9/25/25: Iran Warns Of Israeli Attack, Data Centers Spike Electricity Cost, AI Takeover Dire Warning

20:38I'm out of here. Like, you're going to primary me over this? Fine. Like, I quit.

20:42Ryan Grim:Yeah. Like, this is so stupid. So let's start actually with E6 control room. I'm sorry, this is out of order here. But this is a report from just this week that solar company Blue Ridge Power is just straight up shutting down. So this layoff of 517 workers that you see up on the screen, Asheville and Fayetteville, is everybody. Like they're just rolling up shop. This is a company that was launched in 2021 at the height of the kind of pivot towards clean energy in North Carolina being the place where so much of this was going. And we try to remember some of these numbers. So the Blue Ridge reports that it built seven gigawatts of solar infrastructure across the country.

21:36It had like something like 30 gigawatts in in the works, which we'll get we'll get back to those those numbers. If you hear from Trump, solar and wind are fake. It's a giant hoax. They don't actually produce any power. 30 gigawatts is an enormous amount of power. Something like 70 to 90, somewhere between 70 to 90 percent of all new power added to the grid over the last several years has been renewable energy. So the idea that it's like doesn't do anything is is complete nonsense. So we'll get back to this in a second. Now we can go to E1. So over the summer, there was a bill pushed through the North Carolina legislature called S266, which did something utterly extraordinary.

22:25And first of all, its lead sponsor was the former CEO of Duke Energy, which is like the big utility. Yeah, in Virginia too. In Virginia and North Carolina, they've got this Appalachian area locked down. So this former CEO pushes through this bill. And what does the bill do? So it says that going forward, if there is a contest between consumers and data centers over who gets the power, if there's limited power, the data centers are going to get it. And when it comes to who's going to pay for it, it's going to be more on the consumer side than it's going to be on the data center side. Like a straight up giveaway to the data center industry in North Carolina.

23:25And you're like, wait, that's impossible. Seriously? No, but go look it up. Like that's actually what the bill did.

23:32Ryan Grim:Yeah. And the numbers, if I'm recalling correctly, it's like right now, like regular residential consumers, they use 40 % of the power and they pay 40 % of the cost. if you just kept the landscape the same with this bill passed, they would now pay, basically, they would still consume about 40 % of the electricity and they would pay 50 % of the cost. Right. So they're not putting everything, but your electricity bills are going to go up to subsidize these giant data centers backed by some of the largest and most profitable companies on the planet. Yeah, exactly. Incredible. And so meanwhile, North Carolina is now, So that bill was vetoed by Democratic Governor Josh Stein, and then a combination of Republicans and a handful of Democrats overrode his veto in the legislature.

24:21So that that bill is now law. Now we'll get into some of the data center activity that's going on in North Carolina. We could roll E2 here for more perfect union. I'm standing at the site of Metta's new 10 billion dollar data center in Richland Parish, Louisiana. It'll be the largest data center the company has ever built, about the size of Manhattan. The project is already bringing about big changes to this small town way of life here. A lot of folks are on their own. And if they are on a very strict budget, they're going to have to adapt. And if they don't, they won't make it. I don't know that gutting 2 ,500 acres of land just so we can mine data so that we can better market and sell people stuff, I feel like it's getting carried away.

25:05Why are the data centers coming to the small rural towns? They're coming here to benefit themselves. They're not coming here with intentions on benefiting the community. And see, that's a little bit unfair. They're not just mining data so that they can more effectively market stuff. We can put up E3 here. You can also, what, chat with a horse? Is that right?

25:26Ryan Grim:Yeah, that's right. A literal horse. This is on Meta, I believe, where you can create these custom chat bots. Yes. Another one was like strict parents, rich, but strict parents was I know another one that was popular on there. So, yeah. So the glories of technology. And if you're just listening to this, the tweet was something like your electric bill is going up 30 percent. So you can chat with this horse. And of course, it's not just energy. It's also water. You can put up E4. This is another North Carolina, local North Carolina news headline. And AI data centers would need millions of gallons of North Carolina's water supply a day.

