the naval blockade and sanctions while the sanctions remain on the books. He argued that Asian countries and Gulf states would then start investing in Iran to entangle their economies, so the sanctions would slowly decay. He called it suboptimal, a choice that avoids political cost rather than a real settlement.
Trita Parsi said he has not yet seen enough evidence that the risk of war with Iran has receded in a credible and sustainable way. He said the situation is somewhat better than a few weeks earlier, when a war seemed possible at any time, but that Trump's statements about moving away from military action were not yet backed by changes in troop readiness or asset movements to the region.
Trita Parsi said a walkaway without a formal nuclear deal is not the worst scenario for Israel, but it is not good for them either. He said it leaves a hole in the sanctions wall and gives Israel another opportunity to push for the escalation and confrontation it has pursued for 25 years. A full deal, he argued, would make that much harder.
Nvidia's CEO Jensen Huang has confirmed a romantic relationship with Alice Wang, whom US government memos describe as a Chinese intelligence asset. Reporting from the New York Times says their association went back to at least 2023, and Wang's WeChat showed photos of the two together. The hosts note that most of the new reporting rests on the memos circulated inside the Trump administration, not on new original journalism.
The Financial Times reported that several US government agencies pushed for Singapore-based Megaspeed to be added to the entity list, but the Commerce Department, which chairs the panel, sought over the past two months to delay new inclusions. Commerce Secretary Howard Lutnick was also sent a memo in spring about Alice Wang and took no action, according to Axios. The hosts point out that Lutnick's former investment fund still has a large stake in Nvidia, held now by his children.
The hosts said that each of Anthropic's seven founders is selling exactly $250 million worth of shares when the IPO takes place, which is the cash they will receive. Their overall shares are expected to be worth about $18 billion each. The company was first priced to target a valuation of $1.4 trillion and was then repriced at $1.5 trillion after strong investor demand.
The report, produced by an outside evaluator with access to Anthropic's internal data between August 12 and October 29, 2026, estimates that Anthropic is improving its models on well-specified tasks at around twice the rate it did between March 2024 and October 2025. It attributes most of that acceleration to Anthropic's own use of AI in R&D, with high uncertainty. The hosts treat it as evidence that AI may be helping build AI, the pattern long described as recursive self-improvement.
Our own transcript of the public episode, with a timestamp about every 90 seconds. Tap a timestamp or the link button on a paragraph to copy a link that opens this page there.
0:00Welcome to Navarre Live. Coming up tonight, Trita Parsi explains why Trump may walk away from Iran. We'll take another look at the Nvidia spy scandal. Was it all a psyop? And the dark side of Amthropix IPO. I'm Michael Walker. I'm joined tonight by Richard Haymes, who I sense was sort of giggling throughout my trails, but maybe you weren't. Giggling? No, me? Maybe you weren't. Maybe you weren't giggling. This is Richard Haymesmes do your own research we've let him out of that little dungeon where he's looking up to that camera up there first time out here yeah um very exciting we're talking to lots of themes that i know you're very interested in we're playing to your strengths today we're talking china we're talking ai yeah i think it's also appropriate because these are all
0:41stories in which no one agrees on the facts. They're all murky stories where there's a real genuine confusion at play. Yeah, disagreement over facts as opposed to disagreement over opinions, which is actually more interesting. Yeah, yeah. Before we get started, a reminder that our next downstream IRL is less than a week away, and tickets are still available. That will be at Earth Theatre in Hackney on the 16th, and will feature Owen Jones in conversation with Ash Sarkar. In Owen's new book, he argues that the West is in decline, but not because of migrants, Muslims or a woke mob, as Richard Hames would have you believe. And rather, it is the self-inflicted result of elite hubris and decades of neoliberalism.
1:20So you can attend that and see who you agree with, Owen or Richard. My reason is actually much more esoteric than that. They're much more confusing than that. Because of the rise of China in social media. I am actually writing a book about collapse at the moment. You'll have me on to discuss it at some point. I'm sure. I don't know, don't assume anything. No, we will. back in october there was widespread speculation that trump and netanyahu were planning a renewed assault on iran um tucker carlson who is a crackpot but who also happens to have some good connections to trump related circles even suggested the u.s might be about to deploy a nuclear bomb in iran, since the midterms, there has been a fairly dramatic change in tune from the White House and those around it. So it
2:11started with a tweet from J.D. Vance the day after the midterms. He said, the reason we lost is the war. The president himself said so last night. If you'd like to argue it was worth it that losing the Senate for the next six years was worth it, then do so. Argue it was a worthwhile cost to pay. Agree or disagree, that's at least honest. But everyone knows why we lost, and any honest conversation about where we go from here must start there. Then, over this weekend, Donald Trump said this. So you're not gonna escalate the war now that the midterms are over? Well, I'm not looking to. I don't think there's any reason to. I don't think there's any reason to. Look, the whole- They still have nuclear material, right?
2:50They have nothing. It's buried under thousands of tons of rock. And we have, through Space Force, which I created, we have telescopic lenses that are unbelievable. I can tell the color of the red dots on your tie. And we watch every one of those sites, the three that we blew up, and also Pickaxe Mountain. So I think Pickaxe Mountain is another place where they think that there might be nuclear programs operating. But he's saying, our satellites are so good, we know that there's nothing going on there. On Monday, Trump expanded on his Iran stance in this press conference. More oil has come out of the Strait over the last week than at any point in the history of the Harmo Strait. So we're doing very well. But at this point, it won't be necessary. Taking action. It doesn't seem to be a very necessary
3:46thing, but we're locked and loaded, ready to go for any reason that changes. Israel and the entire Middle East are protected from Iranian danger and threats, and we're seeing momentum toward the end of this conflict. I think it's going to be ending fairly soon. So Trump looked this up. He's actually correct about the amount of oil coming through the Strait of Hormuz. So back in October, the Wall Street Journal reported that there was more oil being exported through the Strait than there had been before the outbreak of war. Those ships were traveling under the protection of the U.S. military. But back when that article was written, oil prices remained high as traders were unsure how long that situation could be sustained. So there were $108 a barrel on the 20th of October when
4:33that report came out. However, after those latest comments by Trump, traders appeared to be lightening up. So by Tuesday, oil prices had fallen below $90, the lowest price since August. So they seem to be reassured that yes, oil is moving through the strait and no, a hot war is not about to begin, at least in the very near future. So what does this mean happens next? Well, an optimist might hope that Trump would now be able to sign some kind of peace deal with Iran, reintegrating Iran back into the world economy and ending Trump's botched war once and for all, we can hope. However, analysts suggest a third option, neither outright war nor a proper peace, is the most likely outcome. Trita Parsi from the Quincy Institute has a new piece out called The Swiss Cheese Solution,
5:23How Trump May Walk Away from Iran. And I caught up with Trita earlier today to find out what this Swiss cheese solution might look like. But I started by asking him whether he is now less concerned than he once was that there will be a renewed hot war between the US, Israel and Iran. Not really, not yet. We've seen indications that things may be moving in away from military action, but that is frankly, mostly based on Trump making statements that he's more interested in seeing what is structurally changing and physical matters such as, you know, readiness of troops, assets being moved to the region. And on that front, I've not yet seen sufficient data points that suggest to me that we have reached an inflection point in which the risk of war is receding in a credible and sustainable
