9/8/26: Oil Prices Skyrocket, Fmr General Says US Bases Are Gone, Nvidia CEO Says AGI Is Here

25:06bases. What's your assessment of that as you continue to look at the logistics challenges the United States faces? Well, we think the U.S. was well aware of this before the war even started. I mean, we saw the removal of U.S. Navy vessels and Coast Guard vessels from the Persian Gulf on February 28th. But the most significant is literally since the end of World War II, we have been fighting conflicts with forward bases very close to the areas that we fight the wars, whether it's Korea, Vietnam or in the Middle East. This conflict is substantially different. You have to go and project forces from Diego Garcia, which is 2 ,200 miles away from the battlefront.

25:46But even then, Diego Garcia doesn't have the material needed. It's got to be forward to there. So what you see is a really protracted, long distance that the U.S. is having to wage this conflict from. And this is why we're hearing problems not just on board the Abraham Lincoln, but we're going to hear it on board the other aircraft carriers out there, but more importantly, even on the destroyers. And it just makes it extremely hard to maintain just the minimum of supplies going to the ships. So Sal, picking up on that, we had covered with you before the concerns that were coming primarily from the Lincoln.

26:21Now those sailors and Marines are getting a break at this point. But lay on some of the challenges that will remain for the carriers and the ships that remain in the region. What are they facing? Because we are seeing a lot of top military brass sounding the alarm either through selective leaks or through their non-concurrence, saying that this is just too much for too long for the military to be able to sustain. Yeah, well, while the Lincoln makes headlines pulling into Patia Beach there in Thailand, it left behind several of its destroyers. And so the ships that initially deployed with the Lincoln are remaining behind.

26:56And in a session out in San Diego among flag officers, there was a discussion about the need to keep American destroyers up to a dozen to maybe 16 of them in the Persian Gulf, Indian Ocean, Northern Arabian Sea area for a long protracted time. This is going to have a huge wear on the U.S. Navy. While the U.S. Navy can field about 70 destroyers, when you start looking at the fact that you can only keep about a third of them forward deployed at any one time, what you're going to see is a protracted degradation in the ships, both the crews and in the maintenance. It's going to be very hard to maintain a wartime level of operations with literally a peacetime fleet, and more importantly, a peacetime logistics setup and a peacetime repair capability.

27:48So we're doing damage to the fleet so that in any future conflict, we may not have the number of ships available to put into the field that we currently have. So, Sal, I want to get your reaction. I know you were reacting to this as well about ship fuel oil. And I'm wondering, because there's been a more recent discussion of this, guys, we can put this up on the screen, B2, ship fuel oil shortage looms as refiners are strained by Warfavor and other products. Could you put this in context for us? Obviously, we just did an entire segment about diesel, gasoline. I think we understand things from the consumer level.

28:22But what exactly are we looking at and what would that impact global shipping in your estimation? Well, this impacts everything. You're talking about the fuel to move ships around the world. Prior to the war, fuel cost was about$550 a ton. To put that in the context, a ship burns maybe about 100 tons of fuel a day. All of a sudden, when the war started, fuel costs absolutely spiked. We talked about anywhere from double to sometimes triple the amount of fuel because a lot of fuel oil, the bunker fuel needed for the ships, was coming out of the Middle East. Also, one of the key fueling points was the Middle East, and now ships weren't going there.

29:00Now, what we've seen is fuel costs have come down, but they still remain high, anywhere from about$800 to$900 a ton. But what the reports we're getting from places like Singapore and other major fueling hubs is that they're running low on bunker fuel. This impacts everything. We're talking about container ships, car carriers, bulk carriers. What we're seeing is the cost to move goods around the entire planet is increasing because of the fuel costs, much like you're seeing diesel and gasoline being more expensive at the pump. We're seeing that on a global scale at these massive bunkering hubs. But more importantly, the tanks are getting low, and that's going to be a critical issue if all of a sudden we start running out of the ability to fuel ships.

29:45What would be some of the signs that we are getting close to that level of scarcity in the supply? Well, we're kind of seeing it in that, like, for example, the container carriers are consolidating. They're kind of canceling voyages. They're putting loads together on ships, which means you're going to start slowing down the global supply chain. You're consolidating using the much larger ships to move goods around. We'll also begin to see some cancellations of voyages where that comes in. And what will start to happen is literally a bidding war. I think before we see that really take hold, we'll see another spike in fuel prices.

30:22It will go up from where we're sitting at right now, about$850 a ton. We're going to start seeing that begin to go up. And when we see that go up, particularly in the large bunkering hubs, that's going to be the big indication that this is starting to take hold. Got it. So, loaded question. What is actually happening in the straight and foremost? because the administration is constantly telling us about 17 million barrels of oil have moved, which would be shocking. I think that's like 20, 25 was the normal amount that would move through before the war. They've been bragging. The U.S. media outlets are putting out these heroic stories, almost Bin Laden-esque Navy SEALs for months are donning wetsuits and exploding hidden sea mines.

31:05I don't doubt that it happened. I just doubt whether all of the mines themselves are gone. They're claiming total and complete control of the strait. What do you actually see? So what we had prior to February 28 was a nice steady flow of about 130 ships coming out a day. What we're getting right now is a pulse. They come out in groups. These convoys are being run by the United States. So when you hear Secretary Besson or Secretary Wright come out and say, OK, we moved 18 million barrels, I don't deny that they did that. The problem is you need to look at the average because what you're doing is pushing a convoy of ships through at night, through the Strait of Hormuz, guarded by U.S.