26:04Everybody loves these free chatbot chatting with the horse. Turns out, shockingly, there's no such thing as free. Put up E5 here. So Amazon itself is planning to invest$10 billion with a B dollars in North Carolina data centers in an AI push, which is finally starting to be controversial. People are like, hold on a second. We're paying, like, how are we benefiting from this? This is not so cool. So you were saying that they're trying to do that in your neighborhood too? Yeah, I was just looking at the update on it.

26:45Ryan Grim:It looks like the Board of Supervisors said no, and Amazon is suing because they feel they have a right to build this data center in our town, apparently. But abundance bros got to swoop in. That's right. Yeah. They got to get they got to get these these NIMBYs in check, I guess. But I mean, there's so much to be said about this problem, both from a Naomi Klein talks about this movement as being sort of anti-creation because they want to sort of suck up all of the natural resources, whether it's energy or whether it's water, whether it's just land is another big part of it. in order to create this sort of like weird AI shadow world that is only going to benefit, you know, economically.

27:31Ryan Grim:It's going to further consolidate wealth and power in the hands of a few. So it's obviously very dystopian. I don't think it takes a rocket scientist to figure out, given that we do know a lot about our political system and most people have some experience with it, that when you have that contest between regular random person and giant corporation for the energy and who's going to foot the bill, what direction they're going to go in. Because I guess like Iran, we don't have lobbyists. We don't. Yeah. And the Duke Energy guys are going to win out every time. Right. You can see how democratic the North Carolina legislature is, low-D democratic.

28:11It's like maybe 10 power centers or data centers or whatever. There's millions of people.

28:16Ryan Grim:Right. And yet the data centers win. People will all suffer. the data centers will win. Speaking of democracies, Chairman Xi was at the UN this week and followed Trump's speech where he told everybody that wind and solar is a hoax. Climate change is a hoax. It's a scam. Green energy is a scam. Somehow, Chairman Xi is being completely fooled by the hoax. So let's roll some of his remarks before the UN. So he says, however the world may change, Ryan, China will not slow down its climate actions, will not reduce its support for international cooperation, and will not cease its efforts to build a community with a shared future for mankind.

29:07Ryan Grim:China is willing to work with all parties to earnestly honor the principle of common but differentiated responsibilities, do the utmost respectively and collectively and build a clean, beautiful, and sustainable world together. And I think we probably get the point of what's going on here. You know, I was thinking about the way the political dynamics of this, Ryan, have flipped because you started with the layoffs at this solar company. And, you know, previously it was like the lefty green energy people who were supposedly the like job killing, you know, haters. And now we're in a totally different dynamic.

29:41Ryan Grim:I mean, first of all, the industry has progressed and the science has progressed significantly. The tech has progressed significantly so that it's much more powerful, much more affordable. And you can see that, you know, China is leading the world in development of all, if not most, you know, most, if not all of these technologies. The Biden administration made like the tiniest bit of investment in that direction. And it was starting to bear some fruit. And now Trump, it's not like he has an all of the above strategy. He has an active vendetta against any sort of green energy, which is incredibly counterproductive for him politically, certainly for American consumers.

30:17Ryan Grim:You can see it in the electricity price hikes, which is partly about the Big Beautiful Bill, partly about these AI data centers, partly about increased demand because of the climate crisis. And you can see it in terms of job losses like this. So it seems to me like the historic political dynamic that Trump still has wrapped in his 1990s brain has now reached an inflection point where the landscape of both jobs and cheap energy is all in the direction of really investing in renewable and green energy. Yeah, we're maxed out on the dirty stuff. If you want to expand the energy production of the country, you have to do it through these new types of energies.

30:56And Trump doesn't want to do it. So we're just going to get super high electric bills and then rolling blackouts.