6:22manner. It does not mean, however, that we're not in a slightly better situation than we were a couple of weeks ago, in which there was extremely high tensions and a fear that the war could break out any time. That, I think, appears to currently not be the case. But a broader situation in which we can say that the risk of war has really receded, I don't see yet. I suppose in terms of what's changed, the midterms, obviously, Trump has lots of freedom still as the president when it comes to foreign policy. There's not that much that sort of the Senate and the House of Representatives can do other than scrutinize. I suppose it does seem like there has maybe been a shift in attitudes within the White House itself, though, especially sort of people around Donald Trump feeling more willing and able to speak out
7:09actively against the Iran war. I'm thinking above all of J.D. Vance. These are people who want to have the chance of becoming president in two years' time and so are going to be very unwilling, I think, to go along. Well, I will hypothesise, maybe unwilling to go along with a renewed hot war in this case. Do you think that's, you know, does that change things in any meaningful way? That does change things. The question is, is it sufficient? It is extremely helpful, undoubtedly. As you pointed out, J.D. Vance's position, and particularly his tweet, has given cover for a lot of others to speak out more forcefully in certain internal meetings and recognizing, of course, that the warnings that had been given about the tremendous political
7:52cost this war would inflict, not just on Trump, but on the GOP as a whole, have now proven to be not only true, but beyond what even some of the worst case scenarios were that people were looking at. So this has undoubtedly emboldened a lot of people. It has probably also put Trump a little bit on the back foot of these things. But it's also important to understand that because of the fact that the window for starting the war is closing in a couple of weeks because of factors such as the Democrats taking over Congress, potentially defunding the war already by December 10th. We may not have Prime Minister Netanyahu in a couple of weeks' time. All of these things also means that those who are pushing for war are pushing much harder
8:37right now because they know that the window will close and their chances of being able to get this war will probably reduce after January in a more significant way because of the manner that Trump's maneuverability will be limited by some of these other factors. In your latest piece, you set out three possible options. So one is sort of a renewed war. The other is some kind of peace deal. And the third is what you call a Swiss cheese strategy. Could you explain what the Swiss cheese strategy would look like? A bit more of a tongue twister than I was expecting there. So look, let me first start off by saying I don't think that that option is optimal in any way, shape or form. It is absolutely suboptimal
9:19to a negotiated settlement. However, because of the political cost of a deal that Trump is very weary about, he knows that it's going to be difficult for him to get what he wants on the nuclear front. He will be attacked for saying that, similar to what was said about the MOU, that this is a document that essentially manifests the US's failure and Iran's victory in this war. All of these factors that I think makes Trump very hesitant about striking a deal. This old naivete or innocence he had about striking a deal with Iran, in which he seemed to be completely oblivious to the political cause, changed after the MOU and the reactions that he received. And on the other hand, whether he wants to or whether he can,
10:05a third round of a war would be extremely difficult, extremely dangerous. But the rejection of the war does not in and of itself mean that the likelihood of a deal increases. Trump may very well go for a third option, which is kind of like the status quo, but with one very significant difference. It is a walkaway option, but without having the blockade and the sanctions in place. What he does right now is that he has the status quo and he has the blockade. The blockade is an act of war. As long as the blockade is there and the American Navy is there shooting at Iranians and Iranians are shooting at them, we are constantly on the verge of a larger war. But in this option, he would imperceptibly start
10:51to reduce the intensification of the implementation of the blockade and the sanctions as a whole. He would truly walk away, not just away from the strait, but also from the blockade. In that scenario, a new equilibrium can emerge, one in which there's enough of a sanctions infrastructure for Trump to claim that the sanctions are still there, but there's enough Swiss holds, essentially, in that sanctions wall that the Iranians do not have any incentives to challenge the status quo. And you would probably see more countries, not just Russia and China, and to a certain extent, India trading or even investing in Iran, but you will start seeing more Asian countries do so. And you will probably start to see some GCC countries
11:35starting to invest in Iran in order to entangle their economies, in order to create economic interdependence between them, not from a straight investment standpoint, but from a security standpoint, similar to what you saw in Europe after World War II, in which economic interdependence was a key element of the peace. So you will start to see countries essentially feeling that the risk of being caught or being punished by American sanctions will start to go down. And essentially, the sanctions war will start to decay. It won't be a compromise. The sanctions will still be on the book. But that's the only type of a walkaway option that I think Trump would like because it doesn't have the same political cost. That is plausible in my view, because if
12:23you have walkaway, but with the blockade, you're not walking away. You're still at war. It's just a different level of intensity of that war. There's an attractiveness to that. I know, as you say, it would be better to have a full, proper peace deal and there'd be reintegration into the world economy in a more formal way. But if he were to sort of just nudge, nudge, wink, wink, stop enforcing the blockade, that would presumably provide quite a lot of economic relief to the people of Iran, over 90 million of them. Presumably, that would mean that sort of the Strait of Hormuz could start to reopen, inflation in the rest of the world might fall, and we would move away from the risk of a renewed hot war. So, you know, that's better than the position we're currently
13:01in right now. Oh, certainly. I suppose an issue that other countries might have in terms of are we now safe to trade with Iran is that Donald Trump isn't going to be president forever. So if it's sort of I'm not going to enforce the sanctions, so feel free to break them, what happens if then in two years' time there's a different president and they turn around and say, look, India, you've been, you've been, you know, contravening our sanctions for the past two years, and we're going to fine you now. And I suppose if there was, you know, if it was the case that the Democrats sort of started using this as a partisan stick to beat Trump with, you could, you could imagine sort of in theory that they're
13:40being a sort of more hawkish Democrat after Trump, maybe not the most likely option. But do you think that mere possibility could stop this all working? Undoubtedly, a lot of countries are going to look at things and say, well, look, if we're actually going to start investing in Iran, that's a lot of money and a lot of risk, particularly if there's uncertainty of what will follow Trump. But I think what will happen is that when some of the larger countries start to be more active in the Iranian economy, it gives a bit more cover for smaller countries to take some risks as well. And then the next president of the United States, Democrat or Republican, will have to ask themselves, if you actually start implementing the sanctions in a more forceful way,
14:23you are restarting a conflict that you saw Trump could not even finish militarily. Is that a headache you want? The United States is going to deal with a lot of headaches internally and externally. Do you want to add one more to that long list? Are you going to let this one pass? Because if you start it, you're going to have some sort of a conflict or war. If you want to go for a deal, which would take a lot of political courage, you will have to withstand the criticism of striking a deal that many will say, ultimately, the Iranians emerged somewhat of a winner of. Because both of those options are so unattractive, the more cowardly option is to take this Swiss cheese path that I just laid out.