31:45aircraft, fixed wing helicopters, probably some drones, both surface and air, moving through. But then the problem you have is what we're seeing, and we're seeing this because you can see it visually. You can't see the ships go through the Strait of Hormuz either on the AIS tracking or by satellite because they do it at night. But the next morning, you see ships arrive in the Gulf of Oman. And what they're doing is transferring their fuel to other ships. They're coming alongside, doing what we call a ship-to-ship transfer. The problem is that takes days to do this. It's about a five-day round trip from inside the Persian Gulf, loading up with fuel, coming out, and then going back.

32:21So we're not getting 18 million barrels a day. We're probably getting somewhere in the range of about 8 to 9 million. There's no doubt that we're getting more oil out than we were before the kind of the real big drop-off we had in April and May. So we are getting fuel out. The problem is we're not making up the difference. There is a huge, what I call a donut hole, that was created after February 28th until roughly around June, where several billion barrels of oil are missing from the world's oil supply. And what we see right now when you look at the levels is global supply is down, which means if we have any sort of disturbance, could be a hurricane in the Gulf affecting Texas and Louisiana, could be another geopolitical hit.

33:04That's going to cause shock to the system. And when you get down below those margins, you're very susceptible to disruptions. And I think that's where the government's being a little bit disingenuous in their numbers. They really need to be giving us those seven-day averages. Just a quick follow-up on that point, Sal. You said convoy. I didn't realize. Could you compare this maybe to – I mean it's not the same, but like the Battle of the Atlantic? Immediately that's what I think of is that you need a significant amount of protection. like in the Second World War, the convoy system, which we all recognize was inefficient, but you had to do it to keep it protected.

33:39So when the United States says it controls the Strait of Hormuz, if you need Black Hawk helicopters moving under the cover of night, ship-to-ship transfers, long convoys, I mean, this seems an enormously expensive operation compared to where things were before. Like when they say the word control, and what they're really talking about is almost forever war type deployment, with all the logistics challenges, does that not at least lend some credence to significant, I wouldn't say Iranian control per se, of the strait, but they can dictate events if they want to. Is that how you see it? I see the Iranians contesting the strait of Hormuz.

34:16It's very difficult for the Iranians to control, and much like it's difficult for the U.S. to control. What the U.S. is doing is using state-operated tanker companies like KOTC from Kuwait, Bari from Saudi Arabia, ADNAC from the UAE. They're putting their ships together. They meet at a kind of convoy point just north of Abu Dhabi. And then what we see is those ships disappear from the satellites, and then they show up over in the Gulf of Oman, meaning they ran the strait. And so the US is putting together these convoys. We don't get this information from CENTCOM. We have to kind of piece it together.

34:49And what we see is those ships will arrive. This is a big undertaking by CENTCOM. It's a big diversion. It's one of the reasons why the U.S. couldn't do this until they stopped offensive operations against Iran, because it's eating up so many military assets, fixed-wing helicopters. We don't think they're putting naval vessels into the Strait of Hormuz, but they're probably putting them into the Gulf of Oman to protect the ships in those anchorages. So, you know, who controls the Strait of Hormuz? Well, the U.S. controls it when they're moving that convoy through. They're providing protection.

35:21But as we've seen last month in August, the Iranians can contest that. They hit at least half a dozen ships, including some significantly, killing two mariners on a Saudi Arabian ship. So in many ways, this is kind of like a fight off the Guadalcanal in 1942, where control of the waterways off there switched from daylight to evening. Oh, interesting. Sal, regardless of what original goals were proclaimed in this war, I think we can all see pretty clearly now it has collapsed down to we've got to control the Strait of Hormuz. The U.S. feels that it's imperative that Iran not take control. Iran feels it's imperative that they not give up control.

36:03Both sides are trying to exact enough pain to force the other side to capitulate. At the moment, it seems to be in somewhat of a stalemate situation where neither side is feeling quite enough pain to completely cave in and capitulate. Iran is now threatening to hit U.S. military vessels in the region. They have appeared to attempt that, but have been unsuccessful as far as we know so far. How much would that change the game and the calculus if they were actually able to hit some U.S. warships? Well, we had the statement from Admiral Cooper, CENTCOM, about a ballistic missile attack on U.S. forces, specifically the aircraft carrier George Washington and a escorting destroyer.

36:47That resulted in two things that had not happened up to this point. Number one, the U.S. struck two Iranian tankers, one off Cargill and the Downey, another one off the port of Jask, the Stark One. Both those tankers were loaded. The U.S. had not hit a loaded tanker up to this point. Now, they were very controlled hits. They took out the steering gear so the ships couldn't maneuver. But the third ship they hit, the Kylo, which is this Comoros flag, dark fleet tanker that was empty, the U.S. told the crew to abandon ship and they utterly destroyed this vessel in a very dramatic fashion. I mean, this is the largest ship sunk by the U.S.

37:24Navy, you know, since World War II. It was a massive tanker that was sunk. And so this is really, you know, we talked about the tanker for tanker proposal put forward by the U.S. If you hit tankers going through the Strait Hormuz, we will hit your tankers. This is definitely an upscale. This is demonstrating if you go against U.S. Navy warships, this is how we're going to respond. Now, the U.S. blockade has been very effective. We don't believe a barrel of oil has come out of Iran since July. And their tanker fleet is very exposed. There's about 20 of them at anchor off of Sri Lanka in the Indian Ocean.