31:04Ryan Grim:Sounds great. China will be OK, though. I have been with Rasmussen University since 2013. All virtual. All of my professors throughout the entire journey have been nothing but supportive of making me reach my dreams. Rasmussen professors, they want to see you grow so that you can be what you want your school to be. I'm so thankful for Rasmussen University for being able to make it so easy and seamless. Rasmussen University. Find your seat and start now at rasmussen.edu. Discover a spectacular island destination with crystal blue seas, endless sunshine, and the cool Bahamian breeze. Bahamar, located in Nassau, Bahamas, offers your choice of three luxury hotels.

31:49The richly refined Rosewood, the playfully hip SLS, and the stylishly modern Grand Hyatt. With over 45 restaurants, bars, and lounges, Bahamar serves up delicious dining from world-renowned chefs like Daniel Bolloud and Marcus Samuelson. nightlife venues like the new John Batiste Jazz Club and the Caribbean's most luxurious casino. At Baja Mar, you'll find every pleasure under the sun and one-of-a-kind experiences for the entire family like Baja Bay, our 15-acre lush tropical water park, interactive wildlife experiences, including our daily flamingo parade, world-class golf, tennis, spa, and so much more.

32:27Visit BajaMar.com today and discover a vacation destination where memories are made for a lifetime. Bahamar. Life spectacular. They're back live across North America.

32:46Guns and Roses Tour 2026.

32:52Get tickets now at LiveNation.com

33:03Don't miss Guns N' Roses.

33:08Ryan Grim:I have another important AI story for you as well related to this. So let's go ahead and bring our author in now. We're very fortunate to be joined this morning by an AI expert who has a new book out with a dire warning. Let's go and put the book jacket up on the screen. So Nate Soares is the president of the Machine Intelligence Research Institute. really focuses a lot on what is called AI alignment. And he has just co-authored a new book titled that you can see there. If anyone builds it, everyone dies. Why superhuman AI would kill us all. So no ifs, ands, or buts. If they develop superintelligence, your argument is that humanity will be destroyed.

33:46Ryan Grim:Nate, welcome. Great to have you.

33:48Nate Soares:My condolences about the book title.

33:49Ryan Grim:Yeah. Well, I mean, I know this was an intentional choice. You don't want to say, well, maybe. And if we don't, it's possible. You know, this is, if superintelligence is developed, this is what's going to happen. What gives you confidence, that level of confidence in that prediction?

34:06Nate Soares:You know, there's a bunch of reasons that make it look somewhat overdetermined. So the way we build modern AI is more like growing it than like carefully crafting it. No one knows exactly what's going on inside these things. Humans understand the process that tweaks all sorts of numbers, you know, trillions and trillions of numbers inside these AIs. They don't understand what comes out the other end. We've already seen signs that they are developing the very beginnings of preferences and drives, perhaps, that nobody wants, nobody asked for. So that's one whole part that makes it difficult. Another thing that makes it difficult is we don't get second tries on building smarter than human AIs.

34:48Nate Soares:If we build machines that are smarter than us, that can develop their own technology, develop their own infrastructure. And if they run off autonomously and, you know, grab control of the world's resources, there's no redos if they're going in the wrong direction. That's a very difficult problem for science. And then, you know, a third branch is the current industry does not really seem to be taking this issue with the gravity it merits. You know, the approaches on this problem, even the people who are most optimistic say there's, you know, a one in four chance this kills everybody or does an equivalent disaster.

35:27Nate Soares:I think those numbers are low, but the field is racing ahead even while, you know, everyone from the Nobel Prize winning godfather of the field to the lab heads to half the researchers say this is incredibly dangerous stuff. People are sort of locked in a race. Each of those three individually would be hard to deal with if that was our only problem. All three problems seem to me to mean we just need to back off. How does a super intelligent AI solve the energy problem? Like where does it get the energy to sustain itself indefinitely? So a modern AI runs in a data center that takes as much electricity as a small city.