15:03And again, I want to emphasize, I don't think it is good enough. I don't think it is attractive enough. It's just, I have limited faith in the political courage of politicians these days to actually make the right choices. And if there is a choice that is not necessarily the right choice, but certainly is not the worst choice, then I can see them going for it. And I can see a lot of other people being at least drawing a sigh of relief that at least the worst has been avoided. I would say that some of the objections may actually, frankly, come from Europeans, as they're likely not going to be able to go back into the modern economy. Not that they will be kept out, but they will be last in line. They're not going to get any preferential treatment. But there will also
15:42be voices in Europe in which there will be a lot of concerns from the nonproliferation standpoint. And the nonproliferation issue will not be resolved in this scenario. So as a result, there will constantly be this lingering threat of some form of a conflict at some point, which again, creates a degree of instability in that scenario. Not at all as bad as the current status quo. But again, part of the reason why it is suboptimal to an actual deal that fully resolves the issue. And the big loser in this situation, as far as I understand, would be Israel, would it? Because you've got an Iranian situation where the government hasn't fallen, which is what Netanyahu wanted in the first place, and where the economy is not so strangled as it once was. But there is also no
16:26form of formal nuclear deal like there was under Obama, which means that presumably, you know, Tehran will rush to get nuclear weapons. And then, you know, that's the nightmare scenario for Tel Aviv. I would say it's not the best scenario for the Israelis, but it's also frankly not the worst scenario for the Israelis. The worst scenario for the Israelis if there actually is a deal and the sanctions are officially off the books and there is a political open endorsement for integration of Yvonne and trade with Yvonne, etc. This one is not bad, is not good from their standpoint in the sense that there's a hole in that wall, but it gives them another lease of life. It gives them another opportunity to be able to push
17:06for the type of escalation confrontation that they have been pushing for 25 years. If you actually have a deal, that will make it much, much more difficult. So given the circumstances, part of the reason, again, why this may end up becoming the most likely scenario is because no one will win fully, no one will lose fully. Briefly on the midterms from a different angle, the Democrats now have a lot more sort of high profile figures who have a very different stance to the Middle East, to sort of the democratic sort of establishment, El-Sayed sort of above all, then obviously Mamdani, although he's been in that role for a while now. Do you think there's a possibility where we have a government in the United States
17:45in two years time and a president with a completely different approach to the Middle East, and that that can be a real opportunity for, I suppose, a bit of a reset? I think that likelihood undoubtedly exists, particularly if you have the El-Sayed, Mamdani wing of the party really take control of the party, which is not where we are yet. There's still some huge challenges. Some of those decisions will likely be made soon, for instance, whether someone like Jeffries will remain the leader in the House, whether someone like Schumer will remain the leader in the Senate. We just saw yesterday that 11 Democrats did not even show up to vote in regards to Pete Hexett's impeachment. And they could have made all of the difference if they just simply showed up.
18:35They didn't even bother to show up. So I think we have to be very clear-eyed that the big tent of the Democrats means that there's a lot of hawks and neocons in that party as well, and many of them in leadership positions. So there needs to be a significant cleaning up of that party if we are going to go in the direction that you mentioned. We have seen presidents try to go in that direction. Some progress at times have been made. But there is a very, very forceful bureaucratic inertia and interests, financial interests, military interests to keep things as they were before. The one thing that is really a change now that may change that equation in a way that we have not seen before is that ultimately the United States suffered
19:19a military defeat, not just in a political sense, but in a military sense, and that a lot of the American bases in the region right now are inoperable, which means that it's not just going back to the old status quo. You actually have to pay to rebuild these bases. And that raises a new question. Are you willing to go that far? A losing strategy, or are you going to pay more money for it? That was Trita Parsi speaking to me earlier today. This week, the spy scandal of the century broke. NVIDIA has opened up an investigation into its CEO, Jensen Wong, after he acknowledged an affair with a woman suspected of being a Chinese spy. So as we discussed on Monday, this was an explosive story. The head of America's most important artificial intelligence company hooking up with an apparent spy working for America's principal geopolitical rival.
20:17But was this story all that it seemed? Daniela Gabor is an economist friend of the show and posted this interesting theory on Twitter. So she said, blowing up Huang's open secret affair same week that Democrat tsunami swept midterms tells us new Silicon Valley strategy. Ramp up national security angle Amadei been pushing to get restrictions on Chinese AI open models eating their lunch and bailouts if bubble explodes. So she's saying the reason this is coming out now um it's not because you know this has suddenly become public she's saying this was an open secret that this relationship has been going on for a long time um and the reason we're finding out about it now is precisely because there is a play maybe by companies like anthropropic to try and sort of support the hawks in Washington, D.C. by
21:06delegitimizing, essentially, Yen-Sung Wang, who's been one of the people saying, let's sort of work openly with China, and I suppose, sort of creating a bit of a red scare about relationships with China. So that's the theory there. And I think it's very interesting. I think it's worth taking seriously. And I think it is quite clear that the reason it's all come out now is not a coincidence, right? Most of the reporting on the scandal is based on memos that were circulating in the Trump administration for at least six months. So some of them going back to last year, but some of the most high profile memos from Marco Rubio were in spring this year. And then it comes out just after the midterms, right? Surely not a coincidence.
21:46There are also some reasons to believe the relationship between Jensen and Alice Wang was not just something that had lasted for a long time, but was something of an open secret. So reporting from the New York Times this week says their association went back until at least 2023. Alice Wang had posted loads of pictures of herself with Jensen on her WeChat, and we showed you some of those on Monday, which is hardly something you'd do if you were trying to keep a relationship, you know, secret, right? It doesn't necessarily seem like something you'd do if you're a spy. And this is significant, associations between Alice and Jensen Wang had already been made public last year. So in October 2025, the New York Times reported on how Alice
22:26Wang had socialized with Yen-Sin Wang and that her company Megaspeed was being scrutinized back then by US officials for its ties to China. Now that report, it's very interesting by the way, found a lot of circumstantial evidence that Megaspeed had been helping China circumvent export controls on chips. So the claim was that the Singaporean company would buy chips from NVIDIA, because NVIDIA is allowed to export chips to Singapore, but not to China. But then Megaspeed would rent those to AI firms, or even smuggle the chips into China. So it's a front, essentially. Now, if you read that piece from last year, the evidence, as I say, is pretty compelling. They have people sort of turning up to the offices
23:08and there isn't sort of anyone really working there. We say, oh, actually, all the people we work for are in China. And there's sort of boxes packed up full of chips that seem like they're going to be sent on somewhere else. However, that piece didn't claim Alice Wang was a Chinese intelligence asset. I mean, it didn't say they were, or they didn't say that she was in a relationship with Jensen. So this was all the personal details they had on their relationship. So this is what was known in October last year. So they say, it's not clear when Mr. Wang and Ms. Wang, who were not related, first met. She was mingling with a crowd of tech executives just after midnight at the Taipei bar in early June of last
23:43year when she offered to ask Mr. Wang of NVIDIA to join them, recalled Parker Schmidt, an executive who was with the group. She said, I bet you guys I can get Jensen here. And then Jensen immediately showed up, Mr. Schmidt said, referring to Mr. Wang by his first name. Photos posted to LinkedIn by NetFunder, a data software provider where Mr. Schmidt is chief executive, show Mr. Wang at the bar with the group. So I suppose as to the idea that this was an open secret, they were having an affair, this here is not exactly conclusive. It doesn't go nearly as far as the stories from the past week. So what new information was uncovered between now and then? What changed? Why did this suddenly become a story this week
24:23when maybe large parts of it were known already? and then? What changed? Why did this suddenly become a story this week when, you know, maybe large parts of it were known already? Well, nearly all the new reporting appears to be based on US intelligence memos. So memos which were circulated within the United States government, not really sort of original research by journalists, it's them getting tipped off that these memos were getting sent around. And those memos included the claim that Jensen and Alice Wang were in a romantic relationship, which Jensen Wang has now admitted, he's confirmed that, and they referred to Alice Wang as a Chinese intelligence asset. Now that's of course why the headlines have all been about Jensen Wang's affair with a Chinese spy. This is what's
25:00interesting though. So all of that's clear, right? It's clear that this was in the memos, clear that he had this affair, he's admitted to it. But none of the reporting has been very clear on what it means to say she's the Chinese asset, right? Do the Americans believe that Alice Wang was hired by the Chinese secret service years ago to become a professional spy, like James Bond, and then she goes undercover as a microchip executive and then undercover seduces Jensen Wang just because she's been told by her handlers that she must do so, that this is her target. Or was she already a chip executive courting Jensen when the intelligence services pressured her to flip and give up his secret? So they said, oh, we see you're in a strategically
25:41interesting role and we see that you're sleeping with Jensen Wang, you know, because they've got surveillance everywhere. And then they say, we think you should flip. Maybe they pressure her. Maybe they say, we've got information on your family and there's some sort of blackmail going on from the security services to Alice Wang. It's possible. Or might the truth be something a bit more mundane? Desmond Shum had a fascinating sub-stack on this that I think is worth sharing with you. So he's a Hong Kong-based businessman and author. And he suggests the spy narrative might miss Alice Wang's real business, what she really does. So he writes that Alice Wang's career path didn't suggest a role in intelligence.