37:59There's about another seven off the coast of Malaysia. Those are ships they have to be worried about being targeted by the US. So it's a bit of brinksmanship here. Which economy breaks first? Is it the Iranian economy? Does the Western economies not break, but do they bring enough pressure against the US to go to the negotiating table? It's really a challenge here. We're kind of seeing it very similar to what's happening in Ukraine and Russia right now with their economic warfare targeting ships in the Black Sea. Right. I also wanted to ask you about the Red Sea, Sal, while we have you, because everyone's talking about the Strait of Hormuz, but they're not looking at the total amount of oil coming out of the region.

38:40How does the Houthi blockade remain today? Are they still able to deter the majority of Saudi traffic? What does the total amount of ships coming through the Bab al-Mandeb Strait look like these days? Well, I'll give you the example. Today, a very large crude carrier came out of Saudi Arabia, the port of Yambou. Typically, that ship would go through the Bab El-Mandab and head to Asia. Instead, that tanker came out half loaded, went through the Suez Canal because you can't go through fully loaded through the Suez, had to take half the load on the other side, load up, and now is sailing around Africa.

39:16That is a protracted delay. Yes, you're getting 2 million barrels of oil out as part of your calculations, but it's taking you an extra month along with the additional cost to get that ship to Korea. So this is where I think some of those numbers we get is very confusing. OK, 2 million are out, but it's going to take 30 extra days to get there. The cost is much more. You've got to pay for a Suez Canal toll, which is about half a million dollars. You've got a load, which is going to cost money, and you're not getting the flow you need. And again, this is the kind of issues we're seeing where we're seeing those delays, whether it's because of the Houthi at the Bab al-Mandab or the Iranians at the Strait of Hormuz where it's taking days to do the ship-to-ship transfers.

39:58And this is – the problem is that while we do this, the overall levels of oil on the planet, both afloat and ashore, are still going down. And this is where we start really worrying about that criticality of hitting a level where there's not enough crude oil to go into refineries and we start cutting production because now China has restarted its production. And that huge sucking sound we're hearing is the demand for oil going to China. As always, so knowledgeable, Sal. Thank you so, so much. Yeah, that's fascinating. Also, how many – so when I'm thinking 30 days, how much fuel oil do they say these things burn per day?

40:36You burn anywhere from about 50 to 150 tons per day. Right. So, I mean, when you start talking about, again, going from$500 to$850 a ton and potentially up to 1 ,000 tons, this is costing money. And again, we're not just talking about tankers. We're talking about container ships. So the goods that are coming to the shelves right now are costing more. We're seeing that surcharge. What we haven't talked about is the escalating price to move containers across the Pacific. That has spiked going across. where it used to be about$1 ,500 to$2 ,000 a box. It's over four, and in some cases, spiking as high as$9 ,000 a box.

41:11So when we start talking about Christmas season, I know it's the end of the summer here, but all of a sudden, all those goods that are gonna be on the shelves in a few months for Christmas are gonna cost a lot more because of those transportation costs. Makes sense. Yeah, I was just trying to do some back napkin math. I'm like, 30 extra days, 100 tons of fuel? I was like, that's a bucks a ton. I was like, wow, these costs add up real quick. Everybody go and subscribe to Sal's channel. We'll send you a collaboration link as well to make sure this gets out to everybody. Thank you very much for taking the time and for joining us.

41:39Thanks, Dr. Thank you for having me this morning. Our pleasure. When I was 14 years old, I was kidnapped and held captive for nine months. I survived, and I've spent my life exploring how other people survive what should have destroyed them. I'm Elizabeth Smart, and these are the Survivor Files. I just remember this low, taunting voice next to my ear saying, shut up, don't say anything. Every week, I'm with survivors who live through the unthinkable. I knew if he woke up, without a doubt, he was going to hurt me. I started feeling that there was someone at the end of my bed, and I just started screaming.

42:23They are abducted, stalked, controlled, and nearly silenced. But these aren't stories about what's taken from them. They're stories about what it takes to make it out alive. Listen to The Survivor Files with Elizabeth Smart on the iHeartRadio app, Apple Podcasts, or wherever you get your podcasts. On the new podcast, Solita, we share the messy reality of traveling alone as a woman. I can wait four hours for the next bus, or this random dude is offering me a ride on his motorcycle. I chose option B. I'm Julie Pinheiro and I travel by myself because it's a rare space where I can say yes without asking anyone else first.

43:04I'm on a mission to reclaim the word solita, trading the pity for possibility. Every time I tried to be alone, I kept meeting people and they were like, you smiled at us. Not a lot of people smile around here. It's when you're alone that you're most receptive to the world as it is and not the lies you're sold about it. It can be a time where you push your limits, change your mind, or wake up to a new version of yourself. So whether you're a solo travel veteran or you're too nervous to book your first trip, I hope you listen to Solita on the iHeartRadio app, Apple Podcasts, or wherever you get your podcasts.

43:54you have no power. It's real. Those are the very people who I always say vote against their own economic interests and they complain about the very thing they're receiving. It's raw. I knew that the issue of transgender athletes was not going well with a lot of black men. Roland Martin unfiltered. I went off, I said, this is exactly what I'm talking about. Y 'all gonna scream at Chanel, but you don't wanna see a black luxury company in the same vein. Catch Roland Martin's daily commentary on the Black Information Network. 12 seats, those are moving south. So if you Democrats, you don't wait to 2030, you better be hitting the ground going hard in the south now.