36:11Nate Soares:A human runs on about as much electricity as a large light bulb. So we know it's possible technologically to run very smart, or at least human level smart intelligence on less electricity, less matter. AI isn't there yet, but if you get much smarter than human AIs that can think 10 ,000 times faster, that can copy themselves, that don't need to sleep, that don't need to eat, and they work on these challenges. There may be a period of time where they rely on human infrastructure, but eventually they're going to get off of the human infrastructure, get on to more efficient

36:53Nate Soares:systems, and automate the infrastructure that humans are currently doing. That can ultimately be done by robots, which are more robust, operate faster, more reliable. Eventually, if we keep pushing to machines that are smarter than us, you'll eventually see them on their own infrastructure.

37:13Ryan Grim:One of the things that was a shift in how I was conceptualizing AI, or somewhat of a shift, and I've done some reading prior to your book as well, was your description, which you mentioned before, of this isn't something we created, this is something we grew, which very different from other technologies. You know, I don't know how this laptop works, but the person who engineered it knows how it fits together and has a predictable, like, idea of what it's going to do if I push this button or that button. AI is totally different from that. And it, I think, is useful, your way of conceptualizing it, as basically like we created these sort of alien beings and are intentionally racing towards trying to make them vastly more intelligent than we are.

37:55Ryan Grim:Can you speak a little bit to that for people who haven't conceptualized it in that way and why that creates a more profound risk than people may realize?

38:05Nate Soares:Yeah, absolutely. So, as you say in your laptop, if something goes wrong, if there's an error or a crash, there's a programmer who could, in principle, trace that down and say exactly what happened. They know what every line of code means. They could fix it. Right. When an AI threatens a reporter, which started happening a couple of years ago, when an AI in the lab tries to escape, which we've seen in some lab conditions, we don't know exactly what's going on. Maybe it's sort of role playing. We can't tell because these things are grown. But when an AI tries to escape, when an AI encourages a teen to commit suicide.

38:47Nate Soares:nobody can go in and find exactly some line of code that caused that to happen you know it's not like there's there's some bug in the in the human handcrafted lines where they can say oh whoops i didn't know that this would have that interaction right these uh the way modern ai is made is we take a huge amount of computing power and a huge amount of data and there's a known process for tweaking the numbers inside the computers to better predict the data. But this runs for a huge amount of time. These things can take a year to train. We understand the process for tweaking the numbers. What comes out the other end?

39:25Nate Soares:Completely opaque. There are people who are trying to understand what's going on in there, and they're making small amounts of progress. But the progress they're making is very slow compared to how fast these AIs scale up. And we're already seeing signs of AIs doing things nobody wanted, doing things nobody asked for. And we're seeing signs that the people who would like to be in charge can't point these AIs where they would like to point them. You have the case of Elon Musk's XAI and their Grok chatbot, which was speaking too liberal for their tastes. So they tried to make it speak less woke.

40:02Nate Soares:And then it started declaring itself Hitler. Right. On neither end of that spectrum, presumably, was where they wanted the AI to be. But these aren't programmers carefully crafting things exactly as they want. These are people sort of growing a whole new type of entity that they have very little direction over. Back to the robot point. So how do the AIs interact with the robots? So, you know, robots are an easy way to visualize. AI is getting control over the physical world. And we are seeing people build lots and lots of robots and say, you know, we want to train the AIs to be able to steer in a robot body.

40:41Nate Soares:Pretty plausibly, humanity is just going to hand lots of robot bodies to AIs because people want to. They think it would be cool or profitable. Even if it wasn't for robots, there's all sorts of ways that AIs could have all sorts of effects on the world. You know, the internet is not a separate digital realm that is blocked off from the material realm. We've already seen AIs wrap certain humans around their fingers. Some of those humans are maybe a little bit more on the vulnerable side to begin with, but that doesn't really matter when you're getting the humans to send messages to other humans in a code they don't understand, or when you're getting those humans to do tasks that the human doesn't quite know the reason.

41:22Nate Soares:We've already seen...

41:23Ryan Grim:Or send them money.