26:22So by the way, this is someone who's done business in China for decades and has written sort of a very successful book about it, in fact. He suggests that instead, her career path is more compatible with her being what in Chinese business circles is known as a white glove. And he goes on to explain what a white glove is. So he says, a commercial fixer connects people with money to those who control resources, arranges financing, and helps transactions overcome commercial or political obstacles. Their most valuable asset is access. A white glove performs a more specific function, acting on behalf of others while concealing the identities or interests of the real principles behind the transaction. So that's interesting. The idea
27:08that you're saying this is not necessarily someone who's employed by the Chinese Secret Service, but she is someone who is allegedly, I mean, in this account, doing the bidding, I suppose, of other people. So some of the examples he gave is that you might have a company whereby there's a group of people who want to gain control of the company, but they don't want the scrutiny that comes with that. So, they ask various people to buy up some shares. And behind the scenes, they're kind of controlling the company. But from afar, it looks like this is a company with distributed ownership. And the suggestion is that Alice Wang was involved in this because she was well-connected. She was presentable, intelligent, I assume. And so so she could play that role, well-connected internationally.
27:50And he writes this, my suspicion is that Alice Wang was not a spy who happened to become a businesswoman. She was a commercial fixer, possibly a white glove, whose value lay in connecting powerful interests while keeping their involvement at arm's length. If that is correct, the real story is not how a Chinese woman compromised an American technology billionaire. It is how Chinese commercial interests may have used offshore companies, financing arrangements, and personal relationships to gain access to billions of dollars worth of advanced American technology. Instead of obsessing over an unproven spy narrative, these are the questions investigators should be asking. Who ultimately controlled the procurement network? Who supplied
28:32the money? Who received the computing capacity? And how much did NVIDIA know about the ultimate customers? The evidence points towards a potentially far more revealing story. Billions of dollars worth of strategically sensitive American technology passing through intermediaries whose real principles and economic interests remain obscured. People should stop chasing the sensational spy narrative and start following the money. That is where the answers are likely to be found. So this is from some, by the way, this is not someone who's like a kind of a real pro CCP tanky kind of guy. He's like a Hong Kong businessman who writes actually for lots of Westerners. And I've seen quite a lot of kind of China hawks actually share that piece and say it was very interesting.
29:14This is very plausible. Of course, none of this rules out that also she did have a direct connection to Chinese intelligence. That's possible. If that was in those US intelligence memos, you know, maybe that's the case. But also, you know, US intelligence has their own interests, right? So if they think that Jensen Huang, it's quite possible, I suppose, that the American security establishment is very uncomfortable about the fact that Trump is very relaxed about sending chips to China, right? It might be that they're very uncomfortable about the relationship that Jensen Huang has to Trump and the blind eye that the Trump administration is turning to the export of these chips. And so they think, how can we bring all of, you know,
29:53how can we change the narrative? And they think, well, there are questions about this woman already. So maybe we could sex up the dossier. Maybe we could say that actually, not only is she, you know, I suppose, doing some secret underhand business dealings, but she's also a Chinese spy. All possible. On the theme of following the money, as this businessman suggests we do, Desmond Sherman, he's saying we should do in this situation. The other big mystery in this story has been why the Americans took no action against this company, against Megaspeed. So Megaspeed and Alice Wang, according to that New York Times article last year, they appeared to be helping Chinese companies evade export controls. And this is supposed to
30:40be key to American national security. They said this is a really important technology, so we're not going to send it to China. But there is this company that's helping China smuggle chips to their own company in Singapore, who's helping Chinese companies evade it. And the US is like, oh, blind eye, who cares really? No, these things happen. So why was no action taken? Well, actually, no, this is just an example of no action being taken. So just today Axios has reported that Commerce Secretary Howard Lutnick was sent a memo in spring about Alison Jensen Wang and took no action. And the Financial Times has even reported that the Commerce Department has been delaying Megaspeed's inclusion on a commercial blacklist. That's right up to the
31:18present day. And the FT reports this on that particular story. So they say, several US government agencies pushed for Singapore-based Megaspeed to be included on the so-called entity list of groups connected to China that allegedly posed national security risks. That's according to multiple people familiar with the deliberations. But the Commerce Department, which chairs the panel overseeing the list, sought over the past two months to delay new inclusions, one of the people said. So there are people in US intelligence saying, and this again, this is almost separate from her being a spy. They're saying, look, Megaspeed is being used, this company is being used to evade export controls. Now, whether you're for or against export controls, this is in American law,
31:57right? So you're not supposed to do it. NVIDIA shouldn't be selling chips to a company which is then selling them on to Chinese firms. But that was happening. And the Commerce Secretary is being told, let's blacklist this company because this is what they're doing. And then the Trade Secretary, the Commerce Secretary, Howard Lutnick, is saying, well, I don't know, maybe we could wait a little bit. Maybe there's a bit more information we could get. I don't know. I don't think there's any rush. So why could that have been? Well, some have hypothesized it could have been due to the influence of Alice Wang on Jensen Wang, who then lobbied the US government. Right? So they say this was the genius of the Chinese operation. So they get Alice Wang to
32:36sleep with Jensen Wang. Alice Wang tells Jensen Wang, I know you're friends with Donald Trump. Can you please lobby on behalf of my company, Megaspeed? I would really appreciate that. And, you know, I'm sure we'll have a lovely evening when you feed back the results. Maybe that's what was happening. And maybe, I mean, I don't think that's completely implausible, by the way. However, we don't need quite such a complex explanation. Because the simple fact is this. Every chip sold by NVIDIA, wherever in the world it's sold, is good for NVIDIA's profits. And that obviously benefits Jensen Huang. So Jensen Huang doesn't need someone sleeping with him to encourage him to sell chips to Megaspeed. It's also good for the US stock market. So whatever's good for NVIDIA is good for the US
33:21stock market. They're a huge portion of the S&P at the moment, and the stock market obsesses Donald Trump. But there's another key player here who directly benefits when NVIDIA does well. Commerce Secretary Howard Lutnick. And this is crazy. This should be a huge part of this story. So Howard Lutnick's investment fund, Cantor Fitzgerald, not sure why it's called that, but it owns 1.19 billion pounds worth of shares in NVIDIA. So the guy at the top of the commerce department, who's been getting memos for months, right, saying, we think Megaspeed is helping NVIDIA smuggle chips to China. He's like, well, let's just wait a little bit. I mean, you know, NVIDIA, they're good guys, big company, good for America when NVIDIA do well.