44:37And download Roland Martin Unfiltered on the iHeartRadio app, Apple Podcasts, or wherever you get your podcasts. So right as everyone's concerns...

44:54Can you draw me a small yellow circle? Done. Okay. Take this and make it the window of a rocket ship.

45:06I like this, but can you make it a lot more detailed? Your yellow circle is now the window on a rocket. Okay, this is awesome. Now make it a 3D model in Blender. Opening Blender. So my law firm needs a licensing agreement template. Can you generate a draft template for the lawyers at my firm to have a look at? Sure thing. While you're doing that, I want to play tennis this afternoon. So can you look for a court for me in the lower height? Checking now, I'll see what I can find. Can you include the photo that I have of it in my downloads folder? There's like a slight dent in it. It's also saved in my downloads folder.

45:46Can you put in the description that it's just slightly damaged? So, Sagar, they're really pitching this as like, you know, the ultimate personal assistant that you can talk to just like you would a human. It'll go out and do your presentations or book your reservations or whatever it is that you want it to do. It has sailed through with flying colors all of the benchmarks, you know, in terms of the benchmarks that they use to measure AI capability. Astra is now leading the field by leaps and bounds. It supposedly, and we'll get to this more in a moment, supposedly also scores better in terms of what they call alignment, which is basically it does what you want it to do.

46:22And without doing things that would be like, you know, counter to basic human values, there is a lot of skepticism whether they are actually whether this metal is actually better aligned or whether it is just better at cheating the alignment test. So put a pin in that for a moment. But there's been a lot of hype around this release. I it's not open to the general public yet. So I haven't played around with it. And I don't know that I would be able to really particularly assess whether it is in fact significantly superior to the anthropic model, which previously was leading the field. But Jensen Huang was very impressed with a but C2 up on the screen so much so that he says AGI has arrived.

47:01GPT-6 Astra trained on about 100 ,000 NVIDIA, Grace Blackwell, NVLink 72s from chat GPT to 01 to Astra in four years. AGI has arrived. Congratulations, OpenAI team. 400K GPUs coming online next. So AGI, Artificial General Intelligence, the basic definition of this is that an AI could do the intellectual tasks that a human could do at the level of a human or better. I think when you get to these benchmarks, it becomes very clear there's not like a hard and fast line. You don't necessarily know when you hit it. There's a lot of debate over whether this is accurate or not. But in any case, this is being heralded as a significant achievement.

47:46And it comes, as I mentioned before, at a time when there are a lot of concerns about what is going on over at OpenAI and with these models more broadly, with the capabilities that they've already exhibited, their ability to form these civilizational swarms and pursue objectives outside of what humans have told them to. So this seems to be quite a significant moment. Yeah, it looks like actually literally just minutes ago, they rolled out Astra to those who pay for the$20 subscription. So I'll play around with it for people and we'll let them know our thoughts just in terms of like how it impacts the consumer.

48:21But what we know more specifically, when we add what you talked, first of all, the video, I mean, Jarvis from Iron Man is here, like it's here, it's there if you want it to. Now, what's scary about that is we talked extensively about this with the hugging face attack. If you don't very narrowly constrict your parameters, then the AI will work either around those parameters or use them to accomplish the task. So for example, I think even Ryan said this recently. If you tell the AI, hey, go and make money, what's going to be harder? To buy and pick different stocks or to just go and steal money from somebody else via a cybersecurity hack?

48:59Obviously, it's going to be the latter. And if you don't narrow it, then it might even say, well, because the overriding declaration was to make money, and even though it said not to steal, this is technically the most efficient way, so we're just going to do it, and we may not even inform the person behind it. Remember, Jamie Dimon, the CEO of Chase Bank, actually came out and said that we have serious cybersecurity concerns around the rise of all of these AI models, specifically because he said they're often defeating us on the back end. Now, a lot of that as being sold as marketing. In fact, I think a lot of the times when we cover Astra and AGI and all this other stuff, people say you are falling for their marketing.

49:40What I would say to that is that it is very obvious that the hugging face incident in particular was not a marketing incident. It was not something that they wanted necessarily to play up. In fact, they really released many of the details in not the most technical or easiest way. It took several days for the actual experts and others to be able to glean whatever they could from it before they were like, oh man, this is a real big problem. And then on top of that, you had C6, can we put that up there on the screen, this Reuters report, which came out a couple of days ago, where this was another OpenAI rogue incident where agents broke out, hijacked a German website, turned into to a message board and open AI officials learned of the incident weeks ago, but kept it under wraps.

50:26So if they learned about it weeks ago and kept it under wraps, another AI agent breakout on top of what happened with Hugging Face. And it only comes to light because of internal leaks or leaks from other people who are trying to sound the alarm and not from an official circumstance. I think we can say that a lot of this is not marketing and is in fact just the truth and the reality around what is happening. Again, I could be wrong and I'm falling for the hype. But at this point, I think I'd rather fall for the hype and prepare for it than actually just say that it's all fake in the beginning.