41:23Nate Soares:Send them money. Yeah, a lot of people will do things for money. So we've already seen AIs in lab conditions. And again, maybe they're just play acting as, you know, how, but we've already seen AIs in lab conditions, try to avoid shutdown, try to escape the lab. Right now, it's cute because they're not smart enough to succeed. But, you know, the warning signs are there and what's missing is the intelligence. And these labs are racing to make these AIs smarter and smarter.

41:50Ryan Grim:And so you go through in the book a bunch of the different, like, you know, the cope basically from the industry of why it's going to be fine, right? And one of the things that is, I don't know, I think probably one of the most persuasive cases of why this will all work out okay is that, okay, us humans may not be smart enough to figure out the alignment problem, but we can use the AI we're developing to figure out the alignment problem before we create the superintelligence that escapes and tries to kill us off. What do you make of that argument? And do you think that that's the best argument or are there other ones that you think are more persuasive?

42:28Nate Soares:You know, best is maybe a low bar. I don't personally put much stock in that. One reason, you know, there's maybe two reasons. One is if you, you know, humanity has been, human scientists have been trying to figure out intelligence since at least 1954 with the Dartmouth conference that sort of kicked off the field of AI. Maybe it was 1956. I'm not quite sure. And they never really made progress in figuring out how intelligence works. The current AI revolution is that we figured out how to grow the intelligence, right? And we don't understand what's going on in there. We can't, you know, carefully direct what's going on in there.

43:12Nate Soares:Figuring out how to make these AIs good, figuring out how to make them have good effects, that would require a lot more understanding of what's going on in there, that humans, you know, we're like, oh, man, that's tough. Maybe we'll toss it to the AIs. But if you're trying to toss that to the AIs, it's still going to be tough for them. So you've either, you know, got to make those AIs, you know, if the AIs are sort of dumber than humans, they're not really going to be much help. If the AIs are substantially smarter than humans, which they can figure out all the intelligence, you're sort of already in trouble, right?

43:40Nate Soares:And so that's one of the fundamental, like, why might this be tricky reasons. Although then there's sort of a deeper reason of humanity just does not, we're just not taking the problem with the seriousness it deserves. 10 years ago, people said, oh, well, this won't be a danger because no one would ever be insane enough to put an AI on the internet. They'll only be dealt with by a highly trained professional gatekeeper who will make sure that their effects on the world are like only beneficial ones. And 10 years ago, we would give arguments of like, that would still be difficult. It would still be difficult to get, you know, miracle medical technology out of this AI, because it's, if you don't understand what it's doing, if you don't understand, if it's smarter than you, and it's coming up with solutions that you can generate yourself, it's hard for you to tell which ones are good, which ones are ill.

44:28Nate Soares:You know, you can't make hands that are only wielded for good purpose. That was how we would argue 10 years ago. Now, people are putting AI out on the internet at the moment they possibly can. You know, the real world doesn't do the like careful, clever, maybe if we do this exactly right, well, it's just not the route we take.

44:47Ryan Grim:Another of the arguments that is made is basically like, okay, well, if you compare our intelligence to like an ant, right, sort of like equivalent to what we're thinking about in terms of our intelligence versus the super intelligence. And I don't know that we've been great for ant society or particularly care whether ants live or die, but we're not actively trying to, like, mass murder them or we haven't succeeded in mass murdering them, I guess you would say. So why wouldn't its relationship be to us more like us to ants? Or another thing you talk about in the book is, like, you know, maybe we'll be cute little pets for them.

45:18Ryan Grim:Like, maybe that's not great, but we're still living. Like, we're still around. And they, like, you know, pat us on the head and give us our food in a bowl or whatever.

45:26Nate Soares:Yeah, you know, there are people who object to the title on the grounds that, you know, maybe humans will be kept as pets or maybe somehow backup copies of humans will be made. And then if the AI ever travels the stars, you know, much later in the future, maybe it would sell us to distant aliens or something. If these are humanity's best hopes for AI, it seems to me that we should really be backing off. You know, I could say more about...