34:11It's also good for his investment fund when NVIDIA do well. Now that investment fund, we should say, it's not formally in Howard Lutnick's hands anymore. When he became commerce secretary, he was very professional. He said, we couldn't possibly have conflicts of interest here. So I should distance myself from this important investment fund because it could persuade me, it could affect what I do. Can you guess who he handed the fund to? Was it his wife? It was his children. Oh, nice. So his children now have over a billion pounds in Nvidia shares. And this is the man who's been sort of twiddling his thumbs when he's told that American laws are being evaded. So there's lots to this. I mean, in a way, it's less...
34:58So I suppose lots of the ways that this has been reported on, and again, all of it is speculation, but lots of the ways it's been reported on is that there's kind of lots of trickery and there's people who have been betrayed so has jensen wang has he been trapped by this woman entrapped by this woman and then have the americans been misled and you know it's all this is all coming out. And it does feel like, looking at all the details, everyone has known everything all along, right? Donald Trump does not give a shit about really American national security. Like he'd talk the talk about China, but if the chips are getting sold to Beijing, he does not care at all. He wants the stock market to go up.
35:39So if Jensen Wang says to him, can we sell a bit more chips to China? I know you don't want to sell the chips to China because that's politically difficult, but don't ask too many questions about this mega speed company also it's donald trump so you can say by the way it's run by this beautiful woman but that's amazing it's run by a beautiful woman i make profit stock market goes up what's there to complain about and trump's like yeah i see the argument howard lutnick's like literally i have a billion pounds worth of shares in nvidia and it just seems like win win win win for everyone unless you are a someone who thinks that america's selling chips to china is bad or if you're someone who just thinks the rule of law
36:18is important and corruption is bad but also like a legitimate position to have the nvidia trade is goes well beyond just howardnick or Donald Trump. The Nvidia trade is part of a whole new era of post-pandemic American growth story. As you know, if you subtract the tech stocks that have done fantastically well over the last few years from the American stock market, America has basically been in effectively a recession for the last few years. They need these companies to succeed. They certainly need the stock market to keep on going up before the midterm elections. And now the midterm elections have passed. Okay, well, they didn't vote for us anyway. Then we can dump the stock. Then we can say that we're getting out of this particular trade.
36:59I think that that kind of level of intrigue is maybe the right level of intrigue to understand this on. It's about the movement of forces. There are, of course, individuals in the Trump administration who are profoundly corrupt, who are benefiting in a very duplicitous and probably illegal way from the trades that they're making on the inside, just like there are Democrats who do the same. But it's much more substantial on the Trump side. But there is this, you know, NVIDIA is a much more profound thing than that. It's about the whole composition of American middle-class society and how it understands itself to be wealthy over the last sort of like few years. I will say like on this story in general,
37:38like the one thing that really fascinates me about the AI race is that it's such a volatile and fast-moving technological area. And yet, it doesn't seem like it's ever quite intrigue-filled or complicated or suspicious enough. Like, we've just had the scandal of the century. And now, two days later, we're sort of thirsty for more scandal, for a more intricate kind of explanation for some really quite, in some ways, straightforward facts. What think is is really amazing about this is that you can come up about the jensen twang alice wang story in general is that you can come up with the best most sophisticated account of why ai is a good trade or why nvidia is a good trade or why nvidia is a good trade, or why NVIDIA is a good trade, or why NVIDIA is a bad trade.
38:25You can come up with the story I just came up with about the US middle class and its investment in the stock market, its financialization post-90s, right? Very sophisticated stories, and yet the whole of this story, the whole of this kind of epochal moment in American capitalism, or in the global capitalism, seems to come down to the question of one guy having sex with one woman, right? What I find astonishing about this story is that literally tens of trillions of dollars of value, what people are calling the fourth industrial revolution, hangs in the balance over the question of whether or not one man had sex with one woman. These are the kind of contingencies that are determining history in a very very strange kind of way you know yeah i mean
39:09it's i mean there's something entertaining about it for sure i think you're potentially understating the the significance of the potential corruption in the trump administration because it's not like it's enormous what i is, in terms of why this was allowed to continue to happen, and why it potentially is still continuing to happen. Because it's not the case that American business is united in wanting to sell chips to China. They're a huge part... You know, the important trade is not just for NVIDIA, it's for all the AI companies and all the tech companies. They own NVIDIA. Like, all the companies own NVIDIA. It's all's all circular trading yeah but what i mean is that so so obviously anthropic and open ai two very important tech firms not public yet but sort of important
39:51they explicitly want there to be fewer chips sold to nvidia because there's a divide in the tech sector in the united states between people who want um ai to be that they want their models to be the only option. So they want to have a sort of monopolistic relationship to this technology. So open AI and anthropic. And that means they really see the Chinese as competitors and they want to hobble the Chinese companies. So this in a way is a betrayal to them, right? This is bad for their profit margins. So I think this is one where actually the elites are quite divided on this. And the reason they've swung in the direction of being relaxed about the rules being broken isn't because it's in the sort of objective interest of american capital to sell these chips
40:34to china it's because this particular company was owned by some well owned by one guy who sort of was happy to sell anywhere and also was having this relationship and um the people who happen to be in the white house at the moment have shares in nvidia are friends with jensen huang i think if there was a democratic president and a democratic white house democratic congress at this point in time mega speed would not have been selling chips to china for the past three years i totally agree with you and and since we are kind of like, you know, separating out our American elites, let's also separate out the interests of our Chinese elites, right? So the Chinese state has had chip restrictions put on it for about four years now.
41:16They came in October 2022. So the very beginning of the kind of this fast period of the AI race. Under Biden, right? Because he was the big hawk on AI. Yeah. And which is why you're saying that if the Democrats race. Under Biden, right? Because he was the big hawk on AI. Yeah. Which is why you're saying that if the Democrats were in the power, then there wouldn't be this kind of sale happening. So, China's been under those chip restrictions for about four years. In those four years, China has had a whole slew of really quite dramatic innovations across biotech, across a whole collection of, you know, their cars, their electric cars are now better, solar panels will continue to come down. Like, you know, Chinese technology has been doing really well. It would not surprise me if the
41:54Chinese state is not particularly interested in importing NVIDIA's chips into China because they think they can compete. And part of that is because they think that, unlike, we'll get onto the meter report in a bit later, unlike the kind of the anthropic people who think the AI race is really about the next 18 months or two years. And that's kind of it. That's the end of history. Unlike them, the Chinese seem to believe, Chinese government rather, seems to believe that the AI race, not 10 years, they expect to catch up they expect their chips to be as good as nvidia's chips in the next 10 years are they right about that i don't know but like they don't want there to be nvidia chips in the country because they have baidu's
42:33internal chip system manufacturer they have huawei thing called high silicon who have this new thing called the the tau folding process it's like a three-dimensional chip i don't know how it works um you know they have domestic chip manufacturers then they expect those chip manufacturers to compete with nvidia so they don't want nvidia to be in there but the chinese ai companies do want nvidia you know the whole question so in both in both countries you've got a divide between the ai companies and the chip companies and the state and the state yeah so in in america the chip companies want well the state. And the state. Yeah. So in America, the chip companies want, well, NVIDIA wants to be able to sell anywhere.