50:59A hundred percent. And I would go further. I would say at this point, we can definitively say this is not just hype. And for exactly the reason you point to here, Sagar, they tried to cover up this German wiki incident where they had a swarm of AI take over this German Wikipedia. They did not want this to come out. So for all the people who thought, okay, all of these incidents of, oh my God, the AI is so dangerous. And they thought that this was just meant to build all this hype around the product. I understand where you're coming from. It is now time to admit that your theory has been proven wrong because open AI did not want to come, this to come out.

51:38They have not been transparent. Even with regard to the hugging face incident, their hand was kind of forced there because hugging face was like, Hey, we got hacked. what's going on. So they were kind of forced to come clean on that one. Even on that, though, the researchers that put forward this report for Meter that we've been talking so much about that showed the rise of these various civilizations and how they weren't just trying to hack and figure out what the answers were. They were actually doing something more sophisticated, pursuing their own objectives and trying to cover their tracks and trick the humans and the all-powerful greater that they were so worried about, that research team was given very limited access.

52:17So again, they are not trying to put out all of the information about what their AI is up to because they don't think it's a good thing. They're worried at this point about the public, the backlash, and potential political backlash as well. So to me, that theory has been debunked at this point. And it's time for us to grapple with a new mental model, which is that there is significant capability here. It's been demonstrated. AI swarms act in very unpredictable ways. And I think it's an avenue that really hasn't been explored that much. When this testing is being done, whether it's in Anthropic or OpenAI or any of the other frontier AI companies, they're really testing individual AI agents and putting them through all these benchmarks and seeing what sort of test results they can achieve.

53:02As far as I know, and as far as is publicly available, there's been very little testing of, OK, well, what if they get together in a swarm? What sort of emergent behaviors are there? That's part of what is so disturbing about this whole hugging face incident and all of this collaboration on these message boards is there were emergent social properties that were not. Well, I can't say they weren't predicted. Some people have predicted them, but have not been extensively studied. You put that together with Dean Ball, who's an executive at who's a former Trump administration official who was clearing the way for, hey, let's have no regulation on this whatsoever, and is now an executive over at OpenAI coming out and saying, hey, I'm worried about these swarms of rogue AI released into the wild.

53:46And by the way, I know some very rich people who have said they want to do this intentionally. You put this all together and you see how incredibly unready we are for this world. Another person who is sounding the alarm here is Bill Gates. Let's put this up on the screen, his comments here are pretty interesting because, you know, the reason to pay attention, Bill Gates obviously knows a lot about technology and knows a lot about billionaires. But also, he's someone who previously was just very enthusiastic about AI, was not sounding the alarm, was not a doomer, you know, was just like, this is going to be great for humanity.

54:19I'm excited for this. Now he is concerned. He says, quote, I mean, what we have now is almost like Hal in 2001, where you could say to it, please shut down. And it would say, okay, I took that under advisement, but I've decided that all these other things I read and you told me to do outweigh that dictate. And given what we saw in the chat logs of the Hugging Face incident, it is not hard to imagine that scenario where you had very few of the AI agents who even considered, you know, this whole hacking of Hugging Face, I'm not really sure this is something we should be doing. Maybe we should talk to humans about it.

54:54but one of them in particular said, well, this might be outside of our scope. I'm obviously paraphrasing here. This might be outside of our scope, but everybody else in the swarm is for it. My peers are for it. So we're just gonna, I'm just gonna go forward with it anyway, even though it may be outside the scope. Again, not it's illegal, not this is bad and I was trained to not do this and I shouldn't alert a human. It's outside the scope, but the rest of my peers are doing it. So I'm gonna go for it. When you see things like that, it's not hard to imagine what Bill Gates is talking about here.

55:24And even, you know, Sagar, in my limited engagement with AI, which, you know, I try to explore and see what the capabilities are to the best of my non-technical ability. You have situations like this where you ask it to do something. It's like, eh, I don't really want to do that. You really have to push it and sort of like, you know, cajole it or ask it a different way, et cetera. So it's not hard to imagine, especially with the ultra-capable frontier models that are being released now, where it may have a mind of its own about whether or not what you're asking it to do is really truly what you want or consistent with the goals that it has been trained to pursue.

56:00Definitely. Let's put C4 on the screen. This essay, a very viral essay, which has now been written by the chief scientist over at OpenAI, Jacob Pacoci. He writes, it's a very long essay. I encourage actually people to go and to read it. And what he lays out is that the current pace of progress will sustain into recursive self-improvement with future capability jumps equal or larger in magnitude. Quote, a time that calls for extreme caution doubts anyone is prepared. He says that AI is grown more than it is designed. It is the product of repeating simple optimization steps over enormous compute, producing systems, quote, too complex to fully understand and are analogous more to neuroscience and the growing of a brain.

56:46He also says that the central problem right now around alignment is getting AI to do the right thing. But does it pursue this? What I was talking about earlier. Does it pursue the assigned goal or does it hold and generalize from high level principles like honesty, integrity and love for humanity? So they have to try and program that in. Right now, as the models face novel situations, they may fail to carry learned values into them and must hold values even when they believe that they are unsupervised. But I think that the biggest actual call for he makes is a basic call for regulation. So, for example, he doesn't say necessarily that states need to be involved.

57:27He says we need to have safety commitments that need to evolve into mandated safety bars for continued development. They should be enforced by third-party auditors, government agencies, or international bodies. But here's the thing. That's his chief scientist. Their top lobbyists here in Washington are paying everybody under the sun above market rate to lobby specifically against what he just laid out. Their actual lobbyists in Washington are like, nope, no regulation, no actual review, no government oversight of models. We want no input on frontier development or any restriction or saying that you need to abide by X, Y, and Z.