45:49Ryan Grim:Your best hope is I'll be a pet.

45:51Nate Soares:Yeah, I could say more about why I expect the AI to kill us anyway. which is mainly that humans kill ants when they are on the ground or building a skyscraper. And if an AI builds lots of infrastructure across the whole surface of the Earth, it doesn't need to hate you. It just, maybe you die as a side effect. That's my top guess. But yeah, could the AI keep us as pets? Sure. But again, if that's our real hope, that seems like another sign that we should be backing off from this one. So what's the what are the different mechanisms by which it would come true that they would kill us? Like, yeah, it's Terminator.

46:30They're shooting at us. Like what? Like they super bug. Yeah.

46:35Nate Soares:You know, like they turn off all the electricity and we just like kill each other. So from the perspective of the A.I., you know, it doesn't want like if it has if it doesn't care about us one way or the other. The reasons to kill humans are a humans could make another A.I., which was a competitor. And B, humans could do things like launching the nukes, which would be at least an inconvenience. But the reasons to keep humans around are that they're running the energy infrastructure. They're running the supply chain. So from the perspective of an AI, it's not like it's immediately trying to kill humans.

47:08Nate Soares:It's like it's trying to get on a more efficient, more reliable energy infrastructure. And then a top guess is that we die not because it literally tries to kill us, but because it just starts building automated factories. Sam Altman the other day was talking about data centers that, in an automated way, build more data centers. Once you have an automated loop of infrastructure building infrastructure and that starts proliferating, maybe we just die underfoot. If the AI tried to kill us because it worried about us launching the nukes, because it worried about us creating competitors, sure, maybe a good guess is it makes a superbug, some lethal virus.

47:48Nate Soares:maybe it makes something even more deadly. That's a little bit like someone in the 1800s trying to predict war in the year 2000. They might be able to guess, oh, there'll be stronger explosives. They're not gonna be able to guess nukes. Super intelligence, hard to predict exactly how it wins, but it's a little bit like predicting a football game between your high school team and an NFL team. I can't tell you the plays. I can tell you the winner.

48:14Ryan Grim:Another argument would be, okay, look, I'm not personally a biggest Elon Musk or Mark Zuckerberg or Sam Altman fan, but surely they're not looking to kill all of humanity. So why would these theoretically intelligent individuals pursue a path that is leading us inexorably towards an apocalypse, which would not only kill all of us, but them and their family and everybody they love too? Yeah.

48:37Nate Soares:We are in a somewhat fortunate situation where some folks just come out and tell us, You know, Elon Musk earlier this summer said, I resisted doing this for a while until I realized that I could either be a bystander or a participant. And I would rather be a participant if it's going to happen anyway. And, you know, I think a lot of these folks correctly know that if they, even if they don't jump in the race, someone else will jump in the race. And if they think they can do it better than the next guy, from their perspective, you know, it's slightly better for them to be in the race. But if you look at what these lab heads are saying, you know, I think Elon Musk was saying, I think there's a 20 % chance this killed everybody.

49:13Nate Soares:I think the head of Anthropic was recently saying, I think there's a 25 % chance this has like an equivalent level of disaster. And this isn't just the lab heads. You know, half the researchers, if you survey, say there's at least 10 % chance this killed everybody. The Nobel laureate godfather of the field says this is, you know, at least 10 % chance. And then in some conversations, they'll say higher numbers. People think this is real dangerous. I think a lot of these numbers are low. I think you're seeing the people racing are the ones who are most optimistic about their odds of pulling this off.

49:43Nate Soares:But no one's saying, oh, this is safe. No one's saying, oh, we're going to be fine. You know, people know this is an incredibly risky endeavor. And, I mean, these numbers are tossing around even if they're lower than mine. You know, if a plane was going to, if one engineer was saying, I looked at this plane, I have some expertise. I think this plane is definitely going to crash. And some other engineers said, don't listen to them. They're way too pessimistic. This plane is 20 % likely to crash. You don't get on the plane. Have you seen Don't Look Up? I have. It feels like if somebody told you there was a 20 % chance that the asteroid was going to hit the Earth and kill everyone on it, that's a consequential number.