43:09Yes. And the AI companies actually don't want them to be able to sell anywhere because they want to be the only ones with the NVIDIA chips. In China, the Chinese government wants, say, DeepSeek to use Huawei chips, but DeepSeek probably wants to use NVIDIA. In the short term. Yeah. But in the long term, maybe wants to have a robust Chinese infrastructure so they can have stability. You know, it's the same reason that lots of countries in the last few years since the Ukraine war have started to buy more soda panels. Because you really do want something that is available in your country. You don't have to rely on oil coming in from outside because that might be disrupted, right? And that's precisely the argument
43:42that Jensen Wang has been making publicly as to why there shouldn't be export controls on chips to China is because he thinks that that will encourage them to make their own chips, which compete with NVIDIA and he wants to dominate the world market. Because he says that means America also dominates as well as him having the most profitable company in the history of the world forever, which I suppose is what he wants. On the timing, so I suppose going back to Daniela Gabor's tweet, she's sort of saying that this is coming out now because the midterms just happened and um there's sort of an alliance between some democratic hawks and anthropic who want uh a sort of reset when it comes to export controls on china and sort of blowing this up right now sets the stage for a more hawkish
44:24chance a more hawkish chance, a more hawkish stance when it comes to China and for the Americans and for NVIDIA to stop selling chips basically to their competitor. And on that front, it does appear to be having the desired effect. So this is a tweet from Elizabeth Warren, who's I think the most senior senator on the finance committee. She says, NVIDIA's CEO just confirmed a relationship with a suspected chinese spy congress is back today we have legislation ready to go to stop advanced ai chips from fueling china's military let's get this done um it's quite a hawkish statement that one i'm going to come to you in a minute richard before that though i do want to show another clip of jensen huang because if alice Alice Huang was a spy trying to get information from
45:09Jensen Huang, she might have been disappointed. Because Jensen Huang publicly admits there are some serious gaps in his knowledge. Earlier this year, he spoke to Ezra Klein, who asked him if it mattered if AI de-skilled people. This is what he said. Try to get a kid to do long division right now. You know, the multiplication table is starting to be forgotten. Doing square roots, my goodness. I mean, it's just basic math is being forgotten. Does it matter? That's my question for you. Yeah, I don't think it does. But there must be some set of skills that matter. Oh, yeah, yeah, yeah. But maybe not those. We're going to discover new ones. Just maybe not those. There are a lot of skills that don't matter.
45:55You know, people don't, I mean, my first confession, I actually don't know my address. I don't really believe that to be true. It's completely true and and janine will tell you and laurie will tell you um uh one day i had to pump gas and it was a few years ago and um i they needed my my zip code and i panicked i didn't know my zip code i don't know my telephone number but i forget these things um i can live with it you can imagine alice wang going to her handlers in the chinese secret services and they're saying like what did you learn about the design of the chips you know what did you learn about uh donald trump's views on china policy and she's like he doesn't even fucking know where he lives.
46:47It is amazing, yeah. On the timing thing, I think there is a much simpler explanation, right? Which is that after the Democrats won, there was already this big, big, big slide across the tech world. I looked at the sort of stock map you can get where you can see how much each stock is up and down. It was fully read. This is before the news came out about Jensen Huang. And then it seems like, you know, the information was released somehow, or maybe NVIDIA had a legal challenge to the New York Times reporting, and they were actually, you know what, report it. These are the kind of things that you would do in a sort of crisis communications situation. It's called kitchen sinking. Shout out to my partner for telling me that term.
47:29Like, you know, it's basically you throw the kind of, while there's already something bad going on, you just add in some extra sort of detail, right? You add in some sort of extra bad thing, and you think maybe it'll kind of wash out in the kind of the overall sea of red. Didn't seem to work in this case, right? Maybe they thought it would be interpreted merely as an affair. You mean the admission from NVIDIA? Yeah. He was forced into that by reporting earlier on. Yes, but the admission, saying like, okay, actually it's true. That kind of move, that kind of timing of that move seems to be designed basically to mean that we kind of instinctively put together these two trades. We say, okay, well, everything's going down
48:08because the Democrats might stop data centers in Texas, right? That's making the stats come down anyway. Therefore, we might as well release this piece of information and that will make it drop as well, but it won't make it drop as far as it would otherwise do if this was a separate event, right? Just put everything together. I don't think the affair is going to save NVIDIA's stock price. Not at all, but it might be. If you're saying doing them at the same time is sort of like ripping off the plaster. Exactly, yeah. I think that's a simpler thing. On the question of Alice Huang's exact role, obviously we are massively speculating here, we have no idea, but I think there is a kind of a point
48:42that I would make about intelligence work. Not that I've been involved in intelligence work, but, you know, allegedly. But there is a kind of a point I would make about intelligence work. Not that I've been involved in intelligence work, but, you know, allegedly. But there is a kind of a... People do say Richard Haynes is a cop. Don't they say... Well, I prefer them to say I was a spy. I'm much more excited to be a spy than a cop. No, I think... Yeah, I think cop. Okay, well, you're heading in first. Sorry, go on. I interrupted your point. What were you going to say? You were going to make a serious point. At the very high levels of industrial power. I like cops, by the way. Oh, yeah, sorry, I forgot.
49:07Your Navarro is kind of a resident bootlicker. Yeah, yeah. At the commanding heights of the industry, there's a coordination between the state and commercial interests. These are not separate things. We don't have to divide out, you know, what is a spy and what's just like merely kind of private company's interest, right? At the level of trillions of dollars, at the level of statecraft, at the level of a new Cold War, if that's what we're in, these two things blend together. So it's not like we need to make it a hard and fast addiction. I'm sorry for calling you for a bootlicker. I've called you a cop. I've got very thick skin. I read the comments. I've got to have thick skin reading these comments.
49:50Jesus Christ. Nothing makes you stronger than this. Leading AI firm Amthropic of this week kick-started their initial public offering or IPO. This is the first time members of the public will be able to buy shares in the company and the initial demand is looking impressive. So on Monday Anthropic said they would target a valuation of 1.4 trillion dollars. That's a lot of money. But after strong demand from investors they've repriced the company at 1.5 trillion dollars. This wouldphropic, which was only founded in 2021, the 11th most valuable company on the planet. And Dario Amadei, in his pitch to investors, suggests it's not just those who purchase shares who will benefit from the company's success.