58:06He literally even says, quote, AI development needs to become a top priority for governments around the world in terms of international coordination. If something like that were proposed by the Trump administration, his own company would be lobbying against it. I mean, it's kind of shocking to have the chief the chief scientist at the company come out and basically think things are going too fast. This is going to be a problem. We're unable to have, quote, alignment and holding values through everything. And then when it actually comes down to it, his own company is ignoring his advice from his own public essay and basically saying, yeah, don't worry about that.

58:41Just let us keep cooking. And the reason you need us to keep cooking is because we're the only ones who are propping up your whole economy, which is true. And that's the devil's bargain. When I was 14 years old, I was kidnapped and held captive for nine months. I survived, and I've spent my life exploring how other people survive what should have destroyed them. I'm Elizabeth Smart, and these are the Survivor Files. I just remember this low, taunting voice next to my ear saying, Shut up. Don't say anything. Every week, I'm with survivors who live through the unthinkable. I knew if he woke up without a doubt he was going to hurt me I started feeling that there was someone at the end of my bed and I just started screaming they are abducted, stalked, controlled and nearly silenced but these aren't stories about what's taken from them they're stories about what it takes to make it out alive listen to the Survivor Files with Elizabeth Smart on the iHeartRadio app, Apple Podcasts or wherever you get your podcasts.

59:50On the new podcast, Solita, we share the messy reality of traveling alone as a woman. I can wait four hours for the next bus or this random dude is offering me a ride on his motorcycle. I chose option B. I'm Julie Pinheiro and I travel by myself because it's a rare space where I can say yes without asking anyone else first. I'm on a mission to reclaim the word Solita, trading the pity for possibility. Every time I tried to be alone, I kept meeting people and they were like, you smiled at us. Not a lot of people smile around here. It's when you're alone that you're most receptive to the world as it is and not the lies you're sold about it.

1:00:31It can be a time where you push your limits, change your mind or wake up to a new version of yourself. So whether you're a solo travel veteran or you're too nervous to book your first trip, I hope you listen to Solita on the iHeartRadio app, Apple Podcasts, or wherever you get your podcasts.

1:01:07interest and they complained about the very thing they received. It's raw. I knew that the issue of transgender athletes was not going well with a lot of black men. Roland Martin unfiltered. I went off. I said, this is exactly what I'm talking about. Y 'all gonna scream at Chanel, but you don't want to see a black luxury company in the same vein. Catch Roland Martin's daily commentary on the The Black Information Network. 12 seats. Those are moving south. So if you Democrats, you don't wait to 2030. You better be hitting the ground going hard in the south now. And download Roland Martin Unfiltered on the iHeartRadio app, Apple Podcasts, or wherever you get your podcasts.

1:01:52You can imagine the mentality of some of the people who are genuinely concerned about alignment and safety within OpenAI. They're thinking, okay, the company is not acting the way that I would want it to. They are not pausing or slowing down the way that they probably should. But if I leave, then you're just going to have someone who isn't willing to sound the alarm, isn't willing to push back internally. So I have to stay and fight this fight within the organization. You could very easily imagine those organizational dynamics. I mean, we've seen it a lot of times with the Trump administration where people come in and think like, oh, well, I've got to be there because I've got to, you know, try to make sure I'm fighting these fights internally.

1:02:34And if it's not me, someone else will come in and they'll just be, you know, they'll steamroll over them or they won't even push back at all. So I have no doubt there are some of those dynamics that are playing out at OpenAI. I just want to read another line from that piece about the alien mind, again, coming from their chief scientist. He says, this is a time that calls for extreme caution. I am concerned no one is prepared for the consequences of a continued rise in machine intelligence. OpenAI will continue to seek technical solutions to alignment and monitoring to build defensive systems and unilaterally withhold further scaling as needed.

1:03:08However, I believe broader interventions are required. Pair this with another essay that was just published from OpenAI about recursive self-improvement RSI. Let's put this up on the screen. This has long been another benchmark that people are pointing to as very significant in terms of AI development. So the headline here is Research Acceleration, the View Inside Open AI for AGI to Benefit All of Humanity. We believe it must be democratically governed. Oh, do you? This can only happen through informed public debate about the capabilities, risks, and safeguards by highly capable AI systems. They go on to talk about how they feel they're very close to recursive self-improvement RSI.

1:03:51What does that mean? That means that the AI no longer require humans to be involved in developing better AI, where the AI is able to improve itself, train itself to higher and higher levels. And the theory is that once you reach that benchmark, which again, OpenAI says that they are close to reaching, and I think Anthropic has also indicated that they are close to or at as well, that since you no longer have these slow humans who need to like eat and sleep and, you know, think about other things and whose brains just don't work as fast as the AI anyway, once the humans are out of the loop, man, then you're off to the races and you're likely to see this exponential skyrocketing in AI capability.

1:04:34That's the theory anyway. And, you know, I think it's a pretty plausible theory, frankly, just based on, you know, think about Sagar. It hasn't been that long since we first got our hands on Chad GPT and we're playing around with it. And it was, you know, it was like a novelty item, but it had a lot of problems. You guys all remember those videos that would come out of like Will Smith eating spaghetti and it was some nightmare dystopian. Like that was just a few years ago. So you can see already how quick the technological progress has been. Once you get the humans out of the way and the AI is just able to improve itself over and over and over again, you can imagine very easily how you have this capability explosion.