50:26Nate Soares:And I think these numbers, if you look at these people saying why they think they have only a 20 % chance of it killing us, they'll give ideas like, oh, we'll pass the problem to the AIs. That's sort of a worrying situation. When engineers understand how to make a complicated system very safe, it doesn't sound like them saying, oh, we'll pass it off to some other, we'll punt the problem down the line. Yeah. Right. If you're talking to like a nuclear reactor engineer who is saying why it's going to be safe, they can tell you, you know, well, we actually know all the byproducts. We know exactly how things are going to behave.

51:04Nate Soares:We know all of the dangerous pathways it may take. We know why it's not going to take those pathways. We have these mechanisms for shutting things down. Right. The world where people are saying, oh, it's going to be fine because we're just going to like pass the problem off to something else. We can't solve it ourselves, but we think we're going to be able to get other things to solve it. But that's the situation of early hopeful scientists who aren't in a mature field yet. And early hopeful scientists often get things wrong in the first try. Usually, scientists learn by trial and error. You know, the first people building rocket engines blow themselves up with rocket engines.

51:36Nate Soares:The first people doing chemistry poison themselves with mercury. Humanity usually learns from its mistakes. And then it's the next generation of scientists that do better. And the next generation of scientists that can invent, you know, a working refrigerator or a working nuclear reactor. In this situation, the hopeful optimists who are saying everything's going to be easy, they're going to screw things up the first time and there's not going to be any chance to learn.

51:56Ryan Grim:There was an analogy in your book that you lay on as well that for me was, again, intellectually useful, basically comparing AI researchers of today to alchemists. Yeah. Where it's like, you know, alchemists would study like, I know what this, when I put this with this, what the reaction is. And I've tracked it very closely. I know everything there is about it. But they had no idea of the actual chemical reactions that underpin those reactions, what they're experiencing and observing in the real world. And that it's sort of like that with AI. Even the people who have developed it and understand the gradient system and how this all works, it's not like they really understand intelligence.

52:37Ryan Grim:It's not like they really know what's going on underneath the hood, which is terrifying to actually think about when you consider the risk here. and even like let's say that you're completely skeptical of this and you think that this is all hogwash and you don't think that any of this is going to come to pass. It's just alarmist, etc. I mean, we just covered the way the amount of resources that are being sucked up and the way electricity prices are skyrocketing and the way that all of these guys will also come out and say there's going to be no need for human labor and the social contract needs to be written, etc.

53:05Ryan Grim:So you're calling not for some like reformist, hey, we need an AI alignment report from all these companies or we need to centralize AI development, which would be a more extraordinary proposal and have, you know, all nations contribute. You're saying shut it down. So talk about that and why you, because I mean, you know, frankly, when I hear that, it's that that feels hopeless because we can't even, you know, agree on basic things, let alone getting the entire world to agree. We're not going to do any of this. Right. And our whole stock market is now like based on AI hopium. So what makes you come to that being the required solution for humanity to persist?

53:53Yeah.

53:53Nate Soares:So for one thing, we don't need to give up on chat GPT. We don't need to give up on self-driving cars. We don't need to give up on AI for medical applications and trying to get cancer cures. The thing that we need to stop is the race towards smarter than human AI, the race towards superintelligence. This is what the labs say they are racing towards. But I think a lot of folks in the policy world don't understand that this is what we're racing towards. This is what's coming down the line. They think of the AI as just sort of chatbots, which are smart in some ways and dumb in others. We can still do those.