50:36We look at things like cancer, Alzheimer's, heart disease, mental illness. These will be relics of the past in, I think, less than a decade. And I believe that the company which gets the risk right will also be the company that has the license to operate, that society trusts and society believes can build the benefits in the right way. I think if we do this right, we can create an age of miracles where humanity is able to solve problems that have eluded us for millennia. We have to make sure that all the risks, all the ways that things can go wrong with this technology, we very carefully address those risks as we bring the benefits and as we have a viable business at the same time. And I invite you to join us on the next phase of this journey.
51:40That was very speculative. Nice if it came to pass, you know, cure cancer, cure Alzheimer's within 10 years. Great. As I say, speculative though, what's more concrete is the payout Dario Emede, along with the other six founders of Amthropic, are set to receive on Monday when the IPO actually takes place. Each is selling exactly $250 million worth of shares to investors, so that's what they'll get in cash. But that will only be a fraction of their overall wealth. The shares of the co-founders are expected to be worth $18 billion each. So they're going to be very, very rich people. So is this the beauty of capitalism working its magic? We benefit from cancer cures, Alzheimer's cures, while some innovative founders get rich.
52:22It's win-win. Well, not necessarily. And while investors seem excited about anthropic AI safety experts are having a full-blown freakout this week, Buck Schleggeris, fantastic name, is CEO of Redwood Research, which is an AI safety outfit. He tweeted this yesterday. I think there is about a one in three chance that within the next year, AI companies will build super intelligence. That is AI that is much smarter than humans in every way. If this happens, I think these AIs will probably take over. That is, they seize control of human institutions like governments or they destroy those institutions. I don't want AI takeover to happen for several reasons. But one of them is that there's a serious chance that as part of it, all or most humans will be killed,
53:06either immediately or within a few years. Also concerned was Daniel Cocatayo. He is the former OpenAI engineer who wrote the famous AI 2027 paper. So he says this, unfortunately, it seems like the intelligence explosion will be happening in 2027 after all. Most of this post analyzes the new META report, though other sources of evidence are also important. If they're right, that implies an AI-driven acceleration that is five months ahead of the AI 2027 scenario. I am scared, sad, and angry. So what's provoked these doomy prognoses? Well, as Coccataglio sort of implied in that tweet, the big update in the past week on the AI safety scene has been a new report from the AI company META, or AI safety
53:52company META. So it has a fairly dull title. It's called Independent Assessment of AI R&D Acceleration at Anthropic in October 2026. But the reason it's got AI safety nerds panicked is that it pertains to a fundamental issue in AI development that these people have been warning about for decades. So it's called recursive self-improvement. It's the idea that as AI gets more powerful, it will increasingly be used in the process of AI development. And then ultimately, the AI will become responsible for improving the AI. So AI improves AI, more powerful AI builds more powerful AI, more powerful AI builds even more powerful AI. So you sort of see there's this cycle which ends up in this sort of theory
54:34with an intelligence explosion. And that's when everything gets really weird and potentially very scary. It's been a theory up to now, but is it already happening? This is what META say. So they say, from August the 12th to October the 29th, 2026, META staff performed a high access evaluation at Anthropic to assess the degree to which AI is accelerating AI R&D. If AI systems started to significantly accelerate further AI development, this could lead to a sudden emergence of extreme and risky capabilities. Anthropic offered unprecedented access to internal data on their AI R&D process, including a variety of metrics on pre-training, reinforcement learning, training compute in each domain, benchmark scores for internal models, AI speed-up analyses, forecasts of a future model's
55:22metrics, and interviews with relevant staff. So they're sort of saying, has this sped up? Is recursive self-improvement happening? And they're looking at loads of data over a few days and doing loads and loads of interviews with staff members, basically to ask them how much do you use AI when you're coding future AI agents? Or how much responsibility I think is probably what they're most interested in. Not are you using AI, but are you handing over responsibility to AI for building future anthropic models? And this is their conclusion. So they say, as of October 2026, anthropic is likely improving the capabilities of their models on well-specified tasks at around twice the rate they did between March 2024 and October 2025.
56:02This estimate is given with high uncertainty, so it might be between 1.3% or 3 times the historical rate. A majority of the acceleration today is likely attributable to Anthropic's extensive use of AI to do R&D, though there is significant uncertainty and other causes likely played a large role in acceleration in late 2025. The lack of observed bottlenecks or strong diminishing returns mean there is a significant risk that further capabilities progress could lead to even faster acceleration, though they say they might be missing some bottlenecks. So this is the chart they showed, or they shared, sorry, showing what they think are accelerating capabilities. It's a slightly confusing chart because on the y-axis is a metric that they've basically put together themselves. So it combines
56:43a bunch of metrics, sort of coding capability, maps ability, various things that the AI can do. How capable is it at various things? And then they've sort of made that into a number. And then they show that between April 2024 and October 2024, it was increasing by 15 points a year. Again, it's a bit arbitrary because this is a metric they've made up. But then it began increasing at 29 points a year after that so it's doubled in speed um and as i say from interviews with workers at anthropic um they assess that this acceleration was largely because engineers were handing over large parts of their tasks to ai um richard i know you've read the meter report you're as you're as ai pilled as i am um as i say i never really like graphs where on one of the
57:25axes is a sort of like metric which is made up by the people who've drawn the graph but anyway um i think there are reasons to sort of take seriously the fact that ai development may have sped up and that it may be because um ai is being used to train ai and that that does look a little bit like the stories about recursive self-improvement, which people have been warning about for over a decade. How freaked out should we be about this report? Deeply concerned. Yeah, deeply concerned. I think there's a, going back to what I was saying about the, in the previous story about the NVIDIA, about there being this sort of need to produce in the AI story a more sophisticated or more sort of intrigue-filled
58:06explanation for the world. This has definitely happened on the left with regards to AI. We want there to be something more sophisticated and complicated at stake. We want it to be PR. We want it to be, you know, we want it to be a psyop of some description. It is, to some extent, right, like suspicious that this is coming out a week before the Anthropocene IPO. But, you know, the AI race is moving very fast. There is always some significant event happening quite soon, or has just happened, right? Like, you know, it's very volatile. It's a very fast-paced environment. And so you could always say, well, isn't this suspicious timing for this thing, right? At root here, what is happening is that AI is being,
58:44yeah, is exactly as you say, right? We are into the era of recursive self-improvement. The technology is getting really, really, really good. If you don't believe that, you should use it, right? I think it's very difficult to go through some very basic steps, right? Get a model, get a subscription to Claude, use the most sophisticated model, try and find a thing that you can do on your computer that your model is not better at than you, think about what economic implications that has for you, right? And the reason why I'm sort of insisting on this is not because I'm some sort of, because I have investments in Nvidia, or I have investments in Anthropic, or I'm going to get anything on the IPO, right? I'm not. The reason I'm insisting on it is because
59:23this is going to be i think a major aspect of class politics for the next 10 years and that's what i care about right well also i mean but what do you think about it do you think are you worried about killing us next year because that's what a bunch of these people are worried about so there's no class politics in a well there is there is planet no no class politics beautiful planet. No class politics on a dead planet. Beautiful. No, there is class politics on a dead planet. On a dead, not on a dead, dead planet. Sure. Before we, you know, in the final stage of a dying planet. There's class politics on a dying planet but not a dead one. Yes. Anyway, let's talk about the dying part, if that's what you want to go on. So I think the thing
59:59is that, like, there's a whole range of risks that could come from AI. There are risks that could come from data centers, taking electricity, being noisy, being bad for people to look at, destroying jobs. There are lots of things that happen in data centers. There are risks from AI models being biased and hallucinating. There are risks from AI taking your job because it's actually really, really, really, really, really good at doing things. And that's the one that I am particularly worried about. I am worried about a systematic disempowerment of the working class because capital can simply pay other bits of capital to do the task that labor was previously doing. That's the thing I'm concerned about.