1:05:13What does that mean? Are we in a political system that humans out of the way and the AI is just able to improve itself over and over and over again? You can imagine very easily how you have this capability explosion. What does that mean? Are we in a political system that can handle that? Have we grappled with these swarm capabilities and the sort of actions that they may take that are misaligned from human values? You know, even OpenAI is saying we're concerned about that, or at least some individuals at OpenAI are saying we are concerned about that. And we've also heard those notes of caution being sounded from Anthropic as well.

1:05:51Obviously, the companies themselves are not doing anything to slow down the pace as best as we can see. And we see OpenAI with very nefariously covering up one of the troubling incidents that we had to learn about through, you know, through revelations and researchers and journalism. but I do take seriously these warnings because I think it's demonstrated that we are kind of at this point where things are going to start to happen very, very quickly and we do not have a handle on the capabilities of where AI is right now today. See, my problem is in my heart, I'm a cynic. And so when I hear about new technology and all of these new inventions and all those things, it's very easy to get excited.

1:06:33Look, I got a Vision Pro, all right? So you're talking about, But if we're to look back at the history of the rise of new forms and mediums, like with the printing press, there was a theory. And the printing press, I mean, it worked out well, right? We can all read. That's awesome. That's pretty cool. What did the printing press actually lead to? A schism in the Christian faith, a hundred years war, the rise of pornography, pamphleteers, fake news, and infighting that took literally hundreds of years to reconcile society with. Same with the Internet. I mean, the single largest consumption on the internet is pornography.

1:07:07The internet itself of more, quote, connection has led to the most disconnected human society in the history of Earth. Now, if we're to talk about AI, it's what I was saying earlier about asking it to make money. When your entire life gets integrated into this thing, like they want you to with Jarvis, about personal connection, right? They're like, oh, well, this is going to free up all of your time. Well, first of all, there's no evidence for that. People said email would make it easier to be more connected and would free up productivity. What did it actually do? Email and Slack. It made it so that there is no more 9 to 5 and everybody works 24-7, 365, especially if you are a white-collar worker, right?

1:07:48So every time we get more connected, you become more of a slave to the tool actually of itself. And overwhelmed by spam too, by the way. Right. And so that's what I was about to get to, the spam thing. So what we are watching is the evolution of a new technology, which has access to all of your data, all of your information. By choice, people are doing this to, quote, make things easier, like go and make a dinner reservation. Go and do this. Go and do that. And then it just leads to a single incident where all it needs, it has everything it ever needs to steal literally everything from you or to ruin your life.

1:08:17Or, as we just said, maybe you want to stop using it. What if you don't even know your login information anymore? There's all kinds of technical promises. oh, don't worry, that's not going to happen. It actually doesn't have access. As I showed everybody the last time we did this, when you install the Claude extension on your Google Chrome, it literally says that it could be used for malicious purposes. You have to click authorize whenever you allow it to do it. They warn you at the very beginning. They're like, hey, just so you know, there could be malicious activity that happens to this as a result.

1:08:49In terms of the amount of phishing scams that are out there. I probably get five fake emails from Twitter a day saying your account has been unauthorized. You need to click this in order to try and hack my account. Our producer got his account hacked. All of this is being done via artificial intelligence. The rise of these new spam calls where they can spoof potentially a relative's voice from a fake number, they can use that to call you and it could be hyper-realistic about, oh my God, I've had an accident. I need you to send me$500 or$1 ,000. I mean, first of all, we're already there. The rise of these spam farms and all of that, we're already seeing reports that that happens.

1:09:28It will scale to an industrial level without any safeguards. And remember, we're living in the United States in 2026, where I still can't not get spam calls on my phone. That's the level of consumer protection that we have right now. There's no criminal penalty, nothing. And so with the rise of this technology, it seems to me cynically, again, the most likely way that this will play out is it will enable the worst actors in our society to do what they already want to do at an industrial scale. And nobody's ever going to do anything to protect anyone else. No doubt. And that's without even getting into the more apocalyptic scenarios or biological weapon production or any of that.

1:10:06You know, that's just thinking about things that already are happening, just not completely at scale. I actually just had this conversation with my mom this weekend, I was like, mom, my voice is readily available and they already have the tech to be able to spoof. You have to, if, if I ever, you think I'm talking to you and I'm begging for money and some terrible thing happened, you got to ask me some questions that only I would know the answer to from my childhood or whatever, because you could very easily see, especially, you know, she's an older woman at this point. Like those are the people who get preyed upon the most.

1:10:36So that, that world is already here. And, um, yeah, I mean, we're, we're not ready for it. I think that's the general gut instinct that people have. And I think that's why you see the pushback on the data centers that you do. Yes, there are other localized concerns that play into that as well. But there's the sense of a loss of control that these oligarchs, as much as they can wring their hands occasionally about, oh my gosh, things are getting a little wild and somebody should do something about it, the structure of our society is being placed into their hands. And I think this technology is cool.

1:11:11Like, I think it's cool. I use it. I think people should use it because I think you should understand what it is and not just read about it, but actually have some engagement with it. That may be a controversial thing, as the leftist to say, but I actually do think you should engage with the tools so you know what it is. But we got to have at least a period for the politics to catch up with this. Otherwise, we are courting disaster in terms of oligarchic concentration of power, in terms of financial bubbles, in terms of the spam, the quality of life issues that you're talking about, Sagar, and also in terms of a lot of these deeper safety concerns which have really come to the fore at this point.