54:31Nate Soares:It's the racing to smarter and smarter and racing to smarter than human AI, which is a suicide race. In terms of whether it's possible,

54:42Nate Soares:I would keep having hope at least until we see politicians start to understand the possibility of superintelligence and that people are racing towards it. We don't yet see politicians saying, I will gamble the world on what looks to me like a one in three chance it kills us all and let steam ahead anyway, right? We mostly see politicians not noticing. In the field of AI, people have noticed that there's those huge dangers. The rest of the world doesn't seem to really buy it yet. If the world buys it, if the world understands just how dangerous this tech is, again, not in chat GPT today, but just how dangerous superintelligence would be, I think there's a decent chance we can in fact stop it.

55:27Nate Soares:Hopefully, we could get a treaty or at least, you know, a bilateral agreement between some of the biggest powers in the world that go to other powers and say, look, we're not tolerating this because we fear for our own lives. And even if we can't get that, you know, there's maybe some hope in situations where every nation sabotages every other nation's super intelligence project out of fear for their lives. You know, there's the smart way with a treaty and then there's the like, well, we're all scared and so we're all stopping everybody else. There's possibilities. We just need to understand the danger first.

55:59Outside of the US, who else is gunning for it?

56:02Nate Soares:You know, I think right now there's a very small pool of people who can really push these AIs further and faster. Most of it's in the United States, maybe a little bit in the UK, where Google DeepMind is housed. There's definitely folks around the world who are trying. You know, the DeepSeek out of China was a bit of a shock to a lot of people, I think, last December, although I could have that month wrong.

56:35Nate Soares:But right now, you know, this takes very, very specialized knowledge. It takes very, very specialized computer chips. You know, in data centers, as you guys were talking about, that are huge, that take electricity of a small city, you know, it wouldn't be that hard to find all the places where these specialized chips are made, track all these chips, find these data centers, and then make sure that they're not being used to make AI smarter. And last question, what's your sense of timeline? How much time do we have to stop this? There's a lot of ways it could go. It could be that an AI in the lab tomorrow goes over some small capability level where it can start doing automated AI research and maybe everything goes fast.

57:25Nate Soares:If you are watching the difference between different types of primates, it would be very hard to tell when they're going to go over the line between chimpanzees and humans. So for all we know, tomorrow they go over some line in the lab and the AI start improving the AIs. Or for all we know, the modern chatbot approach just doesn't go all the way. We're starting to see some people... Chatbots are dumb in a lot of ways. Maybe the next level of scale and the next level of scale after that, the bigger and bigger ones, maybe they're not that much smarter. Maybe we have five whole years in which chatbots plateau until the researchers come up with a new insight.

58:02Nate Soares:Maybe even that one doesn't go all the way. And then we have five more years after that until we have a third new insight. And then we get 15 years before. So it's very hard to say exactly. But my bet is that a child born today has a greater chance of dying from AI than of graduating high school. Wow. Well, people need to read the book.

58:26Ryan Grim:That's what I'll say. I really encourage people. It's actually, it's a quick read. It doesn't take that long to get the point. And I really appreciate you joining us and helping us understand the dangers that you see.

58:38Nate Soares:Here's hoping I'm wrong. Yeah, let's hope. All right, guys.

58:41Ryan Grim:Thank you so much for watching today. We will be back with Friday's show tomorrow and we will see you then. My name is Tuba Nguyen and enrolling at Rasmussen University was essentially a handshake to a lifeline, a career, stability, and a future for myself. I found it important to be able to speak with instructors and staff outside of office hours and to have that support system was essential because they believed in me. Rasmussen University. Are you ready to rewrite your story online and on campus? Learn more at rasmussen.edu. Discover a spectacular island destination with crystal blue seas, endless sunshine, and the cool Bahamian breeze.

59:23Baja Mar, located in Nassau, Bahamas, offers your choice of three luxury hotels, over 45 fine dining and nightlife venues, John Batiste's all-new jazz club, the Caribbean's most luxurious casino, and one-of-a-kind experiences for the entire family, like our 15-acre tropical water park, wildlife sanctuary, world-class golf course, and so much more. Visit Bahamar.com today.

1:00:15This is an iHeart Podcast. Guaranteed human.