1:00:44And then, yes, there are absolutely existential risks risks and there is something in me that wants to say you know as ilia zyrkowski put it shut up multiply right eight billion people dead implies 100 billion people dead because all the people after that also don't never come to be and there should be care we should care about that more But that's not the politics that we can actually activate in the world, right? Effective altruism, whatever its virtues, whatever its drawbacks, whatever its vices, is never going to be a mass politics in the time span that we need to activate it. What is a mass politics is the fact that this thing may well replace you at your job, that it may well prevent you from being useful to capital,
1:01:26which, at the end of the day, is the reason why capital keeps you alive. You know, that's the kind of like, the level of which I think we should think about this as a threat. The more extreme ends of the risk distribution, the existential risk stuff, I take somewhat seriously. But it's not the focus, I think, because we can't generate a mass politics around it. Garrison Lovely is right about this, right? He's, you know... I think you can make a mass policy. I think the easiest mass politics to make about AI is you say that people running these companies said it might kill us all. Let's pause. But people don't believe it. More and more people are believing it. But... And I think people are...
1:02:03It's in a way more believable than they're going to take all of our jobs. Because, and I'm like, what these guys who are quite good at predictions, you know, these meter people, they have made very, Daniel Kockatalia has made great calls about AI advancement. Like in 2021, he was writing that sort of, in precisely four years time, there are going to be millions of people signed up to chat GPT. Can I say like, just because there are people often on the left you get this idea oh well this is just hype yeah the only people who have been right are the people who have been accused of being hype in fact the technology is progressing faster than almost anyone predicted so hype is in fact the opposite way around like it's not we don't we're not suffering from an excess of hype
1:02:42in fact we're suffering in a sense from a from a lack of that's book sorry go ahead what i mean is that that's such i think it is actually going to be quite a while before it takes all of our jobs but if these guys are right it could cause a lot of chaos and damage before that because what they've been quite wrong about is sort of how quick this is going to disperse with the economy because they think everyone's like themselves and And everyone who works in Silicon Valley, they're like, they're all using the latest models. They're constantly waking up in the morning thinking, oh, what can I replace with AI now? Because that's what they love. And so they think that everyone else is thinking the same,
1:03:14which is not the case. But what they have been quite good at is, yeah, guessing or estimating, predicting, projecting, whatever, how good the actual technology is going to be. And these guys think that when the technology gets to a certain level, we're not going to be able to control it. And so I think it's plausible that we lose control of AI before it takes many of our jobs. I think that's also possible. Yes. So there are two things that I think people should, in general, keep in mind. I try to keep in mind for myself. One is that model progress is faster than you think it's going to be. Because the exponential is always in some ways surprising, especially when you have an exponential, like in this meter report,
1:03:49is actually itself speeding up. You've got super exponential growth. The other thing to say is that the economics of it is always more confounding and slower than you think it's going to be, right? So on the one hand, I think you're absolutely right about the job part. I know that I still am very worried about systematic disempowerment. I think that's a very real prospect. One thing that people should take away from this and connect to the Anthropic IPO as an event is that if you look at the rate of model improvement, it is going up exponentially, right? The models are increasingly getting better. But if you look at the rate of model improvement per dollar invested, it's actually slowing down.
1:04:37We're getting diminishing marginal returns, which means that because Anthropic has received vastly more than exponential increases in amount of money that people are throwing at it, they're only getting from that vastly more exponential this falling off rate of intelligence increase. It's going up, it's going up faster, but it's not going up as fast as the money is going up. Why does this matter? Because Anthropic, to keep that tendency going up, need a hell of a lot more money. Where do you get a hell of a lot more money from? From an IPO. And I would say, you're saying 1.5 is an impressive sum of money. A couple of months back, it was $2 trillion. That's actually a 25% cut off the price of anthropics.
1:05:10It's cheap, I would say. It's cheap. Buy anthropics. This is not financial advice. In terms of, what's the one... So there'll be lots of people watching this thinking, it's bullshit, it can't count the number of E's in 17, which it can now, by the way. What would you recommend someone does to sort of convince them that AI is a it can now, by the way. What would you recommend someone does if they, to sort of convince them that AI is a big deal, like personal use? So the objective result you should look at is that Astra, or sorry, not Astra, but the model, the internal model that OpenAI have, that is more advanced than Astra, managed to prove the quasi-Riemann hypothesis. So the Riemann hypothesis is very possibly like the largest outstanding problem in all of maths.
1:05:52And it didn't prove that problem centrally, but it managed to constrain the range of possible solutions to it in a way that no human has ever managed to do in the hundred years that has been an outstanding problem. You know, and I did it in about a few hours of compute. So the example I was going to use, which I don't know if I should say, because it might put me out of a job, but you showed me this. So on, was it Google Gemini? If there was a report that I was thinking of sort of covering on the show, and it wasn't a show day, by the way. It wasn't me sort of directly doing the script, uh it was a report into like something from the resolution foundation about pip i think actually and i was like well this is quite long i want to clean my flat so
1:06:35you download the pdf you upload it to google germany and it makes a half hour podcast with these two characters who sound completely human. And they literally just have a conversation about the report where they step by step tell you everything that's in it. And it's like, that is, it's not the same as listening to like, you don't have a parasocial relationship with these people. You don't trust them. You know, like, because lots of like media is about you trusting the person who is speaking to you and you being interested in what that particular person thinks. But the mere fact of the technology that you can put in any report any book and it makes you this like podcast about it not just reading it but sort of like where there's this realistic conversation between
1:07:14two people and to be fair that the podcast that they came out with would have taken me at least a day to sort of script like it's impressive what is so like but deeply peculiar about listening to those podcasts could have done the same thing is that now they've started to add in little false starts. So instead of saying in a kind of robotic voice everything that people are saying, they will do what I just did, right? Which is put on kind of into the middle of my sentence to make it sound more humanistic. They will do, like I just did, you know, you've got kind of the robot making your thing, but the robot is pretending to have these artisanal, you know, little turns of phrase that humanizes them to a certain extent.
1:07:52I can imagine that becoming much more personalized and people developing genuinely parasocial relationships with them. I don't think Google wants you to do that yet, but there will be a podcast service attached to one of the AI girlfriend or boyfriend yeah where you can have this kind of yeah stick to stick to reports about social policy not yes someone to tell you that you're beautiful and worthwhile and have meaning and stuff because that's i would never have anyone tell me that uh i don't know why anyone would there's something about this dynamic what is the i don't know what that's what is it let's well let's investigate that after the show anyway i think you're you've got great worth and you're
1:08:32wonderful and i've really enjoyed the last day i also i also think the same about you um thank you everyone for tuning in come back tomorrow for another live stream from 6 p.m but now you've been watching navarro Media. Good night. This broadcast is brought to you by Navarro Media. Go to navarromedia.com slash support.