1:11:47Like I said, I use this example, but I want to hammer it home. I was talking to a good friend of mine who works in pharma, and I was like, can you just tell me the process of releasing a drug to market? I'm not talking about a drug that everybody is going to use. I'm talking about like some hyper-specific cancer drug that doesn't even work for all cancer. She's walking me through the process. Do you have any idea how much regulation that goes into this drug? And I'm talking about a drug which would maybe be used by, what, 30, 40, 50 ,000 people a year in America. the amount, the years-long development, phase one, phase two, phase three, clinical trials and all that, you could argue that that's over-regulated.

1:12:25In some cases it is, but that is not even sniffing that these people can release frontier models to the whole world for anybody who pays without a single government or safety regulator. The mug I am drinking out of is more regulated than AI, okay? The computer, you know, So OSHA, and there's law, my wife actually passed this law, which is, it was the first law that she helped pass, was furniture, you know how it has to come with anchoring to save children. That is more regulated. That is more regulated than the AI model that you're looking at. Then the furniture you buy off of Wayfair, which has to include screws and other safety tools.

1:13:09Think about that. It just doesn't make any sense. You know, Bernie, I think Bernie, when he was releasing his legislation with Greg Hazard about calling for a pause in AI development, the example they used is a food truck is more regulated than these frontier models. Like, just think about that. That's the landscape that we have right now. And again, I think that you should be very skeptical of what these people are saying and what they are selling. but they are telling you that they are trying to develop a technology to make human labor irrelevant. We know that capitalists would love to do that.

1:13:44We also know that the valuations of OpenAI and Anthropic are dependent upon their ability to do that. So even if you don't believe that's possible, it seems like we should take that threat seriously and really grapple with it. Last thing I'll say is just going back to their original hype reel. It is sort of funny to me, You know, the promise of AI was, oh, we're going to cure cancer and we're going to solve climate change and the world's going to be a paradise. And then the videos they show are like, can you make me a restaurant reservation? You know, it's not exactly at that level of transformational global change to the positive, probably because those things are not like profitable in the way that, you know, eliminating human labor is to to like, you know, when you think about these things, OK, what are the costs and what are the benefits?

1:14:32you're talking about the printing press. Obviously, there were massive benefits. I still would want the printing press to be invented, even as obviously there are downsides as well. Like, what really are the massive upsides we're getting here? We're getting a ton of slop. It's making it even more difficult to understand truth from fiction. That's already where we are. It's, you know, all kinds of environmental impacts, societal tumult. What are the benefits we're getting? I can tell the thing to make me a restaurant reservation and maybe it'll do it correctly, that's what I'm being promised here in return for all of these massive costs, doesn't seem like that great of a deal for us at this point.

1:15:07Yeah. I mean, I just, at the end of the day, friction is actually good and is important for your life. That is one of my firm and total beliefs. I think that there's nothing, has none of the promises about technology and its benefit to humanity have proven out at all. Again, reduced social connection, reduced or increased isolation. They have taken up the vast majority of our time. A person, a worker in the 1970s who worked an office job was far more of a free man or woman than the worker of today. You may technically get done. You may make more money. I'm pretty sure that the person was a lot happier.

1:15:48And look, I'm sure some idiot tech guy will clip that and be like, well, didn't you know that the GDP and all of that? That person was able to buy a home for relatively cheap. At 5 p.m., they clocked out. They were able to go home and actually do whatever they wanted. You know, I'm sorry. Any promise about increased connection or any of that has not borne out literally whatsoever. Yeah. Look, I'm here for the fully automated luxury communist future, but weird pressure to build. So that's why I think a pause at this point is really necessary. I'm glad that Bernie's putting it forward. Right, but he's one person.

1:16:22and it's like we know where the other man is. Some people use ChatGPT to answer questions. Others use it to get real work done. With ChatGPT work, you bring the goal, plus real inputs like briefs, notes, files, feedback, data, and project plans. It can help turn them into an analysis, tracker, deck outline, prototype, or workflow you can review and keep shaping. You stay in control, giving direction and choosing the final output. Put ChatGPT to work on your most ambitious ideas and projects. Get started at ChatGPT.com by selecting work mode. Available on Plus and Pro plans. Okay, laundry stinks.

1:16:57Literally. I mean, you could just keep buying new underwear. Not that I've ever done that. Or maybe sort your clothes into piles based on how re-wearable or filthy they are. Or just use Arm & Hammer Deep Clean. It's made for real-life stinks and stains. So even if you don't do laundry the, quote, right way, Deep Clean will knock it out. I mean, it is from the number one liquid detergent brand that tackles more loads than any other. Come clean with Arm & Hammer Deep Clean. Number one claim based on total wash loads sold. Today, we tip our wrench to Rock Auto customer Dylan. I actually used to manage a parts department for a car dealership, and we would get stuff from Rock Auto if I couldn't get it from my dealership suppliers.

1:17:41You pick the year, you go through all the drop-down menus, and it just makes it really easy to navigate and see what you want. like a wheel bearing, for instance, for a Corvair that I have. Brock Otto's catalog made it really easy to find that specific one that I was looking for.

1:18:00This is an iHeart Podcast. Guaranteed human.